286 110 10MB
English Pages 942 [909] Year 2023
Ashwini Deshpande Editor
Handbook on Economics of Discrimination and Affirmative Action
Handbook on Economics of Discrimination and Affirmative Action
Ashwini Deshpande Editor
Handbook on Economics of Discrimination and Affirmative Action With 48 Figures and 54 Tables
Editor Ashwini Deshpande Professor of Economics Ashoka University Sonipat, Haryana, India
ISBN 978-981-19-4165-8 ISBN 978-981-19-4166-5 (eBook) https://doi.org/10.1007/978-981-19-4166-5 © Springer Nature Singapore Pte Ltd. 2023 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
This volume is dedicated to my “Economics of Discrimination” students of the past two decades. Their insightful questions and critiques challenged me to explore deeper, ultimately advancing my understanding of this difficult and contentious terrain.
Acknowledgments
Thanks to the advisory board for their support in this endeavor, particularly to those who sent their contributions and suggested other contributors. Special thanks to the authors who provided excellent pieces, despite their busy schedules, not to mention the added pressure due to the pandemic. Daniel Challam deserves thanks for his early research assistance and handling of correspondence with authors and Springer.
vii
Contents
1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ashwini Deshpande
Part I
1
Analyzing Discrimination and Disadvantage . . . . . . . . . . . . .
15
2
Taste-Based Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roland Rathelot and Mirna Safi
17
3
Stratification Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lucas Hubbard and William A. Darity Jr.
49
4
Transforming Gendered Labor Markets to End Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Diane R. Elson
69
5
Theories of Discrimination: Transnational Feminism Inderpal Grewal
..........
85
6
Discrimination as Focal Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kaushik Basu
105
7
Inequality of Opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Patrizio Piraino and Josefina Senese
117
Part II Methodological Approaches to Understanding Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..............
133
.................
151
Lab-in-the-Field Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Allison Demeritt and Karla Hoff
235
8
Decompositions: Accounting for Discrimination Gurleen Popli
9
Field Experiments: Correspondence Studies Marianne Bertrand and Esther Duflo
10
131
ix
x
11
12
Contents
Methodological Approaches to Understanding Discrimination: Experimental Methods – Trust, Dictator, and Ultimatum Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Abhinash Borah Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stephanie Seguino
13
Surveys, Big Data, and Experiments . . . . . . . . . . . . . . . . . . . . . . . Dominik Koehler and Nicholas Menzies
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vijayendra Rao
Part III
Intersectionality and Discrimination . . . . . . . . . . . . . . . . . .
261
285 299
323
351
15
Tackling Intersecting Inequalities: Insights from Brazil Naila Kabeer and Ricardo Santos
........
353
16
Race and Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rhonda Vonshay Sharpe
407
17
Caste and Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kalpana Kannabiran
423
Part IV 18
19
20
21
22
Social and Regional Dimensions of Discrimination . . . . . .
441
The Economic Side of Religious Discrimination in France: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christophe Jalil Nordman
443
Socioeconomic Disparities Among Racialized Immigrants in Canada . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Karun K. Karki, Delores V. Mullings, and Sulaimon Giwa
463
The Economics of Discrimination and Affirmative Action in South Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Imraan Valodia and Arabo K. Ewinyu
481
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Timothy Amos
499
Caste and Socoieconomic Disparities in India: An Overview . . . . . Rajesh Ramachandran
517
Contents
23
xi
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christophe Jaffrelot and Kalaiyarasan A.
Part V
Dimensions of Discrimination
.......................
24
Stereotypes and the Administration of Justice . . . . . . . . . . . . . . . . Brendan O’Flaherty and Rajiv Sethi
25
Evidence of Covariation Between Regional Implicit Bias and Socially Significant Outcomes in Healthcare, Education, and Law Enforcement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tessa E. S. Charlesworth and Mahzarin R. Banaji
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Smriti Sharma
543
565 567
593
615
27
Discrimination in Credit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sunil Mitra Kumar
28
Gender-Based Discrimination in Health: Evidence from Cross-Country . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aparajita Dasgupta
649
......
667
29
Dimensions of Discrimination: Discrimination in Housing Saugato Datta and Vikram Pathania
Part VI 30
Affirmative Action
................................
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thomas E. Weisskopf
31
Experimental Evidence on Affirmative Action . . . . . . . . . . . . . . . . Véronique Gille
32
Does Political Affirmative Action Work, and for Whom? Theory and Evidence on India’s Scheduled Areas . . . . . . . . . . . . . Saad Gulzar, Nicholas Haas, and Benjamin Pasquale
633
697
699 719
729
33
Gender Quotas and Representation in Politics . . . . . . . . . . . . . . . . Vidhu Verma
761
34
Caste Quotas in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Amit Thorat
779
35
The Effectiveness of Affirmative Action Policies in South Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rulof Burger, Rachel Jafta, and Dieter von Fintel
797
xii
36
37
Contents
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need of a Rethink . . . . . . . . . . . . . . . . . . . . . . . . . . . Hwok-Aun Lee Stereotype Threat Experiences Across Social Groups . . . . . . . . . . Valerie Jones Taylor, C. Finn Siepser, Juan José Valladares, and Rita Knasel
Part VII 38
Miscellaneous
...................................
Feminism as Racist Backlash: How Racism Drove the Development of Nineteenth- and Twentieth-Century Feminist Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tommy J. Curry
819 841
867
869
Inequality and Inefficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pranab Bardhan
897
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
905
39
About the Editor
Ashwini Deshpande is Professor of Economics at Ashoka University, India. She is Founding Director of the Centre for Economic Data and Analysis (CEDA), Ashoka University. Prior to Ashoka University, Prof. Deshpande was affiliated to the Delhi School of Economics. She is also a Fellow of the International Economic Association, Research Fellow at IZA Institute of Labor Economics, a Fellow of the Global Labor Organization (GLO), and non-resident Senior Research Fellow of United Nations University - World Institute for Development Economics Research (UNU-WIDER), Helsinki. She has published extensively on economics of discrimination and affirmative action, focusing on caste and gender in India. Other areas of her research interest are social identity and economic outcomes, specially labor market outcomes; self-employment; early childhood development; and education. Recipient of the Exim Bank Award for outstanding dissertation in 1994; the VKRV Rao Award for Indian Economists under 45 and the SKOCH Award for Gender Economics, she is the author of “The Grammar of Caste: Economic Discrimination in Contemporary India” and “Affirmative Action in India”, published by Oxford University Press.
xiii
Advisory Board Members
Pranab Bardhan is Professor of Graduate School at the Department of Economics, University of California, Berkeley. He had been at the faculty of MIT, Indian Statistical Institute, and Delhi School of Economics before joining Berkeley. He has been Visiting Professor/Fellow at Trinity College, Cambridge; St. Catherine’s College, Oxford; and the London School of Economics. He held the Distinguished Fulbright Siena Chair at the University of Siena, Italy, in 2008–2009, and was the BP Centennial Professor at the London School of Economics from 2010 to 2011. He has worked in theoretical and empirical research on rural institutions in poor countries, on political economy of development policies, and on international trade. A part of his work is in the interdisciplinary area of economics, political science, and social anthropology. He was Chief Editor of the Journal of Development Economics during 1985–2003. He was the Co-Chair of the MacArthur Foundation-funded Network on the Effects of Inequality on Economic Performance during 1996–2007. He was awarded the degree of DSC (Honoris Causa) by the Indian Statistical Institute in 2013. Kaushik Basu is Professor of Economics and the C. Marks Professor of International Studies at Cornell University, and former Senior Vice President and Chief Economist of the World Bank (2012–2016). From December 2009 to July 2012, he served as the Chief Economic Advisor (CEA) to the Government of India at the Ministry of Finance. Till 2009, he was Chairman of the Department of Economics, and during 2006–2009, he was Director of the Center for Analytic Economics at Cornell. Earlier he was Professor of Economics at the Delhi School of Economics, where in 1992 he founded the Centre for Development Economics in Delhi and was its first Executive Director. He is currently the President of the International Economic Association (2017–2020). He was the fourth President of the Human Development and Capabilities Association, which was founded by Amartya Sen. He has held advisory posts with the ILO, the World Bank, and the Reserve Bank of India and was, for several years, member of the steering committee of the Expert Group of Development Issues xv
xvi
Advisory Board Members
set up by the Swedish Government. He has also served as member of the Board of Directors of the Exim Bank of India. A Fellow of the Econometric Society, Prof. Basu has published widely in the areas of Development Economics, Industrial Organization, Game Theory, and Welfare Economics. In May 2008, he was awarded one of India’s highest civilian awards, the Padma Bhushan, by the President of India. Haroon Bhorat is Professor of Economics and Director of the Development Policy Research Unit at the University of Cape Town. He has co-edited 5 books and published over 150 academic journal articles, book chapters, and working papers, covering labor economics, poverty, and income distribution. He recently co-edited The Oxford Companion to the Economics of South Africa. He studied at the Massachusetts Institute of Technology and was a Cornell University Research Fellow. Prof. Bhorat consults with international organizations such as the ILO, the UNDP, the World Bank, Ratings Agencies, and emerging market fund managers. He is a member of the World Bank’s Commission on Global Poverty and an IZA Research Fellow (institute for the study of labor). He holds a highly prestigious National Research Chair and is Director of the Western Cape Tourism, Trade and Investment Promotion Agency Board. Prof. Bhorat advised two previous South African Presidents and was an Economic Advisor to former Ministers of Finance. François Bourguignon is Director of Studies at the Ecole des Hautes Etudes en Sciences Sociales and is Emeritus Chair at the Paris School of Economics. Originally trained as a statistician, he earned a PhD in Economics at the University of Western Ontario and a PhD at the University of Orleans. He held the position of Director of the Paris School of Economics from 2007 to the end of January 2013. His theoretical and empirical work focuses on income distribution and redistribution in developing and developed countries. He is the author of several books and numerous articles in international economic journals. He has received, during his career, several scientific distinctions and has taught in several foreign universities. He has extensive experience in consulting with several governments and international organizations. From 2003 to 2007, he was Chief Economist and first Vice President of the World Bank in Washington. William A. Darity Jr. is the Samuel DuBois Cook Professor of Public Policy, African and African American Studies, and Economics, and is Director of the Samuel DuBois Cook Center on Social Equity at Duke University. He has served as Chair of the Department of African and African American Studies and was founding Director of the Research Network on Racial and Ethnic Inequality at Duke. Previously, he served as Director of the Institute of African American Research, Director of the Moore Undergraduate Research Apprenticeship Program,
Advisory Board Members
xvii
Director of the Undergraduate Honors Program in economics, and Director of Graduate Studies at the University of North Carolina. Prof. Darity’s research focuses on inequality by race, class, and ethnicity, stratification economics, schooling and the racial achievement gap, North-South theories of trade and development, skin shade and labor market outcomes, the economics of reparations, the Atlantic slave trade and the Industrial Revolution, the history of economics, and the social psychological effects of exposure to unemployment. He has served as Editor in Chief of the latest edition of the International Encyclopedia of the Social Sciences (Macmillan Reference, 2008) and as an Associate Editor of the new edition of the Encyclopedia of Race and Racism (2013). Bhaskar Dutta is Distinguished University Professor of Economics at Ashoka University, India. His research interests include Cooperative Game Theory, Mechanism Design, Formation of Groups and Networks, Social Choice Theory, and Development Economics. He has been Professor of Economics at the University of Warwick since 2000. He has had a long association with the Indian Statistical Institute, where he has taught during 1979–2002. He has also been a Visiting Professor in several universities including the California Institute of Technology, Universitat Autonoma de Barcelona, Universite Cergy-Pointoise, Paris, and University of Graz. He was the winner of the Mahalanobis Memorial Award of the Indian Econometric Society in 1990. He was the President of the Society for Social Choice and Welfare (2014–2016). He is also a Fellow of the Econometric Society, and the Society for Advancement of Economic Theory. He has been Chair, Standing Committee for India and South Asia, as well as a member of the Council of the Econometric Society. He is currently a member of the Council of the Game Theory Society. He has also served as consultant for the World Bank, UNDP, ILO, and ADB. He is Managing Editor of Social Choice and Welfare and Advisory Editor of Games and Economic Behavior. Diane R. Elson is a British economist, sociologist, and gender and development social scientist. She is Professor Emerita of Sociology at the University of Essex and a former Professor of Development Studies at the University of Manchester. She is noted for her work on issues of development and human rights. She was a member of the UN Millennium Project Taskforce and Advisory Committee member for the UNRISD Policy Report on Gender and Development. She is Vice President of the International Association for Feminist Economics. Her research interests include gender and fiscal policy, and gender and international trade. She is the author of the books Male Bias in the Development Process and Budgeting for Women’s Rights: Monitoring Government Budgets for Compliance with CEDAW (Concepts and Tools) and of many other publications and articles. She has also worked as Special Advisor for UNIFEM and is the Chair of the UK’s Women’s Budget Group. She is the winner of the Leontief Prize for Advancing the Frontiers of Economic Thought for 2016, along with Amit Bhaduri.
xviii
Advisory Board Members
Marc Galanter is Professor of Law Emeritus at the University of Wisconsin Law School. Previously he was the John and Rylla Bosshard Professor of Law and South Asian Studies at the University of Wisconsin-Madison and LSE Centennial Professor at the London School of Economics and Political Science. He teaches South Asian Law, Law and Social Science, Legal Profession, Religion and the Law, Contracts, Dispute Processing and Negotiations. He has authored numerous books and articles related to law, the legal profession, and the provision of legal services in India. Prof. Galanter is also an expert on the Bhopal disaster that occurred in Bhopal, India, in 1984. His collection of court documents, newspaper clippings, secondary sources, and photos form the foundation of the “Bhopal: Law, Accidents, and Disasters in India” digital collection maintained by the University of Wisconsin Law School Library. The Bhopal digital archive contains thousands of documents, videos, a timeline, and a bibliography of other works about the Bhopal disaster. Naila Kabeer is currently Professor of Gender and Development at the Gender Institute, London School of Economics and Political Science. Prior to this, she was Professor of Development Studies at the School of Oriental and African Studies (SOAS), University of London. She was also a Professorial Fellow at the Institute of Development Studies, Sussex, where she is currently an Emeritus Fellow. She is interested in various aspects of inequality and how they play out within households, labor markets, and the wider economy. She is also interested in forms of collective action by poor and marginalized groups that seek a more just distribution of power, resources, and political voice and in the relationship between individual empowerment and societal justice. She has worked under various international and bilateral organizations, such as the World Bank, UNDP, UNIFEM, NORAD, SIDA, Oxfam, and with NGOs such as BRAC, PRADAN, and Nijera Kori. She is currently on the board of the Women’s Rights Program of the Open Society Foundations, of the International Centre for Research on Women, and on the advisory committee of the ILO’s Better Works Program. She is also on the editorial committees of a number of journals including Feminist Economics, Development and Change, Gender and Development, Third World Quarterly, and the Canadian Journal of Development Studies. Katherine Newman is Provost and Executive Vice President of Academic Affairs for the University of California system and Chancellor’s Distinguished Professor of Sociology and Public Policy at UC Berkeley. She previously served as the System Chancellor for Academic Programs for the University of Massachusetts system, the James B. Knapp Dean of Arts and Sciences at Johns Hopkins University. Prior to becoming the Dean at Johns Hopkins University, Professor Newman was the Forbes Class of 1941 Professor of Sociology and Public Affairs at Princeton University and the Director of the Institute for International and Regional Studies, the Founding Dean of Social Science at the Radcliffe Institute of Advanced Study,
Advisory Board Members
xix
and the Director of Harvard’s Multidisciplinary Program on Inequality and Social Policy. She taught for 16 years in the Department of Anthropology at Columbia University and for two years in the School of Law at the University of California, Berkeley. She is the author of 15 books on topics ranging from the sociological study of the working poor in America’s cities to school violence on a mass scale and has written extensively on the consequences of globalization for youth in Western Europe, Japan, South Africa, and the USA. Her work on labor market discrimination by caste and religion in India has been cited extensively for the empirical findings it provides from controlled experiments, observational studies, and qualitative interviews. Amartya K. Sen is the Thomas W. Lamont University Professor and Professor of Economics and Philosophy at Harvard University and was until 2004 the Master of Trinity College, Cambridge. He is also Senior Fellow at the Harvard Society of Fellows. Earlier on, he was Professor of Economics at Jadavpur University Calcutta, the Delhi School of Economics, and the London School of Economics and the Drummond Professor of Political Economy at the Oxford University. Prof. Sen has served as President of the Econometric Society, the American Economic Association, the Indian Economic Association, and the International Economic Association. He was formerly Honorary President of OXFAM and is now its Honorary Advisor. His research has ranged over social choice theory, economic theory, ethics and political philosophy, welfare economics, theory of measurement, decision theory, development economics, public health, and gender studies. Prof. Sen’s awards include Bharat Ratna (India); Commandeur de la Legion d’Honneur (France); the National Humanities Medal (USA); Ordem do Merito Cientifico (Brazil); Honorary Companion of Honour (UK); the Aztec Eagle (Mexico); the Edinburgh Medal (UK); the George Marshall Award (USA); the Eisenhower Medal (USA); and the Nobel Prize in Economics. Sukhadeo Thorat is Emeritus Professor of Economics at the Centre for the Study of Regional Development, Jawaharlal Nehru University in New Delhi, India. He was also the former Chairman of the University Grants Commission (UGC) and of the Indian Council for Social Science Research (ICSSR). As Chairman of the University Grants Commission, he assisted the then government under the UPA and Prime Minister, Dr. Manmohan Singh, in its initiatives to take higher education forward. He is also the founder of the Indian Institute of Dalit Studies and served as its Director from 2003 to 2006. His research areas include agricultural development, rural poverty, institutions and economic growth, problems of marginalized groups, economics of the caste system, caste discrimination, and poverty.
xx
Advisory Board Members
He has been awarded an Honorary DLitt, DSc, DS (Doctor of Education), and LLD degrees by 12 universities from India and 1 university from outside India. He has published 19 books and about 100 research papers. He was awarded the Padma Shri in 2008, for his efforts in extending public service to marginalized groups, minorities, and in education. Thomas E. Weisskopf has taught at the Indian Statistical Institute, at Harvard University, and, since 1972, at the University of Michigan (UM), where he is now Professor Emeritus of Economics. From 1996 to 2001 and from 2002 to 2005, he served as Director of the UM’s Residential College. He has authored or co-authored 10 books and has published more than 100 articles in a wide range of journals in the fields of economic development, macroeconomics, comparative economic systems, political economy, and public policy. His early research focused on issues of third world development and underdevelopment, with particular attention to India. In the late 1970s, his research interests shifted to the macroeconomic problems of advanced capitalist economies; among other things, he undertook studies of trends in productivity growth and profitability from a neo-Marxian political-economic perspective. In the 1990s, he worked primarily on problems of economic transition and institutional development in the formerly socialist economies of the East, concentrating especially on the interaction between political and economic change in Russia. In his recent years of research, he has worked on discrimination and affirmative action in the comparative context of the USA and India, and on the growth of economic inequality in each of these two countries.
Contributors
Kalaiyarasan A. Brown University, Providence, RI, USA Timothy Amos School of Languages and Cultures, University of Sydney, Sydney, Australia Mahzarin R. Banaji Harvard University, Cambridge, MA, USA Pranab Bardhan University of California, Berkeley, CA, USA Kaushik Basu Department of Economics and SC Johnson College of Business, Cornell University, Ithaca, NY, USA Marianne Bertrand University of Chicago Booth School of Business, NBER, and J-PAL, Chicago, IL, USA Abhinash Borah Department of Economics, Ashoka University, Sonipat, India Rulof Burger Stellenbosch University, Stellenbosch, South Africa Tessa E. S. Charlesworth Harvard University, Cambridge, MA, USA Tommy J. Curry Philosophy, University of Edinburgh, Edinburgh, UK William A. Darity Jr. Samuel DuBois Cook Center on Social Equity, Duke University, Durham, NC, USA Samuel DuBois Cook Professor of Public Policy, African and African American Studies, and Economics, Duke University, Durham, NC, USA Aparajita Dasgupta Ashoka University, Sonipat, Rai, Haryana, India Saugato Datta ideas42, Boston, MA, USA Allison Demeritt University of Washington, Seattle, Washington, USA Ashwini Deshpande Ashoka University, Sonipat, India Esther Duflo MIT Economics Department, NBER and J-PAL, Chicago, IL, USA Diane R. Elson University of Essex, Colchester, UK
xxi
xxii
Contributors
Arabo K. Ewinyu University of the Witwatersrand, Johannesburg, South Africa Southern Centre for Inequality Studies, University of the Witwatersrand, Johannesburg, South Africa Dieter von Fintel Stellenbosch University, Stellenbosch, South Africa Véronique Gille Université Paris-Dauphine, Université PSL, LEDA, CNRS, IRD, Paris, France Sulaimon Giwa School of Social Work, Memorial University of Newfoundland, St’ Johns, NL, Canada Inderpal Grewal Yale University, New Haven, CT, USA Saad Gulzar Princeton University, Princeton, NJ, USA Nicholas Haas Aarhus University, Aarhus, Denmark Karla Hoff Columbia University, New York, NY, USA Lucas Hubbard Samuel DuBois Cook Center on Social Equity, Duke University, Durham, NC, USA Christophe Jaffrelot CERI-Sciences Po/CNRS, Paris, France King’s India Institute, King’s College, London, UK Rachel Jafta Stellenbosch University, Stellenbosch, South Africa Naila Kabeer Department of International Development, London School of Economics and Political Science, London, UK Kalpana Kannabiran Hyderabad, India Karun K. Karki School of Social Work and Human Services, University of the Fraser Valley, Abbotsford, BC, Canada Rita Knasel Lehigh University, Bethlehem, USA Dominik Koehler World Bank, Washington, DC, USA Sunil Mitra Kumar King’s India Institute and Department of International Development, King’s College London, London, UK Hwok-Aun Lee ISEAS-Yusof Ishak Institute, Singapore, Singapore Nicholas Menzies World Bank, Washington, DC, USA Delores V. Mullings School of Social Work, Memorial University of Newfoundland, St’ Johns, NL, Canada Christophe Jalil Nordman French National Research Institute for Sustainable Development (IRD), LEDa-DIAL (IRD, PSL University & CNRS) & French Institute of Pondicherry, Paris, France
Contributors
xxiii
Brendan O’Flaherty Department of Economics, Columbia University, New York, NY, USA Benjamin Pasquale New York, USA Vikram Pathania University of Sussex, Falmer, UK Patrizio Piraino University of Notre Dame, Notre Dame, IN, USA Gurleen Popli Department of Economics, University of Sheffield, Sheffield, UK Rajesh Ramachandran Department of Economics, School of Business, Monash University Malaysia, Selangor, Malaysia Vijayendra Rao Development Research Group, The World Bank, Washington, DC, USA Roland Rathelot Institut Polytechnique de Paris, CREST, Palaiseau, France Mirna Safi Sciences Po, CRIS, CNRS and LIEPP, Paris, France Ricardo Santos Ricardo Santos UNU-WIDER, Helsinki, Finland Stephanie Seguino Department of Economics Fellow, Gund Institute for the Environment, University of Vermont, Burlington, VT, USA Josefina Senese Boston University, Boston, MA, USA Rajiv Sethi Department of Economics, Barnard College, Columbia University, and the Santa Fe Institute, New York, NY, USA Smriti Sharma Newcastle University Business School, Newcastle, UK Rhonda Vonshay Sharpe Women’s Institute for Science, Equity and Race, Mechanicsville, VA, USA C. Finn Siepser Lehigh University, Bethlehem, USA Valerie Jones Taylor Lehigh University, Bethlehem, USA Amit Thorat Jawaharlal Nehru University, New Delhi, India Juan José Valladares Lehigh University, Bethlehem, USA Imraan Valodia University of the Witwatersrand, Johannesburg, South Africa Vidhu Verma Centre for Political Studies, Jawaharlal Nehru University, New Delhi, India Thomas E. Weisskopf University of Michigan, Ann Arbor, MI, USA
1
Introduction Ashwini Deshpande
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Analyzing Discrimination and Disadvantage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 What Is Discrimination? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Inequality of Opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Methodological Approaches to Understanding Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Field Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Intersectionality and Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Social and Regional Dimensions of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Dimensions of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Keywords
Discrimination · Affirmative action · Race · Racism · Caste · Religion · Gender · Feminism · Inequality · Social identity
Introduction After a long and tumultuous journey, it is both unbelievable and gratifying to finally write this introduction. The original timeline for this project was first slightly disrupted due to my move from one university to another, and then completely disrupted due to the global shutdown caused by the pandemic. It is thrilling that this volume is now ready for publication. The Covid-19 pandemic did not just disrupt the timeline; it also affected the content in that some authors were not able to deliver, and I was not able to find A. Deshpande (*) Ashoka University, Sonipat, India e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_54
1
2
A. Deshpande
replacements, not for lack of trying. As a result, some important topics that were in the original table of contents are not covered here. However, this regrettable absence, which was completely outside my control, is more than compensated by the superb contributions that you are about to read. All the authors, experts in their respective fields, contributed generously and, with a few exceptions, wrote original papers for this handbook, which, in my view, makes this collection highly valuable. I hope the readers concur! The volume contains 38 chapters organized into 7 sections. Another caveat before we move forward. Even though this volume is entitled the economics of discrimination and affirmative action, several of the contributors are not economists and the chapters are not exclusively focused on an economic analysis of discrimination and affirmative action. This is because, as I learnt very early on in my forays into these topics, the problems are too complex to be captured with only one disciplinary lens. Partly this is also the result of the fact that some economist contributors were not able to deliver due to pandemic-induced distress, as well as the fact that some topics might not have been explored by economists with the kind of nuance that is required for a comprehensive understanding. The contributions from many different disciplines are a strength of this volume, as they collectively enhance our understanding of this multiheaded hydra called discrimination. Thus, the title “Economics of Discrimination and Affirmative Action” should be read more appropriately as “Discrimination and Affirmative Action Through an Economic and Interdisciplinary Lens.” As the latter is too long and unwieldy, we will stick to the original title.
Analyzing Discrimination and Disadvantage This section focuses on various frames or lenses through which we can understand discrimination. A closely linked concept is disadvantage or inequality, to fathom which we need to understand the role of the lottery of birth. Persistent and systematic discrimination can lead to phenomena such as stereotype threat, which can explain the underperformance of discriminated-against groups. Under stereotype threat (discussed below in several papers), reminders of negative stereotypes against their group push the performance of group members in the direction of the stereotype. This section helps understand the interconnections between these related but distinct concepts.
What Is Discrimination? What constitutes discriminatory behavior? For economists, discrimination is simply unequal treatment of economic agents who are otherwise equal. Why do individuals discriminate against others in market settings? Is it compatible with economic rationality? Neoclassical economists have made substantial contributions to this discussion starting with Gary Becker’s seminal 1957 book “The Economics of
1
Introduction
3
Discrimination” where he laid out the theory of taste-based discrimination. This theory focuses on individual tastes and preferences, or simply prejudice, such that economic agents prefer to associate with some persons over others. They exhibit their taste for discrimination by acting as if they were willing to pay something, either directly or in the form of reduced income to associate with some persons instead of others. The grandfather of modern economics, Kenneth Arrow, critiqued this in early 1970s and proposed the theory of “statistical discrimination” as an alternative way to understand discrimination. In the context of the labor market, statistical discrimination occurs when employers use workers’ social identity as a proxy for their productivity, in the absence of complete and costless information about their actual individual productivities. Both these theories have had a profound and enduring influence on economists’ understanding of discrimination. The missing chapter on statistical discrimination is one of the major casualties of the pandemic, but Roland Rathelot and Mirna Safi’s chapter (▶ Chap. 2, “Taste-Based Discrimination”) presents its contrast to the theory of statistical discrimination succinctly such that readers would get a sufficiently good sense of the latter. They begin with an update of Becker’s classic theory with a literature review of theoretical and empirical evidence on taste-based discrimination. They go on to link insights from Becker’s theory, which identifies prejudice or animus as the chief cause of discrimination, to insights from other disciplines, most notably social psychology which helps us understand the source of animus and how it can be changed. Their chapter opens the black box of what constitutes the taste for prejudice, as they present insights from sociology and social psychology, starting with Allport’s seminal book “The Nature of Prejudice,” moving on to implicit and explicit prejudice (which is discussed in several papers in later sections), the stereotype content model to the sociohistorical roots of prejudice. Rathelot and Safi not only make a persuasive case for a multidisciplinary understanding of discrimination but also present a review of possible policy interventions to reduce the scope of prejudice in decision-making. This chapter is followed by two chapters that question the mainstream economic theories and offer alternative theoretical frameworks to understand discrimination. Lucas Hubbard and William A. Darity, Jr (▶ Chap. 3, “Stratification Economics”) present a primer on the newly emerging subfield of “stratification economics” which they use to explain the opposition to affirmative action. Stratification economics focuses on group identities and individual positions within-group and hierarchies across groups. It offers a multidisciplinary theoretical framework for understanding discrimination by linking insights from social psychology and sociology. They argue that affirmative action policies are seen as undesirable because they are expected to improve the subaltern group’s relative position. Stephanie Seguino’s ▶ Chap. 12, “Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination” in Part II elaborates on aspects of stratification economics further. This is followed by Diane Elson’s (▶ Chap. 4, “Transforming Gendered Labor Markets to End Discrimination”) exposition of a feminist understanding of labor market discrimination, which argues that gender discrimination in labor markets
4
A. Deshpande
needs to be understood not as a residual nor as irrational but as rooted in institutional structures and the unequal gender division of paid and unpaid work. Activities that people do to make a living and caring for themselves and their families fall broadly into two parts: the productive sphere, counting all economic activities, and the reproductive sphere, which consist of domestic work as well as both physical and emotional care. The productive sphere cannot operate without the smooth functioning of the reproductive sphere. She highlights why labor markets are not neutral arenas but bearers of gender, in that workers in the reproductive economy are disproportionately female and are penalized the labor market, whereas men, who mostly work in the productive sphere, are rewarded. It is this fundamental asymmetry that causes gender inequality in wages, occupational and sectoral distribution, types of contract, etc. Inderpal Grewal (▶ Chap. 5, “Theories of Discrimination: Transnational Feminism”) provides a succinct history of how the academic landscape has shifted and evolved to include experiences of non-European and non-White American women in the broader realm of women’s/gender studies and development research. It started to shift with recognition of the need for interdisciplinary analysis of gender in contrast to the “masculinist” focus of individual disciplines. Grewal explains that early development research made “‘the third world woman’ . . . a homogenous and passive recipient of Western aid and policy,” as abject subjects to be rescued by western development. The questioning of this framework under the broad umbrella of transnational feminism, which called out the tendency to blame non-Western cultures for the subordination of women, led to nuanced analysis of patriarchy. Grewal’s analysis traces the evolution of this branch through the last three decades of the twentieth century and clearly articulates why transnational feminism is so vital to the understanding of gender gaps in different parts of the world. Despite the transnational feminist critique being around for over three decades, readers would be struck by the persistence of the west/rest-dichotomy-mindset within scholars analyzing gender gaps in different parts of the world, along with a near-complete erasure of the role of colonialism in consolidating practices that led to the subordination of women. Kaushik Basu’s (▶ Chap. 6, “Discrimination as Focal Point”) examines how discrimination might arise in markets when economic agents have no innate desire to discrimination but outcomes would be discriminatory when there is some complementarity between different tasks we do. Basu shows that this kind of discrimination is a product of a free market, where social identities (race, caste, and gender) acquire the salience of a “focal point,” a concept used in game theory by Nobel laureate Thomas Schelling. Basu’s analysis suggests that removing all government regulation and barriers to free competition would not eliminate competition. Contrary to the prediction of the taste-based discrimination model, which shows how in the long run, discriminating firms would be driven out of business.
Inequality of Opportunity The next chapter (▶ Chap. 7, “Inequality of Opportunity”) steps back from specific groups to ask and answer the question: to what extent is inequality (between
1
Introduction
5
individuals or groups) a product of the accident of birth, or circumstances that the individual has no control over, and to what extent it is because of differential effort that individuals exert? An understanding of inequality of opportunity must form a part of the essential toolkit to not only understand the sources of inequality but must also inform policymaking. As Patrizio Piraino and Josefina Senese point out, while the idea of unequal opportunity or the lottery of birth is intuitively simple and appealing, translating it into practice, in particular disentangling personal effort from circumstances over which individuals truly have no control, is challenging. Additionally, effort itself is shaped by circumstances of birth. They discuss the implications of the inequality of opportunity framework for affirmative action: opportunities should be available to all and should not be denied to individuals based on factors outside their control. Again, while this might be acceptable in principle, there is massive opposition to any policy that suggests a more equitable distribution of elite positions. Piraino and Senese spell out these conundrums in the context of educational admissions.
Methodological Approaches to Understanding Discrimination Is discrimination a bit like wind, in that we might be able feel it when it is strong, but is largely invisible and not amenable to being captured? It turns out that with data, especially related to monetary aggregates like income or wages, we can measure how much of the gap between two groups can be attributed to discrimination. Gurleen Popli (▶ Chap. 8, “Decompositions: Accounting for Discrimination”) provides a detailed and meticulous exposition of the original Blinder-Oaxaca decomposition method that decomposes average gaps into an “explained” and an “unexplained” component. She follows up the discussion of average gaps with the more recent advances in the literature that look beyond averages to examine the entire distribution of outcomes of interest. Students would find this chapter particularly useful as in addition to discussing the methodology, she highlights how results should be interpreted with caution, as a decomposition is essentially an accounting exercise. Using secondary data and regression-based methods is an important way to gauge the extent of discrimination. However, there are many facets of discrimination, e.g., discrimination during the job or housing search process, that are not amenable to being captured via secondary data. These might often appear as innocuous or nondiscriminatory in the absence of any other information. For instance, if there are walk-in interviews, and an African American candidate is told there are no vacancies, this might not be immediately perceived as instance of discrimination. However, if for the same job a White candidate gets a call back, it would reveal the employer’s racial bias. To uncover these instances, we would need field experiments.
Field Experiments Marianne Bertrand and Nobel laureate Esther Duflo agreed to contribute their muchcited 2017 chapter in the Handbook of Field Experiments (▶ Chap. 9, “Field
6
A. Deshpande
Experiments: Correspondence Studies”) for this volume. Their contribution, along with the paper by Allison Demeritt and Karla Hoff, introduces readers to the tremendous possibilities of lab-in-the-field experiments to understand discrimination. Bertrand and Duflo highlight the advances in the psychology literature that demonstrates the existence of unconscious or unintentional bias, discussed in greater detail in the section “Dimensions of Discrimination.” They elaborate on audit and correspondence studies to measure discrimination in the field (including a discussion of weaknesses of these methods). The chapter provides a very useful tabular summary of major correspondence studies to measure discrimination along various dimensions (race, gender, caste, sexual orientation, etc.) in a number of countries. The literature has grown larger since the publication of their original paper. They go on to discuss implicit association tests (IAT), discussed in the section “Dimensions of Discrimination,” and Goldberg’s paradigm experiments (laboratory versions of audit or correspondence studies) and papers using these methods. They go on to discuss the consequences of discrimination: stereotype threat and underperformance, self-fulfilling prophecies, and endogenous responses to bias. They also discuss a range of other issues, such as whether diversity hurts or help productivity, or whether diversity in leadership positions affects discrimination. Overall, this chapter is a gold standard reference work for anyone trying to understand the complexity of field experiments for estimating discrimination. Allison Demeritt and Karla Hoff’s ▶ Chap. 10, “Lab-in-the-Field Experiments,” focuses on examining mindsets leading to discriminatory outcomes. This approach, which lies at the intersection of social psychology and economics, is gaining steady acceptance among economists who recognize the need to understand what factors shape human thinking in order uncover aspects of discrimination that secondary data cannot measure. Their chapter provides fascinating insights into behavioral causes of discrimination which they illustrate with examples such as police shootings of unarmed Black men, emotional bias among judges, and so on. This takes us to the use of behavioral games or laboratory experiments that are widely used in understanding discrimination. Abhinash Borah’s ▶ Chap. 11, “Methodological Approaches to Understanding Discrimination: Experimental Methods – Trust, Dictator, and Ultimatum Games,” shows how individuals are often not the rational agents that economics presumes them to be, and that social identity (their own, as well as of others they are dealing with) plays a crucial role in how agents interact with one another. Borah begins with explaining the minimal group paradigm, goes on to elaborate how natural identities have been made salient in games, and ends with a discussion of dictator, trust, and ultimatum games that are widely used in the literature on discrimination. The overall insight from this chapter is the powerful result that in-group favoritism and discriminatory behavior towards the out-group guide behavior by economic agents more often than not. Stephanie Seguino (▶ Chap. 12, “Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination”) reinforces this point by emphasizing that racial stereotypes influence decision-making in ways that lead to racial inequality. She summarizes evidence from audit studies that investigate how employers react to the stigma of race and criminality to find that the stigma of being Black
1
Introduction
7
outweighs the stigma of criminality. Her chapter is a close companion to the chapter on stratification economics that appears in the section “Analyzing Discrimination and Disadvantage.” She deals with prejudice, norms and stereotypes, stereotypes and group positions, and finally applies these concepts to a discussion of policing. Dominik Koehler and Nicholas Menzies (▶ Chap. 13, “Surveys, Big Data, and Experiments”) remind us that there is insufficient quantitative data on the lives of lesbian, gay, bisexual, transgender, and intersex (LGBTI) individuals in developing countries. This not only makes the formation of appropriate policies difficult but also does not allow researchers to use existing methods in the literature to assess disparities and discrimination along the axis of sexual identity. The chapter addresses issues of defining LGBTI communities (which will allow for a more accurate numerical enumeration) and how representative samples can be drawn using big data and experimental techniques. They also highlight the importance of recognizing a spectrum when it comes to classification by sexual identity. In the final paper in this section, Vijayendra Rao (▶ Chap. 14, “Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods”) exhorts economists to be more reflexive and to learn from cultural anthropology and qualitative sociology. Economics, which has always tended towards numerical and quantitative techniques, has taken a behavioral turn over the last decade (as Abhinash Borah’s chapter illustrates). Rao provides an overview of mixing qualitative and quantitative methods, a mix that has the potential to analyze questions of discrimination and development. He argues that this will also lead to more empathetic research and reduce the distance between the researcher and the researched.
Intersectionality and Discrimination This section contains three papers on intersectionality or overlapping identities: intersecting identities in Brazil by Naila Kabeer and Ricardo Santos (▶ Chap. 15, “Tackling Intersecting Inequalities: Insights from Brazil”); race and gender by Rhonda Sharpe (▶ Race and Gender); and caste and gender by Kalpana Kannabiran (▶ Chap. 17, “Caste and Gender”). Together they bring home the fact that tackling discrimination and disadvantage along one dimension is hard enough; intersection of identities makes assessment and remedies of disadvantage particularly urgent and challenging. Kabeer and Santos argue that the international development agenda has focused, often exclusively, on singular monetary measure of absolute poverty. Presenting evidence in support of their argument, they make a case for bringing the intersection between material and identity-based inequalities center stage when it comes to an assessment of poverty. They take a historical long-term view of the evolution of intersecting inequalities in Brazil. Rhonda Sharpe suggests that as the USA becomes more diverse, it will be more complex to identify the role of gender, ethnicity, and race discriminatory outcomes. However, gendered racism, which includes stereotypes about intelligence and
8
A. Deshpande
ability, is not only all-pervasive but it also influences how data are collected and reported. She explains how categories like race, ethnicity, and gender are socially constructed, in that norms, behaviors, and roles are assigned to individuals representing certain identities. She presents data from the USA to show why disaggregated data are needed to appreciate the full impact of intersectionality. Kalpana Kannabiran outlines the realities of caste and gender discrimination in India, notwithstanding the constitutional guarantees to a fundamental right to nondiscrimination. She outlines how control over female sexuality was fundamental to maintaining caste purity, for which caste endogamy was essential. She focuses on historically important anti-caste campaigners who argued why the caste-gender complex needed to be dismantled in order to move towards an egalitarian society. Her focus on the history of social reform in India brings home the realization that Indian feminism is not an offshoot of Western feminism but is far older.
Social and Regional Dimensions of Discrimination This section moves to a new terrain by examining experiences of discrimination in diverse settings. This contains five chapters focusing on discrimination against Muslims in France, the Buraku in Japan, immigrants in Canada, caste disparities in India, and discrimination against Muslims in India. Christophe Jalil Nordman’s ▶ Chap. 18, “The Economic Side of Religious Discrimination in France: A Review,” on religious discrimination against Muslims in France is extremely important as it reminds us that an explicit statement and fierce practice of “liberty, equality and fraternity,” the founding principles of modern French society that foster the idea of a French national identity overriding any group identity, have not been able to transcend discrimination towards groups such as Muslims. Nordman explains how the problem is rooted in France’s colonial history. The erasure of all cultural differences in the name of assimilation has contributed to the feeling of alienation among Muslim immigrants. He presents conclusive evidence of discrimination in a variety of arenas over the last two decades. The lessons from France for the rest of the world are very important as we think of antidiscrimination policies: do we need to acknowledge difference and stress diversity or would the implementation of a uniform national identity over time result in the dissolution of group identities? Karun Karki, Delores Mullings, and Sulaimon Giwa highlight socioeconomic disparities among racialized immigrants in Canada (▶ Chap. 19, “Socioeconomic Disparities Among Racialized Immigrants in Canada”), which is one of the top immigrant host countries in the world. As it celebrates multiculturalism, how equal are the socioeconomic outcomes of immigrants relative to native (White) Canadians? The authors present data that point to disparities in a variety of domains. They highlight policy interventions, including changes in the regulatory environment, that would go a long way towards lowering barriers that immigrants face. Imraan Valodia and Arabo Ewinyu (▶ Chap. 20, “The Economics of Discrimination and Affirmative Action in South Africa”) outline why South Africa is so
1
Introduction
9
critical to the study of discrimination: it is the one of two countries whose economy has been shaped by the discrimination of the majority of the population, in contrast to the rest of the world where discrimination is directed towards minorities. It also has a unique history of the legalization of discrimination via apartheid. Valodia and Ewinyu outline a concise history of the 300-year journey of racial discrimination in South Africa, starting from the time European settlers arrived in Cape Town in 1652. They present a number of compelling statistics that show the contemporary dimensions of discrimination and inequality, but equally importantly policy efforts towards racial equality. As the authors remind us, three centuries of discrimination cannot be reversed in three decades, but there are persistent efforts. Timothy Amos (▶ Chap. 21, “Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002”) reminds us that every society has its outcastes as Japan’s Burakumin. He argues that despite a fair bit of Anglophone scholarship on the Burakumin, there is not much evidence on the politics of financial redress which has reduced the inequality between the Buraku and mainstream communities and reduced discrimination against them. His chapter presents the problem in a historical light and shows how active support by the state has alleviated the problems faced by the Burakumin and shaped a new Buraku identity. In India, certain castes were considered “untouchable” and their presence was considered polluting on account of their traditional occupations that were seen as “dirty” and degrading. Untouchability is illegal in independent India, and these castes are beneficiaries of affirmative action, listed in a group of “Scheduled Castes.” Rajesh Ramachandran (▶ Chap. 22, “Caste and Socoieconomic Disparities in India: An Overview”) examines the contemporary nature of caste disparities in several domains after giving a brief overview of the caste system. He highlights that while disparities are omnipresent, there are clear regional differences, which he shows via color-coded maps. The main conclusion is that caste disparities and discrimination are being reproduced in the present and are not simply a hangover from the past. India has many axes of discrimination and disadvantage: religion is the other important one. Christophe Jaffrelot and Kalaiyarasan A. focus on Indian Muslims as the group that suffers the largest disadvantage among all religious minorities in India (▶ Chap. 23, “Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do”). They present data on the trajectories of marginalization in a variety of domains. Their analysis of broad regional differences is instructive as they highlight differences between Dravidian territory in the south, compared to north and eastern India. Contemporary Indian politics is a key factor in the way these disparities have taken shape.
Dimensions of Discrimination This section begins with a highly significant original paper by Brendan O’Flaherty and Rajiv Sethi (▶ Chap. 24, “Stereotypes and the Administration of Justice”) on how stereotypes affect the administration of justice. While their attention is on the USA, their insights are valid globally. They begin by describing stereotypes,
10
A. Deshpande
conscious (explicit) and unconscious (implicit) beliefs, concepts that are a running theme throughout this volume. They focus on police stop searches that are disproportionately targeted towards African Americans. They go on to discuss conflict, cooperation, and clearance, followed by extreme violence (murder and lethal force). The data are startling: in half the states in the USA, Black exposure to deadly force exceeds White exposure four-to-one. Their evidence brings home the point that while racial profiling with a goal towards crime reduction is often seen as legitimate (including by economists), this leads to a discriminatory and unfair judicial system with substantial harm imposed on the targets of profiling, in addition to entrapment of innocents and lowering of public trust. They persuasively argue that this harm must be taken into account when we evaluate the overall impact of stereotypes and racial profiling in the interest of “justice.” This is followed by a broader exploration of the correlation between regional implicit bias and outcomes in the arenas of healthcare, education, and law enforcement by Tessa Charlesworth and Mahzarin Banaji (▶ Chap. 25, “Evidence of Covariation Between Regional Implicit Bias and Socially Significant Outcomes in Healthcare, Education, and Law Enforcement”). Their chapter tries to answer the puzzle of how an increasing stated preference for equality coexists with persistent discrimination in all significant aspects of life – from housing to jobs to healthcare and law enforcement. Banaji’s work has drawn attention to implicit attitudes and beliefs, which might be contrary to explicit attitudes. In order to measure implicit bias, she (along with Greenwald) pioneered the implicit association test (IAT) that has been used in hundreds of studies worldwide over the last two decades to measure implicit bias that affects individuals and institutions alike. This chapter summarizes several of the studies that Charlesworth and Banaji (with co-authors in different teams) have conducted over the years which show the pervasive implications of implicit bias. Given that minorities are targets of discrimination in the labor market, could they escape discrimination by moving towards self-employment? Smriti Sharma (▶ Chap. 26, “Disadvantage and Discrimination in Self-Employment and Entrepreneurship”) explores these issues by reviewing the literature on disadvantaged minorities in entrepreneurship and self-employment. She begins by offering indicative evidence from a diverse group of countries and moves on explaining the sources of gaps. We find that discrimination in other spheres, such as credit markets, explains a part of the gaps. The conclusion is that we need targeted policies, including ways to reduce unconscious bias by decision makers. Sunil Mitra Kumar’s ▶ Chap. 27, “Discrimination in Credit,” delves deeper into one specific arena that Sharma’s analysis highlights: discrimination in credit. He provides a selective review of the literature about discrimination in credit along all axes of social identity: caste, race, and gender, which makes it clear that this kind of discrimination is ubiquitous. The section on laws, rules, and institutions is insightful as it sheds light on structural or institutional discrimination where discriminatory outcomes occur, because institutions do not consciously take into account existing structures of disparity and exclusion. The chapter delves deeper into mechanisms and methods to estimate discrimination in credit.
1
Introduction
11
Aparajita Dasgupta (▶ Chap. 28, “Gender-Based Discrimination in Health: Evidence from Cross-Country”) focuses on gender-based discrimination in health outcomes by providing cross-country evidence. It is a comprehensive and well-rounded analysis covering many dimensions: discrimination by health providers, measurement issues around discrimination, micro-foundations of gendered institutions and economic development – a section that explores drivers of discrimination. Discrimination is costly; Dasgupta summarizes literature that provides tools to measure discrimination. The key insight from Dasgupta’s analysis is that gender discrimination in healthcare within households is a key driver of overall gender-based inequality in health outcomes. This makes policy even more challenging, as it needs to focus on ways to change attitudes, norms, and behavior that are in the domain of the very private household sphere. Saugato Datta and Vikram Pathania (▶ Chap. 29, “Dimensions of Discrimination: Discrimination in Housing”) investigate discrimination in housing via an audit experiment on India’s largest real estate websites. Complementing the findings of the chapter on religious discrimination in the section “Social and Regional Dimensions of Discrimination,” their evidence shows clear discrimination against Muslims in the real estate sector, measured by callbacks to Muslim clients compared to Hindu upper caste clients. There is less clear evidence against discrimination against Scheduled Caste/Other Backward Classes (SC/OBC) clients. However, as the authors point out, these audit studies use names as signals of identity, and name recognition for SC/OBC clients might be weaker, whereas Muslim names are almost universally recognized. Their findings substantiate a number of anecdotal and journalistic accounts of even very famous film personalities reporting difficulties finding housing on account of their Muslim names.
Affirmative Action What does the presence of discrimination do to individuals who are stigmatized or negatively stereotyped? Do we need special, targeted positive discrimination policies? Thomas Weisskopf has been thinking and writing about these issues over several years. In the opening chapter in this section (▶ Chap. 30, “Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?”), he provides a clear and objective assessment of the costs and benefits of positive discrimination or affirmative action policies. He also compares the implications of group versus class based affirmative action policies, viz., is it better to target social identity groups or would it better to focus attention on individuals below a certain income cutoff? After a careful weighing of all factors, he concludes that there is a strong case for positive discrimination or affirmative action policies. Veronique Gille (▶ Chap. 31, “Experimental Evidence on Affirmative Action”) provides a deeper dive into the costs-versus-benefits issues by summarizing experimental evidence on affirmative action. One of the key criticisms of affirmative action is that it is detrimental to the idea of merit. In other words, it might achieve a
12
A. Deshpande
social justice/diversity/representation goal, but it does so at the cost of efficiency. Another key issue relates to the unintended negative consequences of affirmative action. Beneficiaries might face backlash due to added stigma and hostility from their peers. Experimental evidence helps us answer these questions. Gille’s summary indicates that the efficiency cost of affirmative action to the society, if any, is small, as affirmative action enhances the participation of high-performing individuals within the beneficiary group who would not have participated at all. The backlash effect is also not as large as to make affirmative action do more harm than good. Saad Gulzaar, Nicholas Haas, and Benjamin Pasquale’s (▶ Chap. 32, ”Does Political Affirmative Action Work, and for Whom? Theory and Evidence on India’s Scheduled Areas”) comprehensive and systematic analysis of the implementation of the National Rural Employment Guarantee Scheme (NREGS) in Scheduled Areas of India (where political positions are reserved for Scheduled Tribes) throws light on whether affirmative action reduces efficiency in the provision of public goods. Based on a new dataset of 217,000 villages, they show that for their main policy of interest (NREGS), outcomes under reservation are no worse overall outcomes. Additionally, there are large gains for the targeted groups: STs receive 24.1% more workdays in Scheduled Areas. These gains come at the cost of the relatively privileged (non-SC-ST), not other disadvantaged groups. They find that reservations more closely align benefits to each group’s population share and, hence, do not overcompensate for inequalities. They find similar gains in other pro-poor programs, such as a rural roads program (Pradhan Mantri Gram Sadak Yojana). Their results (reservations not lowering efficiency and beneficiaries not displacing the more disadvantaged) echo findings from other papers1 on Indian reservations (not included in this volume). Vidhu Varma’s chapter (▶ Chap. 33, “Gender Quotas and Representation in Politics”) provides a comprehensive discussion on gender quotas and representation in politics. The discussion is across countries, as the issue of gender representation is not confined to any particular set of countries and straddles both rich and poor countries. She summarizes research and evidence on how quotas could be a fast track to equal representation. Overall, while quotas continue to be controversial, the introduction of quotas for women in legislatures and political parties has doubled the average share of female parliamentarians. She underscores the need for garnering better data on impacts, which would shed clearer light on the efficacy of quotas. Amit Thorat’s contribution (▶ Chap. 34, “Caste Quotas in India”) focuses on India and explains the motivation behind the use of quotas as a redressal mechanism. This chapter provides a good follow-up to the chapter by Rajesh Ramachandran, which provides the justification for a policy of positive discrimination targeted towards marginalized caste and tribal groups. Rulof Burger, Rachel Jaffa, and Dieter von Fintel (▶ Chap. 35, “The Effectiveness of Affirmative Action Policies in South Africa”) start their analysis from 1999,
1
Deshpande and Weisskopf (2014) and Bertrand et al. (2010).
1
Introduction
13
i.e., from the postapartheid period. They go into the details of various policies and find that affirmative action policies were successful in reducing between-group earnings gaps at the top end of the earnings distribution. Overall the gaps remain substantial. This is not surprising after Valodia and Ewinyu’s chapter which reminds us that the history of racial oppression is three centuries old. The final paper in this section by Hwok-Aun Lee (▶ Chap. 36, “Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need of a Rethink”) focuses on an instance of affirmative action that favors the politically dominant but economically disadvantaged Bumiputera majority. This policy has substantially remedied existing inequalities by increasing Bumiputera access to resources and aiding their upward mobility. However, the goal of national integration needs to be advanced, including a discussion about when it might be a good time to roll back affirmative action.
Miscellaneous This is not really a whole section but has two papers that are related to the set of complex issues discussed above but do not fit any one section neatly. The chapter by Tommy Curry critiques America’s feminism (▶ Chap. 38, “Feminism as Racist Backlash: How Racism Drove the Development of Nineteenth- and Twentieth-Century Feminist Theory”) with the argument that it has a anti-Black focus. The chapter takes the view that it is not merely blind spots that have made America’s feminism driven by Whites, but it argues that the primary racial target of White feminists has been the Black male. This argument is similar to the ones we encountered in the chapter on transnational feminism. In the concluding chapter of this volume, Pranab Bardhan’s chapter on inequality (▶ Chap. 39, “Inequality and Inefficiency”) takes a step back from social identities that this volume focuses on and reminds us why inequality is not desirable: it is not only ethically distasteful, it can also be economically harmful, belying the traditional economists’ dogma of equity-efficiency trade-off. I hope this rich selection will be widely used by students, teachers, and researchers to advance their understanding and will inspire readers to contribute to this field of knowledge through further research.
References Bertrand M, Hanna R, Mullainathan S (2010) Affirmative action in education: evidence from engineering college admissions in India. J Public Econ, Elsevier 94(1–2):16–29 Deshpande A, Weisskopf TE (2014) Does affirmative action reduce productivity? A case study of the Indian railways. World Dev 64:169–180
Part I Analyzing Discrimination and Disadvantage
2
Taste-Based Discrimination Roland Rathelot and Mirna Safi
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Becker’s Theory, Extensions, and Empirical Assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conceptual Frameworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Empirical Assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Opening the Black Box of Taste-Based Discrimination: Interdisciplinary Insights . . . . . . . . . . . . From Taste to Prejudice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From Varieties of Prejudice to Varieties of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where Does Taste Come From? Investigating the Roots of Prejudice . . . . . . . . . . . . . . . . . . . . . . Back to Taste-Based Versus Statistical Discrimination: The Role of Information and Accuracy of Beliefs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Towards an Integrated Approach to Discrimination: A Multidisciplinary Perspective . . . . . How to Design Public Policies to Reduce the Consequences of Taste-Based Discrimination? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Prejudice Reduction Interventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reducing the Scope of Prejudice in Decision-Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Towards Less Prejudiced Societies? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18 19 19 22 26 26 27 30 33 35 36 36 38 40 41
Abstract
Becker’s (The economics of discrimination. Chicago University Press, Chicago, 1957) book introduces the concept of taste-based discrimination. This chapter first reviews a large literature in economics that aims to complement Becker’s theoretical framework and to test its empirical predictions. Then, the chapter covers the literature in several disciplines (mostly economics, social psychology, R. Rathelot (*) Institut Polytechnique de Paris, CREST, Palaiseau, France e-mail: [email protected] M. Safi Sciences Po, CRIS, CNRS and LIEPP, Paris, France e-mail: mirna.safi@sciencespo.fr © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_1
17
18
R. Rathelot and M. Safi
and sociology) that defines concepts that are close to Becker’s prejudice, and question its sources. The chapter also seeks to connect taste-based discrimination with other theories of discrimination, including statistical discrimination. The final section presents policies that have been proposed to remedy taste-based discrimination and reviews the literature that evaluates these interventions. Keywords
Prejudice · Bias · Animus · Stereotype · Inaccurate beliefs · Ethnic gaps · Gender gaps
Introduction Discrimination is a pervasive issue in most markets, countries, and social spheres. Although both the phenomenon and its consequences for many discriminated populations have been thoroughly documented, researchers and policy makers are still struggling to identify its mechanisms (see Lang and Kahn-Lang Spitzer (2020); Bertrand and Duflo (2017); Neumark (2018) for the most recent reviews in economics). Taste-based discrimination, as proposed by Becker in his seminal book published in 1957, assumes that discriminating agents incur a utility cost when they interact with individuals from the group that they discriminate against. These insights have been influencing research on discrimination for more than half a century and remain a central tenet of most theoretical and empirical studies in the field. This chapter has three main objectives: (i) to update the literature review on taste-based discrimination with the most recent empirical evidence and theoretical developments; (ii) to bridge the gap between the economic theory of taste-based discrimination and the literature on prejudice in psychology, sociology, and political sciences; and (iii) to discuss the implications of the theoretical framework of tastebased discrimination on antidiscrimination policies and interventions. While scholars from different disciplinary backgrounds have extensively studied discrimination as a central mechanism of social inequality in contemporary societies, the difference between the literature in economics and the other social sciences is striking. A large part of the research in economics on taste-based discrimination aims to understand and measure the relationship between prejudice and equilibrium outcomes in various markets. Economists hence conceive and use the concept of prejudice (sometimes called animus) as a largely ad hoc penalty term in the utility function. The literature in social psychology, sociology, and political science has focused on understanding where prejudice comes from, how it is generated, and how it can be changed. If discrimination comes from prejudice, one needs to find out what the determinants of prejudice as an equilibrium variable are. Considering prejudice as a parameter that is exogenous to policies and is stable over time or across socioeconomic conditions would not make sense. If taste-based discrimination is a large contributor to overall discrimination, understanding the long-term dynamics of discrimination
2
Taste-Based Discrimination
19
requires an understanding of how taste-based discrimination and its manifestations in cross-group differences in socioeconomic outcomes can influence the determinants of prejudice. Section “Becker’s Theory, Extensions, and Empirical Assessments” presents the canonical theory of taste-based discrimination, the way the theory has developed until now, and empirical assessments of this theory, mostly as found in the economics literature in economics. Section “Opening the Black Box of Taste-Based Discrimination: Interdisciplinary Insights” returns to the literature in social psychology, sociology, and economics that investigates the origins of prejudice. Section “How to Design Public Policies to Reduce the Consequences of Taste-Based Discrimination” presents available evidence on policies aimed at reducing the consequences of tastebased discrimination by reducing prejudice, either in individuals or in organizational procedures.
Becker’s Theory, Extensions, and Empirical Assessments Conceptual Frameworks Definition and Seminal Contribution Becker (1957) is the seminal reference on taste-based discrimination, at once defining the concept and introducing a formal theoretical framework. In the simplest version of the model which focuses on employer discrimination, employers incur a utility loss when they interact with workers belonging to a non-preferred group (i.e., a group distinguishable from any other group only in terms of characteristics such as gender, race, religion, sexual orientation, etc.). Becker derives the equilibrium in a setup without frictions where minority workers transact with the leastprejudiced employers. The difference between the wages of the preferred and non-preferred groups is determined by the prejudice of the marginal employer. Here is a simple summary of Becker’s model. Denote A the majority group, B the minority group. Firms produce F(L) where L is the number of employees. The good produced is the numeraire. Workers of groups A and B are equally productive and can receive different wages wA and wB. A firm employing LA workers from group A and LB workers from group B will have a profit equal to: ΠðLA , LB Þ ¼ FðLA þ LB Þ wA LA wB LB : Becker introduces a distinction between the profit and the utility of the employer. The utility of the employer has an additional term, due to taste-based discrimination. V ðLA , LB Þ ¼ ΠðLA , LB Þ δLB : where δ is the taste-based discrimination parameter. If δ > 0, the employer is prejudiced against minority workers. If δ ¼ 0, the employer is indifferent between employing workers of groups A and B.
20
R. Rathelot and M. Safi
We can allow the labor supply to be elastic. Assume that the distribution of the reservation wage is the same in both groups with a cumulative distribution function of H(.). There are NA individuals in group A and NB in group B. Group B is a numerical minority, NB < NA. Labor supply at wage w is equal to NAH(w) in group A and NBH(w) in group B. Suppose first that δ is constant among firms. Employers maximize their utilities such that the optimal number of workers of both groups is: F0 ðLA þ LB Þ ¼ wA ¼ wB þ δ Workers choose to work if they are proposed a wage above their reservation wage. The equilibrium wages in both groups are determined by: wB ¼ wA δ wA ¼ F0 ðN A H ðwA Þ þ N B H ðwA δÞÞ In this simple case of a constant δ, the difference δ between wages in group A and group B makes all employers indifferent between hiring a group A and a group B worker. Because wages offered to A workers are higher than those offered to B workers, the employment rate will be higher in group A than in group B. Employers may differ in terms of their prejudice towards B workers, i.e., in terms of δ. The existence of a wage gap will drive firms with low δ to hire only B workers. Firms with high δ will hire only A workers. In this model, the existence of heterogeneity in δ causes sorting between workers and firms as well as segregation across firms. Formally, suppose that the cumulative distribution function of δ is G(.). The equilibrium is determined by wages wA , the wage gap δ*, and the number L* of jobs as determined by the equations: L Gðδ Þ ¼ N B H wA δ
L ¼ N A H wA þ N B H wA δ wA ¼ F0 ðL Þ
Firms with δ < δ* will hire B workers and those with δ > δ* will hire A workers. An important point made by Becker is that the marginal employer δ ¼ δ* is the one that sets the wage gap for the whole labor market. The wage gap does not depend on average prejudice or on the prejudice of the most prejudiced employers. Rather, it depends on the prejudice of the marginal employer, i.e., the employer on the lower tail of the distribution. While the formulation of prejudice in Becker’s model looks ad hoc, as it relies on the direct manipulation of the utility function of employers, this is not where its main contribution lies. The reason Becker’s model remains influential to this day is because it provides a mapping between the distribution of prejudice among employers and the gender gaps in labor-market outcomes.
2
Taste-Based Discrimination
21
How Can Taste-Based Discrimination Persist? One of the main critiques against Becker’s theory was formulated by Kenneth Arrow. A new employer with no prejudice will be able to hire a minority worker and derive a profit of δ*: the current wage gap in the market. If there is free entry to the market for employers, the least prejudiced employers are able to drive out the most prejudiced ones. Thus, under perfect competition, there should eventually be enough unprejudiced employers entering the market to ultimately eradicate the wage gap in the long run. As pointed by Flinn (2015), the impact of higher competition on market wage gaps is mitigated by the assumption made by Becker that companies’ production function have decreasing returns. Decreasing returns means that there needs to be enough unprejudiced employers to eliminate wage gaps completely. Following the emergence of this critique, several papers have proposed frameworks in which taste discrimination would persist. Goldberg (1982) offers a twist on Becker’s model. Instead of suffering from a utility loss due to prejudice against minority workers, employers experience a utility boost when they interact with majority workers. Switching from discrimination to favoritism means that both neutral and nepotistic employers can survive in the long run, even in a perfectcompetition setting. Charles and Guryan (2008) question the underlying assumption that employers’ prejudice would not have an impact on their utility if they were to become workers. Essentially, they argue that prejudiced employers would choose to remain in the market because they would experience a loss in utility due to the presence of minority workers regardless of their role in the market and, therefore, have no incentive to leave it. Imperfect information is often presented as a way to preserve the persistence of taste-based discrimination in the long run. Black (1995) introduces a search model in which minority and majority workers meet firms and decide whether or not to match with them based on the offered wage and on a randomly drawn nonmonetary value for the match. Firms can be prejudiced, in which case they refuse to hire minority workers, or non-prejudiced, in which case they are indifferent. Because search is random, minority workers can meet prejudiced employers who refuse to hire them. Thus, search is de facto more costly to minority workers. As a consequence, they suffer from longer unemployment spells and exhibit lower reservations wages. Minority workers end up with lower equilibrium wages than majority workers and in worse matches on the nonmonetary amenity dimension. As in Becker’s model, prejudiced employers also suffer from their prejudice because they have to wait longer to find a match. Black makes the argument that these employers can still survive in the long run, if the model allows them to differ in their entrepreneurial abilities. There are other notable differences between Becker’s and Black’s models. For example, when the share of minority workers increases, prejudiced employers have lower profits and tend to be replaced by non-prejudiced employers. This reduces the search cost of minority workers and reduces the gaps in labor-market outcomes between the two groups. Bowlus and Eckstein (2002) build on Black’s model and perform a structural estimation to decompose the causes of the wage gap between Blacks and Whites in the USA. The main differences between their model
22
R. Rathelot and M. Safi
and Black’s are that prejudiced employers will hire minority workers whenever their wages are below some threshold and that minority and majority workers are allowed to have different productivity levels. Interpreting empirical wage gaps through the lens of this model leads to several conclusions: (i) more than half of all the US employers experience a disutility when interacting with Black workers, (ii) on average, the disutility amounts to around 30% of White workers’ productivity, and (iii) Black workers are on average 3% less productive than White workers. In their review of the literature, Lang and Lehmann (2012) argue that the model by Black, as estimated by Bowlus and Eckstein, requires an unduly high level of prejudice to match the data. Becker presents other variations on his model in anticipation of the criticism made about the lack of persistence of discrimination in the long run. One variation has been particularly successful in the subsequent literature: customer discrimination. Even when employers have no prejudice and only care about their own profits, customers may discriminate against the goods and services produced by minority workers. Becker frames it as a difference between the money price of service p and the net price p(1 þ d ) perceived by prejudiced consumers, where d > 0 is the coefficient reflecting customer discrimination. Relatedly, Sasaki (1999) introduces customer discrimination in a search model to explain the gender gap in labor markets, and Borjas and Bronars (1989) use a search model with customer discrimination and imperfect information to explain how Blacks and Whites may self-select into self-employment.
Empirical Assessments Testing the relevance of Becker’s and others’ models of taste-based discrimination has been the subject of several strands of literature. Four of these strands are particularly important: the first one directly tests predictions from Becker’s model, the second one assesses the relationship between market competition and discrimination, the third one focuses on customer discrimination, and the fourth one uses an experimental approach to isolate the impact of prejudice. Other papers identify tastebased discrimination indirectly by ruling out other channels.
Evidence on the Relationship Between Prejudice and Gaps A relatively limited body of literature aims to test the direct empirical predictions of Becker’s model. Charles and Guryan (2008) start from the canonical model by Becker and empirically examine the predictions which posit that: (i) the prejudice of marginal employers matter more than that of the average employer, (ii) the degree of prejudice among the most prejudiced employers does not affect the wage gap, and (iii) wage gaps increase along with the share of Blacks in the labor market. The paper uses questions regarding attitudes towards a Black president and support for interracial marriage from the General Social Survey (GSS) combined with wage data from the Current Population Survey (CPS). States with higher shares of Black workers do indeed have larger wage gaps. The amount of prejudice held by
2
Taste-Based Discrimination
23
employers occupying the lower tail of the prejudice distribution explains this wage gap, while wages do not vary with the prejudice of the most prejudiced persons in a state. Extrapolating their results, Charles and Guryan find that around a quarter of the wage gap between Blacks and Whites in the USA is due to taste-based discrimination. Carlsson and Rooth (2016) apply the same strategy to ethnic wage gaps in Sweden. They also find that attitudes of marginal employers explain the wage gap, while the average employer does not matter. Burn (2020) combines data on attitudes toward homosexuality from the GSS and wage data from the Census and the American Community Survey (ACS) and documents a correlation between managers’ prejudice and the wage penalty paid by gay men. Other papers also leverage correlations between survey measures of prejudice and ethnic gaps. Zussman (2013) starts by showing that Jewish sellers discriminate against Arab buyers in the Israeli online market for used cars using a correspondence-study type of experiment. He also runs a survey on Jewish sellers to learn about their beliefs and attitudes towards Arabs (i.e., in a similar way to Charles and Guryan (2008)) and finds that Jewish sellers hold many intolerant views towards Arabs. Interestingly, one particular statement – “the Arabs in Israel are more likely to cheat than the Jews” – allows for determining the sellers that are going to be the most likely to discriminate against Arab buyers. The paper argues that this is consistent with statistical discrimination. One could also argue that, given its correlation with anti-Arab statements, the fact that these statement jointly explain discriminatory behavior might be evidence in favor of taste-based discrimination.
Competition and Discrimination One of the predictions of Becker’s model is that taste-based discrimination should be more persistent in markets that are less competitive. Comparing markets with crosssectional data cannot identify the effect of competition on wage gaps: most of the literature leverages changes in competition within markets. Ashenfelter and Hannan (1986) correlate female employment with measures of market concentration in the US banking sector. The paper innovates by using firmlevel data as an attempt to tackle the issue that cross-industry comparisons pick up differences other than just those regarding market concentration. Berkovec et al. (1998) test whether or not mortgage lenders discriminate when they evaluate mortgage loans. If taste-based discrimination is involved, discrimination should increase in less competitive lending environments. They fail to reject the null hypothesis that no taste-based discrimination exists. Again with regards to the credit market, Cavalluzzo et al. (2002) document that firms owned by Blacks are more often denied credit than owned by Whites. They find that increases in competition in local banking markets reduce racial gaps in denial rates, which is consistent with some taste-based discrimination. Investigating racial gaps in the car loan market, Butler et al. (2022) combine tests of direct predictions of the Becker model and of the influence of competition. Their results support the existence of substantial tastebased discrimination in this market. Levine et al. (2014) also use the stepped-wedged nature of bank deregulation in the USA to investigate how racial gaps in the labor market are affected by the entry
24
R. Rathelot and M. Safi
of new firms. They find evidence in favor of taste-based discrimination: the wage gap between Blacks and Whites decreases when a deregulation takes place. In a similar vein, Hirata and Soares (2020) look into the impact of the trade liberalization that took place in Brazil in the 1990s on the wage gap between Black and White workers. They show that the wage gap decreased more in sectors that became more exposed to foreign competition following the reforms. The gap also decreased more in sectors that were initially more concentrated. Taken together, these results suggest that, in the USA and in Brazil, higher competition has contributed to reducing racial gaps, which is consistent with the existence of taste-based discrimination in these cases. This is not the case for gender gaps in India, however. Ghani et al. (2016) study the impact of reforms aimed at increasing competition in several markets in India on female employment. They find that these reforms are not associated with large changes in female outcomes on the labor market. Another way to study the influence of competition is to look directly at its impact on firm exit as a function of how discriminatory firms are. Weber and Zulehner (2014) use Austrian data to show that firms that start with lower shares of female workers than what is typically found in their industry have lower survival rates. Pager (2016) combines data from a large-sample correspondence study to measure discrimination at the firm level with data on establishment survival. She documents a correlation between discrimination and the probability to fail within 6 years. Using an experimental approach, Doleac and Stein (2013) measure the effect of race in the online market for second-hand electronic goods, creating fake ads that are identical but for the skin color of the hand holding the item to sell. She finds that Black sellers do particularly poorly in thinner markets, a piece of evidence consistent with taste-based discrimination.
Customer Discrimination Customer discrimination has received considerable empirical attention. The main prediction of this family of models is that occupations that are more exposed to customers should suffer from higher discrimination and exhibit higher wage gaps. Kanazawa and Funk (2001) use Nielsen ratings for the US basketball games to show that viewership increases with the share of White players on the field, controlling for other determinants of viewership. Both results point to substantial customer discrimination in that market. Nunley et al. (2015) use a correspondence study to measure racial discrimination towards recent college graduates in the USA. They find higher gaps in callback rates in jobs that require more interaction with customers. Combes et al. (2016) use observational data from France, where African immigrants are both more often unemployed in general and less often employed in sectors with higher exposure to consumers. They draw two predictions from a search model with employer and consumer discrimination: (i) if higher shares of migrants imply lower ethnic gaps, then there is discrimination (i.e., either customer or employer-based) and (ii) if higher shares of migrants are positively correlated with ethnic differentials in the probability of working in consumer-exposed sectors and if there is discrimination, then there is consumer discrimination. Note that the first prediction is at odds with Becker’s, which was tested by Charles and Guryan (2008)
2
Taste-Based Discrimination
25
but is consistent with Black (1995). They use the 1990 French Census and compare the local labor market at the commuting zone level and find support for both the existence of ethnic discrimination and the contribution of customer discrimination to ethnic employment gaps. In the context of discrimination towards Arab workers in the Israeli labor market, Bar and Zussman (2017) combine survey data and a natural experiment, which result in evidence showing that sizable customer discrimination does exist in this market.
Residual Approaches Many papers only isolate taste-based discrimination as a residual. For instance, Dobbie et al. (2018) study discrimination on the British credit market. When lenders maximize short-term profitability, they indirectly discriminate against illiquid populations for which the gap between short-term and long-term profitability is the largest. They use a test that distinguishes this source of bias from taste-based or stereotype-based discrimination in data from the UK and find empirical support for this “incentive-based” model of bias. These results indirectly show that discrimination against immigrants and older applicants is unlikely to be only explained by prejudice. Correspondence and audit studies have been providing reliable evidence for the existence and magnitude of discrimination for decades (Riach and Rich 2002; Bertrand and Duflo 2017; Neumark 2018; Pager and Shepherd 2008; Quillian and Midtboen 2021). Simple designs cannot be used directly to isolate taste-based discrimination from other potential channels. To discover the mechanism underlying discrimination, researchers resort to either heterogeneity across markets (like Doleac and Stein (2013), see supra), or more complex designs, which delineate statistical discrimination and leave taste-based discrimination as a residual. An early example is Nardinelli and Simon (1990). Controlling for players’ ability, they find that baseball cards of White players are more expensive than those of Black players. Hedegaard and Tyran (2018) hire Danish juveniles to perform a real-life task (preparing letters for a mailing) and let them choose their coworker from a pool of two different people who differ in their ethnicity (i.e., Danish-sounding versus Muslim-sounding names). Pay is based on the pair’s probability so that workers should care about choosing their coworkers carefully. They find that subjects are willing to forego an average of 8% of their earnings to avoid interacting with a coworker that has a Muslim name. Caselli and Falco (2020) use a similar design to look into ethnic discrimination of Danish households towards house-cleaning workers. They deliver leaflets that differ on cleaners’ first names, price, and number of stars attributed by fictitious previous customers. They find that Muslim-named cleaners have lower callback rates when price is the same, but lower prices allow them to bridge the gap with Danish-names workers. These papers provide clients/ coworkers with measures of productivity as an attempt to isolate the gap that would not depend on productivity, and find that subject respond to productivity proxies. However, it is difficult to rule out that at least part of productivity remains unobserved, so that part of the gap could be due to statistical rather than tastebased discrimination.
26
R. Rathelot and M. Safi
Wozniak and MacNeill (2020) use a lab experiment simulating a labor market to randomize the amount of information given about candidates to different employers. Part of the discrimination against Black candidates, exerted both by Black and non-Black employers, is due to statistical discrimination, as selection rates are affected by the perceived reliability of the information (e.g., real versus stated performance scores) provided by candidates. The study pins down a residual part of discrimination that remained even after employers were given real performance scores. This experiment provides indirect but solid evidence in favor of taste-based discrimination (only exerted by non-Black employers) in this context. In the context of online auctions, von Essen and Karlsson (2019) randomize the moment at which bidders observe the name of the seller (i.e., either at the beginning of the bidding process or after the auction has been concluded). Winning bidders tend to provide feedback less often to sellers with foreign-sounding names in cases when names are revealed at the end of the auction rather than at the start. This is consistent with tastebased discrimination, as it indicates that prejudiced bidders exit the auction earlier when names are visible. Other papers leverage natural experiments. In Israel, testers and candidates for driving tests are randomly assigned to one another. Bar and Zussman (2019) document that testers are more likely to grant a driving license to candidates of the same ethnicity as them (i.e., Arab versus Jewish) and of the opposite sex. Lavy et al. (2018) use the random allocation of matriculation exams to graders in order to study the effect of students’ religiosity on the grades they receive from graders. They find some evidence for in-group religious bias, mostly driven by religious men, but the magnitude of these effects are small.
Opening the Black Box of Taste-Based Discrimination: Interdisciplinary Insights From Taste to Prejudice Becker conceives prejudice as an aversion or disutility that some people perceive during cross-group interactions. The use of the word taste reflects Becker’s understanding of discrimination as being rooted in individual preferences. His work does not speak much to the sources or the specific content of these preferences. The model only requires them to be utility costs, and Becker focuses mostly on the consequences for market-level outcomes of the presence of these costs. This section traces back the origins of the notion of taste and explores the literature on the sources of prejudice in interdisciplinary perspectives. A few years before Becker publishes his book, a rich literature on racial inequality emerges in sociology and social psychology. Becker mentions this literature in his book and presents his own contribution as complementary. The seminal reference of this literature is Gordon Allport’s book “The nature of prejudice.” The notion of taste used by Becker follows the concept of prejudice defined in Allport’s seminal work (1954). “Ethnic prejudice is an antipathy based upon a faulty and inflexible
2
Taste-Based Discrimination
27
generalization. It may be felt or expressed. It may be directed toward a group as a whole, or toward an individual because he is a member of that group” (p.10). This early definition stresses the duality of the notion of prejudice: both its affective (antipathy) and its cognitive dimensions ( faulty generalization). Allport refers to these two components of prejudice using words such as “attitudes” and “beliefs” (p.12). This conceptualization of prejudice echoes both taste-based and statistical discrimination theories in economics. Unlike Becker, Allport is interested in the “roots of prejudice” and its individual and contextual determinants. Allport’s definition also suggests that the cognitive dimension precedes the emotional one; prejudice entails negative attitudes based on (erroneous) beliefs. Allport’s definition of prejudice has been implemented in surveys that aim to measure emotions or beliefs towards groups using a wide battery of questions often labeled as “attitudinal.” Such questions measure antipathy, discomfort, hostility, and animosity against minorities based on gender, race, ethnicity, migration status, sexual orientation, etc. They also capture people’s beliefs regarding the members of these groups. For example, questions like the following, from the European Social Survey, capture subjective feelings: “Now thinking of people who have come to live in [country] from another country who are of a different race or ethnic group from most [country] people. [..] please tell me how much you would mind or not mind if someone like this married a close relative of yours.” Other questions that ask individuals to report their level of agreement with statements such as “It is usually better for everyone involved if the man is the achiever outside the home and the woman takes care of the home and family,” from the General Social Survey, typically measure self-declared beliefs. Some questions are used to gather largescale internationally comparative data such as with the World Value Survey or the European Social Survey, while others are specifically adapted to national contexts, like with the modern racism scale in the USA (McConahay 1986). These measures are shown to vary across personality traits, socioeconomic, demographic, and geographic factors, and to correlate with political attitudes and electoral behaviors. For example, support for Donald Trump in the USA has consistently been shown to be associated with race-, gender-, religion-, sexual orientation-based measures of prejudice (Shook et al. 2020). One important challenge of this type of research is to adapt these measurement instruments over time when there are intense shifts in the social desirability for expressing prejudice in surveys (Bonnet et al. 2016).
From Varieties of Prejudice to Varieties of Discrimination In Becker’s model, prejudice only varies in intensity. Other disciplines, however, stress that the type of prejudice is also important in shaping intergroup relations and human interactions.
Implicit and Explicit Prejudice Following changes in the legal and cultural contexts of prejudice in modern societies (e.g., the US antidiscrimination legislation in the 1960s), distinctions started to be
28
R. Rathelot and M. Safi
made between blatant or explicit bias versus what has been referred to as subtle, benign, or modern bias (Fiske 2002; Bobo et al. 2012). Since the 1990s, the literature has distinguished between conscious or explicit and unconscious, automatic, or implicit bias. Many studies working on different contexts show that implicit and explicit measures are correlated: consumption decisions (Maison et al. 2004), employment discrimination (Ziegert and Hanges 2005), teachers’ discrimination (Carlana 2019; Alesina et al. 2018), and discrimination in academic committees (Régner et al. 2019). The literature also highlights their distinctive underlying mechanisms. Explicit attitudes/conscious bias entail expressing aversion and endorsing negative beliefs and ideologies, while subtle/unconscious/implicit bias tends to be more associated with deeply rooted negative emotions stemming from early childhood and socialization. Phelps et al. (2000) find that implicit (but not explicit) bias measures positively covary with activation of the amygdala (i.e., the brain structure associated with your ability to feel certain emotions and perceive them in others) for Whites exposed to photos of Black persons. Some findings also suggest that changes in implicit attitudes may depend on emotional reconditioning, whereas changes in explicit attitudes may depend on more cognitive and motivational factors (Rudman et al. 2001). Implicit attitudes are also shown to be related to early intergroup experience and sociocultural background; they represent a trace of past experience that lingers in the unconscious (Greenwald et al. 2002; Rudman 2004).
Types of Emotions Behind Prejudice Beyond the different levels of expression of prejudice, research has mapped the complexity of human feelings involved in intergroup relations (Cuddy et al. 2007). The literature distinguishes between different types of “antipathy”: disgust, resentment, fear, and anger. These different emotions predict different discriminatory outcomes: disgust implies avoidance and neglect, while fear and anger are associated with more exclusionary and aggressive reactions. For instance, anger-based emotions experimentally provoked by threatening masculinity have been shown to give rise to compensatory dominance that may often take the form of violence and harassment of women as well as physical aggression (Vescio and Schermerhorn 2021). Some evidence also shows that anger is the most decisive type of emotion driving the vote for the French far-right party in reaction to the 2015 Paris terror attacks (Vasilopoulos et al. 2019). This variation in the socio-emotional evaluations at stake in intergroup relations advocates for a wide range of prejudice measurements in survey data. For example, GSS data has consistently shown that few Whites embrace African Americans on an emotional level (i.e., feeling sympathy or admiration for Blacks) despite clear decreasing trends in the expression of explicit racism (Bobo et al. 2012). Emotions involved in potentially discriminatory interactions need not be negative or antipathetic. Moderate bias often consists of withholding positive emotions from out-groups, as expressed in the form of basic liking, respect, or “cool neglect” (Fiske 2002). Intergroup bias is relative rather than absolute: if positive emotions towards the in-group are more intense than the positive emotions towards the out-group, this
2
Taste-Based Discrimination
29
discrepancy may be the source of discrimination. Moreover, specific positive emotions such as being patronizing can be associated with durable discrimination, as shown by the rich literature on benevolent sexism (Glick et al. 2000). Research also highlights strong implicit and positive bias towards in-group members (Lowes et al. 2015; Portmann 2020).
The Stereotype Content Model If prejudice is variable in its nature, intensity, and degree of expression, is it still possible to posit a single theory of taste/prejudice-based discrimination? Research on the stereotype content model (SCM) offers a powerful synthesis. In continuity with Allport’s original insights, which consider emotions and beliefs in tandem, Suzan Fiske and colleagues have delved further into the different varieties of prejudice to uncover their stereotypical contents. The concept of stereotype, defined in social psychology as the mental tendency to perceive an individual as being a representative member of the group, is a central faculty of the brain used to navigate social diversity and complexity (Hamilton and Trolier 1986). This literature documents how, despite the wide diversity across groups and societies, people invariably stereotype others along two fundamental dimensions: warmth and competence (Fiske et al. 2002; Durante et al. 2017). The first dimension pertains to the degree to which group members are depicted as “likable” and “friendly,” while the second refers to the degree to which group members are perceived as “capable” and “assertive.” The model builds on evolutionary grounds: warmth and competence are fundamental dimensions of how one perceives others because humans are predisposed to judge strangers’ intentions to harm or help (warmth) and their capacity to act in line with their intentions (competence) (Fiske 2002). Despite these evolutionary bases, the SCM endorses a conceptualization of warmth and competence as representing “cultural beliefs” about each group. They are learned early in life and activated automatically in social interactions. Measuring the contents of stereotypes implies asking individuals about the way specific groups are portrayed by society. In that sense, it is different from measuring individuals’ own attitudes and beliefs about groups. For example, Fiske et al. (2002) ask the following question to capture different groups’ stereotypical levels of competence: “As viewed by society, how competent/confident/independent/competitive are members of this group? We are not interested in your personal beliefs, but in how you think they are viewed by others.” In addition to providing a synthetic model, the SCM is also used as a prediction tool, which can be used to scrutinize intergroup relations. The two orthogonal dimensions of stereotypes lead to a four-dimensional prejudice matrix comprised of envying the up, scorning the down, pitying the down, and in-group pride (Fiske 2011). The model also predicts that groups low on both warmth and competence dimensions tend to be the most discriminated against and subject to “dehumanization.” In many other situations, stereotyping tends to be ambivalent, which may paradoxically render its effects all the more pernicious; stereotype ambivalence is shown to correlated with greater levels of social inequality and group conflict within societies (Durante et al. 2017). This model has been proven to apply to a variety of
30
R. Rathelot and M. Safi
intergroup perceptions (Cuddy et al. 2008), whether targets are elderly people (Cuddy et al. 2005), sexual and gender minorities (Clausell and Fiske 2005; Vaughn et al. 2017), female subgroups (Eckes 2002), racial minorities (Fiske et al. 2009; Lin et al. 2005), or immigrant populations (Kotzur et al. 2019; Lee and Fiske 2006). Recent use of the model seems to indicate that it can also help us to understand attitudes towards categorical intersections (e.g., gender-sexual orientation or racegender) as well as towards the relationship between generic and subgroup stereotypes (e.g., immigrant vs. specific national origin, women vs. housewives). While early empirical evidence is based on the US society (Fiske et al. 2002), recent studies use the model to investigate stereotype contents worldwide (Durante et al. 2017). In addition to its predictive power on attitudes, recent research tends to directly use SCM to interpret the variety of discriminatory outcomes against different minority groups across societies, as observed through experimental design (Thomas 2018; Veit et al. 2021).
Where Does Taste Come From? Investigating the Roots of Prejudice Prejudice and Social Cognition Where does prejudice come from? Allport’s answer is that “erroneous generalization” and “hostility” are a “natural and common capacity of the human mind” (Allport 1954, p. 17). The first source of prejudice would be embedded in humans’ faculty of perceiving themselves and others as members of groups. Group categorization is constitutive of our social nature. It is fundamental to human cognition and applies to the ways in which we perceive and process the complex information that we receive from our environment. The literature on social categorization documents the fact that individuals have more positive affects towards members of their own group, even in experimental designs that exclude mechanisms such as competition for resources. Research on the minimal group paradigm shows that in-group favoritism emerges rapidly after the “experimental” and arbitrary construction of groups and highlights cognitive mechanisms that trigger favoritism (Tajfel 1969). Individuals retain more detailed information about in-group versus out-group members and better remember how in-group members are similar to the self. The cognitive bases of in-group favoritism are often interpreted as beneficial to norm enforcement. Research on implicit attitudes provides some evidence that the source of in-group favoritism is rooted in selfesteem and cognitive consistency (Greenwald et al. 2002). In contrast with in-group favoritism, negative attitudes towards the out-group tend to be more sensitive to the context of intergroup interactions, e.g., competition, demographic, and cultural threat, or contact (Dovidio et al. 2010). Nonetheless, there is some evidence of negative attitudes towards minority group members even in the absence of competition. This is interpreted in terms of the cognitive salience of distinctive characteristics; distinctive features tend to receive enhanced processing, are more memorable, and exert more influence on judgments. Smaller groups are more distinctive and negative behavior is easier to remember when it comes from
2
Taste-Based Discrimination
31
out-groups, which may explain the persistence of negative feelings and attitudes towards them (Bodenhausen and Richeson 2010; Hamilton and Sherman 1989). On the other hand, evolutionary approaches tend to interpret negative emotions towards small-group members as reflecting cognitive adaptation aimed at avoiding “poor social exchange partners” and joining “cooperative groups,” especially in the research on health stigma and diseases (Kurzban and Leary 2001). What implications does this literature have on taste-based discrimination theory? First, it means that taste-based discrimination may be observed whenever there are processes of “meaningful categorizations.” Group labeling – whether inherited, imported, or constructed – is at the core of taste-based discrimination. So rather than group A having specific emotions towards group B, it is in the very existence of groups A and B as meaningful categories in human classification that the roots of taste-based discrimination lie. For example, research shows how legal categories of migrants (such as refugee and asylum seekers, family reunion migrants, undocumented migrants) although constructed by the law, may trigger social categorization dynamics and lead to different levels of prejudice and unequal treatment across these categories (Wyszynski et al. 2020). Second, the taste-based discrimination theory does not ascertain whether taste emerges from in-group favoritism or from out-group antipathy. Comparing levels of discrimination across different minority groups, several studies suggest the existence of a hierarchy of aversion towards different out-groups (Di Stasio and Larsen 2020; Quillian et al. 2019). At the same time, research designs that allow to compare hostility toward specific minority groups to aversion to an out-group whose origin is not identified suggest that in-group preferences are also at play (Jacquemet and Yannelis 2012).
Bringing the Context in: The Sociohistorical Roots of Prejudice While social categorization is a fundamental source of intergroup prejudice, it is not capable of explaining why some groups in some contexts are disliked, stigmatized, and sometimes dehumanized. As research in history and sociology shows, prejudice is inscribed in historical and cultural contexts and tends to be instilled during early childhood and transmitted from one generation to the next. From a historical point of view, group categorization dynamics arise in specific contexts of intergroup relations (Tilly 2005). For instance, group encounters (e.g., newcomers and old settlers), imposition (e.g., police category, administrative category), negotiation (e.g., gangs, peer groups, public debates), and transfers across social settings (e.g., from the family to the workplace, or from one society to another). In these specific contexts of group interactions, social meanings of human differences are sometimes created collectively and most often transformed and altered. For example, categorizations related to humans’ phenotypic traits are variable in their salience and specific contents across historical periods and cultural settings. This is valid for skin color, eye color, weight, and many other phenotypic characteristics. A specific conception of race – which attributes a hierarchical moral worth to specific phenotypes – is relatively recent (nineteenth century) and is related to the scientific and ideological context in which Social Darwinism emerged in the West (Banton 1998). For these categories to become durable and meaningful in
32
R. Rathelot and M. Safi
routine interactions (Tilly 1998), stable correlations with market-based outcomes are key. Sociologists refer to this process as the transition from symbolic categories to social categories (Lamont and Molnar 2002). This distinction invites us to anchor prejudice in sociohistorical contexts. Social psychologists have delved into the micro mechanisms involved in the reproduction of categorical inequality. The key idea here is that human cognition of the “social structure” is fundamentally hierarchical: the social groups that are perceived as occupying the top of the hierarchy not only enjoy better material conditions but also higher status. Status refers to the moral and symbolic value conferred to a social category because of its position in the social hierarchy. Social psychologists document how several forms of status bias affect behavior (Ridgeway 2013). While preferences for associating with in-group members are widespread, status bias more specifically refers to preferences for associating with high-status groups, which lead to the reproduction of inequalities (e.g., if men and women prefer to work for male bosses, or black and white families prefer to live in white neighborhoods). Status bias is also at stake when high-status groups react to some behaviors or situations that they perceive as threatening their group position or as “going too far.” For example, assertive, dominant women may be disliked and judged as being less employable. Status bias implies that past inequality (i.e., stemming at least partly from discrimination) is capable of sustaining prejudiced meanings of categorizations, which are likely to generate future discrimination. Payne et al. (2019) illustrate the legacy of historical discrimination and provide evidence that corroborate its social and psychological determinants. They find that the US states that were more dependent on slavery before the Civil War display higher levels of pro-White implicit bias today, among White residents, and less pro-White bias among Black residents. The association between slave populations and implicit bias is partially explained by measures of structural inequalities such as poverty, segregation, and social mobility across racial groups. History also matters to reduce prejudice. For example, Schindler and Westcott (2020) show that individuals in areas of the UK where more black troops were posted during World War II are more tolerant towards minorities 60 years after the last troops left (i.e., as measured by electoral outcomes as well as explicit and implicit attitudes towards blacks). Increasing contact with minorities within a collaborative context seems to explain the reduction in prejudice. These findings highlight the importance of history in determining current-day levels of prejudice. If the past is somehow still with us when it comes to prejudice, a key mediating mechanism lies in the intergenerational transmission of aversion and antipathy towards stigmatized groups. Research in psychology finds stable associations of prejudice with personality dispositions, such as social dominance orientation and right-wing authoritarianism, that tend to be transmitted from parents to children. Research on the development of prejudice during childhood is valuable in this regard and highlights the importance of family transmission (Duriez and Soenens 2009). While the mother’s role has been documented as crucial, recent research tends to stress the role of fathers on prejudice related to gender and sexual orientation (O’Bryan et al. 2004). Moreover, sons seem more receptive to parental prejudice
2
Taste-Based Discrimination
33
as compared to daughters: they are particularly receptive to their fathers’ attitudes (as shown by Avdeenko and Siedler (2017) on prejudice against migrants). Qualitative and ethnographic research documents the socialization process during which such prejudiced dispositions are transmitted through narratives and ideologies, as well as through everyday practices and behaviors. Intergenerational transmission may also sustain shocks that reduce prejudice. Cultural shifts such as those driven by women’s emancipation movements or LGBTIQ struggles have a long-lasting impact because they are transmitted intergenerationally. This also seems to hold in the case of structural shocks, such as the one described in the case of Black GIs (Schindler and Westcott 2020). Using cohort analyses, the authors show that intergenerational transmission is probably the core mechanism explaining the significance of the association between the presence of black troops and implicit and explicit bias 60 years after this shock in intergroup contact.
Back to Taste-Based Versus Statistical Discrimination: The Role of Information and Accuracy of Beliefs The economics of discrimination has developed around the duality between two theories, taste-based and statistical discriminations. Taste-based discrimination is grounded on the idea that members of the discriminating group experience a disutility when they interact with individuals from the group that is discriminated against. Conversely, statistical discrimination theory is based on the premise that employers or customers have beliefs about the productivity, service quality, or any other out-group members’ characteristic relevant to the social interaction that is considered. Classic papers underpinning the statistical discrimination literature assume these beliefs to be correct, in the sense that they are equal to the actual conditional distribution of the feature of interest in each group. These two theories have traditionally been opposed – taste-based discrimination being deemed irrational and statistical discrimination rational. Recent literature in economics considers the case in which priors are incorrect. Arnold et al. (2018) show suggestive evidence in the context of gaps in bail decision in the USA for Black versus White defendants, relying on the overrepresentation of black defendants showing up in the right-hand tail of the risk distribution. Bohren et al. (2019) provide more direct evidence using a field experiment to explain the gender gap in the evaluation of content posted on Github. The existence of previous positive evaluations makes women’s posts preferred over men’s, while the opposite occurs when posters have received no prior evaluations. In the context of ethnic discrimination towards Airbnb hosts in North America and Europe, Laouénan and Rathelot (2021) study the evolution of prices posted for a property as the number of reviews left by previous users increases. They find evidence in favor of statistical discrimination, half of which could be attributed to inaccurate beliefs. This literature reveals some similarities between statistical discrimination with inaccurate beliefs and taste-based discrimination. The idea of inaccurate beliefs is close to the conceptualization of prejudice, which raises the question of the sources of these inaccurate
34
R. Rathelot and M. Safi
beliefs/prejudice. The emerging literature on stereotypes and its increasing use in economics paves the way to the reconciliation between statistical and taste-based discrimination theories. First, this literature shifts the debate from the accurate/ inaccurate beliefs distinction to the study of the distortions that are grounded in social cognition. In other words, what really matters for discriminatory outcomes is not what the social reality is but rather how people perceive the social reality. For example, recent evidence suggests that while most stereotypes tend to be correct directionally, they have many inaccurate aspects related to underlying cognitive processes: misrepresentations of the social reality are driven by the disproportionate weight of the tales of the distribution in people’s beliefs constitution, group accentuation effects, and the effect of distinctiveness and group representativeness (Bordalo et al. 2016). These sources of inaccuracy are constitutive of stereotypes. They affect both discrimination and self-stereotyping which render them powerful in explaining a large range of gaps. Hence, inaccuracy of information about groups is anchored in the cognitive processes that are at the sources of stereotypes formation. Moreover, research on stereotype also helps bridge the gap between accurate and inaccurate beliefs on the long run. The literature on stereotype threat shows, for example, that salient stereotypes (including erroneous ones) may become accurate ex-post because people are psychologically driven to conform to them (Steele 1997). Recent research in economics suggests that conforming to stereotypes does not only stem from the psychological threat felt because of exposure to hostility, but also from “avoidance” and lack of “positive interactions,” even when minority and majority group members enter into contact (Glover et al. 2017). Research on the contents of stereotype also helps connect statistical and tastebased discrimination theories. Taste-based discrimination can be related to the stereotypical beliefs about warmth as described in the SCM model. In addition, and quite in line with Becker’s insights, competition is a key contextual variable that determines warmth. People attribute warmth to those perceived to be harmless – in the sense that they are not seen as being in direct competition with the in-group for jobs, school admissions, power, and resources. Moreover, the SCM model shows that perception of group-competence, the second component of the model, derives from the real state of the distribution of resources, or what the literature refers to as the social structure, somehow in line with the standard statistical discrimination theory. There is nonetheless a crucial difference: while statistical discrimination is concerned by the beliefs about groups’ productivity, the SCM predicts that it is status (i.e., the symbolic position in the social hierarchy) that triggers stereotypical beliefs about competence. There is hence no need to think that these beliefs are grounded in “real” differences in productivity (or other type of quality-related characteristics). In other words, people attribute competence to those perceived as holding prestigious jobs and being economically successful and not to those “known” to be more “productive.” Hence, the study of the contents of stereotypes highlights the coexistence of taste and statistical sources of discrimination in the stereotypical beliefs about groups. Finally, the focus on the role of stereotypes and their specific contents invites research on discrimination to pay specific attention to the sources of prejudice
2
Taste-Based Discrimination
35
described in the former section. Stereotypes are not only cognitive objects; they are cultural constructions that are shaped by social norms and ideologies. They connect back to power relations inherited from the past and can carry emotional dispositions transmitted from one generation to the next and socialized within different spheres of social closure.
Towards an Integrated Approach to Discrimination: A Multidisciplinary Perspective The previous sections review studies belonging to different disciplinary traditions, which can be used to build a framework to understand the determinants of prejudice. Consider the prejudice held by an assessor belonging to group A towards members of group B. Stereotypical beliefs about competence depend on the perceived socioeconomic position of group B (i.e., status), which depends on current and past prejudice held towards this group, but can also be distorted by cognitive processes underlying stereotype formation. From an economic point of view, it is tempting to model the evolution of competence as a posterior distribution of quality (or productivity) under imperfect information. However, social psychology stresses that stereotypes are conceptualized less as individual beliefs on group B than beliefs about society’s beliefs on group B. For this reason, competence might be particularly sluggish and lagging behind group B members’ actual improvements in socioeconomic status. Warmth can be thought to be even more persistent, as it is mostly about historical and intergenerational transmission. Beyond stereotypes, one needs to specify the relationship between stereotypes and prejudice. While stereotypes may be shared in society, members of different groups do not interiorize them in the same manner: they do not form the same prejudice towards group B. Social psychology has less to say about this mapping, its determinants, and its potential persistence. Sociologists and political scientists tend to highlight that this mapping has to do with the formation and persistence of narratives, ideologies that continue to feed the definition of in-groups and out-groups in society. They also tend to insist the effect of the context of social relations and the degree of separation between groups. More intense contact may speed up the updating of both competence and warmth stereotypes. If group B’s actual socioeconomic status in society is below other groups, contact needs not improve competence. Drawing on this framework and the body of literature, it seems to us that many questions are still at least partially open. • To what extent do individual level variables (e.g., personality traits, family background, and education) affect the mapping between stereotypes and prejudice? • What is the long-term relationship between education and stereotype contents?
36
R. Rathelot and M. Safi
• What role does the meso-context play in alleviating the role of stereotypes as sources of information (e.g., discussions with other people/colleagues during an evaluative decision, debate with people holding different opinions)? • To what extent do perceived socioeconomic gaps mediate the relationship between contact and prejudice? Understanding the determinants of prejudice may help with the evaluation of policy avenues to fight taste-based discrimination and reduce prejudice. • Policies that aim to increase contact between groups in society may only work if actual socioeconomic gaps are lower than those perceived (especially by the majority group). • Enhancing cultural shifts (destigmatization of disadvantaged groups by collective actors such as the medias, politicians, etc.) may be more effective on the long run to fight against the stickiness of the warmth component of stereotypes. The next section focuses on the literature that evaluates policies and interventions against taste-based discrimination.
How to Design Public Policies to Reduce the Consequences of Taste-Based Discrimination? Three kinds of policies have been explored to reduce taste-based discrimination: (1) reducing prejudice at the individual level, (2) implementing procedures that limit the role of prejudice in decision-making (e.g., hiring, promotion, renting, selling, etc.), and (3) relying on key actors (e.g., the media, politicians, artists) to alter the processes that feed prejudice.
Prejudice Reduction Interventions How to design interventions that curb individual prejudice? Normative approaches such as blaming, sanctioning, or even outlawing are often used, but their impact in changing people’s attitudes seems quite limited. These approaches tend to reduce the expression of prejudice rather than people’s feelings or implicit attitudes, and they mainly do so when the sanction comes from “credible” and “valued” in-group members (measured, for instance, by the number of followers on Twitter, as in Munger 2017). Messages endorsed by elite in-group members that prime common religious identity with the out-group are the most consistently effective way to reduce the spread of sectarian hate speech (Siegel and Badaan 2020). Conversely, prominent public figures who endorse negative stereotypes about minority groups may foster anti-minority attitudes, even within the least prejudiced people (Bursztyn et al. 2020). Paradoxically, the effectiveness of normative approaches is related to the
2
Taste-Based Discrimination
37
thick nature of group boundaries and mediated by mechanisms that enhance social closure and in-group favoritism, rather than reducing out-group antipathy per se. As an alternative to normative approaches, social psychologists and behavioral economists have designed and tested interventions that aim specifically at increasing the likability of stigmatized group members. “Emotional training,” which consist in, e.g., simulation encounters or guided meditation techniques are implemented to trigger emotion-regulation strategies as to help individuals fight off their personal implicit and explicit attitudes (Paluck et al. 2021). Evidence suggests that “generating empathy” is one of the most promising avenues for countering the antipathy that is at the source of prejudice. Several experimental designs drawing on this idea have been tested: making connections with other groups (priming individual history and asking people about their family migration history as a way to reduce antiimmigration attitudes, as in Williamson et al. 2020), perspective taking (asking people to put themselves in the shoes of a refugee or a transgender person, as in Broockman and Kalla 2016; Adida et al. 2018; Kalla and Broockman 2020), increasing awareness (revealing bias in the form of feedback after an implicitassociation test as in Alesina et al. (2018), or by the dissemination of scientific findings on biases and discriminatory decisions as in Pope et al. (2018)). These studies measure direct effects on socio-emotional judgments reported in explicit or implicit attitudes. However, most of them use one type of intervention, show shortterm impacts, and are specific to particular contexts. Another type of interventions focuses on increasing social interactions between groups (Pettigrew and Tropp 2006). According to Paluck et al. (2021), interventions based on promoting contact account for one-third of all prejudice reduction research. Findings suggest that intergroup contact correlates with lower levels of intergroup prejudice. Causal assessment is yet rare, and there is a wide variation in the effects across the type of prejudice (e.g., racial versus gender) (Paluck et al. 2019). Paluck and colleagues note a shift in this literature from a focus on increasing contact within specific institutional contexts (e.g., workplaces, schools, neighborhoods) to new interventions based on enhancing real or imagined interpersonal contact. Recent studies use random assignment to groups in sports teams (Mousa 2020; Lowe 2021), classrooms (Scacco and Warren 2018; Rao 2019), student housing (Corno et al. 2019), military service (Bagues and Roth 2020; Dahl et al. 2020; Carrell et al. 2019), and other settings to provide direct evidence on the interpersonal contact effect. The impact of contact can be mediated by affective pathways (e.g., increasing empathy and decreasing anxiety as stressed by Pettigrew and Tropps’ first meta-analyses) or through cognitive and cultural mechanisms (e.g., enhancing knowledge regarding the outgroup; Boisjoly et al. (2006); Paluck et al. (2021); Bursztyn et al. (2021)). Research on gender roles also highlights the effect of counter-stereotypical contact in reducing the prevalence of stereotypes (Breda et al. 2020). Confronted with the wide heterogeneity of findings across research designs, social psychologists have implemented the interventions that are thought to be the most effective in real-life situations. Forscher et al.’s (2017) habit-breaking package combines five components: (1) stereotype replacement (i.e., identify a stereotype and consciously replace it with accurate information), (2) individuating (i.e., gather
38
R. Rathelot and M. Safi
specific information about the individual in a way to prevent group stereotype from affecting decisions), (3) counter-stereotypical imaging (i.e., positively imagining an effective minority candidate fulfilling a position before the process of evaluation), (4) perspective taking, and (5) increasing opportunity for contact. Forscher and colleagues show that the habit-breaking package tends to change explicit and implicit attitudes. These effects seem to be more durable than those measured for single interventions (they find significant effects up to 2 months after the intervention). The durability of the effect mainly pertains to measures such as awareness, concern, and effort about discrimination, while effects on implicit biases seem to fade rapidly (Lai et al. 2016). An important limitation of the literature on prejudice reduction lies in the lack of clarity on the ways in which it connects back to research on discrimination. Most studies focus on the effects of interventions on attitudes – whether explicit, implicit, or both – as proxies for prejudice. The extent to which reducing prejudice alleviates, in its turn, discrimination is rarely assessed. The rare studies that tackle both attitudes and behavior tend to challenge the relevance of reducing prejudice as an avenue for reducing discrimination. For example, Chang et al.’s (2019) experiment on diversity training in the US firms shows a positive effect on attitudes towards women but not on behavioral outcomes (such as promotion some weeks later). Some studies show that significant effects on behavior are only measurable for the least prejudiced agents, at least regarding explicit attitudes (Alesina et al. 2018). Finally, research on attitudes consistency suggests that changes in behavior may provoke an ex-post change in attitudes because people try to shape their attitudes so that they become consistent with their behavior (Olson and Stone 2005).
Reducing the Scope of Prejudice in Decision-Making Instead of addressing prejudice in individuals, another family of interventions aims at implementing rules and procedures that reduce the role prejudice plays in decision-making. Note that this type of interventions can tackle taste-based discrimination as well as statistical discrimination. The literature in organizational sociology, management, and industrial relations has been largely informed by the experience of Equal Opportunity Policies in the USA since the 1960s (Small and Pager 2020). Dobbin et al. (2015) compare many interventions in corporations and find that those designed to control managers’ bias (e.g., mandatory diversity training, job tests, grievance systems) lead to resistance and possible backlash. This research draws on job autonomy (i.e., freedom, independence, and discretion to make decisions) and self-determination theory (i.e., the crucial role of autonomous motivation in learning and working) to explain the mixed and sometime adverse effects of initiatives that limit managerial discretion/power and consequently produce defensive reactions (Dobbin and Kalev 2016; Howell et al. 2017). Some evidence suggests that even the provision of additional information may backfire in a context where prejudice is high. In a study on the credit market, an intervention informing loan officers that have been formerly shown to discriminate against women that the
2
Taste-Based Discrimination
39
latter have higher repayment rates led the officers to exhibit even more intense gender-based discrimination (Montoya et al. 2020). Following these findings on the mixed effects of individual-level interventions, the organizational literature tends to advocate impersonal procedures such as increasing transparency and accountability as more effective methods for fighting against discrimination (Kalev et al. 2006). “Blinding strategies” are another type of policies to reduce the scope of prejudice. Following the seminal study on blind auditions in orchestras (Goldin and Rouse 2000), different interventions have been evaluated, mostly with regards to the labor market, that use procedures designed to remove many of the signals that might trigger taste-based discrimination. Hiding applicants’ names is an example of a powerful blinding strategy, since it has the potential to entirely rule out gender-, racial-, and social class-based prejudice that can be activated by this single piece of information. Qualitative research on coping with discrimination shows that some minority candidates use this kind of strategy when concealing or downplaying racial cues in their application forms (Kang et al. 2016; Safi 2017). Experiments drawing on anonymous resumes have been conducted in several countries, and they provide mixed findings. A French study suggests that anonymous resumes benefit women but can harm ethnic minorities by preventing existing affirmative-action policies from working (Behaghel et al. 2015). A Swedish study also finds that the effects of anonymous resumes tend to be more favorable for women than for ethnic minorities (Åslund and Skans 2012). At odds with the previous two studies, Krause et al. (2012) combine several trials of anonymous resumes in Germany and show that minority candidates may benefit from the blinding process, while women may suffer from it. The effects of blinding strategies seem to be heterogeneous and dependent on context (e.g., the magnitude of discrimination in the hiring process, the existence of affirmative-action or equal-opportunity policies and practices). Beyond the hiring process, blinded versus non-blinded examinations of students display heterogeneous findings that stress the importance of the initial gender composition in the educational field (Breda and Ly 2015). Finally, the evaluation of Ban-the-Box policies in the USA, which are based on blinding the criminal record to prevent employers from asking about workers’ criminal histories, shows detrimental effects on AfroAmerican job applicants. Using a correspondence study, Agan and Starr (2017) document that racial gaps in callback rates increase when Ban-the-Box policies are enacted, largely due to statistical discrimination mechanisms that associate Blackness to criminality. These mixed findings put into perspective blinding strategies as stand-alone policies and suggest the need to better integrate them into a broader strategic framework. A third strategy is to get rid of human evaluators altogether and rely on algorithm-based decisions. Hoffman et al. (2017) study the introduction of an automatic testing technology in several firms. They observe that some managers deviate from the ranking recommended by the test. Contrary to the intuition according to which human managers use discretion to incorporate superior information, they find that these deviations lead on average to hires of lower quality. This suggests that the observed managers could be due to prejudice, and that limiting the scope for manager-driven deviations could both reduce discrimination and increase firm productivity.
40
R. Rathelot and M. Safi
Decisions based on currently used algorithms, which rely on existing data to formulate predictions, can still result in sizable bias. This “algorithm discrimination,” which is non-prejudiced by definition, may be easier to both detect and fix, as compared to taste-based discrimination. Obermeyer et al. (2019) document massive discrimination against minority patients that is induced by a health-prediction algorithm that relies on health expenditure – a highly racially biased variable – as a proxy for the general state of one’s health. Their research justifies rethinking antidiscrimination policies within organizations so as to incorporate well-established insights from the organizational literature (e.g., via regulations and routine audits, increasing transparency in practices, enhancing accountability, etc.) for the purpose of improving algorithm-based decision-making. Research at the intersection of legal studies, organizational sociology, computational studies, and economics provides a promising avenue for building objective and transparent decision-making procedures for alleviating prejudice-based discrimination within organizations whenever it is possible to circumvent the need for human intervention (Kleinberg et al. 2020). Rose (2021) shows how seemingly race-neutral policies regarding technical rules (e.g., that forbid drugs and alcohol, require the payment of fees and fines, limit travel, etc.) for convicted offenders who serve their sentences under “community supervision” at home may in fact be a source of further racial disparities in the US justice system. More specifically, evidence from a reform enacted in the state of North Carolina indicates that harshly punishing offenders for the violation of technical rules does very little to promote compliance or close the Black-White gap in re-offending. While it is true that rule violations are correlated with the overall likelihood of re-offending, the predictive power of these violations is far weaker among Black probationers. Moreover, since Black probationers are more likely to violate these rules for a number of societal reasons, this reform disproportionately harms Black offenders and does nothing to close the racial gap in rearrests.
Towards Less Prejudiced Societies? While interventions such as raising awareness and consciousness, increasing contact, and implementing impersonal procedures clearly speak at the societal level, research is still lacking in clear theoretical and methodological frameworks that enable us to derive hypotheses on the possible effects of these interventions on societies as a whole and to test them empirically. As most meta-analyses document a wide variability in the effects of interventions, the conception of large-scale policies is structurally confronted with intense trade-offs. Such trade-offs are particularly salient when one considers the extrapolation of contact theory to the meso or macro level. Some studies suggest positive and lasting effects of increasing contact on intergroup relations as measured by level of intermarriage (Merlino et al. 2019) (i.e., through quasi-random variation in the share of black students across cohorts within the US schools). Others show that increasing contact may backfire when it triggers competition, as in the case of demographic change in relation to immigration in neighborhoods (Enos 2014). Recent evidence
2
Taste-Based Discrimination
41
also suggests that there is a trade-off between short- and long-term effects of diversity and contact (Ramos et al. 2019). Public policy can specifically focus on building the sociopolitical environment that will trigger the positive rather than the negative effects of such interventions. Allport’s seminal work advocates for the necessity of securing equal status and collaborative framing, while also promoting for intergroup contact (Allport 1954; Pettigrew and Tropp 2008). For example, in contrast with theories regarding the corrosive effect of ethnic fragmentation on trust and social cohesion, high levels of ethnic diversity in contexts of intense social contacts can trigger nation-state building when it is a collaborative endeavor (Bazzi et al. 2019; Bagues and Roth 2020). The literature in cultural sociology can help inform public policy’s aim to create an “atmosphere” that enhances positive effect from reducing prejudice. Building on qualitative fieldwork conducted in different national and local contexts, this literature suggests that efficient antidiscrimination policies need to tackle the “narratives” and “cultural repertoires” that feed prejudice. Some studies advocate large-scale public policies that engage in destigmatization to tackle the specific contents of stereotypes surrounding stigmatized groups (Clair et al. 2016) such as the belief that “immigrants equal the welfare state” or “black people equal violent” or “boys are better in maths.” Changing narratives, removing blame, and promoting equal interactions could be the most effective long-term strategy. This literature builds on examples like the progressive destigmatization of AIDS, cultural shifts towards gender equality, and stigma surrounding obesity. In addition to cultural and meaning-making dynamics, the effectiveness of interventions that build on the role of mass media relies on their capacity to enhance empathy and intergroup contact beyond interpersonal interactions (see, for example, the effect of the Paralympic Games as in Bartsch et al. 2018). Exposure to minoritygroup, nonfictional public figures with which one can identify oneself may also be conducive to generating empathy at the societal level (Alrababa’h et al. 2021). Emerging research also highlights the importance of “role model.” Finally, the literature on the stereotype content model suggests that emotions towards other groups are truly embedded in today’s socioeconomic inequalities and the social hierarchy that derives from them. Equalizing social position through policies that promote social mobility, equal opportunity, and diversity may lead to the reduction of antipathy towards the most stigmatized groups in the long run.
References Adida CL, Lo A, Platas MR (2018) Perspective taking can promote short-term inclusionary behavior toward Syrian refugees. Proc Natl Acad Sci 115:9521–9526, publisher: National Academy of Sciences Section: Social Sciences Agan A, Starr S (2017) Ban the box, criminal records, and racial discrimination: a field experiment*. Q J Econ 133:191–235 Alesina A, Carlana M, Ferrara EL, Pinotti P (2018) Revealing stereotypes: evidence from immigrants in schools. Tech. rep., National Bureau of Economic Research
42
R. Rathelot and M. Safi
Allport GW (1954) The nature of prejudice: 25th anniversary edition. Addison-Wesley, Cambridge, MA Alrababa’h A, Marble W, Mousa S, Siegel AA (2021) Can exposure to celebrities reduce prejudice? The effect of Mohamed Salah on islamophobic behaviors and attitudes. Am Polit Sci Rev 115: 1111–1128 Arnold D, Dobbie W, Yang CS (2018) Racial bias in bail decisions*. Q J Econ 133:1885–1932 Ashenfelter O, Hannan T (1986) Sex discrimination and product market competition: the case of the banking industry. Q J Econ 101:149–173 Åslund O, Skans ON (2012) Do anonymous job application procedures level the playing field? ILR Rev 65:82–107, publisher: SAGE Publications Inc Avdeenko A, Siedler T (2017) Intergenerational correlations of extreme right-wing party preferences and attitudes toward immigration. Scand J Econ 119:768–800, eprint: https:// onlinelibrary.wiley.com/doi/pdf/10.1111/sjoe.12190 Bagues M, Roth C (2020) Interregional contact and national identity. CAGE online working paper series 526, Competitive Advantage in the Global Economy (CAGE) Banton M (1998) Racial theories. Cambridge University Press, Cambridge, MA Bar R, Zussman A (2017) Customer discrimination: evidence from Israel. J Labor Econ 35: 1031–1059 Bar R, Zussman A (2019) Identity and bias: insights from driving tests. Econ J 130:1–23 Bartsch A, Oliver MB, Nitsch C, Scherr S (2018) Inspired by the paralympics: effects of empathy on audience interest in para-sports and on the destigmatization of persons with disabilities. Commun Res 45:525–553, publisher: SAGE Publications Inc Bazzi S, Gaduh A, Rothenberg AD, Wong M (2019) Unity in diversity? How intergroup contact can foster nation building. Am Econ Rev 109:3978–4025 Becker G (1957) The economics of discrimination. Chicago University Press, Chicago, Reprint 1971 Behaghel L, Crépon B, Le Barbanchon T (2015) Unintended effects of anonymous résumés. Am Econ J Appl Econ 7:1–27 Berkovec JA, Canner GB, Gabriel SA, Hannan TH (1998) Discrimination, competition, and loan performance in FHA mortgage lending. Rev Econ Stat 80:241–250 Bertrand M, Duflo E (2017) Field experiments on discrimination. In: Banerjee AV, Duflo E (eds) Handbook of field experiments, North-Holland, pp 309–393, vol. 1 of Handbook of Economic Field Experiments, chap. 8 Black DA (1995) Discrimination in an equilibrium search model. J Labor Econ 13(2):309–334 Bobo LD, Charles CZ, Krysan M, Simmons AD (2012) The real record on racial attitudes. In: Marsden PV (ed) Social trends in American life: finds from the general social survey since 1972. Princeton University Press, pp 38–83 Bodenhausen GV, Richeson JA (2010) Prejudice, stereotyping, and discrimination. In: Advanced social psychology: the state of the science. Oxford University Press, New York, pp 341–383 Bohren JA, Imas A, Rosenberg M (2019) The dynamics of discrimination: theory and evidence. Am Econ Rev 109:3395–3436 Boisjoly J, Duncan GJ, Kremer M, Levy DM, Eccles J (2006) Empathy or antipathy? The impact of diversity. Am Econ Rev 96:1890–1905 Bonnet F, Lalé E, Safi M, Wasmer E (2016) Better residential than ethnic discrimination! Reconciling audit and interview findings in the Parisian housing market. Urban Stud 53:2815–2833 Bordalo P, Coffman K, Gennaioli N, Shleifer A (2016) Stereotypes*. Q J Econ 131:1753–1794 Borjas G, Bronars S (1989) Consumer discrimination and self-employment. J Polit Econ 97: 581–605 Bowlus AJ, Eckstein Z (2002) Discrimination and skill differences in an equilibrium search model. Int Econ Rev 43:1309–1342 Breda T, Ly ST (2015) Professors in core science fields are not always biased against women: evidence from France. Am Econ J Appl Econ 7:53–75
2
Taste-Based Discrimination
43
Breda T, Grenet J, Monnet M, Van Effenterre C (2020) “Do female role models reduce the gender gap in science? Evidence from French high schools,” IZA discussion paper Broockman D, Kalla J (2016) Durably reducing transphobia: a field experiment on door-to-door canvassing. Science 352:220–224, publisher: American Association for the Advancement of Science Section: Report Burn I (2020) The relationship between prejudice and wage penalties for gay men in the United States. ILR Rev 73:650–675, publisher: SAGE Publications Inc Bursztyn L, Haaland IK, Rao A, Roth CP (2020) “Disguising prejudice: popular rationales as excuses for intolerant expression,” NBER Working Papers 27288, National Bureau of Economic Research, Inc Bursztyn L, Chaney T, Hassan TA, Rao A (2021) “The immigrant next door: exposure, prejudice, and altruism,” Working Papers 2021–16, Becker Friedman Institute for Research In Economics Butler AW, Mayer EJ, Weston JP (2022) Racial disparities in the auto loan market. Rev Financ Stud. forthcoming Carlana M (2019) Implicit stereotypes: evidence from teachers’ gender bias*. Q J Econ 134: 1163–1224 Carlsson M, Rooth D-O (2016) Employer attitudes, the marginal employer, and the ethnic wage gap. ILR Rev 69:227–252 Carrell SE, Hoekstra M, West JE (2019) The impact of college diversity on behavior toward minorities. Am Econ J Econ Pol 11:159–182 Caselli M, Falco P (2020) “As long as they are cheap experimental evidence on the demand for migrant workers,” Development Working Papers 466, Centro Studi Luca d’Agliano, University of Milano Cavalluzzo KS, Cavalluzzo LC, Wolken JD (2002) Competition, small business financing, and discrimination: evidence from a new survey. J Bus 75:641–679 Chang EH, Milkman KL, Gromet DM, Rebele RW, Massey C, Duckworth AL, Grant AM (2019) The mixed effects of online diversity training. Proc Natl Acad Sci 116:7778–7783 Charles KK, Guryan J (2008) Prejudice and wages: an empirical assessment of Becker’s the economics of discrimination. J Polit Econ 116:773–809 Clair M, Daniel C, Lamont M (2016) Destigmatization and health: cultural constructions and the long-term reduction of stigma. Soc Sci Med 165:223–232 Clausell E, Fiske ST (2005) When do subgroup parts add up to the stereotypic whole? Mixed stereotype content for gay male subgroups explains overall ratings. Soc Cogn 23:161–181, publisher: Guilford Publications Inc Combes P-P, Decreuse B, Laouenan M, Trannoy A (2016) Customer discrimination and employment outcomes: theory and evidence from the French labor market. J Labor Econ 34:107–160 Corno L, Ferrara EL, Burns J (2019) “Interaction, stereotypes and performance. Evidence from South Africa,” Tech. Rep. W19/03, Institute for Fiscal Studies, publication Title: IFS Working Papers Cuddy AJC, Norton MI, Fiske ST (2005) This old stereotype: the pervasiveness and persistence of the elderly stereotype. J Soc Issues 61:267–285 Cuddy AJC, Fiske ST, Glick P (2007) The BIAS map: behaviors from intergroup affect and stereotypes. J Pers Soc Psychol 92:631–648 Cuddy AJ, Fiske ST, Glick P (2008) Warmth and competence as universal dimensions of social perception: the stereotype content model and the BIAS map. Adv Exp Soc Psychol 40:61–149 Dahl GB, Kotsadam A, Rooth D-O (2020) Does integration change gender attitudes? The effect of randomly assigning women to traditionally male teams*. Q J Econ 136:987–1030 Di Stasio V, Larsen E (2020) The racialized and gendered workplace: applying an intersectional lens to a field experiment on hiring discrimination in five European labor markets. Soc Psychol Q 83: 229–250 Dobbie W, Liberman A, Paravisini D, Pathania V (2018) Measuring bias in consumer lending, working papers 623, Princeton University, Department of Economics, Industrial Relations Section
44
R. Rathelot and M. Safi
Dobbin F, Kalev A (2016) Why diversity programs fail. Harv Bus Rev 94:14 Dobbin F, Schrage D, Kalev A (2015) Rage against the iron cage: the varied effects of bureaucratic personnel reforms on diversity. Am Sociol Rev 80:1014–1044, publisher: SAGE Publications Inc Doleac JL, Stein LC (2013) The visible hand: race and online market outcomes. Econ J 123: F469–F492 Dovidio J, Hewstone M, Glick P, Esses V (2010) Prejudice, stereotyping, and discrimination: theoretical and empirical overview. In: The SAGE handbook of prejudice, stereotyping and discrimination, pp 3–28 Durante F, Fiske ST, Gelfand MJ, Crippa F, Suttora C, Stillwell A, Asbrock F, Aycan Z, Bye HH, Carlsson R, Björklund F, Dagher M, Geller A, Larsen CA, Latif A-HA, Mähönen TA, Jasinskaja-Lahti I, Teymoori A (2017) Ambivalent stereotypes link to peace, conflict, and inequality across 38 nations. Proc Natl Acad Sci 114:669–674, iSBN: 9781611874112 Publisher: National Academy of Sciences Section: Social Sciences Duriez B, Soenens B (2009) The intergenerational transmission of racism: the role of right-wing authoritarianism and social dominance orientation. J Res Pers 43:906–909 Eckes T (2002) Paternalistic and envious gender stereotypes: testing predictions from the stereotype content model. Sex Roles 47:99–114 Enos RD (2014) Causal effect of intergroup contact on exclusionary attitudes. Proc Natl Acad Sci 111:3699–3704 Fiske ST (2002) What we know now about bias and intergroup conflict, the problem of the century. Curr Dir Psychol Sci 11:123–128, publisher: SAGE Publications Inc Fiske ST (2011) Envy up, scorn down: how status divides us. Russell Sage Foundation Fiske ST, Cuddy AJC, Glick P, Xu J (2002) A model of (often mixed) stereotype content: competence and warmth respectively follow from perceived status and competition. J Pers Soc Psychol 82:878–902 Fiske S, Bergsieker H, Russell A, Williams L (2009) Images of black Americans: then, “them,” and now, “Obama!”. Du Bois review: social science research on race, 6 Flinn CJ (2015) Gary Becker’s contributions to the analysis of discrimination. J Demograph Econ 81:45–50 Forscher PS, Mitamura C, Dix EL, Cox WTL, Devine PG (2017) Breaking the prejudice habit: mechanisms, timecourse, and longevity. J Exp Soc Psychol 72:133–146 Ghani SE, Grover A, Kerr S, Kerr WR (2016) Will market competition trump gender discrimination in India? Policy Research Working Paper Series 7814, The World Bank Glick P, Fiske ST, Mladinic A, Saiz JL, Abrams D, Masser B, Adetoun B, Osagie JE, Akande A, Alao A, Brunner A, Willemsen TM, Chipeta K, Dardenne B, Dijksterhuis A, Wigboldus D, Eckes T, Six-Materna I, Expósito F, Moya M, Foddy M, Kim H-J, Lameiras M, Sotelo MJ, Mucchi-Faina A, Romani M, Sakalli N, Udegbe B, Yamamoto M, Ui M, Ferreira MC, López López W (2000) Beyond prejudice as simple antipathy: hostile and benevolent sexism across cultures. J Pers Soc Psychol 79:763–775 Glover D, Pallais A, Pariente W (2017) Discrimination as a self-fulfilling prophecy: evidence from French grocery stores*. Q J Econ 132:1219–1260 Goldberg MS (1982) Discrimination, nepotism, and long-run wage differentials. Q J Econ 97: 307–319 Goldin C, Rouse C (2000) Orchestrating impartiality: the impact of “blind” auditions on female musicians. Am Econ Rev 90:715–741 Greenwald AG, Banaji MR, Rudman LA, Farnham SD, Nosek BA, Mellott DS (2002) A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept. Psychol Rev 109:3–25 Hamilton DL, Sherman SJ (1989) Illusory correlations: implications for stereotype theory and research. In: Bar-Tal D, Graumann CF, Kruglanski AW, Stroebe W (eds) Stereotyping and prejudice: changing conceptions. Springer, New York, pp 59–82, Springer Series in Social Psychology
2
Taste-Based Discrimination
45
Hamilton DL, Trolier TK (1986) Stereotypes and stereotyping: an overview of the cognitive approach. In: Prejudice, discrimination, and racism. Academic Press, San Diego, pp 127–163 Hedegaard MS, Tyran J-R (2018) The price of prejudice. Am Econ J Appl Econ 10:40–63 Hirata G, Soares RR (2020) Competition and the racial wage gap: evidence from Brazil. J Dev Econ 146:102519 Hoffman M, Kahn LB, Li D (2017) Discretion in hiring*. Q J Econ 133:765–800 Howell JL, Redford L, Pogge G, Ratliff KA (2017) Defensive responding to IAT feedback. Soc Cogn 35:520–562, publisher: Guilford Publications Inc Jacquemet N, Yannelis C (2012) Indiscriminate discrimination: a correspondence test for ethnic homophily in the Chicago labor market. Labour Econ 19:824–832 Kalev A, Dobbin F, Kelly E (2006) Best practices or best guesses? Assessing the efficacy of corporate affirmative action and diversity policies. Am Sociol Rev 71:589–617 Kalla JL, Broockman DE (2020) Reducing exclusionary attitudes through interpersonal conversation: evidence from three field experiments. Am Polit Sci Rev 114:410–425 Kanazawa MT, Funk JP (2001) Racial discrimination in professional basketball: evidence from Nielsen ratings. Econ Inq 39:599–608 Kang SK, DeCelles KA, Tilcsik A, Jun S (2016) Whitened résumés: race and self-presentation in the labor market. Adm Sci Q 61:469–502 Kleinberg J, Ludwig J, Mullainathan S, Sunstein CR (2020) Algorithms as discrimination detectors. Proceedings of the National Academy of Sciences, publisher: National Academy of Sciences Section: Physical Sciences Kotzur P, Friehs M-T, Asbrock F, van Zalk M (2019) Stereotype content of refugee subgroups in Germany. Eur J Soc Psychol 49:1344–1358 Krause A, Rinne U, Zimmermann KF (2012) Anonymous job applications in Europe. IZA J Eur Labor Stud 1:1–20, number: 1 Publisher: SpringerOpen Kurzban R, Leary MR (2001) Evolutionary origins of stigmatization: the functions of social exclusion. Psychol Bull 127:187–208 Lai CK, Skinner AL, Cooley E, Murrar S, Brauer M, Devos T, Calanchini J, Xiao YJ, Pedram C, Marshburn CK, Simon S, Blanchar JC, Joy-Gaba JA, Conway J, Redford L, Klein RA, Roussos G, Schellhaas FMH, Burns M, Hu X, McLean MC, Axt JR, Asgari S, Schmidt K, Rubinstein R, Marini M, Rubichi S, Shin J-EL, Nosek BA (2016) Reducing implicit racial preferences: II. Intervention effectiveness across time. J Exp Psychol Gen 145:1001–1016, place: US Publisher: American Psychological Association Lamont M, Molnar V (2002) The study of boundaries in the social sciences. Annu Rev Sociol 28: 167–195 Lang K, Kahn-Lang Spitzer A (2020) Race discrimination: an economic perspective. J Econ Perspect 34:68–89 Lang K, Lehmann J-YK (2012) Racial discrimination in the labor market: theory and empirics. J Econ Lit 50:959–1006 Laouénan M, Rathelot R (2021) Can information reduce ethnic discrimination? Evidence from Airbnb. Am Econ J Appl Econ, forthcoming Lavy V, Sand E, Shayo M (2018) The surprisingly small effects of religion-based discrimination in education. Working paper 24922, National Bureau of Economic Research Lee TL, Fiske ST (2006) Not an outgroup, not yet an ingroup: immigrants in the stereotype content model. Int J Intercult Relat 30:751–768 Levine R, Rubinstein Y, Levkov A (2014) Bank deregulation and racial inequality in America. Crit Finance Rev 3:1–48 Lin M, Kwan V, Cheung A, Fiske S (2005) Stereotype content model explains prejudice for an envied outgroup: scale of Anti-Asian American stereotypes. Personality Soc Psychol Bull 31: 34–47 Lowe M (2021) Types of contact: a field experiment on collaborative and adversarial caste integration. Am Econ Rev 111:1807–1844
46
R. Rathelot and M. Safi
Lowes S, Nunn N, Robinson JA, Weigel J (2015) Understanding ethnic identity in Africa: evidence from the implicit association test (IAT). Am Econ Rev 105:340–345 Maison D, Greenwald AG, Bruin RH (2004) Predictive validity of the implicit association test in studies of brands, consumer attitudes, and behavior. J Consum Psychol 14:405–415, eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1207/s15327663jcp1404 9 McConahay JB (1986) Modern racism, ambivalence, and the modern racism scale. In: Prejudice, discrimination, and racism. Academic Press, San Diego, pp 91–125 Merlino LP, Steinhardt MF, Wren-Lewis L (2019) More than just friends? School peers and adult interracial relationships. J Labor Econ 37:663–713 Montoya AM, Parrado E, Solís A, Undurraga R (2020) Bad taste: gender discrimination in the consumer credit market. IDB Working Paper 1053 Mousa S (2020) Building social cohesion between Christians and Muslims through soccer in postISIS Iraq. Science 369:866–870 Munger K (2017) Tweetment effects on the tweeted: experimentally reducing racist harassment. Polit Behav 39:629–649 Nardinelli C, Simon C (1990) Customer racial discrimination in the market for memorabilia: the case of baseball. Q J Econ 105:575–595 Neumark D (2018) Experimental research on labor market discrimination. J Econ Lit 56:799–866 Nunley JM, Pugh A, Romero N, Seals RA (2015) Racial discrimination in the labor market for recent college graduates: evidence from a field experiment. BE J Econ Anal Policy 15: 1093–1125 O’Bryan M, Fishbein HD, Ritchey PN (2004) Intergenerational transmission of prejudice, sex role steriotyping, and intolerance. Adolescence 39:407–426, publisher: Libra Publishers Inc Obermeyer Z, Powers B, Vogeli C, Mullainathan S (2019) Dissecting racial bias in an algorithm used to manage the health of populations. Science 366:447–453 Olson JM, Stone J (2005) The influence of behavior on attitudes. In: The handbook of attitudes. Lawrence Erlbaum Associates Publishers, Mahwah, pp 223–271 Pager D (2016) Are firms that discriminate more likely to go out of business? Sociol Sci 3:849–859 Pager D, Shepherd H (2008) The sociology of discrimination: racial discrimination in employment, housing, credit, and consumer markets. Annu Rev Sociol 34:181–209 Paluck EL, Green SA, Green DP (2019) The contact hypothesis re-evaluated. Behav Public Policy 3:129–158, publisher: Cambridge University Press Paluck EL, Porat R, Clark CS, Green DP (2021) Prejudice reduction: progress and challenges. Annu Rev Psychol 72:533–560, pMID: 32928061 Payne BK, Vuletich HA, Brown-Iannuzzi JL (2019) Historical roots of implicit bias in slavery. Proc Natl Acad Sci 116:11693–11698, publisher: National Academy of Sciences Section: Social Sciences Pettigrew TF, Tropp LR (2006) A meta-analytic test of intergroup contact theory. J Pers Soc Psychol 90:751–783 Pettigrew TF, Tropp LR (2008) How does intergroup contact reduce prejudice? Meta-analytic tests of three mediators. Eur J Soc Psychol 38:922–934, place: US Publisher: John Wiley & Sons Phelps EA, O’Connor KJ, Cunningham WA, Funayama ES, Gatenby JC, Gore JC, Banaji MR (2000) Performance on indirect measures of race evaluation predicts amygdala activation. J Cogn Neurosci 12:729–738 Pope DG, Price J, Wolfers J (2018) Awareness reduces racial bias. Manag Sci 64:4988–4995 Portmann L (2020) Do stereotypes explain discrimination against minority candidates or discrimination in favor of majority candidates? Br J Polit Sci 52:501–519 Quillian L, Midtboen AH (2021) Comparative perspectives on racial discrimination in hiring: the rise of field experiments. Annu Rev Sociol 47:391–415 Quillian L, Heath A, Pager D, Midtboen AH, Fleischmann F, Hexel O (2019) Do some countries discriminate more than others? Evidence from 97 field experiments of racial discrimination in hiring. Sociol Sci 6:467–496
2
Taste-Based Discrimination
47
Ramos MR, Bennett MR, Massey DS, Hewstone M (2019) Humans adapt to social diversity over time. Proc Natl Acad Sci 116:12244–12249, publisher: National Academy of Sciences Section: Social Sciences Rao G (2019) Familiarity does not breed contempt: generosity, discrimination, and diversity in Delhi schools. Am Econ Rev 109:774–809 Régner I, Thinus-Blanc C, Netter A, Schmader T, Huguet P (2019) Committees with implicit biases promote fewer women when they do not believe gender bias exists. Nat Hum Behav 3: 1171–1179, number: 11 Publisher: Nature Publishing Group Riach PA, Rich J (2002) Field experiments of discrimination in the market place. Econ J 112: 480–518 Ridgeway CL (2013) Why status matters for inequality. Am Sociol Rev 79:1–16 Rose EK (2021) Who gets a second chance? Effectiveness and equity in supervision of criminal offenders. Q J Econ 136:1199–1253 Rudman LA (2004) Sources of implicit attitudes. Curr Dir Psychol Sci 13:79–82, publisher: SAGE Publications Inc Rudman LA, Ashmore RD, Gary ML (2001) “Unlearning” automatic biases: the malleability of implicit prejudice and stereotypes. J Pers Soc Psychol 81:856 Safi M (2017) Promoting diversity in French workplaces: targeting and signaling ethnoracial origin in a colorblind context. Socius 3:2378023117728834 Sasaki M (1999) An equilibrium search model with coworker discrimination. J Labor Econ 17: 377–407 Scacco A, Warren SS (2018) Can social contact reduce prejudice and discrimination? Evidence from a field experiment in Nigeria. Am Polit Sci Rev 112:654–677 Schindler D, Westcott M (2020) “Shocking racial attitudes: black G.I.s in Europe,” SSRN scholarly paper ID 3098282. Social Science Research Network, Rochester Shook NJ, Fitzgerald HN, Boggs ST, Ford CG, Hopkins PD, Silva NM (2020) Sexism, racism, and nationalism: factors associated with the 2016 U.S. presidential election results? PLoS One 15: e0229432, publisher: Public Library of Science Siegel AA, Badaan V (2020) # No2sectarianism: experimental approaches to reducing sectarian hate speech online. Am Polit Sci Rev 114:837–855 Small ML, Pager D (2020) Sociological perspectives on racial discrimination. J Econ Perspect 34: 49–67 Steele CM (1997) A threat in the air: how stereotypes shape intellectual identity and performance. Am Psychol 52:613–629, place: US Publisher: American Psychological Association Tajfel H (1969) Cognitive aspects of prejudice1. J Soc Issues 25:79–97, eprint: https://spssi. onlinelibrary.wiley.com/doi/pdf/10.1111/j.1540-4560.1969.tb00620.x Thomas K (2018) The labor market value of taste: an experimental study of class bias in U.S. employment. Sociol Sci 5:562–595 Tilly C (1998) Durable inequality. University of California Press, Berkeley Tilly C (2005) Historical perspectives on inequality. In: Romero M, Margolis E (eds) The Blackwell companion to social inequalities. Blackwell Publishing, Malden, pp 15–30 Vasilopoulos P, Marcus GE, Valentino NA, Foucault M (2019) Fear, anger, and voting for the far right: evidence from the November 13, 2015 Paris terror attacks. Polit Psychol 40:679–704 Vaughn AA, Teeters SA, Sadler MS, Cronan SB (2017) Stereotypes, emotions, and behaviors toward lesbians, gay men, bisexual women, and bisexual men. J Homosex 64:1890–1911 Veit S, Arnu H, Di Stasio V, Yemane R, Coenders M (2021) The “big two” in hiring discrimination: evidence from a cross-national field experiment. Personal Soc Psychol Bull 48:167–182 Vescio TK, Schermerhorn NEC (2021) Hegemonic masculinity predicts 2016 and 2020 voting and candidate evaluations. Proc Natl Acad Sci 118, publisher: National Academy of Sciences Section: Social Sciences von Essen E, Karlsson J (2019) The effect of competition on discrimination in online markets – anonymity and selection. PLoS One 14:1–18
48
R. Rathelot and M. Safi
Weber A, Zulehner C (2014) Competition and gender prejudice: are discriminatory employers doomed to fail? J Eur Econ Assoc 12:492–521 Williamson S, Adida CL, Lo A, Platas M, Prather L, Werfel SH (2020) Family matters: how immigrant histories can promote inclusion. Am Polit Sci Rev 115:686–693 Wozniak D, MacNeill T (2020) Racial discrimination in the lab: evidence of statistical and tastebased discrimination. J Behav Exp Econ 85:101512 Wyszynski MC, Guerra R, Bierwiaczonek K (2020) Good refugees, bad migrants? Intergroup helping orientations toward refugees, migrants, and economic migrants in Germany. J Appl Soc Psychol 50:607–618. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jasp.12699 Ziegert JC, Hanges PJ (2005) Employment discrimination: the role of implicit attitudes, motivation, and a climate for racial bias. J Appl Psychol 90:553–562, place: US Publisher: American Psychological Association Zussman A (2013) Ethnic discrimination: lessons from the Israeli online market for used cars. Econ J 123:F433–F468
Stratification Economics A Primer and an Explanation on Opposition to Affirmative Action Lucas Hubbard and William A. Darity Jr.
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How Stratification Economics Explains Attitudes Toward Affirmative Action . . . . . . . . . . . . . . . . . A Primer on Stratification Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What Stratification Economics Says About Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How Stratification Economics Explains Affinity for White Affirmative Action . . . . . . . . . . . . How Stratification Economics Explains White Negativism Toward Black Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50 52 52 54 56 61 63 65 65
Abstract
In this chapter, stratification economics provides a vehicle for analyzing attitudes toward affirmative action. The chapter begins with a historical example of an American politician who relied on racial tension and the fears of the dominant group to garner support to promote his campaign. This offers an introduction to the framework and guiding elements of stratification economics – most notably, its emphasis on relative, identity-based, group status. This is followed by a discussion of elemental principles of stratification economics and a summary of a range of its prior applications to explain intergroup disparities, considering outcomes in labor markets, social markets, public good allocations, and more. L. Hubbard (*) Samuel DuBois Cook Center on Social Equity, Duke University, Durham, NC, USA e-mail: [email protected] W. A. Darity Jr. Samuel DuBois Cook Center on Social Equity, Duke University, Durham, NC, USA Samuel DuBois Cook Professor of Public Policy, African and African American Studies, and Economics, Duke University, Durham, NC, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_3
49
3
50
L. Hubbard and W. A. Darity Jr.
Subsequent sections explore the consistency of this approach with the varying attitudes expressed toward various group-targeted policies in America. In particular, historical examples are gathered into two categories for analysis and exploration: policies that become desirable when they no longer risk benefitting the lower-status (subaltern) group’s status and policies that are undesirable because they are expected to improve the subaltern group’s relative position. Finally, the chapter concludes with additional thoughts on the potential for targeted and universal proposals to achieve popular support in order to advance equity. Keywords
Stratification economics · Affirmative action · Racial equity · Discrimination · Group identity · Social exclusion
Introduction With 10 days left before election day in 1990, Jesse Helms trailed in the polls. His opponent, the black democrat Harvey Gantt, was expected to benefit from the fact that a third of North Carolinian voters had grown weary of Helms, who was winding down his third term representing the state as a Republican in the US Senate. Helms went on the attack with “ads subtly priming consciousness of Gantt’s blackness” (Jamieson 1992, p. 94). His initial advertisements hit Gantt’s record on abortion, his affiliation with gay rights leaders, and his financial dealings in becoming a millionaire. These ads made various alterations to Gantt’s face and voice, shifting to black and white to heighten his hue and slowing his diction so that viewers would later identify him as “stupid” and “definitely black” (Jamieson 1992, p. 96). However, it was Helms’ next ad that proved to be a memorable haymaker. A voice over says, “You needed that job, and you were the best qualified. But they had to give it to a minority because of a racial quota.” The narrator is never shown; neither is the face of the “you” in desperate need of employment. What is shown are a pair of white hands, first holding a paper, presumed to be a resume, and then crumpling it as they realize they are not wanted. As the narrator paints the candidates’ attitudes regarding a proposed racial quota law, the hands have disappeared, but not without a closing gesture on the split screen: Note too that in the final frames Gantt is ‘for racial’ and Helms is ‘against quotas.’ Here the shot of Gantt is closer—more menacing—than that of Helms. Before this apposition, the hands have exerted control. The fist once clenched in anger is now clenched in action: crushing Teddy Kennedy’s head and about to encircle Gantt’s! (Jamieson 1992, p. 98)
Helms would go on to win the election with 52.5 percent support, with Gantt garnering a mere 38 percent of the white vote, and the creator of the 30-second spot, Alex Castellanos, would become a mainstay of Republican presidential campaigns
3
Stratification Economics
51
in the 1990s and early 2000s, eventually earning a reputation as “one of the keenest, most cutthroat strategists in the business” (Veis and Naddaf 2007). With what came to be known simply as the “hands” ad, Castellanos had discovered something powerful about the political dynamite that could be used against preferential treatment programs like affirmative action for blacks. Not only are these programs often without wide popular support, but for those politicians reliant on the support of a dominant social group, such issues also can be used as a wedge issue to rally one’s base. A half-century of American politics has proven, beyond a doubt, that nothing motivates white voters like the fear of black folks catching up or surpassing them. Perhaps it is not surprising that white voters in America would bristle at policies that favor other racial or ethnic groups. After all this is consistent with principles of rational self-interest in the context of group identification and position. Indeed, the overwhelming white American vote for former reality TV star, Donald Trump, in the 2016 and 2020 presidential elections, ensued after his preying on anxieties related to a perceived growing racial threat and his mobilization of outright racist rhetoric. When a 2017 white supremacist march took place in Charlottesville, Virginia, leading to mayhem and death, Trump refused to condemn the racist protestors. Instead, he placed them on par with those who were resisting their presence and purpose, saying there were “very fine people on both sides,” while repeatedly flashing what appeared to be the “white power” sign. This perceived threat, of course, was nothing new; it had previously become an essential component of the “Southern strategy” – an effort to shift the white southern vote away from the Democratic party – in the wake of the civil rights movement. In 1981, political consultant Lee Atwater, who successfully led the charge by coding economic policies that appeared to be unfavorable to the working class as policies to preserve white dominance, infamously summarized the approach: By 1968 you can’t say ‘n———’—that hurts you, backfires. So you say stuff like, uh, forced busing, states’ rights, and all that stuff, and you’re getting so abstract. Now, you’re talking about cutting taxes, and all these things you’re talking about are totally economic things and a byproduct of them is, blacks get hurt worse than whites.. . . (Perlstein 2018)
Why is this strategy successful? More broadly, why does the performance and status of other groups matter so much, and how does this lead to individuals – in this case, working class whites – voting against their purported economic self-interest? Why do programs that favor certain groups only become objectionable when they do not benefit the social group in power? Stratification economics offers answers to these questions. This chapter introduces the basic tenets of stratification economics, the subspecialty of economics that blends economics, sociology, social psychology, and history to understand the feedback loops between group identity, political movements, and individual action. The framework of stratification economics enables an exploration of several case studies related to affirmative action in the United States of America. Specifically, stratification economics shows how whites have benefited
52
L. Hubbard and W. A. Darity Jr.
from the existence of preferential programs in the past and why they push back against recent initiatives that bolster other groups, typically blacks. The chapter concludes with some thoughts on the potential routes toward garnering popular support for affirmative action and related policies.
How Stratification Economics Explains Attitudes Toward Affirmative Action A Primer on Stratification Economics Stratification economics fuses economic theory with sociology and social psychology to explain how group identities affect individuals’ behavior, especially when individuals consider protecting both their individual status position and the status position of the group with whom they identify. Relying on the notion that human beings care more about relative position than their absolute position, stratification economics posits, with respect to group identification, “the key relevant comparison group will be an outside racial, ethnic, gender, caste, or religious group” (Darity et al. 2017). An obvious application of stratification economics accounts for racial and ethnic disparities, which “long have been treated as a peripheral object in economics” (Darity et al. 2015). Stratification economics is sufficiently general to apply to all salient instances of group identification and inequality, whether it be on the basis of gender, caste, religious affiliation, as well as race and ethnicity, or at the intersection of combinations of these categories. Any analysis of intergroup inequalities without understanding the effects of these classifications is sorely incomplete. Consider three factors to highlight this: the racialized distribution of wealth, the significance of wealth as the key economic indicator of well-being and opportunity, and wealth’s capacity to be transferred to future generations, both directly or indirectly, to entrench advantage or disadvantage (Tippett et al. 2014). As such, acts of economic discrimination, past or present, have outsized and long-standing effects on who possesses and who is dispossessed. The principles of stratification economics, drawing heavily from Darity et al. (2015), can be summarized as follows: • Individuals behave rationally and in a self-interested fashion. • Individuals form an identity through affiliation and association with a particular social group (or groups). • Social beliefs about their group can affect their sense of affinity as well as their individual productivity and performance. Such effects include stereotype threat, stereotype lift, and stereotype boost (▶ Chap. 37, “Stereotype Threat Experiences Across Social Groups”). • Collectively, members of a particular social group (or groups) seek to maintain and extend their group’s relative performance in hierarchy, in accordance with
3
Stratification Economics
53
Herbert Blumer’s (1958) important work on the sources of prejudice. The greater the perceived difference in status between one’s own group and another, the greater the emphasis an individual will place on their personal position in comparison with others in their social group. The narrower (or the narrowing) of the perceived difference in status between one’s own group and “the other,” the greater the emphasis an individual will place on their social group’s position in comparison with the other; colloquially, this is a backlash effect against perceived progress on the part of the subordinate group. What follows from these tenets is perhaps the most novel element of stratification economics – and its most striking divergence with standard economic theory. Typically, economics treats discrimination as an inefficient irrationality that can be cured by market competition. Gary Becker’s influential taste-based models of employer and employee discrimination lead to the conclusion “that nonproductivity-based differentials in wages must evaporate over time; both approaches deny the persistence of discrimination” (Darity et al. 2017, p. 48). In contrast, stratification economics says, because of the desire to maintain one’s position – and one’s group’s position – in a hierarchy, discrimination is not only persistent but also “rational and functional” (ibid. p. 50). Furthermore, market competition is maneuverable to sustain discriminatory outcomes. What does stratification economics reject? It rejects the conception that economic and related disparities between groups are driven by deficiencies or dysfunctional behavior on the part of the subordinate, or subaltern, group, rather than discrimination, coercion, and inherited advantages. Applying the lens of stratification economics, a number of social phenomena – which run counter to traditional economic theory – suddenly become explicable. Stratification economics provides a concise explanation for the wage gaps between black and white employees (the presence of discrimination against darker-skinned employees, or colorism, in labor markets), as well as an explanation for the employment gap itself (Goldsmith et al. 2006; Pager 2003; Bertrand and Mullainathan 2004). It explains why families invest more in the futures of lighter complexioned offspring – employers, like many agents in racialized anti-black societies, offer those individuals with greater proximity to whiteness outsized rewards for similar accomplishments – and why darker-complexioned black women have lower odds of marriage or remarriage than lighter-complexioned black women (Hordge-Freeman 2015; Rangel 2015; Hamilton et al. 2009). And in the public sphere, it has an answer for why natural disasters like Hurricane Katrina disproportionately harm black individuals and families: a dearth of public services and disaster relief for these groups – and excess services and relief for white families (Price 2008). Moreover, it is consistent with studies conducted outside the oeuvre of the framework, such as research showing why America redistributes public goods less than Europe – the United States is more racially heterogeneous, and dominant groups are less willing to provide “handouts” to subordinate groups (Alesina et al. 2001). Stratification economics also explains more subtle effects. One recent example is the recent trend of whites experiencing increases in their mortality rates – at a time when no other groups in the United States are witnessing declines in their mortality
54
L. Hubbard and W. A. Darity Jr.
rates (at least until the onset of the COVID-19 pandemic). The reason? Their (false) belief that the white relative position is declining – a belief strongly connected, again, to the Trump vote. The belief yields a psycho-emotional threat that has contributed to a decline in mental health resulting in “deaths of despair” from alcohol consumption, opioid addiction, and suicide (Siddiqi et al. 2019). The framework applies neatly and with great effect to political phenomena, some of which are well-known. The “Southern strategy,” mentioned earlier, is especially relevant. The strategy follows the stratification economics framework precisely. A white voter likely will not benefit individually from a given political action, the “abstract. . .totally economic things” Atwater describes. But with its passage, the white would maintain – or even lengthen – the economic lead of his/her “group” over the competing group. And so, the individual supports politicians advancing such a policy or action. Crucially, by standard economic theory, the person may appear to be acting against his/her self-interest by voting for something that does not benefit them individually. But in promoting the relative economic status of the group to which the individual belongs, this act squarely coheres with the core theory of stratification economics. Moreover, there are benefits associated with being part of a dominant social group, regardless whether one is closer to the top or to the bottom within the group. For example, all white Americans have an advantaged position in encounters with the police and the criminal justice system vis-à-vis all black Americans. Given its potential effects on the relative economic status of dominant and subaltern groups in America, affirmative action (and similar policies that target benefits toward particular subsets of the population) is ideal for analysis through the lens of stratification economics. We undertake a preliminary look at how the principles of stratification economics can illuminate the analysis of affirmative action in the United States.
What Stratification Economics Says About Affirmative Action Coined by President John F. Kennedy in Executive Order #10925 and elucidated in subsequent orders from President Lyndon B. Johnson, “affirmative action” promised to not just “open the gates of opportunity” but also provide all citizens with “the ability to walk through those gates,” as Johnson put it in his 1965 commencement address to Howard University (Katznelson 2006). But in the years to follow, when this turn of phrase was turned into policy, it would immediately bear the brunt of staunch resistance and fierce debate. Viewed under the conditions of the larger civil rights movement, in which any progress had to withstand myriad volleys of revanchist attack, this opposition should not be surprising. More than a half-century later, its role applied in higher education – debilitated from decades of decisions that have left affirmative action as a policy justified solely as a facet of diversity – likely will soon be revisited yet again when Students for Fair Admissions v. Harvard reaches its expected destination before the Supreme Court.
3
Stratification Economics
55
Some have proposed that attitudes toward such policies emerge from a complicated confluence of economic self-interest, social demographics, racial affect, and beliefs toward the merits of stratification (Kluegel and Smith 1983). Others presuppose that resistance to affirmative action is more related to the policies themselves. While racial animus is present in the white population, it is not responsible for all of the opposition to affirmative action (Kuklinski et al. 1997). In their analysis, they find that a majority of whites (in both the north and south) support “extra effort” made by the government to benefit black Americans, but more than three-fourths oppose “preferential treatment” for blacks. Yet the arguments waged against affirmative action are well worn territory (Darity 2013). Complaints include its violating the principles of meritocracy, its negative effect on productivity, its leading students to be mismatched in higher education and/or stigmatized, its ineffectiveness in assisting all members of the target population, and the implementation of affirmative action along racial rather than class lines. These complaints all wilt under the heat lamp of investigation. Research shows that “merit” is not an objective scale but a tipped measure, designed to favor male candidates over female peers and altered continuously to exclude black students from public institutions (Uhlmann and Cohen 2005; Cross and Slater 1996). Section III, discussing affinity for white affirmative action, details how the “meritorious” students are in fact the beneficiaries of centuries of favorable conditions and policies. Moreover, claims of decreased productivity following affirmative action initiatives have been debunked at a macrolevel (Conrad 1995). Microlevel analyses have long shown the same, as have anecdotes like that relayed in W.E.B. Du Bois’ The Philadelphia Negro, of a “crank” manager at the Midvale Steel Works breaking a pattern of exclusionary hiring to let black and white mechanics toil alongside one another in his factory: . . .he had a theory that Negroes and whites could work together as mechanics without friction or trouble. In spite of some protest he put his theory into practice, and today any one can see Negro mechanics in the same gang with white mechanics without disturbance. (Du Bois 1967, p. 129)
The “crank” was Frederick Winslow Taylor, who later became widely known as the “father of scientific management.” In short, top business minds had been aware since the 1880s that at least one form of this complaint was unfounded. Some of the latter criticisms of affirmative action detailed by Darity (2013) are worthy of particular attention, in that they possess certain kernels of truth but belie their intentions. True, an affirmative action policy may not help all people in the subaltern group; indeed, it is crafted such that it can “produce and/or enhance ‘the creamy layer’” of the subaltern middle class (ibid, p.221). True, even if affirmative action is implemented, class-based inequality will remain. Do such grumbles justify scrapping the entire policy? Why would the solution not entail supplementing affirmative action with anti-poverty programs, or (additional) affirmative action on the basis of class?
56
L. Hubbard and W. A. Darity Jr.
Stratification economics provides a harsh but consistent answer to these claims, following the principles outlined in the prior section. The desire to curb or alter affirmative action policies that favor black Americans stems from a wish on the part of the dominant group (white Americans) to protect their relative economic status. If affirmative action was to achieve its goals, it would close the gap between white Americans and their subaltern peer group. Scrapping it would alleviate this threat, and shifting its focus to ameliorating class-based disparities would make it much less beneficial to the subaltern group (Darity et al. 2011). Furthermore, it is worth mentioning that the findings of Kuklinski et al. (1997) also align with stratification economics. It follows that a majority of white Americans will say that they support “extra effort” to aid black Americans if they feel that this extra effort does not put their relative status at risk. Likewise, Kluegel and Smith (1983) highlight public polling that shows the white population’s desire for shallow benefits (“help”) to be conveyed to blacks and for substantive benefits to be denied. In stratification economics parlance, anything that genuinely will close the gap between the dominant and subaltern groups will be opposed; insubstantial attempts, however, are unobjectionable and could even be valuable to the dominant group, if the public and political conclusion is that even with the provided help the subaltern group proves incorrigible. Moreover, stratification economics suggests that if the dominant group can improve its status relative to the subaltern group – if it can become more dominant – then such a proposal will receive approval. Indeed, a number of recent investigations show that opposition to targeted, group-based policies arises solely when the subaltern group benefits; then, and only then, are such policies marked with the imprimatur of “unfair” to justify their resentment. To see this in action, one needs only to consider the name of the plaintiff in the higher education court case: “Students for Fair Admissions,” the implication being that the college admission process – replete with preferential admission for offspring and relatives of alumni and standardized tests that most accurately reflect an applicant’s familial wealth – only becomes “unfair” once race is considered. The following two sections spotlight instances throughout American history that reflect this pattern. The first section consists of a series of cases in which whites have benefitted disproportionately from nominally universal policies; the second features situations in which policies that favor blacks have been instituted – and then quickly disbanded or preemptively headed off. Both cases are consistent with stratification economic theory and, contrary to the “grumbles” stated above, point to the greatest driving force of attitudes toward such policies to be racial or outgroup resentment.
How Stratification Economics Explains Affinity for White Affirmative Action Given what is known about its founding, expansion, and maintenance, it makes sense that America would strategically promote the rights of certain groups over others. For the duration of its history, the country has been dominated by a minority
3
Stratification Economics
57
elite – originally the planter aristocracy who became America’s founding fathers – who granted rights to certain groups and not others to preempt unification and revolutionary action from the underclass. During the century and a half leading up to the American Revolution, and the subsequent near-century prior to the Civil War, the settlers and white (male) working class benefitted. The colonial government and the subsequent US federal and state governments propped them up with allotments of land, stolen from the indigenous people, and with status, gained from the stigmatization, abuse, and further dehumanization of the black population, both free and, especially, enslaved. These policies were not labelled with a special term like affirmative action because there was nothing special about them: for a very long time, it was taken as given that policies would affirm the predominantly white power structure in the country. Stokes et al. (2003) highlight the Naturalization Act of 1790 as an initial stroke of favoritism that curried no resistance. Whereas the constitution famously sidestepped the racial hierarchy in play in America, the Naturalization Act directly addressed race, allowing “any alien, being a free white person” a path to citizenship in the country. This declaration, the authors note, “was adopted without a single dissent by the first sitting U.S. Congress” (Stokes et al. 2003). In the second half of the nineteenth century, homesteading, the practice of setting aside cheap or free land for settlement, would continue to consolidate white dominance. With its synthesis of “the dignity and opportunities of free labor” and “social mobility, enterprise, and ‘progress,’” the concept became popular in the run-up to and aftermath of the Civil War (Foner 1981). The passage of the Homestead Act of 1862, along with related subsequent acts – the 1866 Southern Homestead Act, 1873 Timber Culture Act, and 1877 Desert Land Act – launched settlement of western and southern lands. Starting with the 1862 Act, 160 acres of untamed land were available at $1.25 an acre to those willing to cultivate it for 5 years. This “free land,” the sort of tangible, direct help that white Americans would bristle at in the late twentieth century and beyond, was largely viewed not as a giveaway or a handout but as an entitlement. New York Tribune editor Horace Greeley exhorted such ideas in 1842, identifying the need for the lower classes to enjoy “the Right to Labor and to receive and enjoy the honest reward of such labor. . .” (Robbins 1933). In 1846, George Henry Evans of the National Reform Association petitioned congress with the slogan “Vote Yourself a Farm”; the reformers would go on to advance the notion of “the equal right to land,” such that none would be left “dependent on another for the right to work for a living” (Fure-Slocum 1995). Some newspapers did engage in handwringing over these allocations: The Goshen Democrat feared “whole contents of European poorhouses emptied down upon our fertile West,” and the Boston Daily Mail claimed that such a recipient of free land would be a “pauper entail” (Robbins 1933). In the 1800s, homesteading as a policy had faced much southern resistance, with the fear that this expansion of small, independent farmers and their subsequent political weight would lead to greater nationwide opposition of slavery. (Indeed,
58
L. Hubbard and W. A. Darity Jr.
homesteading’s brief unfeasibility was due to the fear that its implementation would eventually lessen the gaps between blacks and whites.) Following the secession of the Confederate states, however, the Homestead Act of 1862 passed resoundingly in both chambers (107 to 16 in the house, and 33 to 7 in the senate). Speaking before congress in 1862, Speaker of the House of Representatives Galusha Grow said that “there has, perhaps, never been a measure before Congress so emphatically approved by a majority of the American people.” And its effects could not be overstated, allowing the freemen who were recipients of this land to, in Grow’s words, “develop the elements of a higher and better civilization” (Anderson 2011, p. 119). One and a half million families received this land asset; Williams (2000) estimates a quarter of the US adult population has descended from this cohort. The families who benefitted from these acts, however, were not equally distributed, given the capital constraints to claim the acreage, and to the larger outlays required to build a farm once occupied (Zinn 1980; Deverell 1988). While 300 acres were distributed to about 400,000 interests, writes Greg Grandin in The End of the Myth, “this was less than half of the acreage private interests acquired through purchase. Within a decade of the act’s passage, large capitalists and regulators had laid claim to the most fertile, best irrigated, and, via railroad lines, best connected portion of public ‘free land’” (Grandin 2019, p. 110). Nominally, homesteading was open to everyone. But the Black Codes of 1865 and other methods of social control – legal or otherwise – wedged blacks geographically, preventing them from settling these western lands and protecting the dominance of white Americans. In Boom Town, his book on the history of Oklahoma City, Sam Anderson details the Oklahoma City Land Run of 1889, in which frontiersmen raced into town at high noon to make claims to the land. The description underscores the attendant pressures black Americans faced in this era of the Jim Crow South – and the west that Jim Crow Southerners would go on to settle: The Land Run was for white men. Not officially, in a legal sense, but in practice. The number of African Americans who made the Run is, like many things about that day, hard to pin down, but at least one historian guesses it was fewer than fifty. Others estimate two hundred, or somewhere near one thousand. In any case, it was a tiny fraction of the human tide— roughly 100,000 strong—that rushed in that day over the prairie. Given the opportunity the Land Run represented, especially for the poor and marginalized, that absence screams volumes. Oklahoma’s settlers were a heavily armed white mob. Black Americans would have known to proceed with caution. (Anderson 2018, p. 163)
New frontiers and their capacity to generate wealth were denied most blacks (section IV, on white negativism towards black affirmative action, details this further). Moreover, Jim Crow laws dictated their capacity to move and live where they already were. In the ten largest American cities in 1880, the typical neighborhood in which an African American lived was just 15 percent black; by 1940, their local neighborhood was 75 percent black (Rothstein 2018). This restriction of movement – the denial of what Roscoe Dunjee, publisher of The Black Dispatch in Oklahoma City, called “the right of natural expansion” – is crucial to generating patterns of segregation (Anderson 2018). Once racial
3
Stratification Economics
59
segregation is established, many seemingly innocuous, race-neutral policies can in fact systemically advantage certain groups at the expense of others; in short, segregation turns these policies into affirmative action by a different name. James Baldwin knew this, skewering the pattern of investment that spread across America in the 1950s with his assertion that “urban renewal means negro removal.” Government-sanctioned actions throughout the twentieth century – invariably positively branded as “renewals” or “revitalizations” – would demolish black neighborhoods and public housing to make way for white investment in the downtown; the effects of such policies, then, were not far from those of destructive race riots, like the 1921 Greenwood riots in Tulsa (Anderson 2018). They achieved some amalgamation of promoting the dominant group’s economic status and weakening the subaltern group’s; regardless of the mixture, the relative status of the dominant group improved. The New Deal heightened the deleterious aspects of housing segregation. In 1933, the Home Owners’ Loan Corporation was created to rescue homeowners near default during the Great Depression, and the loan risk assessment practice of redlining – so termed for the color that would demarcate African American neighborhoods on these maps – was introduced, helping codify the seemingly inherent risks of lending to blacks. The next year, the Federal Housing Administration, or FHA, was formed to help middle-class Americans access the threshold of homeownership. Approval for applicants was similarly racially restricted and consistent: “no guarantees for mortgages to African Americans, or to whites who might lease to African Americans, regardless of the applicants’ creditworthiness” (Rothstein 2018, p. 67). From 1945 to 1959, African Americans received less than 2 percent of all federally insured home loans, a tremendous setback given the outsized role homeownership plays in wealth accrual (Hanchett 2000; Shapiro 2006). Overtly stating its whites-only preference in the appraisal process, the FHA did not outwardly purport to be race neutral. But an additional significant post-World War II policy that – despite its promise of universality – disproportionately benefitted non-black Americans was the Servicemen’s Readjustment Act of 1944, now commonly known as the GI Bill. It comprised “the most wide-ranging set of social benefits ever offered by the federal government in a single, comprehensive initiative,” with spending summing to more than $95 billion between 1944 and 1971 (Katznelson 2006, p. 113). The bill provided avenues to attend college, own homes, and engage in other wealth-building ventures and, in doing so, fostered a new middle class in the country. Both black and white veterans leapt at the chance to participate in GI Bill programs. However, John Rankin, the racist Mississippi congressional representative who chaired the Committee on World War Legislation that workshopped the bill, ensured that the federal dollars would be kept under local and state control through branches of the Veterans Administration. As a consequence, in the southern states where Jim Crow racism ran rampant, the desired funds “could be directed to the country’s poorest region while keeping its system of racial power intact” (Katznelson 2006, p. 125).
60
L. Hubbard and W. A. Darity Jr.
With local control in place – white-staffed state departments and local VA centers who could discourage blacks from taking advantage of the GI Bill’s bounty – the barriers to wealth-building for blacks were born anew, and a universal program quickly became targeted to the dominant group. In Mississippi, whites filed six times as many applications for unemployment payments as black (Katznelson 2006). With most avenues of higher education – the exception being historically black colleges and universities – still closed to blacks, they were forced to turn to crowded, inferior options (“no black college had a doctoral program or a certified engineering program”) and, disproportionately, to eschew collegiate opportunities (Katznelson 2006. p. 133). Of the veterans born between 1923 and 1928, just 12 percent of blacks enrolled in college programs, while 28 percent of their white counterparts did (Katznelson 2006). Job training provided an opportunity in name only: in the first 2 years of training programs in the south, black veterans counted for less than eight percent of the total enrollment, and, rather than earning the living wage the GI Bill allotted, they were occasionally charged a fee for being trained (Katznelson 2006; Frydl 2009). Blacks found themselves the victims of occupational sorting, landed in jobs far beneath their skill level: “Carpenters became janitors; truck drivers dishwashers; communications repair experts porters” (Katznelson 2006). In addition to the housing policies which created a (white) middle class, the New Deal revamped labor laws through the National Labor Relations Act (NLRA) and Fair Labor Standards Act (FLSA), which lifted wages, limited the length of work weeks, and guaranteed the right to unionize and collectively bargain. The Democratic Party, spearheading this legislation, carefully carved out exceptions to the legislation to exclude farmworkers and maids from these crucial benefits and protections: two groups that comprised “more than 60 percent of the black labor force in the 1930s and nearly 75 percent of those who were employed in the South” (Katznelson 2006, p. 22). A 1939 poll of opinions toward the Social Security Act that provided old-age benefits and insurance against unemployment – another New Deal output that (at least initially) excluded agricultural and domestic workers from its benefits – found that 89 percent of the public approved (Amenta and Parikh 1991; Leff 1983). In short, the political resistance from the Southern Democrats targeted the widesweeping nature of these policies; a universal policy was unwelcome because of who it would include. Florida Democrat James Mark Wilcox was unequivocal when discussing the FLSA before congress in 1937: We may rest assured, therefore, that when we turn over to a federal bureau or board the power to fix wages, it will prescribe the same wage for the Negro that it prescribes for the white man. Now, such a plan might work in some sections of the United States but those of us who know the true situation know that it just will not work in the South. You cannot put the Negro and the white man on the same basis and get away with it. (Wilcox 1937)
This is the corollary to the Southern strategy that Lee Atwater described in 1981. Instead of a policy in which blacks are hurt worse than whites, the New Deal time
3
Stratification Economics
61
and again offered policies in which whites gained more than blacks. When these policies were not targeted so, they had less support. But it was under such guidelines that New Deal policies became implemented and highly regarded. They had aimed to build a middle class, and they did; they built a white middle class, locking black Americans out of crucial mechanisms to build wealth and thus fixing their place on a lower economic rung of society.
How Stratification Economics Explains White Negativism Toward Black Affirmative Action For brief windows in American history, the black population has had glimmers of favorable treatment. In 1862, for example, it looked like early waves of freedmen wouldn’t have to wait long or move far to start generating wealth through land ownership; they might receive South Carolina parcels directly, with the undertaking of the Port Royal Experiment in the midst of the Civil War, “predicated on the principle that newly freed men and women would have an opportunity to engage in homesteading on land vacated en masse by southern planters” (Darity and Mullen 2020, p. 130). However, the attempt fell apart under “the constant threat of brutality, rife with irregular pay for the black laborers or no pay at all” (ibid., p. 134). Then General William Tecumseh Sherman’s declaration, now colloquialized as “forty acres and a mule” and first made in Special Field Orders No. 15 before being formally provisioned in an 1865 Freedmen’s Bureau bill, authorized reallocating confiscated and abandoned lands in the Confederate states, a promise that would have allocated approximately 5.3 million acres of land in three states. But for innumerable reasons – not least of which was Andrew Johnson’s occupancy of the White House, his amnesty of former Confederates, and the subsequent return of their property – a key portion of the plan became “null and void, making it impossible for ex-slaves and loyal white refugees to rent up to forty acres with an option to purchase the land and then receive the title” (Darity and Mullen 2020, p. 178). This has been the pattern for black Americans; anytime it appears that a policy might narrow the gap between themselves and the dominant group, the resistance it engenders from the dominant group – writ large or simply from its most powerful actors – means it is inevitably scuttled, shortchanged, or set up for failure. A telling instance is the first (brief, quickly aborted) instance of preferential treatment for black Americans: the operations of the Bureau of Refugees, Freedmen and Abandoned Lands – better known as the Freedmen’s Bureau – following the Civil War. Supplying necessities for those displaced by the war; establishing schools, hospitals, and other institutions; assisting in the process of (re)settlement for those displaced during the war; and providing “a government guardianship for the relief and guidance of white and black labor from a feudal agrarianism to modern farming and industry,” it was a rather unique undertaking. Du Bois, in 1935, called it “the most extraordinary and far-reaching institution of social uplift that America has ever attempted” (p. 219).
62
L. Hubbard and W. A. Darity Jr.
Notably, the bureau did not simply benefit blacks. In certain states in the Deep South, “the bureau extended twice—and in some cases four times—as much relief to whites as to blacks” (Nancy Isenberg, in Grandin 2019, p. 103). Nevertheless, President Andrew Johnson decried the bureau “as a giveaway,” an ostensible double-edged sword that “was both trapping African Americans in a new form of slavery and giving African Americans preferential jobs” (Grandin 2019, p. 106). The Southern Homestead Act of 1866, championed by Oliver Otis Howard, commissioner of the Freedmen’s Bureau, tried to provide blacks with the wealthbuilding mechanism their white peers had seized. It offered up the best public land remaining in Alabama, Arkansas, Florida, Louisiana, and Mississippi to “freedmen and loyal refugees,” although the most fertile and promising land had previously been claimed prior to the war (Lanza 1990; Williams 2000). In total, only 2.9 million acres, or six percent of the total land offered, was transferred before the program was disbanded after a decade. Williams (2000) provides a dispiriting analysis of the demographic data: Estimates from a sample of homestead claims in Mississippi reveal that about 23% of claimants under the Southern Homestead Act were judged to be Black. In that sample, 35% of Black claims were successful compared to 25% of white claims (Lanza 1990). Using these percentages, 5440 of the 27,800 final patents may have been awarded to Black homesteaders. Citing Magdol (1977), only 4000 Blacks even made homestead entries under the Act (p. 160). Either way, the reality is that few homesteads were granted to Black claimants. (Williams 2000)
With its mission and focus aiding predominantly black Americans, the Freedmen’s Bureau faced much resistance. In 1866, the second Freedmen’s Bureau bill, both expanding the bureau and making it permanent, was vetoed by Johnson. A modified version of the bill would eventually pass over yet another Johnson veto, but it was indicative of the staunch resistance the bureau faced throughout its work, as Du Bois writes in Black Reconstruction: Even if it had been a perfect and well-planned machine for its mission, the planters in the main were determined to try to coerce both black labor and white, without outside interference of any sort. They proposed to enact and enforce the black codes. They were going to replace legal slavery by customary serfdom and caste. And they were going to do all this because they could not conceive of civilization in the South with free Negro workers, or Negro soldiers or voters. . .Under these circumstances, the astonishing thing is that the Bureau was able to accomplish any definite and worth-while results. . . (Du Bois 1935, p. 223–224)
Despite “haphazard” financial support throughout the bureau’s short-lived existence – it wrapped up its operations in 1872 – the committee that reviewed the bureau’s performance in 1874 commended its efforts, while noting the resistance the program faced: “No thirteen millions of dollars were ever more wisely spent; yet, from the beginning, this scheme has encountered the bitterest opposition and most relenting hate” (Du Bois 1935, p. 229).
3
Stratification Economics
63
Moving into the latter half of the twentieth century, the honing of polling practices enables an analysis with finer granularity of which policies and proposals continue to spark resistance from the dominant group. As stratification economics predicts, attitudes toward group-based preferential treatment became more hostile as the recipients of such benefits changed. Following the advent of Executive Order #10925 that established “affirmative action” – and, after a 54-day filibuster, the passage of the Civil Rights Act of 1964 – policies targeting or disproportionately benefitting black Americans have received varying levels of support. In certain instances, such support can be inconsistent, if not contradictory. One poll from 1992, for example, showed that 44 percent of respondents thought too much was being spent on “welfare” but only 13 percent thought too much was being spent on “assistance to the poor” (Zinn 1980, p. 579). As racial animus drives much of the animosity toward programs like “welfare,” use of that term – as well as the tinged phrases “quotas” and “reverse discrimination” – can trigger a more negative response in the dominant group (Gilens 1995; Feagin and Porter 1995). More generally, public support for policies benefitting blacks shifts in proportion to the policies’ magnitude and/or directness of help. A 2019 Pew survey showed three-quarters of Americans believed it is important for “companies and organizations to promote racial and ethnic diversity in their workplace”; the same exact survey showed nearly three-quarters saying race and ethnicity should not be taken into account when considering hiring and promotions (Horowitz 2019). Diversity is all well and good, until the necessary steps toward greater inclusion are explicit and the status and privileges of the dominant group become threatened in a job market that is seen as a zero-sum game.
Conclusion Unfortunately, stratification economics does not afford much optimism for effective policy solutions. On the face of it, universal policies that disproportionately benefit the subordinate group might hold the greatest political possibility. There are caveats. Universal policies – and especially universal opportunities – are rarely truly universal. They can be easily derailed by bad actors, in the winding journey from fruition to implementation, to become targeted policies hiding behind a veneer of universality, as was the case with the GI Bill. They can be victims of the cruel calculus of bigotry, as the victims of historical injustices are then deemed lending risks and redlined out of future funding. They can be submerged in the heavy sands of historical injustice, in which barriers are overlooked – the fees that black homesteaders would have to pay, the bigotry they would face as landowners – and the moment for racial change passes. But upon closer inspection, it appears that policies often escape unflagged when they work with and calcify existing patterns of discrimination. When such policies work against the grain of such patterns, however, they become objectionable; while white privilege and discrimination has been “normalized over time, and has shaped
64
L. Hubbard and W. A. Darity Jr.
everyday interactions between blacks and whites from enslavement to the present day,” what some might call “reverse discrimination” is unacceptable (Stokes et al. 2003, p. 14). The period after the Civil Rights Act of 1964 shows us that discrimination falls when “active corrective policies” are implemented (Darity 2005). But since 1968 and the civil rights era, whites in America have, in fact, largely maintained their dominant economic status, slightly widening the racial wealth gap (Kuhn et al. 2020). Despite the initial promise that affirmative action held as an “active corrective policy” for black Americans, attacks on its constitutionality through a series of Supreme Court cases have drastically weakened and broadened its focus such that most of its benefits have been shuttled to white women (Moseley-Braun 1995). It is not unreasonable to expect similar fates to befall policies that directly aim to assist subaltern groups. However, there exist a few scenarios in which gains may be made, somewhat indirectly and unpredictably. First, history has previously provided periods of shock and serendipity in which the subaltern group is the majority or finds itself in the halls of power with the capacity to write the rules for affirmative action. The direct hand that B.R. Ambedkar had in writing the post-independence Indian Constitution – and garnering support for a system akin to affirmative action to benefit the scheduled castes, scheduled tribes, and other backward classes – comes to mind (Weisskopf 2023). Indeed, in moments of great upheaval and crisis, ideas that were previously unfeasible can suddenly achieve popular support. Many of the examples discussed in this chapter – homesteading during and following the Civil War, the New Deal during the Great Depression, the civil rights bills amidst the upheaval of the 1960s – came amidst unrest. Similarly, profit-seeking opportunists have used natural disasters and fomented crises around the world in order to create volatile environments in which economic policies might be changed for their benefit (Klein 2007). While none of these examples (save, perhaps, some consequences of the provisions in the Indian Constitution) led to long-term improvements for subaltern groups, the principle – that in such moments, bold ideas become more feasible – is worth heeding. Second, it is worth considering universal policies in which relative gains for subaltern group members are overshadowed by the individual gains for dominant group members. For example, a federal job guarantee, as proposed by Paul et al. (2018), would not target any specific group, but due to the intersection of race and un/underemployment, individuals in the subaltern group would reap the most benefit. While many individuals in the dominant group will continue to resist this – those that are gainfully employed and would receive no economic benefit and would simply see their group’s relative economic status decline – it is feasible that enough lower-class individuals in the dominant group would receive a substantial direct benefit and support the policy. Indeed, initial polling suggests substantial popular support for such a program (Budryk 2020). However, this is akin to a class-based affirmative action policy. Thus, it will not alleviate all the disparities that have accrued and continue to accrue from race-based discrimination and intergenerational inherited deprivation, as Darity et al. (2011) emphasize.
3
Stratification Economics
65
Third, one could project a “Trojan horse” scenario in which the subaltern group gains in the short term because a previously implemented policy, viewed as innocuous, has beneficial effects that were initially overlooked. For example, job training programs for blacks have more support than job quotas, as the gains they promise are less substantial and tangible (and, notably, such programs are consistent with the view that black exclusion stems from black deficiency, not from discrimination). If these programs were to be improved – perhaps by accurately predicting macroeconomic demands and retraining applicants accordingly – such that they led directly to well-compensated job offers for all in the program, the relative position of the subaltern group would improve. However, over time, it is likely that the dominant group would realize the threat it had missed and rectify, or terminate, such a program. Finally, demographic change presents us one last, Charlie Brown-esque kick at the football. If the dominant group becomes overwhelmed numerically or the subaltern population develops a majority voting bloc, then a path to approval for affirmative action might not have to garner support from white America. A number of problems arise, though, when considering the overwhelming inclination of immigrating individuals to self-identify as white (regardless of skin tone) and that the strength of the self-identity of individuals in the subaltern group scales with the penalty they face for belonging to said group (Darity et al. 2006, 2017). In other words, it is unlikely that the dominant group will disappear soon (let alone peacefully), and the agitation from the subaltern group for such affirmative action policies might weaken as the dominant group becomes less powerful and punitive. As such, stratification economics – and the balance of American history – tells us that to project a feasible route to a group-based affirmative action policy on behalf of a subaltern group in America is rather wishful. Such an outcome is not impossible, but it will require some fortune, the arrival of which will undoubtedly spark a discovery of new depth under the bedrocks of this burgeoning subfield. Given what stratification economics says about the dominant group’s resistance to threats to its status, though, perhaps the best chance of achieving popular support for a policy that lets black Americans “walk through the gates of opportunity” is if white Americans do not think it actually will help black Americans walk through the gates.
Cross-References ▶ Stereotype Threat Experiences Across Social Groups
References Alesina A, Glaeser E, Sacerdote B (2001) Why doesn’t the United States have a European-style welfare state? Brookings Pap Econ Activ 32(2):187–278
66
L. Hubbard and W. A. Darity Jr.
Amenta E, Parikh S (1991) Capitalists did not want the social security act: a critique of the ‘capitalist dominance’ thesis. Am Sociol Rev 56(1):124–129. Retrieved 15 Jan 2021, from http://www.jstor.org/stable/2095678 Anderson H (2011) That settles it: the debate and consequences of the homestead act of 1862. Hist Teach 45(1):117–137. Retrieved 15 Jan 2021, from http://www.jstor.org/stable/41304034 Anderson S (2018) Boom town: the fantastical Saga of Oklahoma City, its chaotic founding, its apocalyptic weather, its purloined basketball team, and the dream of becoming a world-class Metropolis. Broadway Books, New York Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. Am Econ Rev 94(4):991–1013 Blumer H (1958) Race prejudice as a sense of group position. Pacific Sociol Rev 1(1):3–7. https:// doi.org/10.2307/1388607 Budryk Z (2020) Majority of voters support a Federal jobs Guarantee Program. Retrieved 25 Jan 2021, from https://thehill.com/hilltv/468236-majority-of-voters-support-a-federal-jobs-guaran tee-program Conrad C (1995) The economic cost of affirmative action. In: Simms M (ed) The cost and benefits of affirmative action. Joint Center for Political and Economic Studies, Washington, DC Cross T, Slater R (1996) Once again, Mississippi takes aim at black higher education. J Blacks Higher Educ 12:92–96 Darity W (2005) Stratification economics: the role of intergroup inequality. J Econ Financ 29: 144–153. https://doi.org/10.1007/BF02761550 Darity W (2013) Confronting those affirmative action Grumbles. In: Wicks-Lim J (ed) Capitalism on trial: explorations in the tradition of Thomas E. Weisskopf. Edward Elgar, Cheltenham, pp 215–223 Darity W Jr, Mullen AK (2020) From here to equality: reparations for black Americans in the twenty-first century. The University of North Carolina Press, Chapel Hill Darity W Jr, Mason PL, Stewart JB (2006) The economics of identity: the origin and persistence of racial identity norms. J Econ Behav Organ 60:283–305 Darity W Jr, Deshpande A, Weisskopf T (2011) Who is eligible? Should affirmative action be group- or class-based? Am J Econ Sociol 70(1):238–268 Darity W Jr, Hamilton D, Stewart JB (2015) A tour de force in understanding intergroup inequality: an introduction to stratification economics. Rev Black Polit Econ 42(1–2):1–6. https://doi.org/ 10.1007/s12114-014-9201-2 Darity W Jr, Hamilton D, Mason P, Price G, Dávila A, Mora M, Stockly S (2017) Stratification economics: a general theory of intergroup inequality. In: Flynn A, Warren D, Wong F, Holmberg S (eds) The hidden rules of race: barriers to an inclusive economy. Cambridge University Press, Cambridge, pp 35–51 Deverell WF (1988) To loosen the safety valve: eastern workers and Western lands. West Hist Q 19(3):269–285 Du Bois W (1935) Black reconstruction in America, 1860–1880. The Free Press, New York Du Bois W (1967) The Philadelphia negro: a social study. Schocken, New York Feagin J, Porter A (1995) Affirmative action and African Americans: rhetoric and practice. Humboldt J Soc Relat 21(2):81–103. Retrieved 15 Jan 2021, from http://www.jstor.org/stable/ 23263011 Foner E (1981) Politics and ideology in the age of the civil war. Oxford University Press, Oxford, UK Frydl KJ (2009) The G.I. bill. Cambridge University Press, Cambridge, UK Fure-Slocum E (1995) Urban poverty and ‘the right to cultivate the earth’: American land reformers in the 1840s. Iowa J Cult Stud 1995:120–132 Gilens M (1995) Racial attitudes and opposition to welfare. J Polit 57(4):994–1014. Retrieved 15 Jan 2021, from http://www.jstor.org/stable/2960399 Goldsmith AH, Hamilton D, Darity W Jr (2006) Shades of discrimination: skin tone and wages. Am Econ Rev 96(2):242–245
3
Stratification Economics
67
Grandin G (2019) The end of the myth: from the frontier to the Border Wall in the mind of America. Metropolitan Books, Henry Holt and Company, New York Hamilton D, Goldsmith AH, Darity W Jr (2009) Shedding ‘light’ on marriage: the influence of skin shade on marriage for black females. J Econ Behav Organ 72(1):30–50 Hanchett TW (2000) The other ‘Subsidized Housing’: Federal aid to suburbanization 1940s–1960s. In: Bauman JF, Biles R, Szylvian KM (eds) From tenements to the Taylor Homes: in search of an urban housing policy in twentieth century America. Pennsylvania State University Press, University Park, pp 163–179 Hordge-Freeman E (2015) The color of love: racial features, stigma, and socialization in black Brazilian families. University of Texas at Austin, Austin Horowitz J (2019) Americans see advantages and challenges in Country’s growing racial and ethnic diversity, May 08. Retrieved 15 Jan 2021, from https://www.pewsocialtrends.org/2019/05/08/ americans-see-advantages-and-challenges-in-countrys-growing-racial-and-ethnic-diversity/? utm_source¼link_newsv9&utm_campaign¼item_317006&utm_medium¼copy Jamieson KH (1992) Dirty politics: deception, distraction, and democracy. Oxford University Press, New York Katznelson I (2006) When affirmative action was white: an untold history of racial inequality in twentieth-century America. W.W. Norton, New York Klein N (2007) The shock doctrine: the rise of disaster capitalism. Henry Holt and Company, New York Kluegel JR, Smith ER (1983) Affirmative action attitudes: effects of self-interest, racial affect, and stratification beliefs on whites’ views. Soc Forces 61(3):797–824. https://doi.org/10.2307/ 2578135 Kuhn M, Schularick M, Steins UI (2020) Income and wealth inequality in America, 1949–2016. J Polit Econ 128(9):3469–3519. https://doi.org/10.1086/708815 Kuklinski J, Sniderman P, Knight K, Piazza T, Tetlock P, Lawrence G, Mellers B (1997) Racial prejudice and attitudes toward affirmative action. Am J Polit Sci 41(2):402–419. https://doi.org/ 10.2307/2111770 Lanza ML (1990) Agrarianism and reconstruction politics: the southern homestead act. Louisiana State University, Baton Rouge Leff M (1983) Taxing the ‘forgotten man’: the politics of social security finance in the new deal. J Am Hist 70(2):359–381. https://doi.org/10.2307/1900209 Magdol E (1977) A right to the land: essays on the Freedmen’s community. Greenwood Press, Westport Moseley-Braun C (1995) Affirmative action and the glass ceiling. Black Scholar 25(3):7–15 Pager D (2003) The mark of a criminal record. Am J Sociol 108(5):937–975 Paul M, Darity W Jr, Hamilton D, Zaw K (2018) A path to ending poverty by way of ending unemployment: a Federal job Guarantee. RSF: Russell Sage Foundation J Soc Sci 4(3):44–63. https://doi.org/10.7758/rsf.2018.4.3.03 Perlstein R (2018) Exclusive: Lee Atwater’s infamous 1981 interview on the Southern strategy, December 07. Retrieved 19 Dec 2020, from https://www.thenation.com/article/archive/ exclusive-lee-atwaters-infamous-1981-interview-southern-strategy/ Price GN (2008) Hurricane Katrina: was there a political economy of death? Rev Black Polit Econ 35(4):163–180. https://doi.org/10.1007/s12114-008-9033-z Rangel MA (2015) Is parental love Colorblind? Human capital accumulation within mixed families. Rev Black Polit Econ 42(1–2):57–86. https://doi.org/10.1007/s12114-014-9190-1 Robbins R (1933) Horace Greeley: land reform and unemployment, 1837–1862. Agric Hist 7(1): 18–41. Retrieved 15 Jan 2021, from http://www.jstor.org/stable/3739802 Rothstein R (2018) The color of law: a forgotten history of how our government segregated America. Liveright Publishing Corporation, of W.W. Norton, New York Shapiro TM (2006) Race, homeownership and wealth, 20 Wash U J L & Pol’y 53. https:// openscholarship.wustl.edu/law_journal_law_policy/vol20/iss1/4
68
L. Hubbard and W. A. Darity Jr.
Siddiqi A, Sod-Erdene O, Hamilton D, Cottom TM, Darity W Jr (2019) Growing sense of social status threat and concomitant deaths of despair among whites. SSM - Popul Health 9:100449. https://doi.org/10.1016/j.ssmph.2019.100449 Stokes C, Lawson B, Smitherman G (2003) The language of affirmative action: history, public policy and liberalism. Black Scholar 33(3/4):14–17. Retrieved 15 Jan 2021, from http://www. jstor.org/stable/41069037 Tippett RT, Jones-DeWeever A, Rockeymoore M, Hamilton D, Darity W Jr (2014) Beyond broke: why closing the racial wealth gap is a priority for national economic security. Report prepared by Center for Global Policy Solutions and The Research Network on Ethnic and Racial Inequality at Duke University with funds provided by the Ford Foundation Uhlmann E, Cohen G (2005) Constructed criteria: redefining merit to justify discrimination. Psychol Sci 16:474–480 Veis G, Naddaf R (2007) The 50 most powerful people in D.C., August 9. Retrieved 12 Jan 2021, from https://www.gq.com/story/fifty-most-powerful-dc-washington-political-aides-journalists? currentPage¼3 Weisskopf T (2023) Affirmative action in the USA and India. In: Handbook on economics of discrimination and affirmative action. Springer, New York Wilcox JM (FL). Congressional Record 82, 2 (1937). (Text from: Congressional record permanent digital collection). Accessed 13 Feb 2021 Williams T (2000) The homestead act: a major asset-building policy in American history (working paper 00-9). Center for Social Development, St. Louis. https://openscholarship.wustl.edu/csd_ research/46/ Zinn H (1980) A people’s history of the United States. Harper, New York
4
Transforming Gendered Labor Markets to End Discrimination Diane R. Elson
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Labor Markets as Gendered Institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trends in Gender Gaps in Labor Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Participation and Employment Rates and Labor Force Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gender Earnings Gaps: Equalizing Up or Equalizing Down? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gendered Occupational and Sectoral Segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gender Inequality in Labor Markets and Income Inequality Between Households . . . . . . . . . Efficiency and Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gender Discrimination and Inefficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Static and Dynamic Efficiency, Micro- and Macro-efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Social Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transformatory Strategies and Labor Market Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70 70 73 73 75 75 76 76 77 78 79 79 83 83
Abstract
This chapter sets out the foundations for a feminist understanding of labor market discrimination. It is argued that labor markets are gendered institutions operating at the intersection of the sphere of production and the sphere of reproduction. Gender discrimination in labor markets needs to be understood not as a residual nor as irrational but as rooted in institutional structures and the unequal gender division of paid and unpaid work. Data on gender gaps in labor markets needs careful interpretation as a reduction in gender gaps does not necessarily imply an improvement in women’s empowerment or well-being. The relation between discrimination and economic efficiency is complex. Labor market institutions which are biased against women may, in some circumstances, and in respect of some objectives, be micro-efficient, but they are not, in the long run, D. R. Elson (*) University of Essex, Colchester, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_4
69
70
D. R. Elson
macro-efficient, especially with respect to achievement of broader social objectives beyond short-run profitability. Labor markets can be transformed through public policy and collective action so that discrimination is eliminated in ways that promote improvements in both quality and productivity of employment. Keywords
Gender · Labor market · Discrimination · Economic efficiency
Introduction This chapter aims to complement other articles in this volume by proposing that gender discrimination in labor markets should be understood as rooted in the operation of labor markets as gendered institutions situated at the intersection of the sphere of production and the sphere of reproduction. Labor markets are structured by practices, perceptions, norms, and networks which are bearers of gender and reinforce gender inequality. Gender gaps in employment may be falling, but the social and economic forces underlying the data need careful examination before conclusions are drawn on the benefits for women. Falls may reflect equalizing down rather than equalizing up or be offset by changes in behavior within households. Mainstream economists have claimed that discrimination is inefficient and will be eliminated through competition. However, women’s disadvantage is deeply rooted in markets, and more transformatory strategies are required to end it.
Labor Markets as Gendered Institutions Mainstream economists tend to approach labor markets as neutral arenas in which buyers and sellers interact. The buyers and sellers may be differentiated by sex, and they may have different endowments and preferences. This is acknowledged to be gender discrimination in labor markets if differences in the hourly earnings cannot be accounted for by differences in variables such as education and on-the-job experience. As Humphries and Rubery (1995) shows, gender discrimination is treated as a residual, stemming from the tastes of employers, to be acknowledged as an explanation when other explanations fail. From this perspective, discrimination against women is a puzzle since it does not maximize profits. There is however a different way of approaching labor markets that begins not from the premise of a neutral arena within which particular individuals may, or may not, be prejudiced against women’s employment but rather from the idea that labor markets are institutions which are bearers of gender. The idea of social relations which are bearers of gender, though not gender-ascriptive, was first put forward by Whitehead (1979). An example of a gender-ascriptive social relation is that between a husband and wife, where the terms husband and wife ascribe male and female gender. The employer/employee relation is not gender-ascriptive in that way. But it is
4
Transforming Gendered Labor Markets to End Discrimination
71
a bearer of gender in the sense that there are social stereotypes which associate masculinity with having authority over others in the workplace (being the boss) and social stereotypes about what is men’s work and women’s work. Such stereotypes are not matters of individual preference but are inscribed in social institutions. The formal and informal rules which structure the operation of labor markets are instantiations of the gender relations of the society in which the labor market is embedded. They reflect existing problems of gender domination and subordination and also the tensions, contradictions, and potential for change which is characteristic of any pattern of gender relations, no matter how unequally power is distributed. Labor legislation, government labor standard inspectorates, trade unions, professional and business networks, systems of job evaluation, systems of organization of work, and pay determination structures, all these are bearers of gender, even if no overt reference is made in their rule books and in the informal discourses that surround them, to gender difference and gender inequality. These issues have been extensively discussed in the pages of Gender, Work and Organization. Pioneering theoretical contributions were made by Acker (1990). For instance, research into pay and job grading systems shows that occupations which are predominantly undertaken by women tend to have grading systems which compress jobs into a narrow range of grades offering few opportunities for advancement. Promotion systems frequently operate on the basis of rules which differentiate those with career potential from the rest on the basis of criteria such as geographical mobility. Payment systems, no matter how much they are codified and rule-based, always have scope for discretion in their application. Research on performancerelated payment systems shows that different criteria of good performance appear to be applied to men and women, even within similar jobs. Several examples are provided by Grimshaw and Rubery (2007) in relation to structures of pay and job grading systems, promotion, and payment systems. The most fundamental way in which labor markets are gendered institutions is in the way in which they operate at the intersection of ways in which people make a living and care for themselves, their children, their relatives, and friends. Activities which make a living are recognized by economists as economic activities which should in principle be counted as part of national production. For brevity, the sum of this largely market-oriented work may be called the sphere of production. But, as feminist economists have pointed out, unpaid, un-marketed caring activities are also critical for the functioning of the sphere of production, since they reproduce, on a daily and intergenerational basis, the labor force which works in the sphere of production. For more discussion see Gardiner (1997) and Folbre and Pujol (Eds.) (1996). Moreover, feminist economists have argued that unpaid caring activities entail work, even though they are not market-oriented. For brevity, the sum of unpaid, caring work may be called the sphere of reproduction (Elson 2013). Labor markets form one of the points of intersection of these two economic spheres (Humphries and Rubery 1984). But they operate in ways that fail to acknowledge the contributions of the sphere of reproduction. Instead, they operate in ways that disadvantage those who carry out most of the work in the sphere of reproduction: women. Labor market institutions are constructed in ways that
72
D. R. Elson
represent only the costs to employers of the time that employees spend on unpaid caring for others. The benefits are not represented. Thus, for instance, employees’ parenting duties are represented as liabilities and not assets to their employers. Labor market institutions are constructed in ways that reflect only the immediate costs of time off from paid work to have and rear children and care for sick relatives and friends. They are not constructed to reflect the benefits of the reproduction and maintenance of a pool of labor from which employers can select their employees. Nor do they reflect the enhanced interpersonal skills which come from parenting and managing a household. The sphere of reproduction produces benefits for the sphere of production which are externalities, not reflected in market prices and wages. Or, as feminist economist Nancy Folbre puts it, labor market institutions fail to face up to the problem of who pays for the kids? (Folbre 1994). Of course, the operation of labor market institutions cannot escape the fact that someone has to pay for the kids. (Just as there is no such thing as a free lunch, there is also no such thing as a free replenishment of the pool of labor.) Most labor market institutions are constructed on the basis that the burdens of reproducing the labor force will be, and should be, borne largely by women. For instance, arrangements for paternal leave are far less widespread than maternal leave, and where they do exist, there are many barriers to men taking up their entitlements, because promotion often depends upon showing commitment to the job and taking paternal leave may be interpreted as a sign of weak commitment to the job. Domestic responsibilities penalize women in the labor market and are a key factor in women’s weak position in terms of earnings and occupations. This has been quantified for Britain and some other European countries by Joshi (1990) and Joshi and Davies (1992). It is sometimes claimed that labor markets adapt so as to allow women to combine paid work with unpaid work – for example, part-time work and home-based work. (There is a huge literature on part-time work and flexible labor markets. A useful review of many of the issues can be found in Jenson et al. 1988.) But this kind of adaptation is generally one-sided, more designed to allow the sphere of production access to workers whose entry into the labor market is constrained by domestic responsibilities than to give weight to the contribution that women’s unpaid work makes to the sphere of production. This is revealed in the way in which these types of work typically do not have contracts which give employees any rights to paid time for meeting their responsibilities in the sphere of reproduction – rights such as maternity leave and time off for caring for sick relatives. Nor does such employment cover women’s needs when they have retired from paid work, since pension rights are also frequently not covered. Instead, the presumption appears to be that someone else (husband, children) will support such workers in their old age. Thus, labor market institutions are not only bearers of gender, but they are also reinforcers of gender inequality. But different institutional configurations give different results: some labor markets are more equal than others. Moreover, improvements can be brought about by public action, that is, by combined action by the state and by groups of active citizens (this concept of public action is developed in Dreze and Sen (1989)). Public action can build upon the potential for change in labor market institutions themselves. Such institutions are not like fortresses of stone;
4
Transforming Gendered Labor Markets to End Discrimination
73
rather they are combinations of overlapping and conflicting practices, norms, and networks, whose seeming solidity at any moment masks subterranean pressures and fissures.
Trends in Gender Gaps in Labor Markets Gender inequality in employment persists, with gaps in male and female participation, employment rates, and employment status, male-female pay differentials, gender segregation by sector and occupation, and gender inequalities in contracts and conditions of work.
Participation and Employment Rates and Labor Force Status Over the twentieth century, the female labor force participation rose significantly in almost all countries. However, in the early years of the twenty-first century, between 2000 and 2018, the global female labor force participation declined from 50.9% in 2000 to 47.9% in 2018 (ILO 2019). As the rate of education enrolment has increased across the world, labor force participation rates have decreased for both young men and women. Despite falling overall rates of labor force participation, the gender gap has narrowed slightly, from 27.6 percentage points in 2000 to 26.5 percentage points in 2018 (ILO 2019). This is because the fall in female labor force participation has been slower than the fall in male labor force participation. There is variation between countries: while this gender gap has narrowed in developing and developed economies, it continues to widen in emerging economies. Note that these three categories are defined by the ILO on the basis of World Bank data. Developing economies are low-income countries with a gross national income (GNI) per capita of US$1005 or less; emerging economies are middle-income countries with a GNI per capita between US$1006 and US$ US$12,235; developed economies are defined as those with high income, with a GNI per capita of US$12,236 or more. While the widening gender gap in participation rates in emerging economies shows that women in emerging economies are still a long way from catching up with men in terms of labor market opportunities, it also reflects the fact that a growing number of young women in these countries are enrolled in formal education, which delays their entry into the labor market – gender gaps in educational attainment have shrunk considerably in these countries (ILO 2018). In developed economies, the gender gap in participation rates is being reduced as women and men in these countries have near equal educational achievements and women face less restrictive social norms regarding paid work. In developing countries, the gap is falling because of rising female participation and diminishing male participation. Diminishing male participation is associated with increasing enrolment of young men in secondary and tertiary education (ILO 2018). The extent to which the reduction in the gap in participation rates reduces gender inequality more broadly depends on the returns accruing to men’s and women’s paid
74
D. R. Elson
work, and the extent to which male decreases and female increases in work in the sphere of production is offset by male increases and female decreases in work caring for family, friends, and neighbors (in the sphere of reproduction). The evidence from time budgets from a wide variety of countries in all regions of the world suggests that the gender division of work in the sphere of reproduction does not change enough to offset rising female and falling male participation in the sphere of production. The result is that women in the labor force typically endure a longer working day than their male counterparts once we add their unpaid work to their paid work (ILO 2016). Further, it is important to note that the increased participation of women in the developing countries can represent distress sales rather than free choices to take up new opportunities. It is the falling earnings of other household members which have propelled poor women into the labor market in many poor communities undergoing economic restructuring (Moser 1996; Beneria and Feldman 1992). A complicating factor is that rising female labor force participation is, in part, a statistical artefact. An unknown proportion of the rise is due to improvements in accounting for women’s economic activity – reflecting to some degree changes in perceptions of what counts as work, both the perceptions of statisticians and enumerators, and perhaps also, the perceptions of respondents to surveys. For instance, International Conference of Labour Statisticians recently reduced the scope of the definition of “employment” to refer only to activities performed for others in exchange for pay or profit. At the same time, it enhanced the definition of “work” to include also the production of services for own use, such as unpaid care work (ILO 2013). In practice, new definitions do not immediately get operationalized, and it takes time for new statistics to be reported. Nevertheless, despite improvements in statistical surveys, a great deal of women’s economic activity is still invisible. The extent to which economic activity is publicly recognized as such is very much bound up with the question of whether the person doing it is paid in cash or acts as unpaid labor in a family-based business. Statistics on labor force status tend to show a considerably higher proportion of women than of men with the status of unpaid family labor (ILO 2016). Even if a woman gets paid for her work, she may be no better off in terms of meeting the needs of herself and any children she has. Labor market participation may in itself involve additional costs – transport, clothing, accommodation, and equipment. Debt may be incurred for start-up costs, and lack of information may lead to underestimation of these costs (particularly where migration is involved). Even if significant costs of participation are not involved, women may find that once they are earning their own income, there is an offsetting reduction in income transfers from nonmarket sources, particularly from the father of their children. Folbre (1994) suggests that this is one of the key downsides of the increase in female labor force participation. Moreover, while labor market participation opens up new opportunities, it also opens up new risks. There is, for instance, a risk of entitlement failure (i.e., a failure to establish command over sufficient resources for survival) owing to loss of employment, a drop in wages, or a rise in prices (Dreze and Sen 1989). One of the
4
Transforming Gendered Labor Markets to End Discrimination
75
important ways in which markets are gendered is in terms of the ways in which they handle risk (such as the risk of job loss). In general, risk-reducing mechanisms have been much more a feature of male forms of market participation – for instance, trade unions, job security rights, social insurance benefits, and business and professional associations. Labor market institutions have typically been constructed on the assumption that women employees were secondary earners who could draw upon the assets and earning of men (male partners, husbands, fathers, brothers, etc.) to cushion them against risk. These institutions have assumed that women have extended entitlements which do not have the force of law but are sanctioned by accepted norms about what is a legitimate claim (Dreze and Sen 1989). Women’s very act of participating in the labor market, however, may weaken their extended entitlements, if it involves stepping outside what have been accepted as the normal roles for women. The possibility of earning an income of their own may empower them to take more decision about their own lives – but it may also cut them off from support by male kin, leaving them on their own and newly vulnerable to market forces.
Gender Earnings Gaps: Equalizing Up or Equalizing Down? Globally, the gender pay gap is estimated to be close to 23%: annually, women working full-time earn 77% of what men working full-time earn (ILO 2016). On an average, in most countries, the gap has decreased from 21.7% to 19.8%. However, progress in reducing the gender pay gap has been slow, in spite of significant progress in women’s educational attainments as well as labor force participation rates. A reduction in earning differentials may take place through equalizing up or through equalizing down. In the case of the former, there is no absolute drop in men’s earnings (which indeed may even be rising, although at a slower rate than those of women): whereas in the latter case, men’s earnings fall absolutely (and indeed, women’s earnings may also fall, though not by as much). Equalizing up is likely to be more beneficial to women than equalizing down. For instance, it may involve lesser risk of domestic violence. But even in the case of equalizing up, the narrowing of the gender wage gap may overstate the benefits to women, because there may be countervailing moves in male-to-female income transfers. For instance, women may be expected to finance a larger proportion of household expenditure from their earnings (Chant 2007). Gender-disaggregated household income and expenditure budgets are needed to put the benefits of any narrowing of male-female earning differentials into perspective.
Gendered Occupational and Sectoral Segregation Globally, over the past two decades, occupational segregation has narrowed slightly due to the movement of women in “mixed” category of jobs (such as leadership and
76
D. R. Elson
managerial roles). However, between 2000 and 2010, there has been a 0.5 percentage point decline in women’s participation in male-dominated professions (like trade jobs, machine operator, etc.). Further, female-dominated occupations remain highly feminized: women remain overrepresented as “clerical, service, and sales workers” and in “elementary occupations” (ILO 2016). This segregation has increased over the past decade due to skill-biased technological change especially in emerging economies. Moreover, women are more likely to be concentrated in the lower segments across job hierarchies in employment. Labor markets that were once characterized by secure, permanent, full-time jobs with pension rights and other benefits have changed as more and more jobs are casualized. Over the years, more women than men have found themselves in marginal, part-time work that is more precarious and unpredictable (ILO 2016).
Gender Inequality in Labor Markets and Income Inequality Between Households Contemporary patterns of economic growth which are structured by market liberalization and globalization are associated in many countries with growing inequality in the distribution of income between households (UNDP 2019). As far as many women are concerned, on the one hand, their bargaining power, in relation to the men in their households, their communities, their networks, and the organizations of their civil society, may often (though not, as we have argued, inevitably) be increasing as a result of their greater participation in labor markets. But at the same time, their households, their communities, their networks, and the organizations of their civil society are more and more at the mercy of global market forces that are out of control, as witnessed in the Asian financial crisis of 1997 and the global financial crisis of 2009 (Elson 2013).
Efficiency and Discrimination Neoclassical welfare economics favors definitions of efficiency which gloss over distributional questions and gloss over the question of who has the power to define efficiency. Instead, the focus is on productive efficiency, defined as achieving maximum output from given resources, and allocative efficiency, defined as achieving that particular form of maximum output, that particular combination of commodities, that will best satisfy given, well-defined, preferences. (The latter is described as the optimal situation for the economy to be in and is consistent with profit maximization in a perfectly competitive economy.) It is not possible to directly test whether an economy has achieved optimality, but it is assumed an economy will move in that direction if prices move closer to marginal costs and wages move closer to marginal productivity. Typically, deregulation has been seen by mainstream economists as something which will assist this and promote economic growth (e.g., see World Bank’s Doing Business Report 2015).
4
Transforming Gendered Labor Markets to End Discrimination
77
Gender Discrimination and Inefficiency Mainstream economists define gender discrimination in labor markets in terms of women getting lower wages, even though they have the same productive capacities as men in terms of education and work experience. This ignores the broader structures of disadvantage, which mean that women are less likely to be educated and to be able to accumulate the same work experience as men. Discrimination against women in employment clearly violates the condition for productive efficiency as defined in mainstream economics, i.e., the allocation of given quantities of resources (in this case, given quantities of human effort of a given quality) to maximize output. Thus, it can be argued that there is an efficiency-loss-based case for interventions in labor markets to counter discrimination, such as equal opportunities legislation. But some mainstream neoclassical economists would probably contest the idea that discrimination was a major reason for gender differences in occupations and wages. They would argue that discrimination against women reduces profits for employers and misallocates resources and that market competition will eliminate such discrimination. An influential labor market textbook instructs students that no strong motive for discrimination against women has been found and concludes that: labor mobility and competition among firms will eradicate discrimination and allocate labor to its most productive uses so minimizing costs (Polachek and Siebert 1993). It is possible to provide explanations for the persistence of discrimination which take account of incomplete information. For instance, the incomplete knowledge of employers about the characteristics of potential employees – employers may systematically underestimate the productive potential of women and thus pay them less and confine them to lower-grade occupations – may result in the so-called error discrimination. But again, as Humphries and Rubery (1995) points out, neoclassical economists would argue that such mistaken behavior would lead to lower profits and therefore would not persist in competitive markets. It is possible, however, to see how such mistaken behavior might persist if the social embeddedness of economic activities is recognized, which implies that male employers are not just economic men, interested only in maximizing profits, but are likely to be interested in perpetuating their advantage in the social and political spheres of life too. Altering their perceptions of the productive potential of women may threaten male employers’ advantages in other spheres of life, outside the workplace, and thus mistaken beliefs may persist because they are part of a wider system of male power. Gender discrimination may also persist because it serves strategies for managing the production process in conditions of uncertainty. It is impossible for employers to specify an employment contract to cover every possible eventuality and to be present to monitor every worker at every stage of the production process. In this context, a variety of strategies for managing production are adopted (Sawyer 1995). Such strategies include systems of monetary incentives and disincentives (e.g., piece rates, bonuses, deductions from wages for poor quality work), systems of monitoring and enforcing discipline; and systems of building the cooperation and improving the
78
D. R. Elson
skills of employees. Existing patterns of gendered power, what Breugel and Perrons (1995) call the gender order, can be mobilized in such strategies. For instance, if the prevailing gender order gives more authority to men in the wider society, then systems of monitoring and enforcing work discipline can mobilize that authority by placing men rather than women in supervisory and managerial positions. If the prevailing gender order facilitates building solidarity and team spirit among men on the basis of excluding (and even denigrating) women, then systems of building cooperation in the workplace can mobilize team spirit by excluding women, formally or informally. It is important to note that such discrimination is only efficient in the sense that it may contribute to maximizing output, given the prevailing gender order. It would be more efficient in terms of production of output to change the prevailing gender order, so that the exercise of authority was not a male preserve and so that team building did not depend on excluding and denigrating people of a different gender. This would allow everyone’s talents to be effectively mobilized and would increase productivity.
Static and Dynamic Efficiency, Micro- and Macro-efficiency The problem is that economies are currently locked into set of institutions which operate in a self-reinforcing and cumulative way to limit the potential gains from employment. One of the reasons that it is so hard to change is that people’s preferences and perceptions are to a large extent endogenous, shaped by the prevailing sets of institutions, and current rules and norms are continually reinforced (Ferber and Nelson 1993; Bowles and Gintis 2000). To discuss these issues further, it is helpful to refer to distinctions between static and dynamic efficiency and microand macro-efficiency (Humphries and Rubery 1995). Static efficiency refers to efficiency at one point in time, and discussion in this chapter so far has been related to this static concept of efficiency. Dynamic efficiency refers to efficiency over a period of time, during which various factors that have to be taken as given in the short run can be changed. Economists usually focus attention on investment to change the stock of skills and equipment to be used in paid employment. It is possible to also focus on investment to try to change the gender order. Mainstream economists already recognize that market forces are less likely to achieve dynamic efficiency than to achieve static efficiency and that discrimination against women and girls in investment in education and training is inefficient in the long run but is inclined to see the root of such discrimination in the behavior of households and families, which are not subject to competitive forces in the same way as firms. The recommended answer is usually public subsidies to the education of girls. But experience seems to show that this by itself is not dynamic enough, since it does little to challenge and transform current gendered norms and perceptions. For instance, occupational segregation remains strong despite closing of gender gaps in school enrolment and even in educational attainment (UNDP 2013). Some of the reasons for this can be explained in terms of the distinction between micro-efficiency and macro-efficiency. Micro-efficiency refers to each economic
4
Transforming Gendered Labor Markets to End Discrimination
79
agent (individual, or household, or enterprise) operating as efficiently as they can in the given circumstances. Macro-efficiency refers to the efficiency of the economy as a whole. There is a strong presumption in neoclassical economics that microefficiency will lead to macro-efficiency, through market coordination based on each individual pursuing their own self-interest as they see it. But this rests on the presumption that information is shared and flows quickly and that institutions can quickly be reshaped. In other words, that history does not matter. Once these very strong assumptions are relaxed, however, it is quite possible for there to be numerous alternative outcomes, in each one of which everyone is doing as well as they can in the circumstances (in technical terms, multiple equilibria) but in many of which the whole society is not doing as well as it could. In other words, there are processes of cumulative causation which lock the society into a lower level of achievement than it could potentially reach. Labor market institutions which are biased against women may, in some circumstances and in respect of some objectives, be micro-efficient, but they are not, in the long run, macro-efficient.
Social Efficiency The definition of efficiency in economics is narrow and ignores other goals that people might have besides maximizing output of commodities, consuming commodities, and enjoying leisure – goals such as dignity, autonomy, self-respect, and pleasure in exercising skills in their work. Within the confines of neoclassical welfare economics, there is no scope for relating efficiency to empowerment goals and judging the efficiency of an economic system in terms of how effective it is in providing everyone with opportunities for self-realization (Elson 1997). There is both theoretical reasoning and empirical evidence that suggest that more inclusive, democratic, and egalitarian ways of organizing employment can both increase productivity and better serve human needs through promoting commitment and cooperative behavior, encouraging workers to share information and give of their best (e.g., Hodgson 1984; Bowles and Gintis 1993, 1997; Pagano 1991).
Transformatory Strategies and Labor Market Standards The problem is that to move the society as a whole, from the lower-level equilibrium characterized by discrimination and lower economic and social productivity to a higher-level equilibrium characterized by absence of discrimination and higher levels of economic and social productivity, requires changes in people’s perceptions and concerted and cooperative action. Without this, economic agents remain locked into an inherited set of gendered roles and behaviors in the context of labor market institutions which are bearers of a preexisting set of gender relations. In order to achieve this, there is a clear need for a transformatory employment policy; that is, a policy which helps to change people’s perceptions of what is possible, beneficial, and fair, fosters cooperative action, and strengthens women’s
80
D. R. Elson
bargaining power in the workplace, the home, and the marketplace. It is in this context that labor market regulation should be seen. It is, of course, impossible to have an unregulated labor market, since all markets require some sort of authority to govern them (Sawyer 1995). The questions are: what kind of regulation? Whose interests does it serve? What kind of market institutions does it promote? What kinds of norms and perceptions does it promote? In most countries basic labor standards on health and safety, minimum pay, and rights of the labor force to organize and bargain collectively are likely to be more important for improving the employment conditions of most women than complex standards relating to equal pay for work of equal value. This does not mean there is no need for specific legislation to make discrimination against women illegal. Such legislation is critical for shaping norms and perceptions – but its immediate value for many women in many countries will be to enshrine an important principle. In practical terms, public policy to set and enforce equal standards for men and women on labor basic rights would likely have more impact for more women. There is a pressing need to challenge the idea that women do not need the same basic employment rights as men because their work is more likely to be part-time, intermittent, or home-based. Simple deterministic models of the labor market which do not allow for uncertainty and endogenous preferences tend to suggest that labor market regulation to improve women’s employment outcomes and reduce gender bias in labor market institutions is likely to reduce women’s employment opportunities. The message is that women can have better employment or more employment – but not both, because better employment means extra costs for employers or (in the case of the self-employed) for customers. The World Development Report for 2013 admits that, on the question of minimum wage legislation, there is empirical evidence suggesting both that minimum wages discourage employment and that they do not (World Bank 2013). This is not surprising. As with all attempts to establish the impact of specific policies, much will depend on the conceptual framework used to generate ideas about the counterfactual and on judgments about the extent of policy implementation and the significance of changes in non-policy variables in the same period. Conflicting empirical evidence can often be traced to conflicting theories and conflicting judgments. Insofar as extending women’s employment rights imposes additional costs on individual employers, it is possible to design policy packages which will minimize such costs. For instance, the costs of maternity leave for women workers (or paternity leave for men workers) could be financed from general taxation rather than from the profits of individual employers. Indeed, the ILO Maternity Protection Convention stipulates that individual employers should not be liable (Anker and Hein 1985). Moreover, once it is acknowledged that the output produced from a given labor force varies with the way in which the production process is managed, the way is open for the idea that raising benefits to employees can lead to increases in quality and productivity (this has led to the development within mainstream economics of efficiency wage theory – see, for instance, Weiss (1990)). As Sawyer (1995) explains, this might happen through improvements in the health and strength of workers (e.g., through better nutrition), through improvements in the willingness to
4
Transforming Gendered Labor Markets to End Discrimination
81
work of workers (e.g., less shirking), or through stimulating improvements in the physical organization of production (e.g., less downtime between operations). The same argument might be made in relation to improvements in the safety of the workplace, job security, time off for family responsibilities, etc. The objection might be raised that if raising wages and improving working conditions (or prices paid to the self-employed) improve productivity and quality, then employers (or customers) would do this without any need for prompting from public policy. But economic agents have bounded rationality and make adjustments through a process of search and discovery in which there can be a good deal of inertia associated with the existence of vested interests and in which perceptions of opportunities are shaped by existing positions (Humphries and Rubery 1995). Thus, public policy to set fair minimum standards in wages and working conditions should not be assumed to be redundant: it may prompt employers and customers to recognize new opportunities for increasing quality and productivity and help to overcome inertia. Of course, there are employers and customers who are not interested in increased quality and productivity: their competitive strategy rests on production of low quality, cheap goods, and they are interested in extracting as much work as possible for minimum return, provided there is surplus labor available. Such strategies frequently impose severe costs on the workforce – ranging from stress, exhaustion, ill health and disabilities, to loss of life. These strategies may be efficient in generating short-term private profits; they are not socially efficient in achieving the wider objectives of human development. The adoption of more socially efficient strategies may however be blocked by powerful groups who, as Folbre (1995) remarks, may be willing to pay a price, in lower efficiency, for continued control over a disproportionate share of output. It is always easier for a sectional interest group to maintain its opposition to change if it can buttress its position with the argument that the status quo is more efficient, hence the importance of carefully distinguishing between different types of efficiency. Economic strategies which treat people as simply instruments of production without regard for their welfare (ignoring, for instance, risks to their health) are widely regarded as unethical, although there are debates about where the line falls between practices which are ethical and unethical. But individual businesses often argue that competitive pressures mean that ethical considerations are a luxury. The role which can be played by public policy to set standards for wages and working conditions is thus twofold: to shape perceptions and set norms about what is ethical and unethical and to create an environment in which all enterprises face similar standards and thus cannot obtain competitive advantage by paying wages which are below subsistence level and undervalue the contributions of employees and enforcing working conditions which endanger the well-being of the workforce. In a global economy, there needs to be international as well as national conventions on what fair minimum standards are. Some of these standards will vary with the context (the fair minimum wage in Bangladesh will be lower than in the United States because subsistence norms and costs differ in the two countries). But there is a case for absolute standards on matters of life or death: for instance, with respect to chemical hazards and fire hazards. The ILO conventions, agreed by representatives
82
D. R. Elson
of governments, workers, and employers, are a source of international labor standards that women’s organizations are using in their organizing, though they are not, of course, panaceas. For instance, research by Women in Informal Employment: Globalizing and Organizing (WIEGO) – a research and advocacy group – notes that 20 years after the adoption of the ILO Convention on home work (1996) only ten countries have ratified the convention, too few workers have realized their rights, and the legal environment around home-based work remains uncertain. Enforcing standards is difficult, even with well-organized and resourced inspectorates, and most developing countries are far from having these. A vital role must be played by collective organizations that bring together groups of workers in trade unions or other forms of organization (whether self-employed or employees) to monitor standards and mobilize in civil society for their enforcement. It is beyond the scope of this article to review the advantages and disadvantages and successes and failures of different types of organizing, from the point of view of women. For an informative discussion of these issues, see Chhachhi and Pitten (1986) and Prugl and Tinker (1997). International conventions and national legislation can play an important role in legitimizing the activities of such groups and providing a focal point for the process of changing the perceptions of the workforce and the wider society about the rights that should be and can be enjoyed by women and men in their employment. In 2013, the first ILO Convention on paid domestic work came into force, and the demand for ratification of the convention is being used as a rallying call by organizations of domestic workers around the world (UN Women 2015). Enforcing standards is always likely to be uneven – it is easier to monitor a few large workplaces than many small ones. So it is important to address at the same time the availability of a wider range of forms of employment. The existence of a large pool of surplus labor with little or no alternative but to accept employment on whatever terms are offered is always likely to undermine the enforcement of decent labor market standards. It is important to explore alternatives to those offered in the competitive struggle for short-run profit. Important contributions to countering discrimination against women in the labor market can be made by provision of employment by the public sector. For instance, research by Fernanda Bárcia de Mattos and Sukti Dasgupta (2017) suggests that India’s National Rural Employment Guarantee Scheme (MG-NREGA) has been instrumental in ensuring paid employment for women and has a positive and significant effect on women’s control over household decisions. ILO’s Global Wage Report 2018–2019 suggests that even though there is much diversity across countries, on average, hourly gender pay gaps are lower in the public sector than in the private sector. A good example of what can be achieved through public policy is Brazil in the first decade of the twentieth century. Between 2001 and 2009, 17 million jobs were created in the country, ten million of them jobs that gave access to the social security system stimulated by the Brazilian government’s package of economic and social policies. Brazilian macroeconomic policy in this period aimed at inclusive growth; and the labor market position of women was strengthened via a doubling in the minimum wage and more investment in labor standards inspection. The outcomes included an increase in women’s labor force participation rates from 54% to 58%, with the proportion of women in jobs that provide social security cards increasing
4
Transforming Gendered Labor Markets to End Discrimination
83
from 30% to 35%, and a decline in gender pay gap from 38% to 29%. Most importantly, this decrease has been secured via an increase in both men and women’s wages, a good example of equalizing up (UN Women 2015).
Conclusion Gender discrimination is deeply inscribed in the institutions and practices of labor markets, especially the ways they fail to account for the benefits to the economy of the unpaid work that women do in the sphere of reproduction and in the ways they undervalue the paid work that women do. Some gender gaps in wages have been falling, but sometimes this is a result of equalizing down rather than equalizing up. Though some mainstream economists argue that gender discrimination is inefficient, gender discrimination can have advantages for businesses, and we cannot rely on competition to eliminate it. While it is important to have specific laws that outlaw discrimination and call for equal pay for work of equal value, these laws are limited in their ability to transform labor markets in ways that can reach all women. Strong labor market standards, public policy to create decent jobs, and women’s collective organization all have vital roles to play in creating truly gender equal labor markets. Acknowledgments I thank Niharika Yadav for her very helpful assistance in the preparation of this chapter.
References Acker J (1990) Hierarchies, jobs and bodies: a theory of gendered organisations. Gend Soc 4: 139–158 Anker R, Hein C (1985) Why third world urban employers usually prefer men. Int Labour Rev 24: 73–90 Beneria L, Feldman S (eds) (1992) Unequal burden: economic crisis, persistent poverty and women’s work. Westview Press, Boulder Bowles S, Gintis H (1993) The democratic firm: an agency-theoretic evaluation. In: Bowles S, Gintis H, Gustafsson B (eds) Markets and democracy. Cambridge University Press, Cambridge Bowles S, Gintis H (2000) Walrasian economics in retrospect. Q J Econ 115(4):1411–1439 Breugel I, Perrons D (1995) Where do the costs of unequal treatment for women fall? An analysis of the incidence of the costs of unequal pay and sex discrimination in the UK. In: Humphries J, Rubery J (eds) The economics of equal opportunities. Equal Opportunities Commission, Manchester Chant S (2007) Gender generation and poverty: exploring the ‘Feminization of Poverty’ in Africa, Asia and Latin America. Edward Elgar, Cheltenham Chhachhi A, Pitten R (eds) (1986) Confronting state, capital and patriarchy-women organising in the process of industrialisation. Macmillan, London de Mattos FB, Dasgupta S (2017) MGNREGA, paid work and women’s empowerment. Working paper no. 230. International Labour Organisation. https://www.ilo.org/wcmsp5/groups/public/ %2D%2D-ed_emp/documents/publication/wcms_613735.pdf Dreze J, Sen A (1989) Hunger and public action. Clarendon Press, Oxford Elson D (1997) Economic paradigms and their implications for models of development: the case of human development. In: Culpeper R, Berry A, Stewart F (eds) Global governance and development 50 years after Bretton woods. Macmillan, London
84
D. R. Elson
Elson D (2013) Economic crises from the 1980s to the 2010s: a gender analysis. In: Waylen G, Rai S (eds) New Frontiers in feminist political economy. Routledge, London Ferber MA, Nelson J (1993) Beyond economic man: feminist theory and economics. University of Chicago Press, Chicago Folbre N (1994) Who pays for the kids? Gender and the structures of constraint. Routledge, London Folbre N (1995) Engendering economics: new perspectives on women, work and demographic change. Paper presented to annual World Bank conference on development economics, Washington, DC Folbre N, Pujol M (eds) (1996) Special issue in honor of Margaret Reid. Fem Econ 2(3):1–152 Gardiner J (1992) Gender, care and economics. Macmillan, London, 1997 Grimshaw D, Rubery J (2007) Undervaluing women’s work. European Work and Employment Research Centre. http://www.njl.nu/uploads/Paper_2007_Jill_Rubery.pdf Hodgson GM (1984) The democratic economy. Pelican Books, Harmondsworth Humphries J, Rubery J (1984) The reconstitution of the supply side of the labour market: the relative autonomy of social reproduction. Camb J Econ 8(4):331–346 Humphries J, Rubery J (eds) (1995) The economics of equal opportunities. Equal Opportunities Commission, Manchester International Labour Organisation (2013) Resolution concerning statistics of work, employment and labour underutilization. https://www.ilo.org/ International Labour Organisation (2016) Women at work. https://www.ilo.org/ International Labour Organisation (2018) Global Wage report. https://www.ilo.org/ International Labour Organisation (2019) World employment and social outlook. https://www.ilo. org/ Jenson J, Hagen E, Reddy C (1988) Feminisation of the labour force: paradox and promises. Polity Press, Cambridge Joshi H (1990) The cash opportunity cost of childbearing: an approach to estimation using British evidence. Popul Stud 44:41–46 Joshi H, Davies H (1992) Day care in Europe and mothers foregone earnings. Int Labour Rev 131(6):561–579 Moser C. (1996) Household responses to poverty and vulnerability, Vols. 1–4, Urban management programme, Washington, DC: World Bank Pagano U (1991) Property rights, asset specificity and the division of labour under alternative capitalist relations. Camb J Econ 15(3):315–342 Polachek SW, Siebert WS (1993) The economics of earnings. Cambridge University Press, Cambridge Prugl E, Tinker I (1997) Micro entrepreneurs and homeworkers: convergent categories. World Dev 25:1471–1482 Sawyer M (1995) The operation of labour markets and the economics of equal opportunities. In: Humphries J, Rubery J (eds) The economics of equal opportunities. Equal Opportunities Commission, Manchester UN Women (2015) Progress of the world’s women 2015–2016. Transforming economies, realizing rights, chapter 2. UN Women, New York. https://www.unwomen.org/en United Nations Development Programme (2013) Humanity divided: confronting inequality in developing countries. UNDP, New York United Nations Development Programme (2019) Human development report. UNDP, New York. https://www.undp.org/ Weiss A (1990) Efficiency wages. Princeton University Press, Princeton Whitehead A (1979) Some preliminary notes on the subordination of women. IDS Bull 10(3):10–13 World Bank (2013) World development report. World Bank, Washington DC. https://www. worldbank.org/ World Bank (2015) Doing business 2015. World Bank, Washington DC. https://www. worldbank.org
5
Theories of Discrimination: Transnational Feminism Inderpal Grewal
Contents Feminist Research and International Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 Postcolonial Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 Race/Ethnicity Critiques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 Diaspora Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 The Mobility Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 Transnational Feminism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Abstract
In order to explain transnational feminism, this chapter will first describe trends in research that led to this feminist intervention. This chapter will then summarize the new interdisciplinary fields of research that contributed to transnational feminism, including postcolonial studies, race/ethnicity studies, diaspora studies, and critical geography. In the final section, this chapter will turn to transnational feminism and the different approaches that are part of this field and its contributions. Keywords
Transnational · Feminism · Postcolonial · Race · Diaspora · Mobility
In the past two decades, the term transnational has become popular in both the theory and practice of research and activism. What used to be called “international” or “cross-cultural” or “global” is often termed as “transnational.” While this
I. Grewal (*) Yale University, New Haven, CT, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_5
85
86
I. Grewal
widespread use of the term means that its meaning can change with dissemination, there are certain broad trends in usage that can be identified. This chapter considers the feminist contributions to this debate and will trace how the term gained traction because of changes in academic research in the 1980s that came in response to social movements, new theories of global social change, and critiques of economic globalization. In academic contexts, especially in the United States, a great deal of work on transnational, even outside of feminism, has come as a paradigm shift. Transnational histories are now a particular subset of the discipline of history, focusing on histories of connection and of regions and oceans, the movements of people across national boundaries, and the making of national boundaries. Transnational work in anthropology focuses on the mobility of people and their cultures, and on the impact and effects of transnational capital that crosses national boundaries, and the making of border zones and borderlands. Geographers, especially in the field of cultural geography, have used a broader notion of mobility to understand the relation between space and place. Sociologists have begun to use this concept, especially with regard to the production of differences of power between groups of women (Patil 2011). Migration studies in social sciences has used the term to understand international migrant and refugee movements. Work in literature and art has analyzed esthetic forms that cross national boundaries and the cultural work done in and by diasporic communities (Parikh 2017). In political science, it is feminist scholars who have used the term to speak of feminist movements, UN and international projects and solidarities, and the hierarchies within them. The academic site of the production of feminist interest in transnationalism has been, for the most part, within US academia, though it is crucial to understand that its influences came from and are used in research and social movements in other regions. Starting in the 1980s, historical understandings of varieties of national and religious feminisms became resonant in research showing how different nationalisms and social movements enabled particular feminisms to flourish, understanding how, when, and why they emerge in particular ways. These shifts came, in US contexts, in the aftermath of civil rights and feminist movements and globally, with the end of European colonial rule in some places, the emergence of new forms of imperialism, combined with the recognition of non-white immigrant populations in the United States and United Kingdom. By the 1980s, as neoliberal economic globalization gained popularity, and as transnational corporations became ubiquitous, the term transnational began to be used within academic research as well. In order to explain transnational feminism, this chapter will first describe trends in research that led to this feminist intervention. This chapter will then summarize the new interdisciplinary fields of research that contributed to transnational feminism, including postcolonial, race/ethnicity studies, diaspora studies, and critical geography. In the final section, this chapter will turn to transnational feminism, the different approaches that are part of this field, and its contributions.
5
Theories of Discrimination: Transnational Feminism
87
Feminist Research and International Problems Feminist movements in the 1960s led to changes in academia, as researchers began to study the lives of women and to argue for changes in policy that paid attention to women’s lives. While research and teaching concerning gender and feminist movements had emerged in departments of English and Humanities in the US by the 1970s, they initially focused on histories and writings of women in the United States and Europe. In the social sciences, as feminists struggled to be visible in academic and policy spaces, they began to examine inequalities between men and women arising from the devaluation and erasure of women’s work. In fields such as economics, feminist economists addressed wage and income inequalities, though this kind of work was often marginalized. Also marginal was the study of inequality outside of Europe and the United States, and the feminist work on global inequality was also seen as outside the mainstream of social science in the 1970s and into the 1980s. However, feminist scholars continued to examine how women were ignored, erased, and devalued in many societies, including within academia. Feminist scholars across the globe joined the effort to understand how inequality, violence against women, and erasure of their contributions to arts and sciences all impacted their status and condition in their societies. Historians and political scientists examined how gender played a role in conflicts and wars and how political violence and militarisms were enabled by ideas of male power and masculinity (Enloe 2016). Women’s studies (now often termed gender studies) emerged because of the need for interdisciplinary analysis of gender, as spaces where new methodologies could be introduced (instead of the demands of the disciplines which were often masculinist) and where activists and artists would connect the academic formation to the feminist movement (Wiegman 2002). Feminists began a critique of the academic disciplines, institutions, and knowledges that excluded women and which were often patriarchal. One area of policy and work that was concerned with international issues (outside of women’s history) was in what was called Development research – the policies and projects that, beginning in the 1960s, international institutions such as the UN and World Bank (among many others, including nations) used to “modernize” the “third world.” Ester Boserup argued that Development and mechanization enabled women’s separation from waged labor leading to their devaluation (Boserup 2007). Maria Mies argued that capitalism used women for cheap labor (especially in their roles as housewives) across capitalist and socialist societies that constituted a sexual and international division of labor (Mies et al. 2014). By the 1980s, feminists argued that women had been left out of Development policy and women’s needs had been neglected, especially because Development focused on industry and science as engines of modernization. However, the subsequent inclusion of women within Development research and practice meant that even feminist research and policy came to believe that capitalism was not an impediment to development and that women could be added to development policy and be modernized. This project came to be known as Women in Development within international organizations, WID; organizations run by women from the Global South organized their own network, DAWN, Development Alternatives with Women for a New Era, to push for their
88
I. Grewal
own agendas. As Shirin Rai argues, WID ignored the social and political contexts of women’s lives (Rai 2013). Some feminist researchers criticized a Development approach that left out both colonialism and capitalism in its analysis and had relied on a normative belief in ideas of labor and productivity based on male standards. Amartya Sen and Martha Nussbaum developed the “capabilities” approach that sought more recognitions of the freedoms and rights that would allow women to succeed (Nussbaum and Sen 1993). Despite these changes, the reliance of Development on notions of modernity as the goal has been subject to critique, and even the subsequent Gender and Development approach, which tried to focus on broader social formations of gender, remained tied to notions of Western modernity as the end-goal of Development. As its critics revealed, GAD made women into unwitting targets of Western policy and global institutions that experimented on their lives without recognition of cultural realities (Kabeer 1994). Decades of Development policy have led to uneven results for women. Critics also fault Development policies for ignoring alternative ways of living and different goals that were central to many communities, including indigenous ones. Thus, some scholars see Development as a form of neocolonialism and Development Economics as a form of power whose belief in modernity and markets as progress needed to be rethought (Escobar 2011). What is relevant to the discussion of transnational feminism is that it emerged from resistance to the way that Development experts thought of the “third world woman” as a homogenous and passive recipient of Western aid and policy. In a now-classic essay, “Under Western Eyes,” Chandra Mohanty offered critical examinations of discourses of “third world women” as stereotypic and abjected subjects to be rescued by Western Development (Mohanty 1984). However, it was not just in Development studies but also in all kinds of research that simplistic ideas of non-Western women were offered. What was common in feminist research was the erasure of colonial histories and colonial policies and the tendency to blame non-Western cultures for the subordination of women. Thus, blame was placed for gendered inequalities and subordination on a culture and a patriarchy, without attention to how culture and patriarchy had been “recast,” institutionalized, and deployed by colonial regimes and nationalist patriarchies to enable colonial rule (Sangari and Vaid 1990). What was left out in such analyses was not just the imposition of Western norms for women’s lives or the homogenizing of women as the same everywhere but also how patriarchal European and American cultures presented themselves as superior and altruistic in rescuing women from their cultures; Gayatri Spivak would pithily call this a colonial project of “saving brown women from brown men” (Spivak 1988) that left out the violent and extractive history of colonial rule.
Postcolonial Contributions Postcolonial critiques such as those by Gayatri Spivak did not only emerge in the US academy but came from many movements, from projects of decolonization and critiques of contemporary colonial histories, and from many global regions and
5
Theories of Discrimination: Transnational Feminism
89
routes. While in the 1980s postcolonial studies scholars in the US were interested in seeing how Western biases misrepresented the lives of women and around the world, as Edward Said would argue in his groundbreaking work, Orientalism, they also worked to break down what would be called “orientalist” ideas in research (Said 1978). They hoped to dismantle the “tradition-modernity” divide that had seeped into social science, global health, international relations, and Development studies – all with profound impacts. They argued that seeing non-Western societies as “traditional” did not account for the profound changes that had created new modern ways of living everywhere. Nor did it encompass the traditional elements in Western societies. Postcolonial feminism critiqued the production of universalized categories and the overwhelming power of US and European knowledge production in the making of empire and even in anti-colonial nationalisms (Narayan 1997; Jayawardena and Zakaria 1986; Shohat 2001). Feminist postcolonial theorists, however, did address the relations across boundaries, across cultures and international domains, which produced different, gendered subjects in relation to global and other projects (national and subnational as well as regional) in historically specific ways. They examined orientalist biases in the representation of women who were not white or European and critiqued the exoticization and sexualization of female subjects (Yegenoglu 1998). Gayatri Chakravorty Spivak, for instance, suggested that the very project of representation in European culture was warped by its colonial history and the problematic concepts and categories that comprised knowledge. Her approach also engaged with thinking of the global circulation of representations of non-Western women (Spivak 1985), especially in the ways that Western women’s subjectivity was constituted by contrast with the presumed unfree nature of the Asian woman. The questions of how modernity made different subjects globally, how patriarchies differed, and how what it meant to be a women (or to have a gendered identity) varied historically and regionally were both critically rethought (Weinbaum et al. 2008). While scholars questions how Africa had been represented by colonial knowledge (Mudimbe 1988), African feminists examined how gender was different in Africa, for instance, and whether feminism could be dissociated from its Western history of representing Africa through racist notions of savagery or backwardness (Oyěwùmí 2003). Postcolonial theory was also concerned with the problems and theorizations of nationalisms and desires and difficulties of decolonization because of how colonial laws and institutions had become embedded. However, though it engaged in a critique of nationalism and nationalist histories, it did not engage as much with internationalism or with diasporic and migrant populations. While postcolonial theory brought to light the problem of development and modernization, geopolitical and national inequalities, and commensurabilities, they also relied on the integrity of national and territorial boundaries. It was the work of artists such as Trinh. T. Minhha, a Vietnamese-American filmmaker and scholar, who bridged the divide between postcolonial studies and the important area of research that comprised race/ethnic studies in the United States (Trinh 1989).
90
I. Grewal
Race/Ethnicity Critiques Trinh. T. Minh-ha’s films and writings suggested that representation of brown and black women in the United States was fundamentally also about problems in colonial esthetics and how Western cultures had come to see Asian women as sexualized and subordinated. As a woman of Vietnamese origin who came to the United States after its war, she examined how colonialism produced representations of difference, racial and cultural, which came from the history of white European anthropology; she argued that women of color artists and writers could create other representations that would resist the colonial ways of seeing. Trinh’s contributions came not just from postcolonial but also from the emerging area of activism and research that had come to be called ethnic studies in the United States. It had begun in the 1960s and took shape in US universities also as Asian American, African American, Native American, and Latino studies. Emerging from the ferment and activism of the 1960s, and with solidarity with “third world movements,” it suggested that the United States itself was a racial nation, with long histories of discrimination. Thus, the United States needed to be understood through both its racial heterogeneity and racial hierarchy. Underfunded and precarious in the academy, scholars in these fields focused on the US history of racial exclusion, seeking to be part of “third world” liberation and arguing for understanding ongoing racisms and the histories of colonization of Native American lands, use of Asian labor in the making of the country, and United States foundational economy of slavery. Engaged in both scholarship and activism, with concerns to educate population within and outside the academy, it changed the conversation in numerous fields of research to engage with racial history and its contemporary formations. For feminist research, this focus on race and racism was critical for examining exclusions by white feminists, for showing how “women of color” or black or brown feminists had to struggle for space and representation (Taylor 1998, Hartman 2008, 2019); in the process, they showed how whiteness itself was a form of privilege. Black feminism emerged as a specific historical formation, a subjectivity forged out of a history of enslavement, Jim Crow, and ongoing racisms that operated in particular ways on the bodies of black women (Combahee 1977, Collins 2008, Hull 1982). Many argued that black women should not call even themselves feminists since the term and the concept referred to notions of equality and modernity that were white and European (Walker 2003, Oyewumi 2004). The race critique in legal scholarship, entitled Critical Race Studies, revealed how the law itself was an instrument of racial power and that it was not a neutral arbiter but rather a structure that enabled the power of white privilege (Bridges 2019). All these critiques changed feminist research in the US. They broke that consensus, produced by the first generation of academic feminist scholarship in the US and UK academies, of a universal woman and a global feminism. Scholars examining the United States as a slave-holding and colonial society began to examine how race divided black and white women and how race was a reason why there could not be one feminism or unified feminist movement. Postcolonial studies feminists from Asia and Africa and the Caribbean argued that colonialism’s violence included the
5
Theories of Discrimination: Transnational Feminism
91
making of gendered subjects through new patriarchies and that African women, for instance, had to produce their own feminism (Oyěwùmí 2003). Other scholars, participating in the uprising for racial justice in the United Kingdom, called for analysis of feminisms as forms of racial and imperial power. Valerie Amos and Pratibha Parmar’s phrase, “imperial feminism” (Amos and Parmar 1984), and Hazel Carby’s “White Women Listen” opened the doors to critiques of Western feminism, based on questions of colonialism, empire, and the racial difference (Carby 2000). Such work led to calls for Western feminism to recognize its own patriarchal and racial formations and limits. In particular, feminists asserted that women were different because of race and culture, and so their needs were different, and colonial histories mean that there was no commonality across women globally. Black feminism, also emerging as a “womanist” rather than “feminist” politics, argued that histories of racism and slavery meant that feminism was different for black women because it involved a struggle against slavery and racism and for civil rights rather than one solely against patriarchy. Thus, a universal concept of feminism, one that said that globally all women were together in a struggle against misogyny and patriarchy, was challenged by black feminisms. Many of the United States-based feminists such as Angela Davis, bell hooks, Audre Lorde, and Gloria Anzaldúa also emphasized that race, class, and sexuality were central to the work of imperialism, capitalism, and race in producing subordination through gender. The influential collection, This Bridge Called My Back, published first in 1981, which brought together writings by “US women of color,” had also become prominent, first in radical anti-racist feminist circles in the United States and also in academic ones, and it brought out questions of racial difference among feminists of different ethnicities and races, suggesting that “women of color” was a new formation that would work against “white” feminism (Moraga 2015). Feminist scholars addressed the relation between race, class, sexuality, and nationalism in terms of the “double” or “triple” oppression caused by these multiple factors; the famous statement by the Combahee River Collective is one example of this work. Additionally, feminist theorists of color from Audre Lorde to bell hooks, to novelists and writers, emphasized the feminist critique of heteronormative family and nationalism. Critical race studies coined the notion of intersectionality, as an answer to understanding the multiple oppressions experienced by black women (Crenshaw). Kimberlé Crenshaw’s (1991) innovative use of this term was a means to address the problem of the gender-versus-race argument, by which women of color were often asked to choose politically between their racial or feminist identities. Ethnic studies also focused on a history of colonialism, particularly the colonization of the Americas and the attempt by white settlers to force Native Americans into servitude, poverty, or removal from their ancestral lands. The concept of “settler colonialism” has become critical in understanding the racial and colonial logic of “manifest destiny” of Westward expansion by European settlers. Ethnic studies also brought attention to Asian American and Latin American histories, examining histories of migration and the making of national boundaries (in Texas and California). Gloria Anzaldua suggested that borderlands were distinct spaces of racial
92
I. Grewal
history and hybridity (Anzaldúa 2012). More recently, research on Muslim American and Arab American groups has both crossed these ethnic divides (because of the long history of black Muslims in America) and produced new questions regarding the making of the “Muslim” as a racialized category in the aftermath of 9/11 (Ahmad 2002).
Diaspora Studies If the division between international and domestic, or that between the national and the international, was maintained within postcolonial and US ethnic and race studies, it was the study of diasporas, particularly of brown and black populations in the West, that revealed the gaps of so much research that took the nation-state and citizenship as unproblematic and settled concepts. Emerging out of the Birmingham School of Cultural Studies in the United Kingdom, from interventions based on race and racism, and drawing on the concept from Jewish communities, diaspora studies was a project of understanding racial Others in the West, on the one hand, and on studying their relation to nationalisms in the West, on the other. This understanding of diaspora, as against white nationalism also came from the formation of “Black Britain” as a political movement emerging from protests against racism and which brought together British Asian and Caribbean communities. In the work of scholars such as Stuart Hall, Paul Gilroy, and Hazel Carby, it also examined how class in Britain had to be understood through considerations of race and racism (Carby 1999; Hall 2018; Gilroy 1991); these scholars along with artists and activists such as Sutapa Biswas and Ingrid Pollard intervened in the formation of a new critique of esthetics based on race. The specificity of diaspora, in its distinction from the place or origin or settlement, was a key idea. In the work of Paul Gilroy, theories of nationalism were modified to suggest that racialized communities did not just assimilate but came to have a double consciousness, as W. E. B. Du Bois had argued, producing strategies of difference and identification that could resist the dominant white nation. Theorizing diasporas as being against dominant nationalisms, the academic work of Black British Cultural Studies joined with postcolonial and race critiques to examine exclusionary white nationalisms as ongoing expression of imperial power. Diaspora studies became also a way to think about the racial diversity of subjects of empire and to show how some communities remained connected across continents. In this latter approach, it linked up with studies of the Jewish diaspora to suggest that diasporas spanned continents and regions and comprised mobile populations and groups. Over the last decades, “diaspora” has become a key term within cultural criticism. It has become widely used to signal a new way of thinking about race, gender, colonialism, and globalization, and migration and enabled the study of emerging and changing cultures resulting from global migrations. Previously understood predominantly in terms of assimilation or acculturation, or as movement from home to host, migrations were seen within diaspora theories as creating racially non-white
5
Theories of Discrimination: Transnational Feminism
93
communities as sites of resistance, rather than within a teleology of assimilation. It has allowed researchers to think of communities as racialized communities struggling against violent white societies and connected to other similar struggles globally rather than solely in terms of their place of origin as well as their difference from both origin and place of settlement. Diaspora revealed that cultures and nations could travel and move and be heterogeneous and movable and that the boundaries of nations could not be mapped onto supposedly fixed territorial boundaries. Feminist scholars such as Saskia Sassan argue that diasporas form “countergeographies of survival” that are central to global cities and their patterns of migration and that in this survival it is important to pay attention to gendered patterns of work in relation to urban space (Sassen 2000). Avtar Brah theorized diaspora not as an identity but as a space, adding a feminist, relational, and geographical dimension to the concept and arguing that diaspora space is the site of intersection of varying formations of difference but with emphasis on gender, race, and class (Brah 1996). Emphasis on race and sexuality, nationalism, and diaspora came from the work of Chicana feminist, Gloria Anzaldua, formulating “border” theory and “borderlands” as a specific project of making racialized subjects in the United States (Anzaldúa 2012).
The Mobility Paradigm Along with an interest in diaspora, the 1980s also saw an interest in “mobility” and what came to be called critical mobility studies and the “mobilities turn” in social sciences. Epistemologies of flows (Appadurai 1996), oceanic connections (Gilroy 1993), intimacies (Stoler 2010), and historical diasporas (Chow 1993) disrupted the traditional disciplinary and interdisciplinary formations of research on the nationstate. While “mobility” was often a matter of engineering and technology research, cultural geographers and anthropologists began work on questions of epistemology and historiography in relation to mobility, not just of people but also of the movements of all sorts of aspects of society that were understood as static or fixed. In particular, this meant a rethinking of boundaries and borders of nation-states as well as of notions of settlement and nativism. Rather than focus on cultures and communities as bounded, could we think of them – as in diaspora studies – as moving? Was movement not formative of all communities, and not just some of them, so that we needed to consider migration as normative rather than exceptional? What was the historiography of movement that had been lost in the histories of settlement that became normative? Mobility research suggested that a bounded and fixed notion of the nation-state was insufficient to understand social and political formations both historically and, in the present, because its boundaries and its people were not fixed: who became a citizen, how nation-states changed by colonialisms, and how national boundaries could not be thought of as unchanging. Thus, the foundational assumptions of settlement as normative to all societies needed to be rethought, and theories of space and movement had to be considered more thoroughly. Doreen Massey’s feminist critiques of conceptualization of place and space argued that space was
94
I. Grewal
dynamic and that place was not frozen in time but was an ongoing process of making meaning of space – turning space into place. Geographers such as John Urry and Mimi Sheller argued for a sociology of mobility, of understanding different kinds of mobilities from tourism to commuting as a key to societies (Urry 2004). Tim Cresswell extended this work into a critique of “sedentarist metaphysics” that he argues pervades much of the ways in which knowledge and power operate, in that it assumes that people, nations, concepts, space, and place are settled within stable ideas of territory and that belonging and rootedness are normal, while mobility and movement are seen as suspect; he saw modernity as a move to deny that people and societies were on the move even before the modern period (Cresswell 2007). For anthropologist Liisa Malkki, this normative notion of rootedness and identity requires a theorization of mobility so that refugees were not understood as people out of place, but within a long history of the movement of people impacted by conflict or other catastrophic events (Malkki 1995). Joining with postcolonial and race critiques, researchers on mobility questioned Western modernity’s colonial claim of new or exceptional power made through narratives of exploration, encounter, globalization, migration, or travel, suggesting that such claims were made by erasing longer histories from other regions.
Transnational Feminism Transnational feminism came out of the political and social ferment of social movements and academic innovations mentioned above. Different theories of transnational feminism drew from critiques and projects of Development and from postcolonial studies the emphasis on histories and ongoing colonial and neocolonial impacts. From cultural geography, it brought the questioning of any distinction between spatial categories such as “local” or “global,” from diaspora studies, the concept of nationalisms as enabled or resisted by populations from afar and from within, as well as the focus on migration. Race/ethnic studies revealed that nations in the “West” were uneven and unequal spaces with histories and communities that stretched around the globe; migrations and displacements produced new solidarities – especially racial ones – across national boundaries. From mobility studies, transnational feminism brought questions of migration as a nation on the move, to thinking of border-crossings and borderlands as particular spaces of transnational memory and movements; finally, it enabled questioning of the notion that territorial nation-states were natural or stable and argued that social worlds and social life were produced by historically shifting connection and movement across borders. It argued against both the claim of a “global sisterhood” between women that would be united in resisting patriarchy (Morgan 1996) or the claim that non-Western patriarchies were the source of all problems for women outside the West. By the new century, transnational feminism or, as Grewal and Kaplan would call it, “transnational feminist practices” became a sprawling field of feminist research, with particular trends, approaches, and differences. There are three main strands of how “transnational feminism” is being used in academic contexts: one as an
5
Theories of Discrimination: Transnational Feminism
95
academic formation within research and teaching on gender that is predominantly focused on knowledge politics, disciplinary questions, and research methodology; this would be what Grewal and Kaplan would call “transnational practices” as a research analytic. A second approach is to understand this as an identity formation and as a form of international solidarity as in calling someone a “transnational feminist” or in positioning a scholar or scholar-activist or activist as being part of “transnational feminism.” A third is a focus on solidarities, inequalities and activism between feminists across national divides, and across academic and activist spaces, and there are divergent approaches to this question of connections. While there is often a blurring of the differences between these three approaches, all three usages reveal that work on transnational feminism emerged to break down binaries such as activist/academic, modern/traditional, domestic/international, and “local-global” and to understand global connections and differences. Many, though not all, of these approaches engage with the anti-colonial and anti-imperial and antiracist politics of feminist activisms that insist on understanding how imperial history and power undergirds feminist politics today, especially as they impact feminist connections and activisms across national boundaries. However, the capaciousness of the term, and its proximity to transnational capital and transnational corporations because of a shared terminology, and its focus on breaking down national boundaries have made for both critical reception and/or uneasy adoption. For some, it suggests the presence of transnational corporations, while others use it because it is precisely because it is a reminder of the context of globalized capital. Some use it as a synonym for international issues, and some as a stand-in for imperialism. But it is because of its interventions in the politics of knowledge, research, and activism that the term has gained resonance. In the anthology Scattered Hegemonies, published in 1994, coedited by Inderpal Grewal and Caren Kaplan, and also in another anthology, Feminist Genealogies, Colonial Legacies, Democratic Futures, by Chandra Talpade Mohanty and M. Jacqui Alexander (1997), all attempted to bridge understandings of subject populations in the United States with those outside it and to think about how histories of colonialism, feminism, racism, and modernity linked gender constructs and feminist movements across nations and nationalisms. Grewal and Kaplan argued that “local-global” spatial constructs erased the constitutive connections across national borders and that transnational forces were critical to making gender, albeit in different ways because of different hegemonic arrangement of power and specific histories of colonialism and imperialism. Working in a related but somewhat different framework, Mohanty and Alexander offered highly influential critiques of categories of “non-Western” women and Western development regimes, calling for transnational feminism as a response to capitalist and colonial as well as racial heteropatriarchies. They focused on the need for solidarity across national divides and the breaking down of activist/academic binaries to think about feminists collaborating for the purpose. Grewal and Kaplan saw transnational feminist approaches as a way to understand the hierarchical and asymmetrical relations among diverse feminisms created by myriad institutions and contexts globally, many of which resulted from European imperial cultures. They sought attention to the connections
96
I. Grewal
and links – rather than commonalities and differences, as in the prevailing comparative approach – between feminisms that produced power relations within them. They argued that these histories made critical differences between women so that transnational feminist practices involved unequal relations within cross-border projects; they argued that histories of modernity and colonialism created networks of power that influenced differing constructions of gender. Along with Mohanty and Alexander, they were critical of the “global sisterhood” assumption voiced by many Euro-American feminists, though others saw such a “sisterhood” as a strategic participation of feminists with their own cultural and political moralities making up a transnational public (Ong 1996). For Chandra Mohanty and Jacqui Alexander, transnational feminism enables understanding of solidarities based on racial difference that have been enabled by a critique of imperial capitalism. They focused on how racial empires have produced conditions of exploitations and how women have forged collective movements in resistance. Seeing their own work as enabled by collaboration across identities (from India and Caribbean) and understanding the divergent racial identities placed on them as they moved to the United States, Mohanty and Alexander were interested in thinking how “democratic futures” could be formed by collaborations and solidarities. With an emphasis on feminist praxis that resists the ways that postcolonial nation-states recuperate colonial gendered and sexualized hierarchies and subordination, Mohanty and Alexander called for activism that is based on resisting colonial and capitalist heteropatriarchies. They argued against “global feminism” that relies on Western norms and the need for a democracy uncoupled from capitalism and engaged with socialism and decolonization. A key aspect of transnational research focused on critiques of nationalism and the impact of colonial making of national boundaries. While some scholars in transnational research argued that the nation form was being overcome by global and regional entities and boundaries (such as the European Union), others critiqued nationalism rather than the nation form (Jayawardena and Zakaria 1986). Some saw the making of new nationalisms within transnational networks, especially in the role of diasporas in nationalist longings, suggesting that the transnational did not transcend nationalism even if it produced international and transnational alliances (Axel 2001; Mankekar 2015). Transnational approaches emphasized both the operations of transnational capital in the present and in the making of these nationalisms. Consequently, these researchers saw the national as the global-national, that is, reliant on global capital while claiming to be independent of it. Transnational feminist research was also critical of the “comparative feminism” approach that reduced the study of feminisms to nationally bounded “areas.” This transnational critique emerged out of a recognition that national comparison and comparative political and literary studies were a colonial project, seldom able to break out of the North-South binary, which was variously metaphorized as moderntraditional, advanced-backward, or mobile-static. Most so-called comparative work operated through comparisons between the Global North and Global South as separate, bounded entities and civilizations that could be mapped onto existing nation-states; transnational studies hoped to overturn this diagnosis of colonial
5
Theories of Discrimination: Transnational Feminism
97
epistemologies in favor of lines cutting across these boundaries and networks that influenced how gender was constructed through relations across nations. In particular, transnational feminist approaches were united in critiquing the erasure of the connections and linkages among empires and nationalisms and among the forms of power that circulated globally to create both colonialisms and anti-colonialisms. Such an understanding was important in showing how something called “foreign” was used to create something called “domestic” (Kaplan 2005), how social movements moved across national boundaries and became meaningful in other places by operations of power (Basu 2016; Savci 2021; Sameh 2019), and how boundaries and networks shift and change. It was because of this trenchant critique of empire, the power of networks (including feminist ones) and transnational capital, that the activist aspect of transnational feminism became an important formation. If some saw transnational feminism as an academic enterprise limited to theoretical debate, Richa Nagar and Amanda Lock Swarr argued for its possibilities in the practice of feminism (Nagar and Swarr 2010). They argued that the “praxis” aspect of these theories, which had applications for feminist organizing, could be understood as centering on alliances across geographies, communities, and class that would have to be long-standing and reciprocal, based on trust and relationships. They continued the focus on networks and the divides between feminists but suggested that these had to be countered by forming relationships between activists and researchers, by collective work on writing and translation across divides. They also focused on the critique of neoliberal capital and empire but suggested that connections could be made to transcend some of the hierarchies of nationalism, class, and gender in order to forge alliances. They saw themselves as doing the transnational activist feminist research that was being called for, but with the goal of making the relations that would create transnational activism across social and political borders. From a very different political and academic space, another strand of transnational research came to the forefront. Margaret Keck and Kathryn Sikkink, examining the advent of feminist organizing in policy and advocacy circles, argued in their book, Advocacy Beyond Borders, that it was feminist networks across national divides that would lead to robust policy changes at the international level, leading to changes in international policy (Keck and Sikkink 1998). They argued that a “transnational advocacy network includes those relevant actors working internationally on an issue, who are bound together by shared values, a common discourse, and dense exchanges of information and services” (Keck and Sikkink 1998, 2). Bridging theories from sociology and political science, this approach drew on research on social movements that crossed domestic and international politics to argue that emerging organizations such as NGOs (nongovernment organizations) could be activist spaces that involve the work of volunteers and create new relationships between women. While this approach did not address the histories of colonization or race that created hierarchical relations between women, it came to be a powerful approach to understanding how feminist international policy was being produced through feminist networks. Later contributions by political scientists, Myra Marx Ferree and Aili Tripp, have continued in seeing feminist practice in what they call the
98
I. Grewal
“transnational arena” as the “intersection of the international and the local” (Ferree and Tripp 2006, p. vii) that came about with the UN sponsored World Conference on Women in Beijing, which had an NGO Forum. They examine the use of organizations and treaties in shaping feminism as a “product of transnational dialogues and disagreements, coalitions and networks” (p. viii) and shaping what they call a “transnational opportunity structure,” by which they mean the organizations, policies, and networks that enable feminism to flourish in both international and local contexts. While they pay attention to inequality between women, they argue that the transnational arena provides a space for feminist organizing and representation of women from the Global South. They depart from the work of Grewal and Kaplan in the critique of “global feminism” as Western hegemony and from Mohanty and Alexander in their critique of global capitalism. Rather they see the global as providing both opportunity for feminist movements and organization to flourish in new networks across national borders. This strand of discussion and debate has been active in understanding questions of human rights and other international mechanisms used by feminists. This focus on transnational networks and activism has also emerged in the “theory” vs. praxis divide, with many researchers arguing that the “theory” approach does not address or encompass what is going on in the activist arenas. Some of these debates follow on Ferree and Tripp’s analysis to argue that global feminism is much more complex and cannot be seen as a space of a neocolonial hegemony (Mendoza 2002). They advocate for a more positive approach to feminist activism that suggests that transnational feminism is much more independent and critical of globalization processes that emerged in the 1980s and critical of the neoliberalism underlying them. Val Moghadam’s research on “transnational activist networks” suggests that these make up a “transnational public sphere” that emerged to address the inequities of globalization, particularly neoliberal capitalism and religious fundamentalisms, from a feminist perspective (Moghadam 2005). Debates on NGOs have proliferated in recent decades, as researchers have examined the nature of this field of activism that emerged at the end of the twentieth century, with some seeing these as another arm of imperialism, others understanding them as transnational activist networks, and yet others arguing that these represent the waning power of many nation-states in the Global South (Kamat 2002; Bernal and Grewal 2014; Thayer 2009). These new approaches to women-focused and feminist NGOs have included analysis of imperialism and power as central to the making of feminist networks. Importantly, it was not just gender but sexuality that came to be understood as transnational, as scholars studied how LGBTQ movements spread and how empires used sexual practices for power, as with the British-era sodomy laws that were imposed in British colonies across Asia, Africa, and the Caribbean (HRW 2008; Ngaruiya 2019). The research on transnational sexuality also engaged with the ways that sexual practices and sexual identities had a variety of histories that produced new visibilities and new terms through the work of modern nation-states and social movements. The impact of imperialism, not just in the past, was understood to be critical in the making of sexuality within national frameworks; nationalism and sexuality were seen as key aspects of histories of sexuality, nationalisms, and
5
Theories of Discrimination: Transnational Feminism
99
colonialisms. Critique of “global gay” formations that assumed that LGBT movements had “freed” oppressed gay people in non-Western and so-called unmodern spaces emerged along with research on sexual tourism, the politics of sexuality at the border (Luibhéid 2002), and the geopolitics of “pink-washing” that claimed that imperial states were more supportive of queer communities (Puar 2017). An important critique of “homonationalism” emerged to suggest that some of those with LGBTQ identities had embraced a nationalism that could be xenophobic and racist (Puar 2017). Within gender and sexuality research, there was an important formulation of “transnational sexualities” which included both a history of imperialism and ongoing power of the West (Alexander 2006) to inform the making of these identities. How had British-era sodomy laws criminalized sexualities and what made them continue into the present? What were the transnational impacts of new Christian evangelical missionaries who advocated also for criminalizing gay sex? How had the biomedical project of controlling the HIV/AIDS pandemic impacted sexualities? While some scholars saw terms such LGBT as Western imports (Massad 2008), others debated how to understand new visibilities created by these terms and what to make of practices of sexuality that existed beyond and in spite of the heteronormativity of colonialisms and nationalisms. The use of sexuality within the Global War on Terror by the United States was examined (Puar 2017), while others examined how HIV/AIDs created new categories of sexuality through public health practices (Dutta 2013). A third area of work was to understand the relation between migration, diaspora, and transnationalism (Manalansen 2003) and understand both the impact of border-crossing on sexual identity and the experiences of queers in diasporic spaces. All of this interest has led to discussion of transnational issues in the study of the history of sexuality (Wiesner-Hanks 2011). More recent directions in transnational feminism have built on much of the earlier work on empire, neoliberal capitalism, Development, global governance, and the rise of religious nationalism, as Anneeth Kaur Hundle suggests (Hundle et al. 2019). Jennifer Nash points out that the perception in the US academy is that transnational research is about international issues, while intersectionality seems to be concerned with the United States and questions of race (Falcón and Nash 2015). Others argue that transnationalism seems to be only about the United States; this view ignores the quite large body of work that is being done by scholars working outside the United States (Fernandes 2013). The question that remains even now is how to think race as a project of many Euro- American empires contending with disjunctive and different imperial histories and anti-colonial movements. Tina Campt and Deborah Thomas, in their special issue for Feminist Review on the theme of “Diasporic Hegemonies,” argue that a transnational feminist analytic helps in understanding the role of race and gender in diasporas (Campt and Thomas 2008). They argue that it enables a recognition of heterogenous forces that shape racism, migration, and displacement as well as the internal relations within diasporic communities; they focus on what they call the “intra-diasporic differences, asymmetries, and limits of diasporic relationality.”
100
I. Grewal
Feminist research had also used transnational approaches to think about migration and the history of colonialism as one about divergences and connections in racial formations, questioning notions of nativism and of belonging or non-belonging based on religion, race, or ethnicity (Ahmad 2017). From the robust work on migration and refugee asylum that questions these categories as constructed to create a sedentary and settled national body, to work that calls itself “critical refugee studies” and “critical security studies” (questioning the frameworks of state security and the making of human security states), transnational feminism questions the construction of nations and nationalisms, borders, and border zones (Noori 2020; Ghosh 2017). Migration research has seen a change from understanding the migrant as an interloper to seeing migration as a normative subject (influenced by mobility studies), whose mobility is not new in the modern world. It has also suggested that immigration as unidirectional move from “home” to “host” needs to be rethought in a context in which people move across multiple borders and as borders also move away from people. Furthermore, the territoriality of the postcolonial nation reveals shifting and mobile boundaries; the nineteenth-century colonial “carving” of Africa and the “partition” of South Asia and Palestine are examples, revealing that boundary control and making enable power over peoples, regions, and economies that preexisted these boundaries and now come to have transnational memories and imaginaries of previous times. Thus, it is important to think transnationally about national boundaries, to see how boundaries are a matter of power, and border zones are policed and militarized to produce a citizenry and a nation-state. Though this chapter has focused on scholarly strands and trajectories of transnational feminism, the term continues to change over time and with usage. Emerging politics and events and changing theories and methods in academia all will impact how terms are used and where they are used. The many debates within feminist research can be attributed to the tendency of feminist movements to be selfreflexive and to attend to changing social movements nationally and globally – which they must if they wish to remain relevant. What is clear is that over the decades since the study of feminism moved into academic sites in the 1970s, what was conceptualized as a feminism has now become understood as feminisms, and feminist research has shifted to understanding the histories of connections and the making of inequalities between gendered and, in addition, sexual subjects globally. Understanding how gendered subjects are produced in their transnational difference and connections remains an important task.
References Ahmad M (2002) Homeland insecurities: racial violence the day after September 11. Social Text 20: 101–115 Ahmad A (2017) Everyday conversions: Islam, domestic work, and South Asian migrant women in Kuwait. Duke University Press Books, Durham Alexander MJ (2006) Pedagogies of crossing. Duke University Press, Durham Alexander MJ, Mohanty CT (eds) (1996) Feminist genealogies, colonial legacies, democratic futures, 1st edn. Routledge, New York
5
Theories of Discrimination: Transnational Feminism
101
Amos V, Parmar P (1984) Challenging imperial feminism. Fem Rev:3–19. https://doi.org/10.2307/ 1395006 Anzaldúa G (2012) Borderlands/La Frontera: the new mestiza, 4th edn. Aunt Lute Books, San Francisco Appadurai A (1996) Modernity at large: cultural dimensions of globalization, 9th edn. University of Minnesota Press, Minneapolis Axel BK (2001) The nation’s tortured body: violence, representation, and the formation of a Sikh “diaspora”. Duke University Press Books, Durham Basu A (2016) Women’s movements in the global era, 2nd edn. Routledge, Boulder Bernal V, Grewal I (eds) (2014) Theorizing NGOs: states, feminisms, and neoliberalism. Duke University Press Books, Durham Boserup E (2007) Woman’s role in economic development. Routledge, London/Sterling Brah A (1996) Cartographies of diaspora: contesting identities. Routledge, London/New York Bridges K (2019) Critical race theory: a primer (concepts and insights). St. Paul, Mn: Foundation Press. https://www.amazon.com/gp/product/B07KMDQFL6/ref=dbs_a_def_rwt_hsch_vapi_ tkin_p1_i0. Butler J (2006) Gender trouble: feminism and the subversion of identity. Routledge, New York Campt T, Thomas DA (2008) Gendering diaspora: transnational feminism, diaspora and its hegemonies. Fem Rev 90:1–8. https://doi.org/10.1057/fr.2008.41 Carby HV (1999) Cultures in Babylon. Verso, London. First edition published by Verso 1999 edition Carby HV (2000) White woman listen!: black feminism and the boundaries of sisterhood. In: Owusu K (ed) Black British culture and society. Taylor & Francis, Abingdon, pp 82–88. https:// doi.org/10.4324/9780203360644_chapter_7 Chow R (1993) Writing diaspora: tactics of intervention in contemporary cultural studies, 1st edn. Indiana University Press, Bloomington Chowdhury EH (2009) Locating global feminisms elsewhere: braiding us women of color and transnational feminisms. Cult Dyn 21:51–78 Collins, Patricia Hill (2008) Black feminist thought: knowledge, consciousness, and the politics of empowerment (Routledge Classics): Hill Collins, Patricia: 9780415964722: Amazon.Com: Books. New York, N.Y: Routledge. https://www.amazon.com/Black-Feminist-ThoughtConsciousness-Empowerment/dp/0415964725 Combahee River Collective (1977) “The Combahee River Collective Statement. United States,” 2015. https://www.loc.gov/item/lcwaN0028151/ Crenshaw K (1991) Mapping the margins: intersectionality, identity politics, and violence against women of color. Stanford Law Rev 43:1241–1299. https://doi.org/10.2307/1229039 Cresswell T (2007) On the move: mobility in the modern Western world. Routledge, London Dingo R (2012) Networking arguments: rhetoric, transnational feminism, and public policy writing. University of Pittsburgh Press, Pittsburgh Dutta A (2013) Legible identities and legitimate citizens. Int Fem J Polit 15:494–514. https://doi. org/10.1080/14616742.2013.818279 Enloe C (2016) Globalization and militarism: feminists make the link, 2nd edn. Rowman & Littlefield Publishers, Lanham Escobar A (2011) Encountering development. Princeton University Press, Princeton Falcón SM, Nash JC (2015) Shifting analytics and linking theories: a conversation about the “meaning-making” of intersectionality and transnational feminism. Women’s Stud Int Forum 50:1–10. https://doi.org/10.1016/j.wsif.2015.02.010 Fernandes L (2013) Transnational feminism in the United States: knowledge, ethics, power. NYU Press, New York Ferree MM, Tripp AM (2006) Global feminism: transnational women’s activism, organizing, and human rights. NYU Press, New York Ghosh S (2017) Relative intimacies: belonging and difference in transnational families. Econ Polit Wkly 52:7–8
102
I. Grewal
Gilroy P (1991) “There ain’t no black in the Union Jack”: the cultural politics of race and nation, 1st edn. University of Chicago Press, Chicago Gilroy P (1993) The black Atlantic: modernity and double-consciousness, reissue edn. Harvard University Press, Cambridge, MA Grewal I, Kaplan C (2001) Global identities: theorizing transnational studies of sexuality. GLQ 7: 663–679 Hall S (2018) Essential essays. Duke University Press Books, Durham/London Hartman S (2008) “Venus in Two Acts.” Small Axe 12, no. 2 (2008): 1–14. Hartman S (2019) Wayward Lives, Beautiful Experiments. New York and London: W.W. Norton Company. https://wwnorton.com/books/9780393357622 Hooks B (1987) Ain’t I a Woman: Black Women and Feminism. London: Pluto Press. https://www. amazon.com/Aint-Woman-Black-Women-Feminism/dp/1138821519 Hull A (1982) Patricia Bell Scott, and Barbara Smith, eds. but some of us are brave. New York, N.Y: The feminist press at CUNY. https://www.feministpress.org/books-a-m/but-some Hundle AK, Szeman I, Hoare JP (2019) What is the transnational in transnational feminist research? Fem Rev 121:3–8. https://doi.org/10.1177/0141778918817525 Jayawardena K, Zakaria R (1986) Feminism and nationalism in the third world, reprint edn. Verso, London/New York Kabeer N (1994) Reversed realities: gender hierarchies in development thought, 4th printing edn. Verso, London/New York Kamat S (2002) Development hegemony: NGOs and the state in India. Oxford University Press, Delhi/New York Kaplan A (2005) The anarchy of empire in the making of U.S. culture. Harvard University Press, Cambridge, MA Keck ME, Sikkink K (1998) Activists beyond borders: advocacy networks in international politics, 1st edn. Cornell University Press, Ithaca Lorde A (1984) Sister outsider: Essays and speeches. Berkeley, CA: Crossing Press, 1984. https:// www.amazon.com/Sister-Outsider-Speeches-Crossing-Feminist/dp/1580911862 Luibhéid E (2002) Entry denied, 1st edn. University of Minnesota Press, Minneapolis Malkki LH (1995) Refugees and exile: from “refugee studies” to the national order of things. Annu Rev Anthropol 24:495–523. https://doi.org/10.1146/annurev.an.24.100195.002431 Manalansen M (2003) Global Divas: Filipino gay men in the diaspora. Duke University Press, Durham. https://doi.org/10.1215/9780822385172 Mankekar P (2015) Unsettling India: affect, temporality, transnationality. Duke University Press Books, Durham Massad JA (2008) Desiring Arabs. University of Chicago Press, Chicago Mendoza B (2002) Transnational feminisms in question. Fem Theory 3:295–314. https://doi.org/10. 1177/146470002762492015 Mies M, Werbner P, Werbner R, Federici S (2014) Patriarchy and accumulation on a world scale: women in the international division of labour, 3rd edn. Zed Books, London Minh-Ha TT (1989) Woman, native, other: writing postcoloniality and feminism, underlining/ margin notes edn. Indiana University Press, Bloomington Modern Girl Around the World Research Group (2009) In: Weinbaum AE, Thomas LM, Ramamurthy P, Poiger UG, Dong MY (eds) The Modern girl around the world: consumption, modernity, and globalization. Duke University Press, Durham Moghadam VM (2005) Globalizing women: transnational feminist networks. Johns Hopkins University Press, Baltimore Mohanty CT (1984) Under Western eyes: feminist scholarship and colonial discourses. Bound 2 12(13):333–358. https://doi.org/10.2307/302821 Mohanty CT (2003) Feminism without borders: decolonizing theory, practicing solidarity, 35275th edn. Duke University Press Books, Durham/London Moraga C (2015) This bridge called my Back, fourth edition: writings by radical women of color, illustrated edn. State University of New York Press, Albany
5
Theories of Discrimination: Transnational Feminism
103
Morgan R (ed) (1996) Sisterhood is global: the international women’s movement anthology, reprint edn. The Feminist Press at CUNY, New York Mudimbe VY (1988) The invention of Africa: gnosis, philosophy, and the order of knowledge, 1st edn. Indiana University Press, Bloomington Nagar R, Swarr AL (2010) Theorizing transnational feminist praxis. In: Critical transnational feminist praxis. SUNY Press, New York, pp 1–20 Narayan U (1997) Dislocating cultures: identities, traditions, and third world feminism, 1st edn. Routledge, New York Ngaruiya Njambi W (2019) What sexuality? Whose knowledge? Mapping “heterosexuality” and “homosexuality” within transnational feminisms. Gend Women’s Stud 2. https://doi.org/10. 31532/GendWomensStud.2.2.001 Noori S (2020) Navigating the Aegean Sea: smartphones, transnational activism and viapolitical in (ter)ventions in contested maritime borderzones. J Ethn Migr Stud 2020:1–17. https://doi.org/ 10.1080/1369183X.2020.1796265 Nussbaum M, Sen A (eds) (1993) The quality of life, reprint edn. Clarendon Press, Oxford, UK/New York Ong A (1996) Strategic sisterhood or sisters in solidarity? Questions of communitarianism and citizenship in Asia. Indiana J Glob Leg Stud 4:107–135 Oyewumi O (2004) African Women and Feminism: Reflecting on the Politics of Sisterhood. Trenton, N.J.: Africa World Press. https://www.amazon.com/African-Women-FeminismReflecting-Sisterhood/dp/0865436282/ref=asc_df_0865436282/?tag=hyprod-20&linkCode= df0&hvadid=312039872799&hvpos=&hvnetw=g&hvrand=762559162626175513& hvpone=&hvptwo=&hvqmt=&hvdev=c&hvdvcmdl=&hvlocint=&hvlocphy=9031535& hvtargid=pla-584586051796&psc=1&tag=&ref=&adgrpid=62149175916&hvpone=& hvptwo=&hvadid=312039872799&hvpos=&hvnetw=g&hvrand=762559162626175513& hvqmt=&hvdev=c&hvdvcmdl=&hvlocint=&hvlocphy=9031535&hvtargid=pla584586051796 Parikh C (2017) Transnational feminism. In: Goyal Y (ed) The Cambridge companion to transnational American literature. Cambridge University Press, Cambridge, UK, pp 221–236 Patil V (2011) Transnational feminism in sociology: articulations, agendas, debates. Sociol Compass 5:540–550. https://doi.org/10.1111/j.1751-9020.2011.00382.x Puar JK (2017) Terrorist assemblages: homonationalism in queer times, anniversary, 10th anniversary edn. Duke University Press Books, Durham Rai SM (2013) The gender politics of development: essays in hope and despair. Zed Books, London Reilly N (2011) Doing transnational feminism, transforming human rights: the emancipatory possibilities revisited. Ir J Sociol 19:60–76. https://doi.org/10.7227/IJS.19.2.5 Said E (1978) Orientalism. Pantheon Books, New York. Sameh CZ (2019) Axis of hope: Iranian women’s rights activism across borders. University of Washington Press, Seattle Sangari K, Vaid S (1990) Recasting women: essays in Indian colonial history. Rutgers University Press, New Brunswick Savci E (2021) Queer in Translation. Durham N.C.: Duke University Press. https://www. dukeupress.edu/queer-in-translation. Sassen S (2000) Women’s burden: counter-geographies of globalization and the feminization of. . . J Int Aff 53:503–524 Shohat E (2001) Talking visions: multicultural feminism in a transnational age, Documentary sources in contemporary art. MIT Press, Cambridge, MA. Paperback – common Spivak GC (1988) Can the subaltern speak? | Gayatri Chakravorty Spivak | academic room. In: Nelson C, Grossberg L (eds) Marxism and the interpretation of culture. University of Illinois Press, Urbana, pp 271–313 Stoler AL (2010) Carnal knowledge and imperial power: race and the intimate in colonial rule. University of California Press, Berkeley. Paperback
104
I. Grewal
Swarr AL (2010) Critical transnational feminist praxis, illustrated edition, SUNY series, praxis: theory in action. SUNY Press, Albany Thayer M (2009) Making transnational feminism: rural women, NGO activists, and northern donors in Brazil. Routledge, New York This Alien Legacy (2008) Human Rights Watch [WWW document]. https://www.hrw.org/report/ 2008/12/17/alien-legacy/origins-sodomy-laws-british-colonialism. Accessed 19 Feb 2021 Trinh, Minh-Ha T (1989) Woman, Native, Other: Writing Postcoloniality and Feminism. Underlining/margin Notes edition. Bloomington: Indiana University Press. Urry J (2004) Tourism mobilities, 1st edn. Routledge, London/New York Walker A (2003) In Search of Our Mothers’ Gardens: Womanist Prose. Reprint edition. Orlando: Mariner Books. Wiegman R (ed) (2002) Women’s studies on its own: a next wave reader in institutional change. Duke University Press Books, Durham Weinbaum, Alys Eve, Thomas, Lynn M, Priti Ramamurthy, Uta G. Poiger, Madeleine Yue Dong, and Barlow, Tani, eds. (2008) The Modern Girl Around the World: Consumption, Modernity, and Globalization (Next Wave: New Directions in Women’s Studies): The Modern Girl around the World Research Group, Weinbaum, Alys Eve, Thomas, Lynn M., Ramamurthy, Priti, Poiger, Uta G., Dong, Madeleine Yue, Barlow, Tani: 9780822343059: Amazon.Com: Books. Durham N.C.: Duke University Press. https://www.amazon.com/Modern-Girl-Around-World-Globaliza tion/dp/0822343053 Wiesner-Hanks ME (2011) Crossing borders in transnational gender history*. J Glob Hist 6: 357–379. https://doi.org/10.1017/S1740022811000374 Yegenoglu M (1998) Colonial fantasies: towards a feminist reading of orientalism. Cambridge University Press, Cambridge, UK/New York ﺝﺩﻟﻴﺔ, J, Jadaliyya (n.d.) Gay rights as human rights: pinkwashing homonationalism [WWW document]. Jadaliyya – ﺝﺩﻟﻴﺔ. https://www.jadaliyya.com/Details/24855. Accessed 19 Feb 2021
6
Discrimination as Focal Point Markets and Group Identity Kaushik Basu
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Markets and Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Art of Creating Optimal Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Policy Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Addendum. Empirical Tests and a Normative Caveat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . New References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
106 107 110 112 113 115 115 115
Abstract
This chapter presents a theory of discrimination for markets in which there are complementarities between different tasks. It is shown that, in such a setting, even when groups are a priori identical, employers will end up discriminating against certain groups. Group discrimination serves the purpose of creating a focal point in a market game. In this model, the free market, far from curbing discrimination, nurtures it and thereby creates the need for purposive policy intervention. It is argued that, with the rise of technology, the problem of discrimination as focal point will get more acute and we will have to think in terms of affirmative action or a system of taxation and subsidy to support groups that get excluded. Keywords
Discrimination · Focal point · Strategic complementarity · Affirmative action
K. Basu (*) Department of Economics and SC Johnson College of Business, Cornell University, Ithaca, NY, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_8
105
106
K. Basu
Introduction For good or for bad, group identity matters in determining market outcomes (Akerlof and Kranton 2010; Sen 2006). Discrimination against certain groups of people and, by its converse, in favor of other groups has been common practice, observed in different societies and through different periods of history. India’s caste system, with its attendant practice of intolerance and effort to marginalize large groups of people, is an example, as is the history of apartheid in South Africa and racial discrimination and slavery in the United States. From an analyst’s point of view, they are often troubling because norms and laws often merge into each other. Some of these heinous practices were explicitly backed by the law as in the case of South African apartheid and US slavery. At other times, such as with India’s caste system through history and now (see Deshpande 2010) or with racial discrimination faced by African Americans in contemporary United States, it was not backed by the law but by social norms, customs, and individual beliefs and preferences. The focus in this chapter will be on discrimination which does not have the backing of law. As a mirror image of this, we often see certain groups benefiting from discrimination in their favor. This has been true of men through long stretches of history and even now in most societies. Similarly, in the United States, the United Kingdom, India during colonial times, or South Africa till recently, if you could choose your skin color, I would strongly recommend white. Where do these discriminatory preferences come from and why have they been so persistent? Without doubt, there must be many explanations for this, ranging from plain, simple bigotry and prejudice to various forms of statistical discrimination that economists have written extensively about.1 While not wanting to take away from those standard explanations, this chapter presents a novel argument whereby discrimination has no innate origins but arises naturally in markets where there happens to be some complementarity between the different tasks we do.2 This kind of discrimination is a product of the free market and the beliefs of ordinary people. The upshot is the group identities of people come to matter in such settings. It is, as I will show, closely linked to the idea of “focal point,” used in game theory (Schelling 1960). In particular, race, caste, and gender become important in equilibrium because they acquire the salience of the focal point. In some ways, the focal point theory of discrimination developed in this chapter is more alarming than other forms of discrimination where one can point to the source, be it human mendacity or the distortions of statistical information. In my analysis, removing government interference and allowing competition in the market to flourish do not remove discrimination, as standard economics had suggested, because it is in fact a product of precisely the free market. The model is based on sufficiently realistic assumptions to make me believe that, while other forms of discrimination no doubt occur, the focal point theory of
1 2
See Arrow (1973, 1998), Phelps (1972), Spence (1974), and Stiglitz (1974). A polar case of this in a development context occurs in Kremer (1993).
6
Discrimination as Focal Point
107
discrimination does play an important role and, as such, deserves greater attention. In the last section, I will argue why, given trends in contemporary labor markets, this is likely to become even more important.
Markets and Discrimination A good starting point is the celebrated paper by Bertrand and Mullainathan (2004). This is often treated as the most compelling empirical demonstration of pure racial bias in labor markets. When we see discrimination, the question always arises as to whether it is bias or reflection of something else that correlates with race or gender or caste, whatever it is one is studying. Thus, if an employer hires more whites than blacks, is it really a preference for whites, or is it merely a reflection of the fact that the employer needs PhDs and white job applicants are more likely to have PhDs? Bertrand and Mullainathan corrected for this by sending out job applications with fictitious resumes to help-wanted advertisements that appeared in Chicago and Boston newspapers. It was soon evident that, controlling for all other things, candidates with white names were far more likely to get callbacks for interviews than those with black names. There were striking results, such as how having a white name is equivalent to 8 years of work experience with a black name. In brief, they had engineered the celebrated “ceteris paribus” condition that traditional economists so often talked about but were seldom able to demonstrate. And the findings were striking.3 What I wish to do here is to question whether this necessarily demonstrates racial bias. Note that for most tasks in life, to conduct them effectively, you need to successfully do other tasks. If you work for a firm’s sales department to promote sales, you need to be able to successfully interact with buyers’ groups and delivery services units. If the buyers’ groups and delivery services units try to shun you, you will not be able to do the sales work you are supposed to do well. And of course, the problem is similar for the buyers’ group. When they reach out to you, they know they will get better services from you if you are trusted and used by the sales department and the delivery services unit and likewise for the delivery services unit. They have to gauge how successful you will be with the sales department and the buyers’ group. This is where race can come to acquire significance even in the absence of any innate racial preference. If you feel Emily – a common white name – is more likely to do your task more effectively, you will prefer to hire Emily over Lakisha. If all three units do that, this becomes self-fulfilling. The white name provides a focal point in a 3
Similar results were reported by Siddique (2008) who sent out applications in India using castebased names. A paper by Thorat, Banerjee, Mishra, and Rizvi (2015) does a similar test for the home rental market in the National Capital Region, in and around Delhi, and record a similar bias against Muslims and Dalit applicants. For empirical studies, using other methodologies, which nevertheless suggest pure racial bias, see Hamilton et al. (2015) and Pager, Western, and Bonikowski (2009). For some engaging research based on legal analysis, see Sander (2006) and Coleman and Gulati (2005).
108
K. Basu
labor market for tasks that exhibit “strategic complementarity” – economists’ term for work contexts where doing one task raises your productivity in another. Since my chapter is a methodological intrusion into that rather barren terrain between economics and sociology and is, at the same time, at deviance from the so-called Chicago school, it may be apt to quote Ken Arrow from an interview he gave to Richard Swedberg in 1988, in which he distances himself from the Beckerian approach but stresses the importance of interaction between economics and sociology: “[A] lot of the environment in which economic transactions take place is social and historical in nature. I do not know exactly how to fit these pieces together but there are, for example, accounts of how special groups have played a distinct role in, say, trade. You have Chinese middlemen in Asia; Jews at certain times; Quakers during one period; and so on. It is clearly their social characteristics that matter” (Swedberg 1990, pp. 136–137). What I am arguing is that this is true but the social characteristics may well be endogenous, the product of equilibrium, however we got there. The basic idea, which shows the relation between discrimination and focal point, can be illustrated with a simple example.4 There are two entrepreneurs, 1 and 2, who have need for certain tasks to be done, and there are n (>2) service operators or laborers who can do these tasks. Suppose, for instance, entrepreneur 1 needs a person to look after his lawn – buy and apply fertilizer, sow seeds, mow, and so on – and entrepreneur 2 wants to lend money to someone. The person who is able to borrow the money can buy fertilizers and seeds easily and so does the lawn work better. And the laborer who gets the lawn contract will be more likely to pay back loans that she takes. The entrepreneurs do not know the underlying causation of what makes a laborer more productive, to wit, the fact that if both reach out to the same laborers, they get better value. This is not unlikely in a real setting where thousands of entrepreneurs reach out to hundreds of thousands of laborers. They realize some are more productive than others and may search for markers of that without quite knowing the fundamental model that drives this. The above paragraph may be summed up as follows. Each of the two entrepreneurs picks one citizen for the task he needs to get done. If he picks a citizen who is not picked by the other entrepreneur, he gets a benefit of x, and if he picks someone the other entrepreneur also picks, he gets y. Given what we said about strategic complementarity: y>x
ð1Þ
The entrepreneurs are not aware of this strategic complementarity. All they know is that they may get x or y, without being aware of what drives the difference. The only critical assumption in this exercise is (1). All other structures of the model can be varied, and we will still get the same essential result. I should clarify that I am not claiming that strategic complementarity is always the case but simply that it is
4
A more complex and also more realistic model is developed in Basu (2015).
6
Discrimination as Focal Point
Table 1 The discrimination game
109 N
N
y n
þ
W
y n
þ
B
y n
þ
W xðn1Þ n xðn1Þ n xðn1Þ n
y n
þ
y n
þ
x
B xðn1Þ n xðw1Þ w
y n
Þ þ xðn1 n
x y n
Þ þ xðb1 b
realistic in many situations, and when that happens, I want to show that a kind of discrimination happens which requires no innate bias and no differences in ability or skill across groups and arises entirely through natural market processes with the concept of the focal point playing a critical role. To convert this to a game, I need to put in a little more structure to the model. Assume that the n laborers are of two different races: w (>1) of them are whites and b (>1) blacks. Hence w þ b ¼ n. Each entrepreneur, in selecting a laborer to get his or her task done, has to use one of the following rules: no discrimination (strategy N), discrimination in favor of whites (strategy W), and discrimination in favor of blacks (strategy B). If she chooses N, it means she randomly picks one of the n citizens, with 1/n probability of each citizen being chosen. If she chooses W, it means each white person faces a probability 1/w of being selected, likewise for strategy B. It is easy to work out the payoffs of the two entrepreneurs depending on the choice each of them makes. This is displayed in the payoff matrix described below, in what I call the discrimination game. Since it is completely symmetric, there is no need to show the payoffs of both entrepreneurs. I show the payoff earned by entrepreneur 1; and, in this game, entrepreneur 2 gets the same (Table 1). To understand the payoff, let us check the top left hand box. Both entrepreneurs choose N, that is, pick a laborer with no attention to race. After one has chosen, the probability that the other person will choose the same laborer is 1/n. When that happens, each gets a payoff of y. The probability that the other entrepreneur will choose someone else is (n 1)/n. When that happens, each gets x. So the expected payoff is y/n þ x(n 1)/n. It is easy to work out the payoffs in the other boxes by a similar reasoning. It is simple to check that this game has three Nash equilibria – (N, N), (B, B), (W, W) – that is, no one discriminates, everybody discriminates in favor of whites, and everybody discriminates in favor of blacks. To check this, note that if the other person chooses N, no matter what you do, you will get the same payoff. So you cannot do better by unilaterally deviating from N. Next check that, as long as y exceeds x, as assumed, and given that, by definition, n > w, the following are true: y=w þ xðw 1Þ=w > y=n þ xðn 1Þ=n, and y=w þ xðw 1Þ=w > x: In other words, if others discriminate in favor of whites, whites will be on average more productive, and so it is in your interest to choose a white to do your task. In other words, (W, W) is a Nash equilibrium. For the same reasons, (B, B) is an equilibrium as well.
110 Table 2 The discrimination game: a special case
K. Basu
N W B
N 5/4, 5/4 5/4, 5/4 5/4, 5/4
W 5/4, 5/4 3/2, 3/2 1, 1
B 5/4, 5/4 1, 1 3/2, 3/2
For those with an aversion to symbols, let me convert the above game to a society in which there are four laborers, two whites and two blacks. And suppose y ¼ 2 and x ¼ 1. By inserting these values, the above discrimination game collapses into the special case illustrated below (Table 2). The three Nash equilibria are now obvious. If others discriminate, you had better do the same. But as always with games with many equilibria, there is a need for a focal point which allows players to coordinate their behavior. What I am claiming is that in markets with strategic complementarity, as just described, race or gender or caste can be the focal point. It is important only because others think it is important. You prefer Emily to Lakisha not because you have a preference for white over black but because all of you need to zero in on some group and it so happens, for reasons of history or whatever, you have settled on whites. One important implication of this is that one popular view, namely, if you leave it all to the market, with no government regulations and intervention, racial and caste discrimination would go away, is not valid. Discrimination arises from the free market. If you want to stop discrimination, you may, in fact, need regulation and conscious affirmative action. And when we go for affirmative action, we must not indulge in the politically correct banter, so often heard, that by doing affirmative action you do not hurt your returns. The truth is that your returns may indeed be diminished by such action. The appeal has to be that even if your return drops, there are certain actions in life which ought to be indulged in for its innate moral goodness. Affirmative action may be one of those. I shall return to this in the last section.
The Art of Creating Optimal Groups As a digression, I may point out that this focal point model of discrimination can be put to some rather Machiavellian uses, such as that of creating your own group and then promoting it. And indeed, such practices are not unknown in the world. Alumni associations and fraternities are good examples of this. They are often used to promote the college or frat label. Thus, we are told how Harvard students are smarter than others, or how Berkeley students are more productive than others, or how Cornell graduates are more creative than others (this one happens to be true), and so on. What the analysis in this chapter shows is that, once such beliefs catch on, they can become self-fulfilling because that belief then serves as a focal point. Most of us, human beings, have multiple identities, race, nationality, language group, ethnicity, gender, and so on (see Sen 2006), and as I just argued, we can also create new identities (see also Basu 2011). Now, if you want to deliberately nurture the view that one of these identities is a mark of greater productivity, this model
6
Discrimination as Focal Point
111
suggests it may be worthwhile to pick on or create a group that is relatively less populous. To prove this, consider the payoff an entrepreneur gets from the equilibrium in favor of whites. This is given by [y þ (w 1)x]/w. The payoff in an equilibrium in which blacks are favored is given by [y þ (b 1)x]/b. Let us call the former whincome, and the latter blincome. Recall b ¼ n w. Hence, as w becomes smaller, whincome rises and blincome falls. Both (W, W) and (B, B) are still Nash equilibria, but the former becomes more and more dominant as the white population becomes smaller. In brief, if you want to promote the idea that a particular group you belong to is more productive, you will be better off if you choose a small group. Among other things, this explains why raising the profile of women is such a hard task. They constitute roughly half the population. Considering the case of nationalities, if you promote the idea that the British are more productive and the idea that Chinese are more productive and people buy into this belief, the British will turn out to be even more productive in a world in which they are believed to be more productive than the Chinese will be in a world in which the Chinese are believed to be more productive. There is often surprise expressed at the fact that Britain was such a small nation that once ruled virtually the world. What is being said is that we should not be surprised. One counterargument to this needs to be kept in mind. If smaller groups are more effective, why not go all the way and proclaim that you as an individual, or one-person group, is more productive? The reason must be that as groups get too fragmented, people cannot hold information about group characteristics in their heads because there are too many groups. This can set a lower bound to how small groups can be without losing advantage. To model this formally will entail a more elaborate analysis, but the intuition behind what I am arguing should be evident. In closing this section, I want to remind the reader that while this focal point model of discrimination is very important, as with all theory, to take it to the real world and to put it to use, we must enrich it with our commonsense and reasoned intuition.5 Hence, the above theory must be combined with other ideas and our own experience before it is put to use or employed in designing policy. It is, for instance, worth reminding ourselves that productivity and even intelligence are also dependent on how a person is treated and on how society views this person’s group. Even if the discrimination is purely a focal point at work, it can leave scars on people, making the ones believed6 to be less intelligent actually behave less intelligently. Unlike in a strategic-form game where a switch from one equilibrium to another can It is not a matter to go into here, but this reference to commonsense and “reasoned intuition” is not a casual side remark. I have argued at length elsewhere that for science to be useful, we must combine it with these skills. Pure analysis of data or pure theory cannot help us help the world till we combine them with reasoned intuition (Basu 2014). 6 Among the most notable findings on this are studies by Ambady, Shih, Kim, and Pittinsky (2001) and Hoff and Pande (2006). See, also, Field and Nolen (2005), Hoff (2015), and World Bank (2015). 5
112
K. Basu
be effected in the twinkling of an eye, in reality, these changes are likely to take time and involve economics and psychology.
Policy Implications The model presented here opens up some policy dilemmas. It should be evident from the above model that GDP or the aggregate payoff earned by all is higher when there is group discrimination. Since a certain amount of multitasking enhances productivity, it is better to have a subset of the population multitask, that is, do all the work, rather than spread tasks thinly across all. Some people may find this result troubling, namely, the fact that discrimination against a group can enhance GDP. In my view, this result is disturbing only if you consider a higher GDP to be sacrosanct. All these results suggest is that you should be willing to forego some GDP to achieve fairness and greater equity across groups. Stewart (2005) pointed out that, in contrast to vertical inequality (that between persons), the normative economics of horizontal inequality (inequality across groups) remains a rather neglected subject. Exactly how one develops this and characterizes the various trade-offs between aggregate well-being and horizontal equality can matter a lot in how choices are made, as we just saw (see also Jayadev and Reddy 2011; Subramanian 2011). Fortunately, the context is simple enough in the discussion that follows that the exact trade-offs will not matter but this is indeed a subject that deserves greater attention in the future. With this in the background, note that there are two ways of achieving equity or an equitable distribution of incomes or payoffs in the present model. The first is straightforward “affirmative action.” Incentivize employers to choose a diverse workforce so that the total work is spread equitably across all individuals. The second is to let a limited number of people do all the work and then tax them and subsidize those who did not get work. This latter will result in a higher per capita income since the workforce (i.e., the people who find work or are called upon to do tasks) will be more productive. There are social scientists who have objected to the latter on the ground that work in itself gives people dignity; so even if one were to get the same ultimate income but without having to work, this may cause a diminished sense of self. I prefer to be cautious with this argument. Through history, there have been the leisure classes – the British landed aristocracy, the Indian zamindars (a brainchild of the British landed aristocracy) – who did precious little work and lived lives of luxury, and there is no evidence of them feeling diminished by the experience. There are also cases, especially relevant to women, where voluntary nonwork is, in fact, a statement of empowerment and an act that enhances agency (Basu 2016). Basically, what people need is a sense of legitimacy for what they earn. There are contexts where to get a dole without getting to work is offensive; it is like being told that you are not fit for work. This never troubled the aristocracy and the leisure classes because they had a sense of entitlement for their ample incomes, even though where that entitlement came from is a puzzle. In a world where either one group gets
6
Discrimination as Focal Point
113
to work or the other does and there is no essential a priori difference between the two groups, for one group to work and the other to be subsidized is a legitimate strategy. Someone has to not work to make those who work more productive; in this setting, there is no indignity in not working. It is in fact a contribution to society. There is no need to resolve the policy dilemma in this chapter, but I want to point out that this problem is going to get more acute, since with the march of technology, relatively unskilled work is steadily shrinking in the world. The share of GDP that accrues to workers as aggregate wage bill has been falling over the last four or five decades. The trend is quite alarming. As I point out in Basu (2016), in 1975, total wage bill as share of GDP in the United States, Japan, and the European Union were 61%, 77%, and 66%. Now they are 57%, 60%, and 56%, respectively (see also Karabarbounis and Neiman 2014). This trend is true almost without exception in all high and middle income countries; and this is a challenge we will have to face up to sooner rather than later. What the model developed in this chapter tells us is that we face a choice – whether to forcefully distribute the limited work thinly across the entire labor force (thereby impairing productivity) or let few people work (and be productive) and then tax them to subsidize the ones without adequate work and with all the time in the world to read philosophy and swim. In itself, the advance of technology is a matter for celebration. What makes it worrying is that, as machines and robots displace workers, the incomes of workers become profits for the owners of the machines and the robots. I have discussed in Basu (2016) how we need to take on this problem head on. What this chapter points to is an additional problem. As work becomes scarce, in markets with strategic complementarity, there will likely be an exacerbation in the problem of group discrimination, with large groups being kept out of the labor market. This chapter provided an explanation of the mechanics of how this will happen and drew attention to the kinds of policy questions that, sooner or later, we will have to confront.
Addendum. Empirical Tests and a Normative Caveat The central idea behind my chapter arose in response to the striking empirical finding of Marianne Bertrand and Sendhil Mullainathan (2004), in which they showed that, in hiring workers, people discriminate purely on the basis of race. This seemed to put an end to the debate as to whether racial discrimination was an outcome of the effort to spot talent that may not be visible but may, in equilibrium, correlate with race or a result pure preference for one race over another. The same study can be done for other group identities, such as gender or sexual orientation or religion or caste. It is the wonderful design of their experiment that establishes that this is pure discrimination and not a case of race or caste being used as a surrogate in search of other indicators that affect productivity. Moreover, we know from later experiments that these discriminations are persistent, especially the discrimination against blacks. There has been a slight abatement in discrimination against Hispanics (see Quillian et al. 2017).
114
K. Basu
The central argument of my chapter was that this empirical finding is, however, compatible with a world in which no one has an innate propensity to discriminate. If I need to hire someone who will need to work with others, I need someone others will want to work with. This strategic complementarity gives rise to the need for a coordination device. This is closely related to the importance of networks in labor markets, the importance of which in job-market discrimination has been seen in the United States and in India (see Royster 2003; Deshpande and Newman 2007). In such circumstances, the preference for a specific race or gender can be the outcome of the need to coordinate. If others make that discrimination, I need to do it as well. We could have favored whites or blacks. There is no innate advantage in preferring any group. Either would serve the purpose. Society has multiple equilibria. Identity is a focal point. Once everyone prefers one group, it is in each individual’s interest to prefer that group. That was the central message of the chapter. I want to use this occasion to add two caveats, concerning empirical tests and the normative implications of my argument. Some of this I have discussed in Basu (2018, Chap. 5). First, it is important to realize that if discrimination in favor of some group is a coordination device, then in certain kinds of labor markets, there will not be any discrimination. What this implies is that to test whether people have an innate taste for a certain race, gender, or any other group identity, in contrast to the discrimination being a pure coordination device, we need a test more fine-grained than what Bertrand and Mullainathan (2004) and some later authors, such as Siddique (2011), have done. Note that certain kinds of work require little interaction. If I am hiring a sales agent, that person has to interact with others to do my job well. Call such jobs “interactive work.” But if I am hiring an artist to paint my portrait, it does not matter how that person interacts with others. Let us call this kind of work “solipsistic work.” If the discrimination in favor of whites is the outcome of pure racial bias, then when we hire labor for solipsistic work, we should not see any discrimination. In other words, to see how much pure racial bias there is in the labor market in a certain society, what we need to do is a two-pronged test, one in which we send out responses to a call for applicants for an obviously interactive job (sales agent, communications personnel, administrative secretary) and another where we send out applications in response to a call for people to do some solipsistic work (portrait artist, home painter, caretaker for pets). If we find positive and same level of discrimination for both types of jobs, we would have established that there is innate discriminatory preference. But if we find less discrimination in the solipsistic job market, it would show that the focal point of games is at play. My conjecture is that we will find the presence of both reasons for discrimination (innate and as a coordination) in most societies but there will be variations across nations, with some nations and societies exhibiting higher levels of innate racial and gender bias. Much of my analysis is a dry, abstract dissection of a deeply troublesome attribute of many societies, namely, the propensity to discriminate against certain groups, generally minorities.
6
Discrimination as Focal Point
115
Second, let me emphasize that group discrimination is deeply unethical and totally objectionable, no matter what the source. There is a lot written about and covered in the social media about degrading levels of discrimination, even in our contemporary world. I may add here that while there is now some awareness of racial and gender discrimination, some of the worst and most scarring ill treatment is meted out to LGBTQ groups (see Lee Badgett 2020).
Conclusion In closing, what I want to emphasize is that the argument that a part of the discrimination may be the result of an arbitrary focal point, where each individual, including each discriminator, is a victim of the circumstance, must not be treated as a mitigation of the normative problem. What recent events around the world and especially in the United States, and the rise of the Black Lives Matter movement, have made is a reminder that the consequence of discrimination, no matter what the source and whether or not we can lay the blame at anyone’s doorstep, is abhorrent. Further, what my analysis shows is that the problem cannot be solved by leaving it all to the market. The Chicago school view that by freeing the market we would cut down discrimination is wrong. We need affirmative action.
New References Lee Badget MV (2020) The economic case for LGBT equality. Beacon Press. Basu K (2018) The republic of beliefs: a new approach to law and economics. Princeton University Press, Princeton. Deshpande A, Newman K (2007) Where the paths lead: the role of caste in postuniversity employment expectations. Econ Polit Wkly, October 13. Quillian L, Pager D, Hexel O, Midboen A (2017) Meta-analysis of field experiments shows no change in racial discrimination in hiring over time. Proc Natl Acad Sci U S A. Royster D (2003) Race and the invisible hand: how white networks exclude black men from blue-collar jobs. University of California Press, Berkeley. Siddique Z (2011) Evidence of caste-based discrimination. Labour Econ 18(S1): S146–S159.
References Akerlof G, Kranton R (2010) Identity economics: how our identities shape our work, wages, and well-being. Princeton University Press, Princeton Ambady N, Shih M, Kim A, Pittinsky TL (2001) Stereotype susceptibility in children: effects of identity activation on quantitative performance. Psychol Sci 12:371 Arrow KJ (1973) Higher education as a filter. J Public Econ 2:193–216
116
K. Basu
Arrow KJ (1998) What has economics to say about racial discrimination? J Econ Perspect 12: 91–100 Basu A (2016) Why Can’t laziness claim agency? The neo-Malthusian streak in contemporary feminism. Presented at conference on “Malthus: food, land, people”. Cambridge University, Cambridge, UK Basu K (2011) Beyond the invisible hand: groundwork for a new economics. Princeton University Press, Princeton Basu K (2014) Randomization, causality and the role of reasoned intuition. Oxf Dev Stud 42: 455–472 Basu K (2015) Discrimination as a coordination device: markets and the emergence of identity. World Bank Policy Research working paper 7490 Basu K (2016) Globalization of labor markets and the growth prospects of nations. J Policy Model 38(4):656–669 Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. Am Econ Rev 94:991–1013 Coleman JE Jr, Gulati M (2005) Response to professor Sander: is it really all about the grades? N C Law Rev 84:1823–1839 Deshpande A (2010) The grammar of caste. Oxford University Press, Oxford, UK Field E, Nolen P (2005) Race and student achievement in post-apartheid South Africa, mimeo: Harvard University Hamilton D, Darity W Jr, Price AE, Sridharan V, Tippett R (2015) Umbrellas don’t make it rain: why studying and working hard isn’t enough for Black Americans. Mimeo, The New School, New York Hoff K (2015) Behavioral economics and social exclusion: can interventions overcome prejudice?. World Bank Policy Research working paper 7198 Hoff K, Pandey P (2006) Discrimination, social identity, and durable inequalities. Am Econ Rev 96: 206–211 Jayadev A, Reddy SG (2011) Inequalities and identities. SSRN working paper Karabarbounis L, Neiman B (2014) The global decline of the labor share. Q J Econ 129:61–103 Kremer M (1993) The O-ring theory of economic development. Q J Econ 108:551–575 Pager D, Western B, Bonikowski B (2009) Discrimination in a low-wage labor market a field experiment. Am Sociol Rev 74:777–799 Phelps ES (1972) The statistical theory of racism and sexism. Am Econ Rev 62:659–661 Sander RH (2006) The racial paradox of the corporate law firm. N C Law Rev 84:1755–1822 Schelling T (1960) The strategy of conflict. Harvard University Press, Cambridge, MA Sen A (2006) Identity and violence. Alfred Knopf, New York Siddique Z (2008) Caste-based discrimination: evidence and policy. Mimeo, Northwestern University Spence M (1974) Market Signaling: information transfer in hiring and related screening processes. Harvard University Press, Cambridge, MA Stewart F (2005) Horizontal inequalities: a neglected dimension of development. In: WIDER perspectives on global development. Palgrave Macmillan, New York Stiglitz J (1974) Theories of discrimination and economic policy. In: von Furstenberg G (ed) Studies in the economics of discrimination. Lexington Books, London Subramanian S (2011) Inter-group disparities in the distributional analysis of human development: concepts, measurement, and illustrative applications. Rev Black Polit Econ 38:27–52 Swedberg R (1990) Economics and sociology. Princeton University Press, Princeton Thorat S, Banerjee A, Mishra V, Rizvi F (2015) Urban rental housing market: caste and religion matters in access. Econ Polit Wkly (50):47–53 World Bank (2015) World Development Report 2015: mind, society, and behavior. The World Bank, Washington, DC
7
Inequality of Opportunity Patrizio Piraino and Josefina Senese
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Inequality of Opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Affirmative Action in Higher Education Through IOp Lenses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
118 119 122 126 127
Abstract
Various circumstances beyond individual control influence the range of options available to individuals and determine what life objectives one can or cannot pursue. They also impact the translation of individual effort into valued outcomes. The inequality of opportunity (IOp) framework helps clarify the role that circumstances have on socioeconomic disparities. This chapter reviews the central insights of IOp theory. It then discusses how affirmative action can be understood as a natural redress policy from an opportunity egalitarian standpoint. Specifically, the key goal of “leveling the playing field” pursued by affirmative action policies can be seen as a form of reshuffling of opportunities from those who enjoy various types of advantages to those who face structural impediments to express their full potential. Keywords
Inequality of opportunity · Affirmative action
P. Piraino (*) University of Notre Dame, Notre Dame, IN, USA e-mail: [email protected] J. Senese Boston University, Boston, MA, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_33
117
118
P. Piraino and J. Senese
Introduction The theory of inequality of opportunity (IOp) recognizes how individual achievements are due to both the circumstances people find themselves in and the effort they exert. This is a crucial distinction, as outcome disparities may be less objectionable if they largely originate from differences in personal efforts as opposed to factors outside individuals’ control (Brunori et al. 2021; Ferreira and Peragine 2015; Kanbur and Wagstaff 2014; Ramos and Van de gaer 2012; Roemer and Ünveren 2017). The IOp framework thus brings to the fore the idea that, in order to make a statement on the “fairness” of a given level of inequality, it is necessary to look at the nature of its determinants. Observed socioeconomic inequalities need not arise from unequal opportunities. Inequality may derive from variation in preferences, work ethics, or choices individuals make. However, it is often the case that people have similar aspirations and dedicate the same amount of time and effort to pursuing their goals. Yet, some groups have a significantly lower chance of achieving those goals due to various types of group-specific barriers to economic success. It is also possible that among individuals who achieve similar socioeconomic status, some groups had to exert much more effort than others to get there. In this case, looking at an equal distribution of outcomes may not necessarily reflect an equitable one. The fundamental difference here is that outcome equality may conceal that some individuals had to earn certain milestones that more advantaged individuals could instead take for granted. To illustrate this, let us consider the distribution of educational outcomes across a cohort of students. Simply looking at inequality in test scores, or completed grades, may not be as useful to gauge the education system’s fairness as would be investigating the various determinants of educational achievement. Intuitively, if students are free to decide what they want to do with their time and choose how much effort to exert in learning, one could argue that they should be held responsible for their outcomes. On the other hand, if factors beyond students’ control affect the levels of effort exerted and its “translation” into educational achievement, it is more difficult to argue that students should bear full responsibility for their success or failure. For example, some low-income students who desire to perform well in high school may also need to work part-time to support their household. In such a situation, not only would these students invest fewer hours studying, but the time spent learning could also be less productive compared to the time invested by their classmates who do not need to work. In other words, the set of opportunities students face and the careers they can ultimately pursue are partly determined by circumstances over which students have little or no say. While intuitively simple, the IOp approach brings along many theoretical and empirical challenges. For instance, there is no consensus on the set of personal characteristics that can be deemed as beyond the domain of individual responsibility. Also, contributors to the existing literature propose various alternative empirical approaches to disentangle the relative role of effort and circumstances in determining individual outcomes. Some of these challenges are due to the fact that many
7
Inequality of Opportunity
119
determinants of socioeconomic success are not easily observable (e.g., motivation, aspirations, etc.). An additional empirical challenge is that what can be identified as individual effort in the data may itself be influenced by unobservable circumstances. Notwithstanding the ongoing debates in the IOp literature, this chapter argues that it is valuable to examine affirmative action policies through IOp lenses. In particular, the key objective of leveling the playing field can be seen in the IOp framework as a form of redistribution of “life chances” from those who benefit from various inherited advantages to those who face structural impediments to express their potential. The rest of the chapter proceeds as follows: section “Inequality of Opportunity” reviews the central tenets of IOp theory and discusses a number of empirical approaches to operationalize the major distinction between circumstance and efforts. Section “Affirmative Action in Higher Education Through IOp Lenses” outlines how IOp clarifies the rationale for a class of affirmative action interventions. To convey the key arguments, this chapter will use the example of policies aimed at increasing access to higher education institutions. This chapter concludes with some reflections on how IOp may be used to better motivate and design affirmative action interventions and for future academic research.
Inequality of Opportunity The starting point of IOp theory is the idea that socioeconomic hierarchies are not inherently unfair. Rather, our judgment of a given ranking in social status should be influenced by the underlying determinants of individual success and by the extent to which higher positions in the social ladder are contended on equal terms. But what do we understand as equal terms? Let us consider the example of access to selective higher education institutions. The legal requirement of nondiscriminatory practices might be met with an admission mechanism that ensures that all eligible applicants are evaluated exclusively on their prior academic qualifications. On the other hand, equality of opportunity would require that being eligible to apply is not enough and that everyone should have a genuine chance to become “qualified” (Arneson 2015). In this example, having equal opportunity implies correcting the competitive advantage that favorable circumstances grant on some applicants relative to others. Arneson (1989) maintained that “The argument for equal opportunity rather than straight equality is simply that it is morally fitting to hold individuals responsible for the foreseeable consequences of their voluntary choices, and in particular for that portion of these consequences that involves their own achievement of welfare or gain or loss of resources” (p. 88). He further contended that truly equal opportunities occur when governments act to reduce the impact of mere luck on people’s prospects for welfare, including external barriers such as educational opportunities and internal obstacles such as variances in decision-making ability. In this context, an important distinction is that between accountability and responsibility. Individual preferences can be seen as not entirely under people’s
120
P. Piraino and J. Senese
control, as they are influenced by resource availability (Cohen 1989). Roemer (1998) explains that whereas some lifestyle choices are clearly a person’s responsibility, others arise from factors beyond their control, and people should not be held accountable for these choices. The key point is that we should be accountable for our own poor judgment only when, given our circumstances, it was plausible to expect to have behaved better. “Thus, for example, we may decide that an adolescent is responsible for a poor attendance record at school, because she decided to cut classes frequently, but nevertheless not hold her accountable for it, by our taking steps to repair her consequent educational deficit, if her circumstances explained her behaviour” (p. 18). Roemer also notes that people’s decisions regarding effort investments can vary throughout their lifespan. Given that people tend to think and act for themselves to a lesser degree when they are younger, adolescents are more easily influenced by their social background than adults. Consequently, if adolescents and children have their own special effort range, context determines their educational choices to a greater extent (Roemer 1998). That is, one also needs to consider the stage of life of different individuals when determining whether they can be regarded as having exerted the same personal effort. The basic intuition of the IOp theoretical framework can be formalized in its most general terms by expressing a given outcome y as a function of both individual circumstances C and the efforts e invested to achieve higher levels of y:1 y ¼ f ðC, eÞ
ð1Þ
A trivial implication of Equation (1) is that individuals facing the same circumstances and applying the same amount of effort would be able to achieve the same outcome.2 Note that neither the process by which outcomes are derived nor the set of opportunities individuals face are explicitly modeled in this framework. Opportunities will need to be inferred by observing the joint distributions of circumstances, effort, and outcomes. It may be obvious to some readers that this simple conceptualization encompasses, in fact, two distinct and independent postulates: (i) The compensation principle, which proposes eliminating those inequalities in y caused by factors that are beyond individual control (ii) The reward principle, which is concerned with how to reward efforts among individuals who face similar circumstances (Brunori et al. 2021; Fleurbaey and Peragine 2013; Ramos and Van de gaer 2012)
1
The model, with some variants, was introduced by Fleurbaey (1994), Roemer (1993), and Van de Gaer (1993). After these seminal contributions, a rich literature has emerged, with both empirical and theoretical developments; see the recent surveys by Ferreira and Peragine (2016), Ramos and Van De Gaer (2016), and Roemer and Trannoy (2016). 2 Consider two identical twins, who share the same genetic endowment and household inputs, who go to the same school and undertake similar activities. If they devote equal amounts of effort in their schoolwork, they should, in theory, attain equivalent schooling outcomes.
7
Inequality of Opportunity
121
Any acceptable measure of IOp would need to adhere to both the compensation and reward principles. However, when these postulates are given explicit empirical formulations, they tend to collide (Bosmans and Özturk 2021). The literature has proposed two main ways to operationalize the compensation principle, resulting in two prevailing empirical approaches for measuring IOp: the ex ante and the ex post approaches. While both methodologies require several analytical choices over many empirical steps, it is sufficient for the purposes of this chapter to clarify the key distinction. In the ex ante approach, equality of opportunity is achieved when the set of opportunities is the same for all individuals, regardless of their circumstances. This amounts to formulating the compensation principle with respect to individual opportunity sets. It demands equalizing circumstances across individuals ex ante. In the IOp literature, a group of individuals endowed with the same circumstances is referred to as a type. Each type can be seen as an opportunity set for individuals facing the same circumstances. Empirically, the ex ante approach estimates inequality between opportunity sets. This is also referred to as betweentype inequality. On the other hand, in the ex post approach, equality of opportunity is achieved when individuals exerting the same level of effort obtain the same outcome. Here the compensation principle is operationalized by demanding lower inequality among individuals with the same effort but unequal outcomes – i.e., ex post compensation. The IOp literature also presents different treatments of the reward principle. Broadly speaking, different papers suggest different ways to think about outcome inequality among individuals endowed with the same circumstances. At the two extremes, there are contributions proposing complete neutrality – i.e., utilitarian reward (Van de Gaer 1993; Fleurbaey 2008) and those introducing inequalityadverse rewards (Ramos and Van de Gaer 2016). Intermediate and agnostic positions are also present in the literature (e.g., Peragine 2004; Fleurbaey and Peragine 2013). Because of these different approaches, empirical researchers have used a variety of methodologies to analyze the aggregate distributions of outcomes, efforts, and circumstances. Most papers utilize a measure of IOp that reflects the principles of (i) ex ante compensation and (ii) utilitarian reward. This is achieved by estimating some version of the between-type inequality index mentioned above (see Checchi and Peragine 2010 for a nonparametric version). This approach computes the inequality in a counterfactual distribution where all disparities are solely due to variation in observable circumstances. Let us call ÿ the outcome distribution resulting from a scenario where efforts do not play any role in the function f – i.e., 100% opportunity inequality. Once this counterfactual is created, the opportunity set faced by different individuals can be inferred by comparing the observed distribution of y to ÿ (Ferreira and Peragine 2015; Ramos and Van de gaer 2021). The closer y is to ÿ, the higher the degree of inequality of opportunity in the observed outcome distribution (Brunori et al. 2021). In other words, the level of “unfairness” in the distribution is proportional to the effect that circumstance variables (which lie beyond individual responsibility) have on individual outcomes. By defining types as groups of individuals sharing the same circumstances, this literature sometimes denotes the inequality due to personal efforts as within-group inequality and the
122
P. Piraino and J. Senese
inequality due to circumstances as between-group inequality—i.e., inequality of opportunity (see Checchi and Peragine 2010). It is also common in the literature to present relative IOp estimates, defined as the share of unfair inequality over total outcome inequality. For a given inequality index I, the relative IOp measure is then equal to I ðÿÞ=I ðyÞ. In order to estimate this class of IOp indices on the microdata typically available in most countries, it is necessary first to classify which observable covariates can be regarded as circumstances. This is not straightforward, as it requires a judgment on which factors can be attributed to individual responsibility as opposed to conditions beyond personal control (Ramos and Van de gaer 2012; Roemer and Trannoy 2016). Moreover, data constraints generally make it impossible to account for all relevant circumstances, which implies that the within-group component is generally treated as a residual. The literature shows that the incomplete observability of circumstances generates a downward bias in IOp estimates. For this reason, the estimated IOp levels in the literature are generally interpreted as lower bound values of the true degree of inequality of opportunity in a given outcome y (Ferreira and Peragine 2015; Kanbur and Wagstaff 2014; Ramos and Van de gaer 2021).
Affirmative Action in Higher Education Through IOp Lenses A key implication of the IOp framework is that inequality resulting from differences in efforts among individuals endowed with the same set of circumstances is consistent with equal opportunities. In such cases, an appropriate policy intervention would be for governments to adopt various forms of redistribution interventions aimed to reduce the fraction of overall inequality that we deem unfair as measured by the IOp index. Any policy aimed at equalizing opportunities, as defined above, would attempt to provide equal chances of attaining a particular outcome regardless of people’s circumstances. That is, achievements should be made accessible to everyone, and opportunities should not be denied based on factors outside people’s control – e.g., socioeconomic status, ethnicity, gender, and religion (Beckley 2002; Cestau et al. 2017; Kellough 2006; Roemer and Ünveren 2017). Translating this ideal to policy practice is far from straightforward. Eliminating barriers to valuable positions (educational institutions, high-pay jobs, promotions, etc.) by simply allowing everyone to apply and evaluating applicants on merit only would not be sufficient from an IOp point of view. This version of “meritocracy” does not consider history, context, and past and present disadvantages. In fact, this notion can contribute to perpetuating the exclusion of historically marginalized groups. Policy design can thus benefit from IOp-informed approaches that try to guarantee that everyone has a genuine chance to realize a valued outcome (Arneson 2015; Kellough 2006; Kodelja 2016; Moses 2011). From this perspective, affirmative action stands out as a particularly useful policy tool for redistributing opportunities. Affirmative action acknowledges that structural disadvantages (including discriminatory practices) have permeated the distribution
7
Inequality of Opportunity
123
of roles and opportunities among different groups. A wide range of measures coexists under the umbrella of affirmative action policies. The most common distinction is that between policies that involve quotas and those that set goals. Quotas are used to reserve a certain number of positions to target group applicants. Goal-based affirmative action policies use instead a scoring mechanism in selection processes that allocate different points (or set different eligibility thresholds) to different groups based on observable proxies for disadvantage. Regardless of the approach, these sets of policies aim to generate an achievement distribution in which disadvantageous circumstances are partially compensated for and individual efforts appropriately rewarded. If well-designed, such policies should have the overall effect of increasing the fraction of outcome inequality that can be attributed to personal choices relative to inherited disadvantages (Lippert-Rasmussen 2020). A domain where the introduction of affirmative action policies has resulted in highly polarized (and politicized) debates is that of access to higher education. Looking at affirmative action through IOp lenses helps appreciating a key rationale for such policies: redressing unfair advantage. That is, preferential admission policies can be seen as a tool to ensure genuine competition for valued positions. This is achieved by redistributing chances from those who have historically benefited from structural inequalities to those who have faced different types of barriers (Arneson 2015; Critzer and Rai 2000; Crosby et al. 2006; Kellough 2006; Sander and Taylor 2012). This section looks at how the theory of IOp can be used to interpret some common arguments presented by both advocates and skeptics of affirmative action policies in higher education. From an IOp perspective, the starting point is to recognize that the lower academic preparedness of certain groups of students, which decreases their likelihood of accessing higher education, is partly a result of disadvantageous circumstances rather than lack of effort. These students had to face adverse conditions earlier in their lives, which led them to be less “qualified” than their more affluent peers. Conditional on having the appropriate support, these students may be able to catch up with the students who appeared to be better prepared at the moment of admission (Arcidiacono et al. 2015; Cestau et al. 2017; Kellough 2006). A key rationale for preferential admission in higher education is thus the offsetting of this accumulated disadvantage. It is an attempt to equalize access in a context where the tools necessary for achieving admissibility are unfairly distributed across sociodemographic strata. When selective institutions implement merit-based admissions, underprivileged communities are often underrepresented. To correct for this, there are several policies that can be implemented to change the allocation mechanism of scarce academic spots. “Corrective policies” are designed to create educational opportunities that would have otherwise been lost due to various social obstacles, including discrimination (Arcidiacono et al. 2015; Fryer and Loury 2005; Tanner 2016). By pursuing a greater representation of people from underprivileged backgrounds, these policies may redistribute scarce higher education openings from advantaged groups to underprivileged ones (Arneson 2015). That is, preferential admission policies will inevitably create a group of “displaced” students. In popular media debates, this is
124
P. Piraino and J. Senese
sometimes labeled as a form of reverse discrimination. Applicants who are displaced by affirmative action beneficiaries may feel unfairly treated and their academic efforts not adequately rewarded. From an IOp viewpoint, it is clear that these arguments misinterpret the reward principle and fail to recognize the principle of compensation. Affirmative action is intended to equalize opportunities by correcting for advantages that some candidates have gained unfairly. Although the displaced students may not be directly responsible for the structural forces that led to other groups being disadvantaged, they nonetheless benefit from these forces. A useful example to appreciate this point is provided by the rationale for affirmative action policies in post-apartheid South Africa. While white South African students are not directly responsible for the institutional racism of the past, few of them would not acknowledge that the set of educational opportunities they benefit from is partly a legacy of that discriminatory system. The IOp framework clarifies that this is a form of advantageous circumstance that justifies compensation. It is important to note that the reshuffling of opportunity brought by affirmative action policies will only benefit a subset of underprivileged students. If affirmative action fails to acknowledge within-group differences in circumstances, it may only benefit the most privileged segments of the target population. This could have the unintended consequence of rewarding privilege, rather than need, within the beneficiary group. That is, one might argue that affirmative action fails to recognize the disadvantage of those who need compensation the most within the underprivileged groups. What is more, if the marginal admitted applicant of the beneficiary group comes from a better-off background than the marginal displaced student, affirmative action may even become regressive. This is called “mis-targeting” in the literature, with reference interventions that give preference to relatively advantaged applicants who are members of a disadvantaged group at the expense of underprivileged members of an advantaged group. Although this scenario is certainly plausible, proper policy design could help mitigate this unintended consequence (Bertrand et al. 2010). Moreover, the existing literature does not find empirical evidence that suggests significant extents of mis-targeting. For instance, Bertrand et al. (2010) studied quotas for different cast groups and their effects on the composition of engineering colleges’ admissions in India. They find that caste-based targeting redistributes education resources from the most economically advantaged to traditionally underserved groups. Similarly, Bagde et al. (2016) examined student enrollment and academic success at engineering colleges in India. The authors concluded that affirmative action increased attendance from the most disadvantaged castes. Nonetheless, both papers admitted that this policy may have excluded other underserved groups, such as female applicants. Kerr et al. (2017) examined the impact of a race-based preferential admission policy in South Africa. Using applicant and admission data from a top university, they show that race was effective as a proxy for economic disadvantage and that beneficiary students had significantly lower socioeconomic status than the displaced white students.
7
Inequality of Opportunity
125
A criticism sometimes raised in discussions about affirmative action in higher education is that preferential admission may stigmatize beneficiaries. Students admitted under preferential rules may be judged harsher as a result of their group identity. Their accomplishments may be undervalued or dismissed and their flaws highlighted. For instance, Deshpande (2019) investigated the stigma associated with caste-based affirmative action among Indian students. While the author found no significant differences in effort or academic attitudes between recipients and non-recipients, upper-caste students judged the performance of affirmative action beneficiaries negatively, indicating the prevalence of discriminatory attitudes toward them. While this type of risk speaks more to the prevalence of discriminatory mindsets rather than a particular policy flaw, it is plausible that an unwelcoming environment for affirmative action beneficiaries may decrease the policy effectiveness (Bertrand et al. 2010; Fischer and Massey 2007; Kerr et al. 2017). It is also possible that beneficiaries find themselves in an environment for which they are unprepared, making them feel isolated and discouraged. Minority or disadvantaged students would be admitted but be more likely to fail or drop out. Thernstrom and Thernstrom (1997) write that “when students are given a preference in admission because of their race or some other extraneous characteristic, it means that they are jumping into a competition for which their academic achievements do not qualify them and many find it hard to keep up” (pp. 405–406). For example, elite academic institutions often teach at a faster pace and assume that students are familiar with a range of topics. Affirmative action beneficiaries may find the coursework excessively challenging and perform poorly in their classes. Relative to peers with similar characteristics but who attend less selective institutions, students enrolled in selective colleges may have a lower chance of graduating. This set of arguments is referred to as the mismatch hypothesis: affirmative action may generate a mismatch between students’ precollegiate skills and institutions according to this skill distribution (Alon and Tienda 2005; Arcidiacono and Lovenheim 2016; Sander and Taylor 2012). Sander and Taylor (2012) emphasize that mismatch may have a cascade effect: affirmative action students who barely pass the freshman courses can be at an even greater disadvantage in subsequent classes, widening the achievement gap with their more advantaged peers. That is, students’ ability to persist in challenging fields/majors is harmed by their initial low credentials. Likewise, Frisancho and Krishna (2016) analyzed the trajectory of a graduating class from a prestigious engineering college in India. Compared to their same-major peers, affirmative action students did not catch up: they fell behind, and performance disparities were greater for more selective majors. While theoretically plausible and empirically true in certain contexts, the mismatch problem can be partly addressed with design adaptations such as extra support interventions both before and after admission. The support may range from targeted bootcamps or other intensive catchup courses to the availability of dedicated wellness coaches and academic mentors. These types of interventions would again be consistent with the compensation principle from an IOp perspective and would help ensure that students can succeed in high-performance contexts when given a “fair chance” – i.e., the right set of tools.
126
P. Piraino and J. Senese
A different type of concern with respect to the effectiveness of affirmative action in higher education relates to distortionary incentive effects. By making hard work and talent less relevant for achieving valued positions, affirmative action would lessen incentives for the beneficiaries to exert more effort. If target groups are rewarded for factors unrelated to their effort, the policy would violate a key principle of IOp theory – i.e., an achievement distribution where individual efforts are appropriately rewarded. However, the opposite can be asserted as well: minority students may respond by exerting more effort by virtue of perceiving a higher chance to be admitted in higher education. An exogenous increase in the probability of admission for a group of historically disenfranchised students increases the expected returns to the effort exerted. In other words, affirmative action creates a situation in which previously unattainable opportunities are regarded as accessible and hence worth the effort (Arcidiacono and Lovenheim 2016; Bok and Bowen 1998; Durlauf 2008; Fryer and Loury 2005; Tanner 2016). Such a scenario would clearly be consistent with the reward principle.
Concluding Remarks Inequality of opportunity postulates that disparities in socioeconomic outcomes are not necessarily unfair. Individuals’ accomplishments should be rewarded when they derive from differential degrees of personal efforts. However, hard work and motivation are not the only factors determining inequality. Various types of circumstances on which individuals have no control affect the range of options and aspirations people can pursue and impact the extent to which a personal effort translates into desirable outcomes. In other words, IOp theory acknowledges that social hierarchies are not intrinsically unjust, but the process by which status is achieved defines the fairness of a certain social ranking. Inequality is more objectionable when it results mostly from variation in factors outside individual control. Since the seminal contributions in the early 1990s, the literature on inequality of opportunity has evolved rapidly. While different theoretical approaches have been presented, most contributions agree that the normative goal is not outcome equality. The objective is an achievement distribution where efforts are appropriately rewarded and unfavorable circumstances properly compensated for. Empirically, this would translate into an outcome distribution where all observable inequalities can only be attributed to personal choices (Ramos and Van de gaer 2012). This basic intuition of IOp theory lends itself quite naturally to the domain of affirmative action policies. Indeed, affirmative action recognizes that discrimination and other exclusionary practices influence how opportunities are distributed among different social groups. Building on this awareness, it attempts to reshuffle the distribution of “opportunity sets” to compensate for unfair advantages and to reward true merit. From an IOp perspective, the goal of affirmative action policies is to ensure genuine competition when the prospects of reaching a valued goal are unequally distributed across social groups.
7
Inequality of Opportunity
127
Focusing on the case of preferential admission to universities, this chapter showed that while affirmative action can reduce the impact of circumstance inequities, it does not represent a silver bullet. The potential for distortionary behavioral responses as well as inadequate institutional support for beneficiaries may significantly limit the effectiveness of affirmative action in higher education admissions. Careful, theory-driven and context-conscious policy design is necessary to avoid unintended negative effects on both beneficiaries and non-beneficiaries. This chapter also argued that preferential admission alone may not be sufficient to equalize opportunities, and additional support mechanisms for beneficiaries may be required depending on the context. From an IOp perspective, affirmative action can be seen as a useful tool for decreasing inequality of opportunity. It should be part of a comprehensive system of corrective measures to improve outcomes for disadvantaged groups. As Barros Francis (1968) said more than half a century ago, “It is clear that the concept of equal [. . .] opportunity, especially in higher education, is essentially meaningless if it is removed from the matrix of the social situation” (p. 312). Yet, despite its flaws, affirmative has opened and continues to open doors. In a global context where exclusion and inequality of opportunity are still prevalent, affirmative action is a valuable step to counterbalance persistent group inequality (Arneson 2015; Moses 2011).
References Alon S, Tienda M (2005) Assessing the “Mismatch” hypothesis: differences in college graduation rates by institutional selectivity. Sociol Educ 78:294–315. https://doi.org/10.1177/ 003804070507800402 Arcidiacono P, Lovenheim MF (2016) Affirmative action and the quality-fit trade-off. J Econ Lit 54: 3–51. https://doi.org/10.1257/jel.54.1.3 Arcidiacono P, Lovenheim MF, Zhu M (2015) Affirmative action in undergraduate education. Ann Rev Econ 7:487–518. https://doi.org/10.1146/annurev-economics-080614-115445 Arneson R (1989) Equality and equal opportunity for welfare. Philos Stud 56:77–93. https://doi. org/10.1007/BF00646210 Arneson R (2015) Equality of opportunity. In Zalta EN (ed) The Stanford encyclopedia of philosophy (Summer 2015 ed.). Stanford University https://plato.stanford.edu/archives/ sum2015/entries/equal-opportunity/ Bagde S, Epple D, Taylor L (2016) Does affirmative action work? Caste Gender College Quality Acad Success India 106:1495–1521. https://doi.org/10.1257/aer.20140783 Barros FJ (1968) Equal opportunity in Higher Education. J Negro Educ 37:310–315. https://doi.org/ 10.2307/2294165 Beckley H (2002) Capability as opportunity: how Amartya Sen revises equal opportunity. J Religious Ethics 30:107–135. https://doi.org/10.1111/1467-9795.00100 Bertrand M, Hanna R, Mullainathan S (2010) Affirmative action in education: evidence from Engineering College Admissions in India. J Public Econ 94:16–29. https://doi.org/10.1016/j. jpubeco.2009.11.003 Bok D, Bowen WG (1998) The shape of the river: long-term consequences of considering race in College and University Admissions. Princeton University Press, Princeton Bosmans K, Özturk ZE (2021) Measurement of inequality of opportunity: a normative approach. J Econ Inequal 19:213–237. https://doi.org/10.1007/s10888-020-09468-1
128
P. Piraino and J. Senese
Brunori P, Ferreira F H G, Peragine, V, Piraino P, Triventi M, Van der Weide R, Bloise F, Gupta R, Gasparini L, Lakner C, Luppi F, Mahler D, Narayan A, Neidhöfer G, Palmisano F, Randazzo T, Rampino T, Serlenga L, Serrano J (2021) Equal chances: equality and intergenerational mobility around the world. Unpublished manuscript. Cestau D, Epple DN, Sieg H (2017) Admitting students to selective education programs: merit, profiling, and affirmative action. J Polit Econ 125:761–797. https://doi.org/10.1086/691702 Checchi D, Peragine V (2010) Inequality of opportunity in Italy. J Econ Inequal 8:429–450. https:// doi.org/10.1007/s10888-009-9118-3 Cohen GA (1989) On the currency of Egalitarian justice. Ethics 99:906–944. https://doi.org/10. 1086/293126 Critzer JW, Rai KB (2000) Affirmative action and the university race, ethnicity, and gender in higher education employment. University of Nebraska Press Lincoln, Nebraska, USA Crosby FJ, Iyer A, Sincharoen S (2006) Understanding affirmative action. Annu Rev Psychol 57: 585–611. https://doi.org/10.1146/annurev.psych.57.102904.190029 Deshpande A (2019) Double Jeopardy? Stigma of identity and affirmative action. Rev Black Polit Econ 46:38–64. https://doi.org/10.1177/0034644619837211 Durlauf SN (2008) Affirmative action, meritocracy, and efficiency. Polit Philos Econ 7:131–158. https://doi.org/10.1177/1470594X08088726 Ferreira FHG, Peragine V (2015) Equality of opportunity: theory and evidence. The World Bank, Washington, DC Ferreira FHG, Peragine V (2016) Individual responsibility and equality of opportunity. In: Adler MD, Fleurbaey M (eds) The Oxford Handbook of Well-Being and Public Policy, Oxford University Press, Oxford Fischer MJ, Massey DS (2007) The effects of affirmative action in higher education. Soc Sci Res 36:531–549. https://doi.org/10.1016/j.ssresearch.2006.04.004 Fleurbaey M (1994) On fair compensation. Theory and Decision 36:277–307. https://doi.org/10. 1007/BF01079932 Fleurbaey M, Peragine V (2013) Ex Ante versus Ex post equality of opportunity. Economica 80: 118–130. https://doi.org/10.1111/j.1468-0335.2012.00941.x Frisancho V, Krishna K (2016) Affirmative action in higher education in India: targeting, catch up, and mismatch 71:611–649. https://doi.org/10.1007/s10734-015-9927-1 Fryer RG, Loury GC (2005) Affirmative action and its mythology. J Econ Perspect 19:147–162. https://doi.org/10.1257/089533005774357888 Kanbur SMR, Wagstaff A (2014) How useful is inequality of opportunity as a policy construct? The World Bank, Washington, DC Kellough JE (2006) Understanding affirmative action: politics, discrimination, and the search for justice. Georgetown University Press, Washington, DC Kerr A, Piraino P, Ranchhod V (2017) Estimating the size and impact of affirmative action in undergraduate admissions at the University of Cape Town. South African J Econ 85:515–532. https://doi.org/10.1111/saje.12174 Kodelja Z (2016) Equality of opportunity and equality of outcome. CEPS J 6:9–24 Lippert-Rasmussen K (2020) Making sense of affirmative action. Oxford University Press, New York Moses MS (2011) Race, affirmative action, and equality of educational opportunity in a so-called “Post-racial” America. Kansas J Law Public Policy 20:413–427 Peragine V (2004) Ranking income distributions according to equality of opportunity. 2:11–30. https://doi.org/10.1023/B:JOEI.0000028404.17138.1e Ramos X, Van de gaer D (2012) Empirical approaches to inequality of opportunity: principles, measures and evidence. J Econ Surv 30:855–883. https://doi.org/10.1111/joes.12121 Ramos X, Van de gaer D (2016) Approaches to inequality of opportunity: Principles, Measures and Evidence. 30:855–883. https://doi.org/10.1111/joes.12121 Ramos X, Van de gaer D (2021) Is inequality of opportunity robust to the measurement approach? Rev Income Wealth 67:18–36. https://doi.org/10.1111/roiw.12448
7
Inequality of Opportunity
129
Roemer JE (1993) A pragmatic theory of responsibility for the Egalitarian Planner. Philosophy & Public Affairs 22:146–166 Roemer JE (1998) Equality of opportunity. Harvard University Press, Cambridge Roemer JE, Trannoy A (2016) Equality of opportunity: theory and measurement. J Econ Lit 54: 1288–1332. https://doi.org/10.1257/jel.20151206 Roemer JE, Ünveren B (2017) Dynamic equality of opportunity. Economica 84:322–343. https:// doi.org/10.1111/ecca.12197 Sander RH, Taylor S Jr (2012) Mismatch: how affirmative action hurts students it’s intended to help, and why universities won’t admit it. Basic Books, New York Tanner J (2016) Towards lifting the burden of stereotype: affirmative action and equality of opportunity. Ethos 29:152–172 Thernstrom AM, Thernstrom S (1997) America in Black and White: One nation, indivisible. Simon & Schuster, New York Van de Gaer D (1993) Equality of opportunity and investments in human capital. Ph.D. Thesis, KULeuven
Part II Methodological Approaches to Understanding Discrimination
8
Decompositions: Accounting for Discrimination Gurleen Popli
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Decomposing the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oaxaca-Blinder Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Issues with the Decompositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment Effect Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Formal Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Going Beyond the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Residual Imputations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reweighting Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditional Quantiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recentered Influence Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
134 135 135 137 139 141 143 143 144 146 147 148 149
Abstract
This chapter summarizes the different regression-based decomposition methods used in the empirical literature to evaluate discrimination. Starting with the decomposition at the mean using the methods made popular in the 1970s by Oaxaca and Blinder, we discuss how the method has evolved over time to look beyond the means, taking into account the entire distribution of the outcomes of interest. We present the formal identification assumptions underlying the decomposition method and discuss cautions that should be exercised in interpreting them and their limitations. We also explain how the “unexplained gap” in the decomposition, often used as a measure of discrimination, relates to the treatment effect literature.
G. Popli (*) Department of Economics, University of Sheffield, Sheffield, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_15
133
134
G. Popli
Keywords
Decomposition · Counterfactual regressions · Distributions · Discrimination JEL Codes
J31 · J71 · C21
Introduction The fact that discrimination exists is not a doubt; it exists and impacts every aspect of our lives, including, but not limited to, economic outcomes (in educational opportunities and outcomes in labor and credit markets), social outcomes (network formations and residential location), health outcomes (mortality rates and access to health services), and interactions with the criminal justice system (Arrow 1998; Lang and Kahn-Lang Spitzer 2020; Small and Pager 2020). Within the discipline of economics, much of the theoretical discussion around discrimination started with the work of Becker (1957), while much of the empirical work has come from within the field of labor economics with the seminal works of Oaxaca (1973) and Blinder (1973), who used wage regressions to quantitatively assess the role of discrimination in explaining the observed wage gaps between men and women (in case of Oaxaca) and between whites and blacks (in case of Blinder, who also looked at the gender wage gaps). Since Oaxaca and Blinder’s work, the literature on regression-based decompositions has seen substantial growth in methodological advancements and their empirical applications. This chapter gives a summary of the different regression-based decomposition methods. Starting with Oaxaca and Blinder’s work, we discuss how the method has evolved over time. Two key limitations of the regression-based decompositions should be stated at the very beginning. First, all the methods discussed here follow the partial equilibrium approach. The wage gaps are decomposed into a part explained by the differences in endowments of the two groups, holding the returns to these endowments constant, and the differences in the returns to endowments across the two groups, holding their endowments constant. This assumes that we can change the endowments without impacting returns to them and vice versa. This is a strong assumption and often not valid. Second, regression-based decomposition methods are an accounting exercise. While we can know various factors contributing to the existing difference between wages (or any outcome of interest), decompositions do not tell us the underlying mechanisms. They can, however, help us confirm an existing hypothesis or form new ones. When we do regression-based decompositions, we are not trying to detect discrimination but are attempting to quantify it. While we have to be cautious in our interpretations that not all the observed wage gaps are discrimination, nothing stops us from concluding that discrimination exists and is buried in it. The way we use regression-based decompositions and the way we interpret them are critical. If done
8
Decompositions: Accounting for Discrimination
135
correctly and used correctly, we can give some accounting of the existing discriminations. Section “Decomposing the Mean” of the chapter discusses the most famous decomposition method – the Oaxaca-Blinder (OB, henceforth) method. This section will also highlight issues around interpretations, including the treatment effect interpretation, and formal assumptions for identification. Section “Going Beyond the Mean” will discuss the different decomposition methods that have been proposed in the literature since the OB work. Section “Conclusion” provides concluding comments.
Decomposing the Mean Oaxaca-Blinder Method What we refer to as the OB decomposition in economics was first employed in demography by Kitagawa (1955) and made popular in sociology by Althauser and Wigler (1972). This method has been used extensively in the empirical economics literature to study mean wage gaps between different groups in the labor market, defined over gender, race and ethnicity, age, disability, and over immigration status, among others. The decomposition has also been used to look at outcomes beyond the labor market, like consumption and expenditure differences, and inequalities over time and space (i.e., across different regions). Below we lay out the basics of this decomposition. Let Ygi be any outcome of interest, for individual i (i ¼ 1, . . ., n) belonging to group g (M, W). For ease of exposition, let the two groups be men (M) and women (W ) and the outcome of interest be wages. Let Χ gi ¼ (Xgi1, . . ., XgiK) be a vector of K covariates which are associated with the outcome of interest. We assume that the outcome of interest is continuous and linearly related to the covariates as: Y gi ¼ βg0 þ ΣKk¼1 Xgik βgk þ υgi
ð1Þ
where βg0 and βgk are parameters to be estimated, and υgi is the error term which is conditionally independent of Χ gi, such that E(υgi|Χ gi) ¼ 0. The difference in the μ means of the outcomes between the two groups, ΔO ¼ Y M Y W , where Y g is the mean outcome for group g, is given as: μ
ΔO ¼ βM0 βW0 þ μ
k
XWk βMk βWk
ΔS ðunexplainedÞ
þ
k
XMk XWk βMk
ð2Þ
μ
ΔX ðexplainedÞ
where Xgk is the mean of covariate Xgik for group g, and βg0 and βgk are the estimated parameters, intercept and slope coefficients, from the regression models estimated separately for the two groups. The mean difference is decomposed into two parts.
136
G. Popli μ
The first term ΔS is the unexplained component, if the outcome of interest is wages; μ this is also referred to as the wage structure effect. The second term ΔX is the explained component; this is also referred to as the composition effect. The composition (explained) effect is the difference in wages due to differences in the observed covariates (endowments) of the individuals across the two groups. For example, the composition effect is the part of the mean gender wage gap that is explained due to observable differences in mean characteristics, XMk XWk , among the men and women, like education and labor market experience. The wage structure (unexplained) effect is the difference in mean wages due to the difference in returns to individual characteristics, βMk βWk . The difference in the intercepts, βM0 βW0, is interpreted as the part of the unexplained gap attributed to group membership. The unexplained component is the gender wage gap that is often associated with labor market discrimination, omitted variables, and unobserved heterogeneity. Using the expression given in Eq. (2), we can investigate both the aggregate and detailed decompositions. Aggregate decomposition is where we are only interested μ μ in the overall wage structure and composition effects, that is, ΔX and ΔS, which helps us understand how much of the observed gap is due to differences in endowments (composition effect) and how much of the gap is due to differences in returns to those endowments (wage structure effect). If we are interested only in aggregate decompositions, we do not even need to separately estimate the wage regressions for μ women. The unexplained component can be simply computed as, ΔS ¼ μ ΔO k XMk XWk βMk . There is, however, often an interest in detailed decomposition, where we wish to know the contribution of each individual covariate, XgiK, to both the wage structure and composition effect. For example, to understand the gender wage gap, it can be of interest to know how much of the composition gap is due to differences in education between men and women and how much is due to differences in the labor market experience between the two groups, as the policy implications from the two can be different. Similarly, we might also be interested in differences in returns to education versus differences in returns to labor market experience when looking at the wage structure effect. The OB decomposition has a counterfactual interpretation, which has been exploited in later methodological innovations. Consider, Y Ci ¼ βM0 þ ΣKk¼1 XWik βMk
ð3Þ
where Y Ci is the counterfactual wage for women obtained by using their own characteristics, but the estimated parameters are from regression for men. The counterfactual wages tell us what women’s wages would have been if they had their own characteristics, but their characteristics were rewarded as men’s are. Using the notion of counterfactual wages, we can write the decomposition in Eq. (2) as:
8
Decompositions: Accounting for Discrimination
YM YW ¼ YM Y
C
137 C
þ Y YW
μ
μ
ΔX
ΔO
ð4Þ
μ
ΔS
C
where Y is the mean counterfactual wage distribution. The first term of Eq. (4) is the explained component and the second term is the unexplained component. The unexplained component is the difference in wages that women would have received if they had their own characteristics but were treated as men are in the labor market and the actual wages of women.
Issues with the Decompositions This section discusses three main cautions that should be exercised or issues that should be kept in mind when interpreting the findings from regression-based decompositions. These cautions apply to the OB decomposition of the mean and all the extensions and methods proposed since then, including those discussed in section “Going Beyond the Mean.” First is the issue of the reference group. To understand this, let’s consider the counterfactual wages generated for women, given by Eq. (3). In this format, it is clear that men are assumed to be the reference group, and male returns to characteristics (the estimated coefficients) are assumed to be nondiscriminatory, that is, we assume that these are the returns that would prevail in the market for both men and women in the absence of discrimination. This is not an innocuous assumption; choice of reference group can alter how the observed wage gap is attributed to the effect of wage structure and composition. The choice of reference group has been discussed in the literature, with various alternatives being proposed. Cotton (1988) gives an excellent graphical illustration of the inherent assumption made when choosing one group as a reference relative to the other group and proposes using a weighted average of the estimated coefficients for the two groups as the nondiscriminatory coefficients. Neumark (1988), on the other hand, proposed using estimated coefficients from a pooled regression for the two groups. Jones and Kelley (1984) propose what is known as the three-fold decomposition: μ
ΔO ¼ βM0 βW0 þ
k
XWk βMk βWk
wage structure effect
þ
k
XMk XWk
βMk βWk
þ
k
XMk XWk βMk
composition effect
ð5Þ
interaction term
The first term, in Eq. (5), is the pure wage structure effect, as before this reflects how much of the gender wage gap results from differences in how women’s characteristics are actually valued in the market and how they would be valued if
138
G. Popli
they had the same rates of return as of men. The second term is the pure composition effect, as before, this reflects how much more women would earn if they had the same characteristics as men, but nothing else has changed. The third term is the interaction term; this is the amount that women would gain if they had the characteristics as men and if these characteristics had returns similar to men. In the empirical analysis, this term is often small and is hard to interpret when we have multiple covariates. For a further discussion of the alternative measures proposed in the literature and a unified framework to compare them, see Oaxaca and Ransom (1994). The second issue concerns the base group or the “omitted group” problem. The omitted group problem is discussed in detail by Oaxaca and Ransom (1999) and Oaxaca (2007). This is a concern if we want to do a detailed decomposition of the wage structure (unexplained) component and have categorical covariates. If we are interested only in aggregate decompositions, then this is not a concern. To illustrate the problem, let’s assume the only covariates we have are sectors of employment: primary, manufacturing, and service. In the regression framework, we include two dummies for two sectors, and one sector is omitted as the base category. Say we arbitrarily set the primary sector (sec1) as the omitted category and include dummies for manufacturing (sec2) and service (sec3) sector in the regression equation. This gives us the wage structure effect: μ
ΔS ¼ β0 M0 β0 W0 þ
X k¼2,3 Wseck
βMseck βWseck
ð6Þ
In the presence of categorical covariates, the difference in the intercepts of the two wage regressions for the two groups can be written as: β0 M0 β0 W0 ¼ βM0 þ βMsec1 βW0 þ βWsec1 ¼ βM0 βW0 þ βMsec1 βWsec1
ð7Þ
When we have categorical covariates difference in intercepts, β0 M0 β0 W0 includes not only the gap attributed to the group membership, βM0 βW0 , but also the gap attributed to belonging to the omitted sector, βMsec1 βWsec1 . The latter will change depending on the sector that is omitted. The omitted group’s issue arises only in the wage structure effect and does not impact the composition effect; neither does this problem arise if there is only one binary dummy variable as a covariate (Oaxaca 2007). However, the problem gets complicated if we have more than one categorical covariate. Further, as Jones (1983) pointed out, the base group problem also exists in continuous covariates that do not have a natural scale, like test scores. Some solutions to the base group problem have been proposed in the literature, see Gardeazabal and Ugidos (2004) and Yun (2005), which involve the normalization of coefficients of the categorical variables. However, there is no agreement on this, these normalizations tend to be sample specific,
8
Decompositions: Accounting for Discrimination
139
and different researchers can use different sets of normalizations. Base groups should be chosen based on context and economic meaning. The third issue is of self-selection within groups and between groups. The most common example of self-selection within groups is of differential selection into the labor force by gender. Self-selection between groups arises when there is an element of choice over group membership, for example, union versus nonunion workers. In the presence of selection (whether between groups or within groups), the estimated coefficients of the wage regression are biased. There are unobserved variables that are correlated with both selection, whether it is the choice of participation in the labor force or the choice of joining unions, and wages. A solution to this problem is to estimate a selection-corrected wage regression, similar to that proposed by Heckman 1979. The selection-corrected wage regression then yields the unbiased estimates of the wage regressions and allows for subsequent decomposition. Let the selection-corrected wage regression be given as: Y gi ¼ βg0 þ ΣKk¼1 Xgik βgk þ σ g λgi þ υgi
ð8Þ
where λgi is the control variable to correct for selection, if the Heckman selection model is used, this will be the inverse mills ratio estimated from the first step of the selection model; and σ g is the estimated coefficient for the control variable. This now changes the decomposition, as the selection-corrected wage regressions have an additional term; the decomposition is now given as: μ
ΔO ¼ βM0 βW0 þ
k
XFk βMk βWk
μ
ΔS ðunexplainedÞ
þ
k
XMk XWk βMk μ
ΔX ðexplainedÞ
þ λM σ M λW σ W where the first two terms are as before and the last term λM σ M λW σ W
ð9Þ is the
difference in the wage gap attributed to differences in selection bias. How the selection term (control variable and the parameter associated with it) is attributed to the different parts of the wage decomposition has implications for interpreting discrimination; for a discussion see Neuman and Oaxaca (2004).
Treatment Effect Interpretation Fortin et al. (2011) show that the wage structure effect estimated from regressionbased decomposition has parallels with the treatment effect literature. To explain this, let us consider union workers (U ) and nonunion workers (N ). The difference in the average wages if everyone is paid according to the wage structure of union members and the average wages of everybody if they were paid according to the wage structure of nonunion members can be conceived as the average treatment
140
G. Popli
effect (ATE) where the treatment is “union” membership, that is, switching workers from being nonunion members to union members. Assuming nonunion workers as the reference group, we can construct counterfactual wages for union workers, these are the wages that the union workers with characteristics XUik would earn if they were paid as nonunion workers: C
Y ¼ βN0 þ
k
XUk βNk
ð10Þ
Using this counterfactual wage, the decomposition can be written as: μ
ΔO ¼ Y U Y N ¼ Y U Y μ
ΔS
C
C
þ Y YN
ð11Þ
μ
ΔX
where the difference between the average wages of the union workers and their C counterfactual wages, Y U Y , is the wage structure effect. Under treatment effect interpretation, Y U , the average wages of the union workers, is the average wages of C the treatment group where the treatment is “union” membership; and Y is then the average wages of the treated (union) workers if they were not treated, that is, they C were nonunionized. Difference between the two, Y U Y , is the “union effect” or the average treatment effect on the treated (ATT). In this framework, the composition μ effect, ΔX , is referred to as the selection bias. The ATT interpretation of the wage structure holds for aggregate decompositions, including extensions that go beyond the mean, under the formal identification assumptions discussed below; however, this equivalence does not always hold for detailed decompositions. The issues of choice of the reference group and the base group problem remain in this interpretation as well; however, this approach gives us a way around the selection issue. For a detailed discussion of equivalence between the treatment effect interpretation for the OB decompositions, see An and Glynn (2019). For the OB decomposition, we had assumed the error term to be conditionally independent of covariates, E(υgi|Χ gi) ¼ 0. Under the treatment effect interpretation, this assumption can be replaced by a weaker assumption of “ignorability.” This helps us address some issues of selection bias. Under the assumption of ignorability, selection bias is allowed as long as it is the same for the two groups. For example, ability, which we do not observe, can be correlated with education, an observed covariate, as long as the correlation is the same in the two groups being considered. This is the “selection on observables” assumption used in the treatment literature. Under this interpretation, we do not need to calculate the composition (explained) component, once we have the unexplained component, we can compute the μ μ μ explained part as: ΔX ¼ ΔO ΔS , which reflects the difference in the distribution of the covariates and the error term. In the treatment effect literature, ATT has a causal interpretation. Even though the wage structure effect is derived under the same conditions as ATT, the causal
8
Decompositions: Accounting for Discrimination
141
interpretation often cannot be extended to the wage structure effect for two main reasons. First, in decompositions, “treatment” is not always a choice or something that can be manipulated easily. For example, while we can conceive union/nonunion membership as a choice, gender cannot be conceived as a choice or something that can be manipulated. Further, as we discussed in the selection issue, group membership is often related to unobserved variables, this means the ignorability assumption is often violated, making the causal interpretation hard. Second, as mentioned in the introduction, decompositions are a partial equilibrium approach. For example, the wage structure effects for union wage gap tells us, holding union workers characteristics (Χ Ui) same, if the returns to their characteristics were the same as that of nonunion workers (i.e., βUk ¼ βNk), then the wage structure effect would be zero. Equating the two parameters is similar to “treating” union workers as nonunion workers, but this treatment is likely to change union workers’ characteristics. When Χ gi is impacted by the treatment, we cannot obtain a causal interpretation. Unless great caution has been taken to include only those characteristics that are unlikely to be impacted by the treatment, for an example of this see Neal and Johnson (1996).
Formal Identification In this section, we lay out the minimum set of assumptions needed to identify the aggregate decomposition. There are further assumptions needed to identify detailed decompositions and some of the extensions discussed in section “Going Beyond the Mean.” For a more technical account of these assumptions, refer to Fortin et al. (2011), who set out the full set of assumptions needed for the identification for all regression-based decomposition methods. Assumption 1: Mutually exclusive groups The population of interest can be divided into two mutually exclusive groups, g (A, B). We are interested in comparing the outcome Ygi of the two groups. In line with the treatment effect literature, YAi and YBi can be interpreted as potential outcomes for individual i. If the individual is in group A (B), then we only observe YAi (YBi). Assumption 2: Structural form A worker i belonging to group g is paid according to the wage structure mg, which is a function of worker’s observable characteristics, Xgi, and unobservable characteristics, υgi: Y Ai ¼ mA ðXAi , υAi Þ and Y Bi ¼ mB ðXBi , υBi Þ
ð12Þ
In a more general framework, we can think of mg as the function linking individual characteristics and their outcomes. Under linearity: Ygi ¼ mg(Xgi, υgi) ¼ Χgiβg þ υgi. There are three reasons why wages can be different between groups. (1) The difference in the wage setting equations mA and mB. For the linear model, this
142
G. Popli
is the difference in the returns to the observed characteristics, βA and βB, or the difference in the returns to the unobservable characteristics. (2) Differences in the distribution of the observable characteristics, Xgi, for the two groups. (3) Differences in the distribution of the unobservable characteristics, υgi, for the two groups. Assumption 3: Simple counterfactual treatment A counterfactual wage structure, mC, is said to correspond to a simple treatment when it can be assumed that mC(XBi, υBi) mA(XBi, υBi), or mC(XAi, υAi) mB(XAi, υAi). This assumption states that we can identify what the distribution of wages for groups A will be if the returns to their characteristics are similar to those of group B. This is the counterfactual we constructed for women, given by Eq. (3), and the counterfactual we constructed for union workers, given by Eq. (10). This assumption, however, also highlights that while we can identify what the distribution of wages for women (union workers) will be if they were paid as men (nonunion workers) are, we cannot identify what the distribution of wages for women (union workers) will be if there were no labor market discrimination (no unions). Assumption 4: Overlapping support No single value of observable, Xgi, or unobservable, υgi, characteristics can identify group membership. This rules out a situation where the covariates in Xgi are different across groups, or there are values that Xgi can take only for one group and not for the other. The issue of overlapping support, or common support as it is also referred to, is well recognized in the treatment effect literature. This is usually not a problem if our focus is only on the decomposition of the mean but becomes an issue when looking at distributions. An example of different covariates across groups is when let’s say we want to compare wages of immigrants and natives, where the country of origin is an essential predictor of wages for the immigrants but is not relevant for natives. Similarly, when we look at gender wage gaps, often women tend to be concentrated into occupations and or industry combinations where there will be no men. For a discussion of this and possible solutions, see Ñopo (2008). It is also possible that the range of values that certain covariates take differ by groups. This was discussed by Barsky et al. (2002), where the authors look at the role of income in explaining the black-white wealth gaps, and found that there are certain regions of income where no blacks are observed. Assumption 5: Conditional independence or ignorability Let Dgi ¼ 1{i g}, where 1{.} is the indicator function for group membership. Ignorability requires Dgi ⊥ υ j X. Intuitively, the ignorability assumption states that the distribution of unobservable characteristics, υ, given observable characteristics, X, be the same for the two groups. In case of the decomposition of the mean and linear specification, it does not require E(υgi|Χgi) ¼ 0, instead all it requires is E(υAi|ΧAi) ¼ E(υBi|ΧBi). In the treatment effects literature, this assumption is also referred to as unconfoundedness or selection on observables.
8
Decompositions: Accounting for Discrimination
143
Going Beyond the Mean There have been several extensions of the OB decomposition. In this section, we discuss some of these extensions. First, we discuss the extension proposed by Juhn et al. (1993), who give a way to account for the unobserved characteristics via a residual imputation method. Second is the reweighting method, proposed by DiNardo et al. (1996) to investigate the distribution of the unexplained differences. The third is the decomposition of the conditional distributions, based on quantile regressions, by Machado and Mata (2005). Fourth, and finally, is the decomposition of unconditional quantiles, using Recentered Influence Functions, proposed by Firpo et al. (2007, 2009). While most of the methods discussed in this chapter allow us to look beyond the mean and are good at obtaining aggregate decompositions, detailed decompositions beyond the mean remain challenging, and where possible, they come at the cost of simplicity and intuitive appeal of the OB method.
Residual Imputations The main contribution of Juhn et al. (1993), JMP henceforth, was to take into account the returns to the unobservable characteristics (also referred to as unobserved heterogeneity) explicitly in the decompositions. Starting with the regression function given by Eq. (1), when we decompose the mean gap between two groups we get Eq. (2). Residuals, υig, are not a part of the decomposition as υg ¼ 0, where υg is the mean residual of group g. However, the residuals υig are interpreted as unobservable characteristics and JMP propose a way to look at how the distributions of these unobserved characteristics differ among the two groups. JMP define the cumulative distribution of wage residuals, conditional on covariates, as θig ¼ F(υig| Xig). This gives us, υig ¼ F1 θig jXig F1 ig , which is the inverse cumulative distribution of wage residuals. Using this definition of residuals, we can rewrite Eq. (1) as: Y gi ¼ βg0 þ ΣKk¼1 Xgik βgk þ F1 ig
ð13Þ
Assuming the two groups to be men and women, there are now three potential sources of the gender wage gap, differences in the observables, X; differences in returns to observables, β; and differences in the distribution of unobservables (residuals), F1 g . To obtain the decomposition, JMP recommend generating two different counterfactuals: K 1 Y C1 i ¼ βM0 þ Σk¼1 XWik β Mk þ FiM
ð14Þ
K 1 Y C2 i ¼ βW0 þ Σk¼1 XWik βWk þ FiM
ð15Þ
144
G. Popli
The first counterfactual, Y C1 i , gives us the counterfactual wages for women, if the observable characteristics are of women and the returns to the observables and the residual distribution are of men. The second counterfactual, Y C2 i , gives us the counterfactual wages for women, if both the observable characteristics and returns to them are of women, but the residual distribution is of men. Using the two counterfactuals, the decomposition is given as: YM YW ¼ YM Y
C1
þ Y
C1
Y
C2
The first term on the right-hand side of Eq. (16),
þ Y
C2
YW
YM Y
C1
ð16Þ , gives us the
difference in wages due to observable characteristics (the composition effect); C1 C2 , gives us the difference in wages due to returns to second term, Y Y observable characteristics (the wage structure effect); and the third term, C2 Y Y W , gives us the difference in wages due to differences in the distribution of the unobservable factors. While proving to be very useful in looking at the role of unobserved characteristics in explaining wage differentials, the JMP method has some limitations which need to be kept in mind. First, if the number of observations between the two groups does not match, it is not clear how to assign residuals from one group to another, that is, how to generate the first counterfactual, Eq. (14). The solution proposed by JMP for this is to replace the ith ranked residual from women’s residual distribution with the ith ranked residual from the residual distribution for men. Second, while theoretically, we need θig ¼ F(υig| Xig), empirically all we can get is θig ¼ F(υig), which implies independence between observed and unobserved characteristics. This can be an unrealistic assumption. Third, the decomposition is path dependent, the order in which we generate the counterfactuals can change the size of the different effects. Lastly, in the empirical analysis, the three components of the decomposition need not add up to the observed mean gap. For a full discussion of the JMP method’s limitations and possible solutions, see Lemieux (2002) and Yun (2009).
Reweighting Methods DiNardo, Fortin, and Lemieux (1996), DFL henceforth, generalized the OB method for the entire distribution of wages. We start by estimating the distribution of observed wages for the two groups, f Y g f ðYjgÞ ¼ f ðYjg, xÞhðxjgÞdx
ð17Þ
where f(Y|g) is the distribution of the wages for group g; f(Y|g, x) is the wage distribution for group g given individual characteristics, X ¼ x; and h(x|g) is the
8
Decompositions: Accounting for Discrimination
145
distribution of individual characteristics for group g. The empirical counterpart to Eq. (17) can be given as: ng
f Yg ¼
j¼1
K
Y g,i Y g,j , for all i ¼ 1, . . . ,ng bg
ð18Þ
where K(.) is the kernel function and ng is the number of observations in group g. The actual distributions are estimated for both groups, let’s say men and women. DFL then propose an estimation of a counterfactual distribution for one of the groups, we show it for women, defined as: f C ðY W Þ f C ðYjMÞ ¼ f ðYjM, xÞhðxjW Þdx ¼ ωðxÞf ðYjM, xÞhðxjMÞdx
ð19Þ
where f C(YW) is the counterfactual distribution of women, such that the distribution of individual characteristics is as that of women, but they are paid as men would be. The counterfactual distribution suggested by DFL is the distributional equivalent of counterfactual wage regression defined in Eq. (3). The counterfactual distribution is simply a reweighted distribution of men, where ω(x) is the reweighting function Þ defined as ωðxÞ hhððxjW xjMÞ . To estimate the counterfactual distribution, we need an estimate of the reweighting function, which using the Bayes rule can be written as: ωð x Þ ¼
PrðWjxÞ=PrðW Þ PrðMjxÞ=PrðMÞ
ð20Þ
To estimate Eq. (20), we pool the data and estimate a probit model, where the dependent variable is the binary gender variable, and the covariates are the individual characteristics, X. Pr(W| x) and Pr(M| x) are simply the predicted probabilities from the probit model, and Pr(W) and Pr(M ) are the unconditional probabilities. Once we have the estimate of the reweighting function, empirically the counterfactual distribution can be estimated as: C
f ðY M Þ ¼
nM j¼1
ωðxÞK
Y M,i Y M,j , for all i ¼ 1, . . . nM bM
ð21Þ
Once we have estimates of the actual and the counterfactual distributions, we can f estimate the distributional composition effects (ΔX ) and the wage structure effects f
(ΔS ) as: f
f
ΔX ¼ f ðY M Þ f C ðY W Þ and ΔS ¼ f C ðY W Þ f ðY W Þ
ð22Þ
The DFL method is simple to implement and intuitive and can decompose statistics other than the mean. The reweighting function proposed by them has been used in the context of many other decomposition methods, including the
146
G. Popli
Recentered Influence Function approach discussed below. While aggregate decompositions are easy to do with the DFL method, detailed decompositions are still an issue. For binary variables, detailed decomposition is feasible, and the authors discuss how to do them in their 1996 paper; however, detailed decompositions for continuous covariates remain a challenge. See Butcher and DiNardo (2002) and Altonji et al. (2012) for some solutions to detailed decomposition within the DFL framework. Other methods, other than DFL, have been proposed in the literature to look at the decomposition of distribution; for one such method, see Jenkins (1994) who takes the Generalised Lorenz Curve approach to estimate both the actual distributions and the counterfactual distributions.
Conditional Quantiles While the DFL method allows us to do aggregate decomposition at the distributional level, detailed distribution beyond the mean remains a challenge. An alternative to looking at the entire distributions and doing detailed decomposition is proposed by Machado and Mata (2005), MM henceforth, who based their decomposition on conditional quantile regressions (Koenker and Bassett 1978). In this, instead of estimating a regression for the mean (OLS), we start by estimating a quantile regression for each group, men and women, given as: Qτ Y g jXg ¼ Xg βτg , τ ð0, 1Þ
ð23Þ
where βτg gives us the returns to the characteristics, X, on the τth quantile of the wage (Y ) distribution. As in the OB method, we next estimate a counterfactual distribution for women if they had their own characteristics but are paid as men would be: Qτ Y C jXW ¼ XW βτg
ð24Þ
Once we have estimates of the actual and the counterfactual quantile regressions, τ τ we can estimate the composition effects (ΔX ) and the wage structure effects (ΔS ) at different quantiles as: τ
τ
ΔX ¼ Qτ ðY M jXM Þ Qτ Y C jXW and ΔS ¼ Qτ Y C jXW Qτ ðY W jXW Þ To do the quantile decomposition, MM suggest the following simulation: 1. Sample τ from a standard uniform distribution. τ 2. Estimate the quantile regression for the τth quantile, and obtain βg .
ð25Þ
8
Decompositions: Accounting for Discrimination τ
3. Compute Qτ Y C jXW ¼ XW βM then computed as residual,
τ
and Qτ ðY W jXW Þ ¼ XW βM . The difference
between the two gives us the wage structure effect τ ΔX
147
τ (ΔS ),
composition effect is τ
¼ Qτ ðY M jXM Þ Qτ ðY W jXW Þ ΔS
4. Repeat steps 1 to 3 M times. MM method is computationally very intensive; further, while it allows for the detailed decomposition of the wage structure effect, we cannot obtain the detailed decomposition of the composition effect. Melly (2005) provides a way to reduce the computation time and do a detailed decomposition of the composition effect at the median. Chernozhukov et al. (2013) provide a further extension of the MM method, giving detailed decomposition for both the wage structure and the composition effect. A key limitation of this method is that the detailed decompositions based on conditional quantile regressions are path dependent, the order in which the various covariates are considered in decomposition can alter their contribution to the explained and the unexplained components.
Recentered Influence Functions Firpo et al. (2007, 2009) proposed a way to look at the unconditional distributions and do both the aggregate and detailed decomposition, by using Recentered Influence Function (RIF) regressions. The RIF regression for the τth quantile, qτ, of the wages, Yg, for group g, is defined as: RIF Y g , qτ ¼ qτ þ τ d g,τ =f Y g ðqτ Þ, . . . τ ð0,1Þ
(26)
where f Y g ðqτ Þ is the density function of Yg computed at quantile qτ, and dg, τ is the dummy variable taking value one if Yg qτ and zero otherwise. The RIF(Yg, qτ) has two properties that make it particularly useful; first, its expectation is the actual τth-quantile, EY[RIF(Yg, qτ)] ¼ qτ; and second, the expectation of the conditional RIF, when conditioning on the vector Xg, is also the actual τth-quantile, EX[Ey[RIF(Yg, qτ)| Xg]] ¼ qτ. Assuming RIF to be a linear function of covariates, we have, RIF Y g , qτ ¼ βτg0 þ ΣKk¼1 Xgik βτgk þ υτg
ð27Þ
where βτg is the vector of coefficients for the τth-quantile, and υτg is the error term. Given the two properties of the RIF function, Eq. (27) is the unconditional quantile regression, which is estimated separately for the two groups, men and women. The difference in the τth-quantile wage for men and women, qτ, M qτ, W, can then be decomposed as follows:
148
G. Popli τ
τ
τ
qτ,M qτ,W ¼ βM0 βW0 þ Δ
τ
k
τ
XWk βMk βWk
τ
ΔS
þ
τ
XMk XWk βMk
k
ð28Þ
τ
ΔX τ
On the left-hand side of Eq. (6), Δ , is the gap in the wages of men and women at τ the τth-quantile. The first term on the right-hand side, ΔS, is the wage structure effect τ at the τth-quantile, and the second term ΔX , is the composition effect at the τthquantile. Like the OB decomposition, we can obtain both the aggregate decomposition and the detailed decomposition, which are path independent. When the RIF is evaluated at the mean of Yg, we get OB decomposition as a special case. The RIF regressions can be biased as the assumption of linearity holds true only locally. To correct for the specification error, the RIF regression is combined with the DFL reweighting function. This requires estimating the RIF regression for women with the reweighting function such that the covariates of women have the same distribution as of men. This yields the specification error and the reweighting error, separately from the composition and the wage structure effect, respectively. Empirically these errors tend to be very small.
Conclusion Oaxaca and Blinder decomposition was first popularized in the 1970s, the use and popularity of this method have not waned over the last five decades. This chapter summarizes the OB decomposition’s main facets, its main limitations, cautions that should be exercised when using it, and the formal identification assumptions underlying it. We also link OB decomposition to the treatment effect literature. This chapter is not an exhaustive discussion of all methods of decomposition and the issues underlying them. For example, we have only focused on continuous outcomes, and not discussed the various methods that have been proposed for limited dependent variables (Fairlie 2005). Instead, this chapter’s focus has been to introduce the regression-based decompositions and highlight, up to date, key innovations in this method. This main limitation of regression-based methods, from the perspective of discrimination studies, is the interpretation of the unexplained gaps. On the one hand, not all of the unexplained gap can be discrimination, there often are unobservable factors at play making the unexplained gap an overestimation of discrimination. On the other hand, the estimated coefficients (return to endowments) from regressions already taken into account the feedback from the market, so any estimated discrimination is likely to be understated. The decomposition methodology has come a long way from the original work of Oaxaca and Blinder; increasingly we have methods
8
Decompositions: Accounting for Discrimination
149
that help us address the issues of unobserved productivity differences, self-selection, and omitted variables. In recent years, in an attempt to separate discrimination from unobserved productivity differences, self-selection, and omitted variables, there has been substantial growth in the experimental literature on discrimination (for a review, see Neumark 2018). While the experimental methods help us identify discrimination more robustly and explore the underlying mechanisms, they have their limitations. Most of the experimental literature in labor economics focuses on hiring, which may not impact earnings, whereas the decomposition methods can look at earnings directly. Acknowledgments I would like to thank Okan Yilmaz and Meng Le Zhang for their helpful comments on an earlier draft of the chapter.
References Althauser RP, Wigler W (1972) Standardisation and component analysis. Sociol Methods Res 1: 97–135 Altonji JG, Bharadwaj P, Lange F (2012) Changes in the characteristics of American youth: implications for adult outcomes. J Labor Econ 30(4):783–828 An W, Glynn AN (2019) Treatment effect deviation as an alternative to Blinder–Oaxaca decomposition for studying social inequality. Sociol Methods Res 50(3):1006–1033. https://doi.org/ 10.1177/0049124119852387 Arrow KJ (1998) What has economics to say about racial discrimination? J Econ Perspect 12(2): 91–100 Barsky R, Bound J, Charles KK, Lupton JP (2002) Accounting for the black–white wealth gap: a nonparametric approach. J Am Stat Assoc 97(459):663–673 Becker GS (1957) The economics of discrimination. University of Chicago Press, Chicago and London Blinder AS (1973) Wage discrimination: reduced form and structural estimates. J Hum Resour 8(4): 436–455 Butcher KF, DiNardo J (2002) The immigrant and native-born wage distributions: evidence from United States censuses. Ind Labour Relations Rev 56(1):97–121 Chernozhukov V, Fernández-Val I, Melly B (2013) Inference on counterfactual distributions. Econometrica 81(6):2205–2268 Cotton J (1988) On the decomposition of wage differentials. Rev Econ Stat 70:236–243 DiNardo J, Fortin N, Lemieux T (1996) Labor market institutions and the distribution of wages, 1973-1992: a semiparametric approach. Econometrica 64(5):1001–1044 Fairlie RW (2005) An extension of the Blinder-Oaxaca decomposition technique to logit and probit models. J Econ Soc Meas 30(4):305–316 Firpo S, Fortin N, Lemieux T (2007) Decomposing wage distributions using recentered influence function regressions. Working paper, University of British Columbia Firpo S, Fortin N, Lemieux T (2009) Unconditional quantile regressions. Econometrica 77(3): 953–973 Fortin N, Lemieux T, Firpo S (2011) Decomposition methods in economics. In: Handbook of labor economics, vol 4. Elsevier, pp 1–102 Gardeazabal J, Ugidos A (2004) More on identification in detailed wage decompositions. Rev Econ Stat 86:1034–1036 Heckman J (1979) Sample selection bias as a specification error. Econometrica 47:153–162 Jenkins SP (1994) Earnings discrimination measurement—a distributional approach. J Econ 61: 81–102
150
G. Popli
Jones FL (1983) On decomposing the wage gap: a critical comment on Blinder’s method. J Hum Resour 18:126–130 Jones FL, Kelley J (1984) Decomposing differences between groups: a cautionary note on measuring discrimination. Sociol Methods Res 12(3):323–343 Juhn C, Murphy KM, Pierce B (1993) Wage inequality and the rise in returns to skill. J Polit Econ 101(3):410–442 Kitagawa EM (1955) Components of a difference between two rates. J Am Stat Assoc 50: 1168–1194 Koenker R, Bassett G (1978) Regression Quantiles. Econometrica 46:33–50 Lang K, Kahn-Lang Spitzer A (2020) Race discrimination: an economic perspective. J Econ Perspect 34(2):68–89 Lemieux T (2002) Decomposing changes in wage distributions: a unified approach. Can J Econ 35(4):646–688 Machado JA, Mata J (2005) Counterfactual decomposition of changes in wage distributions using quantile regression. J Appl Econ 20(4):445–465 Melly B (2005) Decomposition of differences in distribution using quantile regression. Labour Econ 12(4):577–590 Neal DA, Johnson WR (1996) The role of premarket factors in black-white wage differences. J Polit Econ 104(5):869–895 Neuman S, Oaxaca RL (2004) Wage decompositions with selectivity corrected wage equations: a methodological note. J Econ Inequal 2:3–10 Neumark D (1988) Employers’ discriminatory behavior and the estimation of wage discrimination. J Hum Resour 23:279–295 Neumark D (2018) Experimental research on labor market discrimination. J Econ Lit 56(3): 799–866 Ñopo H (2008) Matching as a tool to decompose wage gaps. Rev Econ Stat 90(2):290–299 Oaxaca R (1973) Male-female wage differentials in urban labor markets. Int Econ Rev 14:693–709 Oaxaca RL (2007) The challenge of measuring labor market discrimination against women. Swedish Econ Policy Rev 14(1):199–231 Oaxaca RL, Ransom MR (1994) On discrimination and the decomposition of wage differentials. J Econ 61:5–21 Oaxaca RL, Ransom MR (1999) Identification in detailed wage decompositions. Rev Econ Stat 81(1):154–157 Small ML, Pager D (2020) Sociological perspectives on racial discrimination. J Econ Perspect 34(2):49–67 Yun MS (2005) A simple solution to the identification problem in detailed wage decompositions. Econ Inq 43:766–772 Yun MS (2009) Wage differentials, discrimination and inequality: a cautionary note on the Juhn, Murphy and Pierce Decomposition Method. Scott J Political Econ 56(1):114–122
9
Field Experiments: Correspondence Studies Marianne Bertrand and Esther Duflo
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measuring Discrimination in the Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Audit Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Correspondence Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Correspondence Studies in Other Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Beyond the Résumés . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Limitations of Correspondence Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implicit Association Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Goldberg Paradigm Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List Randomization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Willingness to Pay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consequences of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Self-Expectancy Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Expectancy Effects and Self-Fulfilling Prophecies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discrimination in Politics and Inequality Across Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Benefits of Diversity? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What Affects Discrimination? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Leaders and Role Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Intergroup Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Socio-Cognitive De-Biasing Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Technological De-Biasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
152 157 158 160 169 172 173 178 182 183 184 187 187 192 196 197 200 200 206 209 218
This chapter was originally published in 2016 as Chapter 8 in the Handbook of Economic Field Experiments, and can be accessed at https://doi.org/10.1016/bs.hefe.2016.08.004. Re-published here with permission. M. Bertrand (*) University of Chicago Booth School of Business, NBER, and J-PAL, Chicago, IL, USA e-mail: [email protected] E. Duflo MIT Economics Department, NBER and J-PAL, Chicago, IL, USA e-mail: edufl[email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_16
151
152
M. Bertrand and E. Duflo
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Abstract
This chapter reviews the existing field experimentation literature on the prevalence of discrimination, the consequences of such discrimination, and possible approaches to undermine it. We highlight key gaps in the literature and ripe opportunities for future field work. Section “Measuring Discrimination in the Field” reviews the various experimental methods that have been employed to measure the prevalence of discrimination, most notably audit and correspondence studies; it also describes several other measurement tools commonly used in lab-based work that deserve greater consideration in field research. Section “Consequences of Discrimination” provides an overview of literature on the costs of being stereotyped or discriminated against, with a focus on selfexpectancy effects and self-fulfilling prophecies; section “Consequences of Discrimination” also discusses the thin field-based literature on the consequences of limited diversity in organizations and groups. The final section of the chapter, section “What Affects Discrimination?,” reviews evidence for policies and interventions aimed at weakening discrimination, covering role model and intergroup contact effects, as well as socio-cognitive and technological de-biasing strategies. Keywords
Field experiments · Discrimination · Implicit bias · Audit studies · Correspondence studies
Introduction Black people are less likely to be employed, more likely to be arrested by the police, and more likely to be incarcerated. Women are very scarce at the top echelon of the corporate, academic, and political ladders despite the fact that (in rich countries at least) they get better grades in school and are more likely to graduate from college. While many in the media and public opinion circles argue that discrimination is a key force in driving these patterns, showing that it is actually the case is not simple. Indeed, it has proven elusive to produce convincing evidence of discrimination using standard regression analysis methods and observational data, in the sense in which we define discrimination throughout this chapter: members of a minority group (women, blacks, Muslims, immigrants, etc.) are treated differentially (less favorably) than members of a majority group with otherwise identical characteristics in similar circumstances. However, over the last couple of decades, a rich literature in economics, sociology, political science, and psychology has leveraged experiments (in the lab and in the field) to provide convincing evidence that discrimination, thus defined, indeed exists. We begin this chapter by describing the various experimental methods that
9
Field Experiments: Correspondence Studies
153
have been used to measure discrimination in the field. Overall, this chapter offers staggering evidence of pervasive discrimination against minority or underrepresented groups all around the world. We summarize this chapter and discuss some of its key limitations. If discrimination is as pervasive as evidence suggests, what do existing theories tell us about costs to minority groups and to society overall? The two workhorse models of discrimination in the economics literature give drastically different answers, particularly with respect to the societal consequences. In the first model, developed in Becker (1957) for the context of the labor market, some employers have a distaste for hiring members of the minority group. They may indulge this distaste by refusing to hire, say, blacks or, if they do hire them, paying them less than other employees for the same level of productivity. If the fraction of discriminating employers in the economy is sufficiently large, a wage differential will emerge in equilibrium between otherwise identically productive minority and majority employees and this wage differential will be a reflection of the distaste parameter of the marginal employer for minority workers (Becker 1957; Charles and Guryan 2008). By electing to not hire minority workers, inframargin racist employers will experience lower profits. In fact, if the conditions of perfect competition were satisfied, discriminating employers would be wiped away and taste-based discrimination would disappear.1 This “taste-based” explanation for discrimination stands in contrast with what many economists would view as a more disciplined explanation, which does not involve an ad hoc (even if intuitive) addition to the utility function (animus toward certain groups) to help rationalize a puzzling behavior. In a “statistical discrimination” model (Phelps 1972; Arrow 1973; Aigner and Cain 1977), the differential treatment of members of the minority group is due to imperfect information, and discrimination is the result of a signal extraction problem. As a profit-maximizing prospective employer, renter, or car salesman tries to infer the characteristics of a person that are relevant to the market transaction they are considering to complete with that person, they use all the information available to them. When the personspecific information is limited, group-specific membership may provide additional valuable information about expected productivity. For example, again using the labor market scenario, it may be known to employers that minority applicants are on average less productive than majority applicants. In this case, an employer who sees two applicants with similar noisy but unbiased signals of productivity should rationally favor the majority applicant to the minority one as her expected productivity is higher. While expected productivity will equal true productivity on average within each group, statistical discrimination will result in some minority workers being treated less favorably than nonminority workers of the same true productivity, i.e., will result in discrimination as defined above. In the extreme case where individual signals of productivity are totally uninformative, an employer may
1
Refusing to hire black people could be efficient if the employer knows he cannot work well with them due to his animus, but this does not take away from the fact that businesses that do this should not survive.
154
M. Bertrand and E. Duflo
rationally decide to make offers only to whites if the mean productivity among blacks does not exceed the required threshold. While taste-based discrimination is clearly inefficient (simply consider how it constrains the allocation of talent), statistical discrimination is theoretically efficient and hence more easily defendable in ethical terms under the utilitarian argument. Moreover, statistical discrimination can also be argued to be “fair” in that it treats identical people with the same expected productivity (even if not with the same actual productivity) and is not motivated by animus. In fact, many economists would most likely support allowing statistical discrimination as a good policy, even where it is now illegal (as it is, for example, in the US labor market and real estate market contexts). Unfortunately, as we discuss below, while field experiments have been overall successful at documenting that discrimination exists, they have (with a few exceptions) struggled with linking the patterns of discrimination to a specific theory. Meanwhile, psychologists have made considerable progress in their own understanding of the roots of discrimination on a largely parallel track. The theories they have advanced and the (mainly lab) experiments they have conducted have been helpful in better nailing the microfoundations of discrimination. We believe this body of work blurs the sharp line economists tend to draw between taste-based and statistical explanations. Psychologists’ work on discrimination is embedded in an immense literature that attempts to understand the roots of prejudice, widely characterized as negative evaluation of others made on the basis of their group membership. This literature has looked for the microfoundations of such negative evaluation in a wide variety of areas, including personality development, socialization, social cognition, evolutionary psychology, and neuroscience. Early scholarship in psychology viewed prejudice as a form of abnormal thinking and equated it to a psychopathology (think Adolf Hitler) that could be treated by addressing the personality disorders of the subset of the population that was “diseased.” It was only in the second half of the twentieth century that the prevalent view of prejudice among psychologists became rooted in normal thinking processes (Dovidio 2001), with socialization and social norms being viewed as dominant drivers. Influential work by Tajfel (1970; Tajfel and Turner 1979) demonstrated the key role social identity plays in the process underlying prejudice. Experimental evidence has shown that the assignment of people to groups, even if they are totally arbitrary ones and even if they do not last, is sufficient to produce favoritism for in-group members and negativity toward out group members. At the same time, evolutionary psychology has stressed the importance of social differentiation and the delineation of clear group boundaries as a way to achieve the benefits of cooperation between human beings without the risk of excessive costs, with group membership and group identity emerging as a form of contingent altruism (Brewer 1981). While in-group love might not necessarily imply out-group hate, the same factors that make allegiance with group members important provide grounds for antagonism and distrust of outsiders.
9
Field Experiments: Correspondence Studies
155
In addition, more recent advances in the psychology literature have demonstrated the existence of unconscious, unintentional forms of bias. Modern social psychologists believe that attitudes can occur in implicit modes and that people can behave in ways that are unrelated to or even sometimes opposed to their explicit views or selfinterests (Banaji and Greenwald 1995; Bertrand et al. 2005; Dovidio et al. 1998a, b; Greenwald and Banaji 1995). Neuroscience studies have shown that different regions of the brain are activated in conscious versus unconscious processing, suggesting that unconscious processes are distinct mental activities. For example, the unconscious processing of black faces has been associated with activations of the area of the brain associated with emotions and fear while the conscious processing of the same faces increases brain activity in areas related to control and regulation. Implicit biases are more likely to drive behavior under conditions of ambiguity, high time pressures and cognitive loads, or inattentiveness to the task. Both of these dominant views of prejudice in the psychology literature – as an evolutionary phenomenon making group membership an important component of one’s social identity or as an unconscious automatic negative association triggered by exposure to out-group members – could serve as microfoundations to what the more reduced form “animus-based” models economists have worked with. More importantly, these psychological models make clear that the limited information and decision-making model that drives statistical discrimination might be itself endogenous to conscious or unconscious prejudice against the out-group members. If a social need to positively associate with one’s own group also makes the out-group members feel more distant and unknowable (Brewer 1988), an employer may not invest in collecting information on an out-group member, or decide that the individual signals of productivity for minority group members are totally uninformative, resulting in all minority group members being equally treated as unhirable. Limited de-facto contact between in-group and out-group members will imply that majority employees or coworkers will be fairly ignorant about the quality of minorities; this would mean that employing, electing, or renting to them may seem riskier which, in the presence of risk aversion, will also trigger more statistical discrimination (Aigner and Cain 1977). Unconscious bias may influence the specific criteria or formulae that are used to assess expected productivity (Uhlmann and Cohen 2005), for example, the sense of danger that is implicitly triggered by seeing a black face or reading a black name on a résumé may lead an employer to put too great a weight on docility as a work quality than would be warranted for maximum productivity. Recently, the emphasis on “fit” between a prospective employee and the company as a hiring criterion in technology jobs has raised the spectrum of a new form of subtle discrimination. Similarly, unconscious stereotypes may influence our judgment of the inputs into productivity, with the same level of assertiveness being deemed as good for productivity when coming from men but bad when coming from women (Rudman and Glick 2001). Perhaps most importantly, whether discrimination is taste-based or statistical, it may ultimately result in genuine difference between groups through self-fulfilling prophecies. If the stereotypical woman is not good at math, talented girls may become discouraged and ultimately not become good at math. If teachers or
156
M. Bertrand and E. Duflo
employers assume that students of a particular color are less smart, they will invest less in them. Thus, discrimination, whether it is taste-based or statistical, can create or exacerbate existing differences between groups. Discrimination that starts as tastebased and inefficient can easily morph into the more “justifiable” form. “Valid” stereotypes today could be the product of ambient animus, very much complicating the division between the different theories of discrimination. The chapter proceeds as follows. Section “Measuring Discrimination in the Field” is devoted to the various experimental methods that have been used in the field to measure discrimination, in particular audit and correspondence studies. Audit studies send out individuals who are matched in all observable characteristics except for the one in question (race, criminal record, etc.) to apply for jobs or make purchases; then researchers analyze the responses. Correspondence studies – which represent by far the largest share of field experiments on discrimination – do the same but control for more variables by creating fictitious applicants (often for jobs or apartments) who correspond via mail. We summarize the findings of this body of work (which clearly demonstrate the pervasiveness of discrimination) and discuss its key limitations. In this section of the chapter, we also discuss a few alternative methods to measure discrimination, many of them, such as Implicit Association Tests, Goldberg Paradigm experiments, and List Randomization, having developed in the psychology literature for use in the lab, as well as measures of willingness to pay to interact with minority group members. We argue that these alternative methods deserve more consideration by economists interested in measures of discrimination for their field research. Section “Consequences of Discrimination” reviews the work that addresses the costs of being discriminated against, or stereotyped. In particular, we review the experimental work that has studied how the threat of being viewed through the lens of a negative stereotype can have a direct negative effect on performance. We also review the experimental literature on expectancy effects, the goal of which has been to understand how stereotypes and biases against minority groups may end up being self-fulfilling. We round up the second part of the chapter by reviewing what is a surprisingly thin amount of field-based literature on the costs (and benefits) of the limited diversity in organizations and groups that directly result from discrimination. This allows us to discuss field work that has considered the consequences of discrimination not just from the perspective of the group that is discriminated against, but also from the perspective of society as a whole. The third and final section of this chapter, section “What Affects Discrimination?,” is related to the review of various interventions and policies that have been proposed to undo or weaken discrimination. This section covers topics such as the impact of role models, how contact and exposure to the minority groups may change prejudice, as well as large psychology literature on both socio-cognitive and technological de-biasing strategies. We argue that there is a lot of promising future research that is “ripe for the picking” in this area, given the large amount of theoretical and lab-based work that has not yet been taken to the field.
9
Field Experiments: Correspondence Studies
157
Measuring Discrimination in the Field Earlier research on discrimination focused on individual-level outcome regressions, with discrimination estimated from the “minority” differential that remains unexplained after including as many proxies as possible for productivity.2 The limitations of this approach are well-known. The interpretation of the estimated “minority” coefficient is problematic due to omitted variables bias. Specifically, results of a regression analysis might suggest differential treatment by race or gender even if the decision-maker (say, an employer) never used group membership in her decision of how much to pay an employee. However, it could be the case that race or gender is correlated with other proxies for productivity that are unobservable to the researcher but observed by the employer. It is therefore impossible to conclude that the employer used group membership in her decision-making process using this method. The traditional answer has been to saturate the regression with as many productivity-relevant, individual-level characteristics as are available. But, of course, ensuring that the researcher observes all that the decision-maker observes is a hopeless task. Moreover, adding more and more controls to a regression could ultimately obscure the interpretation of the evidence. Consider the labor market context: minority workers might be best-responding to the discrimination they know to exist and could have simply sorted into industries where there is no or limited discrimination. Hence, finding no racial gap in earnings after controlling for industry or employer fixed effects in a regression may indicate that there is no discrimination at the margin, which is very different from no discrimination on average. Also, as pointed out in Guryan and Charles (2013), the variables the researcher controls for might themselves be affected by discrimination. That is, disadvantaged groups may not have access to high-quality schools because of discrimination, yet they might, given their low human capital accumulation, be paid the “fair market wage.” While one might still be tempted to conclude from this that there is no discrimination in the labor market but instead discrimination in the education market, that might not be right if the minority group’s expectations about labor market discrimination drive their educational decision. In other words, minority group members may decide to under invest in education if they expect that they will not be able to obtain labor market returns for this education. Audit and correspondence methodologies were developed to address these core limitations of the regression approach to measuring discrimination. We review below both types of studies, discuss the extent to which they address these limitations of the regression approach, and also consider new issues they create.
2
For a review of this earlier literature on the narrower topic of labor market discrimination, see Chap. 48 in Altonji and Blank (1999).
158
M. Bertrand and E. Duflo
Audit Studies In the best-known collection of audit studies exploring the extent of discrimination, Fix and Struyk (1993) describe the method as follows: Two individuals (auditors or testers) are matched for all relevant personal characteristics other than the one that is presumed to lead to discrimination, e.g. race, ethnicity, gender. They then apply for a job, a housing unit, or a mortgage, or begin to negotiate for a good or service. The results they achieve and the treatment they receive in the transaction are closely observed, documented, and analyzed to determine if the outcomes reveal patterns of differential treatment on the basis of the trait studied and/or protected by anti-discrimination laws. . .
Discrimination is said to have been detected when “auditors in the protected class are systematically treated worse than their teammates” (Yinger 1998).3 A well-known early example of the audit method is offered by Ayres and Siegelman (1995). In this study, pairs of testers (one of whom was always a white male) were trained to bargain uniformly and then were sent to negotiate for the purchase of a new automobile at randomly selected Chicago-area dealerships. Thirty-eight testers bargained for 306 cars at 153 dealerships. Testers were chosen to have average attractiveness, and both testers in a pair bargained for the same model of car, at the same dealership, usually within a few days of each other. Dealerships were selected randomly, testers were randomly assigned to dealerships, and the choice of which tester in the pair would be the first to enter the dealership was also randomly made. The testers bargained at different dealerships for a total of nine car models, following a uniform bargaining script that instructed them to focus quickly on one particular car and start negotiating over it. Testers were further instructed to tell dealers at the beginning of the bargaining that they could provide their own financing for the car. In spite of the identical approach to bargaining, Ayres and Siegelman (1995) find that white males are quoted lower prices than white women and blacks (men or women). While ancillary evidence suggests that the dealerships’ disparate treatment of women and blacks may be caused by dealers’ statistical inferences about consumers’ reservation prices, the data do not strongly support any single theory of discrimination. Another well-known audit study of the labor market is Neumark, Bank, and Van Nort (1996), which investigates the role of sex discrimination in vertical segregation among waiters and waitresses. Specifically, two male and two female college students were sent to apply in person for jobs as waiters and waitresses at 65 restaurants in Philadelphia. The restaurants were divided into high-, medium-, and low-price categories, with the goal of estimating sex differences in the receipt of job offers in each price category. The study was designed so that a male and female
3
Results from the earliest audit studies can be found in Newman (1978), McIntyre, Moberg, and Posner (1980), Galster (1990), Yinger (1986), Cross, Kenney, Mell, and Zimmerman (1990), James and DelCastillo (1991), Turner, Fix, and Struyk (1991), and Fix and Struyk (1993).
9
Field Experiments: Correspondence Studies
159
pair applied for a job at each restaurant, and so that the male and female candidates were on average identical. The findings are consistent with discrimination against women in high-price restaurants and discrimination in women’s favor in low-price restaurants. Of the 13 job offers from high-price restaurants, 11 were made to men. In contrast, of the ten job offers from low-price restaurants, eight were made to women. In addition, information gathered from restaurants included in the study suggests that earnings are substantially higher in high-price restaurants, meaning that the apparent hiring discrimination has implications for gender-based differences in earnings among waitpersons. Results are interpreted as consistent both with employer discrimination and customer discrimination. Another interesting application of the audit method is Pager (2003) who matched pairs of individuals applying for entry-level positions, and probed the impact of a criminal record, conditional on race. The author employed two black testers who formed a team, and another pair of white testers. Within each team, one auditor was “assigned” a criminal record (this assignment was random and rotating – that is, each tester played the role of an ex-convict at some point).4 In total, 350 employers were audited. The effect of the criminal record was both statistically significant and meaningful in magnitude: 17% of attempts with whites who had a supposed criminal record received a callback, compared to 34% of tries with whites who said they had no criminal record. That is, an equally qualified white candidate was about half as likely to receive a callback if he was believed to be an ex-convict. For black applicants, the effect was notably larger: 5% of attempts with blacks who were supposedly ex-convicts received a call-back, compared to 14% of applications with blacks that had no record, meaning an equally qualified black candidate was about one-third as likely to receive a callback if he had a criminal record. Furthermore, these estimates show that a black applicant without a criminal record was about as likely to receive a callback as a white applicant with a criminal record. Most audit studies do not explicitly test which theory of discrimination has most explanatory power, even if they often informally discuss what forms of discrimination might or might not be consistent with the observed patterns in the data. A notable exception is List (2004) who recruited buyers and sellers at a sports cards market and documented that minority buyers receive lower offers when they bargain for a collectible card. One finding of List (2004) is that in this context lack of information – and the expectation that minorities are inexperienced – drives discriminatory behavior. Experienced dealers discriminate more. Among experienced buyers, final offers to minorities are similar to offers received by white men, but minorities require more time to achieve this outcome. Moreover, List tries to rule out taste-based explanations for the data by combining the field data with results from a dictator game conducted in the lab with these card dealers. He finds that nonwhite males receive roughly as many positive allocations in this game as white males and interprets this pattern as evidence for the absence of taste for discrimination.
Pager argues that, “[b]y varying which member of the pair presented himself as having a criminal record, unobserved differences within the pairs of applicants were effectively controlled for.”
4
160
M. Bertrand and E. Duflo
Of course, while a laboratory experiment is a useful complement to the field study, the behavior of dealers in the dictator game, on its own, does not prove that tastebased discrimination is absent during the actual market transactions.
Limitations of Audit Studies Many of the weaknesses of audit studies have been discussed in Heckman and Siegelman (1993) and Heckman (1998). First, these studies require that both members of the auditor pair be identical in all dimensions that might affect productivity in employers’ eyes, except for the trait that is being manipulated. To accomplish this, researchers typically match auditors on several characteristics (height, weight, age, dialect, dressing style, and hairdo) and train them for several days to coordinate interviewing styles. Yet, critics note that this is unlikely to erase the numerous differences that exist between the auditors in a pair. Another weakness of the audit studies is that they are not double-blind: Auditors know the purpose of the study. As Turner, Fix, and Struyk (1991) note, “The first day of training also included an introduction to employment discrimination, equal employment opportunity, and a review of project design and methodology.” This may generate conscious or subconscious motives among auditors to generate data consistent or inconsistent with their beliefs about race or gender issues. As psychologists have documented, these demand effects can be quite strong. It is very difficult to insure that auditors will not want to do “a good job.” Even a vague belief by auditors that employers treat minorities differently can result in measured differences in treatment. The possibility of such a demand effect is further magnified by the fact that auditors are not in fact seeking jobs (or trying to buy a car for themselves) and are therefore more free to let their beliefs affect the bargaining or interview process.
Correspondence Studies Correspondence studies have been developed to address some of the more obvious weaknesses of the audit method. Rather than relying on real auditors or testers that physically meet with a potential employer or potential landlord, correspondence studies rely on fictitious applicants. Specifically, in response to a job or rental advertisement, the researcher sends (many) pairs of résumés or letters of interest, one of which is assigned the perceived minority trait. Discrimination is estimated by comparing the outcomes for the fictitious applicants with and without the perceived minority trait. The most common (but not the only) way to manipulate the perceived minority trait has been through the names of the applicants (e.g., female names, African-American names, Arabic names, etc.). Outcomes studied in a correspondence study have been mainly, but not exclusively, limited to measuring call-backs by employers or landlords in response to the mailed or emailed fictitious application.5
5
See sections “Retail” and “Academia” for cases of different approaches.
9
Field Experiments: Correspondence Studies
161
The correspondence method presents several advantages over the audit method. First, because it relies on résumés or applications by fictitious people and not real people, one can be sure to generate strict comparability across groups for all information that is seen by the employers or landlords. This guarantees that any observed differences are caused solely by the minority trait manipulation. Second, the use of paper applications insulates from demand effects. Finally, because of the relatively low marginal cost, one can send out a large number of applications. Besides providing more precise estimates, the larger sample size also allows researchers to examine the nature of the differential treatment from many more angles and hence promises to link it more closely to specific theories of discrimination.6 Although Guryan and Charles (2013) call correspondence tests a “significant methodological advance,” and a review of discrimination in the marketplace published about 15 years ago discussed only observational and audit studies (Yinger 1998); the method is actually not that new. Fictitious applications and résumés have been sent to employers in order to uncover racial or religious discrimination nearly half a century ago.7 However, the number of correspondence studies in economics has greatly increased following Bertrand and Mullainathan (2004), who study race discrimination in the labor market by sending fictitious résumés in response to helpwanted ads in Boston and Chicago newspapers. To manipulate perceived race, they randomly assigned very white-sounding names (such as Emily Walsh or Greg Baker) to half the résumés and very African-American-sounding names (such as Lakisha Washington or Jamal Jones) to the other half. In total, they responded to over 1300 employment ads in the sales, administrative support, clerical, and customer services job categories and sent out nearly 5000 résumés. They find that white names receive 50% more call-backs for interviews.
Correspondence Studies in the Labor Market The main results of labor market correspondence tests are reviewed in Table 1. As is clear from Table 1, labor market correspondence studies have by now been carried out in many countries around the world and have focused on a variety of perceived traits that can be randomized on a résumé. Below, we review some of these studies in more detail, focusing in particular on those that have attempted to go beyond simply documenting whether or not differential treatment occurs based on the manipulated traits, and move toward understanding which theory may best fit the patterns in the data. However, one of our bottom lines will be that, unfortunately, the studies have tended to be fairly close replications of Bertrand and Mullainathan (2004) for different populations or contexts. With a few exceptions, the literature has We discuss in section “Limitations of Correspondence Studies” other weaknesses that are shared by the correspondence studies, as well as added weaknesses of the correspondence method compared to the audit method. 7 See Jowell and Prescott-Clarke (1970), Jolson (1974), Hubbuck and Carter (1980), Brown and Gay (1985), and Riach and Rich (1991) for early studies. One caveat is that some of these studies fail to fully match skills between minority and nonminority résumés. 6
162
M. Bertrand and E. Duflo
Table 1 Correspondence Studies across countries Effect (callback ratio) White-toIndigenous ratio: 1.8 Low attractiveness hurts white females Employed to long-term unemployed: 1.25
Paper Galarza and Yamada (2014) Trait: ethnicity; attractiveness
Country Peru
CVs/apps 4820
Vacancies 1205
Eriksson and Rooth (2014) Trait: unemployment duration Blommaert, Coenders, and van Tubergen (2014) Trait: Arabic name
Sweden
8466
–
Netherlands
636
–
Nunley, Pugh, Romero, and Seals (2014) Trait: race
The USA
9396
–
Ghayad (2013) Trait: unemployment duration Bartoš, Bauer, Chytilová, and Matejka (2013) Trait: ethnicity (Roma, Asian, and Turkish)
The USA
3360
600
Employed-tounemployed: 1.47
Czech Rep. and Germany
274 (Czech R.) 745 (Ger.)
–
The USA
6400
1600
Czech-toVietnamese: 1.34 Lower requests for CVs if candidate is Turkish White-toMuslim: 1.58
The USA (largest 100 MSAs)
12,054
3040
Wright, Wallace, Bailey, and Hyde (2013) Trait: religion/ ethnicity Kroft, Lange, and Notowidigdo (2013) Trait: unemployment duration
Dutch-toforeign: 1.62 (unconditional ratio) no difference, if views held fixed White-toblack: 1.18 (unconditional)
1 log point change in unemployment duration: 4.7 percentage points lower callback probability
Theory No
No
No
Inconsistent with statistical discrimination, consistent with taste-based discrimination No
Consistent with attention discrimination
Consistent with theoretical models of secularization and cultural distate theory No
(continued)
9
Field Experiments: Correspondence Studies
163
Table 1 (continued) Paper Baert, Cockx, Gheyle, and Vandamme (2013) Trait: nationality (Turkishsounding name) Bailey, Wallace, and Wright (2013) Trait: sexual orientation Ahmed, Andersson, and Hammarstedt (2013) Trait: sexual orientation Acquisti and Fong (2013) Traits: sexual orientation and religion Patacchini, Ragusa, and Zenou (2012) Traits: sexual orientation and attractiveness Kaas and Manger (2012) Trait: immigrant (race/ethnicity) Maurer-Fazio (2012) Trait: ethnicity
Jacquemet and Yannelis (2012) Trait: race / nationality
Effect (callback ratio) Flemish-toTurkish: 1.03–2.05, depending on the occupation
Country Belgium
CVs/apps 752
Vacancies 376
Theory No
The USA
4608
1536
No effect
No
Sweden
3990
–
No
The USA
4183
–
Heterosexualto-homosexual (male): 1.14 Heterosexualto-homosexual (female): 1.22 Christian-toMuslim: 1.16
Italy
2320
–
Heterosexualto-homosexual: 1.38
No
Germany
1056
528
Consistent with statistical discrimination
China
21,592
10,796
The USA
330
990
German-toTurkish: 1.29 (if no reference letter is included) Han-toMongolian: 1.36 Han-toTibetan: 2.21 English-toforeign names: 1.41 English-toblack names: 1.46
No
No
Consistent with patterns of ethnic homophily
(continued)
164
M. Bertrand and E. Duflo
Table 1 (continued) Paper Ahmed, Andersson, and Hammarstedt (2012) Trait: age Oreopoulos (2011) Trait: nationality (and race)
Country Sweden
CVs/apps 466
Vacancies –
Canada
12,910
3225
Carlsson (2011) Trait: gender Booth, Leigh, and Varganova (2011) Trait: Ethnicity Booth and Leigh (2010) Trait: gender
Sweden
3228
1614
Australia
Above 4000
–
Australia
3365
–
UK
1000+
–
Sweden
1970
985
McGinnity, Nelson, Lunn, and Quinn (2009) Trait: nationality/race
Ireland
480
240
Banerjee, Bertrand,
India
3160
371
Riach and Rich (2010) Trait: age Rooth (2009) Trait: attractiveness/ obesity
Effect (callback ratio) 31 year old to 46 year old: 3.23
Theory No
English nameto-immigrant: ranged from 1.39 to 2.71 (against Indian, Pakistani, and Chinese applicants) Female-tomale: 1.07
No
White-toItalian: 1.12 White-toChinese: 1.68 Female-tomale: 1.28 (femaledominated professions) 2.64 favoring younger candidates Nonobese/ attractive-toobese/ unattractive: ranged from 1.21 to 1.25 (but higher for some occupations) 1.8, 2.07, and 2.44 in favor of Irish and against Asians, Germans, and Africans, respectively Upper caste-toother backward
No
No
No
No
No
No
No (continued)
9
Field Experiments: Correspondence Studies
165
Table 1 (continued) Paper Datta, and Mullainathan (2009) Traits: caste and religion Lahey (2008) Trait: age Petit (2007) Traits: age, gender, and number of children
Country
The USA
CVs/apps
Vacancies
–
France
App. 4000 942
157
Bursell (2007) Trait: ethnicity
Sweden
3552
1776
Bertrand and Mullainathan (2004) Trait: race
The USA
4870
1300+
Jolson (1974) Trait: race and religion
The USA
300
–
Effect (callback ratio) castes: 1.08 (software jobs, insignificant), 1.6 (call-center jobs) Young-toolder: 1.42 Ranged from 1.13 to 2.43 against 25-year-old, childless women Swedish-toforeign names: 1.82 White-toAfricanAmerican:1.5 (1.22 for females in sales jobs) White-toblack: 4.2 for selling positions
Theory
No No
Inconsistent with statistical discrimination No
No
failed to push the correspondence methodology to design approaches to more formally test for various theories of why differential treatment is taking place. Race, ethnicity: Studies of labor market discrimination based on race and ethnic background have been by far the most popular application of the correspondence method to date. While publication bias is always a concern, the results of correspondence studies where the trait of interest is race or ethnicity offer overwhelming evidence of discrimination in the labor market against racial and ethnic minorities. Evidence has been accumulated from nearly all continents: Latin America (e.g., Galarza and Yamada (2014) compare whites to Indigenous applicants in Peru), Asia (e.g., Maurer-Fazio (2012) compares Han, Mongolian, and Tibetan applicants in China), Australia (e.g., where Booth, Leigh, and Varganova (2011) compare whites to Chinese applicants), Europe (e.g., Baert et al. (2013) compare immigrants to nonimmigrants in Belgium), Ireland (e.g., where McGinnity et al. (2009) compare candidates with Irish names to candidates with distinctively non-Irish names), etc. Attempts to adapt the correspondence method to learn more about which theory of discrimination best fits the patterns in the data have been mainly focused on trying to provide corroborative evidence for (or against) statistical discrimination. The most
166
M. Bertrand and E. Duflo
common approach has been to investigate whether the gap in callbacks is responsive to the amount of information provided to employers about the job applicants, as was first done in Bertrand and Mullainathan (2004), in which they studied how credentials affect the racial gap in callback. In particular, Bertrand and Mullainathan (2004) experimentally varied the quality of the résumé used in response to a given ad. Higher-quality applicants had on average a little more labor market experience and fewer holes in their employment history; they were also more likely to have an e-mail address, have completed some certification degree, possess foreign language skills, or have been awarded some honors. The authors sent four résumés in response to each ad: two higher-quality and two lower-quality ones. They randomly assigned an African-American sounding name to one of the higher- and one of the lowerquality résumés. They find that whites with higher-quality résumés receive nearly 30% more callbacks than whites with lower-quality résumés. On the other hand, having a higher-quality résumé has a smaller effect for African-Americans. In other words, the gap between whites and African-Americans widens with résumé quality. While one may have expected improved credentials to alleviate employers’ fear that African-American applicants are deficient in some unobservable skills under a statistical discrimination explanation for the overall discrimination, this was not the case in their data. Bertrand and Mullainathan argue that one simple alternative model that may best explain the patterns in their data is some form of lexicographic search by employers: Employers receive so many résumés that they may use quick heuristics in reading these résumés. One such heuristic could be to simply read no further when they see an AfricanAmerican name. Thus they may never see the skills of African-American candidates and this could explain why these skills are not rewarded.
These findings are replicated in Nunley et al. (2014): Blacks received 14% fewer callbacks compared to whites, and discrimination was not mitigated when productive characteristics were added to a résumé. However, some studies report results that are more in line with the predictions of statistical discrimination models. Oreopoulos (2011) submitted 12,910 résumés in response to 3225 job postings in Canada. First, he compares (fictitious) applicants who had a foreign name, but who attended a Canadian (or foreign) university and had work experience in Canada. The callback rate is 1.39 for Canadian (English-sounding names) versus foreigners if they went to a Canadian university, and 1.43 if they went to a foreign university. However, the callback rate falls dramatically if the foreigners’ job experience was purely international (2.71 callback ratio). Moreover, if candidates who had foreign job experience and education had a Chinese last name with an English first name (Allen and Michelle Wang), their prospects on the job market improved. This raises the possibility that a fraction of the “discrimination” is statistical, for example, with employers making inference about the candidate’s English skills. Perhaps even more striking, Kaas and Manger (2012) sent out 528 pairs of applications in Germany to study the effect of a Turkish-sounding name. The German-to-Turkish callback rate was 1.29 when no reference letter was included. Discrimination was eliminated
9
Field Experiments: Correspondence Studies
167
when a reference letter, containing indirect information about productivity (such as conscientiousness and agreeableness), was added, which the authors interpret as evidence of consistency with statistical discrimination. It is interesting that such “soft information” presented in the reference letter appears to remove the difference in callback rates even though “harder information” presented in a résumé (such as employment history or honors) does not in other studies. It would be interesting to probe this contrast further. Gender: There are fewer studies on gender, and discrimination against women at the callback stage is much less apparent in general. Some studies attempt to show whether the degree (and nature) of discrimination depends on the nature of the profession. Carlsson (2011) sent paired applications for positions of IT professionals, drivers, construction workers, sales assistants, high school teachers, restaurant workers, accountants, cleaners, preschool teachers, and nurses. Overall, women are called back slightly more often than men; in male dominated professions, males have a slight (insignificant) advantage. Booth and Leigh (2010) focused on femaledominated professions (waitstaff, data-entry, customer service, and sales jobs) and found a callback of 1.28 in favor of women. A topic of interest for future work would be to apply the correspondence method to measure the extent to which a bias exists against women with children, or against young women who may have children in the future. To our knowledge only one study, Petit (2007), studies this aspect. In order to shed light on the role of family constraints in gender discrimination, Petit sent résumés for male and female applicants, with or without children, of age 25 or 37. Discrimination against women is detected for young workers in higher skilled positions (in the French finance industry), but not among prime-age workers. Caste and Religion: Banerjee et al. (2009) study the role of caste and religion in India’s software and call-center sectors. They sent 3160 fictitious résumés with randomly allocated caste-linked surnames in response to 371 job openings in and around Delhi (India) that were advertised in major city papers and online job sites. They find no evidence of discrimination against non-upper-caste (i.e., scheduled caste, scheduled tribe, and other backward caste) applicants for software jobs. But, in the case of call-center jobs, they do find larger and significant differences between callback rates for upper-castes and other backward castes (and to a lesser extent scheduled castes) in favor of upper-castes. They find no discrimination against Muslims. The potential impact of religion on job prospects in the USA is explored by Wright et al. (2013). Affiliation with a religion was signaled through student activities that were listed on résumés.8 The control group had no religious identification in their résumé. Compared to the control group, Muslim applications were
8
It is tricky to signal only religion on a résumé. The manipulation through student activities may reveal more than just religion, an issue we will come back to in section “Beyond the Résumés.”
168
M. Bertrand and E. Duflo
24% less likely to receive at least one contact by either email or phone, and they received 33% fewer total contacts than did those from the control group. Unemployment spells: More recently, researchers have applied the correspondence model to better understand patterns of labor market discrimination against the unemployed. In Sweden, Eriksson and Rooth (2014) randomly assigned various characteristics (contemporary unemployment, past unemployment immediately after graduation, past unemployment between jobs, work experience, and number of employers). Long-term unemployment did not harm job candidates’ chances, as long as the applicant had subsequent work experience. However, if the applicant was unemployed in the preceding 9 months, their callback probability fell by 20%.9 In the USA, Ghayad (2013) finds that (current) unemployment spells longer than 6 months are particularly harmful: The rate of interview requests for résumés with similar firm experience drops 1.13 percentage points for each additional month of nonemployment up to 6 months, and once the candidate experienced 6 months of unemployment, interview requests fell by an extra 8 percentage points. Kroft, Lange, and Notowidigdo (2013) relate these results to the inference problem faced by the prospective employers. The authors not only replicate the result that longer employment duration reduces callback rate, but also show that this depends on the labor market conditions. Duration dependence is stronger in tight labor markets, suggesting that employers use the information on the length of unemployment as a signal of productivity, but recognize that the signal is less informative when the labor market conditions are weak.10 Other characteristics: Sexual Orientation and Age: Résumé studies are now also being used to try to detect discrimination in a number of less obvious domains. Literature has tried to estimate discrimination against LGBT candidates; however, most studies have focused only on lesbians and gay men.11 One of the challenges with estimating discrimination against LGBT candidates is how to provide information that identifies a candidate as a member of that minority, when telling such details are not normally solicited in job applications. In Ahmed, Andersson, and Hammarstedt (2013), which was carried out in Sweden, sexual orientation was indicated by the mention of a “spouse” of either gender in the cover letter, and voluntary work in either an LGBT rights organization (gay identity) or the Swedish Red Cross (heterosexual identity). Targeted occupations included male-dominated ones (construction worker, motor vehicle driver, sales person, and mechanic worker), female-dominated ones (shop sales assistant, preschool teacher, cleaner, restaurant worker, and nurse), and more neutral ones (high school teacher). The authors find some mild evidence of discrimination (ratio of 1.14), which ultimately 9
One caveat, as the authors acknowledge, is that not all employers necessarily view the gaps on the CVs as implying unemployment. 10 This may also explain the finding in Eriksson and Rooth (2014) (mentioned above) since that particular study was carried out between March and November 2007, i.e., during the global financial crisis. 11 To the best of our knowledge, no study has been done that specifically looks at discrimination against transgender people using the correspondence method.
9
Field Experiments: Correspondence Studies
169
could be due to the nature of the signaling (e.g., working in LGBT rights, as opposed to the Red Cross, may be seen as a political gesture, not just revealing an identity). In Italy, Patacchini, Ragusa, and Zenou (2012) performed a correspondence study that revealed “homosexual preferences” through internships in progay advocacy groups and found higher discrimination against gay men (1.38) but not lesbian candidates. In the USA, Bailey, Wallace, and Wright (2013) find no evidence of discrimination against gay men or lesbians candidates. The issue of discrimination by age has also attracted some attention, and several papers (Ahmed et al. 2012; Lahey 2008; Riach and Rich 2010) find that younger candidates are generally preferred to older ones. A fundamental issue with this work is that it is hard to argue that age is not necessarily a proxy for productivity. Lahey (2008) tries to control for physical fitness with hobbies (e.g., racquetball is supposed to indicate fitness), but this is ultimately only moderately convincing. Finally, physical appearance has also been studied: Rooth (2009) studies obesity in the Swedish labor market and Patacchini, Ragusa, and Zenou (2012) investigate the beauty premium in Italy. Using manipulated facial photos to show an otherwise identical candidate as obese, Rooth (2009) shows there is a significantly lower callback response for obese people: Obese men had a six percentage points lower callback rate, while the callback rate for obese women was eight percentage points lower. Patacchini, Ragusa, and Zenou (2012) find a small, but significant, beauty premium for “pretty” females (2%); however, they do not find a beauty premium for men. Interestingly, the beauty premium disappears for high-skilled attractive women: Low-skilled attractive women are more likely to be called back than highskilled attractive women. On the other hand, Hamermesh and Biddle (1994) do find the existence of a beauty premium in the United States. We discuss the rationale for a beauty premium further in section “Willingness to Pay.”
Correspondence Studies in Other Settings Rental Markets Correspondence studies in the housing market have very much followed the same approach as those in the labor market. The main findings from the literature are summarized in Table 2. The rental market studies replicate, in methodology and basic results, those in the labor market. The researchers typically identify rental ads and send enquiries, manipulating the trait of interest. Discrimination against Arabic names is found in Sweden (Carlsson and Eriksson 2014; Ahmed and Hammarstedt 2008; Ahmed et al. 2010). Discrimination against blacks and other minority ethnicities in the USA is found in Ewens, Tomlin, and Wang (2014), Hanson and Hawley (2011), and Carpusor and Loges (2006). Discrimination against immigrants (particularly Muslims) is found in Italy (Baldini and Federici 2011) and Spain (Bosch et al. 2010). Discrimination against LGBT people is found in Ahmed and Hammarstedt (2009). Another popular variation, parallel to the labor market literature, has been to provide more information (e.g., job, etc.) about some of the applicants. Positive
170
M. Bertrand and E. Duflo
Table 2 Rental market papers Study Carlsson and Eriksson (2014) Trait: minority status (Arabic name) Ewens, Tomlin, and Wang (2014) Trait: race
Country Sweden
Inquiries 5827
Effect Swedish-to-Arabic (females): 1.37 Swedish-to-Arabic (males): 1.62
Theory No
The USA
14,237
White-to-black: 1.19
Bartoš, Bauer, Chytilová, and Matejka (2013) Trait: minority status (Roma or Asian name) Hanson and Hawley (2011) Trait: Race
Czech Republic and Germany
1800
Czech-to-minority: 1.27 (site available), 1.9 (pooled Asian and Roma names)
Consistent with statistical discrimination, inconsistent with taste-based discrimination Consistent with attention discrimination
The USA
9456
Consistent with statistical discrimination
Baldini and Federici (2011) Trait: immigrant status; language ability Ahmed, Andersson, and Hammarstedt (2010) Trait: minority status (Arabic name) Bosch, Carnero, and Farré (2010) Trait: immigrant status Ahmed and Hammarstedt (2009) Trait: sexual orientation Ahmed and Hammarstedt (2008) Trait: immigrant (race/ethnicity/ religion)
Italy
3676
White-to-African American: 1.12 (varied by neighborhood and unit type) Italian-to-East European: 1.24 Italian-to-Arab: 1.48
Sweden
1032
Swedish-to-Arab/ Muslim: 1.44 (no information), 1.24 (detailed information about the applicant)
No
Spain
1809
No
Sweden
408
Spanish-to-Moroccan: 1.44 (no information), 1.19 (with positive information) Straight-to-gay: 1.27
Sweden
1500
Swedish-to-Arab male: 2.17
No
No
No
(continued)
9
Field Experiments: Correspondence Studies
171
Table 2 (continued) Study Carpusor and Loges (2006) Trait: race/ ethnicity (Arab, AfricanAmerican)
Country The USA (Los Angeles County)
Inquiries 1115
Effect White-to-Arab: 1.35 White-to-black: 1.59, conditional on hearing back, 1.98 unconditional
Theory No
information (e.g., “I do not smoke and I work full time as an architect”) tends to reduce the callback ratios between white and the minority group, while negative information (“I am a smoker and I have a less than perfect credit score”) or small spelling mistakes in the email tend to increase it.
Retail The expansion of online platforms allows researchers to use the correspondence method to also study discrimination in retail markets. There are currently much fewer such studies, but the door is wide open for more to be performed. Zussman (2013) studies the mechanisms behind ethnic discrimination in the online market for used cars in Israel. This chapter uses an innovative, two-stage approach. First, about 8000 paired emails are sent to sellers of secondhand cars. An enquiry coming from somebody with a Jewish-sounding name was 22% more likely to receive a response than an enquiry emailed by an interested buyer with an Arabsounding name. Second, a follow-up phone survey was used to elicit sellers’ attitudes about minorities to tease out potential mechanisms for this effect. The researchers found that Jewish car sellers who strongly disagree with the statement “the Arabs in Israel are more likely to cheat than the Jews” do not discriminate against the Arab buyer, while others sellers do. That is, expectations about the quality of the transactions are correlated to the differential (average) treatment of Arabs. Pope and Sydnor (2011) report evidence from peer-to-peer lending sites. They find that loan listings with an attached picture of a black individual are 25–35% less likely to receive funding than those of white individuals with similar credit profiles. Academia Milkman, Akinola, and Chugh (2012) ran a field experiment set in academia with a sample of 6548 professors. Faculty members received e-mails from fictional prospective doctoral students seeking to schedule a meeting either that day or in 1 week; students’ names signaled their race (Caucasian, African-American, Hispanic, Indian, or Chinese) and gender. When the requests were to meet in 1 week, Caucasian males were granted access to faculty members 26% more often than were women and minorities; also, compared with women and minorities, Caucasian males received
172
M. Bertrand and E. Duflo
more and faster responses. However, these patterns were essentially eliminated when prospective students requested a meeting that same day. The authors argue that their finding of a temporal discrimination effect is consistent with the idea in psychology that subtle contextual shifts can alter patterns of race- and gender-based discrimination (a topic we return to in the last section of this chapter, section “Technological De-Biasing”).
Beyond the Re´sume´s With the rise of the Internet, employers can easily find more information online about a job applicant besides their résumé. A few recent studies enrich the correspondence methodology by allowing employers to search for more (and different) information than that which would typically be available in a résumé. Given the increasing popularity of online social networks, the contribution of Acquisti and Fong (2013) is particularly interesting. They employ the correspondence method by submitting applications to job postings and extend their experiments by creating either personal websites of social networking profiles for the fictitious applicants, which allow employers to gather additional information if they wish to. The additional information that can be gleaned online about the job applicants relates to their religion and sexual orientation. The question the paper is asking is whether extra information available online but not on the résumé leads to discrimination: Would applicants whose identity is not revealed in the application, but who appear to be Muslim (versus Christian) or gay (versus straight) on a popular social network, suffer unequal treatment? To do so, they created distinct online profiles: one profile on a professional network site, and another profile on a social network site where the emphasis is on sharing photographs or leisure-related comments, not job opportunities. The profile on the professional network site was identical across treatments (even the photograph was the same). The name used by researchers (selected after careful testing) was one not commonly associated with a particular race or religion. That is, the name of the “Muslim candidate” was non-Arabic, but the candidate’s religion could be inferred after some search on the social network site. Only the profile on the social network site contained clues (e.g., Christian versus Muslim or straight versus gay). The experiment finds that only a small fraction of employers use social media to conduct additional inquiry about job candidates.12 Given the limited search efforts by employers, the effects of group membership are generally small. The total effect of trait manipulation is not statistically significant: 12.6% of applicants who appeared to be Christian received callbacks, compared to 10.9% of candidates who
12
Measuring the exact number of visits to a social or professional networking profile is not possible for several reasons. However, using Google Adwords “Keyword Tool” statistics and Professional Network “Premiere” account statistics, the authors estimate that at most one-third of the employers tried to access the profile of the candidates.
9
Field Experiments: Correspondence Studies
173
appeared to be Muslim. About 10.6% of candidates who appeared to be straight males received callbacks, and the share of callbacks for seemingly gay males was nearly identical. The strength of this type of study is that researchers are able to study more naturally the impact of traits that traditionally are not revealed on a résumé. While some correspondence tests have tried to signal religious affiliation or sexual orientation through “extracurricular activities” described on CVs, this type of disclosure might reveal more than religion or sexual identity: The employers might be reacting to someone’s activism regarding their religion or sexuality, not their religion or sexuality per se. While Acquisti and Fong (2013) focus on the impact of gay status for males and religion, their methods could be used to study the effect of other interesting and until now mostly unexplored characteristics. For example, would the size of a candidate’s network have an effect? Would employers infer that a “popular” candidate has valuable social skills? Would attractive-seeming candidates receive more callbacks, or would attempts to “choreograph” one’s online presence be viewed as an undesirable trait? Would candidates who reveal their family status be treated differently than candidates who are more private? Clearly, online field experiments offer a rich landscape for studying “what employers want.”
Limitations of Correspondence Studies While correspondence studies address some key weaknesses of the audit methodology, they share other weaknesses with audit studies and have some unique limitations of their own. Both types of studies can only inform us about the average differences in hiring behavior. But we generally think that applicants care about the marginal response. Real job seekers are likely to adjust their behavior during the search process in a strategic manner: In other words, they will not apply for positions in a random fashion. So, while informative about discrimination on average in a given setting, correspondence and audit studies are not informative about discrimination at the margin, when real job seekers have fully optimized their job search strategy to the realities of the workforce. This is related to a criticism raised by Heckman and Siegelman in their contribution to Clear and Convincing Evidence: Measurement of Discrimination in America when they challenge the use of newspaper advertisements in audit studies, referring to previous findings that most jobs are found through direct contact with a firm, or via informal channels like family and friends: [c]ollege students masqueraded as blue collar workers seeking entry level jobs. Apart from the ethical issues involved, this raises the potentially important problem that the Urban Institute actors may not experience what actually occurs in the these labor markets among real participants. (Fix and Struyk 1993)
174
M. Bertrand and E. Duflo
Another drawback of field studies (both audit and correspondence) is that fictitious applicants typically only apply to entry-level jobs. There are a few exceptions, and some of the studies we describe above use applications to skilled and experienced positions. But the bottom line is that many jobs are never advertised, and the extent of discrimination in the workplace overall may be quite different from the discrimination that is measured at the entry point in the labor market. Yet another limitation of field studies (both audit and correspondence) is that the outcome variables that can be studied are typically very coarse. In fact, in this regard, the correspondence studies are inferior to the audit studies. Most of the time, interview invitations or rental offers (“call-back rates”) are the only outcomes captured by field experiments (one exception being Doleac and Stein (2013), who were able to track transactions – sales of iPods through local online markets – all the way to completion). Obviously, because there is no real applicant, the correspondence study methodology cannot be taken to the interview stage, job-offer stage, or wage-setting stage – or to the stage at which people sign a lease on an apartment. Theoretically, all of this can be measured when auditors are used. However, even audit studies do not allow one to track other important outcomes, such as work hours, working conditions, or promotions. The binary outcome in the typical correspondence studies (callback or not) raises important issues about how to conduct some of the analysis. What should be inferred about discrimination for the employers that do not call back any of the fictitious applicants? Is that evidence of “symmetric treatment”? Riach and Rich (2002) argue that if both the majority and minority candidate are rejected, that does not constitute evidence of equal treatment. Only with more continuous outcome variables – ones that typically are not available to the researcher, such as the ranking of the job candidates by the employer – would it be possible to resolve this tension. Both correspondence and audit studies have also raised ethical concerns. Employers’ time is bound to be a scarce resource, and researchers who carry out these studies are using it without the involved parties’ consent. A positive take on this ethical issue is List (2009) who argues that, “[w]hen the research makes participants better off, benefits society, and confers anonymity and just treatment to all subjects, the lack of informed consent seems defensible.” However, many people outside the scientific community would probably disagree. (In fact, List (2009) refers to experiments where subjects are compensated; in the case of correspondence tests, we did not come across experiments where employers were actually compensated for their time.)13
13
The method of correspondence studies has also been taken to the dating market (e.g., Ong and Wang (2015)). We do not review these contributions here because it is a bit difficult to talk about discrimination when referring to the choice of whom to date, but the ethical dilemma of putting fake applications on a dating website also seems particularly acute. As a conceptual side, it is also not at all clear that one needs to create a fictitious profile on dating websites, as it is already possible for the researchers to observe exactly the same information that the user has when making a decision. There is thus no “unobserved” variable biasing the analysis and no information to be gained from fictitious
9
Field Experiments: Correspondence Studies
175
Another underappreciated ethical issue is that when the “applicant” declines an offer, things other than the anticipated consumption of the employer’s attention can occur. The employer may “learn” (become convinced) that applicants with the attributes similar to those of the fictitious candidate are unlikely to accept offers. If this really happens, it is possible that some real job applicants will be treated differently (possibly less favorably) due to prior communication with the researcher pretending to be a job candidate. But it also possible that after observing a rejection or two from fictitious candidates, an employer may end up having the impression that the market is tighter than they thought; screening could then become less intense, which might be beneficial for real jobless candidates (but potentially detrimental for the employers).14 A subtler criticism of the correspondence and audit methods by Heckman and Siegelman (1993) has been recently revisited by Neumark (2012). Heckman and Siegelman (1993) show that a troubling result emerges in audit or correspondence studies because the outcome of interest is not linear in productivity (as it might be for a wage offer), but instead is nonlinear. That is, we think that in the hiring process firms evaluate a job applicant’s productivity relative to a standard, and offer the applicant a job (or a callback) if the standard is met. This nonlinear relationship can raise issues for any inferences of discrimination based on callbacks if employers believe that blacks and whites differ in the variance of their unobserved productivity. Consider for example the case where employers believe that the variance of unobserved productivity is higher for whites than for blacks. The correspondence and audit methods make black and white applicants equal on observable productivity characteristic X1. However, no information is conveyed on a second, unobservable productivity-related characteristic, X2. Because an employer will offer a job interview only if it perceives or expects the sum β1X1 + X2 to be sufficiently high, when X1 is set at a low level the employer has to believe that X2 is high (or likely to be high) in order to offer an interview. Even though the employer does not observe X2, if the employer knows that the variance of X2 is higher for whites, the employer correctly concludes that whites are more likely than blacks to have a sufficiently high sum of β1X1 + X2, by virtue of the simple fact that fewer blacks have very high values of X2. Employers will therefore be less likely to offer jobs to blacks than to whites, even though the observed average of X1 is the same for blacks and whites, as is the unobserved average of X2. The opposite holds if X1 is set at a high value: In this case, the employer only needs to avoid very low values of X2, which will be more common for the higher-variance whites. In other words, Heckman and Siegelman (1993) show that, even when there are equal group averages of both observed and unobserved variables, an audit or correspondence study can generate biased estimates, with spurious evidence of discrimination in either direction, or spurious evidence of its absence. résumés. The exercise can be performed with observational data (See Fisman et al. (2008), Hitsch et al. (2010), and Banerjee et al. (2009)). This makes the ethical concern particularly salient. 14 This particular issue can be at least partially addressed by debriefing employers ex post about them having been part of a research study.
176
M. Bertrand and E. Duflo
Building constructively on this criticism, Neumark (2012) shows that if a correspondence study includes variation in observable measures of applicants’ quality that affect hiring outcomes, an unbiased estimate of discrimination can be recovered even when there are group differences in the variances of the unobservables. Neumark explains how his method can be easily implemented in any future correspondence study. All that is needed is for the résumés or applicants to include some variation in characteristics that affect the probability of being hired.15 Finally, it is remarkable that after literally dozens of correspondence studies, there has been only limited refinement of the methodology to help discriminate between different theories of the differential treatment that is being consistently observed. Employers must try their best to infer future productivity of a candidate based on limited information. That is, applicants who belong to different groups may experience different treatment even if discrimination as understood by Becker (differential treatment is motivated by prejudice) is absent and only statistical discrimination is at play. Attributes beyond those intended by the researcher may be inferred by the recipient. For example, Fryer and Levitt (2004) suggest that black names may “provide a useful signal to employers about labor market productivity after controlling for information on the résumé.” This is clearly true for age, as we noted, but this may also be true for race if the choice of a black name is a political statement by the parent, accompanied by a different attitude toward schooling and obedience. More broadly, as we already mentioned several times, even if in general employers do not see a particular identity as a sign of lower productivity (or want to discriminate based on it), they may infer something from the fact that the person is wearing it on their sleeves. After all, there was no difference in callback rate according to either religion or sexuality when the information was available to the employer online, but not directly reported in the résumé (Acquisti and Fong 2013). The only approach that has been used repeatedly by researchers to try to separate statistical from taste-based discrimination has been to compare differential gaps in outcomes between pairs of minority and nonminority applicants with weaker or stronger productivity attributes on their résumés or applications. As more productivity-relevant information is included on the résumé, average differences in unobservable characteristics between the minority and nonminority applicants are reduced, and statistical discrimination should also be reduced; however, it is clear that this remains a very indirect way to try to isolate taste-based discrimination among employers or landlords. In this regard, a recent paper that breaks the mold of the typical correspondence study and deserves particular attention is Bartoš et al. (2013). This paper is
15
The method rests on three types of assumptions. First, it is based on an assumed binary threshold model of hiring that asks whether the perceived productivity of a worker exceeds a standard. Second, it imposes a parametric assumption about the distribution of unobservables. Lastly, it relies on an additional identifying assumption that some applicant characteristics affect the perceived productivity of workers, and hence hiring, and that the effects of these characteristics on perceived productivity do not vary with group membership.
9
Field Experiments: Correspondence Studies
177
remarkable in its ability to push the correspondence study methodology forward, think beyond the pure callback data, and refine our theories of discrimination. The paper links two important ideas: Attention is a scarce resource, and lack of information about individual candidates drives discrimination in selection decisions – or, in other words, statistical discrimination is an important factor in selection decisions. While the existing models of statistical discrimination implicitly assume that individuals are fully attentive to available information, the paper develops and tests a model in which knowledge of minority status impacts the level of attention to information about an individual and how the resulting asymmetry in acquired information across groups – denoted “attention discrimination” – can lead to discrimination. In particular, the authors argue that in markets such as the labor market where only a small share of applicants is considered above the bar for selection, negative stereotypes are predicted to lower attention. On the other hand, the effect is opposite in markets where most applicants are selected, such as the rental housing market. Bartoš et al. (2013) test for such “attention discrimination” in two field experiments: one in the labor market and one in the rental market, where they can monitor the decision-maker’s information acquisition about applicants through visits to hyperlinks containing résumés and personal websites (respectively). They created personal websites for fictitious applicants and submitted rental applications in the Czech Republic, and job applications in both Germany and the Czech Republic. The advantage of using hyperlinks to résumés and personal sites is that the researchers were able to track the exact number of visitors to the personal profile, and therefore, the share of landlords and employers who allocated additional attention to an applicant. Hence, the study was able to assess whether a minority-sounding name (1) leads to differential callbacks and (2) causes less or more search. Like the prior literature, Bartoš et al. (2013) find evidence of discrimination against minority applicants in both the housing and labor markets. Most interesting, though, are their findings regarding attention allocation. In the labor markets in both Germany and the Czech Republic, employers put more effort in opening and reading résumés of majority compared to minority candidates. In contrast, in the rental housing market, landlords acquire more information about minority compared to majority candidates through their personal sites. The findings can best be explained by a model where attention is endogenously determined by the type of the market. When the choosing entity needs to select “top candidates,” it will allocate attention to candidates belonging to the group that, according to its priors, is stronger. In markets where most candidates are accepted, some kind of a threshold rule might be used, and the choosing entity will want to eliminate the weakest candidates. In that case (e.g., a housing market), more attention would optimally be allocated to members of the group that is a priori viewed less favorably. The model implies persistence of discrimination in selection decisions, even if information about individuals is available and there are no differences in preferences. The model also implies lower returns to employment qualifications for negatively stereotyped groups (as their credentials are less likely to
178
M. Bertrand and E. Duflo
be reviewed). From a policy perspective, the model and results of this paper also highlight the crucial role of the timing of when a group attribute is revealed. Bartoš et al. (2013) represents a great example of how the résumé study infrastructure can be pushed forward to deliver deeper learning, cleaner links to specific theories of why differential treatment is taking place, and suggestions about policies that might be most effective to address it. More efforts along these lines would help revitalize this literature. We now turn to other approaches to measuring discrimination, often more “labbased” and more closely tied to a particular model of the root of discrimination.
Implicit Association Tests The Implicit Association Test (IAT) is a computer-based test that was first introduced by Greenwald, McGhee, and Schwartz (1998). Developed by social psychologists Greenwald, Nosek, and Banaji, as well as other collaborators, the IAT provides a method to indirectly measure the strength of association between two concepts. This test relies on the idea that the easier a mental task is, the quicker it can be performed. When completing an IAT, a subject is asked to classify, as rapidly as possible, concepts or objects into one of four categories with only two responses (left or right). The logic of the IAT is that it will be easier to perform the task when objects that should get the same answer (left or right) somehow “go together.”16 The typical IAT consists of seven “phases,” including practice phases to acquaint the subject with the stimuli materials and rules. Consider for example an IAT designed to assess association strengths between categories of black and white and attributes of good and bad. The practice phases are used with the materials and sorting rules. In the first, subjects would only be presented with faces as stimuli and be asked to assign white faces to one side and black faces to the other; in the second, subjects would only be presented with words as stimuli and be asked to assign pleasant words to one side and unpleasant words to the other. In the test phases, subjects are asked to simultaneously sort through stimuli representing the four concepts (black, white, good, and bad) but with again only two responses (left side or right side). In two of the test phases (the “stereotypical” test phases), items representing white and good (e.g., white faces and words such as wonderful) need to be placed on one side of the screen, and items representing the concepts black and bad (e.g., black faces and words such as horrible) on the other. In the other two test phases (the “non-stereotypical” phases), items representing the concepts of black and good need be placed on one side of the screen, and items representing the concepts of white and bad on the other. The extent to which an individual dislikes
16
See Lane et al. (2007) for an excellent introduction to IATs.
9
Field Experiments: Correspondence Studies
179
black faces (in this case) is then measured by the difference in milliseconds in response time between the stereotypical phases and the nonstereotypical phases.17 Two broad kinds of IAT are pertinent to discrimination: If attitudes or overall preferences are the issue, the category (e.g., black/white) is associated with words that represent good/bad (as in the example we just gave). Alternatively, one may be interested in the association between a category (e.g., male/female) and a particular trait or attribute (e.g., career/family).18 The first kind is called an attitude IAT, and the second a stereotype or belief IAT. Other types include self-esteem IATs (e.g., categories are self and other, and words are either positive or negative). Since the publication of the original IAT, there have been hundreds of IAT studies, many of which try to capture attitudes that could give rise to discrimination (against black people, Muslims, women, etc.), or phenomena more akin to statistical discrimination (women and math, women and career, women and politics, etc.). The IAT has been extremely influential both within and outside academic psychology. Greenwald, McGhee, and Schwartz’s original 1998 article introducing the IAT has 6689 citations in Google Scholar, as of August 2015. The findings of IAT research on discrimination have been cited to propose changing the law, educating judges and students, etc. IATs are used increasingly as a convenient tool to measure whether attitudes respond to any particular intervention, since they can be conducted remotely, with large samples of online participants to experiments. As such, they are often used as endpoints in psychology experiments, as experimentation moves to online platforms.19 There are a number of meta-analyses, review articles, and critical papers on the use of IATs. It is not in the scope of this chapter to review all of this literature; however, the key question that is raised is what the IAT actually picks up and, relatedly, whether it is effectively associated with other predictors of discriminatory behavior, and discriminatory behavior itself. Some individual studies show promising links. For example, implicit bias predicts a more negative judgment of ambiguous actions by a black target (Rudman and Lee 2002), as well as more negative nonverbal “microbehavior” (less speaking time, less smiling, etc.) during an interaction with a black subject (McConnell and Leibold 2001). This is important, as these microbehaviors are often posited to be the channel through which implicit bias would translate into different behavior, even among people who do not report explicit discrimination. Some studies have also shown some mechanisms for those effects, e.g., showing that participants who exhibited greater implicit distaste of black people were more likely to detect aggression in a black (but not white) face (Hugenberg and
17
In practice, of course, a number of choices must be made about how to use the data, and this is reviewed in Greenwald, Banaji, and Nosek (2003). 18 For example, Nosek, Banaji, and Greenwald (2002) find stereotypical associations connecting male terms with traits related to science and a career, whereas female terms are found to be associated with liberal arts and family. 19 For examples of IATs used as end points, see Lai et al. (2014).
180
M. Bertrand and E. Duflo
Bodenhausen 2004). Only a few studies have investigated whether these differences in implicit attitudes are associated with different behaviors in the field. In Atlanta and Boston, doctors with stronger antiblack implicit attitudes were less likely to prescribe thrombolysis for myocardial infarction to African-American patients, compared to white patients (Green et al. 2007). Rooth (2010) tried to relate the behavior of recruiters in a correspondence study in Sweden (focusing on Arab-Muslim versus Christian) to recruiter-level measures of implicit discrimination they collected later. Unfortunately, they were only able to interview 26% of the recruiters they were targeting, but among those, they did find a correlation between implicit distaste of Arabs as measured in an IAT test and the tendency to not call back a resume with an Arab-Muslim name on it. An initial meta-analysis, conducted on 122 research reports, found that the IAT does seem to be capturing something about attitudes, perhaps more accurately than self-reports (Greenwald et al. 2009). They show that there is a strong correlation between implicit and more standard explicit measures. Moreover, the IAT appears to be a better predictor of actual behavior than explicit reports, particularly for sensitive subjects such as racial preferences (for which they have 32 samples with IAT measure, explicit measure, and questions about behavior). However, a more recent meta-analysis by Oswald et al. (2013) questions these initial findings. Using a larger sample (which includes newer studies as well as some studies that were omitted from the earlier meta-analysis), and a slightly different aggregation method, they find much lower correlation of the IAT with various measures of discrimination than had been initially found in the 2009 meta-analysis. Explicit measures perform equally poorly, to be sure, but not much worse. Beyond this debate (which is probably the core one to be had), IATs have been subject to a number of criticisms and questions, mainly regarding their interpretation. First, to the extent that they differ from explicit attitudes, do they reflect something “deeper” about the individuals, and are they more “true” than the selfdescription in any sense, or simply another type of attitude? Interestingly, the metaanalysis by Oswald et al. (2013) shows that there is in fact a strong correlation between different brain activities when seeing black and white faces and the IAT. This suggests that the IAT does reflect something fundamental about psychological processes. But our behavior is mediated by the social environments, exactly as our answer to a question on prejudice is mediated by this environment. So do IATs really identify prejudice or just some raw psychological “material” that we then transform? What does it mean for someone to feel that they are not prejudiced against blacks but have their IAT showing automatic white preferences (Arkes and Tetlock 2004)? On this last question, Banaji, Nosek, and Greenwald (2004) argue that conscious unbiased attitudes cannot be relied upon in all circumstances, and that IATs may capture unconscious attitudes that may be more relevant in explaining behaviors in other circumstances. Hence, they reject the idea that “if prejudice is not explicitly spoken, it cannot reflect a prejudicial feeling” (Banaji et al. 2004). In this respect, though, the low correlation between the IAT and microbehavior is a bit troubling, as this theory would suggest that the unconscious bias translates into actual acts of discrimination via unconscious behavior.
9
Field Experiments: Correspondence Studies
181
Also, do IATs measure prevalent culture or individual attitudes? For example, if a person identifies women with family more than with career, is she making a value judgment or stating, in a sense, a fact of life? Whatever the resolution of these debates, the measured implicit attitudes vary considerably across people, and the robust correlations – between implicit and explicit attitudes, between the different IATs in similar domains – do seem to indicate IATs capture some signal about the individual. This does not mean that the IAT can be considered a reliable measure of the attitude of any particular individual (at best, it measures attitudes with considerable noise). However, it does mean that the IAT may be a good measurement tool for the propensity for groups to discriminate toward each other. In this context, whatever the predictive value of the IAT for behavior, the extent to which it is affected by a particular manipulation is of interest. As economists, we may be more interested in the extent to which attitudes can be influenced (by experiences, the environment, or specific interventions), than in their pure measurement at a point in time. Using IATs as an outcome variable also helps sidestepping the question of whether they represent any deep truth about anybody: While the signal may be noisy, to the extent that there is signal, this may be a useful measurement tool. As noted, after more than a decade of using the IAT mainly as a descriptive tool, studies in psychology started using them as outcomes. For example, Lai et al. (2014) set up a research contest on de-biasing, where teams are given a budget of 5 min to interact with participants, and the outcome is the scores on a black-white attitude IAT. In recent years, economists have also started using IATs as dependent variables. For example, Beaman et al. (2009) design and implement two IATs in West Bengal, India, to measure preferences toward female leaders, and stereotypical association of women with domestic rather than political activities. They then examine the impact of exposure to female leaders on these two measures (we will discuss the results below in section “Goldberg Paradigm Experiments”). Lane et al. (2007) provide detailed and helpful instructions on how to build an IAT. The software that is needed to construct and analyze the test (millisecond software) is available for purchase. IATs can be designed with only verbal or image stimuli for subjects who are not literate (this is what Beaman et al. (2009) use), and although they are more difficult in populations that have had no experience with computers and for older participants, they can be a very useful tool. As studies increasingly use electronic data collection methods (on tablets or notebooks), the extra cost of adding an IAT diminishes. Of course, the debate in psychology does not suggest that the IAT should be considered a “magic bullet,” suitable to replace any other measure of discrimination. In particular, it is probably not a substitute for good measures of actual behavior in policy interventions. Nevertheless, it can be an extremely useful intermediate variable, to understand the mechanisms beyond a result (in Beaman et al. (2009), the final end point of interest is actual voting), or potentially, if collected beforehand, as a covariate of interest. For example, in Glover, Pallais, and Pariente (2015), the IAT
182
M. Bertrand and E. Duflo
is used as a proxy measure of latent employer discrimination (see further details in section “Endogenous Responses to Bias”).
Goldberg Paradigm Experiments Goldberg Paradigm experiments are laboratory versions of audit or correspondence studies. They are named after a 1968 experiment by Goldberg where students graded written essays, which were identical except for the male or female name of their author (Goldberg 1968). This initial experiment demonstrated a bias: Female got lower grades unless the essay was on a feminine topic. Since then, a large literature in psychology has used the Goldberg Paradigm to identify discrimination against different groups, and in particular in the resistance to female leaders.20 In the typical lab experiment, a group of subjects is asked to review a vignette, describing the behavior of a female or male manager (for example), or witness a confederate (male or female) simulating a leadership situation. The participants are then asked to evaluate the leader’s competence, or to say whether they would have liked to have them as leader for a task they may collectively perform. Reviewing a large number of such studies, Eagly, Makhijani, and Klonsky (1992) do not find that, on average, female leaders are evaluated significantly more negatively than male leaders. However, there are some circumstances where they do find that female leaders are evaluated more negatively, for example, when the leadership was carried out in a masculine style (in particular when the leader was projected to be authoritative). This supports Eagly’s hypothesis of “role congruence”: What people dislike is when women behave in a nonfeminine way. Since strong leaders must be assertive, but women must be demure, it makes it difficult for women to be appreciated as strong leaders. The fact that the circumstances are artificial, and answers have minimal stake associated with them, makes those experiments less relevant, on their own, than field-based correspondence tests. But one advantage of the Goldberg-style experiments is that they can be easily, and finely, manipulated, which makes them good outcome measures in field research (or field experiments). They can also be easily added to a standard survey instrument. For example, Beaman et al. (2009) seek to find out how discrimination against female leaders is affected by prior exposure. They administer two Goldberg-style experiments. In one, they ask the participants to listen to a speech by a political leader, which is read either by a female or a male actor (note that it is important that there are several male and female actors). In the second one, they discuss vignette where women or men leaders make decisions that are either promale (investment in irrigation) or profemale (investment in drinking water). Each individual receives a randomly selected version of the speech and vignette. The randomization is stratified by village, and hence by prior exposure to
20 See Eagly, Makhijani, and Klonsky (1992) for a review and meta-analysis of the literature on such resistance to female leaders.
9
Field Experiments: Correspondence Studies
183
a female leader (due to a policy of gender reservation). While this does not tell us the extent to which any single person discriminates, one can learn whether, on average, exposure to a female leader affects the extent to which individuals give lower grades to women in response to the same speech or vignette. Beaman et al. (2009) find that both men and women, but men more than women, tend to discriminate against female leaders (additional results from this study are discussed further in section “Minority Leaders and the Attitude of the Majority”).
List Randomization Like correspondence tests or Goldberg experiments, list randomization (also known as item count technique, unmatched count, or list response) does not provide a measure of individual bias but can provide an estimate of the extent of discrimination in a population. They are a way of eliciting accurate answers to questions of discrimination in the presence of social desirability bias. The idea is to present the subjects with a list of N statements which are generally noncontroversial, but could be true or false (e.g., I had coffee at breakfast; I like popcorn).21 Then, a randomly selected group of people is asked a potentially controversial statement (e.g., “I would be upset if an African-American family moved next door”) on top of the N noncontroversial statements. The subject only states the number of statements with which they agree. Comparing the fraction of yes among those who got N and those who got N þ 1 statements gives a good measure of discrimination. And clearly no one (including the interviewer) will know how a given subject answered the controversial statement. Unlike the IAT, this method will not reveal biases that are unconscious or biases that the subject wants to deny even to themselves, but it will prevent the results from being affected by social desirability bias. Early applications of list randomization to measure discrimination are Kuklinski, Cobb, and Gilens (1997a) and Kuklinski et al. (1997b). Both studies found considerable racial prejudice in the American South using list randomization techniques (though not in the North). Furthermore, they found higher level of measured discrimination using this method than using direct elicitation methods, for example, respondents are more likely to disagree with a statement such as “I am comfortable with a black family moving next door” when asked via list randomization than with a traditional survey. Likewise, Coffman, Coffman, and Ericson (2013) show that stated discrimination against gay populations is much lower in response to a direct question in the control group than when it is elicited through the list randomization method. For example, respondents were 67% more likely to express disapproval of an openly gay manager at work when the question is part of a list than when the question is asked directly.
21
As we explain below, there is a tension in the choice of those questions: For maximum precision, there should be behavior that almost everyone says yes or no to, but then they do not give any cover to the subject.
184
M. Bertrand and E. Duflo
A few papers have used the method to elicit attitudes toward presidential candidates. Kane, Craig, and Wald (2004) find no discrimination against a Jewish presidential candidate (Joe Lieberman). Martinez and Craig (2010) find that few whites in Florida seemed distressed by the possibility of having a black president. However, list randomization revealed much more opposition toward the idea of a female president than that which was reported in opinion polls (Streb et al. 2008). As noted above, several studies suggest that the randomized list technique yields different answers than direct elicitation. In a meta-analysis across 48 comparisons of direct report and list randomization, Holbrook and Krosnick (2010) found that 63% of the estimates for socially undesirable behaviors were significantly larger when elicited through list randomization. On the other hand, responses on nonsensitive behavior tend to be more similar (Tsuchiya et al. 2007). The list randomization method is, however, not without issues. As we alluded to earlier, there is a fundamental tension between precision (which would require having statements to which everybody responds yes or no) and providing “cover” to the subject (which would require the opposite). The implication is that the results from list randomization methods are often quite imprecise. Gosen (2014) has also shown results tend to systematically depend on how many noncontroversial statements are included in the list, although the opposite was found in Tsuchiya, Hirai, and Ono (2007). In summary, list randomization could be a promising method to measure discrimination as it is less subject to social desirability bias, but since few economists have used it,22 more work needs to be done to ascertain its usefulness in the field. In comparison to other indirect methods, list randomization is often more simple to administer (both for surveyors and respondents) but risks having low power (Droitcour et al. 1991). It would be interesting to see more research comparing measures of discrimination obtained through list randomization compared to an IAT or Goldberg style experiment. It would also be interesting to compare how noisy these different measures are. The fact that the list randomization method can only provide an aggregate (and not individual) measure of discrimination complicates its use as an outcome variable (say, for a randomized experiment), but no more than any of the other methods we have already discussed that also only give group-level outcomes, such as the Goldberg-style experiments.
Willingness to Pay A key prediction of Becker’s model of taste-based discrimination is that people should be willing to pay to interact with people of their own group. Somewhat
22
See Karlan and Zinman (2012) for an example of an application in economics.
9
Field Experiments: Correspondence Studies
185
surprisingly this prediction has not given rise to a large literature trying to evaluate the willingness to pay to discriminate. As we noted, the correspondence and audit tests tend to be based on a binary measure (interview or not, hire or not). Until recently, the body of work that came closest to measuring such willingness to pay was a literature on the “beauty premium” motivated by Hamermesh and Biddle’s (1994) finding that workers with better-than-average looks earn 10–15% higher wages. Analogous to the black-wage race gap, the beauty premium could be due to the fact that more beautiful workers are more productive, say because consumers prefer to interact with beautiful people (Biddle and Hamermesh 1998; Pfann et al. 2000), or because beautiful people are more confident. Or employers may be wrong in their belief. Mobius and Rosenblat (2006) set up a laboratory experiment where undergraduates and graduates from Tucuman, Argentina, were randomly assigned into groups of “employers” and “workers.” In the experiment, “employers” had to hire “workers” to perform a maze-solving task. After a practice test (which was recorded and became the digital “resume” of the worker) and a question where the workers estimated the number of mazes they could solve in 15 min, each worker was matched to five employers, who saw either (1) just the resume, (2) the resume and a photo, (3) the resume and a phone interview, or (4) the resume plus an interview, plus the photograph.23 The employers in turn saw five workers, and for each of them decided how many mazes they thought the worker could solve. This estimate contributed to the employer’s own payment. It also entered into the calculation of the actual wage of most of their workers. Mobius and Rosenblat show that productivity at the task is not affected by beauty (as evaluated by 50 high school students on the basis of the photograph), although worker confidence is. A rise of one standard deviation in beauty increases confidence by 13–16%. However, employers are willing to pay more employees who are considered to be more beautiful: In all the treatments where they can see beauty, employers are willing to pay workers more. The premium ranges between 12% and 17% depending on the treatment. Decomposing the beauty premium by comparing treatments, the authors estimate that 15% is due to the confidence channel and 40% each through the visual and oral stereotype channels (the fact that beauty still affects wages when the employer does not see the employee but talks to her on the phone indicates that beauty is correlated with speaking skill, perhaps another feature of the beauty channel). Interestingly, employer’s estimated productivity is not affected by whether or not they know that it will actually contribute to the worker’s wage. This suggests that there is little pure taste-based discrimination in this lab experiment.
The wages of “workers” were affected by the difference between their estimation of the number of mazes and what they actually competed; therefore, they were incentivized to tell the truth.
23
186
M. Bertrand and E. Duflo
Employers give a premium to beautiful people because they believe (wrongly) that they will be more productive. There are a number of limitations of this experiment, not least of all, from the point of view of this chapter that it is a lab experiment. It is also limited to a one-shot interaction at the hiring stage. Nevertheless, it sets an interesting template for what a field experiment leveraging this methodology might look like and in particular does an excellent job laying out the various pieces that are needed to establish discrimination and understand the mechanism behind it. One paper which has recently followed in Mobius and Rosenblat’s footstep is Rao (2013), which seeks to measure the extent to which well-off kids in India discriminate against poorer kids (in order, as we will discuss in more detail in section “Intergroup Contact,” to estimate the extent to which any such discrimination is affected by forced exposure to poorer kids through an affirmative action program in education). To do so, Rao sets up an Ingenious field experiment, based around team selection for a relay race. First, students from a rich private school and a poor public school, who were all present at a sporting event to support their classmates, were randomized in different sessions with different prizes for winning the race (from Rs.50 to Rs.500, which are very high stakes). After mixing for 15 min, they watched a series of one-on-one sprints (most of them pitting a poor student against a rich one), and then each was asked to indicate on a worksheet which of the two he wanted as teammate for the relay race. After these choices were revealed, one of the choices was picked, the teams were formed, and the relay race was run. To make sure that there was a “cost” to picking out a poorer student (if students did not like them), students had to spend 2 h socializing with their teammate, which was announced prior to team selection. This experiment has a number of clever features. It presents children with a real choice, and by varying the stakes, it makes it clear how much (on average) students are willing to sacrifice to avoid interacting with a poor student. The sprint phase entirely and unambiguously reveals ability, so the setup is targeted to pick up pure taste-based discrimination (e.g., dislike of hanging out with a poor teammate). The results show that there is substantial taste-based discrimination in this context: In 19% of the cases where the poor students is the fastest, rich students prefer to pick the rich student as a teammate anyway.24 Discrimination does decline as the stakes increase: discrimination falls from 35% with the lowest stake, to 27% with the intermediate stake, and 5% in the highest stake. Fitting a structural model to the data, Rao estimates that, for students without prior exposure to poor classmates, the distaste of interacting with a poor student is worth Rs.37. That is, a student is willing to give up to Rs.37 in expected prize money to hang out with a rich student rather than a poor one.
24
In contrast, if a rich student is the fastest, he is picked 97% of the time, and among two students of the same background, the fastest is picked 98% of the time.
9
Field Experiments: Correspondence Studies
187
Consequences of Discrimination Self-Expectancy Effects Stereotype Threat and Underperformance Models of statistical discrimination explain the differential treatment of disadvantaged groups in, say, hiring decisions, due to employers’ inability to perfectly predict a given worker’s future productivity and hence their rational decision to assign some weight to the average productivity of the worker’s racial group. For example, African-Americans as a whole are categorized as less productive than whites, and employers take this supposed average difference in productivity into account when deciding whether or not to hire any African-American job candidates, given their inability to precisely predict each specific candidate’s future productivity. While discrimination emerges under the logic above as a consequence of average differences in productivity across groups, research in social psychology has provided convincing evidence for the reverse causal channel. In particular, the simple process of categorizing or “stereotyping” some groups as less productive appears to cause these groups to be less productive. This research suggests that individuals from some groups may suffer negative performance outcomes (such as lower test scores or less engagement with academics) because of the burden of the “stereotype,” or “stereotype threat” (Steele and Aronson 1995).25 The key conjecture is that the threat of being viewed through the lens of a negative stereotype can create an anxiety that disrupts cognitive performance. In a seminal study, Steele and Aronson (1995) demonstrated in a lab setting that inducing stereotype threat – by asking test takers to indicate their race before the test – significantly undermines African-Americans’ performance on intellectual tasks. They also showed that reducing stereotype threat – by convincing test takers that the test was not being used to measure their abilities – can significantly improve AfricanAmericans’ performance, dramatically reducing the racial gap. Numerous lab studies have since replicated the effects of stereotype threat both with respect to social identities other than race (e.g., gender, income class, etc.) and with respect to mediating outcomes (such as blood pressure, heart-rate variability, performance expectations, effort, etc.).26 Defined by Steele and Aronson (1995) as a “risk of confirming, as self-characteristic, a negative stereotype about one’s group.” 26 While most of these lab studies have been conducted by social psychologists, a few have been performed by economists. For example, Dee (2009) ran a lab experiment with student-athletes and nonathlete students at Swarthmore College, randomly assigning some of them to a treatment that primed their awareness of a stereotyped identity (i.e., student-athlete). He finds that the treatment reduced the test-score performance of athletes relative to nonathletes by 14%. Also, Hoff and Pandey (2006) present evidence from a caste priming experiment in Uttar Pradesh, India. Among 321 high-caste and 321 low-caste junior high school male student volunteers, there were no caste differences in performance in an incentivized maze-solving task when caste was not publicly revealed, but making caste salient created a large and robust caste gap in performance. However, the mechanisms for the underperformance in this case seem quite different from the hypothesized 25
188
M. Bertrand and E. Duflo
The rest of the social psychology research on the stereotype threat has also focused on documenting methods that can undo or undermine the threat of the stereotype. One line of research has addressed the underlying message of the stereotype – that stereotyped individuals are inherently limited because of their group membership. Thus, if participants can be convinced that intelligence is not a fixed trait, but a malleable quality that can be increased with hard work and effort, they may be less prone to stereotyping. Levy, Stroessner, and Dweck (1998) present evidence consistent with this idea. Descriptively, they find that people holding an entity theory of intelligence (i.e., intelligence is a fixed trait) made more stereotypical trait judgments of ethnic and occupational groups than those who believed that intelligence is malleable. Moreover, in a small lab experiment, they found that manipulating implicit theories affected level of stereotyping, at least temporarily. In the experiment, 155 introductory to psychology students were randomly assigned to a fake “scientific” article that either presented evidence for an entity (fixed) or an incremental (malleable) view of personality. After reading the article, they were presented with questions in which they had to rate the extent to which a series of 15 traits accurately describe certain occupational groups (teachers, doctors, lawyers, and politicians) and ethnic groups (African-Americans, Asians, and Latinos). To try to reduce the likelihood of participants recognizing the link between the article and the questions, the researchers told participants that the questions were for a separate study, and that they would be asked questions on the content of the article later. The experiment found a small but significant effect: Those who read the article that argued for fixed personality were less likely to believe traits can change, and more likely to rate stereotypical traits as highly descriptive of their respective groups. Similarly, in a lab experiment, Aronson, Fried, and Good (2002) assign white and black students to one of three conditions to assess the impact of an intervention designed to reduce stereotype threat. In two conditions, students were asked to write a letter of encouragement to a younger student who was experiencing academic struggles. In one of these conditions, students were prompted to endorse a view of intelligence as malleable, “like a muscle” that can grow with work and effort. In the second condition, students endorsed the view that there exist different types of intelligence. The third condition served as a control condition in which students were not asked to compose a letter. Several days after the intervention, all students were asked to indicate their identification with and enjoyment of academics. Results showed that black students in particular were more likely to report enjoying and valuing education if they had written a letter endorsing malleable intelligence. In addition, grades collected 9 weeks following the intervention were significantly
mechanism in the social psychology literature. In particular, the authors find that when a nonhuman factor influencing rewards received for the maze-solving task (a random draw) was introduced, the caste gap disappeared. The results suggest that when caste identity is salient, low-caste subjects anticipate that their effort will be poorly rewarded.
9
Field Experiments: Correspondence Studies
189
higher for blacks in the malleable intelligence condition. Whites showed a similar, though statistically marginal, effect.27 While most research on stereotype threat, and how to undo it, has taken place in the lab, a few interesting field studies have also been conducted by social psychologists. Schools, where test-taking and performance measurements are part of normal operations, have provided a natural setting for much of this field research. Good, Aronson, and Inzlicht (2003) performed a field experiment to test methods for helping female, minority, and low-income adolescents overcome the effects of stereotype threat and, consequently, improve their standardized test scores. Specifically, seventh-grade students in the experimental conditions were mentored by college students who encouraged them either to view intelligence as malleable or to attribute academic difficulties in the seventh grade to the novelty of the educational setting. Results showed that females in both experimental conditions earned significantly higher math standardized test scores than females in the control condition. Similarly, the students – who were largely black or Hispanic and low-income adolescents – in the experimental conditions earned significantly higher reading standardized test scores than students in the control condition. Blackwell, Trzesniewski, and Dweck (2007) based a field experiment on the laboratory finding on “malleable intelligence.” They randomly selected half of a group of 95 mainly African-American and Hispanic seventh graders to participate in an 8-week training on the theory of malleable intelligence based on interventions that had been successful in the lab (25 min per week, in the students’ classroom). In the control condition, students also received small group coaching, but not on this theory. Students in the experimental conditions obtained higher grades. Good, Aronson, and Harder (2008) also conducted a field experiment where they explored stereotype threat and its negation in high-level college math courses that typically serve as gateways to careers in math and science. Male and female students in the last course of an advanced university calculus sequence were given a practice test containing items similar to those found on the Graduate Record Examination (GRE) standardized test. All students were told that the test was “aimed at measuring your mathematical abilities” (stereotype threat), but half of the students additionally were assured that “this mathematics test has not shown any gender differences in performance or mathematics ability” (stereotype threat negation). Test performance was higher for women than men in the stereotype threat negation condition but was equivalent in the stereotype threat condition. In a related field study, Cohen et al. (2006) reduced the black-white GPA gap among low-income middle school students by affirming the students’ self-concepts (and presumably inoculating them from stereotype threat) at the beginning of the school term. The intervention is very light touch: Students are asked to write a series
27
It is important to note that for this experiment, while the randomized interventions took place in the lab, outcomes are measured (1) on naturally occurring tasks outside the lab and (2) quite a long time after the interventions took place; both of these features are important strengths of this experiment compared to the standard “stereotype threat” lab experiments.
190
M. Bertrand and E. Duflo
of short essays focusing on a self-affirming value. The authors first found short-term impacts, and then, most remarkably, fairly large long-term impacts: Over 2 years after the intervention, the average GPA of an African-American student who participated in the essay writing was 0.24 higher than that of the control group (Cohen et al. 2009). Most of the effects are concentrated among those who were initially low achieving. These are remarkable numbers, especially since the effects of most education interventions tend to fade. The authors speculate that the long-term benefits may be due to the fact that initial psychological state sets out a selffulfilling trajectory. Cohen’s work has since then been replicated and extended in other similar contexts, and Dweck and Cohen’s initial insights have helped jump-start a subfield of psychology called “mindset studies.” Yeager et al. (2014) experiment with “wise feedback,” an intervention where high school students are given critical feedback on their written work, alongside with a note emphasizing the high standard of the teachers and the belief that the student can succeed. The intervention reduces mistrust among African-American students and improves the quality of the final product.28
Identity and Preferences The “stereotype threat” literature can be viewed as part of a broader literature on how self-identity considerations may affect behavior and preferences of disadvantaged groups and ultimately may perpetuate gaps in economic outcomes. The same way that women may do poorly on a math test when reminded of their gender (due to the anxiety-inducing burden of the stereotype of “girls not being good at math”), they may also show low-risk preferences when reminded of their gender (if nurtured with the behavioral norm that “girls should not take risk” by gender-biased parents and/or teachers). Against the backdrop of a large literature in social psychology that has tested the self-categorization theory and the cognitive mechanisms through which it operates,29 a few recent papers in economics have leveraged the lab environment to learn more about how various social identities relate to preference parameters, such as risk, time, and social preferences. For example, Benjamin, Choi, and Strickland (2010) explore the effect of racial and gender category norms on time and risk preferences. In a laboratory setting, they study how making salient a specific aspect of one’s social identity affects subjects’ likelihood to make riskier choices, or more patient choices. From a methodological perspective, the study consists of temporarily making more salient (“priming”) a certain social category (as is done in the “stereotype threat” literature) and seeing how the subjects’ choices are affected. For example, the gender identity salience
28
A background paper written for a White House conference gives a good overview of the literature on mindset studies (Yeager et al. 2013). 29 See Reicher and Levine (1994), Forehand, Deshpandé, and Reed II (2002), and LeBoeuf, Shafir, and Bayuk (2010).
9
Field Experiments: Correspondence Studies
191
manipulation is done through a questionnaire included in the beginning of the experiment in which subjects are asked to identify their gender and their opinion regarding living on a coed versus single-sex dormitory floor. The study uncovers some interesting patterns with respect to racial identity. For example, priming a subject’s Asian-American identity makes the subject more patient. Hence, an AsianAmerican identity might partly contribute to the higher average level of human capital accumulation in this racial group. However, making gender salient appears to have no significant effects on either men’s or women’s patience, or their level of risk aversion. Of course, it is possible that the priming performed in this experiment was too weak to temporarily affect preferences. In other words, it is difficult to affirmatively conclude from these nonresults that gender identity norms are not culturally reinforcing whatever biological differences may exist between the sex in the willingness to take risks. Another lab study aimed at assessing how social preferences are affected by gender identity is Boschini, Muren, and Persson (2009). The question under study here is whether gender identity priming affects subjects’ level of altruism. The experiment consists of comparing behavior in a dictator game for subjects whose gender identity has been primed versus not primed. The results indicate that the priming does affect behavior but only when the subjects are assigned to mixedgender groups. Moreover, the effect is driven by males: men are sensitive to priming and become less generous in a mixed-gender setting when primed with their male identity. Women do not appear to respond to the treatment. As far as we are aware of, no field experiment exists on how social identity affects preferences and behaviors outside of the mindset literature discussed above, which focuses on education and on adolescents. It seems worthwhile for future research to consider such work. Interventions might be designed to emphasize a “default” social identity that may be counterproductive for that social group’s performance against an “alternative” social identity. For example, while deciding to work hard toward completing college coursework for a young black father might be uncool because it is “acting too white,” the decision might resonate much more when his identity as a “father” is being primed. Moreover, specific interventions might be designed to simply undo or undermine the power of the social identity norms when they work toward reinforcing differences in behaviors and outcomes between groups. If women decide against applying for a job in a high-risk but also high-return occupation because of internalized conservative social norms about “what is appropriate work for a woman,” it might be possible to undermine the pull of this conservative norm with counteractive “messaging,” in the same spirit as what has been done to undermine the burden of the stereotype in the “stereotype threat” literature. Such interventions might be particularly powerful if the timing of the counteractive “messaging” is close to when women are making these important career choices (e.g., when applying for school or on a job search website, or when considering which contact in their LinkedIn network to reach out to).
192
M. Bertrand and E. Duflo
Expectancy Effects and Self-Fulfilling Prophecies Pygmalion and Golem Effects Suppose minority and majority workers have similar inherent abilities. How could differential beliefs about their abilities persist? One explanation is that employers’ beliefs that minorities are on average less productive are self-reinforcing (Arrow 1973; Coate and Loury 1993). This could happen for two reasons. First, minority and majority workers may rationally make different skill investment or effort choices in the face of the beliefs of their employers. A minority worker may see less value in investing in her skills if she knows that the employers will be slow in updating their beliefs, and hence less likely to promote her. Second, the employers themselves may invest less in the minority workers (e.g., investing in training) if they do not believe that the workers will be “up to the task.” In both cases, employers’ beliefs about minority workers will be self-fulfilling. The social psychology literature offers multiple demonstrations of such selffulfilling prophecies.30 Interesting, most of these demonstrations took place in the field. Earlier work on self-fulfilling prophecies in the social psychology focused on how heightened expectations can be self-fulfilling. In a seminal study, Rosenthal and Jacobson (1968) conducted a field experiment in a US public elementary school (Oak School). Teachers were deceived into believing that a set of one-fifth of their class were expected to develop (“blossom and spurt”) much faster than the rest, as measured by IQ points (supposedly measured by the Harvard test of inflicted acquisition). In fact, this set was randomly selected. The main outcome measure was an IQ test (Test of General Ability), administered at the start of the school year (pretest) and at 4 months (end of first semester), 8 months (end of second semester and of first year of school), and 20 months (end of second school year with a different teacher). Rosenthal and Jacobson showed that the students for whom teachers had raised expectations had faster IQ gains than control students in the same classes (the treatment children gained 12 IQ points over the course of the year, and the control children gained 8), with the biggest effect on first and second grade children by the end of the first year. Rosenthal and Jacobson dubbed this boost in achievement driven by teachers’ beliefs the “Pygmalion effect.” The Pygmalion effect in the classroom was subsequently studied intensively,31 and criticized extensively. Snow (1995) reanalyzed the data from the original
The first self-fulfilling prophecy to be investigated extensively in psychology was the experimenter effect. The experimenter effect refers to the possibility of the experimenter influencing subjects to respond to the treatment in a way that conforms to the experimenter’s expectations. Rosenthal (1963) summarized a dozen experimenter-effect studies and wondered whether similar interpersonal expectation effects occur among physicians, psychotherapists, employers, and teachers. 31 See Dusek, Hall, and Meyer (1985), Rosenthal and Rubin (1978), which was one of the first metaanalysis in psychology and was based on 345 studies, and Rosenthal 1994, which updates it for reviews. 30
9
Field Experiments: Correspondence Studies
193
experiment, highlighting that approximately 35% of the IQ observations fall out of the normal range and that there are several observations which have rapid growth in pretest and posttest scores (e.g., increasing from 17 to 110). He finds that the expectancy effect on IQ disappears when these outlier scores are omitted. A relatively recent review of the literature (including a balanced review of the metaanalysis and the various critics) concludes that while the Pygmalion effect in the classroom is real, it is probably fairly modest (Jussim and Harber 2005). Since this early work, social psychologists have demonstrated the self-fulfilling nature of leaders’ expectations in several other field settings and have tried to better understand the underlying mechanisms. Rosenthal (1994) and Eden (1992) provide a review of much of the work in this literature. For example, Eden and Shani (1982) replicated the original design and results of Rosenthal and Jacobson (1968) in the Israeli Defense Forces. But they also concluded, based on additional survey work to complement the randomized controlled trial, that leadership behavior was a key mediator in generating the Pygmalion effect. Also using the Israeli Defense Forces as a field, Eden and Ravid (1982) interestingly combined expectancy and self-expectancy manipulations in a single study. Trainees included 60 men in the first half-year of military duty enrolled in a 7-week clerical course divided into five training groups, each instructed by a commander. To produce the Pygmalion effect, a random quarter of each instructor’s trainees were described to the instructor as having high success potential. Another random quarter were told directly by a psychologist in a brief personal interview that they had high success potential, in order to induce high self-expectancy. The remaining trainees served as controls. Learning performance was significantly higher in both high expectancy groups than in controls, confirming the Pygmalion hypothesis and the additional hypothesis that inducing high self-expectations similarly enhances trainee performance. Interestingly, while several instructors were unexpectedly relieved midway through the course, the hypothesized performance differentials continued even though the authors abstained from refreshing the expectancy induction among the substitute instructors, reflecting the possible durability of expectancy effects. Finally, Eden and Ravid (1982) also showed that equity considerations among the trainees likely played a mediating role: Trainees in both of the high expectancy conditions reported feelings of over-reward, which may have motivated them to improve their performance. While the Pygmalion literature shows the self-fulfilling nature of raising leaders’ expectations, another branch of this literature also demonstrated the self-fulfilling nature of lowering those expectations. Psychologists have dubbed this the “Golem effect.” There have been far fewer studies on the Golem effect than the Pygmalion effect, given the trickier ethical issues associated with lowering leaders’ expectations (Reynolds 2007). This challenge has led to research designs that are not quite as “clean” as those used to demonstrate the Pygmalion effect. For example, Oz and Eden (1994) randomly led treatment-assigned squad leaders (n ¼ 17) in a military unit to believe that low scores on physical fitness tests were not indicative of subordinates’ ineptitude, while control squad leaders (n ¼ 17) were not told how
194
M. Bertrand and E. Duflo
to interpret test scores. Tests indicated that low-scoring individuals in the experimental squads improved more than those in the control squads. While the researchers employed a respectable research design and were cautious to abide by ethical standards, the sample was extremely small and the researchers did not directly attempt to lower supervisors’ expectations. Given the challenge of doing research on the Golem effect, an alternative approach in the literature has been to rely on natural variation in leaders’ expectations, rather than exogenously varying those expectations. For example, Babad, Inbar, and Rosenthal (1982) studied expectation effects among physical education student-teachers. They found that pupils about whom they imparted high expectations to the instructors performed best (i.e., the standard design for a demonstration of the “Pygmalion effect”). However, they also found that pupils toward whom instructors naturally harbored low expectations performed worse than those regarding whom they had high or intermediate natural expectations, consistent with a “Golem effect.” A recent paper (Kondylis et al. 2015) demonstrates the power of self-fulfilling prophecy. In villages, either women or men were randomly selected to learn a new technology and teach it to others. Women retained more information from the training, and those who were trained by them did in fact learn more. But women ended up performing much worse in terms of the number of farmers they convinced, because other farmers perceived women as less able, and hence paid less attention to their messages.
Endogenous Responses to Bias While the Pygmalion and Golem effects demonstrate the self-fulfilling nature of leaders’ expectations about performance on that performance, they are not directly tied to discrimination. Are leaders’ biases against some groups also endogenously affecting the performance of these groups? Two recent field studies in the economics literature provide what we believe is the first field-based answers to this question. Conceptually, these studies follow a very similar research approach to that in Babad, Inbar, and Rosenthal (1982) to demonstrate the relevance of self-fulfilling prophecies as an explanation for persistent differences in performance between different groups of workers or students. Specifically, rather than “artificially” priming leaders to vary their level of bias, the analysis relies on randomly assigning those trainees to leaders who are known to have different levels of bias. To be clear, the limit of this design compared to the preferred “Pygmalion design” is that any unobserved factors that are systematically correlated to different levels of biases among leaders cannot be formally ruled out as an explanation for the findings. However, the two papers below take several ingenious steps to deal with this concern. Glover, Pallais, and Pariente (2015) studied cashiers in a French grocery store chain, a sizable share of who were of North African and Sub-Saharan African origin. They assess whether cashiers performed worse on the days when they were assigned to a manager who was more biased against their group. They measured managers’
9
Field Experiments: Correspondence Studies
195
bias toward workers of different origins using an IAT test. The cashiers in these stores worked with different managers on different days and had virtually no control over their schedule, allowing the authors to use the quasi-randomness of the schedules to assess the causal effect of being paired with a more biased manager. To address the difficulty raised above of manager bias being correlated with some other manager characteristics that might also affect employee performance (for example, more biased managers might also be less skilled), they use a difference-in-difference methodology, comparing the change in minority workers’ performance under biased and nonbiased managers with the change in nonminority workers’ performance under these two types of managers. They find that on days when they are scheduled to work with biased managers, minority cashiers are more likely to be absent. When they do come to work, they spend less time at work; in particular, they are much less likely to stay after their shift ends, and they scan articles more slowly and take longer between customers. Glover, Pallais, and Pariente (2015) also report interesting complementary survey evidence to better understand mechanisms. They do not find that minority workers report that they dislike working with biased managers more, or that biased managers dislike them, or that biased managers make them feel less confident in their abilities. However, they do find evidence that biased managers put less effort into managing minority workers. Minority workers report that biased managers were less likely to come over to their cashier stations and that biased managers demanded less effort from them. Consistent with this, they find that the effect of manager bias grows during the contract, perhaps as workers may learn that they are not being monitored by biased managers. Lavy and Sand (2015) estimate the effect of primary school teachers’ gender biases on boys and girls’ academic achievements during middle and high school, as well as on the choice of advanced-level courses in math and sciences during high school. In particular, they tracked three cohorts of students from primary school to high school in Tel-Aviv, Israel. They measured teachers’ gender biased behavior by comparing their average marking of boys and girls in a “non-blind” classroom exam to the respective means in a “blind” national exam marked anonymously. For identification, the authors rely on the conditional random assignments of teachers and students to classes within a given grade and a primary school. They compare outcomes for students that attended the same primary school but were randomly assigned to different teachers, who have different degrees of stereotypical attitudes. They find that being assigned to a more gender-biased teacher at early stage of schooling has long-run implications for occupational choices and hence likely subsequent earnings in adulthood. Specifically, teachers’ overassessment of boys in a specific subject has a positive and significant effect on boys’ achievements in the national test on that subject administered during middle and high school, while it has a significant negative effect on girls’. In addition, assignment to primary school math teachers favoring boys over girls encourages boys and discourages girls from engagement in advanced math courses offered in high school.
196
M. Bertrand and E. Duflo
Discrimination in Politics and Inequality Across Groups A direct consequence of discrimination in politics and other leadership positions is that there are fewer members of the discriminated group with the power to act in the interests of others in their group. In a standard median voter world, the underrepresentation of women or other subordinate groups in politics would matter less, as elected politicians would endeavor to represent the interest of the median voter. But if politicians cannot commit to a particular political platform, and their group membership eventually determines the type of policies they will implement, then the lack of representation at the top means that the underrepresented groups in society will get worse outcomes (Besley and Coate 1997; Pande 2003). This would also occur if the absence of a leader means that the underrepresented groups find that they cannot express their preferences in the political arena. The best evidence on the consequences of discrimination in politics comes from studies that have evaluated what happens to the underrepresented groups when they finally gain political representation. A few observational studies have exploited exogenous shocks to representation due to close elections; a few other papers have also studied nonrandomized variation in mandates.32 There have also been a set of studies that exploit the random selection of places that have to elect a leader from a historically underrepresented group (caste, tribe, or gender) in India’s local governments. Comparing villages that were randomly selected to receive either a male or female head, Chattopadhyay and Duflo (2004) find that female leaders spend more on goods that women prefer, compared to those that men prefer. Beaman et al. (2010) replicate the results over a longer time period. Using a dataset that covers a larger number of states, they find that the results persist over time, and that investments in drinking water (a preferred good for women) continue to be higher even after the seat is not reserved anymore and women have (generally) left power.33 Iyer et al. (2012) show that greater female representation (in local governments) is related to more crimes against women; using a household-level crime victimization survey in Rajasthan, they however show that the increase is not due to an actual increase in the amount of crimes, but rather greater willingness to report such crimes. Finally, Chattopadhyay and Duflo (2004) further find that village leaders from scheduled castes invest more in scheduled caste hamlets.34
32 See Pande (2003), Clots-Figueras (2009), Clots-Figueras (2011), and Rehavi (2007) for examples. 33 Bardhan, Mookherjee, and Parra Torrado (2010) compare places before and after reservation and do not find a difference in what leaders do, but since there seems to be a lingering effect of quota on pro-female policies, this finding might not be so surprising. 34 Dunning and Nilekani (2013) find little impact of the reservation on distribution of goods by ethnic group and a strong impact of parties, but they use a regression discontinuity design strategy rather than focusing on a states where the assignment is random.
9
Field Experiments: Correspondence Studies
197
Benefits of Diversity? Discrimination leads to less-diverse firms, legislative assemblies, etc., but does diversity in itself matter for society? What are the implications of the low diversity that discrimination may generate for the performance of organizations and society in general?
Does Homogeneity Hurt or Help Productivity? A long literature in political economy and development has tended to emphasize the cost of diversity, in particular ethnic diversity. If members of different groups do not like each other, diversity creates holdups, breeds conflicts, makes it difficult to agree on public good provision, etc. Lang (1986) proposes that the roots of discrimination are communication difficulties across different groups (what he calls “language communities”). A similar argument is made by Lazear (1999). In that view of the world, segregation arises naturally, because homogenous groups are more productive (since communication within them is faster and easier). More homogeneous groups will create a trusting environment where people can work better together. While the minority will suffer as a result, the short-run equilibrium is efficient, and policies directly aimed at increasing diversity would be socially counterproductive. The role of policy should be instead to diminish language barriers between groups (through the education system, for example). Others have emphasized the benefits of diversity, and potential drawbacks of “homophily” (or the tendency to want to associate only with people like oneself). One powerful argument is that similar people will tend to have similar information and perspectives, and if people only interact with people like themselves, lots of valuable information will be not transmitted across groups. Arguments along these lines have been made, more or less formally, in the human resources and management literature. More formally, Golub and Jackson (2012) show that, when agents in a network prefer to associate with those having similar traits (homophily), it may take a very long time for participants in a network to converge to a consensus. Ultimately, there is thus a trade-off between the cost of communication and collaboration and the benefits of diverse viewpoints, which means that diversity (and hence homophily) may in theory hurt or improve productivity (Hamilton et al. 2003; Alesina and La Ferrara 2005). While there is a large nonexperimental literature on the impact on diversity on public good provision35 and a sizable lab experimental literature,36 the field experimental literature is more nascent. There are nevertheless a few interesting recent papers that we review below. Hoogendoorn and Van Praag (2012) and Hoogendoorn, Oosterbeek, and Van Praag (2013) experimentally varied the composition of teams of undergraduate 35 36
See Alesina and Ferrara (2005) for a review of the literature on diversity. For example, Woolley et al. (2010) and Engel et al. (2014).
198
M. Bertrand and E. Duflo
student required to start a business venture as part of a class. In teams of 12, students start up, sell stock, and run a real company with a profit objective and shareholders for a year. They ran the experiment on 45 teams and 550 students. The composition of teams was varied by gender (men only, women only, or mixed) and ethnicity (the fraction of non-Dutch ethnicity varied from 20% to 90%). Students then had a year to choose their venture, elect officers, conduct meetings, produce, sell, make money, and liquidate. This was a field experiment; the program played out over a year, and the incentives to do well were very high. Students’ ability to graduate, their grades, and potentially some money were all on the line. There is a clear benefit to gender diversity in the experiment. The performance as a function of the share of women in the team is inverse U-shaped, with the peak reached when the share of women is approximately 0.55. The authors attribute this effect to greater monitoring in gender-diverse groups. This in itself is an interesting finding as this is not a mechanism that is emphasized in the theoretical literature: Perhaps when communication is too easy, workers become more complacent. For ethnic diversity, the result is more subtle: Hoogendoorn, Oosterbeek, and Van Praag find that the marginal effect of increasing diversity on performance is zero or perhaps even negative when the teams are at least 50% Dutch. However, once the teams are less than 50% Dutch, further increases in the share of other groups are associated with better performance. The authors also identify evidence for the different channels proposed in the theoretical literature (including not only higher communication costs in more diverse groups, but also more diverse knowledge in more diverse teams), but these results are not extremely precise. Hjort (2013) analyzes a natural experiment where a flower firm in Kenya randomly assigned workers to teams. Kenya offers a context with heightened ethnic tensions, and where the level of distrust among different groups may be particularly high. In the experiment, an upstream worker distributed flowers to a team of two downstream workers. The upstream worker earned w per flower packed, and the downstream worker 2w per flower packed. Hjort finds that, conditional on productivity, upstream workers distribute fewer flowers to teams when one or both are not from his ethnic group, at the cost of lower wages for him, and lower production overall. Furthermore, within mixed teams, upstream workers give more flowers to the worker from their same ethnic group. Interestingly, the output gap between homogenous and ethnically mixed teams doubled during the period in 2008 when ethnic conflict intensified. In response to this, the firm introduced team pay for the downstream workers (not randomized) and subsequently experienced an increase in the productivity of the ethnically mixed teams. Also in Kenya, Marx, Pons, and Suri (2015) randomly assigned enumerators to pairs, and each pair to a supervisor. The job of the enumerator was to make contact with a household and administer an intervention. They find that homogenous pairs have higher productivity, and they attribute that to higher trust in those teams. However, when a pair is further matched with a supervisor of the same ethnic group, the productivity is lower (not higher). The contrast between the (negative) impact of diversity in horizontal teams and the (positive) impact in vertical relationships in the Marx, Pons, and Suri (2015) experiment hints at a different potential
9
Field Experiments: Correspondence Studies
199
negative impact of discrimination: In-group preference may create room for cheating and corruption. In their setting, the coethnic supervisor was willing to let the enumerators cheat.
Discrimination and Corruption Prendergast and Topel (1996) provide a theoretical analysis of the influence of favoritism on optimal compensation and extent of authority for the manager. The point extends further than the firm: Discrimination may lead to misallocation of resources by politicians (to members of their ethnic group), or conversely to willingness to put up with corrupt or incompetent politicians from one’s own group (rather than less corrupt ones from another group) (Key 1949; Padro i Miguel 2007). More generally, voters’ preferences for a group may diminish the role of issues in campaign, and by implication the quality of government (Dickson and Scheve 2006). Besley et al. (2013) provide nonexperimental evidence of this effect: They show that in Sweden, after the social democratic party imposed gender balance by requiring that all candidates be selected in a “zipper” pattern (one man/one woman), the quality of the male candidates greatly increased (they call this the “crisis of the mediocre man”). Experimental evidence of the link between homophily and the quality of politicians or corruption level is rare. One interesting experiment took place in Uttar Pradesh, India’s most populous state, where the rise in caste-based politics has been accompanied by a staggering criminalization of politics (Banerjee et al. 2010). On the eve of the 2007 election, 206 of the sitting members of the legislative assemblies had a criminal case pending against them (Banerjee and Pande 2009). Prior to the 2007 election, the authors conducted a field experiment in which villages were randomly selected to receive nonpartisan voter mobilization campaigns (street plays, puppet shows, or discussions). One type of campaign encouraged citizens to vote on issues not of caste, while the other encouraged them to not vote for a corrupt candidate. They found that the caste campaign led to a reduction of the (reported) votes on caste, and to a reduction in the vote share going to candidates with a criminal record. It thus seems that successfully reducing discrimination (in this case, to be more precise, reducing lower caste group members’ tendency to systematically discriminate against higher caste candidates) does lead to an improvement in the quality of elected leaders. The natural experiment in India discussed in section “Discrimination in Politics and Inequality Across Groups” that introduced quotas for women in politics shed interesting light on this question as well. In the short run at least, reservation for women politicians reduced bribe taking (Beaman et al. 2010). Of course, in the short run, quotas do not increase competition (since on the contrary the pool is reduced to women only, whereas it was initially open to women and men) and the observed reduction of corruption could be due to inherent characteristics of women, or to their lack of experience. However, quotas do tend to increase political competition in the medium run, because once a woman leaves office and her seat is open, she (or her relative) has the option to run again, but the field is now more open to competition
200
M. Bertrand and E. Duflo
than if she were a traditional incumbent. Moreover, when Banerjee et al. (2013) collected data on what happens in previously reserved places, they found that female incumbents whose seats became free were less likely to run than male incumbents whose seat became free, but that this effect disappeared when they considered not only the incumbent, but also the incumbent and their family. In other words, the probability to elect someone from the incumbent’s dynasty remains the same in places that just have experienced a quota or not. Also, they found that the probability of reelection of someone from the incumbent’s family is more sensitive to past performance in the previously reserved places. Thus, the best politicians’ dynasties are more likely to be reelected after reservation, and the worst ones less likely to. To the extent this effect persists, it does suggest that policies that constrain voters to vote outside of their “comfort” zone may improve the quality of the decision-making process overall even after these constraints are lifted.
Law of Small Numbers Even if discrimination does not lead to outright corruption, it may restrict the pool of available candidates. Research shows that the leader quality matters both for firms (Bertrand and Schoar 2003) and for countries (Jones and Olken 2005). If discrimination implies that leaders have to be selected from a relatively small pool, it reduces the chance that the most talented person will be picked, and it thus may have negative productivity consequences. The empirical evidence (even nonrandomized) on any such consequence of discrimination is thin at best: Ahern and Dittmar (2012) and Matsa and Miller (2013) examine the impact of the Norway 2006 law which mandated a gender quota in corporate board seats. They both find negative consequences on profitability and stock prices. However, these are short-run impacts. It could be that women are temporally less effective because they are less experienced, or that they maximize something else other than short-run shareholder value, which may turn out to be profitable in the long run. Unfortunately, we do not see an experiment on this, nor can we think of an obvious design for one. But this would be a very intriguing avenue for further research.
What Affects Discrimination? Leaders and Role Models One effect of discrimination against a group is that few leaders emerge from it into the mainstream. This has potentially three consequences. First, mechanically, fewer people from this group are in a position to make a decision regarding others. To the extent that leaders discriminate against members from other groups, discrimination will persist. Second, the majority group may be reinforced in their belief that the minority group is incapable of success, since they have rarely, if ever, observed success of the minority in practice. And third, members of the minority group may
9
Field Experiments: Correspondence Studies
201
then feel that either they are incapable of succeeding, or that the world is rigged against them, and there is no point in even trying.37 Given all this, discrimination could be lessened by forcing exposure to leaders from groups that are traditionally discriminated against, which is often achieved through quotas. This section reviews the evidence for this and also points out the gaps.
Does Diversity in Leadership Positions Directly Affect Discrimination? Mechanically, discrimination may breed discrimination, because the decisionmaking power is concentrated with the majority group. For example, if managers are mostly males, they may tend to favor other males in their recruiting or promotion decisions. This may happen because they themselves discriminate (consciously or unconsciously) or because they know more males and are more likely to promote or hire people they know or who are more similar to them.38 This tendency is part of the rationale for requiring a certain fraction of women on corporate boards, in academia, or in appointment, evaluation, and promotion committees. It is, however, not obvious that minority group leaders, or committees that contain such minority leaders, would necessarily favor others from the minority: Faced with their own discrimination, they may feel the need to go to lengths to avoid being perceived as biased. In several observational studies, women were not inclined to judge other women more favorably than men.39 In group decisions, there may also be a response of other members of the committee, who may try to “undo” any agenda they perceive (rightly or wrongly) the minority group members to have. The empirical evidence of the impact of minority representation on selection committees largely comes from a series of very interesting papers by Bagues, Zinovyeva, and their coauthors. Bagues and Esteve-Volart (2010) examine the impact of the gender composition of the evaluation committee for the entry exam in the Spanish judiciary on the success of women in that exam. A causal study is made possible by the fact that people are randomly assigned to a committee. They find that women are less likely to succeed at the exam when the committee they are assigned to has more women, while the opposite is true for male candidates. Additional evidence in the study suggest that these results might be driven at least in part by the fact that female evaluators tend to overestimate the quality of male candidates. Zinovyeva and Bagues (2011) and Bagues, Sylos-Labini, and Zinovyeva (2014) present interesting evidence from randomized academic evaluation committees in Spain and Italy, respectively. In both countries, candidates for promotion appear in 37
See Lockwood and Kunda (1997). Bagues and Perez-Villadoniga (2012) provide evidence from entry exams into the judiciary in Spain that support the latter effect; Bagues and Zinovyeva (2015) show that the former effect also applies in the case of academic promotions. 39 See Booth and Leigh (2010) for an audit study in Australia, where they find no interaction between the gender on the résumé and the gender of the recruiter. See also Broder (1993) for similar evidence in the context of NSF proposal reviews, and Abrevaya and Hamermesh (2012) for referee reports. 38
202
M. Bertrand and E. Duflo
front of a centralized committee to be qualified. Files are assigned to randomly composed committees. In the Spanish case, the authors find no impact of an additional female committee member on the promotion likelihood of female candidates. In the Italian case, they find a negative effect: In a five-member committee, with each additional female member added to the committee, the success rate of female applicants relative to that of male applicants decreases by around two percentage points. Analyzing the voting records, they find that both (1) the same female candidate is scored on more harshly by females than by males, and (2) male committee members grade female candidates more harshly when there are women on the committee, perhaps because they are trying to compensate for a perceived bias in favor of women on such committees (even though in reality the opposite appears to be true given (1)). This evidence on academic and recruitment committees is striking and suggests that some type of affirmative action may in fact hurt promising female candidates. It would be interesting to see if it also carries through in other settings, such as management or political decisions. Bursell (2007) is an audit study that makes some progress in this direction (although the specific comparison it focuses on is itself not experimental). He sent 3552 applications to 1776 jobs in Sweden, including applications to more skilled positions, such as senior/high school teachers, IT-professionals, economists, and engineers, and compares, among other things, the callback rates for applicants with Swedish-sounding and non-Swedish-sounding names according to the name of the CEO. He finds, consistent with the evidence above, that when the “CEO of a company has a foreign sounding name, the applicants with a Swedish sounding name have a 2.4 times higher probability to receive a call-back. If the CEO has a Swedish sounding name, the probability is 1.7 times higher” (Bursell 2007).
Minority Leaders and the Attitude of the Majority Even if there is no direct effect of having women or minority members in on leadership decisions, it could still affect discrimination against the minority because those minority individuals in leadership positions will change the beliefs, or precision of the beliefs, of the majority about the competence of the minority group. In a working paper version of Beaman et al. (2009), the authors propose a model where taste and statistical discrimination reinforce each other. Suppose that there are strong tastes (or social norms) against having a female leader. Then, it is very likely that citizens have never observed a female leader in action. This makes female leaders riskier as a group: Even if citizens believe that female leaders are equally competent on average, they have much more precise priors about male leaders, and to the extent they are risk averse, this will lead them to avoid women leaders. This is of course reinforced if citizens start with the prior that women are less competent: They will never have the occasion to find out that in fact they are wrong. In this world, forcing exposure to minority leaders (political leaders, board members, colleagues in academic departments, students at top colleges, etc.) will have a persistent negative impact on discrimination, even if it does not affect the underlying
9
Field Experiments: Correspondence Studies
203
taste for the community, simply by affecting beliefs about the competence of the minority group.40 The impact might be reinforced if the image of what constitutes a good leader also evolves in response to what people see. Eagly and Karau’s (2002) “role incongruity” theory stipulates that one reason why people prefer male leaders is that the traits associated with leadership (strength, assertiveness) are not traits that are associated with women under prescriptive gender norms (such as being nice, accommodating, etc.). Yet, as people get to see many (not just one or two token) women leaders who are strong, but also nice, or who have an effective, but also more accommodating, leadership style, they may change their attitude toward female leaders. As inconsistencies between the female gender stereotype and qualities associated with leadership diminish, so will prejudice toward female leaders. Of course, a potential force in the opposite direction is the possibility of backlash against minority leaders, in particular if there is a perception that they got into their leadership role because of special treatment (Coate and Loury 1993). Beaman et al. (2010) study a natural experiment in the context of local electoral politics in India and are able to provide more evidence for the mechanism underlying the persistent effect of temporary affirmative action policies. In a context where local village councils are randomly selected, by rotation, to be forced to elect female leaders, the authors show that after a cycle of reservation (and even more when the same place happened to be reserved for a woman for two cycles in a row), more women run, and are elected, on unreserved seats. While there could be many reasons for this (including the fact that women may have become more willing to run, or that networks of women may have been created), they provide evidence that it is probably at least in part due to a change in attitude in the villages that were previously subjected to reservation. They collect evidence on attitudes in various ways: with a Goldberg-type experiment, and with two IATs, one for like or dislike for female leader (a more “hardwired” attitude that their model takes no stance on) and one for a stereotype associating women with domestic activities and men with leadership activities (in the spirit of the assessing whether “role incongruity” effects diminished). They find that the experience with the past quota does not affect preferences (as measured by the taste IAT), although it tends to harden stated preferences against women in leadership. However, citizens (particularly men) update on measures of perception of women’s competence. For example, their rating of a speech pronounced in female voice converges to that of a speech given by a male voice if they have been exposed to a quota either in this cycle or in the previous cycle. Moreover, the stereotypical IAT also shows a decrease of the stereotype that associates women with domestic activity rather than with leadership. This provides reasonably strong field evidence that exposure to role models from another group In an observational study, Miller (2014) finds evidence consistent with such effects in the context of affirmative action programs in the USA. US government contractors are forced to hire minority workers. Miller finds that, after an establishment is no longer subject to such affirmative action because it stops being a government contractor, the black employment share nevertheless continues to grow.
40
204
M. Bertrand and E. Duflo
affect attitudes. The evidence seems quite robust: Bhavnani (2009) also finds that women continue to be more likely to be elected after a seat was reserved for a while (in urban Maharashtra). Banerjee et al. (2013) also find, in Rajasthan, that women are more likely to run (and win) on a previously reserved seat. Although there is a vast laboratory literature that test the role-incongruity theory and its implications, and laboratory studies that show, for example, that college students asked to screen candidates for a typical male job (e.g., finance manager trainee) are less likely to discriminate against a female résumé after reading an editorial documenting women’s success in this type of job (Heilman and Martell 1986), we are not aware of field experiments that investigate these types of effect in other contexts (e.g., exposure to a female or black manager, minority teachers, etc.). It would be valuable to establish whether such impact on the majority’s attitude can also be documented in some of these other contexts.
Role Models, Aspirations, and the Attitude of the Minority Leaders issued from disadvantaged groups, in addition to their direct decisionmaking power and to the effect they may have on the opinion of the majority, could also serve as role models and trailblazers. They might affect the attitudes of the minority group about their own ability to succeed, or their aspirations to do so. Seeing successful women or blacks may lessen stereotype threat (as discussed in section “Minority Leaders and the Attitude of the Majority”), or the belief among those groups that society is rigged against them so there is no point in trying anyway. In both cases, exposure to role models may increase effort and lead to better outcomes for the minority, even without direct changes in the majority attitude (though this could of course trigger subsequent change in the majority’s beliefs and attitudes as well). However, as noted by Lockwood and Kunda (1997), these positive effects might be moderated by how fixed minority group members view their ability, and how personally relevant and attainable they consider the achievement of the role models. As in the case of the impact of exposure on the attitudes of the majority, there is both a descriptive and a laboratory literature on this question. The observational literature looks for correlation between outcomes (e.g., teen pregnancy) and measure of stereotyping (e.g., IAT) to naturally occurring exposure (e.g., black teachers). For example, Dasgupta and Asgari (2004) show that women who have been exposed to female teachers and role models are more likely to associate women and leadership. A laboratory experiment literature explores the extent to which exposure to stereotypically feminine role model in a science, technology, engineering, and math (STEM) career (Betz and Sekaquaptewa 2012) or exposure to a nonstereotypical computer science role model (Cheryan et al. 2011) increases the likelihood that girls will present themselves as interested in STEM. Interestingly, and maybe in line with the cautionary note in Lockwood and Kunda (1997), these studies show that a role model that simply belongs to the minority group might not be sufficient, and that the “type” of role model appears to matter significantly. Cheryan et al. (2011) find that exposure to a nonstereotypical computer science role model (e.g., someone who dresses fashionably, enjoys sports and hanging out
9
Field Experiments: Correspondence Studies
205
with friends, and watches “normal” TV shows, such as The Office) through a “getting to know each other” task in the lab increases female subjects’ beliefs of succeeding in the field, and that this is true regardless of the gender of that role model. Similarly to the discussion in section “Identity and Preferences” on “social identity,” the authors argue that this is because women feel the stereotypical characteristics of a computer science major (e.g. social isolation, obsession with computers, and social awkwardness) do not align with what they see as their female gender role (e.g., helping others, having social skills, and attending to physical appearance), and a nonstereotypical role model, whether man or woman, can thus influence young girls’ preferences and beliefs. On the other hand, Betz and Sekaquaptewa (2012) find that counterstereotypic-yet-feminine STEM role models (as signaled by characteristics such as wears makeup and pink clothes, enjoys reading fashion magazines, etc.) discourage middle school girls’ success expectations in STEM relative to gender-neutral STEM role models (as signaled by characteristics such as wears dark-colored clothes and glasses, enjoys reading books, etc.). The authors find that this is particularly the case for girls who did not identify with STEM subjects and conclude that this subgroup of girls viewed the combination of both STEM-success and femininity as unattainable. Reviewing either literature fully is outside the scope of this chapter, but the mixed results of the lab experiments provide interesting directions for field research. Here again, however, the field experiment work so far appears to be quite limited. Beaman et al. (2012) study the same randomized natural experiment for women in leadership positions in India and look at the impact on girls’ educational attainment and career aspirations. They give evidence of impact on parents’ hopes for daughters. Compared to never-reserved villages, parents in reserved villages were more likely to state that they would like their girl to graduate or study beyond the secondary school level, and more likely to state that they would like their daughter to have a career. Parental aspirations for boys did not change. Furthermore, in villages with reservation, girls were more likely to stay in middle school, which cannot be directly attributed to any direct action of the leader because middle schools are not under their jurisdiction. This is therefore strongly suggestive of a causality running from role model to change in aspirations and actual change in behavior. Overall, this literature seems to us surprisingly thin, compared to the larger literature on “horizontal exposure” (e.g., roommates or classmates) which we discuss below in section “Intergroup Contact.” Part of the explanation for this is practical: There is probably more naturally occurring variation in peer groups than in supervisors, leaders, or teachers. Another issue is that, in many settings, female teachers or leaders may take actions that can translate directly into behavioral changes for female students (or trainees) even absent any effect on aspirations, so that isolating the impact on minority aspirations is tricky. While this was not the case in the quota experiment in India, it could have been. Nevertheless, we suspect that the lack of more field studies in this area is also a reflection of too little attention devoted to this important and exciting topic, and that much more probably can be done to explore how exposure to role models affects minority groups’ aspirations.
206
M. Bertrand and E. Duflo
Intergroup Contact Allport (1954) is often credited with the development of the contact hypothesis, also known as Intergroup Contact Theory. The premise of Allport’s theory is that, under appropriate conditions, interpersonal contact is one of the most effective ways to reduce prejudice. The theory, which was originally devised for encounters across racial and ethnic lines, states that if majority group members have the opportunity to communicate with minority group members, they are able to understand and appreciate them. As a result of this new appreciation and understanding, prejudice should diminish. Allport’s proposal was that properly managed contact between the groups should reduce prejudice and lead to better interactions. In particular, Allport held that reduced prejudice will result when four features of the contact situation are present: equal status between the groups in the situation, common goals, intergroup cooperation, and the support of authorities, law, or custom. Much of the psychology literature on the contact hypothesis has focused on lab experiments that have helped refine Allport’s original theory. An unresolved issue in psychology is whether specific conditions for the contact situation are needed to ensure that contact will have the theorized effect. For example, is it important for the contact to take place in a cooperative environment with peers of equal status (e.g., two roommates in a university dorm working together on a math homework) for contact to be effective at reducing discrimination? In a meta-analysis combining observational and experimental studies of the intergroup contact theory, Pettigrew and Tropp (2000) find that intergroup contact reduces prejudice in 94% of the 515 studies reviewed. Their meta-analysis also suggests that the contact effect generalizes to a broad range of minority groups (not just racial and ethnic minorities but also the elderly, the mentally ill, LGBT, etc.) as well as a broad range of contact settings (schools, homes, etc.). Pettigrew and Tropp (2000) also assess whether the optimal conditions for contact stated by Allport are necessary for positive contact outcomes. They find that the inverse relationship between contact and prejudice persists, though not as strongly, even when the contact situation is not structured to match Allport’s conditions. Hence, while Allport’s conditions may not be necessary for prejudice reduction, some combinations of them might be relevant. Psychologists are still debating and investigating the specific negative factors that may prevent intergroup contact from diminishing prejudice. While much of the experimental research on the contact hypothesis has taken place in the lab, there have also been quite a few field experiments. Green et al. (forthcoming) identify 56 field experiments. The best known field work within economics has focused on contact between college roommates. Sacerdote (2000) was the first to exploit the random assignment of roommates at college for a study of peer effects on test scores. More relevant to us, Boisjoly et al. (2006) leveraged random roommate assignment at Harvard to study the impact of shared experiences at college on opinions about the appropriateness of keeping affirmative action policies. They find that white students who are randomly assigned African-American roommates are significantly more likely to endorse affirmative action. Hence, mixing with African-Americans tends to make individuals
9
Field Experiments: Correspondence Studies
207
more empathetic to them. They also find that white students who were assigned roommates from any minority group are more likely to continue to interact socially with members of other ethnic groups after their first year. What remains unclear from Boisjoly et al. (2006) is whether contact to a minority roommate reduced stereotype or bias. Empathy might increase even if bias is unaffected. Burns, Corno, and La Ferrara (2015) leveraged the same design to get at this question, and hence their paper is closest to a field test of the contact hypothesis. Specifically, they exploit random assignments of roommates in double rooms at the University of Cape Town to investigate whether having a roommate of a different race affects inter-ethnic attitudes, but also cooperative behavior and academic performance. They find that living with a roommate from a different race significantly reduces prejudice toward members of that group, as measured by an IAT. The reduction in stereotype is accompanied by a more general tendency to cooperate, as measured in a Prisoner’s dilemma game, but smaller effects on trust, as measured in a trust game. The paper also reports interesting results on grades. Black students that are assigned a nonblack roommate experience higher GPAs; white students that are assigned a nonwhite roommate experience lower GPAs. Related findings are reported in Shook and Fazio (2008). Participants were white freshmen who had been randomly assigned to either a white or an African-American roommate in a university college dormitory system. Students participated in two sessions during the first two and the last 2 weeks of their first quarter on campus. During these sessions, they answered questions about their satisfaction and involvement with their roommates and completed an inventory of intergroup anxiety, as well as an IAT test. Automatically activated racial attitudes (as measured with the IAT) and intergroup anxiety improved over time among students in interracial rooms, but not among students in same-race rooms. However, participants in interracial rooms reported less satisfaction and less involvement with their roommates than did participants in same-race rooms. Several field experiments in psychology have also examined contact effects between classmates. In particular, psychologists have looked at cooperative learning techniques – which are designed so that students must teach to one another and learn from one another and place a strong emphasis on the academic learning success of each member of the group – and tested whether these help reduce prejudice (Johnson and Johnson 1989; Slavin 1995). The rational is that because cooperative learning encourages positive social interactions among students of diverse racial and ethnic backgrounds, it creates some of the conditions hypothesized in Allport as beneficial to reducing discrimination: as students work cooperatively, they have the opportunity to judge each other on merits rather than stereotypes. Slavin and Cooper (1999) provide a review of the field evidence on cooperative learning, which has been generally supportive of cooperative learning being a useful tool to improve intergroup relations. For example, Slavin (1977) and Slavin (1979) study one particular approach to cooperative learning, called Student Teams Achievement Divisions. Under this approach, the teacher presents a lesson, and students then study worksheets in four-member teams. Following this, students take individual quizzes, and team
208
M. Bertrand and E. Duflo
scores are computed based on the degree to which each student has improved over his or her own past record. The team scores are published in newsletters. He finds that the students who had experienced such cooperative learning over periods of 10 to 12 weeks gained more cross-racial friendships than did control students. In a follow-up one year later, Slavin (1979) found that the students that experienced cooperative learning named an average of 2.4 friends outside their own race, compared to an average of less than one in the control group. Also, Slavin and Oickle (1981) found significant gains in Whites’ friendships toward AfricanAmericans as a consequence of using the same cooperative learning method, but, interestingly, no difference in African-American friendships toward Whites. A recent paper in economics also brings the contact hypothesis to the classroom, but under conditions that are not tailored to be as optimal as possible for prejudice reduction to occur, at least as hypothesized by Allport. Starting in 2007, some elite private schools in Delhi were required to offer places to poor students. Rao (2013) exploits this policy change and uses a combination of experimental and administrative data to study whether exposure of rich students (from 14 private schools in New Delhi) to poorer students affects (1) tastes for socially interacting with or discriminating against the poor, (2) generosity and prosocial behavior, and (3) learning and classroom behavior. Core to his identification strategy is a comparison of outcomes for treated and non-treated student cohorts within a school. Rao also exploits a second identification strategy that is closer to a randomized design. Some schools in his sample used the alphabetic order of first name to assign students to study groups and study partners. Hence, in those schools, the number of poor children with names similar to a given rich student provides plausibly exogenous variation in personal interactions with a poor student. This second identification strategy is obviously more appealing as a test of the contact hypothesis, because it focuses more centrally on changes in personal interactions between students, and rules out other confounds (such as changes in teacher behavior, changes in the curriculum, etc.). Rao (2013) finds that economically diverse classrooms cause wealthy students to discriminate less against other poor children outside school. As discussed in section “Willingness to Pay,” Rao’s approach to measure discrimination is quite unique. First, he relies on a field experiment in which rich participants select teammates for a relay race and are forced to reveal how they trade-off more-athletic poor students versus less-athletic rich students. Using this measure of discrimination, Rao finds that exposure to poor students at school reduces discrimination by 12 percentage points. Rao also conducts a second field experiment. He invites students to attend a play date at a school for poor students, and elicit incentivized measures of their willingness to accept. He finds that having poor classmates makes students more willing to attend the play dates with poor children. In particular, it reduced the average size of the incentive that is required to attend the play date by 19%. Having a poor study partner (e.g. contact alone) explains 70% of the increase in this “willingness to play.” When Rao (2013) turns to how exposure to poor students affects pro-social behavior and learning in the classroom, he finds that having poor classmates makes students more prosocial, as measured by their history of volunteering for
9
Field Experiments: Correspondence Studies
209
charitable causes at school, as well as their behavior in dictator games conducted in the lab. The findings reveal that exposure to poor students does not just make rich students more charitable towards the poor; instead, it affects generosity and notions of fairness more generally. Finally, Rao shows that exposure to poor classmates has limited effects on the wealthy students’ test scores: while he detects marginally significant but meaningful decreases in rich students’ English test scores, he finds no effects on Hindi or Math scores, or on a combined index over all subjects. The studies reviewed above suggest that that inter-group contact is an effective tool to reduce prejudice, even though more work remains to be done to ascertain the specific conditions under which contact will be most effective. Yet, some recent work in psychology (Dixon et al. 2012) suggests new angles through which the contact hypothesis should be evaluated, and more specifically, the possibility that its impact on the ultimate goal of achieving a more inclusive society might be less obvious than what would immediately appear. One of the observations made under this new line of work is that prior research on the contact hypothesis has paid little attention on how the minority group reacts to contact, with the focus being on how contact changes prejudice level among majority group members. In this context, Dixon et al. (2012) mention a few observational studies suggesting that, while majority group members may demand more social change towards inclusiveness subsequent to inter-group contact, minority group members may actually become less demanding of social change, as they perceive that discrimination and social injustice have lessened. A few recent studies (Saguy et al. 2009; Dovidio et al. 2009; Glasford and Calcagno 2012) provide lab results consistent with this observational data, with minority group members under the contact condition appearing lulled into believing that the majority group is more just-minded than it really is. If these effects are real, one can easily imagine how contact may backfire at the societal level, with the theoretically more powerful advocates for the minority group (for example, African-Americans at Ivy League Universities experiencing positive inter-group contact) decreasing their level of political activism. At the very least, this provocative new research in psychology suggests that future field work on the inter-group contact hypothesis should be more systematic in collecting evidence on how minority group members react to contact, and broadening the definition of a successful intervention outcome.
Socio-Cognitive De-Biasing Strategies In the absence of direct contact, is it possible to teach individuals to become less biased against the minority group? We start with the discussion of a field experiment in Rajasthan, India, which offers a cautionary tale about how easy it might be to simply tell people to overcome their stereotypes. Banerjee et al. (2013) set up a large-scale randomized experiment designed to test whether citizens can learn from others’ experiences about the quality of female leaders. This is an environment where, we have already shown, there is a large bias against the ability of women to be decision-makers. Using high-quality
210
M. Bertrand and E. Duflo
street theater troupes, they set up a street play followed by a discussion of the performance of local leaders a few weeks ahead of the 2010 panchayat (local government) election. Following up on the work we discussed previously in section “Minority Leaders and the Attitude of the Majority,” which showed that direct experience with a female leader does change attitude towards female leaders (and willingness to vote for one), this study sought to test whether the process could be accelerated by providing citizens objective information that, in fact, women and men are about equally good at carrying out a key task in the local government. The experiment took place in 382 panchayats: in randomly selected ones, a street play emphasized the importance of the local leader in making key decisions, and encouraged citizens to vote for a competent leader. It then showed information on the average performance of all leaders in providing employment under the flagship employment guarantee scheme. In another group of villages, the play and the information was almost the same, but the script of the play emphasized the fact that citizens are often biased against women leaders, but that women also can be good leaders. The statistics provided on leader performance were also disaggregated by gender (as it turns out, women do about as well as men in the sample districts). There are two main results: First, the play and information campaign, when it does not emphasize gender, does appear to move priors. More candidates enter and the incumbent is less likely to enter and to win. For example, the incumbent vote share declines by 6 percentage points (or a remarkable 60%) in villages where the general campaign was run. Moreover, the vote share for the incumbent become more sensitive to past performance in places where the gender-neutral campaign was run. Second, however, these effects disappear in places where the campaign introduces the “gender” theme: in those villages, there is very little effect of the intervention on any outcomes (including on the probability that a female runs or wins, or on the vote share for women). It is as if, when citizens understood that the campaign was about convincing them to consider women, they lost interest. These findings underscore the challenge of fighting discrimination in an environment where discrimination is rife.41 It is possible that this experiment failed because it did not pay enough attention to the structure of the bias and ways to overcome it. Over the last 20 years, social psychologists have designed and tested in the laboratory setting a series of strategies to reduce bias and stereotypical thinking. These include (following the categorization in Paluck and Green (2009)): consciousness-raising, targeting emotions through perspective-taking, targeting value consistency and self-worth, expert opinion and accountability interventions, as well as re-, de- and cross-categorization techniques. Consciousness-raising strategies are inspired by the large body of work (in particular the IAT literature) suggesting that prejudice can operate without the
41
Note that the effect of reservation in this sample on the probability that a woman runs or wins after the reservation is cancelled is still positive, as in West Bengal or Mumbai: so the results are not due to the fact that people in Rajasthan are so hell bent against women that they cannot learn about them. It just appears they cannot learn about them from this intervention.
9
Field Experiments: Correspondence Studies
211
person’s awareness or endorsement of it. The most promising consciousness-raising strategies emerging from the psychology literature to date include counter-stereotype training and approach-avoidance training. For example, in Kawakami et al. (2000), lab subjects received extensive training in negating specific stereotypical thinking towards elderly people and skinheads (young individuals with closed-cropped or shaven heads who typically wear heavy boots, are often part of the working-class, and stereotypically perceived as aggressive). In the elderly stereotype negation condition, subjects were instructed to respond “NO” on the trials in which they saw a picture of an elderly person paired with an elderly stereotypic trait and “YES” when they saw a picture of an elderly person with a non-stereotypic trait. In the skinhead stereotype negation condition, subjects were to respond “NO” on trials in which they saw a picture of a skinhead paired with a skinhead stereotypic trait and “YES” on trials in which they saw a picture of a skinhead paired with a non-stereotypic trait. Kawakami et al. (2000) show that such training in negating stereotypes was able to reduce the stereotypical activation. These results were obtained even when participants were no longer instructed to “not stereotype,” and, importantly, for stereotypic traits that were not directly involved in the negation training phase. This reduced activation level was still clearly visible 24 h following the training session.42 Dasgupta and Greenwald (2001) report on two experiments where they examined whether exposure to pictures of admired and disliked exemplars can reduce automatic preference for white over black Americans and younger over older people. In Experiment 1, participants were exposed to either admired black (e.g., Denzel Washington) and disliked white individuals (e.g., Jeffrey Dahmer), disliked black (e.g., Mike Tyson) and admired white (e.g., Tom Hanks) individuals, or nonracial exemplars. Immediately after exemplar exposure and 24 h later, they completed an IAT that assessed automatic racial attitudes and two explicit attitude measures. Exposure to admired black and disliked white exemplars significantly weakened automatic prowhite attitudes for 24 h beyond the treatment but did not affect explicit racial attitudes. Experiment 2 provided a replication using automatic age-related attitudes. Also, Wittenbrink, Judd, and Park (1997) examined the effects of watching videos of African-Americans situated either at a convivial outdoor barbecue or at a gang-related incident. Situating African-Americans in a positive setting produced lower implicit bias scores. Kawakami, Dovidio, and Van Kamp (2007a) perform another lab experiment on negating stereotypical associations but focus on outcomes that are closer to those we might wish to affect in the real world. Participants first underwent gender counterstereotype training, by pairing male faces with words like “sensitive” and female faces with words like “strong.” They next evaluated four applications (résumés and cover letters) for a position as “chairperson of a District Doctor’s Association.” All of the applicants were qualified, but two had male names and two had female names
42
Two follow-up studies outside of Kawakami’s lab have partially replicated but partially qualified the original findings. See Gawronski et al. (2008).
212
M. Bertrand and E. Duflo
(counterbalanced so that half the participants saw a particular résumé with a male name and the other half saw that same résumé with a female name). The evaluation of applicants involved two separate stages: judging the applicants along 16 different dimensions (eight stereotypically masculine traits like “risk-taker” and eight feminine traits like “helpful”) and then simply choosing the best candidate. Some participants made the trait judgments first and chose the best candidate second, while other participants completed the two tasks in the opposite order. Among participants who had received no training, only 35% chose a woman for the job. In contrast, among participants who had undergone counter-stereotype training, 61% chose a woman.43 Kawakami et al. (2007b) also found that participants can change their implicit biases and unreflective social behaviors through “approach-avoidance” conditioning. In this study, white and Asian participants repeatedly pulled a joystick toward themselves when they saw black faces and pushed it away when they saw white faces. In pulling the joystick in, it was as if participants were bringing the perceived image closer, or “approaching” it. This training significantly reduced participants’ implicit bias on the IAT. The same “approach-avoidance” conditioning training has also been shown to be promising (in the lab) to deal with the stereotype threat. Kawakami et al. (2008) report the beneficial effects for female undergraduates of repeatedly “approaching” math-related images (“e.g., calculators, equations”). After the training, those who initially reported that they did not like math and were not good at it tended to identify with and prefer math on implicit measures, as well as to answer more questions on a math test. In a series of follow-up studies, Forbes and Schmader (2010) replicated Kawakami et al. (2008), but built a longer delay (24–30 h) between the de-biasing training and the math test, and also compared the relative effectiveness of approach-avoidance training with counter-stereotype training. They found that gender-math counter-stereotype training seemed more effective than approach-avoidance training. Women trained to associate the phrase “women are good at” with math-related words exhibited increased working memory as well as improved performance on math questions from the GRE (a graduate-level standardized test). Because emotional states can influence the expression of prejudice, psychologists have hypothesized that interventions that encourage the perceiver to experience the emotions of the minority group might be effective de-biasing strategies. What does it feel like to have your intelligence automatically questioned, or to be trailed by
43
Interestingly, these effects were only observed when the task of choosing the best candidate came second, after the trait evaluation. When this choice task was first, only 37% of those who had undergone the training chose a female candidate. A similar pattern emerged when the order of the tasks was switched, in that participants were consistently biased on the first task and de-biased on the second, regardless of which task actually came first. One possible explanation for this effect is that participants seem to recognize that the researchers are trying to debias them, and then try to correct for this perceived influence by deliberately responding in more stereotypical ways, at least at first. Once they have an opportunity to explicitly counteract the debiasing, they stop trying to resist the training and then the effects emerge. Subsequently, they respond in counterstereotypical ways.
9
Field Experiments: Correspondence Studies
213
detectives each time you walk into a store? Perspective-taking involves stepping into the shoes of a stereotyped person, and can be useful in assessing the emotional impact on individuals who are constantly being stereotyped in negative ways. There are now multiple studies attesting to the merits of perspective taking as a strategy for reducing intergroup bias. Some have linked perspective-taking to decreased activation and application of negative group stereotypes (Galinsky and Moskowitz 2000; Todd et al. 2011); others have shown that adopting the perspective of a particular out-group target leads to more positive evaluations of other individual members of the target’s group (Shih et al. 2009) and of the target’s group as a whole (Stephan and Finlay 1999). For example, Todd et al. (2011) conducted a series of lab experiments examining the impact of perspective-taking on several outcomes: automatic evaluations, approach-avoidance reactions, and behaviors displayed during face-to-face interactions. In one of the experiments, participants watched a video depicting a series of discriminatory acts directed toward a black man versus a white man. As they watched the video, participants either adopted the black man’s perspective or they attempted to remain objective and detached (control group). The researchers included two different perspective-taking conditions in this experiment. Some participants tried to imagine the black man’s thoughts, feelings, and experiences (other condition) as they watched the video; others tried to imagine their own thoughts, feelings, and experiences as if they were in the black man’s situation (self condition). After watching the video, participants completed a variant of the IAT that assesses automatic evaluations of black relative to white Americans. Subjects in both of the perspective-taking conditions (other and self conditions) exhibited significantly weaker pro-white bias than the control subjects. Strategies targeting value-consistency and self-worth rely on the theory that individuals’ desire to maintain consistency between valued cognitions and behaviors or protect their self-worth may be leveraged to lead them to repress their prejudice (Paluck and Green 2009). De-biasing strategies in this area have leveraged cognitive dissonance and self-affirmation theories. For example, in a lab experiment, Leippe and Eisenstadt (1994) apply cognitive dissonance theory to get subjects to see prejudice as inconsistent with their own values: College students softened their antiblack positions on social policies and reported more egalitarian attitudes and beliefs after agreeing to write a public statement in favor of problack policies. Also, Fein and Spencer (1997) report that lab subjects who have “self-affirmed” by circling values that were most important to them were more likely to give positive ratings to a Jewish job candidate.44 A body of research in social psychology suggests that prejudice and discrimination might also be influenced by expert opinion and greater accountability to others for one’s beliefs and behaviors. Levy, Stroessner, and Dweck (1998) show that telling lab subjects that experts believe that personality is malleable reduces stereotyping against minority groups. Dobbs and Crano (2001) report that subjects allocated more points to a fictitious out-group when they were required to justify their 44
Note that participants who were Jewish were excluded from this part of the study.
214
M. Bertrand and E. Duflo
allocations to others; similarly, Bodenhausen, Kramer, and Süsser (1994) show that students involved in a school disciplinary case were less likely to stereotype the student if they believed they would be accountable to their peers for their evaluation of the case. Individuating is another socio-cognitive de-biasing strategy that involves gathering very specific information about a person’s background, tastes, hobbies, and family, so that one’s judgments will be based on the particulars of that person, rather than on group characteristics. This approach is grounded in the social identity and categorization literatures and essentially is a de-categorization effort, where subjects are instructed to focus on the individual rather than the group (Brewer 1988; Fiske and Neuberg 1990). Lebrecht et al. (2009) provide an interesting take on the individuation exercise. In their study, two groups of Caucasian subjects were exposed equally to the same African-American faces in a training protocol run over five sessions. In the individuation condition, subjects learned to discriminate between African-American faces; specifically, they received “expertise training” with other-race faces – defined by the authors as a procedure that improves observers’ ability to individuate objects within the training domain and hence reduce the degree to which other-race faces are stereotyped. In contrast, in the categorization condition, subjects learned to categorize faces as African-American or not. Subjects in the individuation condition, but not in the categorization condition, showed improved discrimination of African-American faces with training. Also, subjects in the individuation condition, but not the categorization condition, showed a reduction in their implicit racial bias. For the individuation condition only, the degree to which an individual subject’s implicit racial bias decreased was significantly correlated with the degree of improvement that the subject showed in their ability to differentiate African-American faces. Other de-biasing strategies inspired by the social identity and categorization literatures include re-categorization and crossed-categorization techniques, where participants are encouraged to think of people from different groups as part of one subordinate group using cues such as same shirt colors or shared prizes, or participants are made aware of their common membership in a third group. Such re-categorization and cross-categorization efforts have shown some success in reducing favoritism for the in-group and improving cooperation between groups (Dovidio and Gaertner 2000; Gaertner et al. 1999). An exciting recent study in the socio-cognitive de-biasing area is Lai et al. (2014), who sought to determine the effectiveness of various methods for reducing implicit bias. Structured as a research contest, teams of scholars were given 5 min in which to enact interventions that they believed would reduce implicit preferences for whites compared to blacks, as measured by an IAT, with the goal of attaining IAT scores that reflect a lack of implicit preference for either of the two groups. Teams submitted 18 interventions that were tested approximately two times across three studies, totaling 11,868 nonblack participants. Half of the interventions were effective at reducing the implicit bias that favors whites over blacks. Most effective were the following interventions: (1) participating in a sports game in which all of the teammates were black while the opposing team was all-white and engaged in unfair
9
Field Experiments: Correspondence Studies
215
play, and being subsequently instructed to recall how their black teammates helped them while their white opponents did not; (2) reading a graphic story in which one is asked to place oneself in the role of the victim who is assaulted by a white man and rescued by a black man; (3) practicing an IAT with counterstereotypic Black (e.g., Michael Jordan, Martin Luther King, Jr.) and counterstereotypic White (e.g., Timothy McVeigh, Jeffrey Dahmer) exemplars. A concern one may have about the relevance of this lab evidence for the field is that it can only document fairly short-term effects (up to 24 h), and hence might be of limited relevance to the real world. However, even such a short time frame might be relevant to some important decisions that have been shown to be subject to bias, such as human resource managers’ decision on whether to pass on a given résumé, or teachers’ grading decisions. Therefore, we believe that even short-term effects could be of real-world relevance. What this lab evidence does not allow us to assess, however, is how these shortterm impacts would differ if the same person (e.g., an HR manager) was repeatedly exposed to such de-biasing strategies (e.g., every time he or she sits down to start reviewing résumés, or grading exams). Some other de-biasing work in psychology has taken seriously this concern about one-shot, short-term interventions and has asked whether related strategies can be built to produce enduring reductions in bias. Work by Devine and a series of co-authors is of particular interest. Devine (1989) proposes a habit-breaking approach to prejudice reduction and likens implicit biases to deeply entrenched habits developed through socialization experiences. “Breaking the habit” of implicit bias therefore requires learning about the contexts that activate the bias and how to replace the biased responses with responses that reflect one’s non-prejudiced goals. Devine and colleagues (Devine and Monteith 1993; Plant and Devine 2009) argue that the motivation to break the prejudice habit stems from two sources. First, people must be aware of their biases and they must also be concerned about the consequences of their biases before they will be motivated to exert effort to eliminate them. Second, people need to know when biased responses are likely to occur and how to replace those biased responses with ones more consistent with their goals. Devine et al. (2012) develop and test a longer-term intervention to help people reduce implicit biases and “break the prejudice habit.” The participants were 91 nonblack introductory psychology students, who completed a 12-week longitudinal study for course credit. The key elements of the intervention were as follows. First, to ensure awareness of their bias, all participants completed a measure of implicit bias and received feedback about their level of bias. People assigned to the treatment group were also presented with a bias education and training program, the goals of which were to evoke a general concern about implicit biases and train people to eliminate them. The program lasted 45 min. The education component likened the expression of implicit biases to a habit and provided information linking implicit bias to discriminatory behaviors across a wide range of settings (e.g., interpersonal, employment, and health). The training component described how to apply a variety of bias reduction strategies in daily life. The training section presented participants with a wide array of strategies (covering many of the
216
M. Bertrand and E. Duflo
strategies discussed below, such as taking the perspective of stigmatized others, imagining counterstereotypic examples, training in negating stereotypical associations and individuation) as well as opportunities to engage in positive interactions with members of the minority group (e.g., inter-group contact). This enabled participants to flexibly choose the strategies most applicable to different situations in their lives. Following the intervention, treated participants had lower IAT scores than control group participants at 4 and 8 weeks after the intervention; moreover, the effects at 4 and 8 weeks were not systematically different from each other, indicating that the reduction in implicit race bias persisted over time. These data provide the first evidence that a controlled, randomized intervention can produce enduring reductions in implicit bias. The intervention created no changes in the participants’ reported racial attitudes, but it did affect participants’ concern about discrimination and their awareness of their personal bias. Also, concerns about discrimination emerged as a moderator for the interventions’ effects. The intervention appears to have raised concerns about discrimination at week 2, and the biggest reduction in implicit bias in the treatment group was among those subjects who experienced growing concerns. Despite the large amount of both theoretical and lab-based work in psychology on these various socio-cognitive de-biasing techniques, it is remarkable how few evaluations of these techniques have been performed in the field. Paluck and Green (2009) perform a thorough literature search of the randomized field evidence on the de-biasing techniques listed above. While the number of field experiments they identify is non-trivial (71), much of the work they survey is not directly guided by the psychology literature or directory transposable into the specific lab-based strategies reviewed above. Moreover, very few of the existing field studies are designed to track changes in behavior outcome measures. The modal existing field study also involves a very short-term follow-up (often within the day) and takes place in a classroom setting with a student population, hence quite “lablike” even if not explicitly in the lab. By far most common have been interventions relying on various forms of entertainment (books, movies, cartoons, etc.) to create a persuasive narrative aimed at altering stereotypical thinking. In many cases though, the entertainment content is not based on the specific psychological theories that have guided the lab work, and it is hence difficult to make a direct link between the lab and the field evidence. For example, Paluck and Green (2009) identify several randomized field experiments performed in schools to measure the impact of reading on prejudice. This work suggests reduction in self-reported bias associated with reading content that portrays contact between children who are similar to the studied population and children of different race (e.g., intergroup friendship), as well as reading content that emphasizes a minority characters’ individual characteristics rather than group membership (e.g., individuating). But is also possible that reading interventions might be effective because of the emotional reaction they induce through perspective-taking (e.g., putting oneself in the shoes of the minority character in a book), or because they are a channel to communicate social norms (e.g., descriptions of what others are doing and hence what the reader should do).
9
Field Experiments: Correspondence Studies
217
Paluck and Green (2009) also identify a few instruction-based (rather than narrative-based) field interventions. In this case again, though, the content of the interventions is rarely directly guided by the lab evidence, and a lack of theoretical foundations may explain in part a lack of impressive findings. One exception is Lustig (2003), which evaluates a training program in Israel that aims to encourage perspective-taking and empathy to reduce prejudice against Palestinians among Jewish twelfth graders. Lustig (2003) reports encouraging findings among Jewish students who were asked to write an essay about the Israeli-Palestinian conflict from the Palestinian viewpoint. The randomized field studies designed to directly test consciousness-raising, value consistency and self-worth, as well as re-, de- and cross-categorization techniques can essentially be counted on one hand.45 All have been performed on student populations and have produced mixed results. At the same time, we are confident that hundreds of anti-prejudice interventions directly inspired by the lab-based literature described above must be taking place yearly not only in schools but also in business and government settings, but are not being rigorously evaluated. It would be of first-order importance for researchers to strike up partnerships with organizations interested in better understanding the value of the diversity training programs they are investing resources in, both in terms of their immediate impact on bias and their ultimate impact on organizational performance. Human resource departments, police departments, and courtrooms are only a few of the possible real-world settings where a much-needed field validation of this large lab-based literature could be performed. For example, the U.S. Department of Justice is funding the development of a curriculum for police staff that reflects on the Fair and Impartial Policing perspective. This training program applies the modern science of bias to policing: it trains officers on the effect of implicit bias and gives them the information and skills they need to reduce and manage their biases. The curriculum addresses not just racial/ ethnic bias, but biases based on other factors such as gender, sexual orientation, religion, socio-economic status, etc. Officers are taught skills, inspired by the lab-tested methods described above to reduce and manage their own biases. Social psychologists from around the nation who conduct the research on human biases are members of the team that help design the curriculum. While this program has been implemented with various target audiences (recruits/patrol officers, first line supervisors, mid-level managers, command staff and law enforcement trainers), to our knowledge it has not been the subject of a rigorous evaluation. As another example, there has been much discussion in the recent years about how the socio-debiasing techniques described above could be used to de-bias judges and jurors. Kang et al. (2012) discuss possible ways to import these techniques to the courtroom. They argue that:
45
See Houlette et al. (2004) for re-categorization, Rokeach (1971) and Rokeach (1973) for value consistency, Katz and Zalk (1978) and Katz (2000) for cognitive retraining, and Lustig (2003) for perspective-taking.
218
M. Bertrand and E. Duflo
In chambers and the courtroom buildings, photographs, posters, screen savers, pamphlets, and decorations ought to be used that bring to mind counter-typical exemplars or associations for participants in the trial process . . . for jurors, then, . . . the hope would be that by reminding them of counter-typical associations, we might momentarily activate different mental patterns while in the courthouse and reduce the impact of implicit biases on their decision-making.
Also, Elek and Agor (2014) show how de-biasing strategies could be feasibly brought to the courtroom with simple alterations to the standard instructions delivered by the judge to the jury, such as including a recognition of the universality of bias and explicit encouragement of perspective-taking.
Technological De-Biasing Stanovich and West’s (2000) distinction between System 1 and System 2 cognitive functioning provides a useful framework for organizing both what scholars have learned to date about effective strategies for improving decision-making and future efforts to uncover improvement strategies. System 1 refers to our intuitive system, which is typically fast, automatic, effortless, implicit, and emotional. System 2 refers to reasoning that is slower, conscious, effortful, explicit, and logical. People often lack important information regarding a decision, fail to notice available information, face time and cost constraints, and maintain a relatively small amount of information in their usable memory. The busier people are, the more they have on their minds, and the more time constraints they face, the more likely they will be to rely on System 1 thinking. In the many situations where we know that decision biases are likely to plague us (e.g., when evaluating diverse job candidates, estimating our percent contribution to a group project, choosing between spending and saving, etc.), relying exclusively on System 1 thinking is likely to lead us to make errors. Also, when the basis for judgment is somewhat vague (e.g., situations that call for discretion, cases that involve the application of new, unfamiliar laws, etc.), biased judgments are more likely. Without more explicit, concrete criteria for decisionmaking, individuals tend to disambiguate the situation using whatever information is most easily accessible – including stereotypes (Dovidio and Gaertner 2000; Johnson et al. 1995). Similarly, certain emotional states (anger, disgust) can exacerbate implicit bias in judgments of stigmatized group members, even if the source of the negative emotion has nothing to do with the current situation or with the issue of social groups or stereotypes more broadly (DeSteno et al. 2004; Dasgupta et al. 2009). Interestingly, and perhaps counterintuitively, happiness may also produce more stereotypic judgments, though the exact reasoning for this is unclear and the stereotypic judgements can be consciously controlled if the person is motivated to do so (Bodenhausen et al. 1994). Circumstances that are tiring (e.g., long hours, fatigue), stressful (e.g., heavy, backlogged, or very diverse caseloads; loud construction noise; threats to physical
9
Field Experiments: Correspondence Studies
219
safety; popular or political pressure about a particular decision; emergency or crisis situations), or otherwise distracting, can adversely affect judicial performance (Eells and Robert Showalter 1994; Hartley and Adams 1974; Keinan 1987). Specifically, situations that involve time pressure (Van Knippenberg et al. 1999), that force a decision-maker to form complex judgments relatively quickly (Bodenhausen and Lichtenstein 1987), or in which the decision-maker is distracted and cannot fully attend to incoming information (Gilbert and Gregory Hixon 1991; Sherman et al. 1998) all limit the ability to fully process case information. Decision-makers who are rushed, stressed, distracted, or pressured are more likely to apply stereotypes – recalling facts in ways biased by stereotypes and making more stereotypic judgments – than decisionmakers whose cognitive abilities are not similarly constrained. For instance, Correll et al. (2002) have used videogames in the lab to assess the effect of race on shoot/do not shoot decisions of targets that are either holding guns or holding nonthreatening objects. While subjects are instructed to shoot the armed targets and not shoot the unarmed targets, subjects make errors and these errors are systematically correlated with the race of the target: They disproportionately shoot unarmed blacks and do not shoot armed Whites. Subsequent work has shown this “shooter bias” to be exacerbated when respondents are tired (Ma et al. 2013), rushed (Payne 2006), or cannot see well (Payne et al. 2005). Some of these circumstances are unavoidable during actual policing. However, any staffing and scheduling steps that minimize officer fatigue could also curb some of these racial disparities. Danziger, Levav, and Avnaim-Pesso’s (2011) field study of sequential parole decisions made by experienced judges provides another interesting illustration. Their sample is 1112 parole board hearings in Israeli prisons, over a ten-month period. The rulings were made by eight Jewish-Israeli judges, with an average of 22 years of judging behind them. Every day, each judge considers between 14 and 35 cases, spending around 6 min on each decision. They take two food breaks that divide their day into three sessions. All of these details, from the decision to the times of the breaks, are duly recorded. They record the judges’ two daily food breaks, which result in segmenting the deliberations of the day into three distinct “decision sessions.” They find that the percentage of favorable rulings drops gradually from 65% to nearly zero within each decision session and returns abruptly to 65% after a break. The researchers attribute their results to the repetitive decision-making tasks draining the judges’ mental resources. When drained, the judges start suffering from “choice overload” and start opting for the easiest default choice, which is to deny the prisoner’s request. The more decisions a judge has made, the more drained he or she is, and the more likely the judge will make the default choice. Taking a break replenishes him or her. However, the researchers did not find any evidence that the timing of the decision affected discrimination: judges treated the prisoners equally regardless of their gender and ethnicity, as well as the severity of their crime. Casey et al. (2012) study how one could build on this knowledge of what triggers System 1 versus System 2 thinking to help technologically de-bias the courtroom. For example, the authors discuss how jurors might be allowed more time on cases in which implicit bias might be a concern by, for example, spending more time reviewing the facts of the case before committing to a decision; similarly, courts
220
M. Bertrand and E. Duflo
may review areas in which judges and other decision-makers are likely to be overburdened and consider options (e.g., reorganizing court calendars) for modifying procedures to provide more time for decision-making. Also, jurors may be asked to commit to decision-making criteria before reviewing case-specific information and courts may develop protocols that identify potential sources of ambiguity. Furthermore, courtrooms could consider the pros (e.g., more understanding of issues) and cons (e.g., familiarity may lead to less deliberative processing) of using judges with special expertise to handle cases with greater ambiguity. A lot of the possible strategies that Casey et al. (2012) discuss for the courtroom setting could naturally be applied to other real-world setting where biases in decision-making have been documented. Another strategy for moving toward System 2 thinking might be, in settings where data exists on past inputs to and outcomes from a particular decision-making process, to have decision-makers construct a linear model, or a formula that weights and sums the relevant predictor variables to reach a quantitative forecast about the outcome. Researchers have found that linear models produce predictions that are superior to those of experts across an impressive array of domains (Dawes 1971).46 In general, the use of linear or more complex machine learning models can help decision-makers avoid the pitfalls of many judgment biases, yet this method has only been tested in a small subset of the potentially relevant domains. With better knowledge of why discrimination occurs in a particular setting, it will become easier to design appropriate technological de-biasing strategies. As we discussed earlier in section “Limitations of Correspondence Studies,” Bartoš et al. (2013) convincingly demonstrate racial gaps in attention allocation by HR managers. Once they see a minority name on a résumé, they pay less attention to that résumé. These findings confirm the merit of requiring separate rankings of applicants from non-minority and minority groups (or across gender lines) followed by a comparison of leading candidates across the groups. One can think of this rule as providing quotas in the pre-selection process. We do not know of any systematic evaluation of such a strategy. Also, since the earlier a decision-maker learns a group attribute, such as name, the larger the asymmetry in attention to subsequent information such as education or qualification, the findings in Bartoš et al. (2013) strengthens the case for suppressing the signals of a group attribute during the part of the selection process. This particular technological approach has been receiving quite a lot of attention from policymakers in the recent years and has been evaluated in the field. In particular, the large number of correspondence studies have raised interest in the possibility of using “blind” hiring procedures. In some recruiting circumstances, the full hiring process can take place anonymously. Goldin and Rouse (2000) famously showed that American orchestras conducting blind auditions hired more women. In most
46
The value of linear models in hiring, admissions, and selection decisions is highlighted by research that Moore et al. (2010) conducted on the interpretation of grades by graduate admission officers.
9
Field Experiments: Correspondence Studies
221
other cases, though, only the first stage of the recruitment is made anonymous: this is the case in anonymous application procedures, such as the masking of identifying characteristics in résumés at the first selection stage. In several European countries, pilot studies of the impact of such anonymization of résumés have been conducted, including relatively large-scale field experiments in France, the Netherlands, Sweden and Germany. These experiments are summarized in Krause, Rinne, and Zimmermann (2012b). Only a subset were truly randomized and we focus our discussion on this subset.47 In 2010 and 2011, the French government initiated an experiment, which was implemented by the French public employment service. It involved about 1000 firms in eight local labor markets and it lasted in total for about 10 months (Behaghel et al. 2014). Among volunteer firms, résumés were either transmitted anonymously or nonanonymously. The experiment’s main findings can be summarized as follows. First, women benefit from higher callback rates with anonymous job applications – at least if they compete with male applicants for a job. Second, and most interestingly, migrants and residents of deprived neighborhoods suffer from anonymous job applications. Their callback rates are lower with anonymous job applications than with standard applications. This adverse effect on minority candidates is the exact opposite effect to what policymakers had hoped, and a surprising result given existing evidence from correspondence testing in France (Duguet et al. 2010), which shows discrimination against minority candidates for some jobs, no discrimination for others, but never discrimination against majority candidates. Behaghel, Crépon, and Le Barbanchon (2014) explain these surprising results by the selfselection of firms that agreed to participate in the field experiment. Among firms that were contacted to participate in the experiment, 62% accepted the invitation. While participating firms were very similar to refusing firms in most observable dimensions, there was one significant exception: Participating firms tended to interview and hire relatively more minority candidates (when using standard résumés). The anonymization therefore prevented selected firms from treating minority candidates more favorably during the experiment. Hence, the results of the experiment cannot be viewed as representative of what anonymization might have achieved if it had been mandated to all firms. Methodologically, this paper offers a valuable illustration of one danger when trying to generalize the findings of a field experiment. External 47
Åslund and Skans (2012) analyze an experiment conducted in parts of the local administration of the Swedish city of Gothenburg between 2004 and 2006. Based on a difference-in-differences approach, the authors find that anonymous job applications increase the chances of an interview invitation for both women and applicants of non-Western origin when compared to standard applications. These increased chances for minority candidates in the first stage also translated into a higher job offer arrival rate for women, but not for migrants. In the Netherlands, two experiments took place in the public administration of one major Dutch city in 2006 and 2007 (Bøg and Kranendonk 2011). The experiments focused on ethnic minorities. The lower call-back rate for minority candidates with standard applications disappears with anonymous job applications. With regard to job offers, however, the authors do not detect any differences between minority and majority candidates – irrespective of whether or not their résumés are treated anonymously.
222
M. Bertrand and E. Duflo
validity is far from guaranteed if there is sizable room for selection or self-selection of subjects into the experiment (Heckman 1992; Allcott 2015). Another large-scale randomized field experiment took place in Germany in early 2010 (Krause et al. 2012a). The publication of a correspondence testing study for Germany (Kaas and Manger 2012) triggered a lively public debate about discrimination in the hiring decisions of German firms.48 Against this background, the Federal Anti-Discrimination Agency initiated a field experiment with anonymous job applications in Germany to investigate their potential in combating hiring discrimination. This experiment was also subject to selection in participation, with eight organizations voluntarily joining the experiment. The characteristics that were made anonymous include the applicant’s name and contact details, gender, nationality, date and place of birth, disability, marital status and the applicant’s picture.49 Unlike the French study, the authors find that the anonymization leads to less discrimination against minority groups. Moreover, anonymizing applications is not too difficult administratively, with standardized application forms that are completed by the applicants appearing as the most effective and efficient way to make applications anonymous.
Conclusion We have organized this chapter along three overarching themes: the measurement of discrimination, the consequences of discrimination, and factors and policies that may help undermine it. It is apparent from our review of the existing field experiments under each of these themes that there remain more unanswered or unexplored questions than there are settled ones. By far the bulk of the field experiments that have been conducted in this area relate to the measurement of discrimination using the correspondence method. This body of work has demonstrated how remarkably pervasive the differential treatment of minority groups is throughout the world (at least in the labor market and rental market). These studies, most often focusing on a single minority group in a single country, have been important in generating debates in the local media and local public opinion and, from that perspective, each has added value. In many fields of inquiry, researchers shy away from replication, but this is refreshingly not the case here – most likely because demonstrating differential treatment in the given country The study finds that applicants with a Turkish-sounding name are on average 14 percentage points less likely to receive an invitation for a job interview than applicants with a German-sounding name who are otherwise similar. In small- and medium-sized firms, this difference is even larger and amounts to 24 percentage points. 49 The study was further designed to assess the practicality of different methods to remove identifiers from applications; practicality was assessed from interviews with the HR specialists at the firms. Four methods were considered: (1) standardized application forms in which sensitive information is not included; (2) refinements of existing online application forms such that sensitive information is disabled; (3) copying applicant’s nonsensitive information into another document; and (4) blacking out sensitive information in the original application documents. 48
9
Field Experiments: Correspondence Studies
223
seemed a sufficiently important goal. On the other hand, researchers’ ability to push the correspondence method further to go beyond pure measurement of differential treatment has been more limited. Disappointingly, there has been minimal methodological innovation in the way correspondence studies are being carried out. The main innovation might have been in leveraging the method to study differential treatment across other characteristics than race, gender or ethnicity, such as in the set of recent studies using the method to study discrimination against the long-term unemployed. While one might conclude from this that the correspondence method might have reached its full potential, recent papers such as Bartoš et al. (2013) which demonstrate how it can be used to study the dynamics of discrimination (endogenous attention allocation in this case) suggest there remain unexplored avenues for more creative uses. Perhaps because so much of economists’ attention has been devoted to using field experiments to measure the extent of discrimination, there has been much less activity in designing creative ways to better document either its consequences or ways to undermine it. The dearth of field-based evidence on these last two themes is particularly striking given the rich theoretical and lab-based literatures (mainly in psychology) that such work could build upon. On the topic of consequence of discrimination, we are heartened to see a few recent papers such as Glover et al. (2015) that develop a creative field design to demonstrate how discrimination can be self-perpetuating. We believe that the last theme in our chapter, interventions to undermine discrimination, is particularly ripe for more field experimentation. It is striking that most of the research in economics on this question has centered around the contact hypothesis and exposure effects, while so many other strategies to de-bias people have been proposed by psychologists and evaluated in the lab. We strongly encourage researchers to take on this work in the near future. Creating more partnerships with organizations that are willing to provide the testing ground for different de-biasing strategies will be particularly useful for this work to move forward. More generally, while field experiments in the last decade have been instrumental in documenting the prevalence of discrimination, field experiments in the future decade should aim to play as large of a role in isolating effective methods to combat it.
Cross-References ▶ Evidence of Covariation Between Regional Implicit Bias and Socially Significant Outcomes in Healthcare, Education, and Law Enforcement ▶ Stereotypes and the Administration of Justice
References Abrevaya J, Hamermesh DS (2012) Charity and favoritism in the field: are female economists nicer (to each other)? Rev Econ Stat 94(1):202–207 Acquisti A Fong CM (2013) An experiment in hiring discrimination via online social networks. Available at SSRN: http://ssrn.com/abstract¼2031979 or https://doi.org/10.2139/ssrn.2031979
224
M. Bertrand and E. Duflo
Ahern KR, Dittmar AK (2012) The changing of the boards: the impact on firm valuation of mandated female board representation. Q J Econ 127(1):137–197 Ahmed AM, Hammarstedt M (2008) Discrimination in the rental housing market: a field experiment on the internet. J Urban Econ 64(2):362–372 Ahmed AM, Hammarstedt M (2009) Detecting discrimination against homosexuals: evidence from a field experiment on the internet. Economica 76(303):588–597 Ahmed AM, Andersson L, Hammarstedt M (2010) Can discrimination in the housing market be reduced by increasing the information about the applicants? Land Econ 86(1):79–90 Ahmed AM, Andersson L, Hammarstedt M (2012) Does age matter for employability? A field experiment on ageism in the Swedish labour market. Appl Econ Lett 19(4):403–406 Ahmed AM, Andersson L, Hammarstedt M (2013) Are Gay men and lesbians discriminated against in the hiring process? South Econ J 79(3):565–585 Aigner DJ, Cain GG (1977) Statistical theories of discrimination in labor markets. Ind Labor Relat Rev 30:175–187 Alesina A, La Ferrara E (2005) Ethnic diversity and economic performance. J Econ Lit 43:762–800 Allcott H (2015) Site selection bias in program evaluation. Q J Econ 130(3):1117 Allport GW (1954) The nature of prejudice. Addison-Wesley, Cambridge, MA Altonji JG, Blank RM (1999) Chapter 48: race and gender in the labor market. In: Handbook of labor economics, vol 3(C). Elsevier, Amsterdam, pp 3143–3259 Arkes HR, Tetlock PE (2004) Attributions of implicit prejudice, or “would Jesse Jackson ‘fail’ the implicit association test?”. Psychol Inq 15(4):257–278 Aronson J, Fried CB, Good C (2002) Reducing the effects of stereotype threat on African American college students by shaping theories of intelligence. J Exp Soc Psychol 38(2):113–125 Arrow KJ (1973) The theory of discrimination. Discrim Labor Markets 3(10):3–33 Åslund O, Skans ON (2012) Do anonymous job application procedures level the playing field? Ind Labor Relat Rev 65(1):82–107 Ayres I, Siegelman P (1995) Race and gender discrimination in bargaining for a new car. Am Econ Rev 85(3):304–321 Babad EY, Inbar J, Rosenthal R (1982) Pygmalion, Galatea, and the golem: investigations of biased and unbiased teachers. J Educ Psychol 74(4):459 Baert S, Cockx B, Gheyle N, Vandamme C (2013) Do employers discriminate less if vacancies are difficult to fill? Evidence from a field experiment. IZA discussion paper, pp 1–30 Bagues MF, Esteve-Volart B (2010) Can gender parity break the glass ceiling? Evidence from a repeated randomized experiment. Rev Econ Stud 77(4):1301–1328 Bagues M, Perez-Villadoniga MJ (2012) Do recruiters prefer applicants with similar skills? Evidence from a randomized natural experiment. J Econ Behav Organ 82(1):12–20 Bagues M, Zinovyeva N (2015) The role of connections in academic promotions. Am Econ J Appl Econ 7(2):264–292 Bagues M, Sylos-Labini M, Zinovyeva N (2014) Do gender quotas pass the test? Evidence from academic evaluations in Italy. Scuola Superiore Sant’Anna, LEM working paper series, 14 Bailey J, Wallace M, Wright B (2013) Are Gay men and lesbians discriminated against when applying for jobs? A Four-City, internet-based field experiment. J Homosex 60(6):873–894 Baldini M, Federici M (2011) Ethnic discrimination in the Italian rental housing market. J Hous Econ 20(1):1–14 Banaji MR, Greenwald AG (1995) Implicit gender stereotyping in judgments of fame. J Pers Soc Psychol 68(2):181 Banaji M, Nosek BA, Greenwald AG (2004) No place for nostalgia in science: a response to Arkes and Tetlock. Psychol Inq 15(4):279–310 Banerjee A, Pande R (2009) Parochial politics: ethnic preferences and politician corruption. CEPR discussion paper no. DP6381 Banerjee A, Bertrand M, Datta S, Mullainathan S (2009) Labor market discrimination in Delhi: evidence from a field experiment. J Comp Econ 37(1):14–27
9
Field Experiments: Correspondence Studies
225
Banerjee A, Green D, Green J, Pande R (2010) Can voters be primed to choose better legislators? Experimental evidence from rural India. Unpublished manuscript, available at http://www. povertyactionlab.org/node/2764 Banerjee A, Duflo E, Imbert C, Pande R (2013) Entry, exit, and candidate selection: evidence from India. Mimeo, 3ie Grantee final report Bardhan PK, Mookherjee D, Torrado MP (2010) Impact of political reservations in West Bengal local governments on anti-poverty targeting. J Global Dev 1(1) Bartoš V, Bauer M, Chytilová J, Matějka F (2013) Attention discrimination: theory and field experiments. CERGE working paper. ISSN 1211–3298 Beaman L, Chattopadhyay R, Duflo E, Pande R, Topalova P (2009) Powerful women: does exposure reduce bias? Q J Econ 124(4):1497–1540 Beaman L, Duflo E, Pande R, Topalova P (2010) Political reservation and substantive representation: evidence from Indian Village councils. In: Berry S, Bosworth B, Panagariya A (eds) India Policy Forum 2010–11, vol 7. Sage Beaman L, Duflo E, Pande R, Topalova P (2012) Female leadership raises aspirations and educational attainment for girls: a policy experiment in India. Science 335(6068):582–586 Becker GS (1957) The economics of discrimination. University of Chicago Press Behaghel L, Crépon B, Le Barbanchon T (2014) Unintended effects of anonymous resumes. CEPR discussion paper no. DP10215 Benjamin DJ, Choi JJ, Joshua Strickland A (2010) Social identity and preferences. Am Econ Rev 100:1913–1928 Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. Am Econ Rev 94(4):991–1013 Bertrand M, Schoar A (2003) Managing with style: the effect of managers on firm policies. Q J Econ 118(4):1169–1208 Bertrand M, Chugh D, Mullainathan S (2005) Implicit discrimination. Am Econ Rev 95:94–98 Besley T, Coate S (1997) An economic model of representative democracy. Q J Econ 112:85–114 Besley TJ, Folke O, Persson T, Rickne J (2013) Gender quotas and the crisis of the mediocre man: theory and evidence from Sweden. IFN working paper Betz DE, Sekaquaptewa D (2012) My fair physicist? Feminine math and science role models demotivate young girls. Soc Psychol Personal Sci 3(6):738–746 Bhavnani RR (2009) Do electoral quotas work after they are withdrawn? Evidence from a natural experiment in India. Am Polit Sci Rev 103(01):23 Biddle JE, Hamermesh DS (1998) Beauty, productivity and discrimination: Laywers’ looks and lucre. J Labor Econ 15:172–201 Blackwell LS, Trzesniewski KH, Dweck CS (2007) Implicit theories of Intelligience predict achievement across an adolescent transition: a longitudinal study and an intervention. Child Dev 78(1):246–263 Blommaert L, Coenders M, van Tubergen F (2014) Discrimination of Arabic- named applicants in The Netherlands: an internet-based field experiment examining different phases in online recruitment procedures. Soc Forces 92(3):957–982 Bodenhausen GV, Lichtenstein M (1987) Social stereotypes and information-processing strategies: the impact of task complexity. J Pers Soc Psychol 52(5):871 Bodenhausen GV, Kramer GP, Süsser K (1994) Happiness and stereotypic thinking in social judgment. J Pers Soc Psychol 66(4):621 Bøg M, Kranendonk E (2011) Labor market discrimination of minorities? Yes, but not in job offers. MPRA paper no. 33332 Boisjoly J, Duncan GJ, Kremer M, Levy DM, Eccles J (2006) Empathy or antipathy? The impact of diversity. Am Econ Rev 96(5):1890–1905 Booth A, Leigh A (2010) Do employers discriminate by gender? A field experiment in femaledominated occupations. Econ Lett 107(2):236–238 Booth AL, Leigh A, Varganova E (2011) Does ethnic discrimination vary across minority groups? Evidence from a field experiment. Oxf Bull Econ Stat 74(4):547–573
226
M. Bertrand and E. Duflo
Bosch M, Carnero A, Farré L (2010) Information and discrimination in the rental housing market: evidence from a field experiment. Reg Sci Urban Econ 40(1):11–19 Boschini A, Muren A, Persson M (2009) Constructing gender in the economics lab. Technical report. Stockholm University, Department of Economics Brewer MB (1981) Ethnocentrism and its role in interpersonal trust. In: Scientific inquiry and the social sciences, vol 214. Jossey-Bass Publishers, San Francisco Brewer MB (1988) A dual process model of impression formation. In: Wyer R, Srull T (eds) Advances in social cognition, vol 1. Lawrence Erlbaum Associates, Hillsdale, pp 1–36 Broder IE (1993) Review of NSF economics proposals: gender and institutional patterns. Am Econ Rev 83:964–970 Brown C, Gay P (1985) Racial discrimination: 17 years after the act. Policy Studies Institute Burns J, Corno L, La Ferrara E (2015) Interaction, prejudice and performance. Evidence from South Africa. Working paper Bursell M (2007) What’s in a name? A field experiment test for the existence of ethnic discrimination in the hiring process. SULCIS WP, 7 Carlsson M (2011) Does hiring discrimination cause gender segregation in the Swedish labor market? Fem Econ 17(3):71–102 Carlsson M, Eriksson S (2014) Discrimination in the rental market for apartments. J Hous Econ 23: 41–54 Carpusor AG, Loges WE (2006) Rental discrimination and ethnicity in names. J Appl Soc Psychol 36(4):934–952 Casey PM, Warren RK, Cheesman FL, Elek JK (2012) Helping courts address implicit bias. Technical report. National Center for State Courts, Williamsburg Charles KK, Guryan J (2008) Prejudice and wages: an empirical assessment of Becker’s “the economics of discrimination”. J Polit Econ 116(5):773–809 Chattopadhyay R, Duflo E (2004) Women as policy makers: evidence from a randomized experiment in India. Econometrica 72:1409–1443 Cheryan S, Siy JO, Vichayapai M, Drury BJ, Kim S (2011) Do female and male role models who embody STEM stereotypes hinder women’s anticipated success in STEM? Soc Psychol Personal Sci 2(6):656–664 Clots-Figueras I (2009) Are female leaders Good for education? Evidence from India. Mimeo, Universidad Carlos III de Madrid Clots-Figueras I (2011) Women in politics: evidence from the Indian states. J Public Econ 95(7–8):664–690 Coate S, Loury G (1993) Antidiscrimination enforcement and the problem of patronization. Am Econ Rev 83:92–98 Coffman KB, Coffman LC, Marzilli Ericson KM (2013) The size of the LGBT population and the magnitude of anti-gay sentiment are substantially underestimated. NBER working paper no. 19508 Cohen GL, Garcia J, Apfel N, Master A (2006) Reducing the racial achievement gap: a socialpsychological intervention. Science 313(5791):1307–1310 Cohen GL, Garcia J, Purdie-Vaughns V, Apfel N, Brzustoski P (2009) Recursive processes in selfaffirmation: intervening to close the minority achievement gap. Science 324(5925):400–403 Correll J, Park B, Judd CM, Wittenbrink B (2002) The police Officer’s dilemma: using ethnicity to disambiguate potentially threatening individuals. J Pers Soc Psychol 83(6):1314–1329 Cross H, Kenney G, Mell J, Zimmerman W (1990) Employer hiring practices: differential treatment of Hispanic and Anglo job seekers. Urban Institute Report 90–4. Urban Institute Danziger S, Levav J, Avnaim-Pesso L (2011) Extraneous factors in judicial decisions. Proc Natl Acad Sci 108(17):6889–6892 Dasgupta N, Asgari S (2004) Seeing is believing: exposure to counterstereotypic women leaders and its effect on the malleability of automatic gender stereotyping. J Exp Soc Psychol 40(5): 642–658
9
Field Experiments: Correspondence Studies
227
Dasgupta N, Greenwald AG (2001) On the malleability of automatic attitudes: combating automatic prejudice with images of admired and disliked individuals. J Pers Soc Psychol 81(5):800 Dasgupta N, DeSteno D, Williams LA, Hunsinger M (2009) Fanning the flames of prejudice: the influence of specific incidental emotions on implicit prejudice. Emotion 9(4):585 Dawes RM (1971) A case study of graduate admissions: application of three principles of human decision making. Am Psychol 26:180–188 Dee TS (2009) Stereotype threat and the student-athlete. NBER working paper no. 14705 DeSteno D, Dasgupta N, Bartlett MY, Cajdric A (2004) Prejudice from thin air: the effect of emotion on automatic intergroup attitudes. Psychol Sci 15(5):319–324 Devine PG (1989) Stereotypes and prejudice: their automatic and controlled components. J Pers Soc Psychol 56:5–18 Devine PG, Monteith MJ (1993) The role of discrepancy-associated affect in prejudice reduction. In: Mackie DM, Hamilton DL (eds) Affect, cognition, and stereotyping: interactive processes in group perception. Academic, San Diego, pp 317–344 Devine PG, Forscher PS, Austin AJ, Cox WTL (2012) Longterm reduction in implicit race bias: a prejudice habit-breaking intervention. J Exp Soc Psychol 48(6):1267–1278 Dickson ES, Scheve K (2006) Social identity, political speech, and electoral competition. J Theor Polit 18(1):5–39 Dixon J, Levine M, Reicher S, Durrheim K (2012) Beyond prejudice: are negative evaluations the problem and is getting us to like one another more the solution? Behav Brain Sci 35(6):411–425 Dobbs M, Crano WD (2001) Outgroup accountability in the minimal group paradigm: implications for aversive discrimination and social identity theory. J Pers Soc Psychol 27:355–364 Doleac JL, Stein LCD (2013) The visible hand: race and online market outcomes. Econ J 123(572): F469–F492 Dovidio JF (2001) On the nature of contemporary prejudice: the third wave. J Soc Issues 57(4): 829–849 Dovidio JF, Gaertner SL (2000) Aversive racism and selection decisions: 1989 and 1999. Psychol Sci 11(4):315–319 Dovidio JF, Gaertner SL, Isen AM, Rust M, Guerra P (1998a) Positive affect, cognition, and the reduction of intergroup bias. In: Sedikides C, Schopler J, Insko CA (eds) Intergroup cognition and intergroup behavior. Lawrence Erlbaum Associates, Mahwah, pp 337–366 Dovidio JF, Gaertner SL, Validzic A (1998b) Intergroup bias: status, differentiation, and a common in-group identity. J Pers Soc Psychol 75(1):109 Dovidio JF, Gaertner SL, Saguy T (2009) Commonality and the complexity of “we”: social attitudes and social change. Personal Soc Psychol Rev 13(1):3–20 Droitcour J, Caspar RA, Hubbard ML, Ezzati TM (1991) The item count technique as a method of indirect questioning: a review of its development and a case study application. In: Beimer PB, Groves RM, Lyberg LE, Mathiowetz NA, Sudman S (eds) Measurement errors in surveys. Wiley, Hoboken, pp 185–211 Duguet E, Leandri N, L’Horty Y, Petit P (2010) Are young French jobseekers of ethnic immigrant origin discriminated against? A controlled experiment in the Paris area. Ann Econ Stat 99–100: 187–215 Dunning T, Nilekani J (2013) Ethnic quotas and political mobilization: caste, parties, and distribution in Indian Village councils. Am Polit Sci Rev 107(01):35–56 Dusek JB, Hall VC, Meyer WJ (1985) Teacher expectations. Lawrence Erlbaum Associates, Hillsdale Eagly AH, Karau SJ (2002) Role congruity theory of prejudice toward female leaders. Psychol Rev 109(3):573 Eagly AH, Makhijani MG, Klonsky BG (1992) Gender and the evaluation of leaders: a metaanalysis. Psychol Bull 111(1):3 Eden D (1992) Leadership and expectations: Pygmalion effects and other self-fulfilling prophecies in organizations. Leadersh Q 3(4):271–305
228
M. Bertrand and E. Duflo
Eden D, Ravid G (1982) Pygmalion vs. self-expectancy: effects of instructor- and self- expectancy on trainee performance. Org Behav Human Perform 30:351–364 Eden D, Shani AB (1982) Pygmalion goes to boot camp: expectancy, leadership, and trainee performance. J Appl Psychol 67(2):194 Eells TD, Robert Showalter C (1994) Work-related stress in American trial judges. J Am Acad Psychiatr Law Online 22(1):71–83 Elek JK, Agor PH (April 28 2014) Can explicit instructions reduce expression of implicit bias? New questions following a test of a specialized jury instruction. Available at SSRN: http://ssrn.com/ abstract¼2430438 Engel D, Woolley AW, Jing LX, Chabris CF, Malone TW (2014) Reading the mind in the eyes or Reading between the lines? Theory of mind predicts collective intelligence equally well online and face-to-face. PLoS One 9(12):e115212 Eriksson S, Rooth D-O (2014) Do employers use unemployment as a sorting criterion when hiring? Evidence from a field experiment. Am Econ Rev 104(3):1014–1039 Ewens M, Tomlin B, Wang LC (2014) Statistical discrimination or prejudice? A large sample field experiment. Rev Econ Stat 96(1):119–134 Fein S, Spencer SJ (1997) Prejudice as self-image maintenance: affirming the self through derogating others. J Pers Soc Psychol 73(1):31 Fiske ST, Neuberg SL (1990) A continuum of impression formation from category- based to individuating processes: influences of information and motivation on attention and interpretation. Adv Exp Soc Psychol 23:1–74 Fisman R, Iyengar SS, Kamenica E, Simonson I (2008) Racial preferences in dating. Rev Econ Stud 75:117–132 Fix M, Struyk RJ (1993) Clear and convincing evidence: measurement of discrimination in America. Urban Institute Press Forbes CE, Schmader T (2010) Retraining attitudes and stereotypes to affect motivation and cognitive capacity under stereotype threat. J Pers Soc Psychol 99(5):740 Forehand MR, Deshpandé R, Americus Reed II (2002) Identity salience and the influence of differential activation of the social self-schema on advertising response. J Appl Psychol 87(6): 1086–1099 Fryer RG, Levitt SD (2004) The causes and consequences of distinctively black names. Q J Econ 119(3):767–805 Gaertner SL, Dovidio JF, Rust MC, Nier JA, Banker BS, Ward CM, Mottola GR, Houlette M (1999) Reducing intergroup bias: elements of intergroup cooperation. J Pers Soc Psychol 76(3):388 Galarza FB, Yamada G (2014) Labor market discrimination in Lima, Peru: evidence from a field experiment. World Dev 58:83–94 Galinsky AD, Moskowitz GB (2000) Perspective-taking: decreasing stereotype expression, stereotype accessibility, and in-group favoritism. J Pers Soc Psychol 78(4):208 Galster G (1990) Racial discrimination in housing markets during the 1980s: a review of the audit evidence. J Plan Educ Res 9(3):165–175 Gawronski B, Deutsch R, Mbirkou S, Seibt B, Strack F (2008) When “just say no” is not enough: affirmation versus negation training and the reduction of automatic stereotype activation. J Exp Soc Psychol 44(2):370–377 Ghayad R (2013) The jobless trap. Job market paper. pp 1–39 Gilbert DT, Gregory Hixon J (1991) The trouble of thinking: activation and application of stereotypic beliefs. J Pers Soc Psychol 60(4):e509 Glasford DE, Calcagno J (2012) The conflict of harmony: intergroup contact, commonality and political solidarity between minority groups. J Exp Soc Psychol 48(1):323–328 Glover D, Pallais A, Pariente W (2015) Discrimination as a self-fulfilling prophecy: evidence from French grocery store. Working paper Goldberg P (1968) Are women prejudiced against women? Society 5(5):28–30 Goldin C, Rouse C (2000) Orchestrating impartiality: the impact of “blind” auditions on female musicians. Am Econ Rev 90:715–741
9
Field Experiments: Correspondence Studies
229
Golub B, Jackson MO (2012) How homophily affects the speed of learning and best response dynamics. FEEM working paper Good C, Aronson J, Inzlicht M (2003) Improving adolescents’ standardized test performance: an intervention to reduce the effects of stereotype threat. J Appl Dev Psychol 24(6):642–662 Good C, Aronson J, Harder JA (2008) Problems in the pipeline: stereotype threat and women’s achievement in high-level math courses. J Appl Dev Psychol 29(1):17–28 Gosen S (2014) Social desirability in survey research: can the list experiment provide the truth? PhD dissertation, Philipps-Universität Marburg Green AR, Carney DR, Pallin DJ, Ngo LH, Raymond KL, Iezzoni LI, Banaji MR (2007) Implicit bias among physicians and its prediction of thrombolysis decisions for black and white patients. J Intern Med 22(9):1231–1238 Green S, Green DP, Dias K, Paluck BL (forthcoming) The contact hypothesis re-examined (in progress) Greenwald AG, Banaji MR (1995) Implicit social cognition: attitudes, Selfesteem, and stereotypes. Psychol Rev 102(1):4 Greenwald AG, McGhee DE, Schwartz JLK (1998) Measuring individual differences in implicit cognition: the implicit association test. J Pers Soc Psychol 74(6):1464 Greenwald AG, Banaji M, Nosek BA (2003) Understanding and using the implicit association test: I. An improved scoring algorithm. J Pers Soc Psychol 85(2):197–216 Greenwald AG, Andrew Poehlman T, Uhlmann EL, Banaji M (2009) Understanding and using the implicit association test: III. Meta-analysis of predictive validity. J Pers Soc Psychol 97(1): 17–41 Guryan J, Charles KK (2013) Taste-based or statistical discrimination: the economics of discrimination returns to its roots. Econ J 123(572):F417–F432 Hamermesh DS, Biddle JE (1994) Beauty and the labour market. Am Econ Rev 84:1174–1194 Hamilton BH, Nickerson JA, Owan H (2003) Team incentives and worker heterogeneity: an empirical analysis of the impact of teams on productivity and participation. J Polit Econ 111(3):465–497 Hanson A, Hawley Z (2011) Do landlords discriminate in the rental housing market? Evidence from an internet field experiment in US cities. J Urban Econ 70(2–3):99–114 Hartley LR, Adams RG (1974) Effect of noise on the Stroop test. J Exp Psychol 102(1):62 Heckman J (1992) Randomization and social policy evaluation. In: Manski C, Garfinkel I (eds) Evaluating welfare and training programs. Harvard University Press, Cambridge, MA, pp 201–230 Heckman JJ (1998) Detecting discrimination. J Econ Perspect 12:101–116 Heckman JJ, Siegelman P (1993) The Urban Institute audit studies: their methods and findings. In: Fix M, Struyk R (eds) Clear and convincing evidence: measurement of discrimination in America. Urban Institute Press Heilman ME, Martell RF (1986) Exposure to successful women: antidote to sex discrimination in applicant screening decisions? Organ Behav Hum Decis Process 37(3):376–390 Hitsch GJ, Hortaçsu A, Ariely D (2010) Matching and sorting in online dating. Am Econ Rev 100: 130–163 Hjort J (2013) Ethnic divisions and productions in firms. CESifo working paper series Hoff K, Pandey P (2006) Discrimination, social identity, and durable inequalities. Am Econ Rev 96: 206–2011 Holbrook AL, Krosnick JA (2010) Social desirability in voter turnout reports: test using the item count technique. Public Opin Q 74:37–67 Hoogendoorn S, Van Praag M (2012) Ethnic diversity and team performance: a field experiment. Tinbergen Institute discussion paper 2012–068/3 Hoogendoorn S, Oosterbeek H, Van Praag M (2013) The impact of gender diversity on the performance of business teams: evidence from a field experiment. Manag Sci 59(7):1514–1528 Houlette MA, Gaertner SL, Johnson KM, Banker BS, Riek BM, Dovidio JF (2004) Developing a more inclusive social identity: an elementary school intervention. J Soc Issues 60(1):35
230
M. Bertrand and E. Duflo
Hubbuck J, Carter S (1980) Half a chance?: a report on job discrimination against young blacks in Nottingham. Commission for Racial Equality Hugenberg K, Bodenhausen GV (2004) Ambiguity in social categorization the role of prejudice and facial affect in race categorization. Psychol Sci 15(5):342–345 Iyer L, Mani A, Mishra P, Topalova P (2012) The power of political voice: Women’s political representation and crime in India. Am Econ J Appl Econ 4(4):165–193 Jacquemet N, Yannelis C (2012) Indiscriminate discrimination: a correspondence test for ethnic homophily in the Chicago labor market. Labour Econ 19(6):824–832 James F, DelCastillo SW (1991) Measuring job discrimination by private employers against young black and Hispanic seeking entry level work in Denver metropolitan area. Unpublished report. University of Colorado, Denver Johnson DW, Johnson RT (1989) Cooperation and competition: theory and research. Interaction Book Company, Edina Johnson JD, Whitestone E, Jackson LA, Gatto L (1995) Justice is still not colorblind: differential racial effects of exposure to inadmissible evidence. Personal Soc Psychol Bull 21(9):893–898 Jolson MA (1974) Employment barriers in marketing. J Mark 38 Jones BF, Olken BA (2005) Do leaders matter? National Leadership and growth since world war II. Q J Econ 120(3):835–864 Jowell R, Prescott-Clarke P (1970) Racial discrimination and white-collar workers in Britain. Race Class 11(4):397–417 Jussim L, Harber KD (2005) Teacher expectations and self-fulfilling prophecies: knowns and unknowns, resolved and unresolved controversies. Personal Soc Psychol Rev 9(2):131–155 Kaas L, Manger C (2012) Ethnic discrimination in Germany’s labour market: a field experiment. Ger Econ Rev 13(1):1–20 Kane JG, Craig SC, Wald KD (2004) Religion and presidential politics in Florida: a list experiment. Soc Sci Q 85(2):281 Kang J et al (2012) Implicit bias in the courtroom. UCLA Law Rev 59(5):1124 Karlan DS, Zinman J (2012) List randomization for sensitive behavior: an application for measuring use of loan proceeds. J Dev Econ 98(1):71–75 Katz PA (2000) Intergroup relations among youth: summary of a research workshop. Research summary. Carnegie Corp, New York Katz PA, Zalk SR (1978) Modification of children’s racial attitudes. Dev Psychol 14(5):447 Kawakami K, Dovidio JF, Moll J, Hermsen S, Russin A (2000) Just say no (to stereotyping): effects of training in the negation of stereotypic associations on stereotype activation. J Pers Soc Psychol 78(5):871 Kawakami K, Dovidio JF, Van Kamp S (2007a) The impact of counterstereotypic training and related correction processes on the application of stereotypes. Group Process Intergroup Relat 10(2):139–156 Kawakami K, Phills CE, Steele JR, Dovidio JF (2007b) (close) distance makes the heart grow fonder: improving implicit racial attitudes and interracial interactions through approach behaviors. J Pers Soc Psychol 92(6):957 Kawakami K, Steele JR, Cifa C, Phills CE, Dovidio JF (2008) Approaching math increases math¼me and math¼pleasant. J Exp Soc Psychol 44(3):818–825 Keinan G (1987) Decision making under stress: scanning of alternatives under controllable and uncontrollable threats. J Pers Soc Psychol 52(3):639 Key V (1949) Southern politics in state and nation. University of Tennessee Press Kondylis F, Mobarak M, Yishay AB, Jones M (2015) Are gender differences in performance innate or social mediated? Working paper Krause A, Rinne U, Zimmermann KF (2012a) Anonymous job applications in Europe. IZA J Eur Labor Stud 1(1):1–20 Krause A, Rinne U, Zimmermann KF (2012b) Anonymous job applications of fresh Ph. D. Economists. Econ Lett 117(2):441
9
Field Experiments: Correspondence Studies
231
Kroft K, Lange F, Notowidigdo MJ (2013) Duration dependence and labor market conditions: evidence from a field experiment. Q J Econ 128(3):1123–1167 Kuklinski JH, Cobb MD, Gilens M (1997a) Racial attitudes and the “new south”. J Polit 59(2): 323–349 Kuklinski JH, Sniderman PM, Knight K, Piazza T, Tetlock PE, Lawrence GR, Mellers B (1997b) Racial prejudice and attitudes toward affirmative action. Am J Polit Sci 41(2):402–419 Lahey JN (2008) Age, women, and hiring: an experimental study. J Hum Resour 43(1):30–56 Lai CK, Marini M, Lehr SA, Cerruti C, Shin J-EL, Joy-Gaba JA, Ho AK, Teachman BA, Wojcik SP, Koleva SP et al (2014) Reducing implicit racial preferences: I. A comparative investigation of 17 interventions. J Exp Psychol Gen 143:1765–1785 Lane KA, Banaji MR, Nosek BA, Greenwald AG (2007) Implicit measures of attitudes. In: Wittenbrink B, Schwarz N (eds) Understanding and using the implicit association test: IV. Guilford, New York, pp 59–102 Lang K (1986) A language theory of discrimination. Q J Econ 101:363–382 Lavy V, Sand E (2015) On the origins of gender human capital gaps: short and long term consequences of teachers’ stereotypical biases. NBER working paper no. 20909 Lazear E (1999) Language and culture. J Polit Econ 107(6):S95–S126 LeBoeuf RA, Shafir E, Bayuk JB (2010) The conflicting choices of alternating selves. Organ Behav Hum Decis Process 111(1):48–61 Lebrecht S, Pierce LJ, Tarr MJ, Tanaka JW (2009) Perceptual other-race training reduces implicit racial bias. PLoS One 4(1):e4215 Leippe MR, Eisenstadt D (1994) Generalization of dissonance reduction: decreasing prejudice through induced compliance. J Pers Soc Psychol 67(3):395 Levy SR, Stroessner SJ, Dweck CS (1998) Stereotype formation and endorsement: the role of implicit theories. J Pers Soc Psychol 74(6):1421–1436 List JA (2004) The nature and extent of discrimination in the marketplace: evidence from the field. Q J Econ 119(1):49–89 List JA (2009) Informed consent in social science. Science 322(5902):672 Lockwood P, Kunda Z (1997) Superstars and me: predicting the impact of role models on the self. J Pers Soc Psychol 73(1):91 Lustig I (2003) The influence of studying foreign conflicts on students’ perceptions of the IsraeliPalestinian conflict. Unpublished Masters thesis, University of Haifa Ma DS, Correll J, Wittenbrink B, Yoav Bar-Anan N, Srirarm, and Brian A. Nosek. (2013) When fatigue turns deadly: the association between fatigue and racial bias in the decision to shoot. Basic Appl Soc Psychol 35:515–524 Marit Rehavi M (2007) Sex and politics: do female legislators affect state spending? Unpublished manuscript, University of Michigan Martinez MD, Craig SC (2010) Race and 2008 president politics in Florida: a List experiment. Forum 8(2) Marx B, Pons V, Suri T (2015) Homogeneous teams and productivity. Unpublished manuscript. http://www.novasbe.unl.pt/images/novasbe/files/INOVA_Seminars/Vincent_Pons.pdf Matsa DA, Miller AR (2013) A female style in corporate leadership? Evidence from quotas. Am Econ J Appl Econ 5(3):136–169 Maurer-Fazio M (2012) Ethnic discrimination in China’s internet job board labor market. IZA J Migr 1(12):1–24 McConnell AR, Leibold JM (2001) Relations among the implicit association test, discriminatory behavior, and explicit measures of racial attitudes. J Exp Soc Psychol 37:435–442 McGinnity F, Nelson J, Lunn P, Quinn E (2009) Discrimination in recruitment. Equality research series. The Equality Authority, Dublin McIntyre S, Moberg DJ, Posner BZ (1980) Preferential treatment in preselection decisions according to race and sex. Acad Manag J 23(4):738–749 Milkman KL, Akinola M, Chugh D (2012) Temporal distance and discrimination: an audit study in academia. Psychol Sci 23(7):710–717
232
M. Bertrand and E. Duflo
Miller C (2014) The persistent effect of temporary affirmative action. Job market paper Mobius MM, Rosenblat TS (2006) Why beauty matters. Am Econ Rev 96:222–235 Moore DA, Swift SA, Sharek ZS, Gino F (2010) Correspondence bias in performance evaluation: why grade inflation works. Personal Soc Psychol Bull 36(6):843–852 Neumark D (2012) Detecting discrimination in audit and correspondence studies. J Hum Resour 47(4):1128–1157 Neumark D, Bank RJ, Van Nort KD (1996) Sex discrimination in restaurant hiring: an audit study. Q J Econ 111(3):915–941 Newman JM (1978) Discrimination in recruitment: an empirical analysis. Ind Labor Relat Rev 32(1):15–23 Nosek BA, Banaji M, Greenwald AG (2002) Harvesting implicit group attitudes and beliefs from a demonstration web site. Group Dyn Theory Res Pract 6(1):101–115 Nunley JM, Pugh A, Romero N, Alan Seals R (2014) An examination of racial discrimination in the labor market for recent college graduates: estimates from the field. Working paper Ong D, Wang J (2015) Income attraction: an online dating field experiment. J Econ Behav Organ 111(C):13–22 Oreopoulos P (2011) Why do skilled immigrants struggle in the labor market? A field experiment with thirteen thousand resumes. Am Econ J Econ Pol 3(4):148–171 Oswald F, Mitchell G, Blanton H, Jaccard J, Tetlock PE (2013) Predicting ethnic and racial discrimination: a meta-analysis of IAT criterion studies. J Pers Soc Psychol 105(2):171–192 Oz S, Eden D (1994) Restraining the golem: boosting performance by changing the interpretation of low scores. J Appl Psychol 85:314–322 Padro i Miguel G (2007) The control of politicians in divided societies: the politics of fear. Rev Econ Stud 74(4):1259–1274 Pager D (2003) The mark of a criminal record. Am J Sociol 108(5):937–975 Paluck EL, Green DP (2009) Prejudice reduction: what works? A review and assessment of research and practice. Annu Rev Psychol 60:339–367 Pande R (2003) Can mandated political representation increase policy influence for disadvantaged minorities? Theory and evidence from India. Am Econ Rev 93(4):1132–1151 Patacchini E, Ragusa G, Zenou Y (2012) Unexplored dimensions of discrimination in Europe: religion, homosexuality and physical appearance. Unpublished manuscript: http://www.frdb. org/upload/file/FRDB_Rapporto_PATACCHINI.pdf Payne BK (2006) Weapon bias: split-second decisions and unintended stereotyping. Curr Dir Psychol Sci 15:287–291 Payne BK, Shimizu Y, Jacoby LL (2005) Mental control and visual illusions: toward explaining race-biased weapon misidentifications. J Exp Soc Psychol 41(1):36–47 Petit P (2007) The effects of age and family constraints on gender hiring discrimination: A field experiment in the French financial sector. Labour Econ 14(3):371–391 Pettigrew TF, Tropp LR (2000) Does intergroup contact reduce prejudice? Recent meta-analytic findings. Reduc Prejudice Discrim 93(114) Pfann GA, Biddle JE, Hamermesh DS, Bosman CM (2000) Business success and businesses’ beauty capital. Econ Lett 67(2):201–207 Phelps ES (1972) The statistical theory of racism and sexism. Am Econ Rev 62:659–661 Plant EA, Devine PG (2009) The active control of prejudice: unpacking the intentions guiding control effects. J Pers Soc Psychol 96:640–652 Pope DG, Sydnor JR (2011) What’s in a picture? Evidence of discrimination from Prosper.com. J Hum Resour 46(1):53–92 Prendergast C, Topel R (1996) Favoritism in organizations. J Polit Econ 104:446–461 Rao G (2013) Familiarity does not breed contempt: diversity, discrimination and generosity in Delhi schools. Job market paper Reicher S, Levine M (1994) Deindividuation, power relations between groups and the expression of social identity: the effects of visibility to the out-group. Br J Soc Psychol 33(2):145–163
9
Field Experiments: Correspondence Studies
233
Reynolds D (2007) Restraining golem and harnessing pygmalion in the classroom: A laboratory study of managerial expectations and task design. Acad Manag Learn Educ 6(4):475–483 Riach PA, Rich J (1991) Testing for racial discrimination in the labour market. Camb J Econ 15: 239–256 Riach PA, Rich J (2002) Field experiments of discrimination in the market place. Econ J 112(483): F480–F518 Riach PA, Rich J (2010) An experimental investigation of age discrimination in the English labor market. Ann Econ Stat (99–100):169–185 Rokeach M (1971) Long-range experimental modification of values, attitudes, and behavior. Am Psychol 26(5):453 Rokeach M (1973) The nature of human values, vol 438. Free Press, New York Rooth D-O (2009) Obesity, attractiveness, and differential treatment in hiring: a field experiment. J Hum Resour 44(3):710–735 Rooth D-O (2010) Automatic associations and discrimination in hiring: real world evidence. Labour Econ 17(3):523–534 Rosenthal R (1963) On the social psychology of the psychological experiment: the experimenter’s hypothesis as unintended determinant of experimental results. Am Sci 51:268–283 Rosenthal R (1994) Interpersonal expectancy effects: a 30-year perspective. Curr Dir Psychol Sci 3: 176–179 Rosenthal R, Jacobson L (1968) Pygmalion in the classroom. Urban Rev 3(1):16–20 Rosenthal R, Rubin DB (1978) Interpersonal expectancy effects: the first 345 studies. Behav Brain Sci 1(3):377–386 Rudman LA, Glick P (2001) Prescriptive gender stereotypes and backlash toward agentic women. J Soc Issues 57(4):743–762 Rudman LA, Lee MR (2002) Implicit and explicit consequences of exposure to violent and misogynous rap music. Group Process Intergroup Relat 5(2):133–150 Sacerdote B (2000) Peer effects with random assignment: results for dartmouth roommates. NBER working paper no. 7469 Saguy T, Tausch N, Dovidio JF, Pratto F (2009) The irony of harmony intergroup contact can produce false expectations for equality. Psychol Sci 20(1):114–121 Sherman JW, Lee AY, Bessenoff GR, Frost LA (1998) Stereotype efficiency reconsidered: encoding flexibility under cognitive load. J Pers Soc Psychol 75(3):589 Shih M, Wang E, Bucher AT, Stotzer R (2009) Perspective taking: reducing prejudice towards general outgroups and specific individuals. Group Process Intergroup Relat 12(5):565–577 Shook NJ, Fazio RH (2008) Roommate relationships: a comparison of interracial and same-race living situations. Group Process Intergroup Relat 11(4):425–437 Slavin RE (1977) How student learning teams can integrate the desegregated classroom. Int Educ 15(6):56–58 Slavin RE (1979) Effects of biracial leaning teams on cross-racial friendships. J Educ Psychol 71: 381–387 Slavin RE (1995) Cooperative learning: theory, research, and practice, 2nd edn. Allyn & Bacon Slavin RE, Cooper R (1999) Improving intergroup relations: lessons learned from cooperative learning programs. J Soc Issues 55(4):647–663 Slavin RE, Oickle E (1981) Effects of cooperative learning teams on student achievement and race relations: treatment by race interactions. Sociol Educ 54(3):174–180 Snow RE (1995) Pygmalion and intelligence? Curr Dir Psychol Sci 4:169–171 Stanovich KE, West RF (2000) Advancing the rationality debate. Behav Brain Sci 23(5):701–717 Steele CM, Aronson J (1995) Stereotype threat and the intellectual test performance of African Americans. J Pers Soc Psychol 69(5):797–811 Stephan WG, Finlay K (1999) The role of empathy in improving intergroup relations. J Soc Issues 55(4):729–743 Streb MJ, Burrell B, Frederick B, Genovese MA (2008) Social desirability effects and support for a female American President. Public Opin Q 72(1):76–89
234
M. Bertrand and E. Duflo
Tajfel H (1970) Experiments in intergroup discrimination. Sci Am 223(5):96–102 Tajfel H, Turner JC (1979) An integrative theory of intergroup conflict. Soc Psychol Intergroup Relat 33(47):74 Todd AR, Bodenhausen GV, Richeson JA, Galinsky AD (2011) Perspective taking combats automatic expressions of racial bias. J Pers Soc Psychol 100(6):1027 Tsuchiya T, Hirai Y, Ono S (2007) A study of properties of the item count technique. Public Opin Q 71:253–272 Turner MA, Fix M, Struyk RJ (1991) Opportunities denied, opportunities diminished: racial discrimination in hiring. The Urban Institute Uhlmann EL, Cohen GL (2005) Constructed criteria redefining merit to justify discrimination. Psychol Sci 16(6):474–480 Van Knippenberg AD, Dijksterhuis AP, Vermeulen D (1999) Judgement and memory of a criminal act: the effects of stereotypes and cognitive load. Eur J Soc Psychol 29(2–3):191–201 Wittenbrink B, Judd CM, Park B (1997) Evidence for racial prejudice at the implicit level and its relationship with questionnaire measurements. J Pers Soc Psychol 72:262–274 Woolley AW, Chabris CF, Pentland A, Hashmi N, Malone TW (2010) Evidence for a collective intelligence factor in the performance of human groups. Science 330(6004):686–688 Wright BRE, Wallace M, Bailey J, Hyde A (2013) Religious affiliation and hiring discrimination in New England: a field experiment. Res Soc Stratif Mobil 34:111–126 Yeager DS, Paunesku D, Walton GM, Dweck CS (2013) How can we instill productive mindsets at scale? A review of the evidence and an initial R&D agenda. In: White paper for White House meeting on “Excellence in education: the importance of academic mindsets” Yeager DS, Purdie-Vaughns V, Garcia J, Apfel N, Brzustoski P, Master A, Hessert WT, Williams ME, Cohen GL (2014) Breaking the cycle of mistrust: wise interventions to provide critical feedback across the racial divide. J Exp Psychol Gen 143:804–824 Yinger J (1986) Measuring racial discrimination with fair housing audits: caught in the act. Am Econ Rev 76:881–893 Yinger J (1998) Evidence on discrimination in consumer markets. J Econ Perspect 12:23–40 Zinovyeva N, Bagues M (2011) Does gender matter for academic promotion? Evidence from a randomized natural experiment. IZA discussion paper 5537 Zussman A (2013) Ethnic discrimination: lessons from the Israeli online market for used cars. Econ J 123(572):F433–F468
Lab-in-the-Field Experiments Insights into Discrimination and Ways to Reduce It
10
Allison Demeritt and Karla Hoff
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Insights into Behavioral Causes of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Misperception by Observers: Shooter Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Emotional Bias Among Judges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implicit Bias of Teachers Against Females in Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bias in Work Attribution When Signals Are Ambiguous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Naturalized Relations of Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ways to Reduce Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dearth of Field Experiments on How to Reduce Prejudice and Discrimination . . . . . . . . . . . Overcoming Racial Divides with Alliances and Cooperation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcoming Ethnic Divides with Integrated Classrooms and Living Situations . . . . . . . . . . The Influence of Radio on Ethnic Divides in Rwanda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
236 239 239 241 243 244 246 248 249 249 252 254 255 257 257
Abstract
Lab-in-the-field experiments have played a major role in identifying behavioral causes of discrimination and ways to reduce discrimination. Such experiments have made it possible to isolate, among the many causes of discrimination, how cognitive mechanisms such as stereotyping, implicit bias, and in-group favoritism contribute to unequal experiences and opportunities. Evidence from natural experiments shows that mandates and incentives, established norms, and the likely desires of judges, teachers, and tenure committees to be fair do not “fix” the biases revealed in lab-in-the-field experiments. To perceive is to categorize, A. Demeritt University of Washington, Seattle, Washington, USA K. Hoff (*) Columbia University, New York, NY, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_18
235
236
A. Demeritt and K. Hoff
and the prototypes and cultural meanings associated with a category affect perception and judgments and can lead to discrimination. The phenomenon of cultural association that drives discrimination is the central focus of this chapter. Hoff, Demeritt, and Stiglitz (The other invisible hand: The power of culture to spur or stymie progress. Manuscript in preparation for Columbia University Press, 2022) call it schematic discrimination and define it as discrimination based on widely shared cultural schemas or “mental models” about specific types of people. It is distinct from taste-based and statistical discrimination because it may be neither consciously chosen nor efficient, and it is inclusive of, but broader than, implicit discrimination because it can occur either at or below the level of consciousness. Lab-in-the-field experiments have been used to evaluate ways to reduce schematic discrimination. Interventions that give people the experience of collaborative intergroup contact have substantially reduced schematic discrimination across a variety of social contexts. Keywords
Implicit bias · Cognitive resources · Cultural mental models · Schematic discrimination · Shooter bias · Stereotype
Introduction Imagine a country with a majority and a minority ethnic group. We’ll call the two groups “M” and “mn.” Members of the majority have limited interaction with members of the minority. Neither likes the other, but the majority is particularly likely to project anger toward the minority and to discriminate against them, since the majority sees the minority as undeserving of their country’s bounty. Now imagine that a single member of the minority hits it big: he lands on one of the country’s most beloved soccer teams and becomes its star player. Almost immediately, discrimination against the minority group declines in regions where people are strident fans of the soccer team. This isn’t rational, since the minority group contains significant diversity: one person – no matter how good his dribbling or finishing skills – does not change the probability of what other members of the hated mn group are bound to be “like.” Nevertheless, the attitudes and behaviors of many people in group M shift as a result of a single individual in group mn. Like other exemplars of diverse categories, the soccer player has an extraordinary power: he is but one of millions, but he changes the perception and likeability of an entire group of people. Sound unrealistic? It’s not; the story is actually true. It describes what happened in Liverpool when a Muslim footballer, Mohamed Salah, joined the UK’s Liverpool football club in 2017 and became its star member. (Soccer in the USA is called football in Europe.) The Muslim prayer that Salah performed after scoring made it clear to the public that he was Muslim, and his devotion to his faith both on and off the field brought Islamic practices into the spotlight. In Britain, Muslims are one of the most discriminated-against groups and the object of about half of all hate crimes.
10
Lab-in-the-Field Experiments
237
Analysis of police data found that hate crimes in the county containing Liverpool decreased by 16% after Salah joined the Liverpool team. Analysis of Twitter posts found that prejudiced speech against Muslims by Liverpool soccer fans fell to half the rate they were before Salah joined the Liverpool team (Alrababa’h et al. 2021). After one of Salah’s successful matches, his fans sang: If he scores another few Then I’ll be Muslim, too. If he’s good enough for you, He’s good enough for me. Sitting in a mosque, That’s where I wanna be.
Salah single-handedly changed the stereotype of Muslims and increased how much British society valued people belonging to the category of Muslim. How did this happen? This chapter examines lab-in-the-field experiments that try to get “inside the minds” of individuals to shed light on some of the mechanisms driving discrimination and the policies that can reduce it. Lab-in-the-field experiments are distinguished by three features: (1) a standardized paradigm for analyzing a problem, e.g., a behavioral game, a memory test, or an evaluation of an image; (2) participants who are representative of a relevant population, e.g., the experiments might draw on personnel officers to understand labor market discrimination or on teachers to study discrimination in education; and (3) a naturalistic setting to try to replicate a real decision-making context. Membership in a group is generally correlated with many factors – for example, race is often correlated with income and education, as well as dress and demeanor. This makes it difficult to isolate the factors driving discrimination where differences in treatment are observed. The advantage of lab-in-the-field experiments for assessing discrimination is that they are able to isolate the effect of membership in a group. These experiments permit analysts to compare, for example, how individuals treat White compared to Black men in situations where the distribution of all features except race is the same in the two groups. The study of discrimination in economics has evolved as the field has broadened its understanding of human decision-making. In standard economics, decisionmakers are rational actors with exogenous preferences. Rational actors, by definition, can commit only two kinds of discrimination – taste-based and statistical. This is because rational actors costlessly make all rational calculations and never process information incorrectly. In the late twentieth century, the psychologists Daniel Kahneman and Amos Tversky introduced into economics a new model of the decision-maker – an intuitive actor. This “behavioral” decision-maker thinks intuitively, automatically, emotionally, and in terms of associations (or “fast”), not rationally (or “slow”) (Kahneman 2003). This decision-maker may discriminate without deliberately choosing to and without awareness of his bias. Unconscious discrimination is called implicit discrimination (Bertrand et al. 2005).
238
A. Demeritt and K. Hoff
In the twenty-first century, researchers documented a kind of discrimination that cannot be reduced to an expression of preferences as economists understand them (as in taste-based discrimination), since it affects perception. Nor is it reducible to implicit bias, since it can influence conscious thinking and may not go away even when individuals are thinking deliberately (“slow”). This form of discrimination arises from individuals’ use of cultural mental models or “schemas” regarding what people belonging to certain groups are “like.” Cultural environments, including organizations, clubs, communities, and states, develop many social constructs that members absorb as they are socialized into the culture. Some of these constructs create associations between individuals belonging to particular groups and the traits that these individuals are thought to have. For example, people often think of librarians as quiet, rule-bound types and of computer programmers as nerdy young males. Of course, neither of these constructs (stereotypes) is universally true. Nevertheless, they creep into thinking – thus, one might feel surprised to meet an outgoing and carefree librarian or a socially skilled, older, and female tech entrepreneur. Social constructs bring cultural associations, meanings, and values into both conscious and unconscious thinking processes. Recognition of this kind of discrimination reflects cross-disciplinary work – not only with psychology but also with sociology, anthropology, cognitive science, and political science. Psychologists have long recognized that people see the world through categories (Bruner 1957). Humans are born with a cognitive system for categorizing the social world into groups (Kinzler and Spelke 2007). But which categories a society constructs – such as Muslim/Christian, “White”/“Black,” or straight/gay – and what meanings emerge for the categories depend on history, social structure, and institutions (e.g., Acharya et al. 2016; Alesina et al. 2013; Hoff and Stiglitz 2010). Humans without any exposure to the culture of a particular group who were dropped into their world would notice physical differences between people, but it is the sociocultural workings of a society that create most of the categories and other “tools of thinking” (such as stereotypes) that people in a culture use. No matter how “slow” and deliberately people think, the cultural categories and other conceptual tools that they use to organize their thoughts influence what they see and how they think. A wide literature on discrimination driven by cultural mental models exists (e.g., Beaman et al. 2009; Rao 2019; Lowe 2021; Sarsons et al. 2021), but there is no widely used term for it. Hoff, Demeritt, and Stiglitz (2022) call this form of discrimination schematic discrimination. Its defining feature is that it is unconscious or conscious discrimination based on cultural mental models (equivalently, cultural schemas). Individuals adopt and internalize many of the mental models of the communities that they are a part of (Berger and Luckmann 1966). The mental models affect how they process information and the meanings that they give to their encounters with the world (Bruner 1990; DiMaggio 1997). People come to believe that certain associations between categories and traits are relevant (even fundamental) to a particular class of decision – such as hiring, criminal sentencing, or evaluation on a test. Correll et al. (2015) invented the term “stereotypic vision” to
10
Lab-in-the-Field Experiments
239
capture the way stereotypes affect visual processing – an example of this appears in the next section. Recognizing the role of cultural mental models in how humans process information and construe situations makes clear why new exemplars will influence behavior. When an individual has new exemplars of a category – such as Salah as an exemplar of a Muslim – the meaning of the category changes. Schematic discrimination might be described as coding a person by a category that he fits into and behaving toward him as if the code summarizes his traits. The coding means that the decision-maker no longer thinks about the person as an individual but rather as a member of the given category. The decision-maker will not expend high effort to see or find counter-stereotypical information (von Hippel et al. 1993; Correll et al. 2015). Lab-in-the field experiments show how labeling a person by gender, race, or social class may change how the person’s performance is evaluated (e.g., in the Goldberg paradigm in Beaman et al. 2009) or, literally, how the person is seen. Jennifer Eberhardt and her colleagues (Eberhardt et al. 2003) asked White college students to draw, as carefully as they could, a picture of a face while its image was onscreen. The images had been created by using software that morphed a color photograph of a Black man’s face with a White man’s face. For half of the students, the man was described as White; for the other half, he was described as Black. The labels led individuals to draw different kinds of pictures. Figure 1 shows drawings by two of the students. These students are not representative of the group but instead of the subset of participants who were identified from their responses to a survey as holding the belief that individuals’ traits are fixed, not malleable. It is this group whose drawings reflected most strongly the racial labels given to the faces. The bottom left figure was drawn in a case where the face at the top was labeled Black; the bottom right figure was drawn in a case where the face at the top was labeled White. The figures illustrate that seeing is not a passive process. Instead, seeing is a constructive process that entails selection and use of social concepts and categories, such as “Black” and “White.” The remainder of this chapter has two parts. The first part examines the implications of social constructions for discrimination. The second part looks at ways to reduce discrimination.
Insights into Behavioral Causes of Discrimination Misperception by Observers: Shooter Bias A long series of highly publicized police killings of unarmed Black men has generated research on how individuals make decisions to shoot. The psychologist Joshua Correll and colleagues invented a videogame simulation of these high-stakes situations. They recruited college students, community members, and police officers to participate in lab experiments to assess whether systematic misperceptions lead to the unjustified shooting of young Black men (Correll et al. 2002, 2006, 2015). In the
240
A. Demeritt and K. Hoff
Fig. 1 Sample drawings by two university students of the face at the top of this figure when it was labeled as the face of a Black man or a White man. (Source: Eberhardt et al. (2003))
videogame, young men unexpectedly appear on the screen, one at a time, holding either a gun or an innocuous object, such as a wallet. The objects that the men in the videogame are holding are small but discernable if one focuses on them. The men are either African American or White, and half of each group are holding a gun. The goal of the player of the videogame is to shoot all the armed targets and not shoot any of the unarmed targets. The results of about 20 studies with community members and college students consistently show the same bias, called shooter bias: participants shot unarmed targets more frequently when they were Black and failed to shoot armed targets more frequently when they were White. White and African American participants
10
Lab-in-the-Field Experiments
241
displayed equivalent levels of shooter bias. As measured by brain activity, the more that participants differentiated the targets by race and processed Black targets as if they were threats, the greater their shooter bias in the videogame. Correll et al. (2015) tracked participants’ eye gaze during the shooter task. The center of the retina produces the highest visual resolution of any point on the retina. As focus moves away from the center of the retina, visual acuity decreases steeply. Stereotypes influenced visual processing. Participants responded to stereotypic targets before attaining the visual clarity they achieved for counter-stereotypic targets. In other words, visual clarity was a choice (but very likely one that was made unconsciously) that helps explain why participants were more likely to shoot an unarmed target if he was Black rather than White (the percentages were 16.4% for Black targets and 12.6% for White targets, p ¼ 0.03). For armed targets, the pattern was reversed: more participants chose not to shoot if an armed target was White (12.0%) rather than Black (9.2) (p ¼ 0.03). Stereotypes had a top-down influence on visual processing; they reduced the probability that individuals focused on the small object in a target’s hand before acting on a decision to shoot or not shoot the target. Police officers receive extensive training with firearms throughout their careers. The training promotes their ability to focus on the central features of a situation and ignore irrelevant features. Unlike the community members, the police who participated in the shooter task experiment did not show bias in their decisions to shoot or not shoot: their expertise eliminated the impact of race on the decision to shoot a man who appeared on a computer screen with a small, identifiable object in his hand. However, the police were affected in another way – speed: they fired at an armed target more quickly if he was Black, and they decided not to shoot an unarmed target more quickly if he was White. Do experts always exert the extra cognitive effort needed to marshal the attention and control needed for unbiased judgment? The next section describes a natural experiment that shows that they do not.
Emotional Bias Among Judges This section considers a natural experiment that reveals how emotions amplify the power of stereotypes to induce discrimination. It shows how upset losses in college football games influenced the sentences that judges imposed on Black juveniles convicted of crimes. If one thinks of the naturally occurring data as the outcome of an experiment, then the “treatment” is an unexpected loss by a prominent football team in the state. The natural experiment has key features of a lab-in-the-field experiment: participants face similar cases (a juvenile accused for the first time of a crime) and make a one-dimensional decision (the length of the sentence). The advantage of this experiment being natural rather than lab-in-the-field is that the incentives to make the “right” decision are enormous. The stakes bear on the deepest values of most of the participants: justice, the freedom of the defendant, and the legal order. The outcomes show whether the incentives that decision-makers face in real life are enough to “fix” the biases that occur in lab-in-the-field experiments.
242
A. Demeritt and K. Hoff
First, a bit of background: there is strong loyalty to sports teams in much of the world. In the USA, around 200 million viewers tune in to watch the regular college football season. Upset losses (i.e., losses by a team when it was expected to win) lead many fans to become angry and upset. Evidence of the strength of this reaction is that an upset loss in the USA leads to a 10% increase in the rate of at-home violence by men against their wives or girlfriends (Card and Dahl 2011). In contrast, losses that are expected, as well as wins (both expected and unexpected), have little-to-no impact on domestic violence. In Louisiana, devotion to the Louisiana State University football team is ingrained in the culture. Weddings and other big social events are often scheduled around the dates of LSU games. Ozkan Eren and Naci Mocan (2018) investigated the effect of upset losses of the LSU football team on the sentences that judges gave juveniles in Louisiana. Cases are randomly assigned to judges in each district court. The judge determines a sentence length for each convicted juvenile, regardless of whether he/she is placed in custody or ponrobation. The data used in this study is the length of the sentences meted out to first-time delinquents, age 10–17, who were convicted of a single offense and sentenced between 1996 and 2012. The judges had broad discretion in sentencing except in cases of murder and rape, and so those cases were excluded. This left a sample of about 10,000 cases. The authors define an upset loss as a loss of a game that the Las Vegas point spread favors the football team to win by four or more points. Controlling for this point spread, the outcome of a college football game is as good as random (Card and Dahl): losses are not correlated with bad weather, economic shocks, or anything else likely to put people in a bad mood. Thus, the researchers could use upset losses of the LSU team to tease out how being angry and upset amplifies the power of widely shared stereotypes to induce discrimination. Compared to periods during the regular season when the team had no games, in a week immediately following an upset loss, the judges discriminated against Black juveniles. The sentences imposed on Black juveniles increased, on average, by 43 days. This is an 8% increase in the average sentence of 514 days. The reaction of judges was stronger when they had stronger ties to LSU (it was their alma mater) and when the game was more consequential (it would have a big effect on the team’s chance of obtaining a national championship). The impact of an upset loss for White defendants was one-tenth as large (about 5 days) and statistically not different from zero. The effects on sentencing decisions persisted over the entire week following a Saturday game, but didn’t carry over to the following week. Non-LSU games had no impact on judges’ behavior. When LSU had not experienced an upset loss, Black juveniles did not receive longer sentences than White juveniles. The race of the judge also had no impact on the sentences he gave. What can explain the emergence of discrimination in the week immediately following an upset loss? A plausible explanation is that judges were upset, so they unconsciously looked for someone to punish. A cognitively easy target was a negatively stereotyped group – Black juveniles. The title of the paper sums up the result: “Emotional judges and unlucky juveniles.”
10
Lab-in-the-Field Experiments
243
Implicit Bias of Teachers Against Females in Science Stereotypes entail cognitive associations between categories – such as a race or gender – and attributes, such as violence or intelligence (Dovidio et al. 1986; Gaertner and McLaughlin 1983). This insight led to the development of the Implicit Association Test (IAT). It is an imperfect yet useful means of assessing implicit attitudes without relying on self-reports (Greenwald et al. 1998). In many cases, IAT scores predict prejudice and discrimination. Michela Carlana (2019) took the IAT to the field of junior high school teachers in Italy to assess whether their implicit bias about gender and science affected their students’ achievement. The IAT entails sorting words into one of the two buckets as quickly as possible. Each bucket is for two categories. Subjects complete two versions of the test – one with and one without the stereotypical alignment of the categories. In the GenderScience IAT, the screen with the stereotypical alignment is: Male
Female
Scientific fields
Humanities
The screen with the counter-stereotypical alignment is: Male
Female
Humanities
Scientific fields
Most people perform the sorting task faster and with fewer errors in the stereotypical alignment. People who hold the stereotype will tend naturally to collapse the two categories on each side of the upper screen shown above into one category: left becomes “male and academic fields associated with men”; right becomes “female and academic fields associated with women.” An implicit bias against females in science shows up as a difference between the time it takes an individual to complete the stereotypical and the counter-stereotypical versions of the IAT. The larger the gap, the greater the individual’s bias. For a person who does not have an implicit gender-science association, the difference in time and accuracy in performance between the two conditions is predicted to be zero. Two features of the Italian junior high school system made it possible to isolate the effect of long-run exposure to teachers biased against women in science. First, the assignment of a teacher to a class of students is as good as random and is not changed for the entire 3 years of junior high school. Second, classes are balanced: an effort is made to have a similar ability distribution of students within each class. With the strong support of the junior high school principals, 80% of all math and literature teachers working in 91 junior high schools in Italy completed the GenderScience IAT. These teachers taught a total of 47,000 students. Carlana found that
244
A. Demeritt and K. Hoff
female students’ scores on a standardized math test declined if their math teachers held the stereotype against women in science, as measured by the IAT. (There was no effect on the test scores of male students.) Having a biased teacher caused one-sixth of the gender gap on the standardized test. At the end of junior high school, students in Italy decide whether to enter an academic or a vocational high school. Math teacher bias against females in science increased by 11% the likelihood that a female student self-selected into a vocational high school. One mechanism through which gender-biased science teachers influenced their female students was to lower their self-confidence in math.
Bias in Work Attribution When Signals Are Ambiguous Joint projects are increasingly important in the workplace (Lazear and Shaw 2007). How to assign credit for joint work is, thus, a growing issue. Employers cannot perfectly observe the contribution of each member of a group to the group’s output. Instead, they have to infer each person’s contribution. Evidence of bias in the attribution of work accomplishments is suggested by two observational studies of the top economics departments in US universities and by a lab-in-the-field experiment in the USA and India (Sarsons 2017; Sarsons et al. 2021). The observational studies and lab-in-the-field experiment are described below.
Observational Study of Promotion to Tenure in Economics Departments In the economics departments of the top 30 PhD-granting universities in the USA, the gender gap in getting tenure is large: 75% of men and only 52% of women are granted tenure (Sarsons 2017, Table 1). Papers published in economics journals are the main basis for tenure decisions. Objective indicators of the quality of the papers are the journal rankings and the number of times a paper has been cited. Since there is no gender gap in either the number or the quality of candidates’ publications, the gender gap in promotion to tenure is puzzling. The obvious explanation might seem to be taste-based discrimination. But the evidence does not support this explanation. Men and women with the same number of single-authored papers were granted tenure with the same probability. If some employers have a distaste for tenuring women, one should see women who write solo-authored papers disproportionately denied tenure. The gender gap arises in the treatment of coauthored papers. Women who coauthor with men have lower tenure rates than men with a similar number and quality of coauthored and solo-authored papers. In the top 35 US PhD-granting universities, one additional coauthored paper is correlated with a 7.4% increase in men’s tenure probability, but only a 4.7% increase in women’s tenure probability (Sarsons et al. 2021, Table 2). The gender gap is especially pronounced for papers that a woman coauthors with one or more men. For men, it largely does not matter whether their papers are solo-authored or coauthored. The difference in the treatment of women’s and men’s coauthored papers accounts for 65% of the gender gap in
10
Lab-in-the-Field Experiments
245
promotion. In this analysis, average paper quality, citations, primary field, PhD institution, and seniority of coauthors are controlled for. The results of the observational study suggest discrimination. But they do not prove it, since there are possible confounds: it could be that tenure committees observe each coauthor’s contribution to a paper and know that, on average, the women up for tenure have contributed less than the male coauthors. To isolate the influence of gender on the assignment of credit for performance in group work, Sarsons and colleagues designed and implemented a lab-in-the-field experiment in which (in one of the treatments) players could perfectly observe each worker’s contribution to the team’s output.
Lab-in-the-Field Experiments on Assessing Performance The experiment has two steps. In step 1, university students (called “job candidates”) complete tasks individually and are individually rewarded. (The tasks are vocabulary search and numerical search tasks, for which prior experiments showed that there is little or no gender difference in performance.) Then the job candidates are randomly put into male-female pairs. In step 2, human resource workers from the USA and India (called “recruiters”) choose a job candidate from a pair under one of the two conditions: either they see the individual scores of the male member and female member of the pair or they see the sum of their scores. (If the woman has scored, say, 4 out of 5 and the man has scored 3 out of 5, the recruiters would see the score 7 out of 10 for the pair.) The recruiters know that there was no interaction between workers and that they were randomly paired. Recruiters also know that their payoffs in the game will depend on the productivity of the job candidates they choose. Thus, the recruiters have an incentive to recruit the more productive member of the pair. Under the first condition, where recruiters received clear signals of individual performance, they make unbiased use of the scores. A candidate’s gender does not affect the recruiters’ choice. This mirrors the finding that men and women receive equal credit for solo-authored papers in real tenure decisions in the top 35 US economics departments. But under the second condition, where recruiters see only the sum of a man’s and a woman’s score, male recruiters are more likely to pick the male candidate, and female recruiters are more likely to pick the female candidate. The odds ratio for an unbiased decision-maker would be 1. But the actual odds ratio that a male recruiter chooses a female candidate ranges from 0.72 to 0.92 (depending on the task and the controls); the odds ratio that a female recruiter chooses a female job candidate ranges from 1.2 to 1.4. The individual scores that make up the joint score are analogous to each person’s contribution to a coauthored research paper. Since most economics professors are men, the results of the lab-in-the-field experiment can explain economics department tenure outcomes as the result of in-group favoritism in the assignment of credit for group work. The finding implies that gender bias in male-dominated fields tends to perpetuate male domination of the fields. This lab-in-the-field experiment did not, however, test for schematic discrimination, since the tasks of the job candidates
246
A. Demeritt and K. Hoff
(vocabulary search and numerical search tasks) are not culturally categorized as men’s or women’s tasks. Ambiguous signals of performance are common in many settings and require decision-makers to make inferences, which can introduce bias. Gabriela Farfan, Alaka Holla, and Renos Vakis (2021) investigated whether teachers’ judgments of a student’s ability were biased by the student’s socioeconomic status (SES). Teachers at public schools in Lima, Peru, were randomly assigned to watch one of the two videos that portrayed a fourth-grader, called Diego, playing at home and in his neighborhood playground. In one of the videos, Diego was portrayed as from a low-SES household: he played in a low-income housing development, and his parents were described as high school-educated, blue-collar workers. In the other video, Diego was portrayed as from a high-SES household: he played in a middleclass neighborhood, and his parents were described as college-educated professionals. All the participants in the experiment then watched the same video of Diego taking an oral exam. His performance gave ambiguous signals: he answered some difficult questions correctly and some easy questions incorrectly. The teachers who had earlier seen the video of low-SES Diego judged his test performance to be below the fourth-grade level, while those who had seen the video of high-SES Diego judged his performance to be close to fifth-grade level. Associations with SES functioned like a lens through which participants saw Diego. It changed what they noticed and remembered about his performance on the oral exam.
Naturalized Relations of Power One of the ways that powerful and oppressive groups have historically legitimized their acts of oppression is by creating categories of people and representing them as ranked according to a natural hierarchy. Skin color was not initially an organizing principle in the colonies that became the USA (see Hoff and Stiglitz 2010, Online Appendix). The Africans who were sold into slavery and brought to the Americas were members of many different ethnic groups with distinct languages and no shared identity. In the colonies that became the USA, all individuals, including enslaved people, were initially accorded some rights and some obligations: they could sue or be sued in court, had to do penance in the parish church if they had illegitimate children, and could earn money of their own. Slave codes, based on the principle that enslaved persons were property without any rights of their own, were enacted by the US colonies only beginning in the late seventeenth century. The codes were a response to the large increase in the profitability of slave labor and the threat to the wealthy elite of alliances between poor, landless descendants of Euro-Americans and African Americans. To maintain slavery, the elite adopted a “divide-and-conquer strategy,” creating a sharply different legal regime for Euro-Americans and African Americans. The slave codes forbade schooling, church attendance, and land ownership to slaves. In some parts of the South, a person who taught a slave to read and
10
Lab-in-the-Field Experiments
247
write could be punished by death. The codes also forbade marriage across the “color line”: the two groups were to be kept in different worlds. Out of this creation of ranked categories of people, the fiction emerged of the natural discontinuity between a superior “White” race descended from Europeans and an inferior race of darkskinned people descended from Africans. The caste system in South Asia was also designed to support the exercise of power by elite groups. Although it had its origins in ancient India, it was transformed by ruling elites under the Mughal Empire and the British Empire. Under the British Empire, caste for the first time became a rigid system of division in South Asia (Dirks 2001), even though there are no discernible natural physical markers of caste (Deshpande 2011). Like African Americans under the slave codes, the lowest castes – historically called Untouchables and today called Dalits – were not allowed to own land or sit inside a schoolhouse. They are also denied the right to draw water from wells used by the upper castes (to emphasize the idea that Untouchables were polluted). The Constitution of India, which came into force in 1950, made Untouchability illegal. Evidence of a new social order is visible to every schoolchild in the stipends publicly distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in the political process. Yet children are also likely to encounter the traditional order of caste, segregation, and untouchability in their own experiences, through the fables they learn, and in the continued insults and atrocities levelled against upwardly mobile Dalits. The two social orders coexist in uneasy tension in India – the legal one in which “we are all equal now,” as some villagers remark, and another in which caste retains a firm grip on the mind. To assess the impact of caste identity on individuals’ performance, Hoff and Pandey (2006, 2014) undertook a lab-in-the-field experiment in a north Indian state (Uttar Pradesh) where the high castes are particularly dominant. Hoff and Pandey recruited 6th and 7th grade male students from the low castes (Dalits) and the high castes and taught them how to solve mazes. Performance in the experiment was measured by the number of mazes solved in two 15-minute sessions. The participants were paid one rupee for each maze they solved. The participants were randomly assigned to treatments that varied the salience of caste identity. In some treatments, participants sat in a classroom with a total of three high- and three low-caste boys. In this setting, low-caste boys solved on average as many mazes as high-caste boys as long as the participants’ castes were kept private. But in the treatments in which caste was publicly revealed, low-caste boys solved more than 20% fewer mazes than they did when caste was kept private. Publicly revealing caste increased the proportion of low-caste children who could not solve a single maze from 1% to 11% (Hoff and Pandey 2014, Fig. 6). This is the main channel through which making caste salient impaired the performance of the low caste. This is an example of stereotype threat, which has been widely studied for race and gender. There was one treatment in which the manipulation of caste cues reduced the performance of the high-caste boys. This was the treatment in which they sat in a high-
248
A. Demeritt and K. Hoff
caste segregated classroom and their caste identity was revealed. Segregating the high caste increased the proportion of high-caste boys who learned how to solve mazes, but significantly and substantially reduced the number of mazes that they solved (Hoff and Pandey 2014, Fig. 4). All available evidence – from the proportion of the high-caste boys who failed to learn how to solve mazes (which declined from 8% to 2%) to a direct test of self-confidence in another study (Hoff and Pandey 2005) – shows that the caste-segregated treatment did not impair the self-confidence of high-caste boys. The most plausible interpretation of the decline in high-caste performance in the caste-segregated treatment is an “entitlement effect”: segregation is a marker of high-caste dominance and evokes a sense of entitlement, reducing the need that the high-caste individuals feel to excel. As the Indian sociologist André Béteille notes, the caste system assigns social preeminence by birth, not performance (Béteille 2011, II [2003], p. 11). As of this writing, caste segregation remains in many villages a marker of high-caste privilege. It is also an issue over which high and low castes struggle. In a national survey in India in 2006, 45% of Dalits in the Hindi-speaking belt of India reported that some streets in their villages were off-limits to Dalits (Girard 2018). The lab-in-the-field experiment on the effect of priming a stigmatized identity on maze-solving was replicated in Beijing, China (Afridi et al. 2015). The subjects were elementary schoolchildren from two social categories: (a) households classified as urban Beijing households (a privileged category) and (b) households classified as rural non-Beijing (a disadvantaged category in Beijing). The household registration system in China, known as hukou, classifies citizens based on the birthplace of their parents or grandparents. It favors those who are categorized as local urban residents in housing, jobs, access to schools, and public benefits. Unlike categories of gender, class, and caste, hukou is a transparently social creation. The experimental findings show that priming hukou shifts performance in ways that mirror the way that social groups are ranked. In this sense, social identity “makes up people.” When the negative stereotype of a group lowers the performance of group members, there can be a self-reinforcing vicious circle over time. Hoff and Stiglitz (2010) present a model in which an effect of a negatively stereotyped identity is to bias individuals’ perceptions of their own performance: the members of the “inferior” group give themselves less credit than they otherwise would; this lowers their self-confidence relative to that of other groups, which lowers their performance relative to other groups and appears to validate the belief in the ranking of groups. Hoff and Stiglitz call this an “equilibrium fiction,” since the objective evidence sustains a false belief. Such fictions can make a society rigid.
Ways to Reduce Discrimination As the studies in the previous section demonstrate, prejudice and discrimination are prevalent in many societies and affect many different types of people. A pressing question is what to do about it. The remainder of this chapter addresses this question.
10
Lab-in-the-Field Experiments
249
Dearth of Field Experiments on How to Reduce Prejudice and Discrimination Are there means to influence individuals’ schema-based biases? Literature reviews in the area of prejudice and discrimination show that the limitations of existing studies preclude a conclusive answer to this question (Paluck and Green 2009b; Bertrand and Duflo 2017; Baldassarri and Abascal 2017; Paluck et al. 2019). In their 2009 review, which is notable for its breadth and methodological focus, Paluck and Green (2009b) found that only a small fraction of the hundreds of studies they reviewed were able to speak to “what works” to reduce prejudice in real-world settings. The vast majority used non-experimental or quasi-experimental methods unable to establish causality, if the methods were evaluated at all. Of even the best studies that Paluck and Green (2009b) include in their analysis, many evaluated interventions that lasted a day or less, more than 80% used non-representative samples as participants, the majority lacked behavioral outcome measures, and half had fewer than 100 participants and thus lacked statistical power. All in all, the authors conclude that there is a paucity of research to support scientific inference and generalization (p. 357). The entire genres of interventions developed in scientific laboratories, non-profits, government bureaus, and consulting firms – such as diversity and sensitivity training – have never been subjected to rigorous evaluation. Particularly given the dearth of field experimental work, lab-in-the-field experiments have an important role to play in shedding light on how to reduce discrimination in the real world.
Overcoming Racial Divides with Alliances and Cooperation Designing interventions to reduce discrimination requires an understanding of its roots. This chapter focuses on schematic discrimination. Recall that it is based not on hate or rational assessments, but instead on the cultural associations with categories into which people are put and the salience of the categories. Central questions are: Why do individuals categorize others by race and ethnicity? When are these categories salient? Until recently, race (like age and sex) was assumed to be a “primitive” dimension on which people individuate others – something humans use to encode or categorize people across all social contexts. But studies in the twenty-first century undermine this belief. The psychologists Kurzban, Tooby, and Cosmides (2001) hypothesized that the ease with which modern people encode race is a byproduct of cognitive machinery adapted to a different skill: detecting coalitions, alliances, and other patterns of cooperation.1 To test the idea that coalitional allegiance is what the
1
The hypothesis is that humans have a biological trait to encode people by their coalitions, alliances, and other patterns of cooperation, but not by their race. The ability to quickly compute and revise coalitional cues would have helped early humans negotiate the social world successfully. Since it gave individuals a survival advantage, the trait would have been selected for in human evolution.
250
A. Demeritt and K. Hoff
mind is actually encoding when it registers race, Kurzban et al. used a tool called a memory confusion protocol. It is a surprise memory test in which mistakes shed light on whether and how the test-taker is categorizing people. Here is how they used the tool. The participants in the experiment, who were college students (primarily Euro-American and Asian American), were told that two basketball teams had been in a fight. Subjects viewed photos of the eight members of the two teams along with statements that each man had made. The statements provided enough information for participants to infer which of the two teams an individual was on – they served as “cues” to coalition. As each team was comprised of two Black men and two White men, race was not associated with alliance. The heart of the protocol was a surprise memory test in which the participants viewed the photos again and were asked to recall who said what. Because people more easily confuse individuals whom they have coded as members of the same category than those coded as belonging to different categories (e.g., a citizen of Verona is more likely to confuse a Capulet with a Capulet than a Capulet with a Montague and vice versa), the pattern of misattributions in the memory test reveals the participants’ categorization schemes. When all of the eight men were dressed identically, subjects encoded the men by race twice as strongly as they encoded them by coalition. That is, they were more likely to misattribute a statement to someone of the same race as the actual speaker than to someone who was on the same team. But something striking happened when researchers manipulated the photos to give the team members colored uniforms signifying their team membership (a visual cue akin to skin color). Subjects then encoded the team coalition twice as strongly as they did race. This result suggests that racial perception and categorization are not fundamental (by contrast, sex encoding was shown to be robust in a nearly identical set of experiments). Instead, the extent to which humans perceive and categorize individuals by race depends on whether race predicts social relationships. The results of these experiments bear further investigation. First is the issue of replication: one recent study replicated the results (Pietraszewski 2021), but another obtained a substantially lower effect (Voorspoels et al. 2014). Second, the participants in these studies were college students. While this population represents an important part of the “field,” since racial discrimination is a problem on college campuses, racial encoding within this group is likely to be less ingrained and more malleable than in other populations. Third, the domain of basketball is strongly racially integrated and very rivalrous compared to other settings. Cueing this context likely facilitated subjects’ cognitive downshifting of race and upshifting of coalitions based on shirt color. Whether race could be “erased” (Kurzban et al. 2001) as easily in other arenas remains to be seen. Still, a key implication is that by changing the experiences people have and the alliances they observe, it is possible to start changing the significance of race. Recent lab-in-the-field experiments provide support for the “alliance” hypothesis, which holds that the root of ethnic and racial discrimination lies in a lack of exposure
10
Lab-in-the-Field Experiments
251
to interethnic cooperation. In Spain, university students in two port cities with ethnolinguistic diversity – Bilbao in the north and Valencia in the east – played a nested public goods game (Espinosa et al. 2019). The game is designed to discern the breadth of the social circle (local, global) within which people are willing to contribute to the collective good. In both cities, subjects readily contributed to a local public good that would benefit their co-ethnics. But whereas ethnic diversity increased contributions to a global public good by 38% in Valencia, it decreased contributions by 27% in Bilbao. A second experiment revealed that the mechanism was lower expectations of cooperation among the subjects in Bilbao. The authors point to the different histories of the north and east of Spain to explain this result. Although both cities are integrated across ethnic groups, Bilbao (in the north) has a history of political conflict, whereas Valencia (in the east) does not. In the north, residents’ day-to-day experiences led them to believe that individuals outside their ethnic group would not cooperate. In the east, historical experiences gave residents a more optimistic lens on interethnic cooperation that influenced how they played the game. Similar results regarding the effect of cooperative experiences were obtained in Mumbai, India, a city notorious for ethnic violence. Tusicisny (2017) had Hindu and Muslim slum-dwellers, many of whom had been implicated in ethnic riots, play a public goods game. He used the game not only to measure cooperation but also to manipulate it – he randomly assigned subjects to experience cooperation, or a lack of cooperation, with outgroup members before their own behavior was measured. Experiencing a short and superficial cooperative experience with an outgroup member caused ethnically mixed groups to produce as many public goods as homogeneous ones. Surprisingly, the effect was just as strong among voters of two extremist parties implicated in ethnic riots. Saumitra Jha’s (2013) analysis of Indian towns yielded the same conclusion about collaborative intergroup contact using historical data. Port cities that developed institutions to promote interethnic cooperation in the medieval period had less religious violence even centuries after the institutions were abolished. Taken together, these experiments lend credence to the view that race and ethnicity are not fundamental to human relations; rather, it may be a lifetime of experience in a society in which race is a predictor of coalition and cooperation (e.g., who interacts with whom, who likes whom, who shares with whom) that causes people to “see” race so strongly. In these experiments, the malleability of behavior and what might be called “preferences” for cooperating versus discriminating is striking and violates the assumptions of standard economics. Yet it is not surprising from the perspective of behavioral economics: humans’ bounded rationality and the influence of cultural categories on thinking lead people to revert to the use of stereotypes and other associations at the slightest need to determine who is “in” and who is “out.” The studies suggest that creating opportunities for people to experience, or even merely to witness, cooperation between historically divided groups can reduce
252
A. Demeritt and K. Hoff
discrimination by changing the associations that individuals have in mind. Such processes may help explain why, for example, even a brief exposure to cross-ethnic participatory politics in a war-torn country can change patterns of social cooperation and improve cohesion in ways that persist after a development program ends (e.g., Fearon et al. 2009). Indeed, the most methodologically robust studies on intergroup contact suggest that cooperation can decrease prejudice and discrimination (Paluck et al. 2019; Lowe 2021). The remainder of this chapter will discuss lab-in-the-field experiments that shed light on how policy can reduce discrimination. Can it change the lenses that people use to see the world?
Overcoming Ethnic Divides with Integrated Classrooms and Living Situations Several studies have used lab-in-the-field methodology to study the effect of intergroup contact in classroom and living settings. The studies arrive at conclusions that are mostly optimistic. Like the studies discussed above, they underline the importance of cooperative experiences and expectations for reducing discrimination. In Bosnia-Herzegovina, four high schools in the city of Mostar had a history of segregation by two main ethno-religious groups: Muslim Bosniaks and Catholic Croats. After the war, which killed or displaced millions, two of the four schools were partially integrated, and students were randomly assigned to a school. Researchers had students play a public goods game in ethnically homogeneous and heterogeneous groupings (Alexander and Christia 2011). Mixed-ethnicity groupings decreased students’ contributions only among students attending segregated schools. In other words, those attending desegregated schools exhibited a greater propensity for intergroup cooperation. Although the study did not look inside the schools to understand how the experience of integrated schooling might have affected students’ minds, a plausible explanation is that the integrated schools generated positive interactions that improved cooperative expectations among students. Another piece of evidence from the study supports this view: students attending integrated schools made higher contributions to the public good when they were able to sanction one another in the game. Institutions reduced the potentially adverse effects of diversity by ensuring that non-cooperators could be punished. These results suggest that ethnic discrimination does indeed have roots in expectations about cooperation. Gautam Rao (2019) found a similar pattern when he examined how wealthy students’ preferences and behaviors were influenced by interactions with poor students. A legal case in India led to a court order requiring 395 private schools to reserve 20% of their seats for students from poor households and to integrate these students into the same classrooms as the students from high-income families. Welloff students who were schooled with poor classmates were more charitable as measured by their volunteer activity. In a dictator game, they shared more of the money they were given than did students schooled only with their well-off peers, and this was true regardless of whether they played against a rich or poor partner. In other
10
Lab-in-the-Field Experiments
253
words, the experience of having poor classmates powerfully shaped fundamental notions of fairness. In a fascinating elaboration, students were asked to pick team members for an incentivized relay race. This allowed Rao to put a price on discrimination: when well-off students had to choose between more-athletic poor students and less-athletic rich students, past exposure to poor classmates reduced well-off students’ discriminatory behaviors by 12 percentage points. A key aspect of Rao’s study is that the significant behavior change of the well-off students came about not simply from being in classrooms with poor students, but through the mixing that occurred in small study groups – a setting that particularly emphasized interaction and cooperation. Implicit Association Tests (IATs), discussed above, are another powerful tool used in lab-in-the-field experiments to study the impact on implicit attitudes of living with, or near, members of racial and ethnic outgroups. Corno et al. (2022) found that random assignment to a Black roommate in a South African university reduced White students’ negative stereotypes of Black people and increased inter-racial friendships while also improving Black students’ academic outcomes. Barnhardt (2009) examined the consequences of a policy that caused Muslim and Hindu families to be housed together in a public housing complex in India. A range of levels of diversity within four-unit complexes allowed her to show that greater exposure to Muslim households improved Hindus’ explicit attitudes about Muslims and reduced the implicit bias of Hindu children against Muslims. Scacco and Warren (2018) reached a less optimistic conclusion about the impact of interactions in classroom settings, and yet the policy implications of their study are the same as those of the studies discussed above. The authors conducted a remarkable field experiment in the riot-prone city of Kaduna, Nigeria, that gave more than 800 young Christian and Muslim men 16 weeks of valuable computer training. They randomized recruitment into the program, assignment to a religiously homogeneous or heterogeneous classroom, and assignment to a co-religious or non-co-religious learning partner. Then they measured discrimination through two behavioral games: a dictator game and a destruction game. The dictator game measures altruism and norms of fairness, and the destruction game is its mirror image – it was designed to mimic aspects of riot behavior in which participants can choose to obtain a small personal benefit at a large cost to another person. At the end of the training, students assigned to mixed classrooms discriminated significantly less against outgroup members than subjects assigned to homogeneous classrooms. But there was a catch: students in mixed classrooms did not actually lessen their discrimination compared to a third group of control students who didn’t participate in the program at all. Rather, students in homogeneous classrooms actually began discriminating more after they finished the training. Unlike in the previous studies, where social contact increased cooperation, in this study, it seems that time spent with in-group members increased discrimination against outgroup members. The study thus yields the same policy implication as the prior studies but reaches it for a different reason: mixed groups may be helpful because they reduce opportunities for in-group bonding that strengthens discrimination.
254
A. Demeritt and K. Hoff
The Influence of Radio on Ethnic Divides in Rwanda A final example demonstrates that even fictional examples of intergroup cooperation can change the associations that inform behavior. In Rwanda, colonial practices divided a population along ethnic lines and fostered resentment between the two groups, known as Hutus and Tutsis. In 1994, the tensions erupted in a civil war. As many as one million people and a shocking 70% of the Tutsi population were slaughtered in just a few months. In terms of the percentage of the population killed, it was one of the worst genocides in recorded history. Analysis of a natural experiment demonstrated that a radio station that broadcast inflammatory messages calling for the extermination of the Tutsi minority incited violence (Yanagizawa-Drott 2014). A field experiment demonstrated that an edutainment radio soap opera could shift attitudes in the other direction by changing people’s perception of appropriate social norms regarding marriage and trust with outgroup members (Paluck and Green 2009a). In a series of lab-in-the-field experiments, Blouin and Mukand (2019) provided more precise evidence of the soap opera’s impact on interethnic relations. They used a number of cutting-edge methodological tools to investigate whether people living in areas that received the soap opera radio emissions exhibited systematically different behaviors than those living in areas that did not. The first test was an “ethnic salience” test (similar to the memory confusion protocol mentioned earlier) in which subjects viewed pictures of Hutu and Tutsi men along with statements about each (e.g., “This person owns a blue bicycle,” or “This person really likes bananas but dislikes guava”). Then subjects encounter a surprise recall test asking them to identify which statements were made by which individuals. The idea is that people living in communities where ethnicity is very relevant are more likely to notice (“encode”) ethnicity while observing the picture/statement pairs. Then, in the recall test, those individuals are expected to make more “within ethnicity” than “between ethnicity” mistakes – that is, they will be more likely to confuse a Hutu with another Hutu than with a Tutsi. The test’s critical measure is the ratio of “within category” to “total” misattributions; higher ratios are thought to reveal a higher level of ethnic identification in a community. The test showed that ethnic relevance was lower in regions that had received the emissions of the government’s soap opera; exposed individuals were 10–13 percentage points less likely to categorize others on the basis of their ethnicity. The soap opera appears to have altered private behavior, as well as public behaviors. One of the key criticisms levelled at Rwanda’s nation-building campaign is that it is doing little to change interethnic sentiments; detractors argue that people simply mask their true feelings and pretend to get along in order to avert government attention. The researchers showed this was unlikely to be the case. They asked villagers to play a trust game. Two people play the game together. One person is randomly selected to be the sender and the other to be the receiver, and players are told the rules of the game. Player 1 is given the money equivalent of about a day’s wages and told to transfer as much as she’d like to Player 2 and keep the rest. An assistant matches the amount transferred, and then Player 2 chooses how much of
10
Lab-in-the-Field Experiments
255
that sum to keep and transfers the remainder back to Player 1. Thus, in this game, if Player 2 is trustworthy, both players gain if Player 1 makes a large transfer to Player 2. In the public version of the game, Player 1 knows that the amount she transfers will be made public and thus reveal how trusting she is. In the private version of the game, Player 1 knows that her transfer will never be made public by the experimenters; thus, she need not fear community disapproval if she shows that she has low trust. Villagers living in regions not exposed to the soap opera offered on average 25% more in the public version of the game than they did in the private version of the game. But there was no such gap in offers made by villagers in regions that had received the soap opera emissions. Moreover, on average, interethnic trust offers in the private trust game were 47% higher in areas that had received the emissions from Radio Rwanda. The results of these experiments, like the others described in this section, suggest that our mental models of “other” racial and ethnic groups are malleable and depend on our experiences. This is a critical insight for policy. If ethnic preferences shift with experiences, changing the experiences that people have as a result of public institutions and policies has the potential to improve ethnic relations. The studies support Kurzban et al.’s (2001) view that what looks like ethnic discrimination may in fact be discrimination based on perceived cooperative potential. Direct experience with an outgroup, or a means of punishing non-contributors, reduced discrimination against outgroup members in all of the studies examined here. While the universe of high-quality research designs does not yet provide sufficient evidence to know whether, and to what depth, intergroup contact might reliably reduce prejudice (Paluck et al. 2019), the findings of existing lab-in-the-field studies support a cautiously optimistic outlook.
Conclusions This chapter shows how lab-in-the-field experiments have revealed the influence of categories, stereotypes, and prior experiences on perception and discrimination. Sociologists, anthropologists, and psychologists have long known that humans categorize everything and everyone they see. Until recently, behavioral scientists across disciplines lacked a scientific way to identify the conceptual structures in human minds, and, therefore, economics took limited account of the large role of categorization in discrimination. In the rational actor models of discrimination – i.e., taste-based and statistical discrimination – it is difficult to see how policies can reduce discrimination and improve relations between groups in society. Much has changed with the advent of a behavioral model of discrimination and with new methods that precisely measure attitudes and stereotypes. Lab-in-the-field experiments that use memory tests, IATs, and behavioral games can quantify bias and predict discrimination. Humans see the world through socially created lenses in which people are categorized and associated with traits that are emblematic of the category in the observers’ minds. These associations may persist indefinitely without an exogenous shock. Social scientists in the twenty-first century can study the biases
256
A. Demeritt and K. Hoff
Table 1 Three kinds of discrimination
Conscious only Both conscious and unconscious/ implicit Biased perception Updating of beliefs
Rational actors Taste-based discrimination ✓
Statistical discrimination
Intuitive actors with cultural mental models Schematic discrimination
✓ ✓
✓ ✓
Yes, but with Bayesian updating, the process may be slow
Sources
Exogenous, unexplained
Imperfect information about individuals in the presence of known and validated differences between demographic groups
Consequences
Discrimination based on preferences
Discrimination stemming from a desire to maximize efficiency. Increases in information about individuals and/or reductions in group-level differences reduce statistical discrimination
Updating may be slow or may not occur at all unless there is emotional engagement that evokes empathy Historical events; social patterns and social constructs; a desire to wield power over other groups; the desire to hold positive attitudes about one’s own group Discrimination stemming from cultural associations. It occurs at or below the level of consciousness and may be unintended. It is reduced by helping people experience collaborative intergroup contact, which tends to increase intergroup empathy and subjective expectations of intergroup cooperation
in human judgments and perceptions scientifically. Lab-in-the-field experiments help them get inside the heads of decision-makers to identify the causal mechanisms that drive discrimination. Lab-in-the-field experiments are a central tool to study how discrimination can be reduced by creating new exemplars and interactions that overturn existing associations and promote cooperation across group divides. Table 1 presents a typology of discrimination: taste-based, statistical, and schematic. This chapter has emphasized schematic discrimination. In most of the examples, categories, stereotypes, narratives, and identities function as lenses that mediate human experience. The lenses influence discrimination by shaping perception and construal.
10
Lab-in-the-Field Experiments
257
Cross-References ▶ Discrimination as Focal Point ▶ Experimental Evidence on Affirmative Action ▶ Taste-Based Discrimination
References Acharya A, Blackwell M, Sen M (2016) The political legacy of American slavery. J Polit 78(3): 621–641. https://doi.org/10.1086/686631 Afridi F, Li S, Ren Y (2015) Social identity and inequality: the impact of China’s hukou system. J Public Econ 123:17–29. https://doi.org/10.1016/j.jpubeco.2014.12.011 Alesina A, Giuliano P, Nunn N (2013) On the origins of gender roles: women and the plough. Q J Econ 128(2):469–530 Alexander M, Christia F (2011) Context modularity of human altruism. Science 334(6061): 1392–1394 Alrababa’h A, Marble W, Mousa S, Siegel A (2021) Can celebrities reduce prejudice? Estimating the effect of mohamed salah on islamophobic attitudes and behaviors. Am Polit Sci Rev 115(4):1111–1128 Baldassarri D, Abascal M (2017) Field experiments across the social sciences. Annu Rev Sociol 43(1):41–73. https://doi.org/10.1146/annurev-soc-073014-112445 Barnhardt S (2009) Near and dear? Evaluating the impact of neighbor diversity on inter-religious attitudes. Working paper: 56 Beaman L, Chattopadhyay R, Duflo E, Pande R, Topalova P (2009) Powerful women: does exposure reduce bias? Quarterly Journal of Economics 124(4):1497–1540. https://doi.org/10. 1162/qjec.2009.124.4.1497 Berger P, Luckmann T (1966) The social construction of reality. Anchor Books Bertrand M, Duflo E (2017) Field experiments on discrimination. In: Duflo E, Banerjee A (eds) Handbook of economic field experiments, pp 309–393. North-Holland, Amsterdam Bertrand M, Chugh D, Mullainathan S (2005) Implicit discrimination. Am Econ Rev 95(2):94–98. https://doi.org/10.1257/000282805774670365 Béteille A (2011) The André Béteille Omnibus. Oxford University Press, Delhi Blouin A, Mukand SW (2019) Erasing ethnicity? Propaganda, nation building, and identity in Rwanda. J Polit Econ 127(3):1008–1062. https://doi.org/10.1086/701441 Bruner J (1957) On perceptual readiness. Psychol Rev 64(2):123–152. https://doi.org/10.1037/ h0043805 Bruner J (1990) Acts of meaning. Harvard University Press, Cambridge MA Card D, Dahl GB (2011) Family violence and football: the effect of unexpected emotional cues on violent behavior. Q J Econ 126(1):103–143. https://doi.org/10.1093/qje/qjr001 Carlana M (2019) Implicit stereotypes: evidence from teachers’ gender bias. Q J Econ 134(3): 1163–1224. https://doi.org/10.1093/qje/qjz008 Corno L, La Ferrara E, Burns J (2022) Interaction, stereotypes, and performance: evidence from South Africa. Am Econ Rev 112(12):3848–75. https://doi.org/10.1257/aer.20181805 Correll J, Park B, Judd C M, Wittenbrink B (2002) The police officer’s dilemma: using ethnicity to disambiguate potentially threatening individuals. J Pers Soc Psychol 83(6):1314–1329. https:// doi.org/10.1037/0022-3514.83.6.1314 Correll J, Urland GR, Ito TA (2006) Event-related potentials and the decision to shoot: the role of threat perception and cognitive control. J Exp Soc Psychol 42(1):120–128. https://doi.org/10. 1016/j.jesp.2005.02.006
258
A. Demeritt and K. Hoff
Correll J, Wittenbrink B, Crawford MT, Sadler MS (2015) Stereotypic vision: how stereotypes disambiguate visual stimuli. J Pers Soc Psychol 108(2):219–233. https://doi.org/10.1037/ pspa0000015 Deshpande A (2011) The Grammar of Caste: Economic Discrimination in Contemporary India. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780198072034.001.0001 DiMaggio P (1997) Culture and cognition. Annu Rev Sociol 23:263–287 Dirks ND (2001) Castes of mind. Princeton University Press, Princeton Dovidio JF, Evans N, Tyler RB (1986) Racial stereotypes: the contents of their cognitive representations. J Exp Soc Psychol 22(1):22–37. https://doi.org/10.1016/0022-1031(86)90039-9 Eberhardt JL, Dasgupta N, Banaszynski TL (2003) Believing is seeing: the effects of racial labels and implicit beliefs on face perception. Personal Soc Psychol Bull 29(3):360–370. https://doi. org/10.1177/0146167202250215 Eren O, Mocan N (2018) Emotional judges and unlucky juveniles. Am Econ J Appl Econ 10(3): 171–205. https://doi.org/10.1257/app.20160390 Espinosa MP, Fatás E, Ubeda P (2019) Linguistic diversity and out-group discrimination in bilingual societies. J Behav Exp Econ 81:102–127. https://doi.org/10.1016/j.socec.2019.06.002 Farfan G, Holla A, Vakis R (2021) Poor expectations: experimental evidence on teachers’ stereotypes and student assessment. The World Bank https://doi.org/10.1596/1813-9450-9593 Fearon JD, Humphreys M, Weinstein JM (2009) Can development aid contribute to social cohesion after civil war? Evidence from a field experiment in post-conflict Liberia. Am Econ Rev 99(2): 287–291. https://doi.org/10.1257/aer.99.2.287 Gaertner SL, McLaughlin JP (1983) Racial stereotypes: associations and ascriptions of positive and negative characteristics. Soc Psychol Q 46(1):23. https://doi.org/10.2307/3033657 Girard V (2018) Don’t touch my road. Evidence from India on affirmative action and everyday discrimination. World Dev 103(C):1–13. https://doi.org/10.1016/j.worlddev.2017.10.008 Greenwald AG, McGhee DE, Schwartz JLK (1998) Measuring individual differences in implicit cognition: the implicit association test. Journal of Personality and Social Psychology 74(6): 1464–1480. https://doi.org/10.1037/0022-3514.74.6.1464 Hoff K, Pandey P (2005) Opportunity is not everything: how belief systems and mistrust shape responses to economic incentives. Econ Transit 13(3):445–472. https://doi.org/10.1111/j.14680351.2005.00230.x Hoff K, Pandey P (2006) Discrimination, social identity, and durable inequalities. Am Econ Rev 96(2):206–211. https://doi.org/10.1257/000282806777212611 Hoff K, Pandey P (2014) Making up people: the effect of identity on performance in a modernizing society. J Dev Econ 106:118–131 Hoff K, Stiglitz JE (2010) Equilibrium fictions: a cognitive approach to societal rigidity. Am Econ Rev 100(2):141–146. https://doi.org/10.1257/aer.100.2.141 Hoff K, Demeritt A, Stiglitz JE (2022) The other invisible hand: The power of culture to spur or stymie progress. Manuscript in preparation for Columbia University Press Jha S (2013) Trade, institutions, and ethnic tolerance: evidence from South Asia. Am Polit Sci Rev 107(4):806–832. https://doi.org/10.1017/S0003055413000464 Kahneman D (2003) Maps of bounded rationality: psychology for behavioral economics. Am Econ Rev 93(5):1449–1475. https://doi.org/10.1257/000282803322655392 Kinzler KD, Spelke ES (2007) Core knowledge. Dev Sci 10(1):89–96. https://doi.org/10.1111/j. 1467-7687.2007.00569.x Kurzban R, Tooby J, Cosmides L (2001) Can race be erased? Coalitional computation and social categorization. Proc Natl Acad Sci 98(26):15387–15392. https://doi.org/10.1073/pnas. 251541498 Lazear EP, Shaw KL (2007) Personnel economics: the economist’s view of human resources. J Econ Perspect 21(4):91–114. https://doi.org/10.1257/jep.21.4.91 Lowe M (2021) Types of contact: a field experiment on collaborative and adversarial caste integration. American Economic Review 111(6):1807–1844. https://doi.org/10.1257/aer. 20191780
10
Lab-in-the-Field Experiments
259
Paluck EL, Green DP (2009a) Deference, dissent, and dispute resolution: an experimental intervention using mass media to change norms and behavior in Rwanda. Am Polit Sci Rev 103(4): 622–644. https://doi.org/10.1017/S0003055409990128 Paluck EL, Green DP (2009b) Prejudice reduction: what works? A review and assessment of research and practice. Annu Rev Psychol 60(1):339–367. https://doi.org/10.1146/annurev. psych.60.110707.163607 Paluck EL, Green SA, Green DP (2019) The contact hypothesis re-evaluated. Behavioural Public Policy 3(02):129–158. https://doi.org/10.1017/bpp.2018.25 Pietraszewski D (2021) The correct way to test the hypothesis that racial categorization is a byproduct of an evolved alliance-tracking capacity. Sci Rep 11:3404. https://doi.org/10.1038/ s41598-021-82975-x Rao G (2019) Familiarity does not breed contempt: generosity, discrimination, and diversity in Delhi schools. Am Econ Rev 109(3):774–809. https://doi.org/10.1257/aer.20180044 Sarsons H (2017) Recognition for group work: gender differences in academia. Am Econ Rev 107(5):141–145. https://doi.org/10.1257/aer.p20171126 Sarsons H, Gërxhani K, Reuben E, Schram A (2021) Gender differences in recognition for group work. Journal of Political Economy 129(1):101–147. https://doi.org/10.1086/711401 Scacco A, Warren S (2018) Can social contact reduce prejudice and discrimination? Evidence from a field experiment in Nigeria. Am Polit Sci Rev 112(3):654–677. https://doi.org/10.1017/ S0003055418000151 Tusicisny A (2017) Reciprocity and discrimination: an experiment of Hindu-Muslim cooperation in Indian slums. Polit Psychol 38(3):409–426. https://doi.org/10.1111/pops.12340 Von Hippel W, Jonides J, Hilton JL, Narayan S (1993) Inhibitory effect of schematic processing on perceptual encoding. J Pers Soc Psychol 64(6):921–935. https://doi.org/10.1037/0022-3514.64. 6.921 Voorspoels W, Bartlema A, Vanpaemel W (2014) Can race really be erased? A pre-registered replication study. Frontiers in Psychology 5. https://doi.org/10.3389/fpsyg.2014.01035 Yanagizawa-Drott D (2014) Propaganda and conflict: evidence from the rwandan genocide. Q J Econ 129(4):1947–1994
Methodological Approaches to Understanding Discrimination: Experimental Methods – Trust, Dictator, and Ultimatum Games
11
Abhinash Borah
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Minimal Group Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dictator Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trust Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ultimatum Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
261 264 270 274 278 281 282 282
Abstract
The chapter provides an overview of the literature that uses experimental implementations of the dictator, ultimatum, and trust games to study patterns of discrimination. The focus is on work using lab-induced and natural identities to study group-sensitive behavior. Methodological issues along with results and insights are emphasized. Keywords
Dictator game · Ultimatum game · Trust game · Discrimination · Identities · Experiments
A. Borah (*) Department of Economics, Ashoka University, Sonipat, India e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_17
261
262
A. Borah
Introduction Although nothing in the way of the foundations of rational choice theory demands that individual behavior be exclusively selfish, research and teaching in the neoclassical tradition has proceeded mostly with the assumption of selfish behavior as the benchmark. That this is a descriptively inaccurate assumption has never been in doubt. However, early criticisms of this approach were often met with the response that this is a useful approximation of human behavior and, as such, a reasonable working assumption. Whether the last claim is valid or not, of course, is an empirical question. The dictator, ultimatum, and trust games, and the elicitation of behavior in them through lab and field experiments have formed the discipline’s most tried and tested methodology in answering this empirical question – as it turns out markedly in the negative. Together, they have provided strong evidence suggesting that a considerable fraction of decision makers have socially minded preferences, or social preferences, for short. One reason behind the appeal of these games and their popularity in the literature is their simplicity. All three are two-player games. In the dictator game, one of the players, appropriately termed the dictator, is endowed with a fixed amount of a desirable divisible good (say money). It is up to the dictator to decide a split of the endowment between herself and the other player, i.e., decide what amount of the endowment, if any at all, she wants to give to the other player. In essence, therefore, the dictator game is simply a decision problem. The earliest experimental evidence of this game in this form comes from Forsythe et al. (1994) and, subsequently, numerous other experiments have been run based on it. These experiments show that, on average, dictators tend to give about 20% of the endowment to the other player (Camerer 2011). Given that players are matched anonymously in these experiments, the evidence suggests that individuals do not care about just their own selfish outcomes, but their preferences extend to concerns like fairness and altruism. The story gets a little more involved in the ultimatum game. In it, one of the players (the proposer) makes an offer to split the endowment with the other player (the responder). The responder then has to decide if she wants to accept or reject the proposal. In the former case, both players get the specified amounts under the offer, but in the latter, both walk away with nothing. The first set of experimental results for this game came from Güth, Schmittberger, and Schwarze (1982). In general, one finds that the median offers of proposers are around 40–50% and mean offers are 30–40%. Offers of 40–50% are rarely rejected. Offers below 20% or so are rejected about half the time (Camerer 2011). The interpretation of outcomes in the ultimatum game is a little more subtle than the dictator game. Here, positive offers need not necessarily signify concern for the other person’s outcomes in a deep sense. Instead, it may simply capture a rational response to the fact that she may reject small offers. On the other hand, rejection of positive offers by the responder signifies social attitudes. Presumably, such offers are rejected because the responder has distributional concerns and cares about the fairness of the allocation.
11
Methodological Approaches to Understanding Discrimination:. . .
263
As far as the trust game goes, in its standard form, the first player (the investor) once again is given an endowment. She can send the second player (the trustee) any fraction of that endowment. The amount sent is multiplied by the experimenter (typically by a factor of three). The second player then has to decide how much of the multiplied amount to return to the first player. This simple game, first introduced by Berg, Dickhaut, and McCabe (1995), provides an elegant way to measure trust and trustworthiness – specifically, the amount sent by the first player measures trust, while the amount returned by the second player measures trustworthiness or reciprocity. Berg, Dickhaut, and McCabe (1995) found that investors invested about 50% of their endowment, whereas the average amount returned by the trustee was about 95% of the invested amount. In the experimental implementation of these games, in their standard form (as mentioned above), details about the identities of the two players – for instance, their gender, religion, race, etc. – remain unspecified and anonymous. The only identity attributed to subjects in them is the superordinate one of simply being human. Therefore, the results that we get in them provide us an insight into the social attitudes of individuals when they view these interactions at the level of dealing with abstract individuals. However, real-world interactions involving the expression of attitudes like altruism, fairness, trust, and reciprocity are often not viewed through the prism of abstract individuals. Individuals in such interactions have identities associated with them. Very often such identities are perceived through the prism of an “us” versus “them” or an ingroup–outgroup divide, because of which a decision maker’s social attitudes may vary depending on the identity of the individual/s she is interacting with. If it is someone from her perceived ingroup, positive sentiments may be particularly heightened. On the other hand, if it is someone from her outgroup, these sentiments may be much more subdued or nonexistent. In other words, such attitudes and sentiments may have a discriminatory nature to them. Luckily, the structure of all three games is flexible enough that experimenters can incorporate such aspects of identity into them to study groupsensitive and discriminatory attitudes. When incorporating identities in such experiments, there are two main design choices that experimenters have drawn on. The first is to endogenously create artificial identity categories as part of the design of the experiment. The most prominent approach in this regard is the minimal group paradigm. The other is to use naturally existing real-world identities like gender, caste, religion, ethnicity, nationality, and the like. When this second approach is used, experimenters often rely on a technique called priming to make such identities salient in the experiment. This technique involves exposing subjects to some feature of the identity sought to be invoked through text or other audiovisual means or by making them perform a task to make this aspect of their identity salient. The chapter starts in section “Minimal Group Paradigm” with a discussion of the minimal group paradigm and how it provides insights on intergroup discrimination. To illustrate how this approach has been used in experiments that use simple games of the type discussed here, we will focus in detail on a leading paper by Chen and Li (2009) that uses this approach. Section “Natural Identities” highlights how natural
264
A. Borah
identities have been used and made salient in experiments that have employed these games. After that, Sections “Dictator Games,” “Trust Games,” and “Ultimatum Games” will discuss more learnings and insights about discrimination developed in experiments using, respectively, the dictator, trust, and ultimatum games.
Minimal Group Paradigm The minimal group paradigm (MGP) originated in social psychology. As is well known, social psychologists have for long emphasized that many aspects of individual behavior and personality may be socially determined. Specifically, to them, the notion of a group or group identity is not a superfluous concept in the sense of simply being a by-product of individual interaction, interests, and interdependence. Rather, groups exert a constitutive psychological influence on individual attitudes and behavior. In making this case, social psychologists have assembled an impressive body of evidence. The minimal group paradigm studies by Tajfel et al. (1971) are a leading example of this. They were designed with the purpose of identifying the minimal conditions for intergroup discrimination. In their original studies, they wanted to add variables incrementally to see at what point group favoritism would occur. To that end, the groups in their control setting were so stripped down and (apparently) empty of psychological significance that no discrimination was expected. In particular, subjects were divided into groups on the basis of trivial, ad-hoc, and random criteria such as, for instance, the result of a coin-flip. The group classification in this control setting became popularly known as the minimal group. The conditions under which a group is minimal include: (1) random nonoverlapping group assignments based on trivial tasks; (2) no interactions between subjects; (3) anonymous group membership; and (4) no link between chooser’s self-interest and her choices (Tajfel and Turner 1986; Chen and Li 2009). Despite this meaningless categorization, the authors found surprising evidence for ingroup favoritism and discriminatory behavior towards the outgroup. In many of the experiments, subjects were asked to anonymously divide a fixed sum of money between a member from their ingroup and one from their outgroup, who remained anonymous except for their group membership. In such settings, subjects chose to allocate as much as 70% to the ingroup member. As Turner writes, “it appears that imposing social categorizations on people (even on an explicitly random basis) produces discriminatory intergroup behavior, intragroup cohesion in the form of more positive attitudes toward and more reported liking of ingroup than outgroup members, ethnocentric biases in the perception, evaluation, and memory of ingroup and outgroup, and an altruistic orientation toward ingroup members” (Turner 1985, pp. 83–84). In these minimal group experiments, the choice problem facing the decision maker is a pure allocation task of dividing a sum of money between two other individuals. Her choice has no bearing on her own outcomes. Chen and Li (2009) extend this class of experiments to relax this assumption. They elicit group-sensitive social attitudes based on choices in simple games of the type we have been discussing here which feature nontrivial trade-offs between own outcomes and
11
Methodological Approaches to Understanding Discrimination:. . .
265
those of others. In so doing, they synthesize the social psychology approach with that of the social preferences literature in economics. Since this work has important conceptual lessons, we will now delve into it at some length and focus on the design of the experiment before going on to their findings. The experiment had five treatments and one control. In the treatment sessions, there were four stages. The first was a group assignment stage. The second was a collective problem solving stage among members within a group to enhance the salience of group identities. The third was an other-other allocation stage in the spirit of the MGP studies. The fourth was a set of two-person games that included the dictator game and other games that measured reciprocity. While subjects in different treatments participated in different stages, subjects in the control group participated only in the fourth stage. All five experimental treatments contained the group assignment stage. Two different group assignment methods were tried. In the Original Treatment, subjects were shown five pairs of paintings by two artists, Paul Klee and Wassily Kandinsky, without being told the artist of each painting. They were then divided into two distinct groups depending on their preferences over these paintings. To measure the difference between such a group assignment based on painting preference and random assignment, two treatments with random group assignments were also implemented – a Random Within treatment and a Random Between treatment. In this assignment, each participant randomly drew one envelope from a stack, where each envelope contained a Maize or Blue slip. The color of the slip in their draw determined their group assignment. The second stage designed to enhance group identity involved communication within the members of the group via an online chat program. This was implemented in all three of the Original, Random Within, and Random Between treatments. Specifically, subjects were provided with two additional paintings and had to answer two questions about the artists who painted them. They were given 10 min to communicate with other members from their group via the online chat program to help them answer the question correctly. These answers were incentivized with correct answers being rewarded. To assess the impact of this task, a No Chat treatment was added that did not include this stage. In the third stage of the Original, Random Within, Random Between, and No Chat treatments, each subject was asked to allocate a given amount of experimental currency between two anonymous participants under three situations: (i) both of them were from her group, (ii) both were from the other group, and (iii) one was from her group and the other from the other group. Together, the second and the third stage were employed to enhance the group identity induced in the first stage. To see their impact, a No Help treatment was added, which did not include these two stages. Further, a comparison of the No Chat and No Help treatments allowed the authors to identify the effects of the other-other allocation task in the third stage. A key question that interests researchers in such studies where lab-induced identities are assigned is about what aspects of the design make these identities salient. In this context, it is worth paying particular attention to the different treatments that Chen and Li (2009) introduce to get a handle on this question. It is something we will come back to below.
266
A. Borah A
Fig. 1 Positive Reciprocity
E
N
B 725, 0 R
L
400, 400
750, 375
A
Fig. 2 Negative Reciprocity
E
N
B 750, 750 L
800, 200
R
0, 150
Finally, in the fourth stage, subjects made decisions in 24 two person sequential games drawn from Charness and Rabin (2002) to analyze the impact of group identity on social preferences and economic outcomes. Of these 24, there were 5 versions of discrete dictator games that required the dictator to choose between: (400, 400) vs. (750, 400); (400, 400) vs. (750, 375); (300, 600) vs. (700, 500); (200, 700) vs. (600, 600); and (0, 800) vs. (400, 400). In these pairs, the first payoff is that of the recipient (Player A) and the second of the dictator (Player B). In addition, there were sequential games of the form shown in Figs. 1 and 2. In these games, player A has to first choose between enter (E) and not enter (N ). If she chooses N, the game ends. On the other hand, if she chooses E, the game comes to player B, who has to then choose between L and R. Observe the difference between the game in Figs. 1 and 2. In Fig. 1, A’s choice to enter is associated with
11
Methodological Approaches to Understanding Discrimination:. . .
267
good intentions, for if she were to choose N, then B’s payoff would have been 0, whereas following E, it is at least 375. On the other hand, in Fig. 2, A’s choice to enter reflects bad intentions since by entering she reduces B’s payoff from 750 to no more than 200. Hence, the first game may induce positive reciprocity on the part of B who may be willing to choose R and a lower payoff for herself in order to reward A for her good intentions. On the other hand, the second game may induce negative reciprocity on the part of B as she may be willing to choose R and a lower payoff for herself in order to punish A for her bad intentions. In each game, subjects were randomly matched with another participant and were randomly assigned the role of player A or B. In all treatments except Random Between, each subject made decisions under two situations: (i) the match is from the ingroup and (ii) the match is from the outgroup. In order to interpret the results, Chen and Li (2009) draw on the social preferences literature and consider a preference representation for player B in the spirit of the representations in Charness and Rabin (2002) and Fehr and Schmidt (1999). In the context of social preferences, the following type of preference representation has been extensively considered: uB ðxA , xB Þ ¼ xB ðρr þ σsÞðxB xA Þ where r ¼ 1, s ¼ 0 if xB xA, and r ¼ 0, s ¼ 1, if xB < xA. The interpretation of xB xA is that it represents a situation of advantageous inequality for B. In this situation, a socially minded B may feel a sense of guilt and she may be inclined to be charitable towards A. The parameter ρ measures B’s sensitivity towards advantageous inequality, i.e., her sense of charity. On the other hand, xB < xA captures a situation of disadvantageous inequality for B and, in this case, she may feel a sense of envy. The parameter σ measures B’s sensitivity towards disadvantageous inequality, i.e., her sense of envy. A priori, we may expect that ρ > 0 and σ < 0. Chen and Li (2009) adapt these class of preferences to accommodate concerns about group identity. Specifically, they model B’s preferences by: uB ðxA , xB Þ ¼ xB ðρð1 þ IaÞr þ σ ð1 þ IbÞsÞðxB xA Þ where I ¼ 1 if A is from B’s ingroup and I ¼ 0 if A is from B’s outgroup. In other words, the parameters, a and b, capture the additional ingroup effect for charity and envy, respectively. For example, when B receives a higher payoff than A, the parameter ρ measures the charity effect for an outgroup match, while ρ(1 þ a) measures the charity effect for an ingroup match. With that background, we can use the choices of subjects in player B’s role in the experimental games to estimate the preference parameters. The table below from Chen and Li (2009) provides the maximum likelihood estimates of these parameters (Table 1). The top Panel A provides the estimates of the charity and envy parameters for the control group. The charity parameter, ρ, is 0.427 for the control group, while in the Original treatment, it is 0.323 for outgroup matches and 0.474 for ingroup matches.
268
A. Borah
Table 1 Estimates of charity and envy parameters Panel A Control (N ¼ 536) Panel B
Treatment (N ¼ 1896)
Charity ρ 0.427 (0.022)a Outgroup charity ρ0 0.323 (0.021)a
Envy σ 0.049 (0.025)b Outgroup envy σ0 0.112 (0.019)a
Ingroup charity ρ0 (1 þ a) 0.474 (0.018)a
Ingroup envy σ 0 (1 þ b) 0.008 (0.021)
Identity parameters a b 0.467 0.931 (0.112)a (0.192)a
Notes: Panel A reports estimates for the control sessions without identity, while panel B reports estimates for treatment sessions with identity a Significant at the 1% level b Significant at the 5% level c Significant at the 10% level
Similarly, the envy parameter, σ, is 0.049 for the control group, while in the Original treatment, it is 0.112 for outgroup matches and 0.008 for ingroup matches. The effect of group identity on charity is measured by the parameter a ¼ 0.467 and that on envy is given by parameter b ¼ 0.931. This indicates that when participants have a higher payoff, they show a 47% increase in charity concerns towards an ingroup match, compared with an outgroup match. Similarly, when participants have a lower payoff, they show a 93% decrease in envy for an ingroup match. Hence, these results allow us to reject the null hypothesis that group identity has no influence on distribution preferences. Specifically, participants’ charity (envy) towards an ingroup match is significantly greater (less) than that towards an outgroup match. To investigate group sensitive attitudes towards reciprocity, the games of positive and negative reciprocity outlined above were separately considered. In positive reciprocity games, it was found that participants were significantly more likely on average to reward ingroup than outgroup members for good behavior. In the logit estimates, player B’s likelihood to reward A in such games was found to increase by 18.6% if A happened to be from B’s ingroup. At the same time, in some of these games rewarding A necessitated that B’s payoff falls below A, i.e., B faces disadvantageous inequality. In those cases, it was found that the negative effect of envy on positive reciprocity is, in fact, stronger towards an ingroup match. In negative reciprocity games, subjects were also found to be significantly more forgiving towards misbehavior from an ingroup than an outgroup member. In such games, player B is 12.8% less likely to punish A for misbehavior if A happened to be from her ingroup. Further, B’s likelihood of punishing decreases with advantageous inequality, and this effect is stronger for ingroup members. Finally, we consider the important question of what generates group effects. The authors exploit the different treatments and games in their design as well as a postexperiment survey, where subjects were asked to report the strength of their ingroup
11
Methodological Approaches to Understanding Discrimination:. . .
269
attachment on a scale of 1–10, to answer this question. The first question of interest in this regard is whether the group assignment method matters for the results. To answer it, they compared results in the Original treatment, where group assignments were made based on painting preferences, and the Random Within treatment, where groups were formed based on random assignments. No significant difference in the proportion of participants who differentiated between ingroup and outgroup matches were found between these two treatments. Further, the two methods of group assignment did not have any significant impact on self-reported group attachments. Therefore, at least for the class of games analyzed here, random assignments seem to do just as well to generate the group effects and, given its simplicity, can be preferred. The next question that deserves attention is whether categorization into groups by itself is sufficient to produce group effects, or is it necessary for group members to interact and build solidarity by, say, helping each other like in the online chat task. To analyze the effect of the online chat stage, the No Chat and Original treatments can be compared. It was seen that there was no significant difference in the proportion of participants showing ingroup favoritism or outgroup discrimination in these two treatments. At the game level, there was evidence in only one of the 24 games that the online chat stage leads to significantly stronger ingroup favoritism. The chat stage, however, was found to generate significantly higher self-reported group attachment, which can be considered as the affective aspect of the group identification process. The other-other allocation task was also found to have no significant effect on participant behavior, nor the self-reported attachment to groups. Finally, on comparing the No Help and Original treatments, it was found that a (weakly) significantly larger proportion of subjects from the No Help treatment made group contingent decisions, compared to the Original treatment. The authors suggest that this may be because group effects induced by categorization have a tendency to deteriorate over time. So this larger proportion might be due to the fact that the game stage followed right after the group assignment stage in the No Help treatment. Next, we focus a bit on work that has looked into the connection between patterns of discrimination induced via lab-generated identities and their relationship with natural identities. Kranton and Sanders (2017) perform such an exercise. They conduct MGP style experiments with simple dictator game like allocation tasks as in Chen and Li (2009). However, unlike the latter, they do not just estimate average behavior. Rather, using a within subject design, they study individual behavior and are able to tightly classify individuals as “groupy” and “non-groupy.” Individuals are deemed to be “groupy” (“non-groupy”) when their estimated utility-type is different (not different) when allocating income to someone in their group than when allocating income to someone out of their group.1 They then use subject-specific data to investigate psychometric, demographic, and economic correlates of this groupy versus non-groupy distinction. Interestingly, they find that non-groupy participants are not distinguished by any basic demographic, nor by any of the Big Five
1
For instance, suppose they are willing to sacrifice their income to reduce advantageous inequality when interacting with an ingroup partner but not when they are interacting with an outgroup partner.
270
A. Borah
personality factors, but are distinguished by lack of political affiliation. Independents are least likely to be groupy, and this difference is significant in their nationwide US sample. On the other hand, participants who are groupy are more likely to identify as Republicans or Democrats, more so the former. Another interesting observation that emerges from their experiment is that participants in the US counties with large drops in the share of employment in manufacturing also are more likely to be groupy. These findings are reaffirmed in the experimental findings reported in Kranton et al. (2020). In this experiment, along with the MGP treatment, there was also a political treatment which divided subjects into groups based on their political leanings. They find that political party members showed more ingroup bias than independents who professed the same political opinions. At the same time, the greater bias of these individuals was also present in an MGP treatment with lab-induced identities. In other words, groupys (non-groupys) remain so irrespective of whether group classifications are made based on natural affiliations or lab-induced ones. Another interesting contribution using the MGP paradigm is Currarini and Mengel (2016), who attempt to uncover the relationship and interplay between homophily and ingroup bias. They set up an experiment that records subjects’ behaviors towards ingroup and outgroup members in social preference games of the type in Charness and Rabin (2002) under two different matching institutions. Subjects were first randomly assigned a group (RED or BLUE), after which two different matching treatments were employed. Under the first matching protocol (EXO), one’s partner (RED or BLUE) was determined randomly. Under the second (ENDO), each subject’s partner preference and their willingness to pay (WTP) for the same were elicited and were then paired with either a RED or BLUE member (with probabilities based on their WTP). The authors find that even under their MGP set-up, in the ENDO treatment, 45% of the subjects were homophilous with a positive WTP. Further, they find that whereas in the EXO treatment there is significant evidence of ingroup bias, with subjects 34% more likely to act positively reciprocal and 39% less likely to act negatively reciprocal in ingroup matches compared to outgroup matches, aggregate ingroup bias either diminishes or totally vanishes (statistically) in ENDO. Hence, to the extent that participants’ expectations are correct, the strong preference for homophily cannot be entirely explained by the anticipation of ingroup biases. The authors, therefore, contend that the nature of the matching institution, specifically the ability to self-select into groups, may produce a shift in behavior and a reduction in ingroup bias. In other words, and perhaps counterintuitively, their evidence suggests that homophily may not be brought about by the anticipation of ingroup bias. Rather, the possibility to express homophilous preferences may end up reducing such biases.
Natural Identities The chapter will now look at ways in which natural identities have been used in lab or field experiments to determine group-based attitudes in the dictator, ultimatum, and trust games. The key question that confronts experimenters in this regard is how
11
Methodological Approaches to Understanding Discrimination:. . .
271
to make natural identities salient in an experiment. This is a nontrivial task for several reasons. First, there is no guarantee that such identity-based attitudes are cognitively salient to subjects at the time they participate in the experiment: how does the experimenter make them salient? Second, the strength of identity-based attitudes may vary across people, even if they are all marked by that dimension of identity. For instance, we are all marked by race, but our race sensitive attitudes are not of the same magnitude. If these differences matter to the question that the experimenter is interested in, then how does she elicit this information? Third, people’s natural identities are multidimensional. If the experimenter’s goal is to activate only one or two dimensions of that identity, how does she go about doing that? We will now look at a few design choices that have been employed by experimenters to get at these questions in the context of the games under consideration. One prominent way in which natural identities can be made salient in experiments is by using the family names of subjects. In the context of many identities, family names reveal the specific identity category to which an individual belongs. For instance, in the context of the caste system in India, an individual’s name very often reveals the caste to which she belongs. As it turns out, names not only reveal identities but given how central names are in the context of social cognition, also produce nontrivial identity based effects. To illustrate this, consider the following experiment reported in Fershtman and Gneezy (2001) that sought to examine ethnic discrimination in Israeli Jewish society by simultaneously using the trust, ultimatum, and dictator games. The two major Jewish ethnic groups are Ashkenazic Jews (European and American immigrants and their Israeli-born offspring) and Eastern Jews (Asian and African immigrants and their Israeli-born offspring), with the former better placed on multiple dimensions compared to the latter. To check for discrimination, the authors formed two groups by enlisting students with typical ethnic names from two different sets of universities. A student from the first group was then randomly matched with a student from the second group to form pairs. Each subject in a matched pair was told the name of their partner, which in the Israeli context, provides an indication of ethnic affiliation. The other thing that names reveal is the gender of the subjects, a fact that the experimenters also utilize. The authors find that in the trust game, out of an endowment of NIS 20, the average transfer to an Ashkenazic male partner was 15.15 whereas the average amount transferred to an Eastern male partner was 8.06. When attention was restricted to male senders, the transfer amounts were 17.16 and 5.62, respectively. Further, the mistrust of Eastern origin men was found to be common among men of all ethnic origins, including Eastern origin ones. At the same time, no evidence was found that for any given amount received, Eastern male partners sent back an amount that differed significantly from that received by an Ashkenazic male partner. This, therefore, rules out statistical discrimination as an explanation for the difference in amounts transferred to Eastern male partners. Could the explanation be in terms of a taste for discrimination? To test this the authors implemented a dictator game with the same matching protocol as in the trust game. The results of the dictator game showed that although there is some differential treatment of groups by ethnicity,
272
A. Borah
there is no clear systematic taste for discrimination. The distribution of transfers to the Ashkenazic players was only marginally different from the distribution of transfers to Eastern players. The average transfers were similar. Therefore, the conclusion from simultaneously interpreting the results of the trust and dictator games is that the discrimination seen in the former is because of mistaken ethnic stereotypes associated with Eastern origin subjects and neither attributable to a taste for discrimination nor statistical discrimination. Interestingly, another dimension of ethnic stereotyping came out from the analysis of the ultimatum game that was administered using a similar matching protocol. Evidence of ethnic discrimination was found in this game, with Eastern players receiving larger transfers than Ashkenazic players. This discrimination, the authors contend, was the outcome of a common ethnic stereotype in Israeli society according to which men of Eastern origin are believed to react more harshly if treated unfairly. Finally, another key finding of the paper is the gender differences that were observed. First, no evidence of ethnic discrimination towards women was found in the trust game when they played the role of trustees. Second, Ashkenazic women are less trusted than Ashkenazic men, whereas Eastern women are more trusted than Eastern men. Further, women’s trust in their game partners was not found to be based on ethnic affiliation or on gender. Accordingly, women trust Ashkenazic male players less than men do and trust Eastern male players more than men do. A technique that experimenters often use to make natural identities salient in experiments is priming whereby some aspect of identity that is relevant for the study is made cognitively salient for subjects in a subtle way. One way of doing this is by exposing them to textual or audiovisual representations associated with that identity, or by making them perform a task that makes aspects of that identity salient. For instance, consider the following dictator game experiment conducted by Fong and Luttmer (2009) that was designed to examine the role of racial group loyalty on generosity in the United States. Specifically, the study wanted to see how race affected giving to Hurricane Katrina victims. In the study, to experimentally manipulate subjects’ beliefs about race, income, and worthiness of hurricane victims, they were first exposed to an audiovisual presentation with photographs of people after the hurricane along with an audio story about the residents of two cities (Slidell, LA and Biloxi, MS) that were hit by Katrina. The presentation consisted of a slide show with eight photographs of victims that were mainly White in one treatment condition, and mostly Black in the other. The photos showed devastation caused by Katrina such as flooding and demolished houses as well as relief aid received by victims. The audio information that accompanied the pictures varied along the lines of economic status, political leaning, religiosity, crime rates, and other demographic information for the city and was chosen so as to be potentially correlated with subjects’ perceptions about the racial composition of the city. Another aspect of design that comes out in this experiment is the need that experimenters feel in many situations to elicit subjective measures of identitybased attitudes to check whether variations in these attitudes have explanatory power. For instance, in the context of this study, it may be the case that subjective racial identity may matter more than objective race. There are different ways in
11
Methodological Approaches to Understanding Discrimination:. . .
273
which experimenters have tried to elicit such attitudes. One way is to administer attitudinal surveys measuring explicit attitudes. However, the drawback with them is that respondents may censor themselves in such surveys so as not to appear bigoted. At the other extreme are implicit measures like the implicit association test (IAT). In this experiment, the authors used a different approach. They asked subjects a simple question to measure subjective racial identification: “How close do you feel to your ethnic or racial group? Very close, close, not very close, not close at all.” Sometimes using a simple measure like this can be useful as it is less demanding to administer than an IAT. At the same time, it presumably does not suffer from the problem of respondents giving biased answers. Subjects then played a dictator game in a laboratory setting in the following form. They were asked how they would like to split $100 between themselves and a charity (Habitat for Humanity) that benefits Katrina victims in the city they saw in the presentation. Interestingly, the authors found no effects of victims’ race on giving, on average. However – and this is why subjective racial attitudes matter – they did find that subjects who report feeling close to their racial/ethnic group give substantially more when the victims are of the same race than those who respond that they do not feel close to their group. Whites who reported being “close” or “very close” to their ethnic or racial group gave roughly $17 less when seeing pictures of Black victims rather than White ones. In contrast, Whites who said they were “not very close” or “not close at all” gave roughly $13 more in response to pictures showing Black victims. On the other hand, Blacks who reported feeling close to Blacks gave $16 more in response to pictures showing Black victims. Blacks who did not feel close to Blacks gave $72 less in response to pictures showing Black victims. These differences are all significant at the 1% level. Another way in which identities have been primed in this class of experiments is by making subjects perform a task or participate in an activity. McLeish and Oxoby (2011) provide an example of this in the context of an ultimatum game experiment which studied the effect of social interactions and salience in social identity on cooperativeness and expectations. They prime the subjects, all students of the same university, by asking them to write for 10 min about either a positive experience with a student from the same university (identity-priming treatment), or a positive experience with a student of another university (distinctiveness-priming treatment), or detailed directions from their dorm room or parking space to the laboratory (no-prime/control group). The first two treatments worked to bring attention to shared or distinct identities, while the control group was designed to not affect the salience of any identity. Subjects then played the ultimatum game. The authors find that subjects are most cooperative and have the highest demands in the identitypriming treatment, and are least cooperative and have the lowest demands in the distinctiveness-priming treatment. This shows that a salient social identity and greater social interactions increases willingness to cooperate, but also increases expectations. Yet another way in which group sensitive attitudes can be elicited in experiments is by drawing on people’s social networks in the real world. An example of such an approach is Glaeser et al. (2000), which attempts to study discrimination as it relates
274
A. Borah
to social capital using a trust game. To that end, they conducted an experiment to measure trust and trustworthiness for individuals in different nodes in a friendship network. The subjects were recruited from an introductory economics course at Harvard University and participated in a trust game. At the experiment site, subjects were paired with another subject in the order of arrival (those who arrived together could play together, raising the likelihood that those who knew each other would play with one another). Observe that the design ensures that the subjects know the identity of the other subject in their pair. It is precisely this knowledge that creates variation in social connection with some subject pairs being well acquainted with one another while others being relative strangers. This allows the experimenters to study how different levels of social connection introduce discrimination in trusting behavior. After being paired, subjects jointly filled out a social connection survey which included nine questions about the social links between the partners (for example, mutual acquaintances). They were then separated and played a trust game in which the experimenter doubled the invested amount. Half of the recipients were given the opportunity to send a message (nonbinding promise) to the sender of their intended future action before the sender decided how much to give to the recipient. Their results support the hypothesis that social connection between the sender and the recipient increases both trust and trustworthiness. A one standard deviation increase in the amount of time that the pair knew each other raised the amount sent by 80 cents (approximately 19%) and return ratio by 5% (significant at the 90% and 95% level, respectively). Similarly, the number of common friends had a positive, though statistically insignificant, effect on the amount sent and the return ratio; ten extra friends were found to raise the return ratio by 2.6%. The authors also find a small but insignificant negative effect on the amount sent, but a large and very significant effect on the return ratio when testing whether individuals from different countries trust each other less. Race also was found to have a significant negative effect on the return ratio. 92% of the cases where the recipient sent back nothing occurred when the individuals were of different races. Finally, it was seen that an individual’s social capital – e.g., having better educated parents, working fewer hours for pay, having more friends, etc. – strongly predict the amount of money that as senders they receive back from recipients.
Dictator Games In this section, lessons from other experiments in the literature that have used the dictator game to study group-based discriminatory attitudes are highlighted. One important question that has interested researchers is about how young in life do parochial and discriminatory attitudes emerge. Fehr, Glätzle-Rützler, and Sutter (2013) study this question through a lab dictator game experiment. In their experiment, each subject was matched with an anonymous partner belonging to the same age cohort, and had to choose between two allocations in the three binary choice problems listed below:
11
Methodological Approaches to Understanding Discrimination:. . .
275
• Prosocial game (giver, partner): (1,1) or (1,0) • Envy game: (1,1) or (1,2) • Sharing game: (1,1) or (2,0) They classify subjects as egalitarian if they choose the allocation (1,1) in both the prosocial and envy games; altruistic if they choose (1,1) in the prosocial game and (1,2) in the envy game; and spiteful if they choose (1,0) in the prosocial game, (1,1) in the envy game, and (2,0) in the sharing game. In order to study the development of parochialism, they implemented an ingroup and an outgroup condition across subjects. While the recipient in the ingroup condition was known to be from the same class (his or her identity remained secret, of course), the recipient in the outgroup condition attended another school, but was in the same grade. Their experimental results show a strong decrease in spitefulness with age. Further, egalitarian considerations peak around the age of 8–11 years. As children move into their adolescent years, they are increasingly influenced by efficiency concerns and altruism becomes more important, with altruistic types being the modal type among 16–17 year olds. When it comes to parochialism, the noteworthy finding of the study is that it emerges in adolescence. What is more, it is particularly prominent among the altruistic types with significant ingroup favoritism evident in this group at the age of 14–15 years, and spitefulness is significantly stronger towards outgroup members from the age of 12–13 years onwards. On the other hand, for egalitarian types, there is no significant difference between the ingroup and outgroup conditions. In other words, there is joint development of altruism and parochialism starting with adolescence. The data from the envy game also reveals that the decline with age in the relative frequency of choosing (1,1) is much steeper for the ingroup than the outgroup condition. This indicates that as subjects get older, they are relatively more willing to accept disadvantageous inequality in the ingroup than in the outgroup condition. Finally, it was seen that females were more frequently classified as egalitarian than males. Continuing with the theme of gender differences in discriminatory behavior, Angerer et al. (2017) conducted an experiment with 824 children from Meran, a bilingual city in Italy, to examine how discrimination rates differ with gender for primary school children. Every subject is anonymously matched with two others from the same school grade but different schools, with the two recipients being from German-speaking and Italian-speaking schools, respectively. This is a variant of the dictator game, and the subject is given four tokens which they have to divide between the two students, and keep nothing for themselves. Although the majority of the students divide the tokens equally, 24.5% of the boys and 16.7% of the girls allocated three of four tokens to the receiver from their own language group. The authors conclude that gender differences in discriminatory behavior emerge early in life, with stronger discrimination by men found in studies with adults having roots in early childhood. In another experiment with children, Friesen et al. (2012) study ethnic discrimination in Canadian children. The children are made to play the dictator game three times, each time with 12 stickers that they have to divide among themselves, a White
276
A. Borah
recipient, an East Asian recipient, and a South Asian recipient. Identities are made salient by showing subjects photos of the recipients before the game. The authors find that only 16% of White children and 21% of South Asian children favored each of the other groups, while the percentage was 31% for East Asian children. They conclude that children from the dominant White identity tend to favor White recipients, while East Asian students do not favor recipients from their own ethnic identity and may, in fact, show outgroup favoritism. This reiterates a message we learned from Fong and Luttmer (2009) – it is not ethnicity per se but rather ethnic identification that may drive discriminatory behavior. An important question relates to how an individual’s different identities work to guide group-sensitive behavior. Ravetti et al. (2019) study discrimination in the presence of multiple identities – specifically, ethnicity, and union membership – with coal miners in Mpumalanga Province of South Africa. The experiment’s sample selection relied on three criteria: union membership, ethnolinguistic identity, and gender. Participants were randomly paired, and each subject played an anonymous dictator game posing both as a dictator (50 rounds) and as a recipient (50 rounds). To further study the effect of varying information sets, the authors had three treatments. In the first, dictators had no information about the recipient (first 10 rounds). In the second, the dictator received information regarding the recipient’s ethnolinguistic identity and union status (next 20 rounds). Finally, in the last 20 rounds, the dictators had the option of signaling their own union membership and ethnic status to the recipient after being informed of their counterpart’s characteristics. This treatment was meant to test whether the givers would try to convey their status to recipients (for example, when they were being more generous to a certain group due to their affiliation). The authors found that union and ethnic identity operate in contrasting fashions. Unionization acts as a factor of workers solidarity, with unionized dictators making relatively more generous offers not just when matched with other unionized workers but also with nonunionized workers. On the other hand, ethnicity drives discrimination among workers, with dictators from the ethnic majority group making relatively less generous offers to not just matches from the ethnic minority groups but their own group as well. An interesting question pertains to how group-based attitudes vary with economic circumstances. Aksoy and Palma (2019) conducted a two-period lab-in-the-field experiment with coffee farmers in a small, isolated village in Guatemala. The first period was before the harvest season (scarcity period) and the second period was during the harvest season (abundance period). In both periods, subjects played a sequence of games, one of which was the dictator game. They employed a 2 2 within-subjects design with subjects playing four rounds of the dictator game with: (1) an ingroup recipient and (2) an outgroup recipient; and during (1) abundance and (2) scarcity periods. The groups were formed based on natural village identities. They found that in the abundance period, subjects gave significantly higher amounts to ingroup members (about 33% of the endowment) than to outgroup members (about 22% of the endowment). In the scarcity period, however, this favoritism fades, with a statistically significant increase in giving towards the outgroup member (about 31% of the endowment) whereas giving towards the ingroup member does
11
Methodological Approaches to Understanding Discrimination:. . .
277
not change. They conclude that scarcity eliminates the ingroup bias in prosocial behavior, notably by an increase in giving towards the outgroup rather than a decrease in giving towards the ingroup. Another important question pertaining to discrimination is whether it is driven more by positive attitudes towards the ingroup or negative ones towards the outgroup. In a field experiment, Grosskopf and Pearce (2017) attempt to answer this question by studying discrimination based on ethnicity among the poorest people in England. Subjects are first asked to divide a sum of money between two strangers; they then play the standard dictator game with another stranger. In both rounds, the ethnicity of the stranger(s) was varied. The ethnic identity of the recipient was revealed to the subjects through the surname of the recipient, which was of either Western European or English origin (e.g., Smith), or of Muslim origin (e.g., Islam). They found that subjects give around half as much to recipients with Muslim surnames as compared to those with English surnames or those who are anonymous. This evidence allowed them to conclude that observed discriminatory attitudes are guided more by outgroup negativity rather than ingroup favoritism. Kolstad and Wiig (2013) attempted to study whether education reduces ingroup favoritism by conducting a field experiment in Luanda, Angola. The experiment consisted of two rounds of an anonymous, one-shot standard dictator game. In the first round, subjects had to divide a sum of money between themselves and someone from the same credit group, and in the second round, they did so with someone from an outside credit group. The results support ingroup favoritism among microcredit clients. Moreover, there is a positive causal effect of education on ingroup favoritism, which goes against the belief that education broadens an individual’s worldview. Ben-Ner et al. (2009) investigated favoritism for ingroup versus outgroup along multiple identity categories (body type, political views, nationality, religion, etc.) in four alternative contexts, one of which was the dictator game (The other contexts were working with another person on a critical project, commuting with another person, and sharing an office with another person). They found that ingroup members are treated more favorably than ones from the outgroup in all the four context and in nearly all identity categories, with family and kinship, political views, religion, sports-team loyalty, and music preference being the significant identity categories that produce discrimination. Surprisingly, the effect of gender was found to be insignificant. This section concludes by noting an important insight from Ockenfels and Werner (2014), who investigated the role of beliefs on ingroup favoritism using a dictator game. In keeping with existing literature, they found that if the identities of the dictator and recipient are publicly known, then dictators discriminate in the favor of ingroup members. However, there is substantially less ingroup favoritism if the dictator is informed that the recipient is unaware of the shared group membership. Moreover, in an environment where dictators are ex ante uninformed about the recipient’s identity, when given the choice, they prefer to avoid information about it if it guarantees that their identity is concealed from the recipient. These results, therefore, indicate that ingroup favoritism is not just a simple function of shared
278
A. Borah
group identity but rather the belief of the dictator about the recipient’s belief (that is, the dictator’s second-order belief) is also an important driver of it.
Trust Games Some further lessons learnt from trust games are presented in this section. To understand the role of religiosity in trust, Chuah et al. (2016) conducted a binary trust game experiment with “low” and “high” stakes. To study discriminatory attitudes, they used different social identities, with religion and religiosity being their main variable of interest.2 They measured religiosity based on the denominationrobust eight-item instrument by Rohrbaugh and Jessor (1975) which takes into consideration different dimensions of religion (such as belief, ritual, etc.) and delivers an individual’s score between 0 and 32. The authors found that senders of all levels of religiosity believe that receivers of high religiosity are trustworthier and discriminate by being more likely to trust them more than receivers of no or lower religiosity. They also found that religiosity enhances the ingroup favoritism shown by senders towards receivers of their own religious affiliation, showing higher trust for ingroup members. Interestingly, they show that religious participants believe that those belonging to some faith are trustworthier, but only invest more trust on those of the same religion. Finally, they find that religiosity is positively associated with the general willingness of senders to discriminate across a range of nonreligious social identities. Overall, they conclude that interpersonal similarity in religiosity promotes trust. Burns (2012) conducted a standard trust game experiment in six public schools in Cape Town with Black, White, and mixed race students. The race of the partner was made salient through photographs. The experimental results show that Black proposers make significantly lower offers (25% of their endowment on average) than non-Blacks proposers (38%), and Black responders receive significantly lower offers (27% of the proposer’s endowment on average) compared to non-Black responders (37%). Lower offers are Black responders is evident in the behavior of all proposers, including Blacks proposers. These lower offers may be attributable to an expectation that Black responders would remit less than others on average. (The actual results show Black responders remitted 26% of the tripled offer compared with 37% for non-Blacks.) Notably, non-Black proposers with Black partners are significantly more likely to make a zero-offer, preferring to “opt out” of any interaction with a Black partner, or make significantly lower offers, even after taking expectations into account. Finally, they find that responders are significantly less likely to make any return offer to a Black proposer. This suggests a systematic pattern of distrust towards Black partners, surprisingly even by Black proposers, which the author partially attributes to mistaken expectations.
2
The subject pool consisted of students in China, Malaysia, and the UK.
11
Methodological Approaches to Understanding Discrimination:. . .
279
Falk and Zehnder (2013) conducted a trust game field experiment with about 1000 residents, representing all of the 13 districts of the city of Zurich. In the experiment, proposers could condition their investment decision on the residential district of the responder. The authors find that trust levels measured in terms of the amount invested differ systematically across the residential districts of the responders, i.e., there is trust discrimination across districts. These differences are accounted for by differences in beliefs about district-specific levels of trustworthiness. In terms of significant correlates, the richer the district, higher is the level of investment, and the higher the fraction of foreigners and levels of religious fragmentation, the lower are investments. The results also reveal that different levels of investment into districts are significantly correlated with the respective willingness to repay on the part of the responders. This indicates that proposers correctly anticipate the relative trustworthiness of inhabitants of different districts and discriminate on the basis of this belief. Finally, it was found that there is ingroup favoritism with proposers investing significantly higher amounts into their own districts compared to other districts. Haile, Sadrieh, and Verbon (2008) employed the trust game to investigate the effect of heterogeneity in income and race on cooperation in South Africa. Each participant played the trust game, and decided how much to transfer as a sender (from an initial endowment of 20 Rand) and transfer back as a receiver (for each of 11 possible outcomes). Additionally, each subject also reported the amount they expected as a return on their own transfer as a sender as well as the amount expected as an investment in the role of a receiver. The authors varied the information that was given to the participants regarding their counterpart; in the “information treatment,” information on both race and income level were given, whereas no such information was given in the “no information” treatment. The authors find no significant differences in behavior when no information is provided. However, when the information is available, it significantly affects trust behavior. Specifically, the low-income subjects from both racial groups invest significantly less in partnerships with the highincome subjects of the other racial group than in any other partnership. The authors attribute this behavior to cross-racial envy. Etang, Fielding, and Knowles (2011) studied the impact of social distance on trust and trustworthiness using experimental data from Cameroon. Subjects played the standard trust game as senders with a partner, who could be from the same village as them or from a different village. They found that senders send significantly more money when the responder is from the same village (74% of the endowment on average) compared to when they are from a different village (63%). This effect was also seen with gender, education, and membership of rotating credit groups. As for reciprocity, there was no significant difference between the amounts sent back to ingroup versus outgroup members. They conclude that trust diminishes with social distance. Meier et al. (2016) conducted trust game experiments in high schools in two neighborhoods in the metropolitan area of Palermo, Italy, to study the effect that an environment of crime has on trust and trustworthiness. The subjects were similar in terms of ethnicity and religion. What distinguished them was whether they were
280
A. Borah
from a high or low Mafia area. They authors found that in the trust game, students who were from a neighborhood with high Mafia involvement displayed lower generalized trust and trustworthiness. At the same time, they showed higher ingroup favoritism. These findings support the argument that the historical informal institution of organized crime can undermine current institutions. Tsutsui and Zizzo (2014) examined the effect of group status on trust levels. Subjects were randomly assigned to one of two groups – “Blue” or “Red”/“Not Blue.” The “Not Blue” group in the instructions was referred to as “not belonging to any group,” or “outsiders to the group,” reflecting lower status as compared to the “Blue” group. The groups were of different sizes (four or eight individuals), reflecting minority or majority status. In stage 1, subjects played the standard trust game with no information about their matched partner’s group. In stages 2–4, they played three games as proposers and three as receivers, receiving information about the average rate of return and investment on a round-by-round basis, as well as with information on the anonymous partner’s group affiliation. Before stage 2, their WTP/WTA for trading group membership was elicited. The experimental findings show that subjects of low status (“Not Blue”) are less trusting of others of low status, and that minority subjects return more to majority subjects. Subjects like being in majorities and dislike being in a low status group. Further, subjects who value their group more return comparatively less to insiders relative to outsiders. Hargreaves Heap and Zizzo (2009) implement a similar design involving the trust game with artificial groups as the one above. In their case, the artificial groups are labelled “Blue” and “Red.” They too find that trust falls with groups because of discrimination against outgroup members. Burnham, McCabe, and Smith (2000) studied the effect of wording on trust to examine the hypothesis that human subjects have a preconscious friend-or-foe mental mechanism for evaluating the intentions of another person. They experimented with a variant of the trust game under conditions of both single play and repeat play. To implement the two treatments – friend and foe – subjects were primed by using the word “partner” or “opponent,” respectively. They found that in the single-play condition, partners and opponents are equally trusting, but partners are significantly more trustworthy than opponents. On the other hand in repeat play, partners are most trusting than opponents, and marginally significantly more trustworthy than opponents. However, they find that both trust and trustworthiness erode over time. Finally, note that one can also find examples in the literature where the presence of salient groups does not produce discriminatory trust attitudes. For instance, Bouckaert and Dhaene (2004) investigated interethnic trust and reciprocity with a sample of male small businessmen, who were either of Turkish or Belgian ethnic origin and owned small businesses in the city of Ghent. The participants played the standard trust game with each pair of matched subjects knowing each other’s first name (which signaled their ethnic origin). Their findings suggest that the average trust and the average reciprocity of Belgian and Turkish participants are independent of ethnic affiliation and, significantly, independent of the ethnic origin of the opposite party.
11
Methodological Approaches to Understanding Discrimination:. . .
281
Ultimatum Games In this section, the focus is on some additional insights from experiments involving the ultimatum game. Ferraro and Cummings (2007) conduct an ultimatum game experiment with Hispanic and Navajo subjects in Albuquerque, New Mexico, with matched pairs bargaining over $10. The innovation in the design was that before playing the game, the subjects in each session were placed together in a room to allow them to observe the ethnic composition of their session. So, even though they did not know the ethnicity of their partner in the game, they knew the composition of the group. There were four experimental sessions: all Hispanic, majority Hispanic, all Navajo, and majority Navajo. The authors find that Hispanic responders have higher minimum acceptable offers than Navajo responders. They also find that both groups tend to discriminate against the other group: their mean reservation prices increase with an increase in the proportion of subjects from the other ethnic group. As for the proposers, the authors find that while playing with their own ethnic group, Navajos make significantly lower offers than Hispanics ($3.83 on average as compared to $4.90 respectively). Finally, they note that the ethnic composition of the group affects the offers made: Hispanic offers decrease with an increase in the proportion of Navajo subjects, while Navajo offers increase with the proportion of Hispanic subjects. Based on subjects’ elicited beliefs about likelihood of proposals being accepted, it is also evident that statistical discrimination does not play a role in the behavior of Hispanic proposers and is rather taste based. Boarini, Laslier, and Robin (2009) look at a transcontinental ultimatum game to better understand bilateral bargaining when one of the parties is from a high-income country (France) and the other from a low-income one (India).3 The two transcontinental treatments are French proposer to Indian receiver (FtoI), and vice versa (ItoF), with two other within-country benchmark treatments. In each treatment, subjects played six one-shot ultimatum games with a $10 endowment. The results of the experiment indicate, first, that on average, Indian senders were more generous toward French receivers ($3.92) than French senders toward Indian receivers ($2.63). Indeed, both French and Indian senders send lower offers when the receiver is Indian ($2.63 and $3.53, respectively), and with French receivers, Indian senders offer more money than French senders ($3.48). Second, they find that when low offers (less than $3) are made, the rejection rate by Indian receivers (19% for French senders and 25% for Indian senders) is far less than French receivers (55% for Indian senders and 60% for French senders). It was also evident that offers decreased to Indian receivers over each round of the game. Overall, bargaining in the FtoI treatment mostly ended up with unequal splits of money in favor of French subjects, while nearly equal splits were the most frequent outcome in ItoF treatment interactions. The authors attribute the observed differences in the experiment to the fact that the value of the experimental currency (US Dollar) in terms of the goods it can buy is
3
The experimental sessions were conducted in the years 2002 and 2003.
282
A. Borah
more in India than in France, and, on average, the experimental participants in France would be wealthier than those in India – with both these observations reiterated as part of the experimental design. Finally, we conclude by referring to an ultimatum game experiment by Eckel and Grossman (2001) that tries to study how attitudes of and towards different sexes affects bargaining behavior. In this study, there are three possible pairings: either subjects know the sex of their partner and are paired with a subject of the same or the opposite sex, or the sex of their partner remains unknown. The experiment is carried out in eight repetitions of the ultimatum game, with a sum of $5 to be divided. Consistent with known insights from the literature, they find that women’s proposals are on average more generous than men’s, regardless of the sex of the partner, and women respondents are more likely to accept an offer of a given amount. What is noteworthy is that offers from female opponents were significantly more likely to be accepted, a result they refer to as chivalry. At the same time, women paired with women almost never fail to reach an agreement, a result they term solidarity.
Cross-References ▶ Field Experiments: Correspondence Studies ▶ Gender-Based Discrimination in Health: Evidence from Cross-Country ▶ Lab-in-the-Field Experiments ▶ Taste-Based Discrimination Acknowledgments I thank Ananya Sen (Pomona College), Reem Qamar (Ashoka University), and Saujanya Bharadwaj (Ashoka University) for excellent research assistance.
References Aksoy B, Palma MA (2019) The effects of scarcity on cheating and in-group favoritism. J Econ Behav Organ 165:100–117 Angerer S, Glenn Dutcher E, Glätzle-Rützler D, Lergetporer P, Sutter M (2017) Gender differences in discrimination emerge early in life: evidence from primary school children in a bilingual city. Econ Lett 152:15–18 Ben-Ner A, McCall BP, Stephane M, Wang H (2009) Identity and in-group/out-group differentiation in work and giving behaviors: experimental evidence. J Econ Behav Organ 72(1):153–170 Berg J, Dickhaut J, McCabe K (1995) Trust, reciprocity, and social history. Games Econ Behav 10(1):122–142 Boarini R, Laslier J-F, Robin S (2009) Interpersonal comparisons of utility in bargaining: evidence from a transcontinental ultimatum game. Theor Decis 67(4):341–373 Bouckaert J, Dhaene G (2004) Inter-ethnic trust and reciprocity: results of an experiment with small businessmen. Eur J Polit Econ 20(4):869–886 Burnham T, McCabe K, Smith VL (2000) Friend-or-foe intentionality priming in an extensive form trust game. J Econ Behav Organ 43(1):57–73
11
Methodological Approaches to Understanding Discrimination:. . .
283
Burns J (2012) Race, diversity and pro-social behavior in a segmented society. J Econ Behav Organ 81(2):366–378 Camerer CF (2011) Behavioral game theory: experiments in strategic interaction. Princeton University Press, Princeton Charness G, Rabin M (2002) Understanding social preferences with simple tests. Q J Econ 117(3): 817–869 Chen Y, Li SX (2009) Group identity and social preferences. Am Econ Rev 99(1):431–457 Chuah SH, Gächter S, Hoffmann R, Tan JHW (2016) Religion, discrimination and trust across three cultures. Eur Econ Rev 90:280–301 Currarini S, Mengel F (2016) Identity, homophily and in-group bias. Eur Econ Rev 90:40–55 Eckel CC, Grossman PJ (2001) Chivalry and solidarity in ultimatum games. Econ Inq 39(2): 171–188 Etang A, Fielding D, Knowles S (2011) Does trust extend beyond the village? Experimental trust and social distance in Cameroon. Exp Econ 14(1):15–35 Falk A, Zehnder C (2013) A city-wide experiment on trust discrimination. J Public Econ 100:15–27 Fehr E, Schmidt KM (1999) A theory of fairness, competition, and cooperation. Q J Econ 114(3): 817–868 Fehr E, Glätzle-Rützler D, Sutter M (2013) The development of egalitarianism, altruism, spite and parochialism in childhood and adolescence. Eur Econ Rev 64:369–383 Ferraro PJ, Cummings RG (2007) Cultural diversity, discrimination, and economic outcomes: an experimental analysis. Econ Inq 45(2):217–232 Fershtman C, Gneezy U (2001) Discrimination in a segmented society: an experimental approach. Q J Econ 116(1):351–377 Fong CM, Luttmer EFP (2009) What determines giving to Hurricane Katrina victims? Experimental evidence on racial group loyalty. Am Econ J Appl Econ 1(2):64–87 Forsythe R, Horowitz JL, Savin NE, Sefton M (1994) Fairness in simple bargaining experiments. Games Econ Behav 6(3):347–369 Friesen J, Arifovic J, Wright SC, Ludwig A, Giamo L, Baray G (2012) Ethnic identity and discrimination among children. J Econ Psychol 33(6):1156–1169 Glaeser EL, Laibson DI, Scheinkman JA, Soutter CL (2000) Measuring trust. Q J Econ 115(3): 811–846 Grosskopf B, Pearce G (2017) Discrimination in a deprived neighbourhood: an artefactual field experiment. J Econ Behav Organ 141:29–42 Güth W, Schmittberger R, Schwarze B (1982) An experimental analysis of ultimatum bargaining. J Econ Behav Organ 3(4):367–388 Haile D, Sadrieh A, Verbon HAA (2008) Cross-racial envy and underinvestment in South African partnerships. Camb J Econ 32(5):703–724 Hargreaves Heap SP, Zizzo DJ (2009) The value of groups. Am Econ Rev 99(1):295–323 Kolstad I, Wiig A (2013) Does an educated mind take the broader view? A field experiment on in-group favouritism among microcredit clients. J Socio-Econ 45:10–17 Kranton RE, Sanders SG (2017) Groupy versus non-groupy social preferences: personality, region, and political party. Am Econ Rev 107(5):65–69 Kranton R, Pease M, Sanders S, Huettel S (2020) Deconstructing bias in social preferences reveals groupy and not-groupy behavior. Proc Natl Acad Sci U S A 117(35):21185–21193 McLeish KN, Oxoby RJ (2011) Social interactions and the salience of social identity. J Econ Psychol 32(1):172–178 Meier S, Pierce L, Vaccaro A, La Cara B (2016) Trust and in-group favoritism in a culture of crime. J Econ Behav Organ 132:78–92 Ockenfels A, Werner P (2014) Beliefs and ingroup favoritism. J Econ Behav Organ 108:453–462 Ravetti C, Sarr M, Munene D, Swanson T (2019) Discrimination and favouritism among South African workers: ethnic identity and union membership. World Dev 123:104604
284
A. Borah
Rohrbaugh J, Jessor R (1975) Religiosity in youth: a personal control against deviant behavior? J Pers 43(1):136–155 Tajfel H, Turner JC (1986) The social identity theory of intergroup behavior. In: Worchel S, Austin W (eds) Psychology of intergroup relations. Nelson Hall, Chicago, pp 7–24 Tajfel H, Billig MG, Bundy RP, Flament C (1971) Social categorization and intergroup behaviour. Eur J Soc Psychol 1(2):149–178 Tsutsui K, Zizzo DJ (2014) Group status, minorities and trust. Exp Econ 17(2):215–244 Turner JC (1985) Social categorization and the self-concept: a social cognitive theory of group behavior. In: Lawler EJ (ed) Advances in group processes: theory and research, vol 2. JAI Press, Greenwich, pp 77–122
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
12
Stephanie Seguino
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Economic Stratification and Stereotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Allport and Blumer on Prejudice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Norms and Stereotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereotypes and Group Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereotypes, Discrimination, and Policing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
286 287 288 289 290 292 295 296
Abstract
Economists studying discrimination often assume rational conscious decisionmaking, ignoring the role of stereotypes – generalizations about groups – that can influence decision-making in ways that lead to racial inequality. Stereotype formation is not accidental but rather is shaped by a society’s prevailing stratification system. In stratified societies, negative stereotypes about subordinated groups are internalized within individuals in their daily actions and decisionmaking in the form of implicit or unconscious bias. This mechanism serves to reproduce and reinforce racial hierarchies, based on stereotype-induced discrimination and its intergenerational transmission of racial bias that leads to discrimination. This chapter explores these topics and applies them to our understanding of the sources of racial disparities in policing. Keywords
Stereotypes · Implicit bias · Racial discrimination · Racial profiling · Stratification · Policing S. Seguino (*) Department of Economics Fellow, Gund Institute for the Environment, University of Vermont, Burlington, VT, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_47
285
286
S. Seguino
Introduction The research conducted by economists that explores race and gender discrimination in labor markets relies primarily on regression analysis techniques that control for individuals’ productivity-related characteristics. Although a race “residual” continues to be observed in wage regressions, some economists dispute the conclusion that racial bias factors into hiring and wage decisions. Instead, they argue that inequality is due to the existence of “unobserved productivity differentials.” Still others note, however, that race is “more than a dummy variable” – that the race variable, in other words, is insufficient to identify the causal mechanisms, sometimes unquantifiable, by which race matters and that racial inequality is produced (Figart 1997). The debate has been rendered lifeless and adrift, in part due to a lack of a welldeveloped theoretical framework for understanding racial bias beyond Becker’s “taste for discrimination.” Research based on field experiments – in particular, job audits – provides a welcome addition to the toolkit in the study of racial discrimination in labor markets as well as in other domains such as housing and bank lending. In one of the earliest studies of this kind, Devah Pager and Bruce Western (2009) matched teams of testers who then applied for entry-level jobs in New York City. Their results reveal how employers respond to applicants who are equally qualified but vary by race, ethnicity, and criminal record. Applicants with a criminal record were less likely to be called back for an interview than those without a record. But black applicants without a criminal record received a smaller percentage of callbacks than white applicants with a criminal record. Given racial disparities in the criminal justice system, negative stereotypes about the criminal background of all black applicants could well have dampened the employment opportunities of even those without a criminal record. The Pager and Western study highlights the influence of negative racial stereotypes on economic opportunities, especially for African Americans. “Ban the box” efforts that had begun years earlier as a result gained momentum. “Ban the box” is a slightly misleading term; the policy does not deny employers access to applicants’ criminal background information. It simply defers criminal history until later in the process, to give all applicants a fairer chance to be judged on their qualifications. From a social psychology perspective, this is a good thing. Research shows that what we learn first about another person colors how we see the information we learn next – the primacy effect (Willis and Todorov 2016). By moving the criminal history to the end of the application, those reviewing applications can avoid reacting to negative stereotypes they might hold about the formerly incarcerated long enough to also absorb information on the applicant’s actual job qualifications. Some studies evaluating the impact of “ban the box” on racial bias, however, have revealed the deep reach of negative (antiblack) racial stereotypes. A disturbing finding from one study is that with the adoption of “ban the box,” racial discrimination worsened for black applicants without a criminal record although callback rates for formerly incarcerated black applicants increased (Doleac and Hansen 2016). Put differently, in the absence of information on the criminal
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
287
background of the applicant, employers use race as a proxy for criminal history. Employers are, in a word, racially profiling. The benefit of the job audit methodology used in these studies is that it allows us to isolate the effect of racial stereotypes on employer decision-making. The results suggest that the underlying determinants of discrimination derive at least in part from antiblack and anti-brown racial norms and stereotypes, a subject matter that has been understudied by economists who instead typically assume rational and thus conscious decision-making on the part of employers. Although audit studies are a step forward in providing evidence of racial discrimination, that is not sufficient. Also needed is a theoretical framework for understanding the origin of and motivation for negative racial norms and stereotypes. Statistical discrimination has been the go-to explanation, based on the notion that generalizations about the characteristics of a group are used in place of an examination of an individual’s actual characteristics because information is costly to obtain. This approach, too, begs the question of why at least some portion of statistical discrimination is based on faulty stereotypes that remain unchanged in the face of evidence to the contrary. In this chapter, I seek to shed light on this important lacuna in discrimination research, first by positing stratification theory as a framework for understanding the origins and perpetuation of negative and inaccurate racial norms and stereotypes. Secondly, I apply this to an analysis of traffic policing data in the United States. Finally, I discuss research from social psychology experiments that serve as an alternative methodology to shed light on racial stereotypes and discrimination.
Economic Stratification and Stereotypes Stratification economics is the study of the processes and institutions that result in a hierarchical economic and social ordering based on ascriptive characteristics. Unlike mainstream economics, whose unit of analysis is the individual, stratification economics emphasizes the role of groups and the structural forces that shape intergroup inequality. Stratification economics explores the intentionality of a dominant group to maintain its privileged economic position, a distinctly different approach than the mainstream’s lens which typically lays the causes of intergroup inequality at the feet of the “deficits” of groups lower on the social and economic hierarchy. The mechanisms by which intergroup inequality is transmitted over time rely on a stratification infrastructure, the latter comprised of racial norms and stereotypes that are internalized at the individual level and act as a “stealth” factor in reproducing hierarchy (Jost and Banaji 1994; Augoustinos and Walker 1998; Seguino 2021). These norms and stereotypes are constructed to advantage dominant group(s) and disparage subordinate groups in ways that justify the latter’s exclusion from access to and control over material resources. Before turning to an elaboration of the concepts of norms and stereotypes from social psychology, I discuss the underpinnings of economic stratification based on intergroup rivalry.
288
S. Seguino
Allport and Blumer on Prejudice Gordon Allport published The Nature of Prejudice (1954) at a time when the Civil Rights movement in the United States was gaining momentum. Social psychologists define prejudice as a negative attitude toward people in a distinguishable group, based solely on their membership in that group. The characteristics assigned to members of that group are negative and applied to the group as a whole. Allport’s research represents a second wave of prejudice theory in which he argued that it is rooted in normal processes of categorization and stereotyping. Allport and subsequent social psychologists have argued that the human mind cannot avoid creating categories, putting some people into one group based on certain characteristics and others into another group based on their different characteristics (Dovidio and Gaertner 2010). We can consciously manage only a small amount of the information we are exposed to. Categorization is thus an adaptive mechanism to conserve on cognitive resources required to process information, it is argued. The output of categorization is stereotypes – generalizations about groups, based on membership in that group. (The concept of statistical discrimination is related to this aspect of information processing and will be discussed later in this chapter.) Although it is normal for humans to categorize, Allport argued that racial animus and prejudice are due to faulty stereotypes. The antidote, according to Allport, is intergroup contact under specific conditions. That contact should take place such that the various groups possess equal status and common goals. This structured contact is hypothesized to give those holding prejudicial views the opportunity to confront and revise their faulty stereotypes. His work led to the emergence of contact theory. Allport did not explore the source of “faulty” stereotypes. Herbert Blumer, however, did. In his 1958 article, “Race Prejudice as a Sense of Group Position,” Blumer argued that prejudice is a tool of stratification (Blumer 1958). Central to Blumer’s theory is the idea that prejudice is not an individual feeling but rather is an outcome of how racial groups form mental constructs of themselves and of other groups, resulting in a sense of relative group position. To characterize the outgroup at the same time also defines one’s own group. Prejudice and the resulting mental constructs serve as constituent components of the infrastructure that facilitates the exploitation and exclusion of subordinate groups. Blumer identifies four “feelings” present in the racially dominant group: (1) a feeling that the dominant group is superior, (2) a feeling that the subordinate group is intrinsically different and alien, (3) a feeling that the dominant group has a proprietary right to prized economic assets and privileges, and (4) fear and suspicion that the subordinate harbors designs on the material assets and privileges of the dominant group. Blumer’s work is a prescient contribution to economic stratification theory. His work spawned the emergence of threat theory, whereby it is theorized that the larger the size of the subordinate group, the more the dominant group feels a threat to its own interests, exacerbating the negative attitudes toward the subordinate group.
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
289
Norms and Stereotypes Stereotypes are oversimplified generalizations about groups of people. They may be positive or negative. In either case, the stereotype is a generalization that does not take individual differences into account. I focus on racial stereotypes and outcomes in this chapter, but a similar application of stereotypes to gender inequality has been made (Seguino 2013). What do we mean by racial stereotypes? Racial stereotypes in stratified societies describe the ways in which racial groups presumably differ, usually in ways that justify the racial division of labor and unequal access to and control over material resources. Stratification-inducing stereotypes typically describe members of the dominant group as harder working, more intelligent, and less inclined than other groups to criminal behavior. Stereotypes thus legitimize the “superiority” or greater deservingness of white – often male – control, undervaluing men and women of subaltern groups, and white women. This is not to suggest that women of the dominant group (white women) do not benefit from their interfamilial relationships with white men, thus participating in and benefiting from the racial hierarchy. Another way to think about this is that white women, though part of the subordinate group in a gender stratified society, possess the protective factor of whiteness in a racially stratified society. A fuller development of stratification theory is needed, however, to understand how racial and gender stratification systems interact, complement, or substitute for each other. Negative racial stereotypes of the subordinate group contribute to and reinforce hierarchical racial norms. Norms are rules, often informal, that specify acceptable behavioral, social, and spatial boundaries for members of racial groups. The effect of such norms is strong because the consequence of violating norms is social approbation. One of the strongest drivers of human behavior is our need for belonging. Norms thus exert a powerful effect on behavior, related to our desire to avoid the risk of rejection and social exclusion. Norms and stereotypes are mutually causative. On the one hand, stereotypes contribute to racial discrimination, for example, in access to good jobs, based on the cultivation of beliefs that subordinate groups are less intelligent and less hardworking and other negative attributes. This contributes to job segregation, with subordinated racial groups in the lowest wage, least secure jobs and dominant racial groups holding the highest quality jobs with greater managerial and supervisory responsibility. A norm – a behavioral expectation – emerges whereby the unwritten rule is that members of the subordinated group are tasked to “stay in their place” with members of the dominant group expected to be the “rightful” sole occupiers of positions of high decision-making responsibility and power. Norms also shape the perceptions and expectations of subordinate group members regarding what jobs and positions they might feasibly aspire to. To the extent norms restrict the subordinate group’s access to important positions and resources, negative stereotypes are reinforced. This observation is linked to social role theory which, although focused on gender roles, is instructive in highlighting the impact of everyday observations of the differences in dominant and subordinate group roles on a group’s assessments of
290
S. Seguino
what its proper role is in society. Movement of subordinate groups into roles they had previously been excluded from or for which they are underrepresented can help to dispel negative stereotypes. There is mounting evidence, for example, that samerace teachers are beneficial to students from underrepresented racial groups in the United States. Gershenson et al. (2021) find that black students who had just one black teacher by third grade were 13% more likely to enroll in college and those who had had two were 32% more likely.
Stereotypes and Group Position If, as Blumer argues, a dominant group works to intentionally maintain its superior position, concrete mechanisms are required to achieve that goal. Brute force is of course one means. But “turf maintenance” (Darity et al. 2010) may be more subtly and extensively accomplished through the creation and perpetuation of negative stereotypes of the subordinated group(s). Stereotypes inform everyday decisionmaking that produces and reproduces the power and privilege of the dominant group, influencing who gets bank loans, who is elected to political office, and what jobs are filled and by whom. And, because stereotypes influence individual decision-making, often at an unconscious level, they act as a “stealth” factor in enforcing a hierarchical system. A deeper understanding of the role of stereotypes as a mechanism to preserve group position can help to inform theoretical and empirical research on discrimination. Social psychologists have developed a series of theories of the determinants of stereotype construction. Earlier theories related the construction of stereotypes to ego-justification, that is, as a mechanism to preserve self-esteem. This individualistic view (whereby stereotypes emerge at the individual level) has been disputed, with social psychologists subsequently arguing instead that stereotypes serve the function of group and system justification (Jost and Banaji 1994; Augoustinos and Walker 1998). Group justification theories posit that stereotypes emerge to protect the status of a social group as a whole. System justification theory posits that stereotypes emerge to justify existing systemic social arrangements at the expense of (out)group interests. Stereotypes, in other words, serve the ideological function of reinforcing the dominance of one group over other(s) by “explaining the poverty or powerlessness of some groups and the success of others in ways that make these differences seem legitimate and even natural” (Jost and Banaji 1994: 10). This theoretical stance stands in contrast to the view that stereotyping is merely a means to economize on cognitive resources. Further supporting the notion that the function of stereotypes is to legitimize unequal systems, evidence shows that stereotypes are not evenly distributed across groups. Rather, it is the subordinate groups that are more likely to be stereotyped (Banaji and Greenwald 2016). Further, those with more power (ability to control) show a greater tendency to stereotype than those with low power (Glaser 2015). In one study, for example, mangers and subordinates in the hotel industry assessed the
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
291
suitability of job applicants. They were primed by being shown stereotyped information about the applicants. Subordinates paid more attention to the individual applicants’ attributes than did managers, whose judgments of the outgroup were influenced by the consistency of the association of the stereotypes with the applicants (Guinote and Phillips 2010). Recent research has demonstrated the unconscious nature of stereotyping (commonly referred to as implicit bias), which may explain the sometimes unwitting participation of broad swaths of society in perpetuating negative racial stereotypes without the dominant group visibly wielding its power (Banaji and Greenwald 2016; Eberhardt 2019). The dissemination of stereotypes occurs not only at the level of the family and friends but also in our broader culture – via media, religious organizations, and political institutions, to name a few – the preponderance of which are headed by members of the dominant group. The now famous “doll test,” which revealed that even very young children hold negative racial stereotypes, speaks to the depth and breadth of our exposure to stereotypes that delegitimize and pathologize members of subordinate groups (Clark and Clark 1939). Social role theory argues that simple observation on a daily basis of social hierarchies (whites as managers, blacks as janitors; women as caretakers, men as breadwinners), facilitated by job and spatial segregation, normalizes stereotypes and indeed provides evidence of their veracity. Stratification theory then posits a very different approach to understanding discrimination than mainstream economists who instead emphasize statistical discrimination. The latter is assumed to be based on rational decision-making by actors who assess the credentials or qualities of, for example, job seekers, based on group averages rather than the individual’s attributes. The underlying theory is that information is imperfect and incomplete and is costly to acquire for profit-maximizing agents. Imperfect information means that relevant qualifications or attributes cannot be easily observed. Even in cases where assessments are based on some kind of assessment (SATs, personality tests, etc.), some aspects of the job may require skills or traits not easily discerned in tests (e.g., honesty, integrity, drive), with the result that employers revert to stereotypes that may lead test scores to be discounted. Not all adherents to statistical discrimination theory hold that stereotypes are an accurate reflection of group averages. Loury (2002), among others, posits that people start with stereotypes – correct or incorrect – for historical reasons. In this case, stereotypes become self-fulfilling and self-reproducing. In sum, from a stratification theory perspective, stereotypes and the resulting implicit bias are situated in individuals who transmit bias and discrimination in their everyday behaviors. Collective action is not needed. All that is required is for the dominant group to have sufficient ability to influence the cultural production of stereotypes. With members of dominant groups overpopulating the culturally and politically important roles of movie producers, media moguls, newspaper editors, politicians, and CEOs, everyday social interactions result in a diffuse but extensive dissemination of negative stereotypes about the outgroup. Discrimination metastasizes throughout the social body, long after evidence of any explicit bias has diminished.
292
S. Seguino
Stereotypes, Discrimination, and Policing Stereotypes contribute to discrimination because they are the lens through which we evaluate others. There is thus a good deal to learn from the application of social psychology research on stereotypes as applied to racial disparities in the criminal justice system and, in particular, race data in traffic policing. Traffic policing is the most common interaction people have with law enforcement. Because traffic stops generate large data sets, they provide a useful window into racial discrimination in other aspects of policing that can help us to understand the impact of stereotypes on discrimination in other domains, such as labor markets, healthcare, and housing. Numerous studies document the higher stop, arrest, and search rates of black and Latino/a drivers as compared to white drivers in the United States (Baumgartner et al. 2018; Pierson et al. 2020). An analysis of a national sample of almost 100 million traffic stops finds that black drivers nationally are about 50% more likely to be stopped by the police than white drivers (Pierson et al. 2020), although in some jurisdictions the disparity is much higher. For example, Baumgartner et al. (2018) who analyzed data on 20 million North Carolina traffic stops found that black drivers were 63% more likely to be stopped even though, as a whole, they drive 16% less than white drivers. Taking into account less time on the road, police were 95% more likely to stop black as compared to white drivers. Stop rate disparities in and of themselves do not prove bias toward black drivers, in part because of the benchmarking problem, that is, the lack of precise knowledge of the racial composition of drivers on the road. Post-stop outcomes, such as search rates by race, offer a more reliable means to assess racial discrimination in policing. This is because researchers know with certainty both the number of searches by race and the number of traffic stops by race. There is overwhelmingly consistent evidence that black drivers in the United States are searched at a substantially higher rate than white drivers, and similar, although less consistent evidence of oversearching of Latinx drivers as compared to white drivers. While national data indicate that black drivers are twice as likely to be searched as white drivers, in some jurisdictions, the disparity is much higher. For example, in Vermont, black drivers are 3.5 times more likely to be searched than white drivers, and in the town of Brattleboro, Vermont, the black/white search rate ratio reaches 8.8 (Seguino et al. 2021). Here, too, it may be argued that this disparity is not necessarily evidence of bias if drivers of color are found to more frequently be transporting contraband. To address this, researchers have identified several tests for racial discrimination in traffic policing. The “veil of darkness” test is based on the hypothesis that police are less likely to know the race of a motorist before making a stop after dark than they are during daylight. The test for racial discrimination in the decision of whom to stop (i.e., racial profiling) is based on a comparison of the racial shares of traffic stops made during daylight to the racial shares made after dark. Pierson et al.’s (2020) national study found that black stop rates are higher during the day than at night,
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
293
consistent with a hypothesis of racial bias in the decision to stop vehicles. (It should be noted that the “veil of darkness” test may not be valid for smaller towns in which the police may position themselves under streetlights or other lit areas at night, thus enabling them to discern the race of the driver.) Another approach to identifying the role of bias in search rate disparities is the “hit rate” test, derived from a game theoretic framework. The underlying model on which the hit rate test is based assumes that in the absence of racial bias, officers pursue a search strategy that maximizes the number of successful outcomes (hits), where a successful search outcome is defined as uncovering some contraband such as illegal drugs, weapons, or stolen goods. The hypothesis is that in equilibrium, absent bias on the part of officers, hit rates will be equal across racial groups. Hit rates that are lower for some racial groups than others are indicative of biased decision-making in whom to search. This is because a lower hit rate for some racial groups (e.g., blacks) than others means police officers are wrong more often in their decision to search black drivers than those of other racial groups. Police departments could improve the allocation of police resources by reducing searches of those groups with lower hit rates until hit rates equalize across racial groups. The failure to do so implies racially biased behavior. Numerous studies have found evidence of racial bias in the decision to search, based on the hit rate test (Baumgartner et al. 2018). An interpretation of those results is that the police rely on a lower threshold of evidence to initiate a search of a vehicle with a black driver than with a white driver. In those cases, the evidence suggests oversearching of black drivers and/or undersearching of white drivers. Finally, some studies rely on logit regressions to assess the probability of a search, controlling for other factors in addition to the race of the driver that might explain the decision to search. Studies control for the age and gender of the driver, the reason for the stop, and the day and time of the stop, and some studies include data on the race, age, gender, and tenure of officers conducting the search. If, even after controlling for other factors, race is a statistically significant predictor of a search, then that provides additional evidence that the race of the driver, independent of these other factors, influences traffic policing. Similar analyses have been conducted on the probability of finding contraband during a search (Seguino et al. 2021). Where these tests reveal racial discrimination in traffic policing, they do not necessarily identify the source of the bias. Is it located in the individual officer who may have a “taste for discrimination” – that is, who is explicitly or implicitly biased? Is it a function of leadership’s deployment decisions, with communities of color more heavily monitored than white neighborhoods? Are the disparities due to a few “bad apples” or to an organizational environment in which bias is condoned or even encouraged? It is possible that case studies can help us answer some of these questions. When the tests reveal racial bias, regardless of what level of an organization is responsible for that bias, an underlying factor is negative racial stereotypes – whether held by police leadership or by specific officers. Police make decisions with incomplete information, and stereotypes help fill that void. But the tests just described are not direct tests of stereotype-driven bias.
294
S. Seguino
We can, however, turn to social psychology experiments to help illuminate the role of racial bias in decision-making. Perhaps the most famous experimental design is the Implicit Association Test (IAT). In this computer-administered test, participants rapidly categorize two opposing concepts (e.g., bad/good, honest/dishonest) with a person’s perceived race (e.g., black/white, although tests exist for other racial groups as well). Pairings more aligned with our underlying stereotypes and thus implicit biases are easier and thus faster to make, while contradictory pairings (not aligned with our unconscious biases and stereotypes) are harder to make, leading us to respond more slowly. The faster one associates positive words with white and negative words with black, the stronger the automatic preference for whites is assumed to be. Almost 75% of those who have taken the IAT in the United States exhibit strong automatic white preference (Banaji and Greenwald 2016). The IAT assesses attitudes, but does that translate into discriminatory behavior? In a metaanalysis of 122 research reports, Greenwald et al. (2009) find that IAT scores are predictive of bias in consequential situations, such as medical treatments, hiring, and voting. With regard to policing, shooter bias tests similarly reveal underlying implicit (antiblack) racial biases (Glaser 2015). In these tests, participants are positioned in front of a video monitor with hands on a control stick and are shown a series of photographs of black and white men holding either a gun or a harmless object like a cell phone. Participants are instructed to shoot when they observe that any of the men is perceived to be holding a gun. The results indicate that participants tend to shoot black men more quickly and erroneously shoot more unarmed black men than white men. Correll, Park, Judd, and Wittenbink (2007) further investigated whether stereotypes cause shooter bias. The authors primed participants in the study, manipulating the strength of association between blacks and guns by having participants carry out the shooter bias experiment, but with different proportions of armed blacks and whites. (Priming is used in social psychology experiments to activate an association.) Priming, that is, exposure to a stimulus, is often so brief as to be inaccessible to the conscious mind, even while it registers in the unconscious mind. This was followed by repeats of the task, with results showing that those who had been exposed to more black shooters in the previous experiment were more likely to show shooter bias. Glaser and Knowles (2008) found that IAT scores that measure the strength of association with between blacks and whites and weapons was a good predictor of shooter bias. Still other experimental designs reveal the depth of negative racial stereotypes that inform policing and the racially disparate results observed in traffic stop studies. Wilson, Hugengbeg, and Rule (2017) asked subjects to evaluate people’s perceptions of physical stature. They found that when asked to rate the height, weight, and strength of young black and white men from photographs showing only their face, black men were consistently rated as taller, heaver, and stronger than white men. The authors then evaluated whether this racial bias was related to their perception of the
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
295
capacity to do harm. They found that white participants rated black men as capable of doing more harm than white men of the same physical stature. The long history of scientific racism that both dehumanized people of African descent and implanted notions of racial inferiority persists and influences our perceptions today. A series of text messages exchanged by police officers in San Francisco in 2016, for example, described black and other minorities as “wild animals, cockroaches, savages, barbarians, and monkeys” (Eberhardt 2019: 145). To test the impact of such imagery on behavior, Goff et al. (2008) showed study participants a video of officers surrounding and beating a suspect whose racial identity could not be clearly determined. Some were led to believe that the suspect was white and others that the suspect was black. Participants who had been primed with animal imagery were more likely to believe the beating was justified but only when the suspect was black. In still another study of this kind, nonblack study participants were shown a series of faces and asked to imagine that the person shown had behaved aggressively toward a police officer but was not carrying a weapon. Study participants indicated that the police would be justified in using more force to subdue the black men than white men (Eberhardt 2019). Glaser (2015: 86) summarizes these research findings: “Whether we like it or not, the implicit stereotypes our culture feeds us, associating blacks with aggression, danger, weapons, put all of us at risk of committing discriminatory acts like racially biased, wrongful shooting.” The impact of stereotypes and biases is not limited to shooter bias but instead is likely representative of their much broader impact on discriminatory behavior in other domains.
Conclusion Economists studying discrimination often assume rational conscious decisionmaking. While stereotypes – generalizations about groups – can influence decisions, there has been little exploration – at least by economists – of the origin or content of stereotypes and their resulting impact on social norms. Social psychologists have explored these topics in great depth. Norms and stereotypes are the lens through which we evaluate others. Stratification theory further helps us to understand how intergroup inequality is produced and reproduced. In stratified societies, negative stereotypes about subordinated groups, and resulting norms, are internalized within individuals in their daily actions and decision-making. This mechanism serves to reproduce and reinforce racial hierarchies, based on stereotype-induced discrimination. As such, economists’ assumption of exogenous preferences is flawed. Rather, social psychology research provides a foundation for understanding the endogeneity of our perceptions and thus decisionmaking, determined in large part by the cultural environment in which we live. It
296
S. Seguino
offers a means to understand the intergenerational transmission of racial bias, leading to discrimination. The production of negative racial stereotypes and the resulting effect on norms create the invisible infrastructure of racial discrimination. Absent a centralized authority and racial “policing” to enforce the racial hierarchy, this is a critical mechanism by which dominant groups maintain their privileged position.
References Allport GW (1954) The nature of prejudice. Addison-Wesley, Reading Augoustinos M, Walker I (1998) The construction of stereotypes within social psychology: from social cognition to ideology. Theory Psychol 8(5):629–652 Banaji, MR, Greenwald, AG (2016) Blindspot: Hidden biases of good people. Random House, New York Baumgartner F, Epp D, Shoub K (2018) Suspect citizens: what 20 million stops tell us about policing and race. Cambridge University Press, Cambridge, UK Blumer H (1958) Prejudice as a sense of group position. Pac Sociol Rev 1(1):3–7 Clark K, Clark M (1939) The development of consciousness of self and the emergence of racial identification in negro preschool children. J Soc Psychol 10:591–599 Correll J, Park B, Judd C, Wittenbink B (2007) The influence of stereotypes on the decision to shoot. Eur J Soc Psychol 37:1102–1117 Darity W, Mason P, Stewart J (2010) Stratification economics: economics and social identity. Mimeo. https://www.aeaweb.org/conference/2011/retrieve.php?pdfid¼213 Doleac J, Hansen B (2016) “Does ban the box” help or hurt low-skilled workers? Statistical discrimination and employment outcomes when criminal histories are hidden. NBER working paper 22469 Dovidio J, Gaertner S (2010) Intergroup bias. In: Fiske S, Gilbert D, Lindzey G (eds) Handbook of social psychology. Wiley, Hoboken, pp 1084–1121 Eberhardt J (2019) Biased: uncovering the hidden prejudice that shapes what we see, think, and do. Penguin Books, New York Figart D (1997) Gender as more than a dummy variable: feminist approaches to discrimination. Rev Soc Econ 55(1):1–32 Gershenson S, Hart C, Lindsay D, Papageorge N (2021) The long-run impact of same race teachers. NBER working paper 25254 Glaser J (2015) Suspect race: causes and consequences of racial profiling. Oxford University Press, Oxford, UK Glaser J, Knowles E (2008) Implicit motivation to control prejudice. J Exp Soc Psychol 44:164–172 Goff P, Eberhardt J, Williams M, Jackson M (2008) Not yet human: implicit knowledge, historical dehumanization, and contemporary consequences. J Pers Soc Psychol 94(2):292–306 Greenwald A, Poehlman T, Uhlmann E, Banaji M (2009) Understanding and using the implicit association test: III. Meta-analysis of predictive validity. J Pers Soc Psychol 97(1):17–41 Guinote A, Phillips A (2010) Power can increase stereotyping: evidence from managers and subordinates in the hotel industry. Soc Psychol 41:3–9 Jost C, Banaji M (1994) The role of stereotyping in system justification and the production of false consciousness. Br J Soc Psychol 33:1–27 Kofi K, Guryan C (2008) Prejudice and wages: an empirical assessment of Becker’s the economics of discrimination. J Polit Econ 116(5):773–809 Loury G (2002) The anatomy of racial inequality. Harvard University Press, Cambridge, MA Pager D, Western B (2009) Investigating prisoner re-entry: the impact of conviction status on the employment prospects of young men. US Department of Justice 228584. https://www.ojp.gov/ pdffiles1/nij/grants/228584.pdf
12
Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
297
Pierson E, Simoiu C, Overgoor J et al (2020) A large-scale analysis of racial disparities in police stops across the United States. Nat Hum Behav. https://doi.org/10.1038/s41562-020-0858-1 Seguino S (2013) Toward gender justice: confronting stratification and power. Géneros 2(1):1–36 Seguino S (2021, forthcoming) Macroeconomic perspectives on stratification. In: Hamilton J, Dixit A, Edwards S, Judd K (eds) Oxford research encyclopedia of economics and finance. Oxford University Press, Oxford, UK Seguino S, Brooks N, Autilio P (2021) Trends in racial disparities in Vermont traffic stops 2014–19. University of Vermont/Cornell University, Burlington/Ithaca Willis J, Todorov A (2016) First impressions: making up your mind after a 100-ms exposure to a face. Psychol Sci 17:592–598 Wilson J, Hugengbeg K, Rule N (2017) Racial bias in judgements of physical size and formidability: from size to threat. J Pers Soc Psychol 113(1):59–80
Surveys, Big Data, and Experiments How Can We Best Learn About LGBTI Development Outcomes?
13
Dominik Koehler and Nicholas Menzies
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Definition: What Are “Lesbian,” “Gay,” “Bisexual,” “Transgender,” and “Intersex”? . . . . . . . . Sexual Orientation, Gender Identity and Expression, and Sex Characteristics . . . . . . . . . . . . Operationalizing the Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Asking About SOGI in Surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimating the Population Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data from the USA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data from Other Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nepal’s 2011 Census . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methodological Overview: How Developmental Outcomes for LGBTI People Have Been Measured Through Surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nepal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Australia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lessons Learned from Other “Hard to Survey” Populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Potential of Big Data in Building LGBTI Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Approaches to Measure Discrimination and Exclusion . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
300 301 302 304 305 308 308 310 310 311 311 313 314 316 317 318 319 319 320
Abstract
There is little rigorous quantitative data about the lives of lesbian, gay, bisexual, transgender, and intersex (LGBTI) people in developing countries. This makes the This chapter was originally published in 2017 as a working paper by World Bank in its POLICY RESEARCH WORKING PAPERS Series, and can be accessed at https://doi.org/10.1596/18139450-8154. Re-published here with permission. D. Koehler (*) · N. Menzies World Bank, Washington, DC, USA e-mail: [email protected]; [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_52
299
300
D. Koehler and N. Menzies
development of policy to improve the welfare of LGBTI people difficult, and it also makes it difficult to know whether sexual orientation and gender identity and expression focused policies and programs are working. Filling this data gap is necessary to understand the development outcomes for LGBTI people. Quantitative data practices exist that can be drawn on to fill the gap, including household surveys, experiments, and big data analysis. Summarizing existing experience, this chapter provides guidance on how to study development outcomes for LGBTI people, by paying attention to the different ways to define sexual orientation, gender identity and expression (SOGIE), and sex characteristics; and collecting samples that allow conclusions to be drawn among LGBTI people, as well as the general population. Keywords
Lesbian · Gay · Bisexual · Transgender · LGBTI · Sexual orientation · Gender identity · SOGI · Development outcomes · Population size · Surveys · Big data · Experiments JEL Classification
C83 Survey Methods and Sampling Methods · O35 Social Innovation
Introduction Each year in the USA, we see an increase in ads and TV spots during Pride month or at megaevents, such as the Super Bowl or the Grammys, that feature happy, wealthy, samesex couples. Seeing these ads, reading newspaper articles about the spending power of same-sex couples, and consuming popular images of lesbian, gay, and trans people on television or in movies, it would be easy to assume that LGBTI people are better off than most. However, appearances can be deceiving and what the media portrays is often far from the lived reality of most LGBTI people, in the USA and beyond. There is a tendency to conflate the experience of one subgroup, wealthier gay men, with the experience of the others tied up in the “LGBTI” acronym, when their opportunities and outcomes are quite different, and what we know about each subgroup varies significantly. The little rigorous data that are available suggest that LGBTI people are often among the most vulnerable in society and on average fare worse than non-LGBTI people.1 Unfortunately, these data are patchy, and so it is hard to say for certain – though theory and qualitative research support this overall finding. Of course, large differences exist across contexts. The scarce quantitative data that do exist are mostly limited to developed countries and focus on urban and cisgender2 male populations. The invisibility in the data
1
Valfort (2017) Cis gender refers to a person whose self-identity conforms to their biological sex, that is, not transgender.
2
13
Surveys, Big Data, and Experiments
301
makes it more difficult to persuade policy makers that LGBTI people should be specifically included in development programs, know how to target such programs to best address their needs, and measure whether such programs are working. To encourage development professionals to address the data gap, this chapter provides an overview of, and some reflections on, existing quantitative data collection efforts.3 The goal is to stimulate and inform future knowledge creation and avoid some obstacles previous studies faced. The chapter addresses issues of defining LGBTI communities, estimating the LGBTI population, the representativeness of samples (including comparing outcomes to non-LGBTI populations), using big data and experimental techniques. Primarily, we hope to reach empirical experts and researchers who are new to the topic to encourage more SOGIE inclusive, rigorous data collection. Secondarily, we hope that those with a background in SOGIE issues might read this and thus be more open to placing an emphasis on quantitative data. By highlighting the specific development outcomes of LGBTI people, policy makers who otherwise would not have recognized LGBTI people as an area of concern might be convinced to design more inclusive development programs.
Definition: What Are “Lesbian,” “Gay,” “Bisexual,” “Transgender,” and “Intersex”? In order to conduct effective research into a particular group of the population, it is necessary to define it. This can be very challenging for LGBTI people for many reasons. LGBTI – as the acronym suggests – is an agglomeration of a number of groups, each of which has their own characteristics and definitional issues (as well as development challenges and outcomes).4 The individual experiences of each subgroup tend to be overshadowed by the more visible and more studied experience of gay men. In particular, the experiences of trans, intersex, and bisexual people tend to be the most understudied.5 The ways in which the terms are understood, expressed, and labeled also significantly varies within and across countries. None of these are unique to LGBTI people as subjects of research, though together they present a
3
Both qualitative and quantitative data are necessary to make informed policy decisions, as well as create the political space to act. Existing data collection efforts have focused more heavily on qualitative data, and we therefore address quantitative data here as one way of encouraging researchers to address this imbalance. 4 Each of these intersects with other markers of identity that impact development outcomes, such as age, race, ethnicity, gender, etc. 5 We also note that there are other (interchangeable and additional) words that people use which are not directly reflected in the “L,” “G,” “B,” “T,” and “I” – which has led to the more recent adoption of “LGBTI+” (along with many other formulations). We celebrate this diversity and fluidity, but given the empirical focus of this chapter we limit ourselves as “LGBTI” as this is where (relatively) more data exist.
302
D. Koehler and N. Menzies
special challenge when it comes to conducting quantitative research that by its very nature emphasizes standardized questions and aggregate answers. Somewhat hesitantly – for fear of seeming to diminish the incredible diversity of sexual orientation, gender identity and expression (SOGIE), and sex characteristics – we present here some broad, higher-level definitions. While we have tried to formulate inclusive definitions, we recognize that some may work in some contexts and not in others. We do this in an attempt to guide common thinking in the drafting of surveys.6 One of the benefits of quantitative data is the ability to collect information on the experience of a large number of people and to compare across contexts. This benefit is enhanced with the use of comparable questions, methods, and analysis. But standardization does come with costs, including glossing over the richness and uniqueness of individual lives. Another risk is that the utilization of common definitions can undermine the collection of data in any given context if people do not understand the terms employed (see below). When it comes to formulating questions, each would need to be thoroughly context-tested prior to any survey and a case-by-case balance struck between local needs and cross-country comparability.
Sexual Orientation, Gender Identity and Expression, and Sex Characteristics Sexual orientation consists of three conceptional elements: (1) attraction, which refers to a person’s enduring emotional, romantic, or sexual attraction to a person of same and/or opposite gender; (2) behavior, which relates to the sex of sex partners (individuals of the same sex, different sex, or both sexes); and (3) self-identification of a person.7 Lesbian (women) and gay (men) refer to people who have an enduring emotional, romantic, or sexual attraction primarily or exclusively to people of the same gender. Homosexuality is another term frequently used to describe this attraction (compared to heterosexual or straight). Bisexual individuals can have the same emotional, romantic, or sexual attraction to another person regardless of gender.8 For the purposes of data collection by way of a quantitative survey, it is important to take into account that these elements do not always align. For example, it is common to find men who have sex with men (MSM), who might also be married to women, and who would not self-identify as gay or bisexual. Here one behavior (sex) is seemingly in contrast to another (marriage to an opposite sex spouse) as well as identity. It is therefore important to include questions related to all three dimensions of sexual orientation in a survey.9
6
For further reference: Yogyakarta Principles on the Application of International Human Rights Law in relation to Sexual Orientation and Gender Identity 7 Park (2016) 8 Other terms include pansexual: attraction to all sexes, genders, and gender identities; and asexual: attraction to none. 9 “Identity, behavior, attraction, and relationships all capture related dimensions of sexual orientation but none of these measures completely addresses the concept.” (Gates 2011)
13
Surveys, Big Data, and Experiments
303
Gender identity refers to a person’s deeply felt, internal identification as a man, woman, or some other gender, which can be different to the sex assigned at birth. Gender expression refers to a person’s outer appearance, speech, social interaction, and how others may perceive this. Transgender refers to an individual whose gender identity is different from the biological sex that was assigned at birth.10 By contrast, cisgender refers to people whose gender identity and sex assigned at birth are the same. Intersex is a scientific term which describes a variety of chromosomal, hormonal, and anatomical conditions or sex characteristics in a person that do not to fit the typical definitions of male and female. In countries around the world, there are many different understandings of sexual orientation, gender identity, and expression.11 Facebook’s 51 options for gender identity is just one indicator of this variety. Further complexity is added to the above definitions for a number of reasons:
Spectrums Sexual orientation, gender identity, and expression are commonly understood not to be binary but rather exist in a series of spectrums.12
Source: World Bank modeled after Center for Gender Santiy 2009
10
There are other terms used to describe gender identity, most commonly transsexual. Terms such as transvestite should not be confused with transgender, since it only refers to a person who dresses in clothes traditionally associated with the other gender. The term also has a negative connotation, with cross-dressers often preferred. 11 These include (among many, many others): third gender, a term common in Asia; two-spirited, a term used by Native Americans to describe a person that can fulfill both gender roles; Hijra a term used in India to describe transgender people mostly transgender-women. Because of its scientific character, this does not necessarily apply to intersex which can be considered as a more static concept compared to sexual orientation, gender identity, and expression. 12 The following diagram is adopted from: Center for Gender Santiy, 2009. “Sex and Gender diagram.”
304
D. Koehler and N. Menzies
Where a person finds themselves on the various spectra can change over time, and only taken together do they compose a person’s sexual orientation, gender identity, and expression.13 Some surveys have asked respondents to place themselves on one or more of these spectrums, while others simply offer a binary set of options for each, e.g., male/female or transgender yes/no. Sexual orientation, gender identity and expression, and sex characteristics also interact with each other. For instance, a transman, a transgender individual who identifies as a man,14 can be attracted to men, women, or both and thus might be straight, homosexual, or bi (or might adopt some other term).
Being Out In addition, people often hide their sexual orientation, gender identity and expression, and/or sex characteristics wholly or in part (i.e., from certain people and/or in certain circumstances). This is known in some places as being “closeted” (not revealing your status) – in contrast to being “out.” “Coming out” is the process in which an individual discloses their sexual orientation, gender identity and expression, and/or sex characteristics within their social surroundings.15 Someone who is “out” does not necessarily disclose their identity to everyone, everywhere, or all the time. Being out or not, and in which contexts, can affect the way in which a person is treated by their family, friends, employer, and society and thus has an impact on their lived experience and development outcomes. There are several ways of trying to measure this, including asking respondents to whom they have revealed their status and asking them how they think others perceive them.
Operationalizing the Definitions Defining the population that is the target of any data collection or knowledge generation effort is a necessary step in designing research projects but the operationalization of definitions can pose a significant challenge especially for hard to survey/hidden populations, such as LGBTI people. Researchers will need to decide which approach to adopt; there is no generalized right answer other than to be driven
13
There might be other dimensions which are relevant to a person’s sexual orientation, gender identity, and/or expression based on cultural difference or other circumstances that have not been taken into consideration for this chapter. 14 National Center for Transgender Equality (2014) 15 Hiding sexual orientation and gender identity are not directly comparable forms of the “closet” and need to be examined and researched on their own terms.
13
Surveys, Big Data, and Experiments
305
by the particular research questions, development problems, and context at hand.16 Following, we will use the example of two surveys to show potential was of addressing this challenge.
Asking About SOGI in Surveys The 2012 EU LGBT Survey by the European Union Agency for Fundamental Rights (FRA) recognized that “(. . .) the probability that a person identifies as LGBT (considering their sexual behavior or general preferences and/or identifying themselves as being gay/lesbian/bisexual) may vary across countries and social contexts, as well as change over time (during the life-course, for example)” (FRA 2013). They developed the following questions to ascertain sexual orientation, covering identification [A4], behavior [A7], and then attraction [A8].17
Source: FRA (2013)
16 17
Tourangeau et al. (2014) FRA (2013)
306
Source: FRA (2013)
Source: FRA (2013)
D. Koehler and N. Menzies
13
Surveys, Big Data, and Experiments
307
In addition, respondents were asked whether they identify as transgender. If they answered yes, they were asked to select between Transgender; Transsexual; Woman with a transsexual past; Man with a transsexual past; Gender variant; Crossdresser; Queer; or Other (with the option to briefly explain the response). This was followed by a set of questions which addressed gender identity on a more subconscious level, allowing researchers to gain a more complete picture of a person’s experience. The second example comes from a survey conducted by the Williams Institute in cooperation with UNDP in Nepal.18 It uses a similar set of questions to the FRA survey to cover the different dimensions of SOGIE; however, it shows one significant variation as a result of taking into account the local context. In the question, 16.1 researchers decided to merge the two concepts, sexual orientation and gender identity, into one question reflecting the local context. The other questions (16, 17, and 18) cover self-identity, attraction (“feelings”), and sexual behavior similar to the FRA.
Source: UNDP and Williams Institute 2014
Two very useful guides on how to survey for both sexual orientation (SMART guide) and gender identity (GenIUSS group) give an in-depth overview of how definitions might be operationalized and outline a number of challenges when surveying LGBTI communities. Both guides were developed for the USA and as such mainly use language that reflects sexual orientation, gender identity, and
18
UNDP (2014)
308
D. Koehler and N. Menzies
expression in that context. Below, we give an overview of how surveys outside of the USA have dealt with these difficulties and what lessons can be learned.
Estimating the Population Size The truth is that we do not know how many LGBTI people there are. Estimating the total number of LGBTI people who live in a given city, region, or country is difficult, not only because of the definitional issues raised above but also because LGBTI people are subject to stigma and marginalization in many countries and therefor often live in hiding.19 As with many hard to survey populations, it is challenging to tell whether LGBTI respondents are in a position to reveal the truth about these sensitive issues. Their trust to any outside entity, especially if it is a government sponsored survey, can be low and hence they are unlikely to disclose their sexual orientation, gender identity, and/or sex characteristics for fear of repercussions. Even in countries where LGBTI people face lower levels of discrimination, they might not be willing to disclose such deeply personal information. Collecting data in a way that does not inadvertently “out” people and being explicit to respondents about privacy measures to protect the confidentiality of participants and their data are critical in encouraging higher response rates and accurate answers. It is important to measure how many lesbian, gay, bisexual, transgender, and intersex persons live in any given location to fully understand the size of the development challenge and the magnitude of responses needed. In simple political terms, it can also be important for making the case that LGBTI people deserve a proportional slice of public and private resources.20 Understanding the size (and other characteristics) of the total LGBTI population is further useful for establishing the ground upon which detailed surveys of a subset of the population can be done.
Data from the USA According to the Gallup, for example, 4.3% of Americans self-identify as L, G, B, or T. This is an increase of 0.8% based on the last estimates from 2012.21 Washington DC has the highest percentage of LGBT-identifying residents at 8.6%, while in South Dakota only 2% self-identify as LGBT.22
19
Tourangeau et al. (2014) Of course, this can cut both ways in situations where the population is numerically small. This is not to undermine the human rights arguments that, no matter the size of a marginalized population, their human rights need to be protected, respected, and promoted. 21 Romero (2017) 22 Gates (2017) 20
13
Surveys, Big Data, and Experiments
309
Source: US Census Bureau 2010
In a study from 2011, the Williams Institute estimated there are roughly 8 million Americans who self-identify as lesbian, gay, or bisexual and 700,000 as transgender.23 Almost twice as many, an estimate of 8.2% (19 million) of the US population, report that they have engaged in same-sex behavior, and 25.6 million or 11% have acknowledged some sort of same-sex attraction.24 This shows the sharp contrasts between LGBT identity (4.3%), sexual behavior (8.2%), and sexual attraction (11%) components within the above definition. A Google US consumer survey found that 5.7% of respondents from a nationwide sample of 10,000 identified as LGBTI.25 It is particularly interesting that more than 10% of people aged 18–24 identified as LGBT compared to only 2–3% of people older than 45. The 2010 US Census asked questions on a person’s household, which allowed the identification of some same-sex households. After accounting for reporting issues, around 650,000 same-sex households were counted or around 0.6% of all US households.26 More recent data (2017) on same-sex marriages shows that there are nearly 1.1 million married LGBTI people in the USA or over 547,000 same-sex couples.27 This provides an alternate way to identify LGBTI people, yet it has significant limitations, including not accounting for single people and couples not sharing the same household.
23
Gates (2011) Ibid. 25 Only half of these reported being out at work. 26 Lofquist et al. (2010) 27 Romero (2017) 24
310
D. Koehler and N. Menzies
Data from Other Countries According to data from a 2016 online survey in 9 EU countries, 5.9% of Europeans self-identify as LGBT.28 Germany has the highest percentage (7.4%), followed by Spain with 6.9% and the UK with 6.5%, while in Hungary only 1.5% self-identify as LGBTI.29 An older survey from the UK (2010) suggests that roughly only 1.6% of the British population identifies as LGB with an additional 3% not able or willing to answer the question. A Norwegian survey from 2010 showed only 1.2% of the population identifies as lesbian, gay, or bisexual.30 A 2014 Australian survey came to the conclusion that 3% of men and 4% of women identified as nonheterosexual.31 A survey from ten Brazilian cities indicates that on average 7.8% of the surveyed men identified as gay and 2.6% as bisexual, totaling to 10.4% of the urban population. Of the surveyed women, 4.9% identified as lesbian and 1.4% as bisexual, resulting in a total of 6.3% lesbian and bisexual women. The National SocioEconomic Characterization Survey (CASEN) from Chile provides one of the few population estimates for transgender people (in this case specifically the gender category “other”). When comparing the responses for “sex assigned at birth” and “gender identity,” 2.7% of the Chilean population can be considered transgender.32 Because of its medical nature, population estimates for intersex people are somewhat more reliable. Widely accepted estimation by Anne Fausto-Sterling, in 2000, estimates that 1.7% of all live births are intersex.33 These examples highlight the broad range of population estimates for LGBTI people in different countries, and even within a given country, estimations can vary significantly over time as the example from the UK shows. To fully understand the development outcomes of LGBTI people, it is important for future research to address this knowledge gap.
Nepal’s 2011 Census In 2011, the Nepalese government included “third gender,” along with male and female in its census. “Third gender” was thought to be a locally resonant umbrella term for sexual orientation, gender identity, and expression. Unfortunately, a number of challenges were encountered which prevented a reasonable estimation of the population size.34
28
Based on a census-representative survey of 11,754 people conducted in August 2016 by Dalia Research. 29 Dalia Research (2016) 30 Gulløy and Normann (2010) 31 Richters et al. (2014) 32 Valfort (2017) 33 Fausto-Sterling (2000) 34 Park (2015)
13
Surveys, Big Data, and Experiments
311
The biggest problem turned out to be definitional ambiguity. Neither respondents nor enumerators had sufficient information on what the term “third gender” meant or who was supposed to be covered under it. Further, there was insufficient training provided to enumerators on how to properly ask the question and respond to questions from citizens. While some of the challenges could have been avoided by providing a guidebook and/or better training, there was also a bigger conceptual problem, revealed through a follow-up survey conducted by UNDP and the Williams Institute in 2014. As mentioned above, the respondents to the later survey – which allowed open-ended identification – used 21 terms to describe their sexual orientation and gender identity. Only 51.4% self-identified as “third gender,” posing the question whether other respondents would have chosen to identify as “third gender” if just that option was provided like it was the case in the census.35 After the Nepalese census attempt, censuses in India and Pakistan have moved forward with similar approaches to include transgender populations. The next census in the UK will likely be the first one in a developed country asking questions about sexual orientation and gender identity. In Uruguay, an effort is underway to conduct a census among transgender people.
Methodological Overview: How Developmental Outcomes for LGBTI People Have Been Measured Through Surveys In the following section, we will discuss four surveys of LGBTI people that have attempted to assess development outcomes: the previously mentioned survey in Nepal; a survey in India; the FRA survey from Europe; and the largest survey of intersex people from Australia. For each survey, we will focus on the methodologies, constraints, and lessons.
Nepal The Williams Institute and UNDP conducted one of the most recent surveys on sexual orientation, gender identity, and/or expression in cooperation with a local Nepalese NGO, Blue Diamond Society.36 Respondents were asked a set of questions, including open-ended questions, on their sexual orientation, gender identity, and/or expression covering the self-identification, behavioral, and attraction dimensions of all three. A number of survey enumerators from the Blue Diamond Society were recruited, trained, and ultimately able to reach 1178 respondents.37 Most of the respondents were recruited through local LGBTI networks based on drop-in centers the Blue Diamond Society manages throughout the country. This 35
UNDP (2014) This included a feedback on the methodology, survey tool, and data analysis. 37 UNDP (2014) 36
312
D. Koehler and N. Menzies
venue-based snowball convenience sample38 survey was developed to fit local circumstances. The sampling strategy is a combination of three distinct recruiting methods: venue-based, snowball, and convenience. Venue-based refers to the dropin centers that were managed by Blue Diamond Society that served as a primary venue to recruit respondents. The survey furthermore employed a snowballing approach where an initial number of subgroup members serve as “seeds” and recruit further respondents of the subgroup until the required sample size is reached. These “seeds” were selected on a nonprobability basis simply because they were conveniently accessible. All three methods are commonly used to survey hard to reach populations, but they have important limitations. In this case, for example, the sampling strategy led to very few women being interviewed. Respondents report an average income that is almost twice as high as the national average in Nepal, and they were also much better educated than the national population (18% had completed higher education compared to 10% nationwide). Further, only 14% undertook agricultural work39 compared to 75% of the national population.40 Leading to a sample that can hardly be considered representative. There are several explanations for this misrepresentation. Most critically, the survey only covered 32 of 75 Nepali districts and was not able to reach the most remote areas. The highest response rates were achieved in Kathmandu, the capital, explaining why the sample is skewed toward the more educated and wealthy. Further, the sampling strategy was built around a community network which is heavily focused on men, contributing to the fact that the survey only reached very few women (87.6% of respondents were assigned male at birth, 12% female, and 1% intersex).41
Methodological Takeaways • Extensive consultation and training are useful when surveying a population that is being marginalized and excluded. In this case, it helped avoid the problems that the Nepalese government encountered when conducting their census. One of the main findings of the survey was the diversity of terms used in Nepal to describe sexual orientation, gender identity, and expression. Providing respondents with locally relevant options to questions about sexuality and gender identity was essential. Designing the survey in a way that allows such diversity needs to be balanced with a way to aggregate data and allow some sort of generalization during data analysis. • Snowball surveys like this provide an opportunity to get an in-depth understanding of discrimination and exclusion in a community (even if extrapolation to the 38
This sampling strategy is a combination of three distinct recruiting methods: Venue-based refers to the drop-in centers that were managed by Blue Diamond Society; the snowballing is a nonprobability approach to sample design where respondents help recruite new respondents; and lastly a convenient sample which in this case refers to the intial group of respondents that were interviewed because they could be easily accessed through Blue Diamond Societies existing networks. 39 Ibid. 40 CIA. The World Factbook 41 UNDP (2014)
13
Surveys, Big Data, and Experiments
313
population at large is difficult) that would be impossible to cover in a census, given space limitations. • Sampling remains a critical challenge. Many LGBTI people in more rural parts of the country were unable to participate in the survey like women who were only partially reached. It is also unlikely that the survey reached many sexual and gender minorities that hide their sexual orientation, gender identity and expression, and intersex characteristics since they would avoid the networks that were used to distribute the survey. Similar issues have been faced with other hard to survey populations; in section “The Potential of Big Data in Building LGBTI Knowledge,” we provide an overview of some of the lessons learned from those populations. While this survey is a milestone in large-scale LGBTI data collection, it illustrates a number of challenges that should be taken into consideration when developing future surveys.
India In India, the World Bank developed a survey with the goal to explore the interconnection between sexual orientation, gender identity, and development outcomes, specifically the relationship between discrimination and socioeconomic status.42 The implementing partner, Amaltas, used a similar community-based approach as employed in Nepal to recruit respondents. Due to a lack of organizations with a rural membership and the overall difficulties of data collection in remote areas, the survey was carried out in urban areas only. Within the L, G, B, and T communities,43 the survey administrators found it particularly difficult to reach bisexual and lesbian respondents. In India, most bisexuals tend to hide their sexual orientation, especially since many of them are married with families. They live out their bisexuality mainly through same-sex behavior, rather than identification. Lesbians, on the other hand, were hard to reach because of a paucity of lesbian-focused community-based organizations that could help carry out the survey. The survey reached 943 respondents in eight Indian cities. The majority of the respondents were “men who have sex with men” (MSM). In India, the term not only is often used as an umbrella term to describe gay identities, but also includes people who engage in same-sex behavior yet would not identify as gay men.44 The survey also reached a high number of hijras45 and a much smaller number of “female-born,” one local term for lesbians.
42
Singh et al. Intersex people were not part of the data collection. 44 Singh et al. 45 Hijra is one South Asian term for male to female transgender identities; it does not cover all trans identities. 43
314
D. Koehler and N. Menzies
The results show high levels of discrimination in the daily lives of the respondents. They are excluded from key services such as banking, housing, and health care. Around one-third of the respondents did not have a bank account and thus could not access any financial services. Similarly, many reported discrimination in the housing market, 13% said they were unfairly denied accommodation, and 15% claimed they were evicted from their homes because of their sexual orientation, gender identity, and/or expression. When attempting to access health care facilities, 15% of the respondents were denied access and another 15% were removed from a facility. One-fourth had been mistreated in government medical facilities.46 Many of the sampling issues that were faced in Nepal were also faced in India, but one additional issue becomes evident in this case: • Given the constraints of the sampling method, only limited comparison to the general population is possible. Meaning that it is nearly impossible to conclude whether LGBTI people are worse off than non-LGBTIO people. For example, the study found that one-third of the respondents did not have access to a bank account, but in order to understand whether this is a problem specific to the LGBTI community a comparable data set for the general population would be needed. One way to respond to this is to include questions on SOGIE status in larger population surveys and thus reach both LGBTI and non-LGBTI people at the same time. But given the likely small number of LGBT people within an overall population, a large sample (of LGBTI and non-LGBTI) people would be needed to pick up enough LGBTI people to be able to undertake a meaningful analysis of the results – especially of the component L,G,B,T,I parts. Further, problems around having respondents reveal themselves and truthfully respond will remain and perhaps be heightened if the survey is not run by trusted local LGBTI partners. Respondents need to be sure that their privacy and security is ensured.47 Hence, it could be more effective to include non-LGBTI people in research projects specific to the LGBTI community, using the same sampling strategy and similar questionnaires to generate comparable data sets. This approach was recently tested by the World Bank in a survey in Thailand (results forthcoming in Fall 2017).
Europe In 2012, the European Union Agency for Fundamental Rights (FRA) collected data on sexual orientation, gender identity, and expression in 28 European countries.48
46
Singh et al. Ibid. 48 At that time, 27 EU member countries as well as Croatia, which officially joined in 2013, participated. 47
13
Surveys, Big Data, and Experiments
315
The online survey reached 93,079 respondents in total. The overarching goal was to explore how LGBT49 people experience fundamental rights.50 FRA designed the survey instrument in 27 languages in close cooperation with a scientific advisory board, made up of representatives from civil society as well as local and international experts. A consortium of Gallup Europe51 and International Lesbian Gay Bisexual Trans and Intersex Association (ILGA) Europe52 implemented the survey. Over a period of 4 months, any person over 18 years53 that identified as LGBT could participate in the anonymous online survey. The online questionnaire and sampling through self-identification provided a number of advantages such as easy distribution across 28 countries, a collection of a large data set, and the development of a tool which could be readily applied in other contexts.54 The questionnaire covered the following areas: • • • • • • •
Public perceptions and responses to homophobia and/or transphobia Discrimination Rights awareness Safe environment Violence and harassment Social context of being an LGBT person Personal characteristics, including age and income group
The survey discovered that despite antidiscrimination laws across the EU, one-third of the respondents felt discriminated against in at least one of the following areas: housing, health care, education, social services, and access to goods and services. Transgender respondents were even more vulnerable, with 35% reporting being attacked or threatened with violence in the previous 5 years. Together with bisexual women, they are also most likely to report income levels in the lowest quartile.55 As with the other surveys, extrapolating from the results to the total LGBT population or non-LGBT people is difficult given the sampling method. Respondents were a segment of the population that had access to the Internet, and frequent LGBT-friendly websites, events, or other services where the survey was advertised.
49
Intersex people were not specifically targeted by this survey. FRA (2013) 51 Gallup is an international survey and consultancy company. 52 ILGA-Europe is the European umbrella organization of civil society organizations working on LGBTIQ issues. 53 FRAs’ focus on respondents over 18 years is very common among these types of surveys. Unfortunately, this makes invisible the experiences of younger LGBTI+ people. 54 FRA (2013) 55 Ibid. 50
316
D. Koehler and N. Menzies
Overall, the vast majority of respondents were younger than 40, mostly gay and well educated. An unusually high number of students answered the survey.56 FRA opted to use an online survey for two reasons: First, the necessary sample size for a representative sample in each country would have needed to be extremely large in order to achieve a robust analysis. They estimated that around 800,000 screening interviews would be needed in order to find 1000 LGBT respondents in each member state. Second, FRA noted that LGBT people are likely to disguise their sexual orientation or gender identity in a screening interview with an unknown enumerator. Hence the self-identifying online survey was chosen, even though the sample would not allow a comparison with the larger population.57 Anonymity was an important aspect of the study design; respondents were not asked for their names or any other personal information. No information about the computer itself was collected (for example, through cookies). While the anonymity provided the respondents with security, it had two major implications. Respondents had to take the whole survey in one go because there was no way of saving progress and continuing at a later stage. Furthermore, it is possible that respondents took the survey more than once since no data about them or their computer was stored to prevent multiple entries. These constraints are inherent of online surveys and should not diminish the overall potential these surveys have especially in areas where the Internet is widely accessible. The involvement of local experts helped in reaching out to large numbers of LGBT people and facilitated the distribution of the survey. The implementing consortium of Gallup Europe and ILGA Europe prepared background notes on the local LGBT communities to target advertisement accordingly.58
Australia In 2015, Australia conducted a survey of 272 intersex people, one of the largest surveys of its kind.59 The survey was structured to be on, with, and for the intersex community, as one means of redressing the historical role of intersex people as subjects of (often medical) inquiry. An online data collection tool was used with both forced-choice (quantitative) and open-ended (qualitative) questions. The survey was circulated through online networks, social media, and general advertising. The survey asked participants what term participants preferred to use to describe themselves to themselves, to family and friends, and when seeking medical assistance. The survey revealed high levels of mental health issues among respondents (60% had thought about suicide and 19% tried it; compared to less than 3% – combined – for the Australian population at large). The intersex respondents also reported having a much higher level of none-completion of secondary school (18%) compared to the
56
Ibid. Ibid. 58 Ibid. 59 Jones et al. (2016) 57
13
Surveys, Big Data, and Experiments
317
general Australian population (2%). Further, a 41% of respondents earned less than $20,000 per year (minimum wage is $34,000).60
Lessons Learned from Other “Hard to Survey” Populations Sexual orientation, gender identity, and/or expression are complex, deeply internal processes.61 Surveys assume that the respondents “know” the information that is being asked about and are willing to reveal it. For sexual orientation, gender identity, and expression, this will often not be the case. That said, there are other identities that are highly personal (e.g., faith), and in many contexts, these are also subject to stigma and discrimination and are therefore sensitive questions that could be potentially dangerous to disclose. Many of the above-described methodological problems in reaching LGBTI communities and comparing their experiences to those of the non-LGBTI population are familiar to researchers looking at other “hard to survey” populations – such as victims of domestic violence, drug users, or sex workers. Past studies and surveys of sexual and gender minorities have built on experiences with other hard to reach populations. What is often needed is a set of culturally sensitive questions that allow respondents to articulate their own self-identification. This, of course, makes meaningful quantitative analysis more complicated. The most challenging part of data collection for any of these groups is defining the target population and sampling them. The Nepal and India surveys used a venue-based snowball convenience sampling method, in which key actors of the community distributed the surveys and each respondent was encouraged to do the same. This method has been used for communities where other standard probability sampling methods are problematic. Respondent-driven sampling,62 another form of chainreferral sampling, takes into account the nature of the referrals and has been used in some cases to correct for bias in the overall network. Online surveys, such as the FRA, are another approach with its own challenges. Surveys of vulnerable groups require a high amount of privacy and trust in order to overcome barriers to nonresponse, inaccuracy, and access. Many LGBTI people face violence as result of their sexual orientation, nonconforming gender identity and expression, and/or sex characteristics similar to many victims of gender-basedviolence. Analogous to surveys on sexual and gender-based violence, protocols and training programs need to be developed to ensure that respondents feel safe in answering surveys, are not put at the greater physical, emotional, or psychological risk by doing so, and can be appropriately referred to counseling and welfare 60
Ebd. In this sense, Intersex might be viewed differently since it is a relatively clear defined medical term, and while Intersex people might face similar discrimination and exclusion they also have unique subset of challenges that LGBT people do not face. 62 Similar methodology to snowballing. A mathematical model of who recruitment process is than used to weight the sample. More information 61
318
D. Koehler and N. Menzies
services as needed. This will also lead to better data as people are more likely to disclose their true experiences when they feel safe. Research also frequently faces issues around motivation – not only for LGBTI respondents but also for a useful comparison group. If respondents do not see the benefits of taking the time to respond, researchers need to apply different techniques to engage. Community involvement in all stages of the research can, therefore, be an important tool to reach the desired number of respondents.
The Potential of Big Data in Building LGBTI Knowledge Surveys are only one tool to collect data on LGBTI people; in recent years, so-called “big data” has become an important source for data. The information that every user is providing by surfing the Internet and using digital services can be a powerful tool to better understand any given population. Social media platforms like Facebook, for example, give users the opportunity to self-identify as LGBTI. According to the companies’ own research from 2015, more than 6 million Americans have come out on Facebook. While this estimation is based on self-identification, Facebook, like many other companies, collects much more data about their users. Peoples’ likes, the places they frequently visit, and the post they write could all be used to develop a much more accurate estimation of how many LGBTI users the social network has. According to an article in the New York Times, from December 2013, 5% of all Google porn searches in the US is for gay porn. Looking at these (gay) porn searches around the world, Google data shows they range from 6% of all porn searches in Central America, South America, Europe, Australia, Pacific Islands, and the Caribbean to 2.5% in Africa, Eastern Europe, and the Middle East. This data is useful, for instance, in showing the simple but often contested fact that LGBT people exist around the world. Yet it does not allow much more substantial conclusions to be drawn as, for example, women search for gay porn as well. Definitional issues remain a problem for big data like for any other source of data. Is Facebook’s self-identification enough, or should other user data be used to identify those who do not publicly state they are LGBTI? Is someone LGBTI just because they search for gay or lesbian porn? While the underlying issues are similar, “big data” does face its own unique challenges when used to better understand the LGBTI population. Nevertheless, it provides important additional sources of data, especially as Internet access becomes more widely available.63
63
One innovative new technique is random domain intercept technology whereby Internet users who incorrectly type in a domain name, are directed to a survey. In 2015, this technique was used by ILGA Global. for the first global online survey of attitudes toward LGBTI people, which attracted over 96,000 responses and will be repeated each year.
13
Surveys, Big Data, and Experiments
319
Experimental Approaches to Measure Discrimination and Exclusion So-called “audit studies” provide an alternative approach to gathering evidence on the development outcomes, discrimination, and exclusion of LGBTI people.64 Experimental methods can overcome some of the challenges noted above yet face their own challenges in terms of generalizability and the general downsides of experiments. Such methods have been used to highlight disparities on the basis of race, ethnicity, and gender, as well as for sexual orientation, gender identity, and expression. Experimental methods have been used to detect discrimination in employment opportunities and access to housing for LGBTI people.65 These experiments have, for example, focused on the job application process and revealed significant bias against openly lesbian and gay applicants, resulting in between a 5% and 40% lower callback rate for interviews, as well as lower wage offers. One 2011 study focused on gay male job seekers and their chances of being invited to an interview. Two fictitious counterfeit resumes were constructed to show applicants with the same qualifications and sent to over 1700 job postings in the USA. One of those resumes was randomly assigned leadership experience in a gay university campus organization, with the “control” resume-listed leadership experience in a nongay organization.66 The study found that “While heterosexual applicants had an 11.5% chance of being invited for an interview, equally qualified gay applicants only had a 7.2% chance of receiving a positive response” (Tilcsik 2011). A similar study was conducted in the United Kingdom. Over 140 “applicants” were in correspondence with 5549 companies. This study focused on the job prospects of gay men and lesbians. Gay men and lesbians had a 5% lower probability of being invited to an interview. Further, the companies that they received call backs from were paying on average less than those that invited heterosexual applicants.67 In Greece, a study unveiled not only a 26% lower chance for a call back for gay and lesbian applicants but also 1.5% lower initial wage offer.68
Conclusion Without greater certainty about the development status of LGBTI people and a rigorous comparison of their welfare to the population at large, it is more difficult to make the case for effective policies and programs – and to know whether any interventions are effective. 64
Pager and Shepherd (2008) Studies for other minority groups have applied similar models for housing, credit and school applications. 66 Tilcsik (2011) 67 Drydakis (2015) 68 Drydakis (2009, 2011) 65
320
D. Koehler and N. Menzies
Rigorous, quantitative data on LGBTI people are scarce, even in developed countries, and even on the most basic issues – such as the size of the population. Surveying LGBTI people has challenges, not least defining the population and the unwillingness of many LGBTI people to reveal themselves to interviewers due to stigma. Lessons can be learned from experience with other “hard-to-survey” populations, where issues of definition, stigma, and access also apply. Through consultation, cultural sensitivity, and training, it is possible to collect goodquality data. Probability sampling presents a particularly challenging issue. As the LGBTI population is relatively small, a general population survey would have to be very large (and costly) to pick up large enough samples of the L, G, B, T, and I subpopulations (even if the issues with getting people to reveal their identities can be overcome). Experimental studies and big data can help triangulate information on some issues and address some of the constraints of population surveys. We encourage survey experts, data analysts and LGBTI development professionals, and the LGBTI community to work together to collect better data and keep focus particularly on issues of definition, sensitive interview protocols, and collecting data that can be readily compared to non-LGBTI populations. Proposed Citation Koehler, Dominik; Menzies, Nicholas (2017). Surveys, Big Data, and Experiments: How Can We Best Learn About LGBTI Development Outcomes? World Bank Acknowledgments The authors would like to thank the following for thoughtful guidance and comments: Georgia Harley, Kristen Himelein, Chloe Schwenke, M. V. Lee Badgett, Clifton Cortez, and the members of the World Bank’s SOGI Task Force. All errors are the authors alone, and the views expressed herein do not represent the position of the World Bank or its executive directors.
References Dalia Research (2016) Counting the LGBT population: 6% of Europeans identify as LGBT. https:// daliaresearch.com/counting-the-lgbt-population-6-of-europeans-identify-as-lgbt/ Drydakis N (2009) Sexual orientation discrimination in the labour market. Labour Econ 16:364 Drydakis N (2011) Women’s sexual orientation and labor market outcomes in Greece. Fem Econ 17:89 Drydakis N (2015) Sexual orientation discrimination in the United Kingdom’s labour market: A field experiment. Hum Relat 68(11):1769–1796. https://doi.org/10.1177/0018726715569855 Fausto-Sterling A (2000) Sexing the body. https://www.amazon.com/Sexing-Body-PoliticsConstruction-Sexuality/dp/0465077145 FRA (2013) EU LGBT survey technical report. http://fra.europa.eu/en/publication/2013/eu-lgbtsurvey-technical-report Gates GJ (2011) How many people are lesbian, gay, bisexual, and transgender? http:// williamsinstitute.law.ucla.edu/wp-content/uploads/Gates-How-Many-People-LGBT-Apr-2011. pdf Gates GJ (2017) Vermont leads states in LGBT identification. https://news.gallup.com/poll/203513/ vermont-leads-states-lgbt-identification.aspx
13
Surveys, Big Data, and Experiments
321
Gulløy E, Normann TM (2010) Sexual identity and living conditions. https://www.ssb.no/a/english/ publikasjoner/pdf/rapp_201038_en/rapp_201038_en.pdf Jones T, Hart B, Carpenter M, Ansara G, Leonard W, Lucke J (2016) Intersex: stories and statistics from Australia. Open Book, London, UK Lofquist D, Lugaila T, O’Connell M, Feliz S (2010) Households and families: 2010. https://www2. census.gov/library/publications/cen2010/briefs/c2010br-14.pdf National Center for Transgender Equality (2014) Transgender terminology. https://www.nawj.org/ uploads/files/annual_conference/session_materials/transgender/transgender_terminology-ncte. pdf Pager D, Shepherd H (2008) The sociology of discrimination: racial discrimination in employment, housing, credit, and consumer markets. Annu Rev Sociol 34:181 Park A (2015) An inclusive approach to surveys of sexual and gender minorities. Report of meeting. http://williamsinstitute.law.ucla.edu/wp-content/uploads/Inclusive-Approach-to-Surveys-ofSexual-and-Gender-Minorities-Nepal-March-2015.pdf Park A (2016) Reachable: data collection methods for sexual orientation and gender identity. Williams Institute, Los Angeles Richters J, Altman D, Badcock PB, Smith AMA, de Visser RO, Grulich AE, Rissel C, Simpson J (2014) Sexual identity, sexual attraction, and sexual experience: the Second Australian Study of Health and Relationships. 11:451 Romero AP (2017) 1.1 million LGBT adults are married to someone of the same sex at the two-year anniversary of Obergefell v. Hodges. Hodges, Los Angeles Singh S, Krishan A, Mishra C, Ghai K. Experienced discrimination and its relationship with life chances and socioeconomic status of sexual minorities in India Tilcsik A (2011) Pride and prejudice: employment discrimination against openly gay men in the United States. Am J Soc 117:586–626 Tourangeau R, Edwards B, Johnson TP, Wolter KM, Bates N (2014) Hard-to-survey populations. Cambridge University Press, Cambridge UNDP, Williams Institute (2014) Surveying Nepal’s sexual and gender minorities: an inclusive approach. UNDP, Bangkok Valfort M-A (2017) LGBTI in OECD countries: a review. OECD social, employment and migration working papers, no. 198. OECD Publishing, Paris
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods
14
Vijayendra Rao
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Economics, Ethnography, and Reflexivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What Can Economics Learn from Ethnography? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cognitive Empathy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Narrative Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Machine Learning and Natural Language Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Studying Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Participation: Respondent as Analyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
324 325 329 330 335 337 340 342 344 346
Abstract
This chapter argues that Economics can learn from Cultural Anthropology and Qualitative Sociology by drawing on a judicious mix of qualitative and quantitative methods to become more “reflexive.” It argues that reflexivity, which helps reduce the distance between researchers and the subjects of their research, has four key elements: cognitive empathy, the analysis of narratives (potentially enhanced by machine learning), understanding process, and participation (involving respondents in research). The chapter provides an impressionistic and noncomprehensive review of mixed-methods relevant to development economics and discrimination to illustrate these points. Keywords
Mixed-Methods · Ethnography and Economics · Narrative Economics · Qualitative Methods · Understanding Process · Participation · Empathy V. Rao (*) Development Research Group, The World Bank, Washington, DC, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_19
323
324
V. Rao
Introduction There are artificial distinctions of method that limit us as social scientists. We economists think of ourselves as quantitative social scientists. Cultural anthropologists and some sociologists see themselves as ethnographers who rely on participant observation and qualitative methods. The reasons for this divide are a vestige of disciplinary history: how scholars are trained and assessed for tenure. By limiting ourselves to one set of methods, we limit the questions we can ask and constrain ourselves in the data available to answer them. The methods lead the questions, rather than the other way around. It did not begin this way. Charles Booth, whose magnificent multi-volume Life and Labour of the People in London (1892–1897), helped set the stage for all social science research, triangulated between data from household surveys, participant observation, and open-ended interviews to create detailed “poverty maps” of London. These maps drew on data, both quantitative and qualitative, that Booth and his team collected from every one of London’s four million residents to visualize where the poor and the rich lived. The maps drew a finely detailed picture of households and their living standards, the nature of their work, and many other characteristics that Booth grouped into three categories: Employment, “Circumstances” (illness, infirmity, family size, etc.), and “Questions of Habit” (what he called problems of “drunkenness and thriftlessness”). The task took 17 years to complete (1886–1903), filled 17 volumes, and had a tremendous influence on economic and social policy and on social science (for more on Charles Booth’s work, see Mary Morgan’s (2019) excellent book. The full set of maps and Booth’s “notebooks” can be explored here: https://booth.lse.ac.uk/map). Over the course of the twentieth century, most social scientists moved away from this grounded, multi-method process of inquiry. Economics, in its quest toward greater scientific precision, became almost entirely focused on the analysis of quantitative data. Other less homogenous disciplines like sociology had sub-disciplines that focused on either qualitative or quantitative methods, while anthropology split into (at least) two fields – cultural anthropology emphasizing participant observation and ethnography and physical anthropology emphasizing the analysis of various types of quantitative data from archeological digs, genetic analysis, and experiments. Psychology moved from observational studies into an entirely experimental direction. These differences in method were a primary characteristic of how the social sciences distinguished themselves from one another. However, even a cursory look at recent social science journals shows that things are changing. Disciplinary differences are increasingly getting bridged. Economics has taken a behavioral turn (Thaler 2016); sociology and political science are getting increasingly interested in mixing qualitative and quantitative methods (Small 2011; Humphreys and Jacobs 2015); psychology is moving toward a mix of experimental, qualitative, and survey methods (Paluck 2010); and mixed methods are being actively used by development researchers and practitioners (e.g., Kanbur and Shaffer 2007). We social scientists have expanded the questions we ask and seem to be
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 325
moving toward a world where methods are subservient to questions. Consequently, we are increasingly mixing methods. This chapter provides a selective overview of mixing quantitative and qualitative methods – which is still a rarity in economics. I argue that qualitative methods offer a powerful set of tools that are of central concern to economists and particularly for economists interested in questions of discrimination and development. They help reduce the distance between researchers and the subjects of their research and can help introduce an element of reflexivity into economics. The chapter does not attempt a full-fledged review of mixed methods, whether in economics or in other disciplines, but rather tries to illustrate various ways in how a judicious combination of qualitative and quantitative methods and tools can result in a more empathetic and richer understanding of human behavior. The examples which I will employ are also impressionistic rather than comprehensive. I will hew closely to topics that are of interest to economists who study discrimination (the topic of this handbook) and development (which is my primary focus of interest) and draw on work from across the world, including my own.
Economics, Ethnography, and Reflexivity At the risk of some oversimplification, we could argue that econometrics focuses on two broad issues. The first is causal inference, strengthened over the last three decades by the extraordinary advances made under the credibility revolution. The second, like most of applied statistics, is about understanding and overcoming the challenges of making robust inferences about larger populations from sample data. But, despite the powerful tools that have emerged over the years under these two important umbrellas, something is left missing. The data employed by an econometrician is often collected by someone else. Even when the econometrician is responsible for collecting it, they usually delegate it to a survey firm focused on fine-tuning the questions, making sure the power calculations are done properly, and deploy research assistants to supervise and monitor the data collection and data entry process. All these steps are necessary to collect data at scale, but the very process of collecting information from many people results in a distance between the researcher and the subjects of their research. This accentuates the preexisting social and cultural distance that exists between the researcher and the researched, which is particularly acute for those who study discrimination or development where the researcher is almost always from a different social class and usually from a different race, caste, and cultural background from those they research. The topics picked by a researcher, consequently, are more likely to reflect their lived reality – their reading of the literature, their notion of what kinds of topics will further their careers – rather than the perspectives of the people they are studying. Ethnography is at the other of this spectrum. The focus is usually on one community, studied intensely by one person who actively participates in and observes the life of the community over an extended period. To illustrate this, let me begin with brief summaries of two ethnographies that employ participant
326
V. Rao
observation – the primary method of cultural anthropologists and of qualitative sociologists trained in the Chicago school of urban sociology. The anthropologist Bhrigupati Singh, in his wonderful ethnography, Poverty and the Quest for Life (2015), draws on 1.5 years of fieldwork with the Sahariya, a community classified as a “primitive tribe” who live in a remote and extremely arid part of the Indian state of Rajasthan. The Sahariya are former bonded laborers who are considered much poorer than the Dalit (“Scheduled Caste”) Chamar community with whom they sometimes share settlements. Singh delves deep into the spiritual lives of the Sahariya trying to understand and describe how they conceive of “wellbeing” and show how different it is from the view of philosophers and social scientists. His descriptions do not ignore the standard analytical categories of caste, class, power, the state, and structural inequality, but Singh attempts to draw the reader into questioning those categories as he tries to understand how the Sahariya see themselves. He questions the tropes describing them in previous work as bound by “superstition” and “subject to supernatural beliefs” which leads one government report to describe their lives as so desolate as to “make their living unthinkable.” Singh’s effort to try to understand this better, to shed light on – as he says – “What makes life unthinkable?” (p. 1). He shows us how it relates to how the Sahariya experience the government in everyday life, the prejudices that they encounter, and their response to this which draws on their rich spiritual life which is very different from dominant religious practices in the region. This leads them to be seen as different and divergent from the development paradigm, which reinforces the discrimination they face both by society and the state. Singh’s description of his research process is worth quoting because it is a nice summary of the ethnographic method (pp. 284–285): “Ethnographic method may be seen as a two-phase process. . .In the first phase, one is a hunter-gatherer, pursuing targets, collecting impressions. . .The next stage of labor is of a more settled cultivator, as we move from impressions to expressions. You plow through what you have gathered, jostling with others who faced similar phenomena and arrived at different thoughts. When impressions are organized and attached to concepts, they turn into thoughts and expression. Concepts are seeds from which impressions grow into thoughts. Our concepts may also be limiting and force us to convey much less than what we saw and felt.” Thus, ethnography is a method that obsesses with process rather than outcomes, that seeks to describe rather than measure, and that attempts to reduce the distance between researcher and researched via a process of deep empathetic understanding, rather than to maintain (what anthropologists see as) a fictive analytic distance. This process of participating and observing has long been the fundamental tenet of ethnography, which I will next illustrate with an older and influential work by the sociologist Elijah Anderson. In his ethnography Streetwise, Anderson (1990) distills 14 years of fieldwork1 (from 1975 to 1989), to study interactions between whites and
1
Combining several 100 interviews with observations from the streets, bars, laundromats, brunches, and parties among other places.
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 327
blacks in the Philadelphia neighborhood he lived in2 which he called the “Village” and the neighboring, largely black, area that he called “Northton.” Drawing on Erving Goffman’s work on Stigma (Goffman 1960) and strategic interaction (Goffman 1969), he drew a picture of how the Reagan-era cuts to public spending, and the decline of the city’s industrial base, led to shifts in how the two neighborhoods interacted with one another. The Village is racially mixed, mainly populated by middle-class whites with a few middle-class blacks – largely academics and researchers – and a few yuppies. Northton was almost entirely black consisting of “stable working-class black families, well-attended churches. . .and young people eager to ‘make something of themselves’.” (pp. 238–239). But to people in the Village (both black and white), Northton had the reputation of being beset by “classic urban ills: drugs, crime, illiteracy, poverty. . .families on welfare. . .” Anderson says that while there was an outward culture of “civility” and tolerance in the Village, they were wary “strangers with whom they must share the public space.” Whites were wary of blacks and gave them “extra scrutiny,” and everyone saw the streets as a “jungle,” especially at night, when “all cats are grey” and everyone seems threatening. All of this challenged the Village’s vision of itself as a civil and tolerant community. Villagers had little “worthwhile knowledge” about black areas and tended to form stereotypes. Blacks in this “picture become the kinds of people they rather not have close to them” and people whom they “try desperately to avoid.” This lack of familiarity with black culture resulted in “prescriptions and proscriptions” about public behavior that were structured to minimize conflict with blacks. And blacks, who happened to find themselves in the Village, often tried to “prove themselves respectable to others they encounter.” Anderson’s book provides a glimpse of a neighborhood just before the gentrifying processes of the 1990s began which priced out black communities from their traditional neighborhoods and shows how racial difference actually plays out among people who generally think of themselves as anti-racist and tolerant. Moreover, Anderson describes how poor black men must negotiate stereotypes as they navigate their way toward making a living in a world where jobs are scarce. It is entirely complementary to the large body of work on the economics of discrimination, both empirical and theoretical, and yet more alive, more grounded, more reflective of the lived reality of both blacks and whites. It describes how statistical discrimination might work but demonstrates that it has strategic and interactional implications which, at least at the time of the book’s writing, were not part of the economics literature. And it is a wrenching narrative of the disproportionate impact that Reaganera cuts and industrial decline had on black communities socially and psychologically and on their economic lives. It is impossible for me to do justice to these rich ethnographies in a couple of paragraphs – they are thick with narrative insight, description, ideas, and theory.
2
I lived in the same neighborhood from 1985 to 1990 which is perhaps one reason why this work resonates so deeply with me.
328
V. Rao
What they have, which economics lacks, is a method that necessitates an intimate relationship with the field. Participant observation attempts to minimize the distance between the researcher and the subjects of their research, and to be acutely conscious of how the researchers’ social class might position how they may see and interpret things, and to bring that reflexivity into their analysis. Central to this is the notion of what the sociologist Mario Small (2018) has called “cognitive empathy,” “the ability to understand a person’s predicament as they understand it.” What participant observation lacks, however, is representative data – it focuses on a case, a village, a neighborhood, a community, with which it deeply engages, but it does not have a robust way of demonstrating that the findings are representative of a larger population. For me, re-reading Anderson’s book more than 30 years after I first read it, his findings are now part of received wisdom; they have been replicated many times and were a central theme even of iconic television shows like The Wire, which is a measure of the book’s validity. But at the time when Anderson was writing the book, he was simply trying to do his best to describe what he observed and experienced in one community, at one period in its history, filtered through his sociological training and his reading of theory, rather than focusing on making a generalizable statement about the urban United States. There is, therefore, a trade-off between the finely filigreed detail of a good ethnography and the focus on a very small sample that it requires and the strong emphasis in survey-oriented fields like empirical economics on making statistically generalizable statements by analyzing data from large representative samples. The sociologist Michael Burawoy (1998), in a brilliant paper on “The Extended Case Method,” argues that there are two models of social science, “positivist” and “reflexive.” He says that in the positivist approach, within which economics would be placed, we limit (p. 5) “our involvement in the world we study, insulating ourselves from our subjects, observing them from the outside, interrogating them through intermediaries. We keep our feet on the ground by adhering to a set of data collection procedures that assure our distance. . .We try to avoid affecting the situation we study, standardize the collection of data, bracket external conditions, make sure our sample is representative.” The reflexive approach characterized by ethnography and participant observation, on the other hand, is a “model. . .that embraces not detachment but engagement as the road to knowledge.” Burawoy (p. 16) calls dialogue “the unifying principle of reflexive science.” It “starts out from a dialogue between us and them, between social scientists and the people we study” (p. 7) and it (p. 5) “embeds such dialogue within a second dialogue between local processes and extralocal forces.” To Burawoy, “reflexive science” is premised on the idea that the interview is not a neutral entity, but an “intervention” where “by mutual reaction. . .we discover the properties of the social order.” He argues that it requires the social scientist to understand the “process,” context, and both the “narrative” and “nondiscursive” dimensions of interactions with the field of study and the respondents within it which may be discovered “through participation, ‘doing’ things with, and to, those who are being studied.” Understanding, and theory, thus emerges in a process of dialogue
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 329
between participant and observer, who in turn engage in dialogue with the scientific community, which in turn re-enters the “wider world of participants.” Burawoy’s purpose in this paper is to make the case for “reflexive science,” arguing that it is at least as scientific as the positivist approach. He, however, sees these approaches as located in entirely separate worlds, so different that they could never meet. My purpose, instead, is to see whether a marriage between reflexive and positivist science is possible and whether a judicious integration of qualitative and quantitative approaches can make economics more reflexive.
What Can Economics Learn from Ethnography? For over a century, the field of economic anthropology has employed ethnography and participant observation to study questions, such as those at intersection between economic and social life, well before mainstream economics got around to them (e.g., Gregory and Altman 1989; Gudeman 2001). Economics is perhaps too embedded within the positivist paradigm to allow ethnography to become a primary tool. However, economics can come closer to ethnography and try to better incorporate at least four key interconnected principles from it. The first, to use Small’s term, is cognitive empathy – to reduce the distance between the researcher and those they research, by trying hard to put the researcher in the shoes of the researched. In the positivist frame, a person who is being “researched” is usually called the “subject,” a term which has implicit within it the vast power differential between scholar and subject. Reducing this power differential is part of the task of achieving greater cognitive empathy – though it may be very hard to put the researcher and researched on an entirely level playing field. The second is the idea that information does not come only in numbers; it can also come in words – via interviews, documents, text, images, video, and other narrative forms. The notion of “narrative economics” got an impetus in economics, particularly in macro and finance, after the publication of Robert Schiller’s (2019) book with that title, but narrative data is still rarely analyzed in the field, particularly by micro-economists. The third is the notion of process. What are the minute details, the slow day-today shifts, that make change happen – the close observation of which is so central to a good ethnography? Empirical economics, particularly since the credibility revolution, has been almost entirely focused on the measurement of outcomes. Understanding the mechanisms that lead to those outcomes has generally been relegated to theory. Structural economics, and some work in causal inference, does attempt to get at mechanisms, but this is usually done either via speculation, by relying on theory to create a story why a certain intervention may have resulted in a certain outcome, or in discrete pieces, by analyzing heterogeneity in impacts or by designing experiments with arms that can tease out some element of how different kinds of stimuli might result in different outcomes. What is not done often is to do the most obvious thing – to keep a close an eye on the ground to observe how change happens by embedding
330
V. Rao
oneself in a site over an extended period to create detailed observations of the process of change.3 The fourth is the principle of participation. Most economic research is primarily a process of extraction – we collect data from subjects, we analyze the data and write a paper, and our respondents never hear from us again. The best ethnographies teach us that the simple act of treating the researched as a co-creator of research rather than a subject can sharply reduce the power differential with the researcher (note, however, that this is not true of all ethnographies which can be just as extractive as any other kind of research). In the next sections of the chapter, I will briefly sketch some of the work done under these broad categories to illustrate how economists might try to come closer these principles in our work.
Cognitive Empathy There is a long tradition in development economics of “village studies,” where economists spend months, if not years, studying, engaging with, and collecting data from a small sample of rural villages. T. Scarlett Epstein (1962) spent several months in the late 1950s studying two villages in South India where she examined the links between the local economy and social structure, comparing a “dry” (arid) and a “wet” (irrigated) village. Her data included a mix of field observations, interviews, and survey data. Her approach to research integrated training in economics and anthropology; as a student of William Arthur Lewis, she was deeply interested in patterns of employment, agrarian relations, and migration, but as an anthropologist trained in the Manchester School, she wanted to probe into the link between caste and social class and how these were affected by agrarian change. Her work on South India is, therefore, suffused by an interesting mix of qualitative and quantitative data and was widely influential. Two economic theorists Christopher Bliss and Nicholas Stern, inspired by Epstein, spent 8 months between 1974 and 1975 in the north Indian village of Palanpur to study a series of important questions: the nature of land tenure arrangements, labor and credit markers, risk, and uncertainty. They were steeped in theory, but they let their conversations, experiences, and observations inform the development of new theoretical models and condition the kinds of survey data they collected. Their book (Bliss and Stern 1982) and the papers they wrote based on their first round of fieldwork were followed by a series of repeated visits (primarily by Stern), their students, and research collaborators that continues to this day (e.g., Himanshu et al. 2018). Interestingly, while all the researchers involved in the
3
I remember being at an interdisciplinary seminar many years ago when an economist was presenting an econometric result and drawing on economic theory to understand how rational choice might explain the result. A sociology grad student raised her hand and asked, “Why didn’t you go back to your respondents and ask them?” The economist had no answer.
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 331
Palanpur project are immersed in the conversations they had in the village, rarely are those conversations reported in their published research which is theoretical and econometric. Their work clearly displays a great deal of cognitive empathy, but it is almost entirely quantitative. Several good economists cut their teeth on the Palanpur project. Among them is Jean Dreze, whom I think of as one of the most cognitively empathetic economists in the world. He has developed what he calls the “Research for Action” method (Dreze 2017) arguing that “Research can help with arguments and evidence that contribute to more effective action.” To this end, he has avoided any funding from the usual places (including my own institution – the World Bank) to avoid any semblance of bias. He lives for much of the time within the poor communities whose perspectives he is trying to represent. And, he has successfully advocated for policies that draw on this bottom-up perspective, most notably the Mahatma Gandhi National Rural Employment Guarantee Act which guarantees a minimum of a 100 days of employment of demand for anyone in rural India. Despite all this, I have rarely seen him explicitly draw on qualitative data, and yet every statement embodies the interests of poor and marginalized people in India. In the late 1980s, to systematically explore what economists and anthropologists could learn from each other, Pranab Bardhan organized a pioneering conference that brought them into a “conversation” on measuring economic change in rural India and explore the “tension” between participant observation and survey methods. Bardhan along with Ashok Rudra (1978) had himself previously engaged in fieldbased research that established the relational, socially embedded, interlinkages between land and labor and credit markers in rural India. Re-reading the edited volume that emerged from the conference (Bardhan 1989), the anthropologist Arjun Appadurai’s (1989) chapter which offers a critique of both anthropology and economics makes some points that resonate well today. He says that even if economists visit the field and engage in participant observation, their work tends to be confined to outcomes rather than process, and they tend to collect data that is “distributional” rather than “relational” (in the sense of understanding the relationships between groups). Drawing on his own fieldwork, Appadurai argues that forcing fuzzy concepts into survey-based measures can cause mismeasurement and that survey techniques need to be developed that can accommodate “fuzzy and approximate” quantitative responses. But, he also says that anthropologists tend to not care about bias – they tend to study more prosperous villages and focus on optimistic responses and on respondents who are outliers rather than those that are at the median. Interestingly, this is the only article in the entire volume that makes an explicit case for mixing qualitative and quantitative methods. T. N. Srinivasan (1989) in his chapter makes an important point (which characterizes some of his own work and that of Bliss and Stern) that “there is nothing inherent in the survey method that precludes it from generating the same information as a village study based on participant observation.” It should be noted that Srinivasan and Clive Bell had engaged in a long period of fieldwork in rural India to study credit markets that resulted in seminar theoretical papers on the subject. Bell (2020) recently published an interesting monograph about their fieldwork.
332
V. Rao
Christopher Udry, who was a student of T. N. Srinivasan, is another economist influenced by anthropology and ethnography who describes his method as “iterative field research” (Udry 2003). He illustrates the method with an explanation of how he and Markus Goldstein conducted research in Ghana. To quote Udry, “An initial hypothesis is refined and clarified through detailed observation, which informs the collection of appropriate data. As the economic environment is clarified during the course of fieldwork, the data collection procedure can be adjusted in response. Finally, the research proceeds to formal statistical analysis and, one hopes, to new hypotheses. This iterative process of moving between theoretical reasoning, informal observation and discussion, data collection and statistical analysis is the locus of creativity in this kind of field research and is its distinguishing feature.” Note the striking similarity with the process of participant observation described by Singh. In the process of doing fieldwork in Ghana, Goldstein and Udry observed that there were distinct differences in the productivity of land across plots that were otherwise very similar. In particular, plots owned by women were less productive than those owned by men, and their respondents attributed this to the fact that women were not able to “invest” in their land by keeping it fallow. They had to keep it constantly cultivated to not have it taken away from them. In the paper that followed (Goldstein and Udry 2008) which was a seminal attempt to demonstrate that institutional factors affect economic efficiency, they demonstrate that this qualitative finding (which they never report in the paper) generalized to a result that less powerful people in the local political hierarchy have less secure tenure rights, and women – as among the least powerful in the community – are particularly hard-hit. Thus, some economists, usually working in development, take inspiration from ethnography and spend a great deal of time in the field – developing insights, observing, conversing, revisiting theory, and collecting several rounds of survey data. But they almost never conduct a systematic analysis of their open-ended interviews, group discussions, field notes, and other types of qualitative information. The research that results is almost always within the standard economics paradigm of theory and econometrics. In my view, this results in an immense loss of valuable information, and the work of Bliss, Stern, Bardhan, Dreze, and Udry would be so much richer if qualitative data were integrated with their quantitative analysis. My own initial experience with trying to do mixed methods work in economics may be instructive as an example of why economists do not publish qualitative analysis. I conducted 6 months of fieldwork in three villages in Karnataka state in South India between 1992 and 1994 with the aim of understanding how sociocultural and economic systems interact to affect living standards within families. The first round of fieldwork was conducted in 1992 with a small team of social workers. We conducted focus group discussions, engaged in several open-ended interviews and participatory appraisal exercises, and administered a (relatively standard) structured quantitative questionnaire to every household in the three villages. In the process of one of our initial discussions with a group of women, one participant’s husband dragged her by her hair out of the room where the discussion was being conducted shouting, “Why are you wasting your time with these people – lunch is
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 333
not cooked yet!” Later we heard that she was severely beaten and faced such violence every day. This led to the decision to make domestic violence (which was at the time not a topic studied much by economists) a focus of the analysis. In the process of trying to elicit responses on this sensitive issue, in the first week, our respondents told us, basically, that life was tough but manageable. Their main problems were with the government – lack of good schools, lack of adequate drinking water, etc. After a week of staying in the village and continuing our interviews, one of the women finally opened up and said, “You have become our friends and we can’t lie to you anymore. Let me tell you the truth. We feel like we are in jail. Our husbands beat us all the time and no one helps when that happens. They spend all the family money on alcohol, and when they come home – no one can help us.” We then explored these issues further through in-depth interviews with men and women who outlined the cold rational calculation behind much of the violence. Based on these discussions, a few key questions on wife-beating were added to the quantitative survey instrument. The analysis of the qualitative and the quantitative data demonstrated the links of domestic violence to issues of control and power within the family, female sterilization (which led to an increased risk of being accused of infidelity), alcoholism, and dowry demands. Since the qualitative work demonstrated, among other things, that domestic violence was often a rational choice undertaken to forcibly extract money from the wife’s family, it also led us to rethink how economists modelled both intra-household and transfer behaviors. Without the fieldwork, we would have never thought of studying domestic violence. Being in the field and engaging in many months of participant observation allowed issues to be probed in the field the moment they were observed. This permitted “surprises” to be easily incorporated into the data gathering process. These surprises were often everyday, even mundane, experiences in the lives of the rural poor and only surprising to relatively affluent outsiders. The malleability of “iterative field research” is the fact that reorienting the analysis based on observations in the field is a key element of the method and adds value to traditional econometric practice by discovering and locating important but understudied issues within the research discourse. Initially my hope was to incorporate the qualitative work, the theory that emerged from it, and the econometric analysis into one paper. Where the qualitative analysis would contribute to the standard positivist frame of generating insights, which would inform a theoretical model that would generate predictions that could be tested for their generalizability with econometric analysis (Rao 1997b). However, my co-author Francis Bloch and I found it difficult to publish this version. Referees did not like the qualitative work, and neither did editors, and we were rejected by three journals. We finally decided to split it into two papers, a mixed-methods empirical paper (Rao 1997a) published in a public health journal (which provides the full range of results) and an economics paper which introduced information asymmetry into a bargaining model, and explained why in that social context husbands might be able to treat their wives as hostages from whom they can extract higher post-marital dowries (Bloch and Rao 2002).
334
V. Rao
The tendency in economics to treat qualitative work, from ethnographies, or open-ended interviews, or case studies, as “anecdote” rather than “data” is, I think, deeply limiting. It constraints the use of information and limits the kinds of questions that can be asked. Nothing illustrates this more than Elinor Ostrom’s Nobel Prize winning work (Ostrom 1990) which is full of insights from careful case studies and qualitative work that she conducted with her research collaborators all over the world. It is these case studies that allowed her to gather the evidence to develop theories refuting the received wisdom that collective action was more likely to fail than succeed because individual interests would outweigh the gains from collective participation (e.g., Olson 1965). Ostrom demonstrated in several insightful analyses of case studies of communities that norms for collective action evolve through repeated interaction, and this repeated interaction creates stable institutions for managing a wide range of common property resources including the management of water and forests. It is important to note that Ostrom was trained as a political scientist, taught in a political science department, and (to the best of my knowledge) did not publish her largely qualitative and experimental empirical work in an economics journal till after she won the Nobel Prize. Yet, her cognitively empathetic approach to understanding how communities managed common resources led to deep insights on human behavior that have had a transformative impact on the social sciences. I would venture to argue that she would never been able to do this work had she had a job in an economics department. Younger economists are increasingly spending time in the field to collect data and test interventions, and some display a great deal of cognitive empathy in how they approach their work. MR Sharan (2021) has written a marvelous account of 10 years of travel in rural areas of the Indian state of Bihar, interacting closely with local activists and trying to understand the inner workings of village government. The book, which is meant for a general audience, is less an ethnography than a memoir, but it brings into sharp focus the challenges faced by citizens, particularly those from disadvantaged groups in accessing government programs. This embedded, empathetic approach is apparent, albeit implicitly, in Sharan’s research (Sharan and Kumar 2021) where he and his co-author Chinmay Kumar analyze data from 100,000 village-level politicians in Bihar to examine the provision of public goods to lower castes. They find that leaders matter – when a lower-level lower-caste representative reports to a high-caste representative, lower castes receive poorer access to public goods. However, a grievance redressal system instituted by the government is accessed more by lower-caste representatives to counter this discriminatory provision. It is clear from this overview of field-oriented economics research that Srinivasan was partially correct. Participant observation can be an extremely valuable tool for economists and can result in quantitative findings that are cognitively empathetic. It does not, however, follow that this is true of all field-based research. Cognitive empathy requires a process of listening and learning from the people one is researching, and reflecting their interests, their insights, and their perspectives in the research through a process of iterative learning. Most work in fieldwork in economics tends to be much more top-down, focused on testing an intervention or
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 335
conducting an experiment which is not based on bottom-up learning. Similarly, not all qualitative work is empathetic – a great deal of it is, unfortunately, perfunctory and careless, employing delegated teams who summarize information from focus groups that replicate the prior biases of the researcher. In a lot of so-called mixedmethods work, qualitative information of this kind primarily serves as means of providing color to “hard” quantitative analysis. On the other hand, field-oriented economists (including me) seem to collect qualitative information that they do not report, or indeed fully analyze, because the profession does not treat it as data worth writing about. This is a mistake. Narrative insights from the deep, grounded interviews and field visits can greatly enrich and complement econometric work and help contextualize findings – they are not merely “anecdotes.” I turn next to a brief overview of research that analyzes data from narratives – open-ended interviews, transcripts of group meetings, social media posts, and the analysis of newspapers – to demonstrate this point.
Narrative Data People don’t talk in numbers; they talk in words. A survey interview where the respondent is directed to take a vague notion and emerge with a precise numerical response, or pick one of a small set of multiple choices when none might exactly fit what they are thinking, is engaging in an unnatural interaction that can lead to a misinterpretation of what they are trying to communicate. Yet almost all the data analyzed by economists comes from interviews of this kind. If economics is to become more reflexive and learn from the people it is trying to study, it needs to get more comfortable with analyzing information from narratives. A nice example of the value of narrative analysis is the use of the technique of “financial diaries” by a team of authors that included an economist, an anthropologist, a development activist, and an expert in finance, Portfolios of the Poor (Collins et al. 2009). Collins et al.’s goal was to understand “money management” of the poor: how people earning less than $2 a day in Bangladesh, South Africa, and India managed their finances, juggling consumption, investments, credit payments, and emergencies. Instead of doing a conventional survey asking questions about credit, savings, and assets, they conducted open-ended, narrative interviews of 250 households in these countries. Each household was visited at least twice a month for a full year, and the research team then reconstructed balance sheets and cash-flow statements from these interviews. This unusual technique revealed several patterns that were previously not understood with conventional survey methods, and they estimated that one-shot surveys were missing about half the financial activities of a household. They found that households had to cope with multiple, overlapping sources of uncertainty and risk and that this required active financial management. A key to being able to cope with financial ups and downs was having the ability to be able to access relatively large sums of money to deal with emergencies and life events. However, respondents faced a “triple whammy” – low incomes, irregular
336
V. Rao
cash flow, and poor access to financial instruments that were unsuitable to deal with these conditions. The key finding was that the “portfolios of poor households,” unlike those of the rich, were “managed to ensure that money can be obtained in the desired amounts at the desired times” (p. 61). The book argues this requires an entirely different set of financial instruments than those that are conventionally available in banks and micro-credit organizations. Jonathan Morduch, the economist in the Portfolios team, then teamed with a finance professional, Rachel Schneider, to use the same “financial diaries” method to examine the finances of poor Americans, also finding that they had precarious lives that were not captured in existing data (Morduch and Schneider 2019). Both books draw on the diary technique to construct “hard facts” but couple this with qualitative analysis of the interview transcripts to provide powerful picture of how poor people in very different parts of the world cope with uncertainty and risk. Sociologists routinely draw on narrative data from open-ended interviews to address important questions. An important example is Michele Lamont’s book on The Dignity of Working Men (Lamont 2000). One of her goals is to examine the “inner logic of racism” by understanding the “grammar of evaluation” used by employed, lower-middle class, white and black men in the United States and white and North-African immigrants in France. Her data consisted of 150 2 hour openended interviews – 30 each with blacks in the United States and North Africans in France and 45 with whites in both countries. She conducted all the interviews herself to avoid interviewer bias and to reduce her distance from her respondents. Her key finding is that “workers judge members of the other group to be deficient in respect to the criteria they value the most.” They do this “boundary work” by constructing “mental maps” of similarity and difference. Working-class white American distinguish themselves from those from higher social classes by seeing them as less disciplined and lacking in “integrity and straightforwardness.” And they see blacks as “lazy” and with “wrong values.” African Americans, on the other hand, consider the upper class as lacking the criteria they value the most – “the caring self.” And they see whites as domineering and less compassionate. Lamont argues therefore that “moral criteria can generate strong intergroup boundaries,” though, at the same time, it can help bridge differences with statements like “there are good and bad people in all races.” In France, on the other hand, white workers, using the language of class solidarity, see the poor and blacks as “part of us,” but say that North Africans lack civility and are “culturally incompatible with the French.” Lamont thus argues that workers’ definitions of “who is part of us” have little overlap with official, government categories. The construction of difference is thus “bounded by the differentially structured context in which people live.” Interviews are, of course, not the only form of narrative data. A project that I was involved with recorded, transcribed, and analyzed 300 village meetings in South India (Sanyal and Rao 2019). The context for this was the 73rd amendment to the Indian constitution that brought deliberative democracy to all two million Indian villages. Village meetings are thus supposed to be spaces where citizens could speak freely about public issues, come to agreement, and work with the village council to
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 337
make decisions and monitor village public goods and common property resources. The villages were sampled, following a natural experiment, to be matched by language and ecology across neighboring districts, which belonged to linguistically different states. Prior to 1956, these matched districts were part of the same state. Since the villages matched across the paired districts, spoke the same language, had the same ecological structure and the same caste structure, and had 300 years of shared history, they belonged to the same linguistic community. Everything else equal, they should have, therefore, had the same patterns of public discourse. However, we found that deliberative meetings were very different in different states – which, because of the natural experiment, we could attribute to state government policy in implementing the 73rd amendment. Thus, we were able to show that deliberation – or what we call “oral democracy” to distinguish it from the kinds of deliberation theorized about in western countries – was possible in poor, highly unequal, contexts with low literacy and that the quality of deliberation was less related to levels of inequality or literacy but determined by state government policy. Moreover, our discourse analysis found that social distinctions were largely equalized within the context of these meetings because village elites were concerned about losing votes by suppressing voice. Thus, individuals from lower castes were just as likely to speak as those from upper castes, and topics raised closely matched the interests of the median voter in the village (Ban et al. 2012). What characterizes these three research projects is their relatively small-N, which reflects the difficulty of analyzing narrative data. The transcripts of 300 village meetings in India, for instance, took me and my co-author 10 years (with about 2 years of active work) to analyze: a process that required deep reading, careful coding, reflection, debate, analysis, and writing. This trade-off is limiting, and in the next section, I will discuss the advantages and disadvantages of machine learning tools and whether they could help us analyze narrative data at scale.
Machine Learning and Natural Language Processing The last decade has seen vast advances in using machine learning (ML) to analyze textual data, a field which is known as natural language processing (NLP). ML and NLP pervade our everyday lives in ways that are seen and unseen – for instance, they underlie how search engines work, target advertising from social media companies, filter spam emails, and create virtual assistants to listen and respond to phone calls and an infinity of other applications. The applications of these methods by governments and firms have been shown to discriminate against minorities and women because they draw on training datasets that are largely populated by dominant groups and by men (Caliskan 2021), which is not an indictment of the methods themselves but of how they have been used. Researchers across a variety of fields, including Economics, have begun to use these tools to study a variety of questions (e.g., Gentzkow et al. 2019b). NLP vastly expands our ability to make sense of high-dimensional textual data which opens whole new areas of research in the social sciences. In particular, it has the potential to
338
V. Rao
revolutionize mixed-methods research by making very large amounts of narrative and textual data amenable to statistical analysis. The question is whether this can bring more reflexivity to economic analysis. To answer this question, it might be helpful to contrast the “hand-made” analysis of village meetings discussed in the previous section with a study that I was involved in (Parthasarathy et al. 2019) that used NLP methods to analyze 100 transcripts of village meetings recorded and transcribed in the Indian state of Tamil Nadu in 2014 (and, consequently, took 6 months rather than 10 years to analyze). The NLP paper studies the degree of gender bias in the village meetings and whether having a woman village president can correct gender bias. The transcripts are analyzed using “topic models,” a method which groups text into assigned numbers of “bags-ofwords.” This allowed us to examine which “topics” were more likely to be raised by women and which by men and which by regular citizens and which by officials. We found that men tend to speak much more than women but that citizens are more likely to speak than officials showing that gram sabhas were active spaces for democratic discourse. We also found that men dominated the discussion with the topics raised by them more likely to be taken up by the next speaker and responded to by officials. But having a (randomly assigned) woman president of the village corrected this gender discrimination. Gentzkow et al. (2019a) also study transcripts of speeches, but in the US Congress, and develop a new estimator to measure the degree of partisanship in these speeches from 1873 to 2016. They find that partisanship sharply increased over the period, particularly after the Republican takeover of Congress in the 1990s and the election of Newt Gingrich as speaker. These patterns are very different from previous work on partisanship in the Congress that used roll-call votes and did not find such a sharp change. Stephens-Davidowitz (2014) also studies US politics looking at trends in racial animus in Google searches from 2004 to 2007 and pinpoints the locations in the United States where this is more likely to be observed. He finds that areas that displayed racial animus during this period were much less likely to vote for Barack Obama in the 2008 election relative to votes they cast for John Kerry in the 2004 presidential election. He estimates that racial bias might have cost Obama 4.2% of the popular vote. Lieberman and Miller (2021) study 36,000 online newspaper articles, and 306,000 comments made on them, in Nigeria and South Africa – both multi-ethnic societies with a history of civil conflict. The question they are interested in is whether news media contribute to nation-building in multi-ethnic societies. Using word counts and topic modelling, they find that when an ethnic group is explicitly named in the newspaper article, it generates more negative references in readers’ comments – referring either to the same ethnic group mentioned in the article or to another ethnic group. In other words, news media can negatively influence inclusive nationalist frames by triggering ethnicity-specific reactions among their readers. The question is if these NLP papers can be classified as “reflexive.” They substantially expand the kinds of data that can be analyzed to study discrimination and polarization and contribute toward our understanding of bias in settings that have not been previously analyzed, but the data is analyzed in a manner similar to
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 339
quantitative survey data without any explicit attempt at cognitive empathy, process understanding, or participation. Comparing my own work in Parthasarathy et al. (2019) and Sanyal and Rao (2019) in analyzing the same topic – Indian village meetings – it is clear that the close reading that Paromita Sanyal and I did of the transcripts, followed by hand-coding, discussion, and iterative analysis (i.e., traditional qualitative work), brought us into closer proximity with the people who participated in these village meetings, than when my co-authors and I let machines do the coding for us. This is, more generally, true of the work done using NLP to study social media. Matamoros-Fernández and Farkas (2021) conduct a systematic review of 104 studies, across disciplines, that study hate speech and racial and gender bias in social media. They find “a lack of geographical and platform diversity” and “an absence of researchers’ reflexive dialogue with their object of study.” This is not by any means an indictment of machine learning but simply emphasizes the point that a set of techniques by themselves cannot move us toward greater reflexivity. This is much more a function of the question being asked, the approach taken to the research, and the data being collected and analyzed. Take the wonderful recent paper on “Folklore” by Michalopoulos and Xue (2021). They begin with an archive of thousands of motifs in folklore from 958 societies across the globe that were coded by the folklorist Yuri Berezkin. They classify the motifs into different “concepts” using a supervised NLP model using a dictionary created by the MIT Media Lab. They find that the concepts correlate with geographic attributes of the regions where they were created – for instance, earthquake-prone regions are more likely to have motifs related to earthquakes. After validating the motifs, they examine the concepts to see if concepts that emerge in these motifs are more likely to reflect contemporary realities. They find “a striking consistency between values derived from folklore and contemporary attitudes related to trust, risk-taking, and gender norms.” Thus, they can relate a society’s “ancestral” cultural heritage to its current norms and behaviors showing that narratives can demonstrate powerful path dependency. Another recent paper by Jayachandran, Biradavolu, and Collins (2021) explicitly compares a reflexive, open-ended qualitative study on gender norms in the north Indian state of Haryana to information collected using a structured questionnaire from the same communities. The main goal of the paper is to understand which questions on gender norms and women’s agency, widely used in the literature, best capture the lived reality of women. For these authors, qualitative work is the “gold standard” for understanding women’s agency, so they begin with a careful qualitative study, where a team of researchers conduct open-ended interviews and code these interviews to score 209 households across several domains by the degree to which women have independent agency. The authors treat this as a woman’s “true agency.” They then relate this score, using ML methods (Lasso and random forest), to find the best match between the qualitative score and outcomes measured in the structured questions to find an index of five measures of agency. This method builds on previous work by Blattman et al. (2016) who also use open-ended interviews to
340
V. Rao
build a validation method to understand the extent of measurement error in quantitative self-reported measures of sensitive questions about crime. There is a greater element of reflexivity in this work and the Folklore paper, because of their attempts to use ML and NLP to bring reflexively collected data (qualitative interviews in the case of the gender paper and a close anthropological reading of folklore narratives in the Folklore paper) into conversation with other kinds of more conventional quantitative data and thereby enrich our understanding of important phenomena. The degree of reflexivity is even more acute in two projects that are underway that are conducting open-ended interviews at scale and then planning on analyzing them using ML and NLP. The first study, “The National Poverty Study” (Alexander et al. 2017), is based in the United States and is conducting a “qualitative census of urban and rural poverty” by interviewing 5000 poor and near-poor respondents, sampled to be representative of poor households in urban and rural settings in the United States. They are employing a semi-structured open-ended protocol that covers domains that include life histories, social and familial relationships, and economic hardship. The second study, which I am involved with, piggybacks on an ongoing panel survey of Rohingya refugees and their Bangladeshi hosts in Cox’s Bazar in Bangladesh (Ashwin et al. 2022). We have been conducting 30-min open-ended interviews with 2000 respondents equally divided between hosts and refugees covering 3 domains – aspirations, well-being, and “belongingness” – to understand how the hosts perceive the refugees and vice versa. Both these studies promise to use NLP and ML methods to analyze open-ended interviews with representative samples of respondents, thus attempting to resolve the small-N limitation that was intrinsic to previous open-ended qualitative work.
Studying Process Empirical economics, particularly after the credibility revolution, seems almost singularly focused on outcomes. Mechanisms are generally relegated to ex-post theory, rather than careful observation of the process by which change happens. Ethnography, participant observation, and open-ended qualitative interviews are eminently suited to filling this gap and to understanding “how” as well as “how much” (White 2008; Bamberger et al. 2010). Yet, despite the increased use of qualitative methods in randomized trials and experiments in adjacent disciplines such as psychology (Paluck 2010), they have not yet taken hold in economics. I will briefly review three studies that integrate serious ethnography with quantitative impact evaluations to demonstrate the added value of incorporating an element of reflexivity into causal inference and demonstrate that it can help both in improving scientific understanding and in informing policy-making. Harkening back to Elijah Anderson’s ethnography of Philadelphia, a team of public health researchers and ethnographers collaborated on a mixed-methods study to find ways of dealing with urban “blight” in poorer areas of the city (Branas et al. 2018). An ethnography of a poor “micro-neighborhood” in the city found it “visibly
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 341
impacted” by drug trafficking and gun violence. Overgrown vacant lots were centers of drug dealing and violent activity and were used to store drugs and guns. Virtually all the residents interviewed in this neighborhood, in comparison with a better-off neighborhood, were very supportive of an intervention to clean up and “green” vacant lots. The team designed an intervention which cleaned vacant lots, planted new grass and trees, and installed fencing. Five hundred forty-one vacant lots were randomly chosen from poor neighborhoods in the city and then randomly assigned to treatment and control groups. After a 38-month period, they found that areas around the treated lots saw a statistically significant 36% drop in perceptions of crime and a 76% increase in the use of these vacant lots for social activities. Overall crime dropped by 13% and gun violence by 29%. In this study, ethnography was used less to track the process of change and more to understand community needs to design an effective intervention to address urban blight and associated gun violence and crime. Blattman et al.’s (2021) recent paper also studies drug-ridden neighborhoods in the city of Medellin in Colombia which has long had a reputation of being at the center of Colombia’s drug economy. The research team, which combined economists and ethnographers, spent 4 years conducting open-ended interviews with dozens of gang leaders, managers, and foot soldiers in 30 gangs asking questions about their 30 gangs on their organization, operations, and rule. They also interviewed several experts, police, prosecutors, and community leaders. Drawing on their fieldwork, they designed a quantitative survey on the services provided by drug gangs in neighborhoods, tax collection, extortion, and the extent to which these were perceived as legitimate by the community. They used a natural experiment where the city was reorganized into 16 wards called communas in 1987. Streets on either side of a communa border were very similar but for the three decades since the reorganization were located at different distances from the police and administrative headquarters of the neighboring communas. Quantitative data were collected from 7000 respondents and 223 low- and middle-income neighborhoods. The qualitative and quantitative data provide revealing insights into the workings of the drug economy. Gangs did not provide public goods, the government does, but gangs provide excludable goods such as security services, informal contract enforcement, and dispute resolution which complement the services that governments provide. Consistent with this, the quantitative data show that the services provided by drug gangs tend to be more frequently closer to communa headquarters. Blattman et al. thus challenge the received wisdom that organized crime “fills a vacuum left by weak states” but argue instead “criminal governance can also be a strategic response to strong state presence.” Integrating an ethnography into the design of an impact evaluation can provide a robust way of examining processes and mechanisms. In 2007, I worked with colleagues to test an intervention to improve the quality of citizen engagement in Indian village governments (Rao et al. 2017). The intervention provided 2 weeks of training to village residents on participatory planning, village government regulations, budgets, and administrative systems, which was followed by monthly followup visits by facilitators over a 2-year period to mobilize citizens and translate their demands and needs into actions. We picked a 100 villages at random and then
342
V. Rao
randomly assigned 50 of them to the intervention. Quantitative data from surveys of villages, households, and key informants were conducted prior to the intervention in 2007 and after the 2-year period in 2009. A 10% sub-sample of the quantitative sample, five treatment and five control villages, were selected for the ethnography, which is a large sample for participant observation. We developed an “embedded reporter” approach where five trained ethnographers were hired to live in these villages (each reporter was assigned to one treatment-control pair) to conduct participant observations. They prepared monthly reports of everything they observed – including economic changes, social changes, and political events. These 240-monthly reports constituted our qualitative data. Early rounds of the qualitative data helped inform the development of the quantitative survey instruments. The quantitative data showed considerable improvement over time across a variety of indicators, but the outcomes between treatment and control villages were not significantly different from each other. The ethnography allowed us to unpack the reasons why the intervention “failed” highlighting the role of variations in the quality of facilitation, lack of top-down support, and difficulties in confronting the stubborn challenge of persistent inequality. We found that while surveys have the advantage of being able to measure predictable outcomes that have impacts that are large enough to be captured, major events sometimes occur that are best investigated on the ground, during the moment they happen, by participant observers. Important shifts occur during points of conflict or during periods of mobilization and coordination. These are not quotidian events and are very hard to predictably measure. Moreover, qualitative investigators can build strong relationships within communities that allow them to “see” differently and thus capture insights that a survey interviewer is unable to do. These three studies, all of which integrate qualitative and quantitative methods into their design at the outset and not as an afterthought, clearly show how much these methods complement each other. Ethnographic work not only helps design better survey instruments, but it can also help design better interventions and provide deep insights into the processes that underlie the changes measured by quantitative data. This obvious point has been made several times, but ethnographic methods still have not become a routine part of the economist’s toolkit, though recently a small number of impact evaluation papers have used open-ended interviews and focus groups to understand mechanisms (e.g., Muralidharan and Singh 2020).
Participation: Respondent as Analyst If cognitive empathy is largely about reducing the social and political distance between the researcher and the researched, participation, or co-production, takes that a step further by actively involving communities in the research process. Robert Chambers and his colleagues at the Institute of Development Studies at Sussex have for several decades now advocated methods broadly labelled as “Participatory Rural Appraisal” that teach elementary graphic and visualization tools to communities so
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 343
that they can analyze and improve their own lives (Chambers 1994) which have had a sizable impact in the practice of community-driven development. Some scholars have made the case for co-produced research, where questions are designed with the input of the communities being studied, members of the community are involved in data collection, and research results are then fed back to the community, arguing that this improves the “rigor, relevance and reach of science” (Balazs and Morello-Frosch 2013). The Centers for Disease Control and Prevention released a “Tools and Techniques” note outlining a method for a more limited form of participation – where research findings are shared via a process of two-way dialogue with marginalized communities, arguing that this builds trust and improves the dissemination of public health advice (McDavitt et al. 2016). Development economists have made a similar case for presenting results from randomized control trials to communities, though some have cautioned that this could affect the integrity of the research process (McKenzie 2011). With all this, the fact remains that it is rarely done. Research in economics – even research that displays a great deal of cognitive empathy – remains largely extractive, and any benefits to the respondents who participate (other than the payments for participating in a survey) are usually indirect, via the long route of research outcomes affecting policy which in turn may have a possible impact on respondents. Recently several projects around the world have been involved in efforts to “democratize data,” by involving people in designing, collecting, and analyzing their own data tailored to answer their own needs (Ada Lovelace Institute 2021). Yuen Yuen Ang (2019), building on Clifford Geertz description of ethnography as “thick description,” has called this “thick data.” In 2014, some colleagues and I co-produced a method, that we called participatory tracking, with tribal communities in South India to facilitate community-level decision-making by democratizing the design, collection, and analysis of data (World Bank 2021). Representatives of over 200 tribal villages engaged in several weeks of deliberations to think about what constituted the good life for them turned those ideas into indicators measured with survey questions and then tested and adapted those questions in their villages so that they did not take more than 30 min to answer. The questionnaires were quite different from those generally used to track poverty with only a 17% overlap in questions with the Indian National Sample Survey. We incorporated the community-designed questionnaire into a tablet-based electronic survey and used a system of video-based training to train community representatives to use the tablet to conduct surveys of their own villages. In our pilot, which covered one district, we were able to conduct a census of 32,000 households in the district in about 6 weeks. Once the survey was conducted, the data were dispatched directly to a cloud server to prevent anyone from tampering with them. The same exercise was repeated the following year. The goal was for villagers to use the data for two purposes, to track changes in the quality public services and in their living standards and to use the data to make better decisions in village meetings (gram sabhas). This was challenging because levels of literacy were low (at about 65%) and numerical literacy was much lower. Thus, we could not use conventional data analytics and visualizations and had to develop ways
344
V. Rao
of showing the data that would be intelligible to someone who could not read or write. So, we co-produced data visualizations with the community, iterating a few times till we were confident that they were widely understood. These data were then presented at village planning meetings, including the village meetings described in earlier sections, and we found that they substantially helped improve the quality of discussion by allowing citizens and officials to focus on the issues rather than debate the facts. (More information is available here https://so-prototype.webflow.io/ categories/democratizing-data) Efforts such as participatory tracking, which are still in their infancy, harness technology to empower communities to conduct their own research designed to study their own priorities. Professional researchers are largely on the sidelines serving mainly as advisors and facilitators. The rise in data literacy around the world, and technology to facilitate data collection, mapping, and visualization, has made such efforts a promising line of action. In effect, we have come full circle and returned to Charles Booth and his idea of using members of the community to collect data on themselves and their neighbors – with their full consent – and to use maps and visualizations to allow people to study to analyze the challenges that they face. The goal is to empower them to conduct research in a manner unmediated by professionals and to act on the research findings to improve their lives.
Conclusion Those of us who do research on topics such as development, poverty, or discrimination study people who are almost always at a considerable economic, social, and cultural distance from us. This raises the important question of “Whose Reality Counts?” (Chambers 1997). Do we focus on trying to meet the current standards of rigor of our discipline, satisfying our peer researchers, and ensuring our objectivity by maintaining an arms-length distance from those whom we research? Or is it imperative upon us to minimize the gap by discovering questions that closely reflect the interests and perspectives of our research subjects via a process of dialogue and use methods and modes of inquiry that as closely as possible reflect their lived reality? Burawoy (1998) argues that research of the first kind is positivist, where objectivity and detachment are the organizing principle, and the second is reflexive where engagement, not detachment, is the primary organizing force. He further argues that “reflexive science” and “positive science” represent entirely different approaches to social science and live in separate spheres. I disagree. I argue in this chapter that there is a middle path between Burawoy’s sharp divide between positive and reflexive science. While Economics is still a long way away from becoming a predominantly reflexive social science, there are four ways in which it can move toward greater reflexivity. All of them could be helped by a judicious mix of quantitative and qualitative methods: greater cognitive empathy, incorporating qualitative and narrative information explicitly into our analytical toolkit, a focus on understanding process as well as outcomes, and facilitating efforts toward greater participation of communities in our research. In this chapter, I have
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 345
drawn on work in economics and the allied social sciences to provide examples of each modality and tried to show how mixing methods might add value. However, there are a few caveats to note if we want to move further on the path toward greater reflexivity. 1) Cognitive empathy – as numerous economists have shown – can result in entirely quantitative analysis which emerges from a deep engagement with the field. Conversely, the analysis of qualitative and narrative data is not necessarily always reflexive. However, the explicit analysis of narratives and their incorporation into the empirical toolkit of economists can move us toward greater reflexivity by permitting the voices of respondents to be directly listened to and learned from, rather than being mediated via the artificial construct of a structured survey instrument. The reluctance among economists to treat open-ended narratives, unless done at scale, as anything other than colorful anecdata is a serious limitation. 2) Gathering good qualitative information, and analyzing it well, requires us to learn from the standards of rigor followed by the qualitative social sciences (see, for instance, Lareau 2021). Simply going to the field, conducting a few interviews and focus group discussions is not good enough. In particular, there is a misconception among some economists that qualitative work is cheap, “quick and dirty,” or easy. Good qualitative and mixed methods work can add quite a lot to the cost of a study. 3) The logic of sampling for qualitative data is not necessarily the same as for quantitative data. It depends on the nature and purpose of the study. For instance, when the purpose is to conduct case studies without necessarily claiming broad representation, this has a different logic (Small 2009) than when qualitative data is collected, for instance, to understand patterns of well-being across a large country which would follow the familiar logic of stratified probability sampling (Alexander et al. 2017). 4) Integrating qualitative and quantitative work can be done in different ways. Following classical inferential logic, small-N qualitative work can be conducted using the case study method to develop hypotheses, which can then be tested for their generalizability, possibly mediated by a theoretical model, with quantitative data collected from a representative sample of respondents (Rao 1997b). The integrated collection and analysis of qualitative and quantitative information can also be analyzed using Bayesian inference (Humphreys and Jacobs 2015). 5) Machine learning and natural language processing are a double-edged sword. They offer tremendous advantages in moving us toward analyzing narrative data at scale. Yet, supervised methods that rely on biased training sets, such as sentiment dictionaries developed for western contexts applied to non-western linguistic cultures, can result in substantial bias. Furthermore, as Woolcock (2021) has argued, when machines are used for analyzing data, narratives have the danger of being analyzed out of context and without nuance, resulting in misinterpretation. In other words, relying on machines without sensitive human
346
V. Rao
intervention has the danger of turning reflexively collected data into non-reflexive analysis. 6) Process, studied carefully with qualitative methods, can be extremely valuable to understand the mechanisms of change and thus complement time-variant quantitative studies, particularly impact evaluations. Understanding process matters not just for research but also for policy, where an exclusive emphasis on policies that can only be assessed using experiments or impact evaluations can sharply limit our capacity to imagine and create a better world (Rao 2019). 7) The potential for the direct participation of “respondents,” “beneficiaries,” and “subjects” in research is vastly unexplored. If our purpose as researchers is to assist in the process by which people become better-off, then more can be done to design and share research with people and even co-create tools so that they can conduct research on themselves without the mediation of experts. This has implications for many things, including the measurement of well-being and poverty. Acknowledgments I am grateful to Harold Alderman, Monica Biradavolu, and Irene Bloemraad for valuable discussions and feedback and to Paromita Sanyal and Michael Woolcock for our many years of collaboration and conversation during the course of which some of these ideas were developed.
References Ada Lovelace Institute (2021) Participatory data stewardship: a framework for involving people in the use of data, September, Creative Commons license Alexander JT, Andersen R, Cookson PW, Edin K, Fisher J, Grusky DB, Mattingly M, Varner C (2017) A qualitative census of rural and urban poverty. Ann Am Acad Polit Soc Sci 672(1): 143–161 Anderson E (1990) Streetwise: race, class and change in an urban community. University of Chicago Press, Chicago/London Ang YY (2019) Integrating big data and thick data to transform public services delivery. IBM Center for the Business of Government, Washington, DC Appadurai A (1989) Chapter 11. Small-scale techniques and large-scale objectives. In: Bardhan P (ed) Conversations between economists and anthropologists. Oxford University Press, Oxford, UK/Delhi Ashwin A, Rao V, Biradavolu M, Haque A, Khan A, Krishnan N, Nagy P (2022) Qualitative Analysis at Scale: An Application to Aspirations in Cox’s Bazaar, Bangladesh. Policy Research Working Papers;10046. World Bank, Washington, DC. © World Bank. Balazs CL, Morello-Frosch R (2013) The three Rs: how community-based participatory research strengthens the rigor, relevance, and reach of science. Environ Sci 6(1):9–16 Bamberger M, Rao V, Woolcock M (2010) Using mixed methods in monitoring and evaluation: experiences from international development. In: Tashakkori A, Teddlie C (eds) SAGE handbook of mixed methods in social & behavioral research. SAGE Publications, Thousand Oaks, pp 613–642 Ban R, Rao V, Jha S (2012) Who has voice in a deliberative democracy? Evidence from transcripts of village parliaments in India. J Dev Econ 99(10):428–438 Bardhan P (ed) (1989) Conversations between economists and anthropologists. Oxford University Press, Oxford, UK/Delhi
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 347
Bardhan P, Rudra A (1978) Interlinkage of land, labour and credit relations: an analysis of village survey data in East India. Econ Polit Wkly 13:6–7 Bell C (2020) Tales of peasants, traders, and officials: contracting in rural Andhra Pradesh 1980–82, International development in focus. World Bank, Washington, DC Blattman C, Jamison J, Koroknay-Palicz T, Rodrigues K, Sheridan M (2016) Measuring the measurement error: a method to qualitatively validate survey data. J Dev Econ 120:99–112 Blattman C, Duncan G, Lessing B, Tobon S (2021) Gang rule: understanding and countering criminal governance. Mimeo 26:12–15 Bliss CJ, Stern NH (1982) Palanpur: the economy of an Indian village. Clarendon Press, Oxford, UK Bloch F, Rao V (2002) Terror as a bargaining instrument: a case study of dowry violence in rural India. Am Econ Rev 92(4):1029–1043 Booth C (1892–1897) Life and labour of the people in London. Macmillan, London Branas CC, South E, Kondo MC, Hohl BC, Bourgois P, Wiebe DJ, MacDonald JM (2018) Citywide cluster randomized trial to restore blighted vacant land and its effects on violence, crime, and fear. Proc Natl Acad Sci U S A 115(12):2946–2951 Burawoy M (1998) The extended case-study method. Sociol Theory 16:1 Caliskan A (2021) Detecting and mitigating bias in natural language processing. In: Artificial Intelligence and Emerging Technology (AIET) initiative. Brookings Institution, Washington, DC Chambers R (1994) The origins and practice of participatory rural appraisal. World Dev 22(7): 953–969 Chambers R (1997) Whose reality counts? Putting the first last. Intermediate Technology Publications, London Collins D, Morduch J, Rutherford S, Ruthven O (2009) Portfolios of the poor: how the world’s poor live on $2 a day. Princeton University Press, Princeton Dreze J (2017) Sense and solidarity: Jholawala economics for everyone. Permanent Black, New Delhi Epstein TS (1962) Economic development and social change in South India. Manchester University Press, Manchester Gentzkow M, Shapiro JM, Taddy M (2019a) Measuring group differences in high-dimensional choices: method and application to congressional speech. Econometrica 87(4):1307–1340 Gentzkow M, Kelly B, Taddy M (2019b) Text as data. J Econ Lit 57(3):535–574 Goffman E (1960) Stigma: notes on the management of spoiled identity. Prentice-Hall, New York Goffman E (1969) Strategic interaction. University of Pennsylvania Press, Philadelphia Goldstein M, Udry C (2008) The profits of power: land rights and agricultural investment in Ghana. J Polit Econ 116(6):981–1022 Gregory CA, Altman JC (1989) Observing the economy. Routledge, London Gudeman S (2001) The anthropology of economy. Blackwell Publishers, Malden/Oxford, UK Himanshu H, Lanjouw PF, Stern N (2018) How lives change: Palanpur, India and development economics. Oxford University Press, Oxford, UK/New York Humphreys M, Jacobs AM (2015) Mixing methods: a Bayesian approach. Am Polit Sci Rev 109(4): 653–673 Jayachandran S, Biradavolu M, Cooper J (2021) Using machine learning and qualitative interviews to design a five-question women’s agency index. Northwestern Global Poverty Research Lab working paper no. 21-104, March Kanbur R, Shaffer P (2007) Epistemology, normative theory and poverty analysis: implications for Q-squared in practice. World Dev 35(2):183–196 Lamont M (2000) The dignity of working men: morality and the boundaries of race, class and immigration. Russell Sage Foundation/Harvard University Press, New York/Cambridge, MA Lareau A (2021) Listening to people: a practical guide to interviewing, participant observation, data analysis, and writing it all up. University of Chicago Press, Chicago
348
V. Rao
Lieberman E, Miller A (2021) Do online newspapers promote or undermine nation-building in divided societies: evidence from Africa. Nations Natl 27:238–259 Matamoros-Fernández A, Farkas J (2021) Racism, hate speech, and social media: a systematic review and critique. Telev New Media 22(2):205–224 McDavitt B, Bogart LM, Mutchler MG, Wagner GJ, Green HD Jr, Lawrence SJ et al (2016) Dissemination as dialogue: building trust and sharing research findings through community engagement. Prev Chronic Dis 13:150473 McKenzie D (2011) Should we share results with respondents? Development Impact Blog, 3 October Michalopoulos S, Xue MM (2021) Folklore. Q J Econ 136(4):1993–2046 Morduch J, Schneider R (2019) The financial diaries: how American families cope in a world of uncertainty. Princeton University Press, Princeton Morgan MS (2019) Charles Booth’s London poverty maps. Thames and Hudson, London Muralidharan K, Singh A (2020) Improving public sector management at scale? Experimental evidence on school governance in India. NBER working paper 28129, Cambridge, MA Olson M (1965) The logic of collective action. Harvard University Press, Cambridge, MA Ostrom E (1990) Governing the commons: the evolution of institutions for collective action. Cambridge University Press, Cambridge, UK/New York Paluck EL (2010) The promising integration of qualitative methods and field experiments. Ann Am Acad Polit Soc Sci 628(1):59–71 Parthasarathy R, Rao V, Palaniswamy N (2019) Deliberative democracy in an unequal world: a textas-data study of South India’s village assemblies. Am Polit Sci Rev 113(3):623–640 Rao V (1997a) Wife beating in rural South India: a qualitative and econometric analysis. Soc Sci Med 44(8):1169–1180 Rao V (1997b) Can economics mediate the relationship between anthropology and demography. Popul Dev Rev 23(4):833–838 Rao V (2019) Process-policy & outcome-policy: rethinking how to address poverty & inequality. Daedalus 148(3):181–190 Rao V, Ananthpur K, Malik K (2017) The anatomy of failure: an ethnography of a randomized trial to deepen democracy in rural India. World Dev 99:481–497 Sanyal P, Rao V (2019) Oral democracy: deliberation in Indian village assemblies. Cambridge University Press, Cambridge, UK Schiller R (2019) Narrative economics: how stories go viral and drive major economic events. Princeton University Press, Princeton Sharan MR (2021) Last among equals: power, caste and politics in Bihar’s villages. Westland Press, Delhi Sharan MR, Kumar C (2021) Something to complain about: how minority representatives overcome ethnic differences. Mimeo, Department of Agricultural and Resource Economics, University of Maryland, College Park Singh B (2015) Poverty and the quest for life: spiritual and material striving in rural India. University of Chicago Press, Chicago/London Small ML (2009) “How many cases do I need?”: on science and the logic of case selection in fieldbased research. Ethnography 10(1):5–38 Small ML (2011) How to conduct a mixed methods study: recent trends in a rapidly growing literature. Annu Rev Sociol 37:57–86 Small ML (2018) Rhetoric and evidence in a polarized society, Coming to terms with a polarized society lecture series. Institute for Social and Economic Research and Policy, Columbia University, New York Srinivasan TN (1989) Chapter 10. On studying socio-economic change in rural India. In: Bardhan P (ed) Conversations between economists and anthropologists. Oxford University Press, Oxford, UK/Delhi Stephens-Davidowitz S (2014) The cost of racial animus on a black candidate: evidence using Google search data. J Public Econ 118:26–40
14
Can Economics Become More Reflexive? Exploring the Potential of Mixed Methods 349
Thaler RH (2016) Behavioral economics: past, present, and future. Am Econ Rev 106(7): 1577–1600 Udry C (2003) Fieldwork, economic theory and research on institutions in developing countries. Mimeo White H (2008) Of probits and participation: the use of mixed methods in quantitative impact evaluation. IDS Bull 39:98–109 Woolcock M (2021) On Folklore and “Folklore” (QJE): forging dialogue, reconciling tradeoffs. Development Impact Blog, 23 February World Bank (2021) Helping communities to gain the ability to collect and analyze their own data. In: Spotlight 1.1. World Development Report 2021, data for development. The World Bank Group, Washington, DC
Part III Intersectionality and Discrimination
Tackling Intersecting Inequalities: Insights from Brazil
15
Naila Kabeer and Ricardo Santos
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Objectives of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conceptualizing Intersecting Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Evolution of Intersecting Inequalities in Brazil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From “Illiterate Agro-Export Outpost” to “Conservative Modernization” . . . . . . . . . . . . . . . . . Shifting Constructions of Inequality in the Brazilian Context: Vertical, Horizontal, and Intersecting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trends in Poverty and Inequality 2002–2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Poverty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Labor Market Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Access to, and Control Over, Land . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Educational Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Explaining the Decline in Intersecting Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From “Conservative Modernization” to “Liberal Neo-Developmentalism” . . . . . . . . . . . . . . . The Politics of Social Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
354 354 355 357 357 358 366 368 373 377 385 388 394 394 397 402 403 404
Abstract
A concern with absolute poverty, defined as a money-metric measure of the ability to meet basic subsistence needs, has occupied a central place within the international development agenda for much of its existence. Efforts in recent N. Kabeer (*) Department of International Development, London School of Economics and Political Science, London, UK e-mail: [email protected] R. Santos Ricardo Santos UNU-WIDER, Helsinki, Finland e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_53
353
354
N. Kabeer and R. Santos
years to promote a more multidimensional understanding of poverty, with a particular focus on human capabilities, have met with some success, and there has been a growing literature on discrimination based on social identities such as race, gender, and ethnicity. Attention to income inequalities waxed and waned in the earlier decades of development planning, but the issue has come on to the agenda in a more sustained way in recent years. The objective of this chapter is to bring these different concerns together by considering the intersection between material and identity-based inequalities and their consequences for the multiple dimensions of poverty and well-being. We aim to do this by exploring the phenomenon of intersecting inequalities in the context of Brazil where it has deep roots in the country’s history. We examine the history of this phenomenon, its unexpected decline in the first decade or so of the twenty-first century and ask how and why this happened. Keywords
Intersecting inequalities · Gender · Race · Ethnicity · Economic policy · Affirmative action · Brazil
Introduction Objectives of the Study A concern with absolute poverty, defined as a money-metric measure of the ability to meet basic subsistence needs, has occupied a central place within the international development agenda for much of its existence. Efforts in recent years to promote a more multidimensional understanding of poverty, with a particular focus on human capabilities, have met with some success, and there has also been a growing literature on discrimination based on social identities such as race, gender, and ethnicity. Attention to income inequalities waxed and waned in the earlier decades of development planning, but the issue has come on to the agenda in a more sustained way in recent years. The objective of this chapter is to bring these different concerns together by considering the intersection between material and identitybased inequalities and their consequences for the multiple dimensions of poverty and well-being. We aim to do this by exploring the phenomenon of intersecting inequalities in the context of Brazil where it has deep roots in the country’s history. We examine the history of this phenomenon, its unexpected decline in the first decade or so of the twenty-first century and ask how and why this happened. The structure of the chapter is as follows. The rest of section “Introduction” elaborates on the concept of intersecting inequalities and examines its manifestation in the Brazilian context. Section “The Evolution of Intersecting Inequalities in Brazil” provides a brief history of the evolution of these inequalities in the Brazilian context and documents its current manifestations. Section “Trends in Poverty and Inequality 2002–2013” uses national household survey data to track trends in key
15
Tackling Intersecting Inequalities: Insights from Brazil
355
socioeconomic indicators between 2002 and 2013. It draws on the relevant data sets from the Brazilian National Survey by Household Sampling (Pesquisa Nacional por Amostra de Domicílios, PNAD). This has been made public and freely available by the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística, IBGE).1 It finds that not only have income poverty and inequality declined over this period but that also there has been a decline in intersecting inequalities. Sections “Trends in Poverty and Inequality 2002–2013” and “Explaining the Decline in Intersecting Inequalities” draw on various bodies of literature to piece together an explanation for how and why this happened. Section “Trends in Poverty and Inequality 2002–2013” explores explanations that focus on economic and social policy during this period while section “Explaining the Decline in Intersecting Inequalities” examines the politics of inclusion. Section “Conclusion” concludes by asking what the Brazilian experience offers by way of generalizable lessons.
Conceptualizing Intersecting Inequalities There have been two broad approaches to the analysis of inequality within international development studies. The first revolves around economic inequality conceptualized in terms of income and is measured at the individual/household level. While the early focus on money-metric measures have been supplemented over time with more multidimensional approaches which encompass inequalities in material wealth and human capabilities, these continue to be measured at the level of individuals. This has been termed a “vertical” model of inequality as it is based on ranking individuals/households by their income, wealth, or human capabilities (Stewart 2002). It can also be described as a class-based understanding of inequality. A second approach revolves around the phenomenon of social discrimination. It uses group-based disadvantage as its entry point into the analysis of inequality where the inequality in question revolves around the social identity of different groups. This gives rise to what has been described as a “horizontal” model of inequality (Stewart, op. cit.) in that the groups in question cut across the different strata that make up the vertical model. Group-based inequalities are the product of cultural norms, values, and practices which serve to routinely disparage, stereotype, exclude, ridicule, and demean certain groups relative to others, denying them full personhood and the ability to participate as equal citizens in the life of their community. The identities in question can take many different forms, but some are socially ascribed from birth, and hence more immutable and harder to shed than others. This, for instance, differentiates identities associated with gender, caste, race, and ethnicity from those associated with age and certain forms of disability.
1
The datasets were sourced from the IBGE website, in the following page: http://www.ibge.gov.br/ home/estatistica/populacao/trabalhoerendimento/pnad2011/microdados.shtm
356
N. Kabeer and R. Santos
The two approaches to inequality thus focus on quite distinct axes of disadvantage: the class distribution of resources and the social devaluation of identity. It is possible to be economically deprived, to lack the means to meet basic needs, without necessarily being despised for it: This underpins the distinction often made between the “deserving” and “undeserving” poor. Similarly, it is possible to face social discrimination, without necessarily facing material deprivation. Gender, for instance, cuts across economic strata so that women are fairly evenly distributed across the economic hierarchy, but generally occupy a subordinate status relative to men within different economic strata. Gender is always associated with discrimination but not necessarily with poverty. However, it is the intersection between class inequalities and inequalities based on identity that serve to define the most severe and often the most enduring form of social exclusion in societies. As Kabeer (2010) noted, groups at these intersection of these inequalities are overrepresented everywhere at the bottom end of the income distribution. Her analysis also highlighted a spatial dimension to these intersecting inequalities: socially excluded groups tended to be concentrated in the most disadvantaged locations – remote and challenging rural terrains or overcrowded and underserved low-income neighborhoods. These social, economic, and spatial inequalities have in turn contributed to political disenfranchisement: Such groups were generally denied voice and influence in collective decisions that affected their lives. The language of “vertical” and “horizontal” inequalities is not always helpful in capturing what is at issue here because of the “grid-like” symmetry evoked by these terms. The language of intersection serves to highlight the fact we are not dealing with an additive model of different inequalities. Rather they reinforce and exacerbate each other, leading to the “sharp discontinuities” and “intensifications” which have been found to distinguish the poor from the poorest in many regions of the world (Lipton 1983). Thus while gender on its own does not signify poverty and social exclusion, the intersection of gender with material inequality is generally associated with an intensification of disadvantage so that women and girls from poor and marginalized groups are almost always the most excluded among excluded groups in a society. Brazil is a useful case study of intersecting inequalities for a number of reasons. For much of the twentieth century, Brazil has been one of the fastest growing economies of the world. Even between 1980 and 2000, when growth had slowed down in much of the world, GDP per capita grew at an annual average rate of 2.5% (Marió and Woolcock 2008). Yet its Gini coefficient remained among the highest in the world throughout this period and fairly constant at around 0.58 and 0.60. In addition, as will be discussed in the chapter, race, ethnicity, gender, and location intersect with the income distribution in Brazil to produce distinct and enduring social and spatial patterns to poverty and inequality. These intersecting inequalities are bound up with the history of colonization in the country and its subsequent efforts at economic development. Their recent declines can thus be seen as a break with a very long history. We turn in the next section to a brief summary of this history.
15
Tackling Intersecting Inequalities: Insights from Brazil
357
The Evolution of Intersecting Inequalities in Brazil From “Illiterate Agro-Export Outpost” to “Conservative Modernization” Portuguese colonization on Brazil began around 1500 with the goal of exploiting its wealth and converting its population to Christianity. The colonizers carried the values of a strictly hierarchical society dominated by a hereditary monarch and aristocratic elite with them into their colony, regarding it as little more than an “illiterate agro-export outpost” (Skidmore 2004). They began by enslaving the Indigenous population to grow sugar but with the decimation of the population by war and disease, turned to the import of slaves from Africa, adding a further layer of “racially inferior people” to an already highly stratified social hierarchy. There were around 3.6 million Africans by 1850 when the slave trade ended, more than anywhere else in the region. They were largely concentrated in the north and north-east of the country which had been the center of the early plantation economy. The colonial settlers subsequently expanded into mining and cattle rearing and then to coffee. The country was under Portuguese rule for three centuries with very little experience of representative self-government (Skidmore 2004). It was governed by Crown-appointed governors-general and bureaucracies till its independence in 1822 when it declared itself an empire. While elections took place periodically in later years, voting was restricted to the small percentage of the population who were literate which, even in the 1920s, was just 25% of the population. One reason for these low levels of education was its suppression under colonialism and prohibition of printing (Skidmore 2004). There were few schools and no universities. Only the children of the very rich among the settlers gained entrance to school and, in the absence of universities, went to Portugal for their university education. The first university was not founded till 1932. Consequently, even into the twentieth century, Brazil had the worst educational record of the major Latin American countries (Birdsall and Sabot 1996). The abolition of slavery in 1888 foregrounded two sets of concerns among the Brazilian elite. One was the disappearance of the pool of cheap, “unfree” labor. The other was the fear that prevailing theories of scientific racism would lead European countries to look down on a country with such a large African and Indigenous population. Both concerns were addressed through a strategy of subsidizing the large-scale immigration of European workers (but forbidding Asian immigration) who would not only add to the workforce but also “whiten” the population and improve the physical and moral characteristics of the Brazilian people (Rezende and Lima 2004). This meant that the newly freed slaves had to compete with European migrants on highly disadvantaged terms which relegated them to the poorest paid jobs in the economy. Brazil’s subsequent development strategy has been described as a process of “conservative modernization” (Marió and Woolcock 2008) entailing the selective incorporation of certain segments of the population into the modernizing sectors of the economy, society, and the political system and the systematic exclusion of the
358
N. Kabeer and R. Santos
rest. Until the 1930s, the economy was dominated by the agricultural sector largely based in the north. Its early efforts to industrialize were boosted by the Depression when a shortage of foreign exchange served to promote domestic manufacture of previously imported products. By the 1950s, the country had embarked on a massive import substituting industrialization program to end its dependence on foreign imports. The government led this process through a corporatist strategy intended to both promote industrialization and to manage the accompanying urbanization and unionization. It created a network of state enterprises to expand its role in the economy, used a large-scale labor bureaucracy to bring unions under its control, and put in place generous pension schemes for federal government employees. Economic growth and industrialization in Brazil was constant and fast until the early 1980s. It was under military rule for much of this period (1964–1985). Economic success partly explained why the military was able to hold on to power for this length of time. However, this was also a period of widening economic inequalities. Industrialization was concentrated in São Paulo and other regions of South-East and South to the neglect of the north and north east. This emerging regional inequality mapped into a racial geography in that Afro-Brazilians were, and remain, concentrated in north and north-east of the country while the majority of whites were to be found in the increasingly prosperous south. The neglect of agriculture meant that between 1960 and 1980, more than 30 million people moved from the countryside into the cities in search of work (Marió and Woolcock 2008). For the better-educated workers and their families, the move into cities held out real prospects of upward mobility and material improvement. But for the poorer majority, it meant finding work in the rapidly expanding informal economy with little hope of better prospects in the foreseeable future. The 1980s saw a slowdown in growth rates accompanied by rising unemployment, high rates of inflation, and further migration into urban areas. Recurrent economic crises gave rise to intensified political mobilization which culminated in the end of military rule in 1985. This was followed by the first democratic election to be held in over two decades and the adoption of a progressive new constitution in 1988. However, successive governments spent much of the decade that followed dealing with hyperinflation and low, sometimes, negative rates of growth. A number of policies directed toward the poor were put in place in the 1990s, and while this led to some decline in poverty, income inequality did not show any sign of change until the start of the new century.
Shifting Constructions of Inequality in the Brazilian Context: Vertical, Horizontal, and Intersecting While racial inequality is closely bound up with income inequality in Brazil, the understanding of the relationship between the two has undergone a number of shifts over time. This in turn reflects the shifting interpretations given to miscegenation, or racial mixture, as one of the defining characteristics of Brazilian society, within both official and popular constructions of national identity. The origins of miscegenation
15
Tackling Intersecting Inequalities: Insights from Brazil
359
go back to the colonial period which was characterized by a high ratio of males to females among the Portuguese settler population combined with relatively unlimited sexual “access” by Portuguese males to female slaves of Indigenous and African origin. The result was a steady increase in the mixed race population. Scientific theories about the inferiority of the nonwhite races flourished in Europe and were incorporated into Brazilian colonial ideology. However, while the racial distinction in Brazil mapped into a social distinction, it was blurred to some extent by the common practice of freeing the offspring of white males and black slave women. This gave rise to mulattos as an intermediate social category of “browns” while “blacks” continued to make up most of the slave population. This creation of what has been described as the “mulatto escape hatch” (Degler 1971) was believed to have forestalled the development of rigid racial boundaries. Instead, it seems to have given rise to the conceptualization of race in terms of skin color rather than descent and as a complicated and continuous rather than a simple, categorical variable. The complex understanding of the relationship between race and inequality in Brazil is clearly illustrated in the debates that ensued as earlier theories about racial inferiority gave way to more contested conceptualizations about how race fitted into the social structure. The discussion of these debates as elaborated by Telles (2004), Skidmore (1992), and others echo the discussion about horizontal, vertical, and intersecting inequalities that we started out with in this chapter. We draw on this work to summarize the different positions in these debates.
From Horizontal to Vertical Inequality According to Telles (2004), the first generation of researchers to contribute to these debates began writing around the 1920s and sought to construct Brazil as a racial democracy in which race made no difference to status and opportunities. They were largely North Americans or had been trained in North America, carried out their research in the North-east of Brazil which had a largely Afro-Brazilian population, and focused their enquiry on horizontal relationships, the relationships between different racial groups. They stressed the fluidity of racial classifications, in marriage and in friendship between people of different races and color and concluded that race was irrelevant to the social hierarchy. Any differences in social status by race that might exist were the legacy of slavery that had only recently been abolished and would diminish over time. The work of Gilberto Freyre (1933) was particularly influential in promoting this vision. He argued that the overwhelmingly white elite had acquired valuable cultural assets from their intimate mixing with the African population whose culture he considered more evolved than Indigenous Brazilians (Rezende and Lima 2004). He juxtaposed the figure of the master and slave, representing Portuguese and African culture, as the two basic poles that constituted Brazilian colonial society. “Brazilianness” was the result of the productive fusion of two antagonistic cultures placed in a hierarchical relation to each other but in which opposites complemented each other harmoniously: “the European culture came into contact with the indigenous, softened by the oil of the African” (cited in Rezende and Lima 2004, p. 759).
360
N. Kabeer and R. Santos
The claim that race was irrelevant to social mobility proved extremely useful to the conservative civil and military elites who ruled Brazil for much of the twentieth century and was incorporated into their nationalist rhetoric as a means of staving off the possibility of racial unrest. There was systematic avoidance of any mention of race by successive governments leading to a major lacuna in relevant data. While race had been included in the first census of 1872 and then in 1890, basic data collection between 1890 and 1940 omitted all reference to racial categories (Skidmore 1992). The second wave of scholarship, starting in the 1950s, were less concerned with the consequences of miscegenation and focused instead on vertical inequalities. These were mainly Brazilian scholars writing within a Marxist framework who researched the white-dominated regions of the South and South east. They found a very different Brazil, with recent European migrants leapfrogging over blacks and mulattos in the labor market, pervasive prejudice and discrimination against nonwhite populations, and rigid racial distinctions and limited social interactions between white people, on the one hand, and the black and brown population, on the other. However, they interpreted racial inequality as a manifestation of the vertical inequalities embedded in the class hierarchy. They argued that because the newly freed black population had not been able to compete on equal terms with the influx of European immigrants that accompanied the abolition of slavery, they had been condemned to “economic maladjustment, occupational regression, and social imbalance” (Rezende and Lima 2004, p. 761). The racial and ethnic specificities which gave rise to group-based inequalities which cut across class were largely ignored, providing the ideological justification for racially and ethnically blind national policies for poverty reduction.
From Vertical to Intersecting Inequalities Much of this debate was carried out within historical and anthropological disciplines and relied on “soft” data, a rich repertoire of laws, travelers’ accounts, memoirs, newspaper articles, and detailed ethnographic accounts of different dimensions of relations among and within different racial categories but unable to establish whether race made a macro difference in social or economic outcomes. Nor was it possible for other social scientists to do so. The harsh political repression that characterized the years of military rule had curtailed the scope for critical research, while the failure of official statistics to include race as a category made it further more difficult to investigate claims about group-based discrimination. Race was reintroduced into the census in 1950 and 1960 but could not be matched with socioeconomic data (Skidmore 1992). It was omitted again in the 1970 census, and it was only in the face of vigorous protest from social scientists, sections of the press, and an emerging group of Afro-Brazilian activists that it was included in the 1980 census. Restrictions began to ease, and the 1976 PNAD was the first time that the government collected and published data on income and employment by race. This new generation of Afro-Brazilian activists had emerged in the 1970s to contest the myth of Brazil’s racial democracy and to protest police brutality,
15
Tackling Intersecting Inequalities: Insights from Brazil
361
mistreatment by public officials, and discrimination in the job market and public places. While they did not enjoy broad support, their arguments were supported by a third wave of scholarship that moved beyond constructions of race relations as either essentially democratic or racial inequality as an accidental feature of class to an analysis of racial and ethnic discrimination as independent of class, but exacerbating its effects. This view of race was made possible by availability of new forms of data. These studies, some of which are summarized below, demonstrated that “100 years after the abolition of slavery, Afro-Brazilians continued to dominate in the lowest economic strata” (Lovell and Wood 1998, p. 105). This was also a period that saw the beginning of attention to the gender as a further dimension of the intersection between class and racial inequality. A number of authors traced the collective struggles of Afro-Brazilian women to navigate their way through the race and class biases of the women’s movement and the patriarchal politics of the black movement in order to affirm blackness and femaleness as legitimate forms of self-identification (Caldwell 2007; Lovell 2000). As Rezende and Lima (2004) observed, Afro-Brazilian women had either been invisible in the debates around the nature of inequality in Brazil or they had been brought in to buttress particular positions. The work of Gilberto Freyre (1933) had located the birth of racial democracy in Brazil in the intimate structure of the plantation house and slave quarters which allowed miscegenation to take place under the aegis of the rural patriarchal family. Ascribing a near-mythical status to the figure of the slave woman, he attributed any violence they had suffered to slavery. Where gender was discussed in the class-based paradigm of inequality, women were simply part of the poor and uneducated black population who had failed to compete with European immigrants for places in the labor market and were thus left with the most poorly skilled jobs. There was no specific reflection at this stage on the situation of black women. As González (2008) points out, the significance assigned to miscegenation as the central core of national identity, buttressing claims about racial democracy, served to “normalize and naturalize” the idea of Afro-descendants as sex objects (in the case of mulatto women) and domestic servants (in the case of darker women), “the necessary physical providers of pleasure, comfort and altruism” (p. 223). As a result, as Caldwell (2007) argues, Afro-Brazilian women are positioned as “the altruistic caretakers of white Brazilians, rather than full citizens and equal participants in Brazilian national culture” (p. 77). Feminists have noted that black women’s experience was marked not only by backbreaking domestic and farm labor associated with their status as slaves but also by “mandatory gender-related functions”: “mammies nursing the children of the white plantation family, the masters’ sexual satisfaction, the young boys’ sexual initiation” (Rezende and Lima 2004, p. 760). Feminist scholarship has drawn attention to how the legacy of this history can still be seen in the occupational structure of the Brazilian economy which relegates a significant proportion of paid domestic service to Indigenous and Afro-descendant women.
362
N. Kabeer and R. Santos
Documenting Intersecting Inequalities in Late Twentieth Century There is now a considerable body of studies documenting how the intersection of inequalities are reflected in different measures of poverty and well-being. Most use national survey data and hence their classifications to capture racial/ethnic identity. Since 1950, the Brazilian census has used color to capture racial/ethnic distinctions. There are five main categories: white or branco who are Euro-descendants; yellow or amarelo, who are descended from Japanese and Korean migrants); brown or pardo who dual/multiple descendants from European and African/Indigenous people; preto, the Afro-descendants; and the category indigenaI or Indigenous people only added in 1991. Because of the very small percentages of the population who are Indigenous or yellow (less than 1% in each case), most studies tend to distinguish between white (including yellow), brown and black (including Indigenous), or just white and nonwhite. We draw on some of these studies to illustrate how intersecting inequalities played out in the Brazilian context in the late twentieth century. We also pay attention where data allows to how Indigenous groups have fared. A first set of findings confirm the marked racial dimension to poverty and inequality. Afro- descendants made up approximately 45% of the Brazilian population in 2000: Between 1995 and 2000, they earned around half of the average income of the white population, a proportion that remained stable over this period. A comparison of the racial composition of income deciles showed that it became increasingly “blacker” with the move from higher to lower income deciles. Black households made up 20% in the highest income decile and increased systematically as income declined so that they made up 70% in the lowest income decile (Marió and Woolcock 2004, p. 16). Racial inequalities were evident in health and education. White life expectancy at birth in 1950 was 47.5 compared to black life expectancy of 40.1. The gap of around 7 years remained the same in 1980 when life expectancies had risen to 66.1 and 59.4, respectively (Lovell and Wood 1998). These differences in life expectancy declined with income level – from 8.59 at the lower end of the income distribution to 2.28 at the highest level. Blacks also lagged behind in education: Primary school enrollment rate for whites was 0.93 and for blacks 0.87 in 1995, but there had been considerable closing of disparities at the level of schooling by 2002 with rates of 0.97 for white and 0.95 for blacks. A second set of findings attested to the overlap between the racial and spatial distribution of poverty. It noted that eight of the ten poorest states were in the north east, which had the highest concentrations of Afro-descendants, while three of the four states in the south east were among the five richest in the country. Nearly 50% of the poor live in the Northeast region although they represent only 30% of the total population. The racial differences in life expectancy at birth noted earlier varied by region. In addition, while the relationship between racial differentials and income levels held across the northern and southern regions of the country, and overall differentials were larger in the poorer north, the relationship was also flatter: It was 7.10 years among high-income groups in the north compared to 8.41 in the south.
15
Tackling Intersecting Inequalities: Insights from Brazil
363
There was also a rural-urban dimension to poverty. Nearly 55% of the poor live in rural or small urban areas although they account for 35% of the population. Among the poor, the majority are found among farm households located in remote, isolated, sparsely populated, and low productivity areas and largely depend on farming and agricultural labor. They are mainly Indigenous (Mario and Woolcock 2004). A third set of studies documented the intersection of inequalities in the labor market. Lovell and Wood (1998) compared the situation in 1960 and 1980. They found an occupational structure that was highly stratified by race, class, and gender, not only both between better-paid white collar and less well-paid blue collar jobs but also within these broad categories. In 1960, 88% of black women were in blue collar jobs compared to 35% of white women. Most of the black women were in the least well-paid and lowest status unskilled manual and personal service categories (74%) and mainly in domestic service (68%). White women in white collar jobs were evenly distributed between professional/technical and clerical jobs; barely any had managerial positions. In 1980, 66% of black women were still in blue collar jobs but while 44% were in unskilled/personal services, the rest had moved into skilled manual labor. The percentage of black women in white collar jobs had increased from 12% to 34%, mainly in professional/technical and clerical. White women had also increased their share of white collar jobs from 48% to 63% but had not made it into managerial/ administrative positions. As far as men were concerned, 82% of black and 62% of white were in blue collar jobs but with more black men in skilled manual work than white (71% and 50%). Similar percentages of both were in unskilled services. Total 37% of white men and only 17% of black men were in white collar jobs, but only white men had managerial positions. By 1980, 46% of white men and 25% of black men were in white collar jobs, including 9.6% and 3.4% in managerial/administrative jobs, respectively. Of the rest in blue collar jobs, the vast majority in both ethno-racial categories were in skilled manual jobs. The occupational gains reported by both men and women in the two groups reflected the expansion of the economy between 1960 and 1980 and the accompanying urbanization of the work force. What can be seen is that, despite expansion of education and opportunities, the black population remained concentrated at the lower ranks of the occupational hierarchy, with black women in particular concentrated in domestic service. Comparison of average monthly wages in 1980 established that white collar jobs paid considerably higher wages than blue collar ones and that unskilled/personal services were the least well-paid of all occupations. Furthermore, white men and women earned more than black men and women, respectively, in each occupational category while black women earned less than black men in each of these occupational categories. For example, focusing on unskilled/personal services, the lowest paid occupational category, white men earned 8917 real in 1980, black men earned 6892 real, and white women earned 4165 real while black women earned 3715 real. Later studies testified to the persistence of the race/gender hierarchy in the labor market. Analysis of 1996 PNAD data showed that while 24.2% of white men and
364
N. Kabeer and R. Santos
34.5% of black men earned less than one minimum monthly wage, proportions increased to 64.6% of white women and 68.7% of black women (Mario and Woolcock 2004). Data on returns to education showed that while earnings rose with education in both 1992 and 2000, it rose more steeply for white men than other groups. For each level of education, the highest earnings were reported by white men and the lowest by black women. Earnings were very similar for white women and black men at lower levels of education but were lower for white women than black women for 11 years of schooling or more. Rezende and Lima (2004) cite work by Telles (1990) who examined the importance of education, gender, age, race, and migratory status for the occupational distribution of workers in the greater metropolitan areas of Brazil and found that gender explained the greatest share of variation in the composition of the informal market. Black women were especially prone to having jobs in the informal sector, even more than would be expected based on the single effects of race or gender, indicating the existence of particularly severe discrimination against nonwhite women. To sum up, studies of labor market inequalities have consistently reported white men as at the top of the occupational hierarchy and black women at the bottom. Black men, on the other hand, often appear to do better than white women. What these studies also confirmed was the continued concentration of black women in domestic work (Lima 2001). As analysis of the 1998 PNAD reports, 36.8% of black women and 27.4% of brown women were still in domestic work. If the “personal services category” was added, then 47.8% of black women in urban Brazil in 1998 were concentrated in these two occupational categories, with the worst conditions in terms of income and working conditions (Rezende and Lima 2004). The data cited so far attest to the significance of intersecting inequalities in determining how different gender/race groups are positioned in vertical hierarchies of income, life expectancy, education, and wages. But before we conclude this section, there is one other point worth noting. There is empirical evidence to support at least some of the claims made in relation to racial democracy in Brazil, leading a number of authors to refer to “Brazil-style racism” (Mario and Woolcock 2004) to differentiate it from the more clear-cut racism of the USA and South Africa, the two countries most often used as comparators. Three examples of this are discussed by Telles (2004). The first relates to ambiguity around racial classification. As we noted, official statistics classify the population by color, using five color categories and relying on self-identification by respondents. Studies have suggested that the population itself has a much richer vocabulary to capture various gradations in color than these five. One survey in 1976, for instance, which asked an open-ended question about color, recorded over 100 terms self-describing color. However, a reanalysis of the data suggested that, in fact, over 95% of the respondents opted for one of six terms (variations on black, brown, and white) with those using other terms (such as purple, dark chocolate, and Pele-colored) being just three or four people. More indicative of the fluidity around racial classification are some of the discrepancies observed by studies. One example of this relates to discrepancies in
15
Tackling Intersecting Inequalities: Insights from Brazil
365
how people choose to classify themselves. Self-classification as well as classification by others will often reflect some combination skin color, appearance, and social status so that that individuals with the same skin tone may classify themselves and be classified differently according to social status. Also the same person may classify themselves differently over time: Changes in self-reclassification between censuses accounted for some of the changes in the racial composition of the population documented at the national level (Lovell and Wood 1998). Telles (2004) notes the particular ambiguity surrounding the “brown” category who, along with the official label of pardo, are also referred to as moreno. In his analysis of 1995 data, he found that those who had classified themselves as moreno in response to the open-ended question distributed themselves between white, brown, and black categories when asked to use the census categories. There was also considerable discrepancy between self and interviewer identification using the census categories. There appeared to be least ambiguity for those with higher education and at the lighter end of the color scale. For instance, there was nearperfect consistency (98%) between interviewer classification among highly educated men and women in the South who classified themselves as white, and only slightly lower for this group in the north (93–94%). At the other end of the color continuum, blacks were more likely to encounter discrepancies than either whites or browns for both higher and less educated groups, but ambiguities appeared to be largest for more educated black women, particularly in Bahia where there was only 22% likelihood that interviewer would classify highly educated women as black. A second form of evidence that appeared to support the idea of a racial democracy related to residential segregation. We have already noted the existence of a spatial geography of race which creates concentrations of whites in the more affluent southern states and of blacks in the poorer north and northeast. However, there has been considerable racial convergence in urban areas: While 51% of white, 42% of blacks, and 37% of browns lived in urban areas in 1960, this had increased to 84%, 80% and 74%, respectively, by 1990. Examining residential patterns in two large urban areas in the South, Telles found considerable evidence of overlapping racial and class segregation: Whites were concentrated in the most affluent central districts while nonwhites were found in the poorer outer districts. However, a more complicated picture emerged at neighborhood level. Here Telles used a dissimilarity index (DI) which measured evenness in the distribution of racial and household income groups across a given neighborhood and an exposure index (EI) which measured the extent to which members of a certain social group are exposed to those of another group by virtue of living in the same neighborhood. The two are not the same: It is possible for whites to be more exposed to blacks in neighborhoods where blacks are a majority, but this has no bearing on the dissimilarity index. He found that the DI for the 10 largest metropolitan areas in Brazil varied from a low of 37 to a high of 48. These are very much lower than estimates from the 10 largest metropolitan areas in the USA which varied from 75 to 93. The EI index ranged from 12 to 59 in Brazil while it varied from 4 to 12 in the USA. In other
366
N. Kabeer and R. Santos
words, white Brazilians were far more likely to live in racially mixed neighborhoods than are US whites and far more likely to be exposed to black people. Residential segregation by race appeared to increase by income, suggesting that higher-income whites were less likely to live in racially mixed neighborhoods. This probably reflected the greater ability of wealthier whites to use the formal housing market to select residence by color than poorer whites who gained access to housing through informal markets. In other words, the lower levels of racial segregation in poorer neighborhoods is less likely to reflect greater tolerance among the poor as it is the precarious housing situation of those “who have little control or concern over the color of their neighbors.” Following on from Brazil’s lower levels of segregation is a third example of what makes race in Brazil different which is the “relatively high levels of racial interaction, including interracial friendship and intermarriage, at least among the poor” (p. 213). This was evident in the data on interracial marriage. Data from 1991 suggested that 23% of marriages in Brazil were between different colors, a much higher percentage than found in other racially structured societies like USA and South Africa. While this is indicative of fairly widespread interracial sociability, it is also the case that intermarriage by whites is mainly with mulattoes. Despite the fact that brown and black people are closer to each other in the vertical distribution of opportunities, there is greater social acceptance of marriage between whites and mulatto. This appears to reflect the preference for lighter rather than darker skin color in the marriage market in Brazil. An important factor in interracial marriage in Brazil is interracial propinquity. As a result, intermarriage occurs mostly in the northeast, where whites are more likely to interact with browns and blacks. Similarly, it is most prevalent among poor whites who are also most likely to be exposed to nonwhites. It is less common among middle-class whites.
Trends in Poverty and Inequality 2002–2013 Bearing in mind these intersecting patterns of inequality in the closing decades of the twentieth century, we turn our attention to the early decades of the twenty-first. As we pointed out in the introduction, this chapter is motivated by the apparent break with history, evidence of declining inequality since 2000 in a country that has historically been characterized by high, and remarkably stable, measures of income inequality. This section of the report uses PNAD data to unpack some of the components that make up this historical reversal and to explore in greater detail how groups that had hitherto been left behind in the country’s progress have fared in an era of declining income inequality. At the start of the period under analysis, Brazil’s population was estimated by PNAD to be 175 million people (of which 51.2% were female) growing to 201 million people (51.4% female) by 2013. As Table 1 indicates, the Brazilian population is largely made up of “whites” and “mixed” groups (53% and 40.5%, respectively, in 2002). “Blacks” made up 5.6%. The share of those classified as “yellow” and “indigenous” are extremely small
Brazil North Northeast Southeast South Center-west Federal District
White 2002 53.2% 27.6% 30.2% 63.2% 82.7% 44.7% 44.1%
2013 46.3% 22.6% 27.4% 54.3% 76.4% 39.5% 44.1%
Yellow 2002 0.4% 0.2% 0.2% 0.6% 0.4% 0.5% 0.4%
Table 1 Brazilian demographics – ethno-racial groups 2013 0.5% 0.3% 0.2% 0.6% 0.5% 0.5% 0.6%
Black 2002 5.6% 4.7% 5.5% 6.7% 3.7% 3.9% 5.8% 2013 8.0% 7.4% 9.6% 8.6% 4.0% 6.6% 7.6%
Mixed 2002 40.5% 67.3% 63.8% 29.4% 13.1% 50.2% 49.3%
2013 45.0% 68.1% 62.4% 36.4% 18.9% 53.1% 47.4%
Indigenous 2002 0.2% 0.2% 0.2% 0.1% 0.2% 0.6% 0.3%
2013 0.3% 1.5% 0.4% 0.1% 0.2% 0.2% 0.3%
15 Tackling Intersecting Inequalities: Insights from Brazil 367
368
N. Kabeer and R. Santos
(0.4% and 0.3%, respectively). During the period under study, there is evidence of a reduction in share of the white population at the national level (46.3% in 2013) and in the different regions (with the exception of the country’s administrative center, the Federal District of Brasilia, where it remained unchanged). This was accompanied by an increase in the share of blacks (8%) and mixed groups (45%). We also note that population was and remained more white/yellow in the south and southeast regions over this period with greater evidence of mixed groups/Indigenous groups in the north and northeast. Blacks appear to be more geographically dispersed although they made up 9.6% of the northeast by 2013. A preliminary analysis of socioeconomic indicators led us to group the white and yellow populations together and those classified as black and mixed together. “Indigenous” Brazilians are usually grouped together with blacks, but they show significantly different values for their socioeconomic indicators, suggesting that they should be object of specific attention. While representing only 0.3% of the Brazilian population in 2013, they do represent 689,000 people, larger than the populations of several countries in the world.
Income Inequalities We start our discussion with income inequality. Table 2 reports on three measures in income inequality at the national level: • The better known Gini Index • The Palma Index (ratio of the 10% highest income bracket of the population’s share of total reported income divided by the poorest 40% share) • Theil’s Generalized Entropy GE(2) measure (additively decomposes total inequality into between group inequality and within group inequality) All three measures confirm there has been a significant decline in income inequality. Figure 1 disaggregates this overall picture by income group. It suggests that while average income rose by 9.5%/annum between 2002 and 2013, the rate of growth was not evenly distributed across the income distribution. It was lower for the bottom 10% of the income distribution, much higher for those in the 10–30% and 30–50% income groups, and then declined steadily at higher levels of income. While this picture is consistent with the picture painted by the vertical inequality indicators, it does suggest that the lowest income group did not benefit from the improvements in income experienced by the poorer half of the population.
Table 2 Indicators of income inequality
Vertical inequality Palma Index Gini Index GE (2)
Income 2002 47.9 0.731 3.237
2013 36.0 0.652 2.191
15
Tackling Intersecting Inequalities: Insights from Brazil
369
Fig. 1 Average yearly growth in income
Fig. 2 Changes in average monthly income by gender, ethno-racial group and region
Figure 2 examines how the rates of change in average monthly income varied by gender, race, and region. It suggests that average monthly income grew faster for women than for men by 1.7 percentage points but that the absolute difference in
370
N. Kabeer and R. Santos
incomes also grew – from around 300 BRL a month in 2002 to around 500 BRL in 2013. A broadly similar pattern is evident for the two major ethno-racial categories: Average monthly income grew faster for the poorer black population than the higherincome white although once again, absolute differences in income grew from 345 BRL in 2002 to 643 BRL in 2013. Growth rates were lowest for the Indigenous minority, and their income levels fell below black earnings in 2013. The regional trends are more complicated. It is clear that the Federal District reported considerably higher monthly incomes in both 2002 and 2013. Despite somewhat higher rates of growth in some of the less affluent districts, the absolute difference in monthly income in the Federal District and the North east, the poorest in both years, grew from 568 BRL in 2002 to 1422 BRL in 2013. Rates of growth were lowest in the North, the second poorest region in both years. In Table 2, we present three different indices which synthesize group inequalities (Gisselquist 2018). The Group Weighted Coefficient of Variation (GWCOV) is a common measure of regional disparities. Weighted by the population size of each group, the GWCOV is less sensitive to changes in the relative position of smaller groups, because they receive a lower weight in the sum, relative to larger groups (Mancini 2008). The second indicator is the Group Weighted Gini Index (GWGini) which, like the standard Gini Index, measures the pairwise weighted sum of differences between the average group measures for the indicator of interest, divided by the overall average. It is more sensitive to changes in relative positions around the population (or sample) mean. The third indicator is the Group Weighted Theil Index (GWTheil) which, as highlighted by Gisselquist (2018), is the most sensitive to the lower end of the distribution. The technical explanation of these indicators is provided in the Appendix. Table 3 below presents the values of the three group-based inequality indicators for three groups: gender, ethno-racial identity (Indigenous, White and Yellow, Black, and Mixed), and region (Northeast, North, Center-West, South, Southeast, and Federal District). All three indicators tell us that, along with overall declines in income inequalities between 2002 and 2013 depicted in Table 2, there have been declines in gender, ethno-racial, and regional inequalities. At the same time, they tell us that ethno-racial income inequalities were and remain larger than income inequalities by gender or region. Figure 3 gives us a more disaggregated picture of the changes in average monthly income for groups at the intersection of gender, ethno-racial identity, and region, with the different groups ranked by their income levels in 2002. A broad reading of Table 3 Indicators of group-based income inequalities Income inequalities Group weighted Coefficient of variation Gini Theil
Gender 2002 0.338 0.169 0.062
2013 0.256 0.128 0.036
Ethnic 2002 0.567 0.175 0.061
2013 0.406 0.147 0.045
Regional 2002 0.250 0.159 0.044
2013 0.229 0.134 0.034
15
Tackling Intersecting Inequalities: Insights from Brazil
371
Fig. 3 Average income by intersecting groups
the figure suggests that average income increased over this period for each of these groups with largely similar ranking in both years, with black/brown and Indigenous men and women at the lower end of the income distribution and white/yellow males at the higher end. While there are some exceptions in the case of a number of nonrepresentative groups (such as Indigenous women resident in the Center-West region and Indigenous men resident in the Federal district), it is worth noting that black/brown women from the Northeast earned the lowest monthly income in both 2002 and 2013 while white/yellow men from the Federal District earned the highest. Thus, while group-level indices suggest that inequalities are higher by ethno-racial identity than by gender, the intersection of race and gender creates what appears to be a fairly consistent income hierarchy by gender and race. The other point to note is the relative flattening of the income curve – the average income gradient between groups reduced from 7.9% in 2002 to 6.6% in 2013, despite an increase in the absolute difference between lowest and highest average group income. Figure 4 offers further insight into the hierarchies defined by intersecting inequalities. Using the idea of the Palma Index, it displays the deviation of the share of the population and of income of the different intersecting groups for 2002 and 2013. The deviation is positive for groups whose share of income exceeded their share of total population and negative for those whose share of income was lower than their share of the population. Positive deviations are largest for white/yellow men from the southeast followed by those from the south and the center-west. Negative deviations are largest for black/brown women from the northeast, followed by black/brown women from the south and black/brown men from the northeast. The shaded area shows all cases where the absolute deviation declined, in other words, the cases
372
N. Kabeer and R. Santos
Fig. 4 Deviation between share of population and share of income (2002 and 2013)
where the share of income came closer to share of population, leading to a more equitable distribution of income. The figure suggests that this was the case for many groups (53% of the groups to be exact), including the most extreme ones, although the positive deviation remains still very high for the most privileged groups (white/ yellow men residing in the Southeast). It also suggests a significant concentration of groups in a window where the difference between share of income and share of population does not exceed 1% – a suggestion that these groups, on average, are not driving group-based inequalities. In fact, the proportion of such groups increased from 75% to 81% of the total. To sum up this section, we have shown that economic growth in Brazil, from 2002 to 2013, was accompanied not only by a decrease of vertical income inequality, but also in horizontal and intersecting inequalities. There has been a decline in regional inequalities in income distribution, in gender inequalities as well as ethno-racial inequalities. However, it is important to note that the decline in ethnoracial inequalities does not include the Indigenous population whose incomes do not appear to be converging toward the average. While an examination of horizontal inequalities suggests that ethno-racial inequality is higher than gender-based inequality, reexamining income distribution from an intersectional perspective suggests women from marginalized ethno-racial groups are clustered around the lower levels of the income distribution, despite a higher proportional increase in their respective average incomes. The suggestion is, therefore, of a possible narrowing of identity-based differences in income due to a more than proportional increase of the income “floor” but not to higher socioeconomic mobility.
15
Tackling Intersecting Inequalities: Insights from Brazil
373
Poverty We turn next to the distribution of poverty across the population. Figure 5 compares BRL estimates of different poverty lines between 2002 and 2016. It shows a nearperfect coincidence between the International (Extreme) Poverty Line (IPL) set by the World Bank and Brazil’s National Extreme Poverty Line. By contrast, Brazil’s National Poverty Line (NPL) is set higher. In our analysis therefore, we will focus on extreme poverty as measured by the IPL and overall poverty as set by the NPL. Figure 6 examines the incidence of extreme and overall poverty for different income groups between 2002 and 2013. It suggests that extreme and overall poverty were concentrated in the lowest 10% of the income distribution in both 2002 and 2013 and both forms of poverty declined over this period. Total 78% of this group were below the poverty line in 2002 declining to 18% in 2013. In addition, we should note that around 37% of the population aged 10+ earned zero income in 2002, a figure that declined to 30% in 2013. Figure 7 shows that, along with the overall decline in poverty, there was a clear decline in the gender gap in the incidence of poverty. While the percentage of men below the extreme poverty line declined from 23% in 2002 to 21% in 2013, the percentage of women declined from 39.6% to 29.4%. Nevertheless, women were still 8% more likely than men to earn less than US$1.25 a day in 2013. Figure 8 reports on the incidence of poverty by ethno-racial groups in 2002 and 2013. There appears to have been a decline in the incidence of poverty for each group so that the gap between the black/brown and Indigenous groups declined vis-à-vis white/yellow groups.
Fig. 5 Poverty lines between 2002 and 2016
374
N. Kabeer and R. Santos
Fig. 6 Poverty by income groups
Fig. 7 Incidence of poverty by gender
Table 4 compares the incidence of extreme poverty between 2002 and 2013 for groups at the intersection of income, gender, and ethno-racial inequalities, focusing on those at the bottom 10% of the income distribution. It finds that there was a
15
Tackling Intersecting Inequalities: Insights from Brazil
375
Fig. 8 Incidence of poverty by ethno-racial group Table 4 Extreme poverty by gender, race and indigenous identity
Below 10% income level Indigenous male Indigenous female Black and mixed male Black and mixed female White and yellow male White and yellow female
Below extreme poverty line 2002 2013 75% 15% 70% 12% 62% 8% 61% 7% 61% 7% 58% 7%
significant decline in extreme poverty for all groups over this period, with particular large declines for Indigenous males and females. Nevertheless, Indigenous men and women continued to face the highest levels of deprivation in 2013 as they had in 2002, with incidence slightly higher among Indigenous men. The next set of estimates examines spatial dimension of poverty in Brazil. Figure 9 suggests declines in poverty in both rural and urban areas with a faster decline in the proportion of rural poor (from 37% to 28.1%) compared to the urban (from 30.5% to 24.7%). This period therefore saw a reduction in the rural-urban poverty gap. Figure 10 shows the regional distribution of poverty. It suggests that while the incidence of poverty went down in all regions, this did not necessarily reduce regional inequalities in poverty. While the largest reduction in poverty was reported by the northeast region, which had been the poorest in 2002, the second largest declines occurred in the south region which had been the richest region in 2002.
376
N. Kabeer and R. Santos
Fig. 9 Incidence of poverty by rural urban locations
Fig. 10 The incidence of poverty by region
Smaller declines in the north, which had been the second poorest region in 2002, meant that it became the poorest region in 2013. The south region was the richest region in 2002 and remained so 2013. The difference in the incidence of poverty in
15
Tackling Intersecting Inequalities: Insights from Brazil
377
the region with highest levels of extreme poverty (north in 2013) and the region with the lowest (south) thus increased in 2013. In summary, our analysis suggests that along with a reduction in absolute forms of poverty in Brazil, the poverty gap between different groups also declined. This decline was evident by gender, by ethno-racial groups, and among groups at the intersection of gender, ethno-racial inequality, and extreme poverty. There was a particularly large decline in extreme poverty among Indigenous groups. The regional distribution of poverty is more complicated. While there was a reduction in the ruralurban gap in the incidence of poverty and while poverty went down in all regions, regional inequalities in poverty may have increased. Smaller declines in poverty in the north, which had been second poorest region in 2002 (behind the northeast), made it the poorest region in 2013 and increased the gap with the richest region, the south.
Labor Market Outcomes Wage Inequalities We next turn to labor market outcomes in 2002 and 2013 and discuss wage earnings as an indicator of progress. Given that wages represented 87% of reported incomes in 2002 and 90% in 2013 according to PNAD surveys, it is not surprising that our estimates closely follow those reported for overall income. Figure 11 examines growth in wages by gender, ethno-racial identity, and region. Wages grew by an average of 9.6/annum over our study period. It grew somewhat faster for women
Fig. 11 Wage levels and growth by gender, race, indigeneity and region (2002 and 2013)
378
N. Kabeer and R. Santos
than men (10.4% and 9.3%, respectively) although women continued to earn lower wages in 2013. The table suggests some decline in ethno-racial inequalities in earnings in that growth in wages were highest for black and mixed groups (11.1%) followed by Indigenous groups (9.9%) followed closely by white and yellow groups (9.6%). Wage levels, however, were and remained considerably higher for white/yellow groups (BRL 701 and 1920 in 2002 and 2013, respectively), particularly compared to those reported by Indigenous groups (BRL 326 and 924). As far as regional disparities were concerned, the slowest rate of growth was reported by the north and the highest by the northeast, but earnings in both 2002 and 2013 were somewhat higher in the north than in the northeast which reported the lowest earnings in both years. The center and southern regions reported growth rates that varied between 8% and 10% but reported higher earnings than the northern regions in both years. The highest earnings, considerably higher than other regions, were reported by the Federal District for both years. Table 5 estimates indices of wage inequality by gender, ethno-racial groups, region, and the intersection of these. The final column needs some explanation. It calculates indicators of group-weighted inequalities for a greater disaggregation, intersecting the three identities: gender, ethnic, and regional. This allows us to highlight the differences between intersecting categories such as “white/yellow men from the south east,” “black/mixed women from the north-east,” or “indigenous women from the Centre-west.” The tables not only suggest that, as with income, ethno-racial wage inequalities are generally larger than inequalities by gender and region but also that the intersection of these inequalities is associated with the greatest wage inequalities. There has been a decline for the group and regional measures of inequality. We also find some suggestion of increasing wage inequalities, particularly when the indicator is more sensitive to the extremes of the income distribution, such as is the case of the GWTheil. We will explore this finding through a further examination of labor market dynamics.
Labor Market Segregation by Occupation and Work Status We turn now to the distribution of occupations across different groups and regions. Table 6 summarizes the average monthly wages in Brazilian Real by occupation in 2002 and 2013 ranked by average wage levels in 2013. It makes a number of points. First of all, occupations at the higher end of the wage distribution generally reported Table 5 Measures of group-based, regional, and intersecting wage inequalities Wage inequalities Group weighted Coefficient of variation Gini Theil
Gender 2002 0.221
2013 0.173
Ethnic 2002 0.527
2013 0.385
Regional 2002 2013 0.265 0.231
0.109 0.017
0.085 0.014
0.167 0.053
0.141 0.037
0.179 0.057
0.139 0.026
Intersecting 2002 2013 0.507 0.573 0.270 0.076
0.300 0.456
15
Tackling Intersecting Inequalities: Insights from Brazil
379
Table 6 Average monthly wages by occupation Occupations Public administration Other industrial activities Other activities Education, health, and social services Transport, storage, and communications Manufacturing Other collective, social, and personal services Trade and repairs Construction Underdefined activities Accommodation and food Agriculture Domestic services
2002 1033.9 970.4 1075.6 768.9 760.8 617.3 522.7 574.3 482.2 348.2 419.7 177.1 205.5
2013 2747.8 2611.6 2496.3 2285.0 1770.7 1519.8 1397.2 1393.7 1371.4 1170.0 1169.0 796.5 685.1
Δ% 9.3% 9.4% 8.0% 10.4% 8.0% 8.5% 9.3% 8.4% 10.0% 11.6% 9.8% 14.6% 11.6%
a slow rate of growth per annum than those at the lower end, leading to some convergence to the national mean. The exception to this pattern is the category “education, health and social services” which grew at 10.4% compared to the 8–9% rates reported for other occupations in the higher end of the table. Second, jobs in public administration reported higher average wages than all other occupations in both 2002 and 2013 (with the exception of “other activities in 2002”). The third point to note is that agriculture and domestic services were the lowest paid jobs in both 2002 and 2013 as they had been the case in the earlier year. Tables 7 and 8 report on the distribution of different groupings across the occupational hierarchy. Table 7 presents simple ratios to capture the class, gender, ethno-racial, and rural-urban inequalities in the distribution of occupations. Table 8 tries to capture how the intersection of these inequalities maps into the distribution of occupations. In Table 7, the ratio of the poorest 40% of the population to the richest 10% captures the class distribution across occupations. The table tells us that poorer workers are severely underrepresented at the higher paid end of the occupational hierarchy and overrepresented at the poorer paid end, particularly in agriculture and domestic services. The ratio of female to male workers suggests greater parity in gender distribution at the better paid end, with women outnumbering men in the education, health, and social services to a markedly greater extent in domestic services at the poorer end – a reminder that, despite of gender segmentation in the occupational structure, women are distributed across the class spectrum. The ratio of black, mixed, and Indigenous workers to white/yellow workers suggests some underrepresentation in better-paid occupations, parity in others, and overrepresentation in yet others. As might be expected, the rural population is underrepresented in most occupations but heavily represented in agriculture. Table 8 takes account of intersecting inequalities and offers clearer insights into the distribution of disadvantage in the labor market. It compares the ratio of men and women from different ethno-racial categories from the poorest end of the income
Public administration 02 13
Other industrial activities 02 13
Other activities 02 13
Transport, storage, and com-munications 02 13 Manufacturing 02 13
Trade and repairs 02 13 Construction 02 13
Underdefined activities 02 13
Accommodation and Agricultufood re 02 13 02 13
Domestic services 02 13
Household income percentile Bottom 40%/top 10% 0.2 0.2 0.5 0.3 0.1 0.2 0.1 0.2 0.4 0.7 0.6 0.8 0.5 0.7 0.6 0.9 2.7 2.8 6.0 .. 1.1 1.5 11.6 7.8 5.5 8.2 Gender Female/male 0.8 0.9 0.2 0.2 0.8 1.0 4.8 4.3 0.2 0.2 0.8 0.8 1.9 2.3 0.8 1.0 0.0 0.0 0.3 0.0 1.4 1.8 0.7 0.6 19.3 18.8 Rural/urban Rural/urban 0.2 0.4 0.5 0.6 0.1 0.1 0.3 0.4 0.2 0.3 0.3 0.5 0.2 0.2 0.2 0.3 0.3 0.5 0.3 1.0 0.2 0.3 9.6 12.4 0.5 0.6 Ethnic group/“white and yellow” Black and mixed 1.2 0.9 0.8 1.0 1.8 0.6 1.4 0.7 1.2 0.9 1.4 0.8 1.1 0.9 1.2 0.9 0.7 1.6 1.0 1.0 1.0 1.1 0.6 1.7 0.6 1.6 Indigenous 1.0 0.7 1.0 0.8 1.0 0.5 1.0 0.9 1.0 0.4 1.0 0.7 1.0 0.9 1.0 0.6 1.0 1.3 1.0 0.0 1.0 0.7 1.0 3.0 1.0 1.6 Intersecting inequalities – ratio of men and women from different ethno-racial groups and poorest households (10–30% in income distribution) to “white/yellow men from top 10% on income distribution” White and yellow male 0.2 0.2 0.5 0.5 0.1 0.2 0.1 0.1 0.5 0.7 0.6 0.8 0.3 0.6 0.7 1.0 2.7 3.2 10.0 2.0 0.8 1.0 9.9 6.1 4.5 5.5 White and yellow 0.2 0.3 0.1 0.1 0.1 0.1 0.4 0.6 0.0 0.1 0.6 0.9 0.8 1.7 0.5 0.9 0.0 0.1 2.0 0.0 1.1 2.9 7.8 4.2 133.0 116.0 female Black and mixed male 0.2 0.2 0.7 0.4 0.1 0.1 0.1 0.1 0.5 0.7 0.5 0.6 0.5 0.6 0.7 0.9 3.3 3.5 8.0 2.0 0.7 1.1 9.5 6.7 7.0 6.0 Black and mixed female 0.1 0.2 0.1 0.0 0.0 0.1 0.4 0.5 0.1 0.1 0.4 0.6 1.1 1.5 0.4 0.8 0.1 0.1 2.0 1.0 1.4 2.7 7.6 4.6 155.0 139.5 Indigenous male 1.3 0.1 0.0 0.0 0.7 0.1 0.0 0.8 0.6 0.6 0.0 0.1 1.3 0.0 0.5 0.5 2.1 1.4 0.0 0.0 0.0 0.0 8.8 11.3 0.0 4.5 Indigenous female 0.0 0.0 0.0 0.0 0.0 0.0 0.7 1.1 0.0 0.0 0.3 0.0 2.0 1.0 1.3 0.9 0.0 0.0 0.0 0.0 2.2 0.6 8.7 10.9 30.0 21.0
Class Poo Minminmmand in hein persBlackare fouagriclu7utrcoffRatios
Education, health, and social services 02 13
Other collective, social, and personal services 02 13
Table 7 This table has some garbled language and I am not sure how it should be captioned Am returning to co-author
380 N. Kabeer and R. Santos
Military 02 13
Employee with work license 02 13 Self employed 02 13
Employee without work license 02 13 Domestic worker with work license 02 13
Domestic worker without work license 02 13 Nonremunerated 02 13
Own production worker 02 13
Own consumption worker 02 13
Tackling Intersecting Inequalities: Insights from Brazil (continued)
Household income percentile 0.0 0.0 0.1 0.1 0.2 0.1 0.4 0.6 1.3 1.2 1.9 2.4 1.3 1.6 9.8 17.0 6.1 5.3 11.9 19.8 2.0 – Bottom 40%/top 10% Gender Female/ 0.5 0.5 1.9 2.0 0.0 0.0 0.8 0.8 0.6 0.6 0.6 0.8 10.8 11.3 24.8 21.0 1.6 2.1 3.6 1.6 0.5 0.0 male Rural/urban Rural/urban 0.4 0.4 0.2 0.5 0.0 0.3 0.3 0.3 1.3 1.4 0.9 1.2 0.5 0.5 0.5 0.7 7.7 8.8 7.2 16.2 0.0 1.0 Ethnic group/“white and yellow” Black and 0.4 0.4 0.7 0.8 1.0 1.3 0.7 0.8 1.1 1.1 1.3 1.3 1.1 1.4 1.7 1.7 1.3 1.2 1.5 2.0 2.0 1.0 mixed Indigenous 0.3 0.3 0.7 0.7 0.0 0.7 0.7 0.5 1.2 1.4 1.0 1.1 1.4 1.6 1.7 1.7 1.3 1.6 2.6 5.5 3.0 3.0 Intersecting inequality: ratio of men and women from different ethno-racial groups and poorest households (10–30% in income distribution) to “white/yellow men from top 10% on income distribution” 1.3 0.1 .. 0.0 18.3 0.5 0.8 1.5 0.1 2.3 0.1 3.0 7.2 5.0 4.0 9.3 2.9 10.5 0.0 2.0 – White and 0.0 yellow male 0.7 0.1 .. 0.0 37.0 0.2 0.5 0.7 0.0 1.1 0.2 30.0 18.5 237.0 23.0 14.9 1.8 33.8 0.0 0.0 – White and 0.0 yellow female 0.0 1.2 0.1 .. 0.1 26.0 0.5 0.7 1.3 0.1 2.8 0.1 4.0 7.5 10.0 4.0 8.6 3.3 9.0 0.0 3.0 –
Ratios
Employer 02 13
Statutory public worker 02 13
Table 8 Distribution of work status by class, gender, ethno-racial identity and region (2002 and 2013)
15 381
Black and mixed male 0.0 Black and mixed female Indigenous 0.0 male Indigenous 0.0 female
Ratios
0.2
1.3
0.7
2.1
0.5
..
..
..
Statutory public worker 02 13
0.8
Employer 02 13
Table 8 (continued)
0.0
0.0
0.0
111.7 0.0
44.7 0.5
48.3 0.2
Military 02 13
0.0
0.2
0.4
Employee with work license 02 13
1.4
1.4
0.8
0.0
0.0
0.0
Self employed 02 13
1.9
1.1
1.0
0.0
0.3
0.2
Employee without work license 02 13
0.0
0.0
31.0
45.3
23.5
13.7
Domestic worker with work license 02 13
60.0
0.0
280.0
0.0
0.0
28.0
Domestic worker without work license 02 13
7.1
14.6
11.3
2.4
1.6
1.8
Nonremunerated 02 13
64.5
11.3
37.0
0.0
0.0
0.0
Own production worker 02 13
0.0
0.0
1.0
–
–
–
Own consumption worker 02 13
382 N. Kabeer and R. Santos
15
Tackling Intersecting Inequalities: Insights from Brazil
383
distribution (10–30%) to men from the privileged white/yellow ethno-racial categories from the top 10% of the income distribution. Predictably, both men and women from the poorer end of the income distribution are severely underrepresented in better-paid occupations (such as public administration and health, education, and social services which are dominated by wealthy men from the privileged ethno-racial groups). There is, however, some variation in the occupational distribution among poorer men and women from different ethno-racial categories. Poorer men from all ethno-racial categories are overrepresented in construction compared to privileged white/yellow men while poorer women are underrepresented. Poorer white/yellow and black/mixed men are overrepresented in the “undefined activities.” All the poorer groups are overrepresented in agriculture relative to privileged white/yellow men and in domestic services, but poorer women from white/yellow and black/ brown categories are heavily overrepresented in this category. Some of the changes between 2002 and 2013 are worth noting: some increase in representation of poorer yellow/white and black/mixed women in “other social services” (but a decline among Indigenous women); some increase in representation of poorer white/yellow male in construction (but a decline among Indigenous men); a large decline in representation of white/yellow as well as black/mixed men in undefined activities along with a smaller decline among equivalent groups of women; an increase in representation of most groups in accommodation/food; a decline in representation of most groups in agriculture with the exception of Indigenous men and women who experienced an increase; and finally a major decline in representation of women from white/yellow and black/mixed groups in domestic service accompanied by a smaller decline among Indigenous women who in any case are not as strongly overrepresented in this occupation. There was some increase in this category of men from white/yellow and Indigenous groups, but they are also not strongly overrepresented. Indigenous women are absent from most occupations with the exception of other social services, trade and repairs, agriculture, and domestic services. To sum up, there has been some convergence in earnings across different occupational categories over time. At the same time, income hierarchies and gender segmentation remain intact in the labor market. Poorer people are concentrated in lower paid occupations while the wealthy are found in the better paid ones. Women are crowded into a limited range of occupations, outnumbering men in education, health, and social services at the better paid end of the hierarchy and markedly more so in domestic services at the poorer paid end. There is less evidence of segmentation along ethno-racial lines with greater evidence of parity between white/yellow and black/mixed. Indigenous groups, on the other hand, tend to be overrepresented in agriculture and increasingly so over time.
Inequalities by Earnings and Work Status Table 9 continues the discussion of inequalities in labor market outcomes, this time ranking work status according to earnings. This ranking did not change between 2002 and 2013. Employers, public sector workers, and military personnel are all higher paid work statuses, earning between 2658 and 5051 Brazilian real a month in
384
N. Kabeer and R. Santos
Table 9 Work status by earnings (2002 and 2013) Work status Employer Statutory public worker Military Employee with work license Self-employed Other employee without work license Domestic worker with work license Domestic worker without work license Nonremunerated Own production worker Own consumption worker
2002 1979.2 1039.4 1150.0 663.9 494.1 363.8 289.7 175.5 4.0 1.2 0.8
2013 5051.5 2686.8 2657.9 1553.5 1373.0 1050.6 868.4 533.1 5.2 2.0 0.0
Δ% 8.9% 9.0% 7.9% 8.0% 9.7% 10.1% 10.5% 10.6% 2.4% 4.8% 100.0%
2013. Next in rank were wage workers with licenses, the self-employed followed by wage workers without licenses. The poorest among the wage earners are domestic workers with or without licenses (earning 868 and 553 real, respectively) while there are further categories of unpaid, own-account, and subsistence workers who barely earn any income. The highest rates of growth in earnings were reported by domestic workers and wage workers without licenses, but, as noted, this did not serve to alter the ranking of work status by earning. Table 9 examines how different social groups are distributed across these work statuses, following the same breakdown as before. As before, Table 7 reports on simple inequalities while Table 8 examines intersecting inequalities. As expected, the ratio of poorer workers to the wealthiest is well below parity in the betterremunerated work status and above parity in the more poorly remunerated statuses, mainly the self-employed and domestic service. The distribution by gender is less clearly associated with rates of remuneration. Women make up twice as many statutory public workers as men and are nearly on par with men regarding the proportion of employees with work licenses. However, the ratio of women to men is 10.8 rising to 11.3 among domestic workers with a license and 24.8 declining to 21.0 among domestic workers without a license. They also proportionally outnumber men among the unpaid family workers and own production workers. As far as the distribution of ethno-racial categories is concerned, the results suggest that white/yellow workers dominate in the better paid categories of employers, somewhat less so as public sector workers and employees with work licenses. They are less likely than other groups to be found in poorly paid categories of employees without work licenses and domestic workers. While black/mixed workers outnumber other groups in the military, both black/ mixed and Indigenous workers outnumber white/yellow workers among workers without licenses. Indigenous workers outnumber the rest among own-production and own-consumption work which are effectively unremunerated. This appears to have increased over time. Rural workers outnumber urban workers in various forms of self-employment and nonremunerated activity – this concentration has increased over time.
15
Tackling Intersecting Inequalities: Insights from Brazil
385
Turning to workers at the intersection of economic, gender, and ethno-racial categories, we find some evidence of improvement in poorer sections of the working population. There appears to have been a move of poorest men and women from different ethno-racial categories out of self-employed status into employer status between 2002 and 2013. This suggests an improvement in returns to their enterprise efforts. There has also been a considerable increase in numbers from all these categories entering the military, some increase in employees with work licenses (except for Indigenous males), and a considerable decline in employees without work licenses for all categories. Of particular interest is the changing composition of domestic work with and without licenses. There seems to have been a rise in the ratio of men in domestic worker with licenses from all ethno-racial categories and of Indigenous women to the reference group and a substantial decline in the ratio of women from poorer white/yellow and black/mixed groups. In addition, there has been a dramatic decline in ratios of women from different ethno-racial categories in domestic work without license. This would suggest that men from poorer households are more likely to enter licensed domestic work while women from poorer households are moving out of domestic work without licenses. Referring back to Table 9, we can see that domestic and other workers with license earn considerable higher wages than those without licenses, confirming that formalization of some of the activities in which poor women and men were involved has contributed to the reduction of group-based and intersecting inequalities. Nevertheless, even with these improvements, it appears that Indigenous women and the rural poor continue to find it difficult to improve their market opportunities. Overall, therefore, as far as work status is concerned, there were clear signals of improvement, both with a convergence of income among market occupations, particularly in relation not only to domestic work but also to informal labor. While women are persistently overrepresented in both “out-of-market” activities and the least paid occupation of domestic workers, they moved into better protected and better paid formal categories of domestic work. Ethno-racial market segmentation, with “white/yellow” men overrepresented in the formal market work statuses and most notably as employers, appears to persist. Among the poorest, the main, and positive, sign of change is one of formalization of the work status, to the benefit of those most disadvantaged by intersecting inequalities.
Access to, and Control Over, Land We turn next to an issue that is likely to have particular relevance to the rural poor, the right to land. The PNAD defined five types of land access: ownership, tenancy, assignee (occupation/use of land with owners’ consent but no share of proceeds), squatter (occupation/use of land without owner’s permission), and partner (authorized occupation/use in return for retribution through share of production). Both squatter and assignee status are regarded as precarious forms of access. Table 10 reports on average nonlabor income associated with these different forms of access. It represents the difference between “income from all sources” and “income from all jobs.” It includes income from not only nonlabor sources, such
386 Table 10 Nonlabor income in BRL
N. Kabeer and R. Santos
Rights to land Owner Tenant Assignee Squatter Partner
2002 109.84 51.00 29.47 50.40 42.88
2013 315.78 222.84 155.17 154.61 117.38
Δ% 38.3% 26.2% 22.1% 36.1% 40.0%
as rents, profits, and interests, but also social transfers. Although full attribution is not possible, it confirms that a relationship between levels of income and security of access. Table 11 examines how access and ownership of land is distributed and how this distribution has changed over time. As far as vertical inequalities are concerned, we can see that poorer households were less likely than richer ones to report ownership of land, somewhat more likely to report tenancy and partner status and considerably more likely to report assignee and squatter status. The table also suggests some increase in ownership among the poor relative to the rich but a decline in all other forms of access. As far as gender is concerned, women appeared almost as likely as men to report land ownership with little change over time. This is remarkable, given widespread gender inequalities in land distribution documented in the international literature. What the table does not tell us is the quantity and quality of land owned by men and women. Women are considerably less likely to be tenants, squatters, and partners but more likely to be assignees. Over time, the likelihood of being tenants, partners, and squatters has increased but the likelihood of assignment declines. Ownership of land does not differ much by rural/urban location though it increased slightly over time. Tenancy and partnership do not appear common in rural areas, and there was a decline in most forms of access. Ethno-racial identities appear to have a bearing on land access. Relative to both white/yellow and black/mixed categories, Indigenous groups appear more likely to report land ownership though this likelihood declined over time. This explains their concentration in agricultural occupations. They were less likely to have tenancy arrangements. Both tenancy status and squatting increased somewhat over time, but the likelihood of being assignees and partners declined. Combining the different forms of inequality, the key findings are that relative to the most privileged groups in the population, only Indigenous men and women were more likely to report landholdings, but this declined by 2013. They were far less likely to be assignees squatters, but many more Indigenous people reported this practice in 2013. Squatting, assignation, and partnerships were more likely to be reported by other groups, but there appeared to have been a decline in most of these practices for most groups. Overall, there is some suggestion of increased formalization of relations regarding rights to land. This seems not only to suggest an easier access to the ownership of land by previously relatively excluded “black and mixed” Brazilians, but also to have been accompanied by loss of ownership among Indigenous groups. Land formalization may have encroached on their ancestral rights. There is a clear suggestion of a persistent discrimination relating to access to land via tenancy agreements, both against women and black, mixed, and Indigenous Brazilians.
Owner Tenant Assignee Squatter Partner Ratios 2002 2013 2002 2013 2002 2013 2002 2013 2002 Household income percentile Bottom 40%/top 10% 0.63 0.83 1.09 0.66 25.38 2.80 18.33 4.89 4.25 Gender Female/male 0.97 0.96 0.46 0.75 1.81 1.45 0.88 1.41 0.62 Rural/urban Rural/urban 1.01 1.09 0.61 0.50 1.50 1.22 1.00 0.85 0.85 Ethnic group/“white and yellow” Black and mixed 0.78 0.88 1.12 0.79 1.97 1.93 5.23 2.72 1.06 Indigenous 1.13 1.06 0.22 0.34 1.21 0.98 0.00 3.50 0.44 Intersecting – group within the 10–30%/capita hh income bracket/“white and yellow male top 10% per capita household income” White and yellow male 0.74 0.82 0.82 0.93 15.10 8.11 3.67 0.00 4.17 White and yellow female 0.39 0.92 0.88 0.00 49.70 21.56 7.33 0.00 2.08 Black and mixed male 0.59 0.81 1.51 0.49 18.80 9.00 27.00 19.00 3.63 Black and mixed female 0.58 0.70 0.42 0.70 31.70 11.11 26.67 27.33 1.83 Indigenous male 1.13 1.04 0.00 0.00 0.00 0.00 0.00 31.00 0.00 Indigenous female 1.13 0.84 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Table 11 Distribution of access and ownership of land by vertical, horizontal, and intersecting inequalities
3.83 0.00 3.46 1.67 0.00 11.00
1.40 0.15
0.87
0.78
2.62
2013
15 Tackling Intersecting Inequalities: Insights from Brazil 387
388
N. Kabeer and R. Santos
This is shown even more clearly among the poorest, with poorer white men relatively better off than all other groups. The apparent move toward more formalized rights to land does not therefore appear to have reduced horizontal and intersecting inequalities in access to land.
Educational Outcomes We conclude our analysis of the PNAD data by examining the distribution of education over time. Education has been singled out as one of the key factors driving the decline in income inequality in Brazil and elsewhere in Latin America. Table 12 reports on a number of summary measures of the distribution of education in 2002 and 2013. It suggests a clear decline in the vertical inequalities in education. We explore this decline in greater depth by disaggregating for different groups and using indicators such as the percentage of children in preschool, the participation rate in formal education of children at 6, and percentage of the population of primary school age (age 6–10) achieving literacy (Fig. 12).
Table 12 Summary measures of educational distribution (2002 and 2013)
Vertical inequality Gini Index Generalized Entropy 2
Fig. 12 Enrolment and literacy by age and household income
Years of education 2002 0.485 0.368
2013 0.423 0.276
15
Tackling Intersecting Inequalities: Insights from Brazil
389
Fig. 13 Enrolment and literacy by age and gender
Figure 13 shows these indicators disaggregated by household income for 2002 and 2013. It is immediately evident that there has been significant overall improvement. The income gradients have declined for all the indicators, most noticeably for the enrollment of not only 6-year olds but also of under-5 in preschool. The effect on the literacy performance of primary school children is also evident, with improvements of nearly 20% in the prevalence of literate children (among the poorest), bridging the income-led gap in education achievement. Figure 14 shows gender equality in under-5 outcomes for both 2002 and 2013, a small advantage for girls in 6-year olds enrolled in school for both years and a somewhat larger advantage for girls in literacy performance in 2002 which persisted in reduced form in 2013. Figure 14 suggests that regional inequalities persist. While there appears to be an increase in all three indicators of educational progress, the north region was lagging in both years of the study. The consistent underperformance and lack of significant improvement in the north region leads to an actual increase in the gap between the enrollment in preschool and primary school and in the prevalence of literacy among primary school age children in relation to the best-off regions of the south and southeast. Figure 15 examines educational inequalities by ethno-racial groups. There is an overall improvement for all three indicators across the three groups. While this seems to have been accompanied by a reduction in the educational gap between yellow/white and black/mixed groups, the gap among children from Indigenous groups seems to have persisted in 2013 with some convergence only observed in relation to children aged 6 who were enrolled in school.
390
N. Kabeer and R. Santos
Fig. 14 Enrolment and literacy by age and region
Fig. 15 Enrolment and literacy by age and ‘racial’ attribution
Table 13 allows us to explore how intersecting inequalities have fared over our study period. It shows clear signs of improvement in both preschool and primary school enrollment with some convergence toward the levels prevailing among boys
15
Tackling Intersecting Inequalities: Insights from Brazil
391
Table 13 School attendance (under 5 and 6 year olds) and literacy (age 6-10) by class, gender, ethno-racial identity and region Percentage in school (by age) Literate (%) Under 5 Age 6 Age 6–10 Ratios 2002 2013 2002 2013 2002 2013 Household income percentile Bottom 40%/top 10% 0.27 0.46 0.85 0.97 0.66 0.80 Gender Female/male 1.07 0.99 1.02 1.02 1.08 1.04 Region/Sul Norte 0.81 0.49 1.01 0.97 0.84 0.83 Nordeste 1.11 0.83 1.05 1.04 0.71 0.85 Sudeste 1.16 0.99 1.08 1.04 0.95 1.01 Centro-Oeste 0.66 0.65 1.00 1.01 0.96 1.00 Distrito Federal 1.24 0.92 1.07 1.03 0.93 1.00 Ethnic group/ “white and yellow” Black and mixed 0.85 0.87 0.94 1.00 0.82 0.91 Indigenous 0.22 0.32 0.49 0.90 0.69 0.74 Intersecting inequalities: ratio of girls and boys from ethno-racial groups in 10–30% of income distribution to boys from white/yellow group in the top 10% of income distribution White/yellow boys 0.24 0.70 0.82 0.90 0.72 0.74 White/yellow girls 0.28 0.77 0.86 0.83 0.75 0.72 Black/mixed boys 0.28 0.74 0.84 0.89 0.60 0.61 Black/mixed girls 0.28 0.80 0.85 0.89 0.69 0.67 Indigenous boys 0.51 0.86 .. 0.71 1.14 1.11 Indigenous girls 0.00 0.73 0.00 0.87 0.91 0.43
from the richest and most privileged ethno-racial group. At the same time, children from Indigenous groups continue to perform worse than others, with Indigenous boys at some disadvantage among 6-year olds enrolled for school and Indigenous girls at a particular disadvantage in terms of literacy achievements among 6–10-year olds. Tables 14 and 15 draw attention to changing trends in education among the working-age population 15–49. A comparison of the younger age groups in the 2 years allows us to see how different groups who will have benefited from efforts to increase education by the government of Lula di Silva (2003–2010) have fared. This becomes particularly apparent in the 20–29 age bracket, the first cohort where most of the members have already completed their education. The tables suggest that years of education (excluding tertiary education), and the proportion of individuals with tertiary education, increased for all age groups in Brazil, and for all social groups, including groups at intersection of different inequalities. The evolution along gender lines is worth noting: While the first indicator suggests that men’s years of education increased more than women’s, particularly within the black/mixed ethnoracial group, the increase of women’s progression to tertiary education exceeded that of men in almost all groups, except for Indigenous people in the 20–29 age bracket. Women already had more years of education than men – while the gap may have reduced among those without tertiary education (particularly relevant here is the
Brazil Male 2002 2013 2002 2013 Years of school, population without tertiary education Age 15–19 7.2 8.3 6.8 8.0 Age 20–29 7.6 9.4 7.3 9.1 Age 30–39 6.5 8.1 6.3 7.8 Age 40–49 5.7 7.1 5.7 6.9 % with tertiary education Age 20–29 4.6% 10.0% 3.8% 8.2% Age 30–39 8.0% 14.3% 7.0% 12.2% Age 40–49 8.9% 13.1% 8.6% 11.0% 8.6 9.8 8.4 7.4 11.8% 16.3% 15.1%
5.4% 8.9% 9.2%
2013
7.5 8.0 6.7 5.8
Female 2002
2.2% 3.1% 1.3%
6.1 6.3 6.2 4.9 7.0% 7.6% 8.9%
6.5 8.1 7.3 5.8
Indigenous 2002 2013
Table 14 Population with and without tertiary education by gender and race (2002 and 2013)
7.5% 11.8% 13.2%
7.9 8.5 7.3 6.6
15.5% 22.0% 19.0%
8.8 10.1 8.9 7.9
White and yellow 2002 2013
1.5% 3.3% 3.3%
6.5 6.8 5.7 4.8
5.6% 7.8% 7.7%
7.9 9.0 7.6 6.5
Black and mixed 2002 2013
392 N. Kabeer and R. Santos
Indigenous Male Female 2002 2013 2002 2013 Years of school, population without tertiary education Age 15–19 5.5 6.0 6.5 7.0 Age 20–29 5.5 7.8 6.9 8.4 Age 30–39 5.9 7.1 6.5 7.4 Age 40–49 4.4 6.3 5.3 5.3 % with tertiary education Age 20–29 0.0% 5.9% 3.8% 7.9% Age 30–39 2.4% 5.6% 3.6% 9.3% Age 40–49 2.8% 5.4% 0.0% 12.5% 8.5 9.8 8.6 7.7 13.1% 19.5% 16.4%
7.7 8.2 7.1 6.6 6.4% 10.6% 13.0%
White and yellow Male 2002 2013
8.5% 13.0% 13.3%
8.2 8.7 7.5 6.5
Female 2002
Table 15 Population with and without education: race disaggregated by gender (2002 and 2013)
17.8% 24.2% 21.3%
9.0 10.3 9.1 8.1
2013
1.1% 2.7% 2.9%
6.1 6.4 5.4 4.7
4.5% 6.3% 6.3%
7.6 8.6 7.3 6.2
Black and mixed Male 2002 2013
1.8% 3.9% 3.6%
6.9 7.2 5.9 4.9
Female 2002
6.7% 9.3% 9.1%
8.3 9.3 7.9 6.8
2013
15 Tackling Intersecting Inequalities: Insights from Brazil 393
394
N. Kabeer and R. Santos
15–19 bracket which falls below tertiary education age), it translated into an increase in the gender gap in tertiary education in 2013. Even though the gaps for the first group reduced, members of the black/mixed and Indigenous ethno-racial groups still lag behind those from the white/yellow groups. It is important to note that, here as well, Indigenous women lag behind men from their groups as well as women from other ethno-racial groups, with the gaps increasing between 2002 and 2013.
Explaining the Decline in Intersecting Inequalities The decline in income inequality in Brazil in the early 2000s, even more so than the decline in poverty, has attracted a great deal of attention within the international development community because it occurred in a country with historically high levels of income inequality during a period that has been characterized by dramatically rising income inequalities in the rest of the world. Our analysis of national household data suggests that the decline in income inequality has benefited historically disadvantaged groups, men, and women at the intersection of class, race, and gender inequalities although Indigenous groups continue to lag behind the rest. The explanations put forward to explain Brazil’s break with its long history of inequalities encompass both policies and politics. In this section, we attempt to pull together some of these explanations in order to draw out what they tell us about the decline in intersecting inequalities, an aspect of social change that has attracted somewhat less attention in this literature.
From “Conservative Modernization” to “Liberal Neo-Developmentalism” While a number of researchers have pointed to the commodity boom of the early 2000s and the strong growth rates that characterized much of that decade as the main explanation for the decline in economic inequalities, Amman and Barrientos (2014) point out earlier periods of growth had not had the same impact: “For example, during the so-called ‘miracle years’ under military rule between 1967 and 1973, Brazil’s growth record, though very impressive by historical standards, could not overshadow the fact that income inequalities were widening and very little progress was being made in tackling entrenched poverty. This is in sharp contrast to conditions since 2003 – strong growth has been achieved hand in hand with a historic reversal of inequality trends” (p. 1). Instead, Amman and Barrientos as well as a number of other authors suggest that the policies adopted by Brazil in the wake of its democratic transition added up to a new development approach, one which combined elements of neoliberal orthodoxy with interventionist measures associated with its previous state-led “developmentalist” tradition to produce what has been termed a “liberal neodevelopmentalist” policy regime (Cornell 2012; Grinberg 2016). It was this
15
Tackling Intersecting Inequalities: Insights from Brazil
395
willingness to move away from the policy orthodoxy that was associated with rising inequalities in much of the rest of the world that allowed Brazil to buck the trend. The early 1990s was a period marked by hyperinflation and continued low rates of growth. It began with the widespread dismantling of various barriers to trade, marking the end of the ISI strategy and opening up the Brazilian economic to foreign competition. Financial markets were also liberalized, leading to a surge in portfolio investment and the appreciation of real exchange rates at a time when domestic companies had lost the protections of the ISI period and suffered from high interest rates. Imports grew at a pace much faster than imports. The adoption of the Plano Real (Real Plan) in 1993–1994 by President Cardoso upheld strong adherence to price stability using the orthodox methods of central bank independence and inflation targeting. While the 1988 Constitution had extended a range of labor rights and protections and encouraged greater attention to redistribution, with a particular focus on the worst off groups and regions, the adverse economic conditions in the years that followed ruled out the possibility of much progress on this. Instead, various reforms were undertaken to make labor markets more flexible, thus increasing informality and doubling unemployment. In addition, the decision to end the indexing of wages to inflation rates and to limit increases in the minimum wage led to a decline in average real wages in the second half of the 1990s (Berg 2010). The left-wing Workers Party that took power in 2001 appeared committed to promoting Brazil’s ability to compete in the global economy and retained the core elements of the Real Plan. The continuity of macroeconomic policy not only served to contain inflation but also in providing a predictable platform for both domestic and foreign investors. At the same time, the government deviated from classic neoliberal prescriptions in a number of important ways. One was interventionist economic policy. The Lula government came to power at a time when more competitive exchange rates and a global commodities boom had given rise to a more favorable macroeconomic environment, but it took steps to shape the pattern of subsequent growth. It adopted a “growth acceleration program” which expanded aggregate demand in the economy through state investment in infrastructure, increased supply of credit by state banks, and expanded investment by state-owned enterprises (Barbosa and Souza 2010). The state supported public enterprises to compete in the open economy and to diversify the domestic productive base. A second deviation from orthodox prescriptions related to labor market policies. The economic recovery generated a large number of new jobs, both in the export sector and the import-competing manufacturing sector. Additional policies were put in place to encourage greater formalization of the economy, including reducing the costs of registering small business and simplifying their tax obligations as well as incentivizing labor inspectors to ensure conformity to labor regulations (Berg 2010). As a result, around 20 million new jobs were created after 2003, most of which were in the formal economy. Of particular significance in the formalization process was the extension by the 1988 Constitution of the right of domestic workers to organize. While their
396
N. Kabeer and R. Santos
precarious conditions meant union membership remains low, the existing unions had been able to use the law and the labor courts to drive the registration of domestic workers so that during the 1990s when formal employment was falling, formalization among domestic workers was on the rise and continued to rise in the 2000s. The third aspect of the policy regime which deviated from the neoliberal orthodoxy was the pursuit of active redistributive policies. Some important social measures reaching out to poorer sections had already been in place in the 1990s: the Unified Health System (1990); the Rural Social Insurance (1993) which aimed at including rural informal workers within the existing Social Insurance Fund; the Family Health Programme (1994) which sought to extend health provision to lower income groups; the FUNDEF program (1994) to upgrade basic education; a noncontributory pension (1996) to provide transfers to poorer older people and people with disabilities; and the Bolsa Escola and PETI programs (1995), offering conditional cash transfers to poor families to reduce child labor and promote children’s health and education. The Lula administration (2003–2010) consolidated all federal guaranteed income programs into Bolsa Família while at the same time expanding the target population to 13 million households. By the end of the 2000s, they reached a quarter of the population, lifted 20 million people out of abject poverty, and nearly halved the proportion of people living below the total poverty line. A particularly important redistributive measure was the policy of increasing the minimum wage above the inflation rate which was adopted by Lula and continued by Dilma Rousseff (the president who succeeded Lula in 2011 and was later impeached). In Brazil, minimum wages provide a floor for labor earnings in formal employment and a benchmark for wage settlements of informal workers (the lighthouse effect). In addition, minimum social insurance pensions and noncontributory pensions were indexed to the minimum wage. The real value of the minimum wage was increased several times, with the bottom quintile of the labor force seeing its incomes rise by 38% between 2003 and 2008. Despite a highly regressive tax system in which the poor paid more taxes than the rich because of reliance on indirect taxes, the expansion of social expenditures could be financed because of the introduction of two new taxes in 1997 which shifted the financing of social policy from reliance on social security contributions and general tax revenues to a more diversified funding mix (Amman and Barrientos). The increasing rates of economic growth in the 2000s increased not only the value and quantity of social insurance contributions, but also the amounts collected from earmarked taxes. These conditions enabled a rapid increase in social expenditure, in absolute and relative terms. Certain features of this policy regime added further to its inclusive nature. One was that the mix of economic and social policies worked to reinforce each other (Amman and Barrientos; Berg). The strong focus on health and education of target groups, particularly the most disadvantaged, not only served to reduce large disparities across states, municipalities, and socioeconomic groups but also fed directly in to the skills and productivity of the labor force. In addition, as Berg points out, it reduced the number of youth entering the labor market every year in search of jobs.
15
Tackling Intersecting Inequalities: Insights from Brazil
397
The combined effects of growth and social policy explain why the rate of income growth was around 19–11%/annum in the lower income deciles compared to around 7–8% in the higher. Studies suggest that income growth among the lower quintiles was driven by a combination of improvements in employment, the rise in minimum wage, and income transfers. The other important feature was that the growth in social expenditure in Brazil and its impact on the incomes of disadvantaged groups led to multiplier effects through their impact on domestic demand. For instance, the IPEA estimated that the GDP multiplier of social expenditure taken as a whole was of the order of 1.37 in the mid-2000s, while the social expenditure multiplier in household income growth rates was higher at 1.85 (cited in Amman and Barrientos, p. 13).
The Politics of Social Change While the preceding analysis outlines the combination of policy measures that helped to bring about a decline in the country’s long-standing inequalities, it poses an interesting question: How did these policies come to be adopted in a country that had enjoyed many periods of growth in the past without seeking to address its high levels of poverty and inequality? Instead, as Skidmore (2004) pointed out, government policies on taxes and expenditures in the past consistently favored the 5–10% of the population who controlled most of the country’s wealth as well as the main levers of its government: “In practice, federal government acts as a powerful channel for redistributing income from those on the bottom to those at the top” (p. 46). In their discussion of the emergence of a new development policy regime in Brazil, Amman and Barrientos refer to the “forging of a broad consensus across the political spectrum, key actors in business, the labour movement and civil society” (p. 6) and to the “depth of the ‘buy-in’ across the political spectrum” (p. 3) that underpinned these policies. They explain this in terms of collective acknowledgment of the exhaustion of the previous economic model “which had effectively ‘locked in’ hyperinflation,” the clever combination of orthodox and heterodox measures that made up the new policy regime, and the “politically adept, consultative and inclusive approaches” (p. 3) of the administrations that began this process after the transition to democracy. But others suggest that the ability to undertake the ambitious redistributive agenda, associated in particular with the Lula regime, rested on favorable political conditions which had been engendered by long-term political processes that began before the transition to democracy. A recent ODI report on countries which had achieved a degree of success in addressing intersecting inequalities highlights a number of facilitating factors: social movements that build the case for, and mobilize around, the policy and politics necessary to address intersecting inequalities; enabling processes of constitutional change; left-wing orientation of parties in power; and efforts to deepen democracy beyond formal electoral politics (Paz Arauco et al. 2014). All of these factors have featured in the Brazilian context, and indeed Brazil was one of the case studies in the report. We elaborate on how these factors have played out in Brazil.
398
N. Kabeer and R. Santos
Social Movements It is clear from a reading of the Brazilian literature that the movement for democracy began sometime in the 1970s, well before democracy was restored. Studies of this period stress both the broad-based nature of mobilization by civil society (Baiocchi et al. 2008) as well as the specificities of forms it took (Dagnino 2007), both factors that had an important bearing on the nature of the transition to democracy and its aftermath. The onset of military rule in Brazil in 1964 was followed soon after by the harsh repression of all opposition. The military’s disregard for human rights and legality lost it the support of some of its traditional allies. Elite sectors of the establishment, including journalist, lawyers, the political opposition, and the Catholic Church, began to confront the military on its human rights record. As Hochstetler (1997) argues, by challenging the legitimacy of the military regime, these traditionally politically active sectors of the elite opened up a space for a grassroots opposition to emerge. Given efforts by the state to contain and de-politicize civil society organizations, neighborhood associations and church groups became the only legitimate places for organizing around popular grievances, leading to a highly decentralized form of protest at the grassroots level. These different streams of opposition to military rule swelled into a growing demand for democracy, but there were interesting divergences in what was being demanded. In their study of attitudes to democracy in the Brazilian population, Rochon and Mitchell (1989) distinguished between attitudes supporting the “procedures of democracy” (removing the military from politics, reducing state control over unions, and strengthening Congress) and those supporting the “substance of democracy” in terms of universal suffrage (people are capable of voting wisely; the vote should be extended to illiterates). They found that most people surveyed in 1972 expressed weak support for either meaning of democracy. This had changed considerably by 1982. For instance, opposition to the military in politics rose from 21% in 1972 to 48% in 1982; support for direct elections went from 57% in 1972 to 82% in 1982 while support for extending the vote to illiterates went from 38% to 56%. What was striking was that while elite attitudes changed in favor of democratic procedures, those from poorer sections expressed greater support for equality of participation in the democratic process. As the authors point out, the literacy qualification was still in place at that time, disenfranchising 25% of the population aged 15+. The Catholic Church played a special role in mobilizing grassroots opposition to military rule through a strategy of setting up of Christian Base Communities within neighborhoods (Cavendish 1994). Although progressive sections of the clergy had been supporting associations of disenfranchised groups since the 1950s, it began to play a far more active role in the 1970s with the rise of liberation theology. The Church increased the safety of activists, helped to make strong connections among community members and nurtured the rise of new social movements. Its influence is evident in a great deal of the social movement literature. Alvarez (1990) pointed it out in relation to women’s movements, Gondim (1989/1990) in relation to neighborhood movements, Bernadino-Costa (2011) in relation to domestic workers’
15
Tackling Intersecting Inequalities: Insights from Brazil
399
unions, Assies (1999) on urban neighborhood associativism, and Della Cava (1989) in relation to Indigenous, neighborhood, and landless movements. Citing Della Cava, Hochstetler points out that the Catholic Church could play this role because “. . .no other institution except for the military, enjoyed a nation-wide network of cadres, a system of communications (even if only door to door) that functioned despite censorship and, unlike the military, a world-wide organization on which it could draw for support and bank on for an international “hearing”” (Della Cava 1989, p. 147). There was another aspect of social movements during this period that echoes some of the earlier literature on the “horizontal sociabilities” that appeared to distinguish relations between different ethno-racial groups at the lower end of the income distribution in Brazil. Dagnino (2007) points to the role of cultural politics in shaping the movement for democracy, extending it beyond the demand for the formal institutions of democracy to the demand for equality of participation that (as the study by Rochon and Mitchell, 1989, had found) informed the understanding of democracy among poor and marginalized groups. She notes in particular how interpretations of citizenship became central elements in building this new inclusive cultural politics. Citizenship was defined, first and foremost, as “the right to have rights,” a definition that challenged previous definition that had been used by the state and dominant classes to exclude large sections of the population: “Class, race, and gender differences constitute the main bases for the social classification that has historically pervaded our cultures and established different categories of people hierarchically disposed in their respective ‘places’ in society” (p. 2474). To be poor in Brazil was not only to be materially deprived but also to be denied human dignity and political voice. The concept of citizenship was mobilized to express both not only the general demand for equal rights embedded in the predominant conception of citizenship but also extended and specified in accordance with specific demands. Citizenship provided a common ground between those struggling redistributive demands, such as housing, land, wages, and so on within their neighborhoods and work places and those challenging cultural rules that denied them recognition, including ethnic minorities, women, sexual rights activists, and human rights activists more generally. In many cases, this common ground served not only to extend rights to those who had been previously excluded but also to give rise to new categories of rights which sought to remedy injustices previously not acknowledged: the right to autonomy over one’s own body, to sexual freedom, and to racial justice. It was this struggle over culture that appears to have been a particularly important hallmark of the Brazilian transition, given the significance of the intersection between economic deprivation and cultural devaluation which characterized its more enduring inequalities. In encompassing the cultural dimension of citizenship, these movements seemed to have effectively politicized the “horizontal sociabilities” that had characterized social relations between different groups at the lower end of the income distribution: “The struggle for citizenship was thus presented as a project for a new sociability: not only for the incorporation of broader citizenship into the political system in the strict sense, but for a more egalitarian format for social
400
N. Kabeer and R. Santos
relations at all levels, new rules for living together in society (negotiation of conflicts, a new sense of a public order, public responsibility, and a new social contract)” (op. cit., p. 2475).
Democratic Practice in Formal Politics While the social movements that became active around the struggle for democracy encompassed many different groups with various kinds of demands, labor organizations emerged as among the most influential. The New Union Movement had its origins in the wildcat strikes that broke out in the late 1970s and workers’ increasing demands for greater shop floor democracy in direct challenge to the older, organized unions which had long been co-opted by the state. The membership of the new movement soon exceeded that of organized labor and became part of the proliferation of civil society organizations that spearheaded the opposition to the military regime. When the military moved to legalize the setting up of new parties, it founded the Partido dos Trabalhadores (PT), or Workers Party, under the leadership of “Lula” da Silva with the objective of promoting an alternative form of politics to the clientelistic relations (“citizenship by concession”) which had taken deep roots in the country’s political culture. The PT adopted the bottom-up participatory model of its allies in the progressive clergy which emphasized the need to build from the base up, the strategy of “basismo.” This began with nuclei of small groups of people who were organized by neighborhood, job category, work place, or social movement (Keck 1992) which met regularly to discuss issues of relevance to the party and the wider movement. The party avoided official ties with any single union or federation or with Catholic Church, the women’s movements, and so on but invited their representatives to join the party and influence its platform. As Guidry suggests, “For both labor and popular leaders, basismo was a radical form of democracy that inverted the traditional top-down, hierarchical form of political relationships in Brazil, creating the space for the direct democracy and conscientizacão stressed in Freirian ideology” (Guidry 2003, p. 91). As Serbin (1999) points out, while the Church was the glue that had held together the progressive groups that opposed the military before 1985, after 1985 the glue became the Workers’ Party. Brazil had its first presidential election in over two decades when military stepped down in 1985. One of the first acts of the newly elected President was to remove the literacy requirement which had disenfranchised a sizeable proportion of the population so that the municipal elections later that year were the first to be based on universal suffrage in the country’s history. The PT won a number of major cities in the 1988 municipal elections, including Porto Alegre and São Paulo, and went on to win further victories at the local level in later years. It was able to use its control of municipal governments to try out various experiments in participatory democracy (Baiocchi 2003). The best known of these was participatory budgeting pioneered by the party in Porto Alegre but adopted in a number of cities that the party governed. The Bolsa Escola program was first developed in the mid-1990s in municipalities governed by the PT and later became the basis of the national Bolsa Familia under Lula’s regime.
15
Tackling Intersecting Inequalities: Insights from Brazil
401
Brazil also adopted a new constitution in 1988 which, as the ODI report pointed out, “represented an explicit effort to break with historically-established patterns of inequality, based on intersections between poverty, race, ethnicity, gender and location, which characterized the region.” This was reflected in strong commitment to social and economic rights inscribed into the constitution. As Hevia-Pacheco and Vergara-Camus (2013) point out, this was a view of rights which was understood in universal terms, granted on the basis of citizenship, a departure from earlier views which linked had rights to an occupation or some other qualification. They also note that what distinguished this constitution from previous ones which had been a largely elite affair was that it was the culmination of “an intense process of social mobilisation from organisations from the whole political spectrum, including from civil society organisations and social movements, which had historically been excluded from these proceedings.” Although the economic conditions prevailing in the country at the time and the neoliberal reforms adopted by the government in response made it difficult for it to live up to these commitments, the constitution nevertheless exercised a great deal of symbolic power. Some efforts to promote policies designed to reach the poor and marginalized were put in place in the 1990s, and while they achieved an extension of social security to rural areas and an expansion in education, they did little to stem rising inequalities associated with the attempts to implement neoliberal austerity programs. Mass mobilization against the reforms led to the election of the PT which had been active in campaigning for the new constitution and could invoke its authority to promote a redistributive program when it took over the government. The massive expansion in social programs undertaken by Lula can thus be seen as an attempt to deliver on the promises of the 1988 constitution. The government took a number of measures which were explicitly targeted to previously excluded groups, including quotas for Afro-descendants in public universities, the Brasil Quilombola program, the National Integral health Policy for the Black Population, the national Policy for the Promotion of Racial Equality, and National Plans for Policies for women. But as the ODI report points out, “the major thrust of its strategy for addressing intersecting inequalities has been through more general programmes for poverty reduction in keeping with the principle of universalism in access to social security enshrined in the constitution.” Various mechanisms for collaboration with civil society were put into place when it came into power. Along with the promotion of participatory budgets, which has received international national attention, less-well known efforts include national state or local councils, which convene conferences as well as more regular council meetings to review policies in a sector or on a specific issue. Delegates to the national conferences were elected in state conferences and expected to represent the agreements reached at state level. In a process that lasts approximately a year prior to the national conferences, municipal and large city conferences are held to review local problems and needs pertaining to specific issues and to forward them to the delegates at the national level. These policy discussions and consultative bodies were mentioned in the 1988 Constitution but have only been implemented by the PT government since 2003.
402
N. Kabeer and R. Santos
Conclusion We set out in this chapter to explore how the decline in income inequalities observed in Brazil in the early 2000s impacted on men and women who were located at the intersection between income inequality and marginalized group identities. Drawing on national data, we found very clear evidence that the lives and life chances of these groups improved during this period. We found that gender, ethno-racial, and regional inequalities in income had all declined, and that along with an overall decline in absolute poverty, there had been a decline in poverty among different groups, including those defined by intersecting inequalities, with a particularly large decline among Indigenous people, who were the poorest section of the population. While there was some convergence in earnings across different occupational categories, the occupational structure remains stratified by class and gender although, among poor men and women, the formalization of work status emerged as a major indicator of improvement in working conditions. There has also been increased formalization of land rights to the benefit of most groups - with the exception of Indigenous men and women. Summing up our findings, we find that the overall decline in income inequalities in Brazil since 2000 has been accompanied by an overall decline in horizontal and intersecting inequalities in income and its largest component, wages, in labor market conditions, in rights to land, and finally in education, a key driver of greater labor market mobility between 2002 and 2013. Progress has been slower for certain intersectional groups than others. Women from different marginalized ethno-racial groups remained clustered at the lower levels of the income distribution and poorer paid jobs. Indigenous men and, to an even greater extent, Indigenous women are in greatest danger of being left behind. Our explorations of various indicators of inequality suggest that this group finds itself with lower incomes and wages than other ethno-racial groups, are lagging behind in education, are crowded into narrower range of occupations and work statuses, and have possibly lost ground in the formalization of land rights. What makes these achievements particularly remarkable is that they occurred at a time when inequalities had been rising in much of the rest of the world. We pointed to the long history of social mobilization that allowed a party into power that had its roots in this history, whose leadership was drawn from the popular classes and who had shown imagination in fostering grassroots democratic practices both before and after it came into power. There are also some important lessons to be learnt from Brazil’s experiment with liberal neodevelopmentalism. First of all, its achievements reflected the attempts of a government to carve out its own path between the macroeconomic discipline enforced by the need to compete in the global economy and the need for strategic intervention to diversify the domestic economy so as to protect it from overreliance on global markets. Second, by defying neoliberal prescriptions with regard to dismantling welfare provision in the interests of labor market flexibility, the Brazilian model showed that it was possible to break with a long-standing history of inequality and to channel some of the fruits of growth to those whose class, gender,
15
Tackling Intersecting Inequalities: Insights from Brazil
403
and race had long positioned them at the bottom of intersecting economic, social, and political hierarchies. Third, and more practically, the Brazilian model demonstrated the synergies that could arise between redistributive policies that invested in the productivity of the labor force and economic policies that sought to develop domestic industrial capacity. The period covered by our study ends before the political crisis that terminated the rule of the Workers Party rule and led to the impeachment of its leader. The Brazilian political system became mired in allegations of corruption that surround both leading figures of the Workers Party as well as of the Opposition that brought it down. It has seen the rolling back of many of the gains of the previous regimes. Attention has also turned to the decline in the country’s economic performance that had set in by the end of our study period. Observers point out that its pattern of growth had not, despite policy efforts, resulted in major productivity gains or an increase in the technology content of products made in Brazil (Costa et al. 2015). Instead, increased prosperity and rising levels of consumption led to massive import of consumer goods, encouraged by an overvalued exchange rate, and a boom in the service economy. Without deeper structural transformation of the economy, it is not clear that growth can be revived. In addition, there are also limits to redistribution that could be achieved, given the balancing act attempted by the Workers’ Party of leaving privilege untouched at the top while attempting to address disadvantage at the lower end. While there was a substantial increase in country’s tax revenues over our study period, regressive indirect taxes continue to account for half of this revenue, and the highest rate of income tax is 28% while the tax burden on capital and finance was described as “mild.” According to one calculation, around 0.2% of the national population owned 47% of declared property and titles (Costa et al. 2015). The Workers Party did not show any inclination toward tax reform since it came to power. Redistribution therefore did not go far enough to empower those at the bottom to become a sufficiently strong political force to hold the Workers Party accountable to those who voted it into power, to prevent its slide into corruption and compromise, and to counter the continued influence of powerful interests who have a stake in maintaining the status quo. Acknowledgments The research for this chapter was carried out with a grant from the International Inequalities Institute at the London School of Economics. A longer working paper version of this chapter can be found in Kabeer and Santos (2017). We would like to thank Evelina Dagnino, Gianpaolo Baiocchi, Rachel Gisselquist, Carla Canela, and Alex Shankland for comments on an earlier draft of the chapter. The usual disclaimers apply.
Appendix Explanation of indicators used: A first indicator used here is the Group Weighted Coefficient of Variation (GWCOV):
404
GWCOV ¼
N. Kabeer and R. Santos R
1 y
r¼1
pr ð y r y Þ
2
1=2
, where yr ¼ n1r
nr i¼1
yir is group r’s mean
value, R is the number of groups, pr is group r’s population share, and yir is the observed value of income or another variable of interest (e.g., years of education or hours of domestic work) of the ith member of group r. 1 GWGini ¼ 2y
R
S
r¼1 s¼1
pr ps jyr ys j where r and s are, eventually overlapping,
group index numbers, R ¼ S are the total number of groups, yk ¼ n1k
nk i¼1
yik is k’s
group averages with k ¼ r,s, pk is group k’s population share, and yik is the observed value of income or another variable of interest (e.g., years of education or hours of domestic work) of the ith member of group k. GWTheil ¼
R
r¼1
pr yyr ðlog yr log yÞ ¼
R
r¼1
pr yyr log
yr y
, where yr ¼ n1r
nr
i¼1
yir is
group r’s mean value, R is the number of groups, pr is group r’s population share, and yir is the observed value of income or another variable of interest (e.g., years of education or hours of domestic work) of the ith member of group r.
References Alvarez SE (1990) Engendering democracy in Brazil: women’s movements in transition politics. Princeton University Press, Princeton Amman E, Barrientos A (2014) Is there a Brazilian model of development? Are there lessons for countries in Africa? WIDER working paper 2014/134. World Institute for Development Economics Research, Helsinki Assies W (1999) Theory, practice and “external actors” in the making of new urban social movements in Brazil. Bull Lat Am Res 18(2):211–226 Baiocchi G (ed) (2003) Radicals in power: the Workers Party and experiments in urban democracy in Brazil. Zed Press, London Baiocchi G, Heller P, Silva MK (2008) Making space for civil society: institutional reforms and local democracy in Brazil. Soc Forces 86(3):2–16 Barbosa-Filho N, Souza JAP (2010) A inflexão do governo Lula: política econômica, crescimento e distribuição de renda. In: SADER, E.; GARCIA, M. A. (eds.). Brasil entre o Passado e o Futuro. São Paulo: Fundação Perseu Abramo / Boitempo, pp. 57–110 Berg J (2010) Laws or luck? Understanding rising formality in Brazil in the 2000s working paper no. 5. ILO Office, Brasilia Bernadino-Costa J (2011) Destablizing the national hegemonies narrative: the decolonized thought of Brazil’s domestic workers’ unions. Lat Am Perspect 38(5):33–45 Birdsall N, Sabot RH (eds) (1996) Opportunity foregone: education in Brazil. Inter-American Development Bank/Johns Hopkins University Press, Washington, DC Caldwell KL (2007) Negras in Brazil: re-envisioning black women, citizenship and the politics of identity. Rutgers University, Piscataway Carla Canelas & Rachel M. Gisselquist (2018) Horizontal inequality as an outcome, Oxford Development Studies, 46(3):305–324, https://doi.org/10.1080/13600818.2018.1508565 Cavendish JC (1994) Christian base communities and the building of democracy: Brazil and Chile. Sociol Relig 55(2):179–195 Costa S, Friza B, Sproll M (2015) Dilma 2.0: from economic growth with distribution to stagnation and increasing inequalities? LASA Forum XLVI(3):21–24
15
Tackling Intersecting Inequalities: Insights from Brazil
405
Dagnino E (2007) Dimensions of citizenship in contemporary Brazil. Fordham Law Rev 75(5): 2469–2482 Degler CN (1971) Neither Black nor White: slavery and race relations in Brazil and the United States. Macmillan, New York Della Cava R (1989) The “people’s Church”, the Vatican, and Abertura. In: Stepan A (ed) Democratizing Brazil. Oxford University Press, Oxford, UK, pp 143–167 Freyre G (1933) The masters and the slaves. A study in the development of Brazilian civilizations. University of California Press, Berkeley Gondim LM (1989/1990). Os Movimentos Sociais Urbanos, A Questao de Organizacao e a Democracia Interna. Revista de Ciencias Sociais 20/21(1/2):31–60 González MT (2008) Race, gender and ethnicity in contemporary Brazil. Lat Am Res Rev 43(1): 219–224 Grinberg N (2016) From populist developmentalism to liberal neo-developmentalism: the specificity and historical development of Brazilian capital accumulation. Crit Hist Stud 3(1):65–104 Guidry JA (2003) Not just another labour party: the Workers’ Party in Brazil. Labour Stud J 28(1): 83–108 Hevia-Pacheco, P. and L. Vergara-Camus (2013) Addressing intersecting inequalities: inclusive political regimes, democratically-elected left-wing governments: the cases of Brazil and Ecuador. Background paper for ODI Report (2014) Strengthening social justice to address intersecting inequalities post-2015. London: Overseas Development Institute Hochstetler K (1997) Democratizing pressures from below? Social movements in new Brazilian democracy presentation to Latin American Studies Association XX international congress, Guadalajara, 17–19 April Kabeer N (2010) Can the MDGs provide a pathway to social justice? The challenge of intersecting inequalities. Institute of Development Studies/MDG Achievement Fund, Brighton/New York Kabeer N, Santos R (2017) Intersecting inequalities and the Sustainable Development Goals: insights from Brazil, International Inequalities Institute working paper no. 14/WIDER working paper 2017/167. London School of Economics/UNU-WIDER, London/Helsinki Keck M (1992) The Worker’s Party and Democratization in Brazil. New Haven, Yale University Press Lima, M. (2001). Serviço de preto, serviço de branco: representações sobre cor e trabalho no Brasil. Tese (Doutorado) - IFCS/UFRJ Lipton M (1983) Labour and poverty, World Bank staff working papers 616. World Bank, Washington, DC Lovell PA (2000) Gender, race and the struggle for social justice in Brazil. Lat Am Perspect 27(6): 85–102 Lovell PA, Wood CH (1998) Skin colour, racial identity and life chances in Brazil. Lat Am Perspect 25(3):90–109 Mancini L, Stewart F, Brown GK (2008) Approaches to the measurement of horizontal inequalities. In F. Stewart (Ed.), Horizontal inequalities and conflict: Understanding group violence in multiethnic societies (pp. 85–105). Basingstoke: Palgrave Macmillan Marió EG, Woolcock M (2008) Social exclusion and mobility in Brazil. World Bank, Washington, DC Marió, Estanislao Gacitúa and Michael Woolcock (2004) Social exclusion and mobility in Brazil Environmentally and Socially Sustainable Development, Latin American and Caribbean Region World Bank Paz Arauco V, Gazdar H, Hevia-Pacheco P, Kabeer N, Lenhardt A, Masood SQ, Naqvi H, Nayak N, Norton A, Sabharwal NS, Scalise E, Shepherd A, Thapa D, Thorat S, Hien Tran D, VergaraCamus L, Woldehanna T, Mariotti C (2014) Strengthening social justice to address intersecting inequalities post-2015, ODI report. Overseas Development Institute, London Rezende CB, Lima M (2004) Linking gender, class and race in Brazil. Soc Identities 10(6):757–773 Rochon TR, Mitchell MJ (1989) Social bases of the transition to democracy in Brazil. Comp Polit 21(3):307–322
406
N. Kabeer and R. Santos
Serbin KP (1999) The Catholic Church, religious pluralism and democracy in Brazil working paper no. 263. Helen Kellogg Institute for International Studies, Notre Dame Skidmore TE (1992) Fact and myth: discovering a racial problem in Brazil, Working paper 173. Helen Kellogg Institute for International Studies, Notre Dame Skidmore TE (2004) Brazil’s persistent income inequality: lesson from history. Lat Am Polit Soc 46(2):133–150 Stewart F (2002) Horizontal inequalities: a neglected dimension of development, Working paper no. 81. University of Oxford, Oxford, UK Telles E (1990) Características sociais dos trabalhadores informais: o caso dasáreas metropolitanas’, Estudos Afro-Asiáticos: 61–81 Telles EE (2004) Race in another America. The significance of skin color in Brazil. Princeton University Press, Princeton
Race and Gender
16
Rhonda Vonshay Sharpe
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Social Constructs: Ethnicity, Gender, and Race . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ethnicity, Gender, and Race Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disaggregated vs. Aggregate Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
408 408 409 413 416 419 419
Abstract
Ethnicity, gender, and race are hierarchical social constructs that segment societies by cultural norms and characteristics that are dynamic and underpin discrimination. As the United States becomes more diverse, identifying discrimination based on gender, race, or ethnicity will be more complexed. An increase in interracial- and intraracial-ethnic and interracial marriages will produce children who are racially ambiguous, that is, children who are difficult to categorize into racial and ethnic categories. Additionally, as society expands cultural norms around gender and gender identity, it may be challenging to discern discriminatory treatment due to gender. Kim (Intersectionality and gendered racism in the United States: a new theoretical framework. Rev Radical Polit Econ, 616–625, 2020) suggests that gendered racism – the intersection of racial and gender stereotypes about ethics, intellect, masculinities, leadership, and nurturing ascribed to race, ethnic, and gender groups – is reinforced by education, legal, and penal systems. These stereotypes also influence how data are collected and reported (Sharpe, We’ve to build the pipeline. What’s the problem? What’s next? The Remix. Rev Black Polit Econ, 191–215, 2019), limiting data availability to identify R. V. Sharpe (*) Women’s Institute for Science, Equity and Race, Mechanicsville, VA, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_29
407
408
R. V. Sharpe
discrimination or other biases. An inclusive and equitable society requires data collection and the disaggregation of data, so intersectional analysis can be used to explain how our complex identities underpin bias and discrimination. Feminist and gender scholars know that gender, ethnicity, and race operate differently across and within groups. Given data limitations, scholars and policymakers must 1) start with data that allow for disaggregation by characteristics outlined in the literature to influence the outcome in question; 2) report findings with a keen eye for the nuanced differences in outcomes; and 3) interpret the findings using an intersectional approach. Keywords
Race · Gender · Intersectionality · Discrimination
Introduction Ethnicity, gender, and race are hierarchical social constructs that segment societies by elevating cultural norms and characteristics. Ethnicity, gender, and race function similarly to prices, becoming metrics for deciding who gets scarce resources, who owns the capital, and who is the labor. The cultural norms and flawed research used to delineate the categories of ethnicity, gender, and race hinder the full participation of many in various markets – education, health, housing, labor, marriage, etc. – that combine to create well-being. This chapter first defines ethnicity, gender, and race. Second, it provides challenges to how research on those subjects is examined, especially research focused on identifying inequities in outcomes. Third, it gives an example of how findings differ if data are disaggregated. Finally, it offers guidance for conducting research that disaggregates data and uses the intersectional framework.
Social Constructs: Ethnicity, Gender, and Race Ethnicity is often defined by customs, language, and ancestry – characteristics associated with culture (American Anthropological Association 2021). Ethnic groups may share similar physical traits, and historical intermarriage patterns have created heterogeneity in appearance for people with a common ethnicity. In the United States, people of Hispanic/Latino ethnicity are of varied racial backgrounds and are defined (US Office of Management and Budget 1997) as follows: Hispanic or Latino. A person of Cuban, Mexican, Puerto Rican, Cuban [sic], South or Central American, or other Spanish culture or origin, regardless of race. The term “Spanish origin” can be used in addition to “Hispanic or Latino.”
16
Race and Gender
409
History has taught us that “Today’s ethnicities are yesterday’s races” (American Anthropological Association 2021). We learn from Painter (2010) and Isenberg (2016) how some ethnic groups – Italians, Jews, Irish, and others – became ethnic groups within the White race. The US Office of Management and Budget (1997) defined racial categories for all federal reporting agencies as follows: American Indian or Alaska Native. A person having origins in any of the original peoples of North and South America (including Central America) and who maintains tribal affiliation or community attachment. Asian. A person having origins in any of the original peoples of the Far East, Southeast Asia, or the Indian subcontinent including, for example, Cambodia, China, India, Japan, Korea, Malaysia, Pakistan, the Philippine Islands, Thailand, and Vietnam. Black or African American. A person having origins in any of the Black racial groups of Africa. Terms such as “Haitian” or “Negro” can be used in addition to “Black or African American.” Native Hawaiian or other Pacific Islander. A person having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands. White. A person having origins in any of the original peoples of Europe, the Middle East, or North Africa. Neil Irvin Painter’s The History of White People chronicles the evolution of “whiteness,” showing that the above categories result from years of discourse around who is White. Such suggests that regardless of the categories used for ethnicity and race, their purpose is to distinguish and control “nonwhite” populations (American Anthropological Association 2021). Like ethnicity and race, gender is a social construct that assigns norms, behaviors, and roles to those individuals perceived to be men, women, boys, and girls. One’s perceived gender may not align with one’s gender identity, i.e., one’s intimate and individual experience of gender. Gender and gender identity are different from sex, which is biological and defined by chromosomes, hormones, and/or reproductive organs. As social constructs, ethnicity, gender, and race vary from society to society and are not static. Because these characteristics are cultural norms, they are dynamic, even if the observed changes are slow. The consequences of cultural norms that define ethnicity, gender, and race create inequities in asset ownership, occupation distribution, and pay (Darity 2005; Bueno 2015; Darity, Hamilton and Stewart 2015).
Ethnicity, Gender, and Race Research Research that examines ethnicity, gender, and race has two effects: (1) expanding what is deemed credible research and (2) exposing inequalities. For many scholars, the credibility of their research has repercussions for their career trajectory. Jacobsen
410
R. V. Sharpe
and Newman (2003) found that female authors are more likely to focus on women in their analysis and to specify gender effects. However, they did not find the same for race/ethnicity, which they attributed to the absence of nonwhite economists publishing in the top journals for industrial relations and labor economics. Mason et al. (2005) found that Black authors are overrepresented in research on racial inequality. Therefore, journals of the International Association for Feminist Economics (IAFFE) (Feminist Economics) and the National Economic Association (Review of Black Political Economy), as well as the Journal of Race and Policy, are invaluable outlets for research that challenges conventional economic thinking about racism, sexism, and economic inequality. Research on race/ethnicity and gender spans some six decades and often does not examine the experiences of those at the intersection of gender and race/ethnicity. A seminal body of work to examine the “full” life of Black women is the edited volume Slipping Through The Cracks: The Status Of Black Women (Simms and Malveaux 1986). The volume covered the full scope of a Black woman’s life – employment, education, family structure, poverty, wealth, health, and economic development (her status in the African diaspora). The volume offered policy recommendations (p. 290), and it is striking how relevant they are today: Proponents of the Women’s Economic Equity Act, for example, failed to note that certain provisions of that act focused on the status of upper and middle-income women while ignoring the status of low-income women, who gain little from insurance reform. A sensitivity to the breadth of the “women’s coalition” might have suggested placing more effort on those provisions of the bill that all women could have supported, such as the daycare provisions. Similarly, efforts at job creation, which usually have a positive effect on the black community, usually focus on public works jobs. Concerns with the status of black women will mean either making public works jobs more available to women or including a broader range of jobs in the job creation efforts.
Wallace (1986) noted that in the economic sphere, Black and White women share many common concerns, but sharp differences in employment, occupation, family structure, and income and wealth status have resulted in more precarious lived experiences for Black women and their communities. Banks (2020) offered a framework for evaluating the unpaid, unseen, devalued collective community work that Black and racially marginalized women provide. She argued that the feminist theory and the theory of the firm do not allow for the multiple forms of oppression, racism, and exclusion from the firm, household, and community, experienced by these women. Banks’ framework elevated the community to be on par with the firm and household; hence, the community is a site of production and a framework for valuing the work of Black women. A common thread in race and gender research is that White women advocate for policies that address the ways gender – being a woman – is oppressive (Malveaux 1985), while Black and other racialized women advocate for policies that address the complexity of their oppression: race, gender, family structure, nativity, etc. (Kim 2020; Saunders and Darity Jr. 2003). Historically, racialized women have found it challenging to support a women’s agenda because it often ignores the needs of the
16
Race and Gender
411
marginalized communities. For example, the feminization of poverty ignores the persistent poverty in Black and Brown communities (Malveaux 1985). Additionally, the causes of poverty for White women and racialized groups are different. Loretta Ross (2015) made a similar point when explaining the origins of the term “women of color.” Fearing their needs would not be addressed by the larger community of women, other racialized women joined Black women to create an agenda for “women of color.” Ross explained that “women of color” was coined as an expression of political solidarity and represented a commitment to collaborate politically with other “minoritized” women to address common oppressions. Over the years, the term has morphed to be a biological term that is shorthand for nonwhite racial or ethnic groups. As a biological term, “women of color” (“people of color”) aggregates the unique experience of Asian, Black, Hispanic, Native American, and multiracial women; it thereby erases identity and conflates the distinct differences for each group of women. Nina E. Banks reminded me that White vs. nonwhite women is problematic for the same reason but also because it explicitly centers White women. Researchers must understand that shared experiences among various groups of women may not lead to similar outcomes. Although using terms like “women of color” and “nonwhite women” is problematic, US survey data often do allow for analysis that considers the “complexities” of individuals. As the United States becomes more diverse, the modern twist on racial discrimination will be discrimination against those who identify at the intersection of race and ethnicity – namely, Blacks who also identify as Hispanic and individuals who are multiracial. According to the US Census, the percentage of married-couple households that were interracial or interethnic grew across the United States from 7.4% in 2000 to 10.2% in 2012–2016 (Rico 2018). That assessment assigned 95.1% of interracial/ interethnic couples to seven categories, four of which list non-Hispanic White intermarried with another racial group or with Hispanic. Non-Hispanic Whites married to another racial group were 54.8% of such couples. Despite historical anti-miscegenation laws preventing the marrying of Blacks to not only Whites (Jenks 1916) but also to Native Americans and Asians, Rico et al. only explicitly listed marriage with Blacks. This reporting may not be the result of the authors’ bias but a reflection of hundreds of years of protecting whiteness that has created a preference for the value of “whiteness,” aka “white privilege.” Nonetheless, the data show a preference for intermarrying with Whites that is likely to replicate the observed racial hierarchy within interracial couples, resulting in consequences for their children. Using history as a guide, interracial couples with a Black spouse and their children will occupy the lower rungs within their communities (Guis 2013). Research suggests that Blacks who identify as biracial are more likely to be described as attractive compared to those who identify as Black only (Reece 2016). Researchers have found that lighter-skinned Blacks have an advantage in the labor market (Goldsmith et al. 2007; Kreisman and Rangel 2015; Price 2010; Robst et al. 2011; Rosenblum et al. 2015), the marriage market (Hamilton et al. 2009), and the judicial system (Gyimah-Brempong and Price 2006).
412
R. V. Sharpe
The largest intermarried combination was nonwhite Hispanics with Hispanics, approximately 40% (Rico 2018). Although Hispanics can be of any race, according to census data, 65% of Hispanics identify as White (https://factfinder.census.gov/ faces/tableservices/jsf/pages/productview.xhtml?src¼bkmk). Seminal research by Reimers (1983) found that Hispanic men earn less, and Cotton (1993) found that Hispanic Black men earned less than Hispanic White men and that Hispanic men earned less than non-Hispanic Black men. Mora and Dávila (2018) found that controlling factors, such as education, experience, immigrant status, and regional differences in cost of living, narrow the wage gap between White and Hispanic men but do not explain the wage gap between White men and Hispanic women (DiAngelo 2012). Sociologist Nancy Lopez (2018) noted that the box checked for race may not reflect the respondent’s “street race,” the race a stranger would assign based on phenotype. The disconnect between one’s street race and the self-identified race will be problematic for lawyers to sue based on racial or ethnic discrimination and for social scientists who research racial or ethnic inequality. As Lopez pointed out, the census and other datasets do not capture ascribed racial characteristics that may influence the street race for Hispanics; such arguably applies to multiracial individuals as well. Therefore, studies that use these data will likely underestimate the occurrence and costs of discrimination and inequality or miss the nuanced differences in outcomes that result from differential treatment based on street race. For example, Meghan Markle and Tia and Tamera Mowry are biracial women from the union of a Black mother and White father. However, Markle’s race is more ambiguous than that of the Mowry twins. Reece (2018) found that mulattos (children from a person of mixed White and Black ancestry) have higher occupational status. If Black-White unions produce more children who are racially ambiguous and who have better social and economic outcomes than children who are phenotypically Black, the differences in these outcomes will be masked in the data. The preference for “White” appearing will be masked in the outcomes and potentially bias downward inequality in outcomes. This would hold for other racial groups as well. Given that characteristics associated with race or hue are likely to be the basis for unequal treatment,1 a solution to this data issue would be to have trained interviewers assign skin shade to respondents or to classify the respondent’s street race. The Multi-City Study of Urban Inequality and the National Survey of Black Americans are examples of such datasets. Collecting such data will be expensive, but without information about salient characteristics of respondents, it may be
1
The Multi-City Study of Urban Inequality and the National Survey of Black Americans datasets provide information for each skin shade. See Darity, William, Jr.; Hamilton, Darrick and Dietrich, Jason. “Bleach in the Rainbow: Latino Preference for Whiteness.” Trans-forming Anthropology, 2005, 13(2), pp. 103–110; Goldsmith, A. H., Hamilton, D., and Darity, W. A., Jr. (2006). Shades of Discrimination: Skin Tone and Wages. The American Economic Review, 96(2)242–245; Rodriguez, Clara E. “The Effect of Race on Puerto Rican Wages,” in Edwin Melndez, Clare E. Rodriguez and Janis B. Figueroa, eds., Hispanics in the labor force: Issues and policies. New York: Plenum Press, 1991, pp. 77–96
16
Race and Gender
413
difficult to prove discriminatory practices against interracial marriages, persons intermarried with Hispanics, or the children from these unions. Kim (2020) offered gendered racism, the intersection of racial and gender stereotypes, as an explanation for the persistent labor market disparities and subsequent inequalities observed for racialized minorities and White women. She suggested that stereotypes about ethics, intellect, masculinities, leadership, and nurturing, which are ascribed to race, ethnic, and gender groups, are reinforced by education, legal, and penal systems and hence sustain institutionalized inequality. Gendered racism may also have implications for research on gender identity and sexual orientation. In addition to the availability of data appropriate for analyzing racial discrimination, data need to be appropriate for analyzing the fluidity of sexual orientation and gender identity (SOGI) (Park 2016). The results of a US Bureau of Labor Statistics (BLS) and US Census Bureau study “did not identify any significant issues that would make collecting SOGI information in the CPS infeasible, though there are many outstanding issues identified in the full study reports that must be studied and addressed prior to any implementation efforts” (Edgar et al. 2018). The concerns expressed were grounded in agency, who can report an individual’s SOGI, and privacy, willingness to report SOGI. The US Census via the American Community Survey and Decennial Census collects information on same-sex households and marriages. These data have been used to evaluate the economic and social well-being of US lesbian and gay couples (see Badgett 1995, 2006; Schneebaum and Badgett 2019). US Department of Education data allow respondents to choose other in addition to male or female. However, like race, a respondent’s choice may not align with their “street gender.” So, while EEOC law prohibits discrimination on the basis of gender identity or sexual orientation, the data do not allow researchers to describe the frequency or costs of discrimination against these groups. Data limitations withstanding, researchers must avoid aggregating data to provide trite analysis of data by race and ethnicity. Data aggregated by race/ethnicity and gender mask disparities in outcomes. Instead, researchers must disaggregate data by the characteristics believed to influence the outcome of interest. The next section provides examples of how aggregate data hide nuanced outcomes.
Disaggregated vs. Aggregate Outcomes Over 50 years have passed since Blacks and women have had their concerns about representation in and treatment by the economics profession addressed by the American Economic Association. While there has been progress in the number of degrees earned by Asians, Blacks, Hispanics, Native Americans, and White women, the economics profession has failed to create a diverse profession that is inclusive. Wu (2017) analyzed data from the Economics Job Market Rumors forum to identify differentials in the way men and women are portrayed. Her findings suggest a “pervasiveness of gender stereotyping” against women. Although her analysis failed to disaggregate the data by race/ethnicity, nativity, or sexual orientation, her
414
R. V. Sharpe
research was a watershed moment for the profession to acknowledge and confront pervasive sexism, racism, elitism, and other toxic and exclusionary behaviors. The AEA climate survey found that 39% of Asian women, 49% of Black and White women, and 54% of Latinas reported experiencing gender discrimination. At the intersection of race/ethnicity and gender, 44% of Asian women, 62% of Black women, and 58% of Latinas reported discrimination due to race/ethnicity and gender. Although men reported low percentages of discrimination due to gender, men did report discrimination due to their race or ethnicity: 22% of Asian men, 43% of Black men, and 15% of Latino men. Much of the research about diversity in economics fails to disaggregate the data using an intersectional approach, i.e., reporting and collecting the data by intersecting race/ethnicity with gender; such failure may mask the nuances in outcomes specific to each race/ethnic and gender group (Sharpe 2019). Consistent with Sharpe and Swinton (2012), Table 1 shows that 63% of men and 63% of women who completed the doctorate in economics in the United States from 1965 to 2015 held an undergraduate degree in economics, while 70% of Black women who completed the doctorate in economics held a bachelor’s degree in economics. Fewer than 50% of the graduates who identified racially as other held a bachelor’s degree in economics. Despite this, the research on factors that influence majoring in economics does not take an intersectional or feminist approach. Studies that do not take an intersectional approach assume that all women and all members of racial or ethnic groups share a monolithic experience. However, data for economics degrees conferred at the bachelor’s level suggest this assumption is flawed. Table 2 provides data for bachelor’s degrees conferred for the 20 years 1999–2018. The growth in degrees conferred for 1999–2008 to 2009–2018 was 32% for women and 4% for men. However, data disaggregated by ethnicity, race, and gender show that Black women earned 5% fewer degrees for 2009–2018 compared to 1999–2008. Despite Hispanic women nearly doubling the number of degrees
Table 1 Undergraduate feeder disciplines for economics by race/ethnicity (percent) 2017
Discipline Total Women Men Black Women Men Hispanic Women Men White Women Men Other Women Men
Mathematics and statistics 7 8 5 4 3 5 8 10 10 13
Economics 63 63 70 61 68 67 63 62 46 47
NonS&E fields 14 15 15 18 12 19 16 13 30 21
Source: Public use 2017 Survey of Doctorate Recipients
S&E other and related fields 14 13 11 16 16 8 12 13 14 12
N/A 1 2 0 2 0 1 1 2 0 7
Total 100 100 100 100 100 100 100 100 100 100
16
Race and Gender
415
Table 2 Economics undergraduate degree production: 1999–2018 Race, ethnicity, and gender Total Women Men Total Genderdifference Asian Female Male Subtotal Genderdifference Black Female Male Subtotal Genderdifference Hispanic Female Male Subtotal Genderdifference Native Female American Male Subtotal Genderdifference Other Female Male Subtotal Genderdifference Temporary Female residents Male Subtotal Genderdifference White Female Male Subtotal Genderdifference
1999–2008 71,872 149,237 221,109 77,365
2009–2018 94,676 210,351 305,027 115,675
Total 166,548 359,588 526,136 193,040
Yeardifference 22,804 61,114 83,918 38,310
Growth 32% 41% 38% 50%
14,737 20,195 34,932 5458
17,208 25,724 42,932 8516
31,945 45,919 77,864 13,974
2471 5529 8000 3058
17% 27% 23% 56%
5141 6686 11,827 1545
4889 9213 14,102 4324
10,030 15,899 25,929 5869
252 2527 2275 2779
5% 38% 19% 180%
4173 7832 12,005 3659
7299 16,272 23,571 8973
11,472 24,104 35,576 12,632
3126 8440 11,566 5314
75% 108% 96% 145%
272 552 824 280
252 603 855 351
524 1155 1679 631
20 51 31 71
7% 9% 4% 25%
3588 7684 11,400 4096
6477 14,178 13,892 7701
10,065 21,862 25,292 11,797
2889 6494 2492 3605
81% 85% 22% 88%
8082 11,852 19,934 3770
20,497 26,452 46,949 5955
28,579 38,304 66,883 9725
12,415 14,600 27,015 2185
154% 123% 136% 58%
35,879 94,436 130,315 58,557
38,054 117,909 155,963 79,855
73,933 212,345 286,278 138,412
2175 23,473 25,648 21,298
6% 25% 20% 36%
Source: Department of Education, National Center for Educations Statistics, Integrated Postsecondary Education Data System
416
R. V. Sharpe
earned for 2009–2019 vs. 1999–2008, the Hispanic gender gap more than doubled. For 2009–2018, White men earned three times as many degrees as White women. The nuanced differences are hidden in the aggregate race and gender data. The negative percentage growth from 1999–2008 to 2009–2019 for Black and Native American women is proof that gender does operate the same for all women. The gender gap within groups suggests ethnicity, nativity, and race do not operate the same across gender. Analysis of the 2019 Integrated Postsecondary Education Data System finds that 39% of the doctorates conferred were awarded to students who identified as US citizens or permanent residents. Women who are US citizens or permanent residents were 13% of doctorates conferred and men were 26% (see Table 3). This gender gap is seen for all race/ethnic groups except for Asians; women earned more doctorates. These data highlight the concern over the lack of Blacks and women pursuing degrees at all levels in economics (Sharpe 2019, 2020a). However, as conversations about diversity in the economics profession increase, there are still tattletale signs of racial and gender bias. For example, reports and panels about diversity in the economics profession talk about solving the “race” or “gender” problem but do not discuss diversity at the intersection of race and gender.2 Since economic research is used for civil rights cases and economists write amicus briefs, who gets a doctorate in economics has implications for US social and economic policy. The data presented in this section provide support for the generation of strategies that take an intersectional approach.
Conclusion Systemic exclusion is not just entrenched inequitable access to quality education, employment, healthcare, housing, and a host of other goods and services but is also rooted in how data are collected and reported (Sharpe 2019). An inclusive and equitable society requires data collection and reporting that center differences that have historically served as a basis for exclusion. Feminist, gender, and race scholars have long known that gender and race operate differently across and within groups. The American Anthropological Association (2021) has a statement about race: A report by Wessel et al. fails to disaggregate women or “minority” to report a count of economist by race, ethnicity, and gender. Wessel, D., Sheiner, L., Ng, M. Gender and Racial Diversity of Federal Government Economists Hutchins Center On Fiscal And Monetary Policy, Brookings Institution, September 2019, https://www.brookings.edu/wp-content/uploads/2019/09/Diversityreport_updated-3.pdf. The panel How Can Economics Solve Its Gender Problem? at the Allied Social Sciences Association meetings in Atlanta did not have a Black, Hispanic or Asian woman participant. January 5, 2019 Atlanta, GA, https://www.aeaweb.org/webcasts/2019/how-caneconomics-solve-gender-problem, 2019. The panel How Can Economics Solve Its Race Problem? mentioned very little about the ways Asian, Black or Hispanic women are treated in the economics profession, January 3, 2020, San Diego, CA, https://www.aeaweb.org/conference/2020/ preliminary/2264
2
16
Race and Gender
417
Table 3 Economics degrees awarded academic year 2018–2019 by degree level Race/ethnicity/residency Total Total men Total women Native American total Men Women Asian total Men Women Black total Men Women Hispanic total Men Women Race/ethnicity unknown total Men Women Temporary resident total Men Women Multiracial total Men Women White total Men Women Race/ethnicity/residency Total Total men Total women Asian total Men Women Black total Men Women Hispanic total Men Women Multiracial total
Award level (count) Associate’s Bachelor’s 2228 28,717 1342 19,567 886 9150 9 47 7 26 2 21 426 3365 228 2049 198 1316 42 1307 24 822 18 485 639 3038 402 2094 237 944 40 763 27 509 13 254 288 5345 142 3039 146 2306 96 950 57 641 39 309 688 13,902 455 10,387 233 3515 Award level (percent) Associate’s Bachelor’s 100% 100% 60% 68% 40% 32% 19.1% 11.7% 10.2% 7.1% 8.9% 4.6% 1.9% 4.6% 1.1% 2.9% 0.8% 1.7% 28.7% 10.6% 18.0% 7.3% 10.6% 3.3% 4.3% 3.3%
Master’s 3463 2020 1443 2 1 1 169 101 68 115 66 49 148 90 58 84 47 37 1973 1057 916 47 32 15 925 626 299
Doctorate 1122 739 383 0 0 0 67 32 35 12 8 4 25 17 8 34 26 8 687 445 242 13 8 5 284 203 81
Master’s 100% 58% 42% 4.9% 2.9% 2.0% 3.3% 1.9% 1.4% 4.3% 2.6% 1.7% 1.4%
Doctorate 100% 66% 34% 6.0% 2.9% 3.1% 1.1% 0.7% 0.4% 2.2% 1.5% 0.7% 1.2% (continued)
418
R. V. Sharpe
Table 3 (continued) Race/ethnicity/residency Men Women Native American total Men Women Race/ethnicity unknown total Men Women Temporary resident total Men Women White total Men Women
Award level (count) Associate’s Bachelor’s 2.6% 2.2% 1.8% 1.1% 0.4% 0.2% 0.3% 0.1% 0.1% 0.1% 1.8% 2.7% 1.2% 1.8% 0.6% 0.9% 12.9% 18.6% 6.4% 10.6% 6.6% 8.0% 30.9% 48.4% 20.4% 36.2% 10.5% 12.2%
Master’s 0.9% 0.4% 0.1% 0.0% 0.0% 2.4% 1.4% 1.1% 57.0% 30.5% 26.5% 26.7% 18.1% 8.6%
Doctorate 0.7% 0.4% 0.0% 0.0% 0.0% 3.0% 2.3% 0.7% 61.2% 39.7% 21.6% 25.3% 18.1% 7.2%
Source: Department of Education, National Center for Educations Statistics, Integrated Postsecondary Education Data System Yet, the concept of race has become thoroughly – and perniciously – woven into the cultural and political fabric of the United States. It has become an essential element of both individual identity and government policy. Because so much harm has been based on “racial” distinctions over the years, correctives for such harm must also acknowledge the impact of "racial" consciousness among the U.S. populace, regardless of the fact that “race” has no scientific justification in human biology. Eventually, however, these classifications must be transcended and replaced by more non-racist and accurate ways of representing the diversity of the U.S. population.
The disaggregation of data removes the racial bias created by analyzing the aggregate “women or men” with White as the norm or the gender bias created by analyzing race/ethnicity categories with men as the norm. When race/ethnicity and gender data are disaggregated, the result can better inform policies that increase economic and social well-being. It also acknowledges the complex identities of each of us, and it allows for the use of intersectional analysis, a theoretical framework for understanding how the complex interplay between various identities influences economic and social well-being. So how does a scholar conduct race and gender research? First, the scholar must start with data that allow for disaggregation by characteristics outlined in the literature to influence the outcome in question. Second, the scholar must think about their own lived experiences and of those in their social network. Third, the scholar analyzes the data disaggregated by these characteristics. Fourth, the scholar reports these findings with a keen eye for the nuanced differences in outcomes. Finally, the scholar interprets the findings using an intersectional approach.
16
Race and Gender
419
Cross-References ▶ Gender-Based Discrimination in Health: Evidence From Cross-Country ▶ Gender Quotas and Representation in Politics ▶ Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination ▶ Stratification Economics ▶ Transforming Gendered Labor Markets to End Discrimination
References American Anthropological Association (2021) AAA response to OMB directive 15: race and ethnic standards for federal statistics and administrative reporting. American Anthropological Association, February 1: https://s3.amazonaws.com/rdcms-aaa/files/production/public/ FileDownloads/pdfs/cmtes/minority/upload/AAA_Response_OMB1997.pdf Badgett MV (1995) The wage effects of sexual orientation discrimination. Ind Labor Relat Rev 48:726–739 Badgett MV (2006) Discrimination based on sexual orientation: a review of the literature in economics and beyond. In: Rodgers WM III (ed) Handbook of the economics of discrimination. E. ELGAR, London, pp 161–186 Bueno CC (2015) Stratification economics and grassroots development: the case of low–income black women workers in Santo Domingo, Dominican Republic. Rev Black Polit Econ 42:35–55 Cotton J (1993) Color or culture?: Wage differences among non-Hispanic black males, Hispanic black males and Hispanic white males. Rev Black Polit Econ 21:53–67 Darity W Jr (2005) Stratification economics: the role of intergroup inequality. J Econ Financ 29:144–153 Darity W Jr, Hamilton D, Stewart JB (2015) A Tour de Force in understanding intergroup inequality: An introduction to stratification economics. Rev Black Polit Econ 42:1–6 Davidson S (2016) Gender inequality: nonbinary transgender people in the workplace. Cogent Soc Scie 2:1–12 DiAngelo R (2012) Chapter 6: What Is Race? Counterpoints, 398:79–86. http://www.jstor.org/ stable/42981486 Edgar J, Phipps P, Kaplan R, Holzberg JL, Ellis R, Virgile M, Nelson DV (2018) Assessing the feasibility of asking about sexual orientation and gender identity in the current population survey. Retrieved from United States Census: https://www.census.gov/content/dam/Census/ library/working-papers/2018/adrm/rsm2018-02.pdf Goldsmith AH, Hamilton D, Darity W Jr (2007) From dark to light skin color and wages among African-Americans. J Hum Resour 57:701–738 Guis M (2013) The effect of interracial marriage on individual-level earnings: an analysis using 2010 PUMS data. J Appl Bus Econ 14:37–41 Gyimah-Brempong K, Price GN (2006) Crime and punishment: and skin hue too? Am Econ Rev 96:246 Hamilton D, Goldsmith AH, Darity W Jr (2009) Shedding “light” on marriage: the influence of skin shade on marriage for black females. J Econ Behav Organ 72:30–50 Isenberg N (2016) White trash. The 400 year untold history of class in America. Viking, New York Jacobsen JP, Newman AE (2003) Do women and non-economists add diversity to research in industrial relations and labor economics? Eastern Economic Journal 575–591 Jenks AE (1916) The Legal Status of Negro-White Amalgamation in the United States. Am J Sociol 21:666–678 Kim M (2020) Intersectionality and gendered racism in the United States: a new theoretical framework. Rev Radical Polit Econ 52:616–625
420
R. V. Sharpe
Kreisman D, Rangel MA (2015) On the blurring of the color line: wages and employment for Black males of different skin tones. Rev Econ Stat 97:1–13 Lopez N (2018) The US Census Bureau keeps confusing race and ethnicity, February 28. Retrieved from https://theconversation.com/the-us-census-bureau-keeps-confusing-race-and-ethnicity89649 Malveaux J (1985) The economic interests of black and white women: are they similar? Rev Black Polit Econ 14:5–27 Mason PL, Myers SL Jr, Darity WA Jr (2005) Is there racism in economic research? European Journal of Political Economy 755–761 Mora MT, Dávila A (2018) The Hispanic–white wage gap has remained wide and relatively steady, July 2. Retrieved from Economic Policy Institute: https://files.eric.ed.gov/fulltext/ ED593384.pdf Office of Management and Budget (1997) Revisions to the standards for the classification of federal data on race and ethnicity. Retrieved from Executive Office of the President, Office of Management and Budget (OMB), Office of Information and Regulatory Affairs, October 30. https:// www.govinfo.gov/content/pkg/FR-1997-10-30/pdf/97-28653.pdf Painter NI (2010) The history of white people. Norton, New York Park A (2016) Reachable: data collection methods. Retrieved from The Williams Institute, March: https://williamsinstitute.law.ucla.edu/wp-content/uploads/SOGI-Data-Collection-Mar-2016.pdf Price GN (2010) Evolution, green beards, and skin hue wage discrimination. World Futures: J Gen Evol 55:341–355 Reece RL (2016) What are you mixed with: the effect of multiracial identification on perceived attractiveness. Rev Black Polit Econ 43:139–147 Reece RL (2018) Genesis of U.S. Colorism and skin tone stratification: slavery, freedom, and mulatto-black occupational inequality in the late 19th century. Rev Black Polit Econ 45:3–21 Reimers CW (1983) Labor market discrimination against hispanic and black men. Rev Econ Stat:570–579 Rico BK (2018) Growth in interracial and interethnic married-couple households, July 9. Retrieved from United States Census Bureau: https://www.census.gov/library/stories/2018/07/interracialmarriages.html Robst J, VanGilder J, Coates CE, Berri DJ (2011) Skin tone and wages: evidence from NBA free agents. J Sports Econ 12:143–156 Rosenblum A, Darity W Jr, Harris AL, Hamilton TG (2015) Looking through the shades: the effect of skin color on earnings by region of birth and race for immigrants to the United States. Sociol Race Ethnicity 2:87–105 Ross L (2015) Here’s a much-needed history lesson on the origins of the term ‘Women of Color’. Retrieved from Every Day Feminism. https://everydayfeminism.com/2015/03/origin-of-termwoc/ Saunders L, Darity W Jr (2003) Feminist theory and racial economic inequality. In: Nelson JA, Ferber MA (eds) Feminist economics today: beyond economic man. The University of Chicago, Chicago, pp 101–114 Schneebaum A, Badgett MV (2019) Poverty in US lesbian and gay couple households. Fem Econ 25:1–30 Sharpe RV (2019) We’ve to build the pipeline. What’s the problem? What’s next? The Remix. Rev Black Polit Econ 45:191–215 Sharpe RV (2020a) Black women economists: at the intersection of race and gender. In: Lundberg EB (ed) Women in economics. https://voxeu.org/content/women-economics Sharpe RV (2020b). Category error. Retrieved from 2020 gender balance index: driving diversity, February. https://www.omfif.org/gbi2020/ Sharpe RV, Swinton OH (2012) Leaks or alternatives: an examination of the pipeline for black economists. Allied Social Science Association Meeting, Chicago
16
Race and Gender
421
Simms MC, Malveaux JM (1986) Slipping through the cracks: the status of black women. Transaction Publishers, New Brunswick Wallace PA (1986) A research agenda on the economic status of black women. In: Malveaux MC (ed) Slipping through the cracks: the status of black women. Transaction Publishers, New Brunswick, pp 293–295 Wu AH (2017) http://calwomenofecon.weebly.com. Retrieved from Gender Stereotyping in Academia: Evidence from Economics Job Market Rumors Forum: http://calwomenofecon.weebly. com/uploads/9/6/1/0/96100906/wu_ejmr_paper.pdf Yavorsky JE (2016) Cisgendered organizations: trans women and inequality in the workplace. Sociol Forum 31:949–969
Caste and Gender
17
Kalpana Kannabiran
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Situating Intersectionality in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Abrahmani-Gender Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Hostile Environments” as an Intersectional Lens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
424 425 428 432 435 437 437
Abstract
Intersectionality is a praxiological tool that helps understand complex and cumulative discrimination along several interlocked and coproduced axes of power, domination, and hegemonies. On the Indian subcontinent, the genealogy of this concept may be traced back to the resistance to the caste-gender complex especially as part of anti-caste movements from the late nineteenth century. The challenge to Brahmanical supremacy by Phule, Savitribai, Tarabai, Ambedkar, and Periyar among a host of others, notably women in Ambedkarite and selfrespect movements, was predicated on freeing women from the thraldom of patriarchal family and kinship practices. This resistance has a continuing and cascading presence. In opening out of the field of intersectionality through translocational positionalities and to transnational and diasporic contexts, this chapter investigates the proliferation of caste discriminatory practices and exclusions in culturally and historically rooted ways in different locales. A key aspect of the caste-gender complex is the deployment of deeply embedded practices of structural violence and atrocity – hostile environments – especially sexual assault and humiliation, which may only be grasped adequately through an intersectional approach. The constitution of India, in setting out the nondiscrimination proK. Kannabiran (*) Independent Sociologist, Hyderabad, India © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_30
423
424
K. Kannabiran
tections, provides the possibility for an elaboration of justice, mindful of the long history of combatting the realities of caste through intersectional approaches to resistance. Keywords
Hostile environments · Bhanwari Devi · Savitribai Phule · Endogamy · Abrahmani-gender complex · Analogous discrimination · B.R. Ambedkar · Radhika Vemula
Introduction The constitution of India guarantees the fundamental right to nondiscrimination on grounds “only of race, caste, sex, place of birth, or any of them.” An important part of this guarantee is the right of access to places of public worship and public resort (Article 15 (1) & (2)). This constitutional delineation provides a point of departure to examining closely the relationship between the different intersecting grounds set out therein – religion, race, caste, sex, place of birth, or any of them. The chapter does not dwell at length on the analogous discrimination against minorities in relation to the Hindu social order, especially Muslims (apart from discrimination on grounds of religion), and against LGBTQI communities. However, the possibilities for deployment of the lens of intersectionality to speak to discrimination and structural violence against religious and sexual minorities is underscored – taking the cue from B.R. Ambedkar’s inclusive approach to the minority question, which hinged on Hindu majoritarian dominance and exclusions rather than on the identity of specific social groups. Further, “the precipitation of conflict through food cultures around beef, the circumambulations around the cow, the rise in practices and technologies of stigmatization and the spiralling collective, targetted, aggravated violence [hate crimes] deployed through a convergence of state and private actors, fold the Dalit and Muslim questions together even as they remain separate and distinct. . .” In this context, the struggles of Dalit women against the majoritarian caste order take on a different meaning altogether, as witnessed in the struggles of Radhika Vemula (Kannabiran 2016). Reservations, Dalit associational freedoms, boycott, atrocity, state violence, and judicial abdication present a complex gendered cascading of Dalit resistance and the reinvention of deeply insurgent motherhood by Radhika Vemula that challenges the Hindutva state, through tenacious public protest and conversion to Buddhism in the face of grief and loss. This chapter presents a review of writing on intersectionality that is relevant to an understanding of the caste-gender complex. As a foundational text, Ambedkar’s 1917 classic Castes in India (2002) is unparalleled. The sections that follow (i) present a brief review of writing that illuminates “intersectionality,” caste, and gender; (ii) recall the reinvention of the Abrahmani (anti-caste)-gender complex; (iii) explore the intersectional possibilities of “hostile environments”; and (iv) offer some concluding observations. The aim of this chapter is to open out the possibilities of engaging theoretically and on the ground with
17
Caste and Gender
425
interlocking systems of discrimination – cumulative discrimination and structural violence – that are powered by the caste-gender complex in the Hindu social order.
Situating Intersectionality in India Intersectionality is a praxiological tool that helps understand complex and cumulative discrimination along several interlocked and coproduced axes of power, domination, and hegemonies. Black women began to use intersectionality as a tool when any linear analysis along a single axis proved inadequate to explain or understand their specific problems (Crenshaw 1991). In her early attempt to set out the signposts for theorizing intersectionality as intersectionality, Crenshaw (1991) explored “the various ways in which race and gender intersect in shaping structural, political, and representational aspects of violence against women of color.” Focusing on the racegender complex, for Crenshaw, served to highlight “the need to account for multiple grounds of identity when considering how the social world is constructed,” providing an elaboration of structural, political, and representational intersectionality. While later scholarship has certainly opened out the field to transnational contexts (Purkayastha 2012) and the Indian diaspora (Reddock 2014) and provided opportunities for a recalibration of the tool to the specificities of contexts in the Global South (Collins and Bilge 2016), Crenshaw’s work provides a critical moment in feminist theorizing on the politics of resistance underscoring the gender-race-sexuality complex in the United States of America. Gender (in nonbinary terms), at intersectionality’s moment of emergence in its present assemblage, was co-constitutive of the concept with race, as was gender-based violence. Emphasizing the point that Crenshaw deployed “intersectionality” as a social justice construct, Collins and Bilge (2016) trace the genealogy of this concept to the early resistance against white supremacy in the United States and point to several instances in the Global South where this tool has been used without naming it as such – in an attempt to “democratize the rich and growing literature on the subject” (2016). The first such instance of Savitribai Phule strikes at the heart of the caste-gender complex. In her article titled “Six Reasons why every Indian must remember Savitribai Phule,” Deepika Sarma writes: Here’s why you should know more about her. She got intersectionality.. . . She organized a barbers’ strike against shaving the heads of Hindu widows, fought for widow remarriage and in 1853, started a shelter for pregnant widows. Other welfare programmes she was involved with alongside Jyotirao include opening schools for workers and rural people, and providing famine relief through 52 food centers that also operated as boarding schools. She also cared for those affected by famine and plague, and died in 1897 after contracting plague from her patients. (Sarma 2015, cited in Collins and Bilge 2016)
Tarabai Shinde’s 1882 text Stri Purush Tulana (A Comparison between Women and Men) (1994), also part of the Satyashodhak tradition, is a second illuminating instance of early Indian writing confronting Brahmanical patriarchy as well as patriarchy within non-Brahman castes.
426
K. Kannabiran
B.R. Ambedkar’s sociology of caste, as set out in Castes in India, sees caste and gender as co-constitutive and predicated on extreme forms of violence and “rituals of humiliation” (Boopalan 2017) of women by men – the makers of injunctions. While the caste system, in his view, was codified by Hindu lawgivers, it was not divinely ordained and therefore not above reproach. The centrality of the caste-gender complex to Ambedkar’s theory of the origin and perpetuation of caste is expressed through his painstaking elaboration of the institution of heterosexual endogamous marriage within Hinduism. Rigid rules of exogamy that prohibited marriage between sapindas (blood kin) and sagotras (clansmen) were enforced through an elaborate system of punishments for infringements. Over this exogamous society is overlaid the principle of endogamy – distinct from racial or tribal endogamy where the group is large and culturally homogenous. In the case of caste, Ambedkar points out a homogenous group is partitioned into smaller rigidly endogamous units, which contain within them rigidly exogamous subunits. This “superimposition of endogamy on exogamy means the creation of caste” (Ambedkar 2002). In order to reconcile two opposing principles of kinship organization and to ensure a balance in the sex ratio in marriageable cohorts, a web of rules governing marriage was instituted. In Ambedkar’s words, “the problem of caste, then ultimately resolves itself into one of repairing the disparity between the marriageable units of the two sexes within it” (Ambedkar 2002). Given the practical difficulties in ensuring parity in sex ratio and the problem of dealing with “surplus women” (when male mortality was higher) and “surplus men” (when more women died within the marriageable cohort), caste society devised three ways of dealing with skewed sex ratio. In the case of surplus women, it was possible to burn her on the funeral pyre of her husband (a difficult proposition, but proven to be possible); it was somewhat simpler to subject her to enforced widowhood, although her allure would pose a threat to the morals of the group. In a double move, therefore, enforced widowhood was accompanied by stripping the widow bare of “anything that might be considered a source of allurement” (Ambedkar 2002) – social death. The third solution was aimed at addressing the problem of surplus men, who could not be burnt on the pyre of the wife, “simply because he is a man” (Ambedkar 2002) – and in a patriarchal society where the man wields authority and is the “maker of injunctions” (Ambedkar 2002), he could not be condemned to celibacy either. The only solution therefore was to find a wife from a younger cohort. Sati, enforced widowhood, and girl marriage were the three uxorial mechanisms through which “endogamy, and by extension caste, is preserved and perpetuated” (Ambedkar 2002). Castes always exist in plural, as the condition for the existence of the caste order is exclusion of groups. Further, enclosure and endogamy face the perennial threat of violation or innovation, both of which cannot be tolerated. There is therefore an elaborate gradation of offences and penalties from violating marriage codes – especially excommunication/boycott – which ensure the formation of new castes (Ambedkar 2002) and ensure the subjugation of women. For as Periyar observed, “[j]ust as Brahminism condemns a very large portion of the working population to shudrahood, so it has condemned women to the servitude of marriage” (cited in Geetha 2003).
17
Caste and Gender
427
The control of female sexuality was the primary tool of surveillance in caste society. As Uma Chakravarti points out in the context of the Peshwai, “[w]hile the sexuality of all women was monitored. . .differing norms were applied to them according to caste with the most rigorous code being reserved for Brahmana women” (Chakravarti 2003). The argument of a gradation of codes for different caste-gender assemblages with differing distribution of power, female autonomy, and patriarchal male control has been made in different contexts, suggesting the need for close attention to the intricate workings of caste-gender complexes. One expression of this is the argument on the “existence of caste-specific patriarchies,” which even “while they may be inflected by the hegemonic interests of a Brahmanical structure” on matters of female sexuality and worth, “cannot be reduced to movements of upward mobility or imitation” (Rao 2003). While Maratha women provide one illustration of this argument, in another context, a close reading of kinship, marriage, and female autonomy among the fisher community, the Mukkuvar in Tamil Nadu, points to a very different crafting of the caste-gender complex. Located at a remove from both Brahman and non-Brahman categories, Christianity among Mukkuvars makes little difference to caste formation secured by rigid marriage prescriptions. In this low-ranking service caste with a “closed” cultural universe, the control by women of not just the hearth but economic transactions and clan affairs as well, notes Kalpana Ram (1992), calls for a recasting of the neat divisions in the debates on separate spheres and the sexual division of labor. Ambedkar outlined an anti-caste praxis that pulled together the historical/genealogical, the existential, and the macrostructural materialist conditions of exploitation that power the deep oppressions of Dalits in Hindu society – pointing specifically to the gender regimes of caste. The political project of annihilating caste and his archaeological method in setting about this task saw Ambedkar reclaim the commons for Dalits (Chavdar tank) by confronting untouchability; proclaiming a resistance against Hindu religion that erected this barricaded social order by publicly setting fire to the Manusmriti; and assembling a different spiritual-intellectual tradition based on fraternity (fellowship) as the core value undergirded by dignity and equality. His purpose in scripting the 22 vows of Navayana Buddhism was to put in place a dialogic context and fundamental precepts tethered to reason and compassion (Rathore and Verma 2011). With respect to conversion to Christianity, Boopalan observes that the translation of the Bible to vernacular languages saw many Dalits embracing Christian scripture, which for them was a move to a proximate spirituality and literacy that was denied them under Hinduism (2017). Sanal Mohan’s (2015) poignant account of Lucy’s conversion spins the caste-gender complex to the center of our consideration of intersectionality, bringing together Lucy’s simultaneous location in transnational contexts – missionary and slave trade, within the Hindu caste order’s slavery practices (Ambedkar 2019), and in the liberatory contexts of conversion for a Dalit woman: Kali. . .was purchased by a European gentleman in 1827 and. . .was supposed to accompany him and his family to Java. She pleaded to the missionaries. . . in Cochin to let her into the mission compound where she had reached from her hideout since she did not want to leave her land forever. . .She was then baptized as Lucy in 1828, who – according to the missionaries ‘continued to grow in grace.’
428
K. Kannabiran
The interillumination is translocational and immersed in plurality. Identity, belonging, and becoming undergirded by gender are central to the construction of selfhood. Savarna men in the South Asian diaspora in the United States appropriated a racialized identity to claim kinship on the basis of whiteness/ Aryan blood, and discrimination against Dalit South Asians in the United States within the South Asian immigrant community is commonplace (Zwick-Maitreyi et al. 2018); Hindu Indo-Trinidadians “still had a negative perception of intermarriage and indeed often treated Douglas as ‘creoles,’ still associated with low castes and polluting in relation to food, marriage, and so forth” (Reddock 2014); in India especially, as this essay attempts to show, the creation of Dalit selfhood was predicated on a rejection of caste, indeed its annihilation. The shifts in economic formations, political regimes, and the interiorities of Dalit lives trigger shifts in practices of discrimination, violence, and resistance – with the fashioning of new weapons and modalities for untouchability-atrocity and new technologies of rule. Floya Anthias (2008) suggests the deployment of the lens of “translocational positionality” that helps move away from categories/groups (caste, gender, class) toward “social locations and processes” and advocates a shift in focus from “identity” to “belonging” which is a relational concept. Grounding her analysis in contexts of migration (itself a “multiplex reality”), she suggests thinking about belonging and identity in a “translocational frame which recognises that people have multiple locations, positions and belongings, in a situated and contextual way” (2008). If examined with reference to the “South Asian” diaspora in the United States, for instance, this category of 4.3 million people includes people with different nationalities, religious affiliations, and importantly caste affiliations having migrated from Nepal, Bangladesh, Pakistan, Afghanistan, Sri Lanka, Bhutan, Tibet, Maldives, and the Caribbean (Zwick-Maitreyi et al. 2018). The framework of translocational positionality helps think through racialized, ethnic, religious, and caste-based dislocations (through class, employment status, practices of segregation, dispossession, detention in camps, and ghettoization, for instance), as well as translocations – across gender plurality. The notion of identity and the politics of belonging have been further extended to an exploration of the “politics of becoming” with specific reference to Dalit women’s political mobilization on the caste-race intersection (Kannabiran 2006), using the work of William Connolly (1996) and Martha Minow (1996).
The Abrahmani-Gender Complex Historically, the caste-gender complex was shaped by several positionalities, each of which either overtly or implicitly interrogated the interlocked oppressions of caste and gender and all of which interrogated the oppression of women in Hindu society. Two of these strands that struck at the patriarchal basis of conjugality are elaborated in this section. The first strand interrogated women’s subjugation within the family – child marriage, enforced widowhood, sati – without overtly interrogating caste. Yet,
17
Caste and Gender
429
the location of these women and the problems they spoke of within savarna society made caste constitutive of the problem of patriarchal conjugality they were resisting – their critique therefore was a critique of the interiorities of savarna society and the confinement of women within. Abrahmani, non-Brahmin, or anti-caste traditions across the subcontinent launched a social critique, interrogating the basis of caste and religion through a resistance against oppressive conjugal arrangements. This is the second strand. The framing of the caste-gender complex in these two different strands was distinctive, although both focused on the question of women’s dignity. Of the early campaigners, Ramabai’s critique of the (savarna) Hindu family was sharp, as was Rukhmabai’s (Chakravarti 1989). The issues reformers raised in the late nineteenth and early twentieth centuries – female infanticide, child marriage and the age of consent, enforced widowhood, and sati – were assembled into a cogent theoretical critique of the caste system by Ambedkar (2002). The resistance to familial practices of subjugation was important also as a critique of British colonialism and the imperial equivocation on the oppression of women in Hindu society (Mani 1990). When Ramabai found that all the 300 infants reported to have been carried away by wolves in Amritsar were girls, she retorted sharply, “even the wild animals are so intelligent and of such refined taste that they mock at British law, and almost always steal girls to satisfy their hunger” (Chakravarti 1989). When in 1887 Rukhmabai was ordered to return to her husband although she had challenged her marriage contracted without her consent during infancy, she wrote to Ramabai, “The learned and civilized judges. . .are determined to enforce, in this enlightened age, the inhuman laws enacted in barbaric times, four thousand years ago. . .There is no hope for women in India, whether they be under Hindu rule or British rule. . .The hard hearted mothers-in-law will now be greatly strengthened and will induce their sons to sue the wives in British courts since they are now fully assured that under no circumstances can the British government act adversely to the Hindu Law” (Chakravarti 1989). In the Satyashodhak tradition, Tarabai Shinde, writing in 1882, angered at the death sentence awarded to a young widow for killing her infant child, asked “[W]hat does stridharma really mean? It means always obeying orders from your husband and doing everything he wants. He can kick you and swear at you, keep his whores, get drunk, gamble with dice and bawl he’s lost all his money, steal, commit murder, be treacherous, slander people, rob peoples’ treasures or squeeze them for bribes. He can do all this, but when he comes home, stridharma means women are meant to think, ‘Oh, Who’s this coming now but our little lord Krishna, who’s just stolen the milkmaids’ curds and milk and. . ..then smile at him and offer their devotion, stand ready at his service as if he was Paramatma himself. But how can people go on believing this idea of stridharma once they have begun to think about what’s good and bad? They’d change their ideas straightaway, won’t they?” (Shinde 1994). Jyotiba Phule, writing in the Satyashodhak tradition, locates enforced widowhood within the caste-gender complex of Aryan (Brahmanical) society, which enslaved women through a range of conjugal practices. In each of these articulations, the
430
K. Kannabiran
question of race is never absent from the reckoning of the caste-gender complex. Writing on “the most delicate subject of enforced widowhood upon Brahmin women,” Phule says, “The partial Aryan institution inconsiderately allows polygamy to males, which causes them to fall into new habits of wickedness. When his lust is satisfied with his legal wives, he for novelty’s sake haunts the houses of public women. . ..In old age in order to obliterate the stigma upon his character, the shameless fellow becomes a religious man and hires public harlots to dance and sing in the temples with a view to venerate the stone idols, for his own satisfaction. After the death of this wicked man, his young and beautiful wife is not allowed by the same Aryan institution to remarry. She is stripped of her ornaments; she is forcibly shaved by her near relatives; she is not fed well; she is not properly clothed; she is not allowed to join pleasure parties, marriages or religious ceremonies. In fact she is bereaved of all the worldly enjoyments, nay she is considered lower than a culprit or a mean beast” (Phule 2002). Intermarriage across barricaded caste lines was not uncommon, as is evident from Savitribai Phule’s account of her rescue of a Mahar woman who had married a Brahmin man and was being stoned by villagers. Sending them to safety to Jyotiba in 1868, she wrote, “When I got to know of this horrible incident, I rushed there and stopped the cruelty by scaring them with possible action by the British government. . .I am sending the couple to you” (cited in Rege 2006). While the “low-caste” woman’s body is sexually violable and commodified in the dominant caste-gender complex, marriage or the conferment of social approval/legitimacy on such unions was clearly not tolerated. Periyar’s rejection of “rituals of humiliation” (Boopalan 2017) embedded in savarna conjugality, and his insistence on mutuality and equality in conjugal relationships was part of his larger philosophy of self-respect and the rejection of caste (Geetha and Rajadurai 1998). With Ambedkar as well, the fusion of politics and conjugality, in which, as Rege argues, there is a prioritization of community and politics over conjugality, “stands out from the middle class discourse of the educated ‘companionate’ conjugality that alienated women from the community without granting autonomy” (Rege 2006). V. Geetha (2003) points out that women selfrespecters addressed conferences and wrote in journals opposing caste and the horrors it perpetrated, but importantly they saw caste “as not merely a division of labor and laborers but as a system that divided women as well.” The assembling of the caste-gender complex and the place of the family in this assemblage is distinctive in the case of Dalit women; crafting an Abrahmani epistemological tradition braided the everyday with the political, the private with the public, and the conjugal with the pedagogical. The Abrahmani-gender complex is insurgent at several levels, as Sharmila Rege demonstrates through her work. Education for girls was the cornerstone of this new imagery. It is useful to recall Jyotiba Phule who established two schools – the first for Shudratishudra girls in 1848 and the second for all girls in 1851 (Deshpande 2002). The Abrahmani shift in the positionality of women in the family reinvents the approach to the public through the medium of education – not as a gift for a more companionate wifehood but as a utopia that promises a new Dalit selfhood – one that
17
Caste and Gender
431
ousts the defining presence of the savarna. Lullabies immersed infants in Babasaheb Ambedkar’s epic courage refashioning the meanings of motherhood – labor, the necropolitics (Mbembe 2019) of caste, and the promise of equality: Sleep well O little one The whole world lies in the shadow of death, the lives of your parents were wasted, ... I’ll feed you doses of your Baba’s courage, The dark night of the enemy has passed, it’s dawn my dear child, In your baby talk sings the bird of equality. (R.D. Gaikwad cited in Rege 2006)
The genealogy of this materialist rendering of the lullaby may be traced back to the political resistance by the first generation of Dalit women in the Ambedkarite movement against the violence of social exclusion – of which carceral labor and the denial of education were at the center – and the annihilation of caste, the goal (Pawar and Moon 2008). Shantabai Dhanaji Dani’s recalls her mother’s oft-repeated statement: “Listen girl, for the poor education is a ray of hope” (Rege 2006: 94–95). The stories of the Varkari saints; the bonding together of spouses, siblings, parents, and children in the face of aggravated caste violence and the perennial anticipation of such violence; and the tenacious, single-minded affirmation of dignity in the face of family violence and caste violence by women presented possibilities for a selfhood that was deeply insurgent and positioned in radically different terms. When Mukta Sarvagod started schooling as a 5-year-old, she was seated at a distance and prohibited from making any physical contact with her teacher and not permitted to help herself to water to clean her slate; she remembered Dalit children who were withdrawn from school by parents unable to bear battery and abuse by savarna teachers. Outside school, her father recounted stories of not being allowed to organize wedding or festival processions, of being given rotten grain in lieu of wages, and of eating bhakris made of greens instead of flour, keeping them perpetually in a state of hunger. Patole underscores this in his account of Dalit cuisine that was born from the historical experience of scarcity and hunger. His invitation to sample the recipes in his book is tempered by the caution that “It is an acquired taste, especially one that has been acquired due to centuries of discrimination” and that “most of [the recipes] do not need oil. Why is that? Because, Dalits could not afford oil, and that is why we have dishes such as the steamed mutke, which is made out of jowar, garlic and coriander. Instead of oil we used beef fat. Then, there is the famous Maharashtrian dish, the rich puran poli, but the Mangs used to substitute ghee with buttermilk for obvious reasons.” Like Muktabai’s father’s retelling of bhakris made of greens, Patole observes that deprivation and starvation over centuries made Dalits “expert foragers” (wild leafy vegetables, bee larvae, and pumpkin leaves) and the practice of “nose-to-tail eating” (Patole 2016). Culinary habits are dense with caste-gender codings within the family, within caste, and in relation to other castes relative to positioning within the hierarchy – the gradation of the body of the animal; the gradation of meat (animal, whether
432
K. Kannabiran
slaughtered for consumption or carrion, the gradation of parts of the animal body); the gradation of vegetables by caste, gender, and marital status (what if at all might a widow eat?); the sequence in which people eat and who may sit together to dine (Kannabiran 2012); and importantly, can the “chuhri” dare to ask for fresh cooked food for her children from “chowdhriji” after slaving for a wedding feast? – to recall Omprakash Valmiki’s Joothan (2007). From the caste-gender complex to state regimes primed by “rules of disgust,” which Redding (2018) defines as “legalized operation of the political emotion of disgust – or put another way, the ‘rule of disgust’” – a cartography of food lays bare the intersectionality of gender-caste, religion, location, politics, choice, coercion, and taste and frames the nationalantinational in binary terms. To return to Muktabai’s story, while there were poor Kunbi families in her village that depended on daily wage labor, the difference was in the fact that a Kunbi woman was never given stale food as wages. Shantabai Kamble’s mother told her stories of Mahars during the plague who were so preoccupied with burying the dead that they could not find time to eat. And when Shantabai returned to her parents after her husband married a second time, her father fended off neighbors who asked that she be sent back saying, “If the girl stays on by herself, well and good – or else let her go wherever she wants to – we are not Brahmans you know” (cited in Rege 2006, emphasis added). Here as well, education was key. She dedicated her book “To Aaye-Appa (mother and father) who worked the entire day in the hot, glaring sun, hungry and without water, and through the drudgery of labour, with hunger pinching their stomach, educated me and brought me from darkness to light” (Rege 2006). As Bama says, “[o]ppressed, ruled, and still being ruled by patriarchy, government, caste, and religion, Dalit women are forced to break all the strictures of society to live” – roaring their defiance and mocking their oppressors, finding through this the courage to revolt (Bama 2005). This knitting together of worlds of labor, economic and social oppressions, hunger, and deprivation in an account of a casteless utopia in Abrahmani traditions in women’s accounts of lives as lived underscores two important aspects of this everyday resistance: the first, the autonomy of women within the family and without, and second, the recalibration of familial gender relations resulting in radically different definitions of the ideal man – father, husband – as someone who must shift from a position of authority to a position of compassion and conviviality.
“Hostile Environments” as an Intersectional Lens Boopalan (2017) approaches the caste-race complex through the lens of wrongs, making four claims: their ritualistic character, their occurrence as aggravated episodes and everyday violence of normal times, their reproduction of violent identities, and the corporeality of wrongs – at the level of the individual body and the socialities that assemble individual selves for the perpetration of wrongs. A close reading of the Scheduled Castes and Scheduled Tribes (Prevention of Atrocities) Act, 1989, and a close examination of the aggravated assaults on Dalit women demonstrate the deeply gendered character of dominant caste wrongs; relations within castes as well were
17
Caste and Gender
433
built on gender wrongs – which, to reiterate Ambedkar’s reasoning, was at the base of the caste system. Bama’s statement, “To bounce like a ball that has been hit became my deepest desire, and not to curl up and collapse because of the blow” (Bama 2000), epitomizes the resistance of Dalit women to the violence of caste. Women’s sexualized bodies are embedded in the Brahmanical caste order in specific ways for Dalit women and savarna women. The latter are arrayed on a scale of untouchability – i.e., the transient/periodic/contingent untouchabilities of the “touchable” savarna woman (the menstruating woman, the new mother, and the widow being sources of pollution). This is distinct from the “untouchable” sexually violable body of the Dalit woman whose position is fixed by their caste location and immutable. This difference, rendered cognizable through an intersectional lens of caste-gender, is fundamental to the caste order. The elaboration of the heteronormative Hindu social order also fits “eunuchs” on this scale as untouchable and stigmatized bodies – riveting gender plurality to the elaboration of intersectionality with reference to caste. Yet despite the stigmatization rooted in transphobia, the caste location of the “eunuch” is material to the experience of un-touchability/ untouchability, as the case may be. Gender in a patriarchal caste order is “defined and structured in such a manner that the ‘manhood’ of the caste is defined both by the degree of control men exercise over women and the degree of passivity [and complicity] of the women of the caste. By the same argument, demonstrating control by humiliating women of another caste is a certain way of reducing the ‘manhood’ of those castes” (Kannabiran and Kannabiran 1991). The caste-gender complex finds its most cogent elaboration in the Scheduled Castes and Scheduled Tribes (Prevention of Atrocities) Act, 1989, and in the amendment to this act in 2015 in the listing of “rituals of humiliation” (Boopalan 2017) now defined as atrocity. The inclusion, through an amendment, of the crimes of economic and social boycott of Dalits sees an infusion of Ambedkar’s concerns into the text of the legislation, as also the recognition of the continuing savarna histories of punitive segregation of dissenting Dalits. The rituals of humiliation and violence proscribed by law include garlanding or parading naked a Dalit person; using force to assault the bodily integrity of a Dalit person, by removing clothes, tonsuring, removing moustaches, painting face or body, or any other similar assault; forced performance of the degrading labor of manual scavenging (which has a predominantly female labor force); and words, gestures, and physical contact of sexual nature without the consent of the woman. Sexual violence and humiliation as evident from this expanded definition of atrocity subjugate male and female bodies in distinct, comparable ways – making the exercise of power and domination contingent on sexual appropriation. The juridical concept of “hostile environments” has its beginnings in India in the case of Vishakha vs. State of Rajasthan (1997) that provided a measure of redress to women against sexual harassment at the workplace. Intersectionality is constitutive of this ruling by the Supreme Court of India, based as it is on the aggravated sexual assault on Bhanwari Devi in Rajasthan, by dominant caste landlords (for a detailed discussion, see Kannabiran 2012). Although there is an amnesia with respect to the originary moment of this concept in the caste-gender complex, it helps understand
434
K. Kannabiran
structural violence and oppression in deeply intersectional ways. Legal cases on caste atrocity that are deliberated upon in courts are often stonewalled by judicial amnesia with respect to atrocities suffered by Scheduled Castes and Tribes and the historical/constitutional background of the SC & ST Prevention of Atrocities Act and its continuing and increasing relevance today. In the process of attempting to wipe out judicial memory on the question of atrocities and structural violence based on caste discrimination, courts have oftentimes taken easy recourse to the language of public morality, setting aside the centrality of constitutional morality to constitutional reasoning. Rhetoric around the “fear” of “false accusations” and “allegations” against “innocent people,” who are at the mercy of “unscrupulous” elements out to trap and “stigmatize” them through arrest and denial of bail by filing cases of atrocity against these “unsuspecting” persons, misrepresents dominant caste dynamics by portraying savarna men in positions of authority/power either as public servants or employers as “unsuspecting,” “innocent” persons. The “unscrupulous elements” are Dalits in precarious employment, if at all. In one case, for instance, the repeated use of terms – like “malicious complaint,” “malicious prosecution,” and “malicious complainant,” speaking of how the act was used “as an instrument of blackmail” to wreak “personal vengeance” and settle scores arising from “personal vendetta,” by “scheming, unscrupulous complainant” – leads to the conclusion that the “above judgments are merely illustrations to show that the abuse of law was rampant” (SK Mahajan 2018). The continuities between courts and public spaces outside are stark. As late as 2015, a public university officially declared a social boycott of Dalit students, soliciting the active intervention of the ruling Bharatiya Janata Party and its student body, leading to the suicide in January 2016 of one of the students boycotted – Rohith Vemula; in March 2016, Shankar, a 21-year-old Dalit man, was hacked to death in broad daylight in Udumalaipettai in Tamil Nadu, and his non-Dalit wife Kausalya murderously attacked by her family for violating caste codes. In April 2016, a young Dalit student in Kerala, Jisha, was sexually assaulted and brutally murdered in her own home, a 150 sq. ft. house in a canal poramboke (common lands) – and there were, mysteriously, no eyewitnesses, nor did anyone attempt to rescue her. In 2018, Dalits in Vadayampadi near Kochi in Kerala were resisting the building of a “caste wall” by the dominant castes, to prevent their entry into common spaces in the village. Within these carceral enclosures of caste, the assault on Dalit women takes on a different meaning. A 19-year-old Dalit woman in Hathras in Uttar Pradesh was assaulted in the most brutal manner and killed by Thakurs from the same village known to her and her family. When her mother found her, she was naked and paralyzed with indescribable physical injuries which included deep injuries on her tongue – to silence her from testifying to the attack on her? Yet, in spite of the grave harms, violent humiliation, and deep trauma, she made a statement naming the perpetrators—she, a young Dalit woman from a landless Valmiki family, whose only protection against chronic anticipated assault and humiliation by the Thakur perpetrators and their henchmen was to stay indoors, never alone, and step out occasionally, never alone. Yet, even this did not give her the guarantee of bare life. Incarceration by caste and annihilation
17
Caste and Gender
435
by caste, as is evident in this case, are “wrongs” encased in “rituals of humiliation” (Boopalan 2017). In facing annihilation by caste, the young woman from Hathras was not the first. Khairlanji, Jisha, the young girls in Budaun, and so on, the roads of this country are strewn with the bodies of Dalit women, young and old, and the bodies of Dalit men. Nor does the proliferation of hostile environments stop with Dalit women. Regimes of carceral governance enact rituals of humiliation and occupation – of Dalit, Adivasi, and Muslim homes, neighborhoods, and homelands, of violent death and collective violence, of desecration of places of worship, of unimaginable suffering, and of humiliation as a condition of existence. The anticipation of violence and murder and of the trauma of silencing – the maiming, the “ontological wounds, psychic scars and existential bruises” (West 1999: 339) – tells the story of a country with a people who gave to themselves a constitution that affirms justice, equality, fraternity, freedom of faith, worship, conscience, liberty, dignity, unity, and integrity.
Conclusion The Equality Labs report on caste in the United States points to the historical emergence of the racialization of caste by the early twentieth-century Hindu diaspora. The failure by South Asians to meet the standards of whiteness set out in the Immigration Act, 1924, according to this report, led to Indian men challenging their exclusion in immigration courts: “the first cases were brought by ‘upper’ Caste immigrants, A.K. Mozumdar and Bhagat Singh Thind, both of whom argued that they passed the whiteness test because they identified themselves as ‘high Caste Hindu, of full Indian blood.’ They explained that because they were ‘upper’ Caste, they had pure ‘Aryan’ blood and that those racial origins were something that they historically shared with Caucasians.” Mozumdar was granted citizenship on this basis (Zwick-Maitreyi et al. 2018). This was after the publication of Ambedkar’s Castes in India. The question of whether caste discrimination is discrimination based on race gained traction in the context of the World Conference against Racism held in Durban in 2001 (Thorat and Umakant 2004). Dalit feminists organizing on the specific questions of Dalit women’s rights were central to this effort. The mobilization of National Federation of Dalit Women (1999) and the Delhi Declaration which affirms that “Dalit women have the right to self-protection in the face of dominant caste male and female aggression, of Dalit male aggression, and of aggression committed by law enforcing machineries of the State” is but one instance of the interweaving of everyday socialities with political society. International soft law protocols and transnational frameworks of intersectionality find a specific enunciation in this context, enabling the forging of alliances across borderlands, to invoke Gloria Anzaldua’s glorious work (1987), speaking truth to power in deeply transgressive ways – from the margins (hooks), redefining the politics of belonging in the process (Yuval-Davis et al. 2006).
436
K. Kannabiran
An understanding of intersectionality as drawing from analogical reasoning (notably in legal interpretation) must negotiate the pitfalls of conflating different categories and examine them in relational terms instead. This understanding may be directed toward a critical praxis that people live out that is “both attentive to intersecting power relations and essentially vital for resisting social inequality” (Collins and Bilge 2016). The core ideas that undergird intersectional frameworks across diverse disciplines, according to Collins and Bilge (2016), are social inequality, power (understood “through a lens of mutual construction”), relationality (that rejects binary thinking), social context (that situates arguments on inequality, power, and relationality), complexity (the need to comprehend entangled social contexts and relations), and social justice (to eliminate discrimination). The constellation of these core ideas may be situated within West’s genealogical materialist analysis (1999) and Ambedkar’s archaeological method (Guru 2012) in addressing the annihilation of caste as the route to self-respect and dignity. Importantly, however, it is demonstrated in the everyday lives and choices of Dalit women resisters across the country. Particularly important is the use of intersectionality as a heuristic device (Collins and Bilge 2016), informed by the intense possibilities of “translocational positionalities” (Anthias 2008) and the “situated politics of belonging” (YuvalDavis et al. 2006) – an analytical assemblage that helps better understand discrimination and inequality and therefore mobilizes and acts better to change the social order. A transformative concept, it also interweaves with ideas of decolonizing knowledge and epistemic disobedience (Mignolo 2009), Southern theory (Connell 2007), and the crafting of epistemologies of the South and resisting epistemicide (Santos 2014). In the Indian/South Asian diasporic/transnational contexts, the majoritarian caste-gender complex produces majoritarian caste supremacist formations at every level through multiplex intersections of caste, race, sexuality, religion, class, and political belief, which together coproduce dominant deeply gendered constructions of citizen-noncitizen (interloper/encroacher/anti-national) rooted in Hindutva ideologies (Kannabiran 2020; see Purkayastha 2012). B.R. Ambedkar, in introducing the draft constitution in the constituent assembly, made a powerful reference to “constitutional morality,” marking it as the basis for a democratic and peaceful body politic. Although he did not elaborate on the idea at any length, he marked a distinction between constitutional morality and “public morality” – the latter containing encrustations of habits of discrimination that the constitution sought to displace and uproot. Opening out the discussion on discrimination with a reference to “constitutional morality” immediately foregrounds the praxiological possibilities of the concept of intersectionality – the resurrection of “constitutional morality” after a hiatus of six decades was occasioned by the decision of the Delhi High Court in the case of Naz Foundation vs. NCT Delhi (hereafter Naz Foundation) which decriminalized homosexuality in 2009, deploying (among other arguments) Article 15 (1) of the constitution of India to rule that the ground of “sex” included sexual orientation and that exclusion of sexual minorities from places of public resort through the criminalization of consensual intimacies/orientation/gender identity was violative of Article 15 (2) (at the time of its drafting focused on practices
17
Caste and Gender
437
of untouchability and segregation of Dalits). Two points are worth noting: first, that Naz Foundation removes considerations of gender from a heterosexual binary construction to gender plural contexts; second, fields of discrimination set out in the constitution overlap and intersect, in a sense co-constituting each other, and may best be understood on those terms. In this case the intersection is demonstrated through the grounds of sex and caste. How might the question of justice – social, economic, and political, as guaranteed in the Preamble to the Indian constitution – be framed anew (See Guru 2008)? How might this transgressive praxis open new ways of engaging from the margins across identities and locations? Intersectionality renders possible the reimagination of the citizenship question by taking note of the critical praxis of resistance that destabilizes the heteronorm and majoritarian dominance – the caste-gender complex – with renewed vigor.
Cross-References ▶ Race and Gender ▶ Stereotype Threat Experiences Across Social Groups
References Ambedkar BR (2002) Castes in India. In: Rodrigues V (ed) The essential writings of B.R. Ambedkar. New Delhi, Oxford Ambedkar BR (2019) Slaves and untouchables. Writings and speeches, vol 5. Compiled by Vasant Moon. Ambedkar Foundation, New Delhi, pp 9–18 Anthias F (2008) Thinking through the lens of translocational positionality: an intersectional frame for understanding identity and belonging. Transloc Migr Soc Change 4(1):5–20 Anzaldua G (1987) Borderlands/La Frontera: the new mestiza. Aunt Lute Books, San Francisco Bama (2000) Karukku. Macmillan, New Delhi Bama (2005) Sangati. (Translated from Tamil by Lakshmi Holmström). Oxford University Press, New Delhi Boopalan SJ (2017) Memory, grief and agency: a political theological accounts of wrongs and rites. Palgrave Macmillan. https://doi.org/10.1007/978-3-319-58958-9. Ebook Chakravarti U (1989) Whatever happened to the vedic dasi? Orientalism, nationalism, and a script for the past. In: Sangari K, Vaid S (eds) Recasting women: Essays in colonial history. Kali for Women, New Delhi, pp. 27–87 Chakravarti U (2003) Reconceptualising gender: Phule, Brahmanism and Brahmanical patriarchy. In: Rao A (ed) Gender and caste. Kali for Women, New Delhi, pp 164–179 Collins PH, Bilge S (2016) Intersectionality. Polity Press, Cambridge Connell R (2007) Southern theory: the global dynamics of knowledge in social science. Allen & Unwin/Polity Press, Sydney/Cambridge Connolly WE (1996) Suffering, justice and the politics of becoming. Cult Med Psychiatry 20: 251–277. Reprinted in Connolly WE (1999) Why I am not a secularist. University of Minnesota Press, Minneapolis, pp 47–71 Crenshaw KW (1991) Mapping the margins: intersectionality, identity politics, and violence against women of color. Stanford Law Rev 43:1241–1299
438
K. Kannabiran
Deshpande GP (2002) Introduction. In: Dehspande GP (ed) Selected writings of Jotirao Phule. LeftWord Books, New Delhi, pp 1–21 Geetha V (2003) Periyar, women and an ethic of citizenship. In: Rao A (ed) Gender and caste. Kali for Women, New Delhi, pp 180–203 Geetha V, Rajadurai SV (1998) Towards a non-Brahmin millennium: from iyothee thass to Periyar. Samya, Calcutta Guru G (2008) Constitutional justice: positional and cultural. In: Bhargava R (ed) Politics and ethics of the Indian constitution. Oxford University Press, New Delhi, pp 230–246 Guru G (2012) Archaeology of untouchability. In: Guru, Gopal and Sundar Sarukkai, the cracked mirror: an Indian debate on experience and theory. Oxford University Press, New Delhi, pp 200–222 Kannabiran K (2006) A cartography of resistance: the national federation of Dalit women. In: Yuval-Davis N, Kannabiran K, Vieten U (eds) The situated politics of belonging. Sage, London, pp 54–71 Kannabiran K (2012) Tools of justice: non-discrimination and the Indian constitution. Routledge, New Delhi Kannabiran K (2016) Of law, resurrection and a future. Seminar 689:82–85 Kannabiran K (2020) The shifting sands of citizenship: dispossessions, constitutional ruptures and borderlands. Sociol Bull 69(3):1–20 Kannabiran V, Kannabiran K (1991) Caste and gender: understanding dynamics of power and violence. Econ Polit Wkly 26(37):2130–2133 Mani L (1990) Contentious traditions: the debate on sati in colonial India. In: Sangari K, Vaid S (eds) Recasting women: essays in colonial history. Rutgers University Press, New Brunswick, pp 88–126. (First published by Kali for Women, New Delhi, 1989) Mbembe A (2019) Necropolitics (trans: Corcoran S). Duke University Press, Durham/London Mignolo WD (2009) Epistemic disobedience, independent thought and de-colonial freedom. Theory Cult Soc 26(7–8):1–23. https://doi.org/10.1177/0263276409349275 Minow M (1996) Comments on ‘suffering, justice and the politics of becoming’ by William E. Connolly, presented as the Roger Allan Moore lecture, 11May, 1995. Cult Med Psychiatry 20:279–286 Mohan S (2015) Modernity of slavery: struggles against caste inequality in colonial Kerala. Oxford University Press, New Delhi National Federation of Dalit Women (1999) Marching into the new millennium: Delhi declaration (Dalit women declare the charter of gender rights and demands), New Delhi, December 1999. Unpublished. National Federation of Dalit Women, Reading Materials prepared for the National consultation on gender and racial discrimination, New Delhi, February 2001. Unpublished Naz Foundation and Others v Government of NCT of Delhi and Others, 2009 (160) DLT 277 Patole S (2016) Why I wrote a book on Dalit food. Express Foodie, 8 September. http://www. expressfoodie.com/main-course/why-i-wrote-a-book-on-dalit-food/. Accessed 26 Feb 2021 Pawar U, Moon M (2008) We also made history: women in the Ambedkarite movement. Translated and with an introduction by Wandana Sonalkar. Zubaan, New Delhi Phule J (2002) Slavery. In: Deshpande GP (ed) Selected writings of Jotirao Phule. LeftWord Books, New Delhi, pp 25–46 Purkayastha B (2012) Intersectionality in a transnational world, Patricia Hill Collins symposium. Gender Soc 26(1):55–66 Ram K (1992) Mukkuvar women: gender, hegemony and capitalist transformation in a South Indian fishing community. Kali for Women, New Delhi Rao A (2003) Understanding Sirasgaon: notes towards conceptualising the role of law, caste and gender in a case of ‘atrocity’. In: Rao A (ed) Gender and caste. Kali for Women, New Delhi, pp 276–309 Rathore AS, Verma A (eds) (2011) B.R. Ambedkar: the Buddha and his Dhamma (A Critical edn). Oxford University Press, New Delhi
17
Caste and Gender
439
Redding JA (2018) The rule of disgust? Contemporary transgender rights discourse in India. In: Hasan Z et al (eds) The empire of disgust: prejudice, discrimination, and policy in India and the US. Oxford University Press, New Delhi Reddock R (2014) “Split me in two”: gender, identity, and “race mixing” in the Trinidad and Tobago nation. In: King-O’Riain RC, Small S, Mahtani M, Song M, Spickard P (eds) Global mixed race. NYU Press, New York Rege S (2006) Writing caste/writing gender: reading Dalit women’s testimonios. Zubaan, New Delhi Santos B d S (2014) Epistemologies of the south: justice against epistemicide. Routledge, New York Shinde T (1994) A comparison between women and men: an essay to show who’s really wicked and immoral, women or men? In: O’Hanlon R (ed) A comparison between women and men: Tarabai Shinde and the critique of gender relations in colonial India. Oxford University Press, New Delhi, pp 73–134 SK Mahajan v. State of Maharashtra, Criminal Appeal No. 416 of 2018 Thorat S, Umakant (eds) (2004) Caste, race and discrimination: discourses in international context. Rawat, Jaipur/New Delhi Valmiki O (2007) Joothan. Translated from the Hindi by Arun Prabha Mukerjee. Samya, Calcutta Vishakha vs. State of Rajasthan (1997) 6 SCC 241 West C (1999) The Cornel West reader. Basic Civitas Books, New York Yuval-Davis N, Kannabiran K, Veiten UN (eds) (2006) The situated politics of belonging. Sage, London Zwick-Maitreyi M, Soundararajan T, Dar N, Bheel RF, Balakrishnan P (2018) Caste in the United States. A survey of caste among South Asian Americans. Equality Labs, USA
Part IV Social and Regional Dimensions of Discrimination
The Economic Side of Religious Discrimination in France: A Review
18
Christophe Jalil Nordman
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Section 1: French History of Immigration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Section 2: Methodologies of Existing Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Job Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wage Gap and Unemployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Labor Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Housing and Rental Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Section 3: Theme-Wise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Job Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Labor Market Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Housing Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Economic Assimilation Related to Marriage and Language Skills . . . . . . . . . . . . . . . . . . . . . . . . . Return Migration from France: Failure or Success? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Section 4: Discussions and Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Society’s Opinion and Drivers of Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Increase or Decrease of Discrimination? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Public Policies: Past and Present . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ways Forward in Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
444 445 446 447 447 448 449 449 450 452 454 455 456 457 457 458 458 460 461 461
Abstract
Religious discrimination in France is an important social issue that has severely impacted the economic lives of the targeted religious minorities, especially in the French National Research Institute for Sustainable Development (IRD), LEDa-DIAL (IRD, PSL University & CNRS) & French Institute of Pondicherry C. J. Nordman (*) French National Research Institute for Sustainable Development (IRD), LEDa-DIAL (IRD, PSL University & CNRS) & French Institute of Pondicherry, Paris, France e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_27
443
444
C. J. Nordman
recent period. Over the years, various studies have been conducted to assess the extent of discrimination and its effects on different socioeconomic dimensions. This chapter discusses this existing literature and proposes a reflection on the resulting established knowledge. I focus on studies published in the last 20 years and make a selection of literature in the field of economics mainly (with some incursions in sociological studies). This chapter covers religious discrimination revealed in domains such as job access, labor market trends, housing sector, language, and other important dimensions. It shows that there exists a relative consensus on the importance, economic consequences, and drivers of religious discrimination in France, especially against the Muslim population. Keywords
Discrimination · Religion · Labor market · Job · Housing · Migration · Muslim
Introduction Religious discrimination in France is an important social issue that has severely impacted the economic lives of the targeted religious minorities. The topic becomes further relevant in the present as these discriminatory trends hinder the promotion of a free, equal, and fraternal society that the French motto puts forward. Even more, recent Islamic attacks in France, in September and October 2020, which follow the very deadly November 2015 ones, have given rise today in public debates and media networks to rapid shortcuts between the failure to integrate foreigners in France and a perceived increase in physical and social violences. One may refer to a media investigation (L’Express, March 2021) on the rise of Islamophobia in France: https:// www.lexpress.fr/actualite/societe/islamophobie_1640167.html (accessed July 19, 2022). Over the years, various studies have been conducted to assess the extent of discrimination and its effects on different socioeconomic dimensions. Efforts have also been made to bring awareness on the topic in the public debate. This chapter discusses this existing literature and proposes a discussion on the resulting established knowledge. For the need of concision, I focus on studies published in the last 20 years, and I make a selection of literature in the field of economics mainly (with some incursions in sociological studies), although I admit that such a topic would necessitate a broader disciplinary view to grasp all its complexity. I also aim at pointing out gaps in this literature that could possibly be addressed in the future. This chapter covers religious discrimination revealed in domains such as job access (according to articles 225–1 and 225–2 of the French penal code, discrimination in hiring on account of religion occurs when job applicants endowed with equivalent records of study and employment, but perceived by the recruiter as belonging to a different religious confession, are not dealt with in the same way), labor market trends, housing sector, language, and other dimensions that impede the integration of religious minorities in the French socioeconomic system. The papers discussed in
18
The Economic Side of Religious Discrimination in France: A Review
445
this review mainly uses the term “integration” in the context of minority cultures being recognized in the French society. Semantically, “integration” occurs when all barriers to full participation in a society have been dismantled (Kymlicka 2012, pp. 14–15); “assimilation” is used to refer to elimination of cultural distinctiveness and entering into the structure of the larger society. But some articles discussed here seem to use them as synonyms. The criteria used for the literature selection include published books, articles, and chapters but also some recent working papers. The chapter is presented in the following manner. Section “French History of Immigration” covers the French historical background of immigration, Section “Methodologies of Existing Studies” summarizes some methodologies used in the selected studies, and section “Theme-Wise Analysis” proposes a theme-wise analysis of these discrimination studies. The last section “Discussions and Concluding Remarks” discusses some of the principal questions asked and implications for public policies and suggests some ways forward in research.
Section 1: French History of Immigration Since the onset of the First World War, France mobilized colonial workers and laborers to fill in the industrial gaps and to sustain the manpower in the economy. The immigrants allowed the war effort to be sustained and economic activity to be maintained, in particular in the industries of weapons. This triggered a movement of immigration from neighboring countries, especially from those that were historically linked to the French culture through colonization, such as Algeria and Morocco. The process existed in France and in other European economies, but, during periods of recession, the immigrants were perceived as a burden on the economy’s labor force. Following the oil crisis of 1973, European countries began to implement restrictive policies of immigration to protect the economic interests of the natives (Fromentin 2013). By 2004, immigrants represented 8.6% of the working population in France, which was more than their proportion in the total population at 7.6% (Fromentin 2013). Though the rates of unemployment fluctuated between natives and immigrants, in 2009, natives being at 8.4% against immigrants at 18.5%, the immigrants mostly occupied low-skilled jobs. Immigration continued at relatively high rates, especially from North African countries through family reunification and refugee schemes even though France imposed tighter migration policies following economic crises in the 1970s (especially the 1973 oil crisis). These schemes led to further immigration from other regions as well, such as sub-Saharan countries and Arabic countries, including Turkey (Meng and Meurs 2009). In fact, the number of immigrants from non-European countries was at times almost higher than that from European countries. A census in 1999 showed that immigrants occupied 7.4% of France’s total population, out of which 45% came from a list of European countries – mainly Spain, Portugal, and Italy – while 39% was from Africa (Meng and Meurs 2009). In terms of immigrants’ repartition, Gubert and Nordman (2009) provide trends between 1990 and 2000 and some projections for OECD countries in the future.
446
C. J. Nordman
They show that, among the North African communities in Europe, the Moroccans are the most numerous: they accounted for 49% in 1990 and 56% in 2000 of immigrants originating from the three Maghrebian countries (Algeria, Tunisia, and Morocco; Gubert and Nordman 2009, Fig. 1). While the numbers of Algerians and Tunisians did not change substantially during the 1990s in France, the numbers of Moroccans increased over the period as a result of family reunification. In terms of geographical distribution, Moroccan nationals are predominantly found in France in 2000, followed by Spain, the Netherlands, and Italy, but are still widely distributed over all European countries. By contrast, emigration from Algeria is very concentrated toward France: 84% of Algerians residing in an OECD country lived in France in 2000. Native French are predominantly Christians. According to the Pew Research Center estimates in 2017 (https://www.pewresearch.org/fact-tank/2017/11/29/5facts-about-the-muslim-population-in-europe/, accessed on July 19, 2022), Muslims amount to 5.7 million in France, i.e., 8.8% of its total population in 2016, out of which most of them are from Maghrebian countries (Algeria, Morocco, and Tunisia). After Catholicism, Islam is the second largest religion in France. Only half of this Muslim population is of Arab descent, while the rest are Turkish and West African, and some are from South(East) Asia. Compared to other European countries, France holds the largest Muslim population which equals to one third of the total Muslim population of Europe (Brookings Institution 2016). With increasing population in France, Muslims are growing at a significant rate in France, generating increasing friction between them and the native population. According to figures reported in Brookings Institution report (2016), the French population is actually predicted to grow, under some projections, to 75 million in 2050 (from 67.4 million in 2020 according to the new census of the Institut national de la statistique et des études économiques – INSEE), which would make France the largest nation among the 25 current members of the European Union. The relatively higher fertility rate of French women of French origin makes up the bulk of the growth differential with countries like the United Kingdom, Italy, and Germany. The latter are indeed expected to experience a steady need for immigrant labor as their national fertility rates fall. In the following sections, I will show that both taste-based and statistical discriminations have led to unemployment issues for Muslim immigrants in various economic sectors. As anti-Muslim discrimination increased, the studies reviewed below pinpoint its impact on different economic well-being indicators. These studies ranged from various economic sectors to multiple methods of survey in which later sections will take up in further detail.
Section 2: Methodologies of Existing Studies I have selected papers here that explore the economic impacts of religious discrimination using various methodologies. These include qualitative interviews of immigrants; experimental designs conducted on firms and companies during their period
18
The Economic Side of Religious Discrimination in France: A Review
447
of hiring; case studies specific to geographic regions of France, especially in the large agglomeration of Paris (Île-de-France); comparative studies between France and other countries; econometric analysis that come up with predictive models; and different types of statistical analysis. Many of them also combine methods to produce more comprehensive results. If categorized thematically, the four main areas of interest discussed below are (1) job access, (2) wage gap and unemployment, (3) labor market functioning, and (4) housing and rental market.
Job Access Discrimination on job access toward immigrants based on their religion is a commonly analyzed topic for France. The literature in this area most often uses experimental methodologies. The aim of these approaches is to evaluate the employer response while hiring an applicant. One method includes creating jobseeker profiles for a CV-based trial. Pierné (2013) and Valfort (2015, 2020) create a bulk of profiles with similar qualifications and backgrounds except for the surname or first name, to essentially test the discriminatory stigma attached to the origin of the person which is revealed by his/her name. They create Christian names, African names with Muslim surnames, and sometimes other combinations. They conduct this test in two stages, once to check the discrimination at the hiring level and once more to check it at the call-back level. At the hiring level, the recruiters do not get to physically meet the applicant which removes the chance of racial bias linked to skin color. Therefore, in this round, the authors can test the discrimination carried out simply by the sound of the name which signals the religious closeness. In the second round, if a candidate with African Muslim origin is called back for an interview, the authors then test the call-back rates to see how many of the Muslim-origin applicants manage to get through as compared to their Christian counterparts. This approach is used by Pierné (2013) and Valfort (2015, 2020) to test the impact of religious discrimination on job accessibility of Muslim aspirants. Results of this exercise will be discussed in section “Theme-Wise Analysis.”
Wage Gap and Unemployment Wage gap and unemployment in France are another important area of interest by researchers. Given the striking numbers of unemployment among Maghreb populations in France, studies use various methodologies to evaluate the depth of discrimination in this field. Due to the high number of variables while calculating the interrelationships between immigration, economic growth, labor market, and situational occurrences, the use of multi-equation econometric modelling is often preferred to quantify the relationships (e.g., Aeberhardt et al. 2010; Fromentin 2013). Richard (2000, 2013) relies on secondary longitudinal datasets to track the unemployment rates at different points in time (the French Echantillon Démographique Permanent (EDP), an INSEE longitudinal dataset equivalent to the English LS and
448
C. J. Nordman
the American NLSY), while Aeberhardt et al. (2010) use a cross-sectional INSEE survey in 2003 (Formation Qualification Professionnelle, FQP). Richard (2000, 2013) combines equations to predict the phenomenon of unemployment, using logistic regressions that allow him to gain a better understanding of the complexity of the mechanisms while estimating the effects of sociodemographic variables on the probability of holding a job.
Labor Market The labor market is thought to be an important domain while studying discrimination in France because it is where discrimination is believed to be the most evident due the history of labor migration in France. Studies in this area include, e.g., Meurs et al. (2006), Silberman et al. (2006), Silberman and Fournier (2008), Bonifazi et al. (2008), Senik and Verdier (2010), Lefranc (2010), Algan et al. (2010), Simon and Steichen (2014), Combes et al. (2016), Killian and Manohar (2016) and Vourc’h et al. (1999). They use a combination of methodologies: comparative analysis, across countries and topics, e.g., earnings, education, and employment; qualitative interviews; and statistical and econometric models. While using comparative analysis (such as in Silberman et al. (2006), Algan et al. (2010), or Killian and Manohar (2016)), the authors include factors of comparison such as European countries and the United States, first- and second-generation immigrants, and assimilation factors, e.g., marriage and access to education. To check the difference of discrimination across generations, the approach used is often a combination of models testing the outcomes before and after a change in certain exogenous parameters (such as the local supply of public housing in Combes et al. (2016) for studying customer discrimination and employment outcomes). Sometimes the studies span across types of occupations by generation (Silberman et al. 2006; Lefranc 2010; Algan et al. 2010). To test labor market discriminations across countries, comparative studies (e.g., Silberman et al. 2006; Algan et al. 2010; or Killian and Manohar 2016) include qualitative interviews of participants from the targeted countries: semi-structured interviews allowed the respondents to be more open-ended and thereby bring forth maximum detail possible of their discriminatory experience. Killian and Manohar (2016) use snowball sampling to include acquaintances of the initial participants and gather a more wholesome picture of the labor market treatments. While comparing the labor market situation within France between natives and the immigrants, regression analysis is most used to estimate the relationship between dependent and independent variables while measuring religious and ethnic disadvantages. Another method used to compare the employment outcomes of immigrants over time is, for instance, synthetic cohort analysis of the French Labour Force Survey (LFS) (Simon and Steichen 2014). In this study, the authors combine cross-sectional LFS from 2003 to 2011 using “pseudo” cohorts of migrants. Finally, Meng and Meurs (2009) study the role of intermarriage in the process of immigrant economic assimilation by running earnings equations of the couples while relying on
18
The Economic Side of Religious Discrimination in France: A Review
449
instrumental variables to correct the endogeneity of marriage (the “sex ratio” for each region-ethnicity cell and the “probability of marrying within one’s own ethnic group”).
Housing and Rental Market Although not investigated as much as the previous three topics, the housing and rental market also showcase discriminatory patterns against immigrants due to their religion. Bonnet et al. (2016), Acolin et al. (2016), and Gobillon and Solignac (2019) base themselves on past studies that show the existence of religious discrimination in this area. The challenge while researching this topic is to be able to differentiate between ethnic discriminatory factors and residential or economic discriminatory ones. In other words, landlords may be reluctant to rent out the house to tenants who come from backward residential areas since they are more likely to be economically indisposed to pay higher rents. Therefore, if an African-origin or Muslim person’s rent application is rejected, it could also be due to economic reasons depending on their residential background and not necessarily to the religious closeness or ethnic aspect. To disentangle these questions, authors use combinations of methodologies, namely, audit testing strategies, randomized experimental designs, and qualitative surveys with probit and multinomial logistic models. The authors start with randomized experimental designs to gather their applicants. Next they send out advertisements online that do not include the level of education, employment, income, or household structure of the applicants. This process is used to minimize the effect of variables on the selection factors of the applicants. In Bonn et al. (2016), a base research is carried out through audit testing to study the association between ethnic and residential effects, and qualitative surveys are conducted to assess the prevalence of these stigmas on the side of the landlords or real estate agents before initiating an experimental design. This allows the authors to have a clearer understanding of the scenario before collecting data to conduct their experiments. In the next stage, authors estimate the determinants of the probability of getting a response and identify correlates that could explain the decision-making process of the agent/landlord. Apart from rental markets, Gobillon and Solignac (2019) also study homeownership rates of immigrants (homeowners vs home sellers). They use econometric decomposition and logit models to estimate the fluctuation of homeownership rates and the probability of the immigrant group’s choice of either buying a home or leaving their home.
Section 3: Theme-Wise Analysis This section discusses the findings of the abovementioned papers (and some additional) in four domains: job access, labor market outcomes, housing market, and return migration. Covered subtopics include unemployment, labor outcomes of first-
450
C. J. Nordman
and second-generation immigrants, gender, religious and ethnic wage gaps, marriage, and the advantage of knowing the French language, entrepreneurship, and skills mismatch.
Job Access This topic is widely covered in Cediey and Foroni (2008), Adida et al. (2010), Pierné (2013), and Valfort (2015, 2020) who look at multiple specific areas such as hiring discrimination, public vs private firms, firm size, gender, religion, and nationality. Hiring discrimination is the most looked at out of all the topics. Using a survey conducted in 2006 and commissioned by the Department of Research and Statistics (DARES) of the Ministry of Labour of France, Cediey and Foroni (2008) broke down the process into various stages to point out exactly when the discrimination was highest. The first stage surveyed was job call-back after the employers saw the applications. They found that, in this stage, applicants from the majoritarian population, i.e., native French, were called back more than those from minorities. This happened simply on the basis of the applicant’s name and profile, without having met them in person. The second stage was the pre-interview round. Here again, three out of four minority applicants were rejected even before appearing for the interview. The third and last stage, the interview round, had an average of only one out of six minority applicants being chosen; the other five were native applicants with the same profile and qualifications as the minority candidates. This pattern was repeated by half of the employers surveyed, while the other half continued to test both kinds of applicants equally. Despite this, only 11% of employers followed the equality treatment of majority and minority applicants without any discrimination. Another finding of Cediey and Foroni (2008) with this survey was that even when a minority applicant was selected over a majority applicant in the initial contact stage, the chances of the minority making it till the end of the selection process was far lesser than when the case was reversed (i.e., when a majority applicant had the advantage in the initial stage and was finally turned down at the end). Overall, the conclusion drawn showed that maximum discrimination against immigrant applicants took place even before meeting them. This showcases the existence of high discrimination based on the family name revealing the person’s religion or ethnicity regardless of physical appearance or voice. In this survey, gender didn’t make a difference when it came to the discrimination against origins which was the same for both North African and sub-Saharan applicants. Similarly, Valfort (2015) found that if a Muslim male applicant shows that he is secular, then his chances of hiring increase even if his CV is not outstanding, proving again that qualification matters less than the religious background of the candidate. In Pierné (2013), the hiring process in real estate businesses (a profession involving high degree of public exposure and contact with customers) shows that there was discrimination against North African-origin applicants regardless of their religion. Whether Catholic, Muslim, or secular, they still faced higher rejection than those with French-origin names. The study aimed to see the interrelation of religion and
18
The Economic Side of Religious Discrimination in France: A Review
451
ethnicity. The results confirmed the existence of origin-based hiring discrimination in which various previous studies also aimed to find out in other countries (for Australia in Riach and Rich 1991; for the United States in Bertrand and Mullainathan 2004). Similarly, while testing religious discrimination, results showed that hiring discrimination happened against closeness to Islam regardless of origin; therefore, ethnicity discrimination depended largely on religious closeness. Another correspondence test in Duguet et al. (2010) showed similar results (16 jobseeker profiles and 1097 CVs in reply to 140 job vacancies in 2006 in the profession of accountant): due to the surnames of candidates, sounding as ethnic origin, there was most discrimination; but if only first names were foreign, it did not make any difference. Therefore, the main discrimination against origin was revealed through surnames, but first name considerations did not appear strong enough to counterbalance the risks of not hiring a person having a good profile. The next subtopics evaluated are intersectional, like gender discrimination within Muslim religious discrimination, firm size, and public private firms. Starting with gender, Valfort (2015) showed that, in many cases, within the Muslim population, women face less discrimination than their male counterparts. Muslim and Jewish women can escape discrimination by showing outstanding CVs, but, for their male counterparts, that does not happen. Muslim men do not benefit from having outstanding CVs. Under that circumstance, the call-back rate for Muslim men was five times lower than that for Catholic men with an equally outstanding profile. Call-back rates for females are significantly higher than males, though this only applies for jobs of assistants. But the higher the professions become in hierarchy of occupations, the more discrimination increases against Muslim women than against Muslim men comparatively. Also, religious Muslim men face more discrimination in comparison to religious Christian men than what religious Muslim women face in comparison to religious Christian women. Another result was that applying through social media platforms does not reduce hiring discrimination because, on most of these platforms, the recruiters can dig up much more in a person’s background. So, whether applying in person or through social media, the discrimination rates were the same. Valfort (2015) provides a comparative table of the results of CV-based trials carried out in France and elsewhere, allowing her to compare the call-back rates of citizens without recent migration in their family background to the call-back rates of citizens who come from Muslim majority countries. This exercise showed that, out of 14 countries (11 being in Europe), France had the highest discrimination against people originating from North Africa, Middle East, and Turkey. The article argues that the findings are coherent with a “theory of menace” (according to sociologist Hubert Blalock in 1967), meaning that when a native person (i.e., employers of French origin) comes in contact with an “exo-group” (applicants with Muslim names), his probability of showing hostility toward it increases with his (perceived) proportion of the persons (applicants) belonging to this “exo-group.” Other findings in Valfort (2015) are that Muslims face discrimination across all firm sizes, despite large firms having enough resources to fight discrimination and to not have to adhere to it. The risk of hiring a practicing Muslim candidate may be offset if the candidate has an outstanding CV, but it doesn’t take discrimination out
452
C. J. Nordman
altogether. Again, Muslim women benefit from this, whereas, ironically for Muslim men, outstanding qualifications may sometimes make it worse. This is explained in the following way: the higher the qualification of Muslim men, the more they are perceived to have tendencies to not take directions. Besides, taste-based discrimination exists equally in both public and private sectors, but when it comes to interview (call-back rates), the private sector is more discriminatory. This could mean that statistical discrimination exists more in private sector. Cahuc et al. (2019) show, however, that the invitation stage (call-back) may be a poor predictor of discrimination, as there can be no initial discrimination at the initial stage, but then it might occur at the final hiring stage which could happen because of the productivity requirements and reduction of vacancies: when there is less productivity requirements, like in the public sector, recruiters do not value high profiles as much, so they tend to choose mediocre profiles without taking the risk of hiring a minority, thereby leading to discrimination at the hiring stage which makes it higher than that in the initial interview call-back stage. Hence, while there is no discrimination at call-back stage against North Africans in the public sector, there is in the private sector. Cahuc et al. (2019) conclude that, based on differences in productivity requirements between public and private sectors, the lack of discrimination at the interview stage in public sector is compatible with the existence of discrimination at the hiring stage, calling into question the ability of correspondence studies to detect unequal treatment of applicants. An interesting study conducted in 2018 aimed to find out the depth of what is termed as “religious penalty” in discrimination which is defined as the increase of discrimination that occurs with being a practicing religious person (Valfort 2020). The study aimed to compare the call-back rates of immigrants of Muslim and Christian culture who originate from the same country and whose religiosity varies from non-religious to religious. To do this, Valfort (2020) includes Muslims and fictitious applicants of Jewish culture in her experiment in order to disentangle whether antiMuslim discrimination is directed at Muslims qua Muslims or at any religious minority. Results showed that Muslims face a general of 6.7% lesser call-back rates than Christians. But non-religious Muslims face significantly lesser discrimination than their religious counterparts. When the religiosity factor is taken into account for Christians, then it actually plays as a catalyst for availing jobs as opposed to religious Muslims who have to submit twice the number of applications to just get a call-back. Additionally, compared to Jews, Muslims have higher rates of discrimination. Thus, two conclusions can be drawn: (i) religious penalty is held against Muslims, but it becomes an additional quality for Christians who are called back twice as much as practicing Muslims and (ii) Muslims are discriminated against due to their Islamic affiliation, not due to their religious minority status.
Labor Market Outcomes The labor market is the next important area that researchers write about since this is where they can find differences of advantage between first- and second-generation
18
The Economic Side of Religious Discrimination in France: A Review
453
immigrants, patterns of unemployment, and influences of social networks and education. First- and second-generation immigrants face different levels of discrimination in the labor market. As far as employment is concerned, Algan et al. (2010) show that all first-generation immigrants have lower employment levels than their Christian counterparts, especially the African immigrants who stand at 15.7% lower. The situation worsens with second generation who experience bigger employment gaps: Africans 47.9%, Turks 41.6%, and Maghrebians 26.7% lower than Christians. Coming to earnings, the gaps increase for first-generation immigrants who are less educated than the native French, and the gap reduces for immigrants from countries that have better education. Yet, unconditionally, all first generations earn less than their native French counterparts. For the second generation, the earning levels improve but only for women from Turkey. In general, second-generation immigrants do not have any significant increase of benefit of employment compared to the first generation even though they hold higher educational qualifications. As claimed in Meurs et al. (2008), employment discrimination not only appears through the rates of job access open to immigrants but also through the type of jobs they hold, which is both vertical and horizontal. Vertical discrimination is within the hierarchical positions in a company, and horizontal is across populations and origins. So even if immigrants do manage to access jobs, they occupy mainly low-skilled jobs. This highlights therefore the existence of discrimination at the workplace. Unemployment rates among Muslims are significantly high in France. As Fredetter (2014) pointed out, “it is safe (and probably something of an understatement) to claim that nearly 60 percent of the immigrant population that the INSEE identifies as facing a 20 percent unemployment rate is Muslim—that is, Muslim immigrants face an unemployment rate of roughly 12 percent. This is twice the unemployment rate of immigrants from Europe.” High levels of unemployment for Muslims may have a direct relation to discrimination. According to Cahuc et al. (2019), high unemployment indicates less job vacancies, so employers have higher chances of discriminating in call-backs against Muslims while choosing candidates because they may be reluctant to take the risk of religious burden. With low unemployment, Muslim discrimination would not be as high as otherwise because employers rely more on profiles and productivity signals. Another angle is the relation between customer discrimination and employment discrimination. As shown by Combes et al. (2016), when consumers do not want to interact with immigrant employees, i.e., customer discrimination, especially in contact jobs, then this can lead to employers being unwilling to recruit immigrants. Hence, their value reduces, causing the differential rate of occupation in contact jobs between Africans and French natives to increase. This study confirmed previous theoretical literature on discrimination in frictional environments: “without frictions, discrimination only affects wages. Under frictions, discrimination in a number of jobs translates into higher chances of unemployment” (Combes et al. 2016, page 110). In a sociological comparative study between North African women in France and Tamil women in the United States, Killian and Manohar (2016) explore the
454
C. J. Nordman
cumulative and interactive effects of four mediating mechanisms: interaction of local labor markets and immigration regimes, education-work experience nexus, social capital in social networks, and racialization. These result in divergent labor market outcomes for North African and Tamil women: compared to Tamil immigrants in the United States, the North African immigrants in France had the advantage of language when they arrived since France was a colonizer of North Africa, yet the Asian immigrants did better in accessing highly paid, professional jobs in the United States. This finding reflected on the educational designs that existed in the United States which basically integrated internship programs and other similar initiatives for students to get in touch with businesses and gain work experience. Because of these education bridging programs, the Tamil immigrants were able to gain work experience and build social networks that later helped them in accessing better jobs than their North African counterparts in France, who did not go through the same advantages during their education. Similarly, Senik and Verdier (2010) find that the work attitudes of the ethnic minorities in France seem to be, to a large extent, related to the availability of their ethnic network; therefore, the importance of social networks is a crucial point in their economic integration in the labor market.
The Housing Market This is another area where discrimination is quite prominent. Research into this subtopic is important because it brings out factors such as regional and residential discrimination and homeownership issues among North African immigrants in France. Acolin et al. (2016) analyze discrimination in five minority groups in the housing sector. The groups surveyed were Northern African, sub-Saharan African, Southern European, Eastern European, and Turks. They find that those with Southern European names did not face significant discrimination though they were required to provide more details than the French applicants. Those with African and Turkish sounding names faced the most discrimination and were 16–22% less likely to get a response from landlords. Previous literature had theorized that the local context can influence the behavior of landlords, i.e., in areas that are already concentrated by a certain ethnic population, their landlords tend to show less discrimination toward new tenants of that same ethnicity. But in this study, discrimination only increased within same ethnic area which could show that discrimination increases in areas where immigrants have lower income levels. Results also point to varying discrimination across regions of France: more discrimination in the North-East, South-West, and Paris region (Île-de-France). Other studies have investigated the discrimination that exists in real estate and the psychologies of real estate agents. Bonnet et al. (2016) look at the interplay of residential and ethnic stigma. They point both discriminatory discourse, which is the prejudice in theory, and discriminatory practice, which is the exercising of the prejudice on the ground. Previous literature had shown that one does not necessarily lead to the other as often people’s attitudes are different than the way they finally act
18
The Economic Side of Religious Discrimination in France: A Review
455
out. Some studies have even shown that people tend to behave exactly the opposite way of what they believe in. So when asked, people from the housing market believed that there exists a huge problem of ethnic discrimination in the business, but when asked if residence could make a difference, they did not believe in that at all. Yet, looking at survey data, results showed that discrimination based on residence was higher than discrimination based only on ethnicity. This highlights that real estate agents are concerned by insolvency issues (lack of financial resources of the tenant) and, therefore, residence matters to them because they associate a neighborhood with the applicant’s financial capacity. But residential discrimination in turn actually influences ethnic discrimination simply because majority of these poorer neighborhoods contain immigrants. Conclusions drawn are that there can be an additive effect of discrimination, e.g., residential discrimination can add to ethnic discrimination. Yet, at the same time, there can be substitutive effects too: sometimes real estate agents simply substitute one for the other (residence and ethnicity) when both factors are closely interconnected. A final area closely studied in this topic is homeownership of immigrants, as a marker of assimilation. Gobillon and Solignac (2019) compare the difference in homeownership rates between natives and first-generation immigrants in France and how this difference evolves over the 1975–1999 period. Entrants into the territory during the period usually suffer more from homeownership because they are the younger population and thus have lesser levels of income which makes it harder for them to get homes in big cities. Leavers during the period, on the other hand, exhibit the opposite effect on homeownership rates: they save up for buying homes in their native countries, so, when they leave France, they create more space in the homeownership business and push the rates up which shows a positive image of assimilation. What is confirmed though is that homeownership is directly related to income levels because lesser wealth accumulation means lesser chances of having a house. The study also compares native stayers with immigrant stayers. Results show that, over time, the gap did widen but at a slower rate, which does indicate a slow positive effect on assimilation of immigrants. Despite this, housing conditions for immigrants are much worse than they are for natives, so even if they have homeownership, the quality is often poor.
Economic Assimilation Related to Marriage and Language Skills Meng et al. (2009) study the role of intermarriage in the process of immigrant economic assimilation. They show that intermarried couples, i.e., native with immigrant, earn 17% more than endogamous couples, i.e., immigrant with immigrant. Knowing the native French language gives an advantage to bigger economic benefits from intermarriage. Similarly, compared to other Muslim immigrants, Maghrebians have higher chances of assimilation since they know French and can hence make better use of labor market networks from their native spouses. More recently, Lochmann (2020) shows that with 100 h of French language training, the probability of entering the labor market increases from 14.5 to 26.6 percentage points. But the
456
C. J. Nordman
idea that language training directly results in an increase in employment rates is contested (Lochmann 2020). Where the training did help though was known as the “information channel” which was essentially the information exchanged between the participants during the course about their daily life and the job market. Apart from this, a positive effect of French courses is found on the job search. At present, further studies are being conducted to test whether the added amendments to the course, such as training focused on practical professional life in France, are helping professional economic integration of immigrants (see the Enquête Longitudinale sur l’Intégration des Primo-Arrivants (ELIPA) and the new Contrat d’Intégration Républicaine (CIR), 2016, which includes a reinforced French language training module).
Return Migration from France: Failure or Success? Finally, the empirical economic literature on return migration is not only small but also highly selective, especially when it comes to France. The situation and impact of returned migrants in origin countries are, however, central to the discussion on the benefits and costs associated with a migration experience. In particular, a returning migrant brings not only financial capital to be invested in his home country but also his experience and skills acquired abroad. Using an original database on return migrants in North Africa (mostly from France), Gubert and Nordman (2011) analyze returnees’ entrepreneurial behavior in Morocco, Algeria, and Tunisia. First, they show that one third of returnees did invest in projects and businesses after return, although this share strongly varies between countries. A lower propensity to invest for Algerians partly results from the fact that a significant share of them went to France in as early as the 1960s and occupied low-qualified positions that did not allow them to acquire any entrepreneurial skill. Second, the probability of becoming an entrepreneur after return seems to be higher for returnees with a first experience as employers or self-employed, for those who received vocational training while abroad, and for those who independently and freely chose to return (as opposed to being “expulsed” by French authorities). Using the same surveys, but looking at subjective well-being indicators of the returnees themselves, Gubert and Nordman (2008) bring support to the idea that returnees do not form a homogeneous group, in the sense that among those who clearly benefited from their experience abroad are the early migrants, i.e., those uneducated migrants who left during the late 1980s and early 1990s. By contrast, those who negatively assess their experience abroad are those young mediumeducated migrants. Focusing specifically on education mismatch, David and Nordman (2017) look at whether return migrants in Egypt (mostly from Anglophone countries) and Tunisia (mostly from France) are more likely to have jobs that correspond to their education level compared to non-migrants. They use a survey conducted on both returnees and non-migrants in 2006 and 2007 (European Training Foundation survey), thus allowing them to compare their respective education mismatch. They find evidence of
18
The Economic Side of Religious Discrimination in France: A Review
457
education mismatch, especially in Tunisia. When looking into the determinants of education mismatch on the Egyptian and Tunisian labor markets, they find a significant positive effect of return migration on the probability of being overeducated. This therefore brings support to previous papers showing that a significant share of returnees from France become entrepreneurs (Gubert and Nordman 2008, 2011). One of the implications of their results on the positive correlation between return migration and overeducation is that returnees might prefer to turn to entrepreneurship in a context where their expectations in terms of use of skills are not met.
Section 4: Discussions and Concluding Remarks Society’s Opinion and Drivers of Discrimination Anti-Muslim discrimination in France (and more generally “Islamophobia”) is a widespread phenomenon that has existed over the years within the French society. It has even, according to some commentators, gained incidence in the recent period due to several deadly Islamic attacks in 2020 (in streets, school, and church). With a survey carried out in 2014 asking the general population about their opinion on antiMuslim discrimination, Valfort (2015) finds that a large proportion of the respondents believed that religious and ethnic discrimination were high. Muslim religion therefore faces a huge taste-based discrimination in the general population. Prior to that, another survey in 2013 showed only 26% of those interviewed had a good/very good image of Islam. From an economic perspective, frictions between natives and immigrants have even more reasons to increase due to immigrants occupying a large portion of the labor market. According to the far-right political party in France, Le Rassemblement National (RN), which gains supporters in national elections years after years (the most recent example being the 2022 legislative election where 88 RN deputies entered the Assemblée Nationale compared to “only” 8 in the previous 2017 mandatory), immigration leaves natives behind in the labor market, and not only during times of economic crisis in the country. Available literature points indeed to increased discrimination during these times. However, demographers and economists agree that France is not (anymore) a country of massive immigration compared to OECD countries (Héran 2017) and is even today relatively “closed” to immigration (Mouhoud 2015: https://www.lemonde.fr/economie/article/2015/09/14/ migrants-dissiper-les-fantasmes_4750063_3234.html (accessed July 19, 2022). See also Le Monde (2019), https://www.lemonde.fr/les-decodeurs/article/2019/02/28/ sept-idees-recues-sur-l-evolution-de-la-france-depuis-trente-ans_5429436_ 4355770.html (accessed July 19, 2022). Hence, discrimination is not necessarily targeted toward Muslims, but toward immigrants in general who are in majority from North Africa and from West Africa (especially Mali, in East of Paris). Ethnic and religious discrimination are very often interrelated in the case of France. Specific anti-Muslim discrimination has been found to increase due to the spread of terrorism in the world, especially since the
458
C. J. Nordman
significant November 2015 attacks in Paris (Bataclan, Saint-Denis Stadium, streets of Paris), and the more recent ones in September and October 2020. Valfort (2015) discusses the drivers of discrimination. She observes that, when French people are asked what they think of the Muslim religion, they have a negative response because of two main reasons: Muslim extremism associated largely with the entire religion and the covering by Muslim women of their heads with the “veil.” Essentially, they believe that Muslim religion has high gender inequality within its customs, and this is looked down upon. But while studying the drivers of Muslim discrimination in the economic sector, Muslim women covering their heads can be another reason of discrimination at the workplace since recruiters feel that they might be less flexible to duties because of it and therefore cast a bad image to clients. Similarly, “male chauvinism” in Muslim men is another driver because Muslim male fanatics refuse to shake hands with women, which becomes a problem in a professional setting. Valfort (2015, page 30) states “Muslim men from North Africa are essentially perceived as ‘hard to manage’, characterized by ‘a complicated relationship to authority’ and for that matter by ‘male chauvinism’.” Certain previous studies discussed above also show that employers believe that religiosity increases the chances of aggressive behavior in Muslims, but this again is seen as a quality when it comes to Christians.
Increase or Decrease of Discrimination? In general, when evaluating the psychological reasons that drive discrimination, studies discussed above (for instance, Cahuc et al. 2019; pages 9 and 11) found that there is an added disadvantage for North Africans. Deep-rooted racial stereotypes from colonial times place North Africans at the bottom of French social hierarchy. The shared history also causes a French general belief that North Africans are a low-educated group and should therefore be more suited for low-skilled jobs. In fact, Cahuc et al. (2019) find that discrimination not only happens against Muslimsounding names (“Mohammad” being the most discriminated name) but also for any name which sounds like it has a foreign origin that cannot be pinpointed. To take matters worse, though some surveys claim that this has reduced since 2017, an increasing Islamophobia due to terrorism certainly adds to the general sentiment of hostility toward Muslims.
Public Policies: Past and Present Since the end of the Second World War, there have been various policies regarding immigration, sometimes to allow immigrants entering the country, such as the family reunification program, and occasionally to curb their inflow such as the ones implemented after the oil crisis. Two major types of policies were in place: immigration and assimilation policies. Other policies have also been put into place by keeping in mind both the “French spirit” (“the country of human rights”) and economic repercussions. Today, the dominant concept in French thinking about
18
The Economic Side of Religious Discrimination in France: A Review
459
the incorporation of the second generation is “Republicanism,” an ideology that emphasizes the equality of all French citizens in their dealings with the state (Silberman et al. 2006; Bourdeau and Merrill 2007). So, among immigrants who have eventually become citizens, France strives to avoid recognizing ethnic differences. Knowing that these immigrant disadvantaged groups are unable to benefit from education as much as their native counterparts, the government has designated certain zones of priority (ZEP) where schools receive additional support for these groups and has undertaken several other “democratization” programs in education. Although the Republican model conceptually aims to provide equality to all its citizens, sometimes this very ideology put into action ends up raising more issues than solving them (Silberman et al. 2006). For example, there is a great reluctance to acknowledge any ethnic divisions which is emphasized due to the secular tradition in the country – the laïcité (principle of separation of civil society and religious society). The famous ruling passed in 2004 against the display of religious symbols in school has led to a lot of backlash from the Muslim community. But, more importantly, it is the refusal to acknowledge minorities which paradoxically creates problems since it prevents ethnic/religious-specific actions from being taken. The equality of all persons, irrespective of “ethnicity” or “origin,” is a fundamental principle in the French Republic. But this, coupled with the French ideology of secularism, is the reason why France does not have affirmative action for ethnic/ religious minorities as even “positive discrimination” would be seen then as discrimination. Yet, as put forth in Cediey and Foroni (2008), after the Second World War, as a result of international treaties, several steps have been taken to curb discrimination in France: the International Convention on the Elimination of All Forms of Racial Discrimination, adopted by the United Nations in 1965 and ratified by France in 1971, and the Discrimination (Employment and Occupation) Convention (No. 111), adopted by the International Labour Organization in 1958 and ratified by France in 1981. In 1972, discrimination was defined as a criminal offense through the AntiRacism Act known as the Pleven Act. Since 2000, with the European Union’s directives, other legal actions began to come up such as the 2001 anti-discrimination law. In 2005, an independent institution (HALDE: Haute Autorité de Lutte contre les Discriminations et pour l’Egalité) was set up for combating discrimination which was later transformed in 2011. It allowed anybody who considers himself to be a victim of discrimination to bring the matter before HALDE. It also conducted campaigns, surveys, and yearly public reports of its work and its findings on discrimination. In the mid-2000s, the government took up measures to improve the economic and social integration of immigrants and “gave them the dignity guaranteed by the Universal Declaration of Human Rights,” e.g., the 2007 Reception and Integration Contract, to provide language training for increasing their chances of participating in the job market (Lochmann 2020). Later, this was replaced by the new Contrat d’Intégration Républicaine (CIR) in 2016 which conducted more extensive training sessions and also provided a certificate of professional skills at the end of the course. Even in the housing context, several subsidies were introduced for first-time house buyers in the form of subsidized loans, though they were not always useful to low-skilled migrants
460
C. J. Nordman
since they lacked financial resources (Gobillon and Solignac 2019). Yet, according to the French Labour Code, France has anti-discrimination laws that clearly state “no person may suffer prejudice in his work or employment by virtue of his origin” which shows that, if not followed and condemned systematically, various provisions exist for curbing the religious and ethnic discrimination. The available literature has presented certain solutions that could further help reduce discrimination in the economy by amending or adding to the existing antidiscrimination policies. According to Killian and Manohar (2016), improving institutional designs and practices can be a helpful solution. Educational degrees could be combined with work experience which will allow networking and help immigrants build social networks. The government could also introduce new public policies to promote the building of social networks (Senik and Verdier 2010) as well as policies targeting the labor market by making more entrepreneurial investments toward immigrants. For example, it could put in place the minimum numbers of immigrants required per employment sector and thereby incentivize them to join. But these policies are again sensitive as they could be seen as affirmative action ones. Some authors claim that, while the civil service accounts for a large proportion of employment opportunities in France, its accessibility to the descendants of immigrants, and most particularly those of North African or sub-Saharan African origin, should be thought as a strategic component of anti-discrimination policies (Meurs et al. 2008).
Ways Forward in Research This literature review and discussion have shown that there exists a relative consensus in the economic literature on the economic consequences and roots of religious discrimination in France. The selection of economic (and some sociological) studies reviewed here covers a wide range of topics, but there are still many areas of future research that they themselves suggest. Regarding the labor market, Valfort (2015) recommends that further steps should be taken to test whether discrimination varies across firm sizes in private jobs. Meurs et al. (2008) suggest tracking the trajectories of immigrant job choices to give insight into their preferences and reasons for not going after them. Algan et al. (2010) suggest evaluating how far public policies have successfully given better outcomes for immigrant assimilation in specific economic sectors. In the housing market studies, Bonnet et al. (2016) recommend combining qualitative insights with systematic survey data to find out whether there is more discrimination when residency becomes a proxy for ethnicity. Gobillon and Solignac (2019) suggest exploring housing transitions to evaluate immigrant access to homeownership and the local municipality’s conditions for them. Other authors who studied the benefits of civil services suggest more research to be conducted to see how they can be helpful for immigrant integration and job satisfaction. Finally, Lochmann (2020) suggests conducting a longer than 3-year-period study to capture the long-term effects of language proficiency on immigrants and how far that helps them access better jobs.
18
The Economic Side of Religious Discrimination in France: A Review
461
Cross-References ▶ Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination Acknowledgments The author acknowledges valuable research assistance provided by Bishwarupa Kar during an internship at the French Institute of Pondicherry in India.
References Acolin A, Bostic R, Painter G (2016) A field study of rental market discrimination across origins in France. J Urban Econ 95:45–60. https://doi.org/10.1016/j.jue.2016.07.003 Adida CL, Laitin DD, Valfort M (2010) Identifying barriers to Muslim integration in France. Proc Natl Acad Sci 107(52):384–390. https://doi.org/10.1073/pnas.1015550107 Aeberhardt R, Fougere D, Pouget J, Rathelot R (2010) Wages and employment of French workers with African origin. J Popul Econ 23:881–905. https://doi.org/10.1007/s001480090266-3 Algan Y, Dustmann C, Glitz A, Manning A (2010) The economic situation of first and secondgeneration immigrants in France, Germany and the United Kingdom. Econ J 120 (February):4–30. https://doi.org/10.1111/j.1468-0297.2009.02338 Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. Am Econ Rev 94(4):991–1013 Bonifazi C, Okolski M, Schoorl J, Simon P (eds) (2008) International migration in Europe: new trends and new methods of analysis. IMISCOE Research, Amsterdam University Press, 325 pages Bonnet F, Lale E, Safi M, Wasmer E (2016) Better residential than ethnic discrimination: reconciling audit and interview findings in the Parisian housing market. Urban Stud 53(13):2815–2833. https://doi.org/10.1177/0042098015596107 Bourdeau V, Merrill R (dir.) (2007) La république et ses démons. Essais de républicanisme appliqué, Editions Ere, coll. “Chercheurs d’ère”, 160 pages Brookings Institution (2016) Being Muslim in France. Brookings Institution Study, Washington, DC. https://www.brookings.edu/wp-content/uploads/2016/07/integratingislam_chapter.pdf Cahuc P, Carcillo S, Minea A, Valfort M (2019) When correspondence studies fail to detect hiring discrimination. IZA DP No. 12653 Cediey E, Foroni F (2008) Discrimination in access to employment on grounds of foreign origin in France: a national survey of discrimination based on the testing methodology of the international labour office. Int Migrat Papers 110(85E):67–78 Combes P, Decreuse B, Laoue’nan, M., Trannoy, A. (2016) Customer discrimination and employment outcomes: theory and evidence from the French labor market. J Labor Econ 34(1):106–170 David A, Nordman CJ (2017) Education mismatch and return migration in Egypt and Tunisia. Space, population and societies, special issue on return migration to Africa, 2017/1. https://doi. org/10.4000/eps.7110 Duguet E, Leandri N, L’horty Y, Peti P (2010) Are young French jobseekers of ethnic immigrant origin discriminated against? A controlled experiment in the Paris area. Annales d'Économie et de Statistique, No.99/100:187–215 Fredetter J (2014) Employment: the Muslim experience in (and out of) the workplace. Constructing Muslims in France: discourse, public identity, and the politics of citizenship. Temple University Press, pp. 103–125. https://www.jstor.org/stable/j.ctt14bt3kx.9 Fromentin V (2013) The relationship between immigration and unemployment: the case of France. Econ Anal Policy 43(1):51–66 Gobillon L, Solignac M (2019) Homeownership of immigrants in France: selection effects related to international migration flows. J Econ Geogr 0(2019):1–42. https://doi.org/10.1093/jeg/lbz014 Gubert F, Nordman CJ (2008) Who benefits Most from migration? An empirical analysis using data on return migrants in the Maghreb. Background paper, MENA Department, Washington DC: The World Bank; MIREM analytical report, MIREM AR2008-03, RSCAS/EUI, Florence, 2008
462
C. J. Nordman
Gubert F, Nordman CJ (2009) Migration trends in North Africa: focus on Morocco, Tunisia and Algeria. OECD J General Papers, 4, pp. 75–108 Gubert F, Nordman CJ (2011) Return migration and small Enterprise development in the Maghreb. In: Plaza S, Ratha D (eds) Diaspora for development in Africa. The World Bank, Washington DC, Chapitre 3, pp. 103–126 Héran F (2017) Avec l’immigration. Mesurer, débattre, agir, Paris, La Découverte, 327 pages Killian C, Manohar NN (2016) Highly skilled immigrant women’s labor market access: a comparison of Indians in the United States and North Africans in France. Soc Curr 3(2):138–159. The Southern Sociological Society. https://doi.org/10.1177/2329496515620644 Kymlicka W (2012) Multiculturalism: success, failure, and the future. Migration Policy Institute, Washington, DC Lefranc A (2010) Unequal opportunities and ethnic origin: the labor market outcomes of secondgeneration immigrants in France. Am Behav Sci 53(12):1851–1882. https://doi.org/10.1177/ 0002764210368100 Lochmann A (2020) L’effet de la formation linguistique sur l’intégration économique des immigrés en France. In: El-Mallakh N, Rapoport H (eds) Migration, intégration et culture: approches économiques, pp 11–15. http://icmigrations.fr/2020/06/08/defacto Meng X, Meurs D (2009) Intermarriage, language, and economic assimilation process: a case study of France. Int J Manpow 30(1/2):127–144. https://doi.org/10.1108/01437720910948447 Meurs D, Pailhé A, Simon P (2006) Persistance des inégalités entre générations liées à l’immigration: l’accès à l’emploi des immigrés et de leurs descendants en France. Population 61:645–682. https://doi.org/10.3917/popu.605.0763 Meurs D, Pailhé A, Simon P (2008) Discrimination despite integration: immigrants and the second generation in education and the labour market in France. In: Bonifazi C, Okolski M, Schoorl J, Simon P (eds) International migration in Europe: new trends and new methods of analysis. IMISCOE Research, Amsterdam University Press, pp 247–269 Pierné G (2013) Hiring discrimination based on national origin and religious closeness: results from a field experiment in the Paris area. IZA J Labor Econ 2(4) http://www.izajole.com/content/2/1/4 Riach PA, Rich J (1991) Testing for racial discrimination in the labour market. Camb J Econ 15(3): 239–256 Richard J (2000) Unemployment among young people of foreign origin in France: ways of measuring discrimination. Rev Eur Migr Int:101–135. https://doi.org/10.3406/remi.2000.1741 Richard J (2013) Unemployment of people of foreign origin in France: the role of discrimination. “Immigration and the life course”. Canadian Stud Population 40(1–2):75–88 Senik C, Verdier T (2010) Segregation, entrepreneurship and work values: the case of France. J Popul Econ 24:1207–1234. https://doi.org/10.1007/s00148-010-0328-6 Silberman R, Fournier I (2008) Second generations on the job market in France: A persistent ethnic penalty. A contribution to segmented assimilation theory. Rev Fr Sociol 49:45–94 Silberman R, Alba R, Fournier I (2006) Segmented assimilation in France? Discrimination in the labour market against the second generation. Ethn Racial Stud 30(1):1–27. https://doi.org/10. 1080/01419870601006488 Simon P, Steichen E (2014) The labour market integration of new immigrants in France. In: Slow motion: the labour market integration of new arrivals in Europe. Migration Policy Institute and International Labour Organization, Washington, DC and Geneva Valfort M (2015) Religious discrimination in access to employment: a reality, antisemitism and islamophobia on the French labour market. The Institut Montaigne. Policy Paper – October 2015, pp 43–57 Valfort M (2020) Anti-Muslim discrimination in France: evidence from a field experiment. World Dev 135. https://doi.org/10.1016/j.worlddev.2020.105022 Vourc’h F, De Rudder V, Tripier M (1999) Foreigners and immigrants in the French labour market: structural inequality and discrimination. In: Wrench J et al (eds) Migrants, ethnic minorities and the labour market. Palgrave Macmillan, pp 72–91
Socioeconomic Disparities Among Racialized Immigrants in Canada
19
Karun K. Karki, Delores V. Mullings, and Sulaimon Giwa
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theoretical Constructs of Dimensions of Socioeconomic Disparities . . . . . . . . . . . . . . . . . . . . . . . . . Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dimensions of Social Disparities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dimensions of Economic Disparities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
464 468 469 469 472 476 477 477
Abstract
Canada’s economic development is built through waves of migrants and immigrant labor. The Canadian labor market, however, is characterized by discriminatory divisions rooted in its history of racism. Discrimination is prejudgment and unjust treatment of individuals and groups based on socially constructed identity markers such as race, ethnicity, gender identity, and immigration status. It is the exclusion of individuals or groups from full participation in the society. This chapter focuses on two dimensions of disparity among racialized immigrants in the Canadian society: economic disparity and social disparity. This chapter begins with an introduction of immigration to Canada. The third section further substantiates the theoretical construct through an extent review of literature on socioeconomic disparities among racialized immigrants in the Canadian labour market, followed by conclusion and recommendations. K. K. Karki (*) School of Social Work and Human Services, University of the Fraser Valley, Abbotsford, BC, Canada e-mail: [email protected] D. V. Mullings · S. Giwa School of Social Work, Memorial University of Newfoundland, St’ Johns, NL, Canada e-mail: [email protected]; [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_21
463
464
K. K. Karki et al.
Keywords
Socioeconomic disparities · Racialized immigrants · Employment discrimination · Canada
Introduction Canada is one of the top immigrant host countries in the world. It is generally viewed as more welcoming to immigrants than other countries and has in many ways a better developed immigration system (Helliwell et al. 2018). The demographic landscape has changed significantly since the late 1960s with the introduction of the points system to government immigration policy in 1967. Some scholars argue that the introduction of the points system was a radical break from the past. Canada eliminated a racist dimension of the immigration policy by opening its doors to immigrants from traditionally non-European countries, historically categorized as undesirable people (Borjas 1993; Wanner 2003). The underlying structure of the immigration policy is very much a continuation of the legacy of colonialism, however. The points system is designed to attract more skilled workers and professionals from around the world who are fluent in English or French. Scholars such as Bauer (2003) and Reitz (2011) conclude that highly educated and skilled immigrants often end up with survival jobs that do not need special skills and are well below their competencies. Current Canadian immigration policy can be argued as a racialized, gendered, and class-based regulatory mechanism that continues to determine who can enter Canada, when, and on what terms. The points system of Canada’s immigration policy has been critiqued widely. For example, Tannock (2011) argues that the “skill-based immigration regime discriminatory and violate core principles of public education provision, unjustly create second-class tires of immigrants officially classified as low-skilled. . . and contribute to a growing problem of brain drain of the highly skilled from sending counties worldwide” (p. 1330). Immigration policy marks “us” through the binary that allows judgment of “them” (Jeyapal 2018). The points system, which still in effect today, is considered racist and continues to uphold implicit biases within the immigration policy framework. The Immigration Act of 1976, implemented in 1978, made Canada a destination for immigrants from countries around the globe. The act’s primary objectives were to promote Canada’s demographic, economic, social, and cultural goals and include priorities of family reunification. Today, Canada’s diverse immigration flows have come to be understood not only as contributors to population growth but also as engines of social transformation and economic growth. This has led to the growth of racially, culturally, and religious Canadian society we can see today in some large urban centers in Canada such as Toronto, Vancouver, and Montreal.
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
465
Fig. 1 Percentage of racialized population, 1981–2016. (Data source: Statistics Canada 2016)
Figure 1 shows a demographic trend of visible minority1 population in Canada. In 1981, about 1.1 million people belonged to a racialized group, representing 4.7% of the total Canadian population (Statistics Canada 1981), while in the 2016 census, about 7.7 million individuals identified themselves as belonging to the racialized2 population, representing more than one-fifth (22.3%) of Canada’s population (Statistics Canada 2016). In 2019, Statistics Canada reported the racialized population as the fastest growing in Canada, with over 75% of new immigrants being members of racialized groups. Canada is often portrayed as a good example of how immigration works. Scholars (e.g., Banting and Kymlicka 2010; Helliwell et al. 2018) contend that, compared to other countries in the Global North, Canada seems much better at integrating newcomers due to its effective and efficient social services, including emergency housing, temporary health services, legal aid, and education. The image Canada projects internationally is of a country more welcoming to immigrants than others, and in fact Canada has, in many ways, a better immigration system. Today, Canada celebrates diversity and multiculturalism as its salient identity. There are, We are not using the term “visible minority” because non-whites are global majority populations, but they are categorized as a minority in countries such as Canada. Instead of “visible minority,” we have used “racialized” to mean people who are not white identified and experienced slavery, colonization, and erasure. In 2007, the United Nations called for Canada not to use the term “visible minority” as the term is racist. The Employment Equity Act defines visible minorities as “persons, other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour.” 2 The term “racialized” is used to acknowledge “race” as a social construct and a way of describing a group of people. According to the Canadian Race Relations Foundation (2008), racialization is the process through which groups come to be designated as different and, on that basis, subjected to differential and unequal treatment because of their race, ethnicity, language, economics, and religion. 1
466
K. K. Karki et al.
100 90 80 70 60 50 40 30 20 10 0 1871
1891
1911
1931
1951
1971
1991
2001
2011
2016
Western Europe(Including British Isles)
United States
Eastern & Southern Europe
Caribbean, Bermuda, Central and South America
Sub-Saharan and North Africa
Asia and Middle-East
Oceania and other
Fig. 2 Migration trend in Canada (1871–2016). (Data source: Statistics Canada, Censuses of population 1871–2016)
however, scholars who criticize Canada’s so-called multiculturalism, arguing that it works better in theory than in practice. In an interview with CBC Radio, Keith Banting, a professor at Queen’s University, states: “while Canada is not immune to the kinds of xenophobic nationalism that have gripped European countries, the Canadian philosophy and practice of multiculturalism is unlikely to lead to the same kind of backlash against immigrants, refugees and multiculturalism” (CBC February 22, 2019 para 8). The migration trend in Fig. 2 shows a sharp decline of Western European immigrants, a moderate decline of US and Eastern and Southern European immigrants, but a sharp increase of Asian immigrants and a moderate increase of other geographical regions. Despite the ever-increasing ethnic population in Canada, the multicultural policy is less effective in implementing its promise of assistance to cultural groups. Canada’s official acceptance of racial and cultural diversity contrasts with the fact that many racialized people report that they frequently experience discrimination and unfair treatment. Statistics Canada (2016) projects the proportion of immigrant population by 2036 in Canada. As shown in Fig. 3, between 55.7% and 57.9% of immigrants are projected to have been born in Asia – mainly in China, India, and the Philippines – while between 15.4% and 17.8% would have been born in Europe by 2036. This would be a reversal of the situation observed in the 1980s. The proportion of immigrants from Africa would continue to increase between 11.0% and 11.9% by
19
Socioeconomic Disparities Among Racialized Immigrants in Canada 2036 (Projected %)
st e er n Eu N ro or pe th er n Af W ric es a te rn Af ric C en a tra lA fri Ea ca st er n Af So W ric ut es a he tA rn si Af a ric & M a id dl eEa st Ea st er n A So si a ut he rn As So ia ut he O as ce tA an si ia a an d ot he rs
e
Ea
Eu rn he
ut
te es
ro p
ro p
e
Eu
ro p Eu
rn So
an
er n W
a
ud rm Be
d
h ut C
ar
ib
be
an
So
N or th
a
er ic Am
er ic
er ic tra
lA
m
Am N or th
C en
a
20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
a
Percent (%)
2011 (Estimated %)
467
Fig. 3 Immigrants living in Canada by region of birth, 2011 (estimated) and 2036 (projected). (Data source: Statistics Canada 2016)
2036. The following figure demonstrates the projection of immigrants living in Canada by region of birth in 2036 (Fig. 4). Helliwell, Layard, and Sachs (2018) state that Canada is ranked seventh happiest country in the world to live in. The authors have considered several compounding factors, including the tightening of immigration policy in the United States and Europe and Britain’s imminent departure from the European Union (Brexit), to show Canada as a home for potential immigrants. Brexit has exacerbated nationalist sentiments, particularly in England. The aftermath has increased tension between the United Kingdom and the EU and “a wider debate over the role of the nation state and the rise of populism in an age of globalisation” (Sandford 2020 para. 5). The Brexit movement gained popularity under the right-wing populism that moved across the globe with the election of Presidents Emmanuel Macron (France), Sergio Mattarella (Italy), and Donald Trump (USA) and Prime Minister Scott Morrison (Australia). The right-wing agenda continues to influence anti-immigrant sentiment worldwide (Sandford 2020). While Canada attracts highly educated and skilled immigrants, many are challenged to find meaningful employment in their industry and are, therefore, excluded in a practical sense from socioeconomic dimensions of Canadian society. With the increasing numbers of racialized immigrants, there has been a marked increase in discrimination on the basis of race, ethnicity, gender, and citizenship status. Since the 1980s, immigrants’ socioeconomic success has been worsening. For example, in 1980, new immigrants were earning 80% of what Canadians earned; by 1996, this number fell to 60% (Reitz 2006). This indicates a brain waste that has steadily increased over the past four decades. The Canadian labor market is polarized, with a
468
K. K. Karki et al.
Fig. 4 Dimensions of socioeconomic disparities among racialized immigrants in Canada
complex axis of gender, race, and ethnicity (Galabuzi 2006). As a result, many racialized immigrants are being channeled into menial, part-time, insecure employment and/or becoming unemployed and/or underemployed. The following section discusses theoretical constructs of dimensions of socioeconomic disparities among racialized immigrants in the Canadian labor market.
Theoretical Constructs of Dimensions of Socioeconomic Disparities We conceptualize dimensions of socioeconomic disparities using theoretical insights from Becker’s (1957, 1993) ideas of the “economics of discrimination” and “human capital.” According to Becker (1957), socioeconomic disparities occur when firms have monopoly power, where certain group of employees are discriminated against based on their identity markers such as race, gender, ethnicity, and citizenship status. He calls this a “taste for discrimination,” meaning that the firm places a disamenity value on employing minoritized workers such as people of color, immigrants, and women. Such disparities and discrepancies exist where employers make generalized assumptions about the worth of employees in terms of their employability in the labor market (Becker 1957) and, therefore, given preference for one group over another. According to Jacquemet and Yannelis (2012), “discriminatory behaviour is
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
469
part of a large pattern of unequal treatment of any member of non-majority groups, ethnic homophily” (p. 824). The authors further argue that ethnic homophily emerges if the discrimination extends to anyone who does not belong to one’s own racial group. In a similar vein, Spence (1973) and Stiglitz (1975) explain labor market acquisition, especially how organizations conduct the screening of their potential employees. According to Spence (1973), the screening hypothesis suggests that educational credentials are a convoluted and expensive way of signaling to employers’ abilities to select the more skilled workers. The notion of screening allows employers screen out applicants based on educational achievements, such as academic degrees, training certificates, skills, and experiences (Stiglitz 1975), and at the same time discriminate against those that they consider undesirables even when they are fully qualified educationally. Becker’s (1993) human capital theory assumes that education, skills, and work experience determine an individual’s potential earning. In another word, the main assumption of this theory is that educational achievements, on-the-job training, and other skills result in an accumulation of knowledge along with the ability to increase work productivity, which is reflected in individuals’ level of earnings. An individual’s human capital serves as a fundamental investment with expected returns. Human capital is an asset to an organization much like a piece of equipment. Considering these theoretical footings, we argue that these theories aptly explain the integration of skilled, racialized immigrants in the Canadian labor market. Studies conclude that skilled, racialized immigrants have lower labor market integration and/or employment outcomes in Canada. Discrimination restricts many immigrants from fully utilizing their skills in occupations commensurate with their professions, resulting not only in the waste of their brain but also a huge loss to the national economy. In the following section, we substantiate the aforementioned theoretical underpinnings through a critical review of existing literature.
Literature Review The dimensions of socioeconomic disparities encompass a spectrum of social and economic barriers, including devaluation of foreign credentials and discrimination based on race, ethnicity, and gender, that often locate immigrants at the very bottom of the skill hierarchy. In the subsequent sections, we discuss separately the dimensions of social and economic disparities.
Dimensions of Social Disparities Dimensions of social disparities refer to the situation when certain groups are deprived of equal opportunities in terms of income, education, employment, healthcare, etc., because of their race, gender, nationality, and other socially constructed identities. In Canada, such exclusionary discrimination can be traced
470
K. K. Karki et al.
back to 1885 when Canada enacted the Chinese Immigration Act, initiating a head tax to discourage Chinese immigration. Subsequently, various immigration acts were introduced based on restriction and the exclusion of undesirable immigrants from similar areas such as Japan and Southeast Asia (Continuous Journey Regulation, 1908). The exclusionary discrimination – deeply rooted in a colonial and racist worldview – has been exercised through institutions, policies, and particular forms of knowledge. Existing research on ethnic diversity in the Canadian labor market has focused on identifying and accounting for discrimination against historically marginalized groups. Reitz (2011) contends that highly skilled immigrants, particularly of Indian, Chinese, Caribbean, or Arab background, are still “driving taxis, mopping floors, bagging groceries, guarding office buildings, delivering pizzas, waiting tables, and working at call centres” (Para 1). This confirms that skilled, minoritized immigrants are not faring well in the Canadian labor market. Oreopoulous (2011) conducted a study by sending thousands of identical fabricated resumes using both ethnic-sounding and English-sounding names. He found that the applicants with English-sounding names were approximately three times more likely to receive a callback for a job interview than foreign-sounding names such as from China, India, or Pakistan. These findings support the theoretical footings of the “taste for discrimination” theory (Becker 1957), namely, that the observed discrimination may arise because of employers’ reluctance to hire people who are from ethnic minority groups. Employers may be concerned that jobseekers with ethnic-sounding names may speak accented English or be unfamiliar with local culture and customs. In the Canadian labor market, racialized immigrants are institutionally and systemically disadvantaged. For example, Reitz (2011) argues that devaluation of internationally acquired credentials by accreditation bodies in Canada is intended to reserve the upper segment of the labor market for Canadian-born citizens. This indicates that race is interpreted as the common denominator in explaining the poor labor market outcomes or socioeconomic marginalization for racialized immigrants. Studies suggest that the Canadian labor market is gender segregated, both in the horizontal and vertical dimensions (Guppy et al. 2019); the former refers to the under (over) representation of a certain group of occupations or sectors not ordered by any criterion, while the latter refers to the under (over) representation of a clearly identifiable group of workers in occupations or sectors at the top of an ordering based on desirable attributes such as income, prestige, and job stability (Bettio and Verashchagina 2009; Kiaušienė et al. 2011). Other studies have explored the organizational positions or hierarchies as an important location of gender segregation. Some feminist critics argue that when gender is added to the economic migration model, women are grossly neglected; gender simply becomes a “control variable,” which Grieco and Boyd (2003) call the “add women, mix, and stir” approach (p. 2). Acker (2009) argues that organizational structure is not gender neutral; all-male enclaves are at the pinnacle of large state and economic organizations, while women are overrepresented at the bottom of organizational structure. Thus, feminist critics argue that the organizational hierarchy has a dual structure, bureaucracy, and patriarchy; the former has its own dynamic, and gender enters through patriarchy.
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
471
Similarly, studies have concluded that the Canadian immigration policy is gendered such that less attention has been given to the implications of skilled immigration schemes for women (Benhabib and Resnik 2009). For example, looking at the economic class immigration, men are disproportionately primary applicants, while women are secondary applicants. The underrepresentation of women as primary applicants raises the concern of gender inequality. While it is also critical to acknowledge that some women immigrate to Canada as primary applicants and bring their male partners as dependents, there has been significant feminization of skilled migrant intake in the health science professions. In another word, more women are recruited for health profession such as nurses, personal support workers, and caregivers. However, they face systemic and institutional discriminations and barriers and racial and ethnic prejudice. Elabor-Idemudia (1999) states that “racially constructed gender ideologies and images often portray Black women as ‘naturally’ suited for jobs in the lowest stratum of a labour market segmented along gender lines” (p. 40). Racialization and feminization of occupations in Canada are suggested as more complex, differentiated, and even hierarchical that have significantly disadvantaged immigrants. This situation has led many immigrants to work in precarious employment and vulnerable conditions. Benach et al. (2014) suggest that precarious and/or vulnerable employment conditions may increase the risk of chronic illness such as cardiovascular, respiratory, and mental health disease, diabetes, and cancer. Thus, gender-based analysis has been an important tool in exploring gender inequities located in the skilled immigration program. Moreover, racialized women are overrepresented in occupations such as salesperson, domestic and related helpers, cleaners, care and professional services workers, office clerks, and motel, café, and hotel and restaurant services workers (Elabor-Idemudia 1999; Premji and Lewchuk 2014). Racialized men, on the other hand, are predominantly overrepresented in occupations such as motor vehicle drivers (e.g., truck, taxi, Uber), factory and warehouse workers, construction workers, gas station attendant, and selfbusinesses (Bauder 2003; Karki 2020; Reitz 2011). Today Canada is undeniably dependent on temporary foreign workers, and their proportion has been on the rise. The vast majority of them can only secure temporary terms of employment through Temporary Foreign Worker Program (TFWP) or the Seasonal Agricultural Workers’ Program (SAWP) in Canada. They are denied the rights to permanent residence and lack freedom of movement between jobs and employers as well as access to rights and entitlements of social citizenship (Arat-Koc 1999). Removing the rights of workers leads to the flexibilization of their labor. Cohen (2006) argues that, while the temporary foreign workers are not technically slaves, and the employers involved do not claim absolute proprietorial rights of them and they cannot be bought or sold, they nevertheless “characteristically do not enjoy full social and civic rights compared with citizen workers” (p. 59). Thus, these disparities have and continued to and continue to negatively impact many racialized immigrants through poverty, economic hardships, deterioration of mental wellness, “high level of household relationship strain and tension, reduction in family time and leisure time, and overall sense of disempowerment at the family/
472
K. K. Karki et al.
household level” (Premji and Lewchuk 2014, p. 134). The impact is more acute for racialized immigrant women given the nuanced and compounded reality of racism, sexism, and immigration status, among others.
Dimensions of Economic Disparities Statistics Canada (2016) shows that immigrants have higher educational attainment compared to Canadian-born population. As shown in the Table 1, immigrant populations are more likely than Canadian-born populations to have completed university degrees. Despite their higher level of education, racialized immigrants are disproportionally relegated to the lower end of the Canadian job market. The unemployment rates for immigrants are higher than their Canadian-born counterparts (see Fig. 5). Moreover, the unemployment rate for immigrant women is the highest among all groups. Statistics Canada (2016) indicates that the unemployment rate for immigrant women compared to similar aged nonimmigrants was almost three times higher. Oreopoulous (2011) found that the median wage for immigrants who landed in Canada between 2001 and 2005 was almost 36% lower than that of native-born workers with similar human capital factors. Similarly, within immigrants, there is a substantial wage gap between men and women. Figure 6 shows the average hourly wages of immigrant cohorts of 2000 and 2001 by immigration class. The hourly wages represent 2 years and 4 years after landing in Canada. The often-cited barriers to employment integration for racialized immigrants are devaluation of credentials achieved outside of Canada, complex and lengthy accreditation/licensing procedures, immigrants’ lack of professional connections, their unfamiliarity with professional and business vocabulary in Canada, and, of course, a lack of Canadian experience (Bauder 2003; Karki 2020; Mullings et al. 2020; Reitz 2011). These artificial barriers put many qualified immigrants at a disadvantaged position, such as downward occupational mobility in the Canadian labor market. Reitz (2011) refers to this phenomenon as the “taxi-driver syndrome,” describing highly educated immigrants working in precarious and low-end employment because of the inability to find meaningful employment commensurate with their skills and qualifications. When new immigrants experience the “taxi-driver Table 1 Level of education among Canadian-born and immigrant populations by sex Level of education High school certificate Trade certificate or diploma College diploma Below the bachelor level University degree Source: Statistics Canada 2016
Males (%) Canadian-born 23.9 17.9 20.8 3.3 20.4
Immigrants 19.3 8.6 15.5 8.6 50.8
Females (%) Canadian-born 22.9 9.2 27.2 4.3 26.6
Immigrants 19.5 5.7 18.5 9.9 49.6
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
473
Fig. 5 Unemployment rates of Canadian-born and immigrant population, 2016. (Data source: Statistics Canada 2016)
Fig. 6 Average hourly wage by immigration class, 2 years and 4 years after landing in Canada. (Data source: Toronto Immigrant Employment Data Initiative 2010)
syndrome” in a host country, it delays or prevents their integration into the host labor market. This leads them to work in jobs that require a lower level of education than they actually possess. This kind of job-education mismatch is known as overeducation. Studies suggest that there is a substantial earnings disadvantage and overeducation among skilled, racialized immigrants in the Canadian labor market. Li, Gervais, and Duval (2006) found that more than 52% of recent immigrants were
474
K. K. Karki et al.
working in jobs for which they were formally overqualified compared to 28% of their Canadian-born counterparts. Sharaf (2013) found a high incidence of overeducation among recent immigrants (i.e., 6 months after they landed in Canada), for example, 76.3% for immigrant males and 71.8% for females. These figures did not improve much after 4 years from arrival, when 70.4% of males and 64.6% females were overeducated. There is an adverse effect of overeducation not only on individual immigrants but also the nation’s economy. The Conference Board of Canada (2004) estimated that the Canadian economy loses up to five million Canadian dollars per annum due to overeducation arising from general lack of thriving. Recent immigrants to Canada struggle in the labor market. The adverse effect of overqualification is associated with a wage penalty for immigrants. Not all immigrants experience employment discrimination identically; systemic racism influences the labor market involvement with respect to employment and career attainment. Immigrants are significantly disadvantaged at the institutional level as the professional associations and licensing bodies do not fully recognize credentials achieved outside of Canada. This indicates prejudice against foreign credentials and ignorance of the true value of these credentials in the Canadian labor market. Similarly, employers become gatekeepers who demand Canadian credentials in order for racialized immigrants to compete for professional and upwardly mobile positions. Bauder (2003) suggests that “professional associations and the state actively exclude immigrants from the most highly desired occupations in order to reserve these occupations for Canadian-born and Canadian-educated workers” (p. 769). We also know that there are nuanced within the proposed Canadian-born and Canadian-educated individuals as racialized Canadian also experience employment discrimination; however, this discussion is beyond the scope of this chapter. Oreopoulous (2011) conducted a study that involved sending thousands of resumes to online job postings across a wide set of occupations and industries in Toronto to investigate factors that affect employers’ decisions on whether or not to contact an applicant for a job interview. The study found that employers significantly discount the foreign-earned credentials and work experience of immigrants. The callback rate of job interviews for applicants with professional experience outside of Canada was almost half compared to applicants with Canadian experience. Similarly, racist perception of whose culture and language is acceptable makes a difference in employment attainment and in which area; racialized immigrants are often hired in the lower end of employment. Skill is not neutral but socially constructed, and it is implicated in the social, cultural, and economic organization. Scholars such as Bauder (2003), Karki (2020), and Reitz (2011) argue that, in the Canadian labor market, immigrants’ credentials, skills, and knowledge are color-coded, gendered, and racialized. Guo (2015) states, “it is the ‘colour’ of the skill associated with immigrants’ skin colour rather than the skill itself which causes deskilled, skilled immigrants” (p. 254). Grugulis and Vincent (2009) maintain that the emphasis on soft skills may support racialized employment discrimination by legitimizing stereotypes about immigrant women while ignoring the structural barriers that limit their potential for employment attainment and mobility. The figure below shows the unemployment
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
475
Fig. 7 Unemployment rate of skilled immigration class (principal applicant and spouse): 6 months, 2 years, and 4 years after landing in Canada. (Data source: Toronto Immigrant Employment Data Initiative 2010)
rate by immigration, class, and gender among immigrant cohorts of 2000–2001 in the period of 6 months, 2 years, and 4 years after landing in Canada. Figure 7 shows that the unemployment rates for principal applicants (both male and female) have declined in a similar proportion over the 4 years of period; however, with respect to spouses, the unemployment rate for men decreased from 34.5% to 9.7%, whereas for women 24.7% to 18.5%. Race and gender inequities are reproduced and perpetuated in Canadian societies. In the context of skilled, racialized immigrants in Canada, the cause of the income differential between immigrants and nonimmigrants is due to the differential values given to their qualification, skills, and experiences. Li (2008) revealed that racialized immigrants’ earning was substantially less than immigrants of British origin (north and west European origin). For example, Chinese and African as well as South Asian immigrants earn $18,000 and $13,000, respectively, less than immigrants from Britain (normally white) with identical human capital characteristics. Racialized women’s employment discrimination is exacerbated as their skills and personal qualities are excluded or undervalued which results in this population being overrepresented in the secondary labor market working long hours in the lower rungs of the labor market where remuneration is low, work is precarious, and benefits are nonexistent. For example, Chui (2011) found that almost 55% of 3.2 million immigrant women were racialized and overrepresented in low-paid, low-skilled jobs characterized by high risk and precarity. Regardless of their qualifications and professional work experience, many racialized immigrant women are relegated to precarious employment, with part-time, flexible hours and no or less employment security, with fewer or no benefits (Chui 2011).
476
K. K. Karki et al.
The major tenet of human capital theory (Becker 1957) is that there is a correlation between education and income. In the context of skilled, racialized immigrants in Canada, for example, the cause of the income differential between immigrants and nonimmigrants is due to the differential values given to their qualification, skills, and experiences. Evidence, however, suggests that this is not always the case; rather, societal factors often operate as perceived meritocracy. As a result, immigrants’ human capital has been grossly undervalued in the Canadian labor market, which pushes them to the lower end of employment.
Conclusion and Recommendations Canada is a major immigrant-receiving country whose reliance on immigration to fill skill shortage to its labor market is not translated into practice. Although there have been a number of government initiatives and credential assessment services in provinces across Canada, evidence suggests that, sadly, all these efforts to address immigrant brain waste are inadequate in magnitude to make more than a small dent in the overall problem (Reitz 2011). When racialized immigrants are denied equal social and economic opportunities, the country will not reap the benefits of the potential of this growing proportion of its population in Canada. As we discussed throughout the chapter, one of the major barriers to occupational integration is a complex, costly, and lengthy accreditation and licensing requirement imposed by professional associations in many regulated occupations. There have been efforts made to ameliorate the issue of credential assessment and licensing; however, there is a lack of coordination among regulatory/licensing bodies, professional associations, and provincial and federal governments. Immigration falls under the federal jurisdiction; however, most regulated and unregulated professions fall under provincial jurisdiction. The lack of coordination between the federal government (immigration) and provincial government (labor market) has contributed to the persistence of inequality in access to professions and trades of internationally trained immigrants. Canada admits 150,000 to 200,000 immigrants for permanent residency annually in order to fill skilled labor market shortages, but this has not translated into ready jobs for internationally trained immigrants. Many immigrants face insurmountable barriers to obtaining jobs in their professions (Premji and Lewchuk 2014). We recommend that the provincial/territorial government develop bridging programs such as employmentrelated language training, sector-specific orientation, mentorship and professional networking, as well as information about licensing procedures both before and after arrival. This may minimize immigrants’ underemployment rates. Thus, an important policy implication is that all levels of governments need effective and efficient coordination in terms of immigrants’ employment integration policies. Similarly, it is also evident that immigrants experience exclusionary discrimination such as racial and gender discrimination. Further, the requirement of Canadian work experience has also been identified as a mechanism for employers with subtle prejudices to justify not hiring immigrants. Li (2001) states, “gender, racial origin, and foreign credentials tend to interact to produce complex outcomes for various
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
477
groups of degree holders” (p. 32). Evidence suggested that Canadian public policy has placed a premium on occupational skills and educational attainment. The systemic discrimination in employment, however, seems to have impeded a better labor market outcome for racialized immigrants. Teelucksingh and Galabuzi (2005) state, “the intensified racial stratification of Canada’s labour market under neoliberal restructuring has led some to observe that what was once described as an ethnic Canadian vertical mosaic in now color coded” (p. 5). Social indicators such as higher rates of poverty, sectoral and occupational concentrations along racial lines, high unemployment and underemployment, higher rate of job-skill mismatch, etc. suggest the need to revisit Canada’s public policy. We recommend that governments, employers, and regulators of professions and trades need to systematically address the issue of employment discrimination by working toward eliminating barriers to access to employment. One way to do it is to revisit the existing policy and strictly implement policies that adopt principles of employment equity. Finally, it was evident that the idea of “inequality regimes” (Acker 2009) in the Canadian labor market produced a complex pattern of race, gender, and class inequalities. For example, downward occupational mobility, occupational segregation, and wage differences illustrate the complex nature of inequality. Moreover, women are found to be significantly disadvantaged. Research suggests that institutionalizing, and thus legitimizing, female leadership increases equality (Haveman et al. 2009). We recommend that the gender pay gap should be reduced not only in policy but also in practice. This is not a quick fix; it requires a very coherent policy response, involving multiple stakeholders such as government organizations, professional associations, educational institutions, licensing bodies, and employers. Premji and Lewchuk (2014) also argue that “promoting stable employment and wellbeing of racialized immigrant women needs to become one of the top policy priorities in Canada and a key benchmark indicator for measuring and achieving a more equitable nation.” (p. 137–138). Particular attention should be given to the low level of wages in professions and sectors that tend to be dominated by women and to the reasons that lead to reduced earnings in professions and sectors in which women become more prominent.
Cross-References ▶ The Economic Side of Religious Discrimination in France: A Review
References Acker J (2009) From glass ceiling to inequality regimes. Sociologie du Travail 51:199–217 Arat-Koc S (1999) Neo-liberalism, state restructuring and immigration: changes in Canadian policies in the 1990s. J Can Stud 34(2):31–56 Banting K, Kymlicka W (2010) Canadian multiculturalism: global anxieties and local debates. Br J Canadian Stud 23(1):43–72 Bauder H (2003) Brain abuse or the devaluation of immigrant labour in Canada. Antipode 35(4): 699–717
478
K. K. Karki et al.
Becker GS (1957) The economics of discrimination. University of Chicago Press, Chicago Becker GS (1993) Human capital: a theoretical and empirical analysis with special reference to education. University of Chicago, Chicago Benach J, Vives A, Amable M, Vanroelen C, Tarafa G, Muntaner C (2014) Precarious employment: understanding an emerging social determinant of health. Ann Rev Public Health 35:229–253. https://doi.org/10.1146/annurevpublhealth-032013-182500 Benhabib S, Resnik J (2009) Migrations and mobilities: citizenship, Borders, and Gender. New York University Press, New York Bettio F, Verashchagina A (2009) Gender segregation in the labour market: root causes implications and policy responses in the EU. Publications Office of the European Union, Luxembourg. Retrieved from https://eige.europa.eu/library/resource/aleph_eige000000203 Borjas GJ (1993) Immigration policy, national origin, and immigrant skills: a comparison of Canada and the United States. In: Card D, Freeman RB (eds) Small differences that matter. University of Chicago Press, Chicago, pp 21–43 CBC (2019, February 22) How did multiculturalism become so central to Canada’s identity? Retrieved from https://www.cbc.ca/radio/thesundayedition/the-sunday-edition-for-february-242019-1.5029453/how-did-multiculturalism-become-so-central-to-canada-s-identity-1.5029456 Chui T (2011) Immigrant women. Statistics Canada catalogue no. 89-503-X. Retrieved from http:// www.statcan.gc.ca/pub/89-503-x/2010001/article/11528-eng.pdf Cohen R (2006) The proletariat at the gates: migrant and non-citizen labour, 1850–2000. In: Cohen R (ed) Migration and its enemies: global capital, migrant labour and the nation- state. Ashgate, Aldershot, pp 39–62 Elabor-Idemudia PA (1999) The racialization of gender in the social construction of immigrant women in Canada: a case study of African women in a prairie province. Canadian Woman Stud 19(3):38–44 Galabuzi GE (2006) Canada’s economic apartheid: The social exclusion of racialized groups in the new century. Canadian Scholars’ Press, Toronto Grieco E, Boyd M (2003) Women and migration: Incorporating gender into international migration theory (Working Paper No. 98-139). Retrieved from http://homes.chass.utoronto.ca/~boydmon/ research_papers/gender_inequality/grieco_and _boyd.pdf Grugulis I, Vincent S (2009) Whose skill is it anyway? Soft skills and polarization. Work Employ Soc 23(4):597–615 Guo S (2015) The colour of skill: contesting a racialized regime of skill from the experience of recent immigrants in Canada. Stud Contin Educ 37(3):236–250 Guppy N, Sakumoto L, Wilkes R (2019) Social change and the gendered division of household labour in Canada. Can Rev Sociol 56(2):145–291 Haveman HA, Joseph PB, Lisa EC (2009) Good times, bad times: the effects of organizational dynamics on the careers of male and female managers. Res Sociol Work 18:119–148 Helliwell JF, Layard R, Sachs JD (2018) World happiness report 2018. Retrieved from https://s3. amazonaws.com/happiness-report/2018/WHR_web.pdf Jacquemet N, Yannelis C (2012) Indiscriminate discrimination: a correspondence test for ethnic homophily in the Chicago labour market. Labour Econ 19(6):824–832 Jeyapal D (2018) Nationalism, neo-liberalism, and the securitization of Canadian immigration policy. In: Harding R, Jeyapal D (eds) Canadian social policy of social workers. Oxford University Press, Oxford, pp 318–340 Karki KK (2020) Once a lawyer, now a security guard. . .and anybody can realize how much it is hurting, right? The experiences of deskilling among skilled minoritized immigrants in Waterloo Region, Southwestern Ontario. Wilfrid Laurier University, Canada Kiaušienė I, Štreimikienė D, Grundey D (2011) On gender stereotyping and employment asymmetries. Res Issues Sociol Res 4(4):84–97 Li PS (2001) The market worth of immigrants’ educational credentials. Can Public Policy 27(1): 23–38
19
Socioeconomic Disparities Among Racialized Immigrants in Canada
479
Li PS (2008) Immigration from China to Canada in the age of globalization: Issues of brain gain and brain loss. Pac Aff 81(2):217–239 Li C, Gervais G, Duval A (2006) The dynamics of over qualification: Canada’s underemployed university graduates. Statistics Canada Analytical Paper catalogue no. 11-621-MIE -No.039 Mullings DV, Giwa S, Karki KK, Gooden A, Shaikh S, Spencer EB, Anderson W (2020) The settlement and integration experience of temporary foreign workers living in an isolated area of Newfoundland and Labrador, Canada. J Int Migr Integr. https://doi.org/10.1007/s12134-02000788-9 Oreopoulous P (2011) Why do skilled immigrants struggle in the labour market? A field experiment with thirteen thousand resumes. Am Econ J Econ Pol 3:148–171 Premji S, Lewchuk W (2014) Racialized and gendered disparities in occupational exposures among Chinese and white workers in Toronto. Ethn Health 19(5):512–528 Reitz JG (2006) Tapping immigrants’ skills: New directions for Canadian immigration policy in the knowledge economy. Evid Policy 11(1):1–18 Reitz JG (2011) Taxi driver syndrome: behind the scenes immigration changes are creating new problems on top of old ones. Literary Rev Canada Retrieved fromhttps://reviewcanada.ca/ magazine/2011/03/taxi-driver-syndrome/ Sandford A (2020, April 7) Post-Brexit guide: where are we now-and how did we get here? Retrieved from https://www.euronews.com/2020/02/11/brexit-draft-deal-first-of-many-hur dles-to-a-smooth-exit Sharaf MF (2013) Job-education mismatch and its impact on the earnings of immigrants: Evidence from recent arrivals to Canada. ISRN Econ 2013:1–14 Spence MA (1973) Job market signalling. Q J Econ 87:355–374 Statistics Canada (1981) Top 10 countries of birth of recent immigrants, 1981. Retrieved from https://www.150.statcan.gc.ca Statistics Canada (2016) Census profile, 2016 Census. Retrieved from http://www12.statcan.gc.ca Stiglitz JE (1975) The theory of screening, education, and the distribution of income. Am Econ Rev 65(3):283–300 Tannock S (2011) Points of prejudice: Education-based discrimination in Canada’s immigration system. Antipode 43(4):1330–1356 Teelucksingh C, Galabuzi GE (2005) Working precariously: the impact of race and immigrant status on employment opportunities and outcomes in Canada. The Canadian Race Relations Foundation. Retrieved from http://www.amillionreasons.ca/WorkingPrecariously.pdf The Conference Board of Canada (2004) Performance and potential 2004–2005: key findings. How can Canada prosper in tomorrow’s world? Retrieved from http://www.conferenceboard.ca Toronto Immigrant Employment Data Initiative (February 2010) Do immigrant class and gender affect labour market outcomes for immigrants? Retrieved from http://www.yorku.ca/tiedi/doc/ AnalyticalReport2.pdf Wanner RA (2003) Entry class and the earnings attainment of immigrants to Canada, 1980–1995. Can Public Policy 29(1):53–71
The Economics of Discrimination and Affirmative Action in South Africa
20
Imraan Valodia and Arabo K. Ewinyu
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Institutionalization of Racial Discrimination in South Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Post-1994 Reforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analyzing Aspects of Employment and Empowerment Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Employment Equity Patterns Among Top Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Higher Education Institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Executive and Legislative Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discrimination Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evaluating the Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
482 483 485 486 487 487 488 491 492 492 493 494 496 496
Abstract
South Africa stands out as one of two countries whose economy has been shaped by the discrimination of the majority of the population. To redress the past, the society has grappled with the contested issue of how affirmative action policies, aimed at the majority, can shape economic outcomes. The chapter begins by summarizing the historic institutionalization of racial discrimination in South Africa, outlining some of the features of the past process that excluded I. Valodia (*) University of the Witwatersrand, Johannesburg, South Africa e-mail: [email protected] A. K. Ewinyu University of the Witwatersrand, Johannesburg, South Africa Southern Centre for Inequality Studies, University of the Witwatersrand, Johannesburg, South Africa e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_22
481
482
I. Valodia and A. K. Ewinyu
black people from active participation in the economy. Various legislations are outlined spanning the period of the 1890s to 1960s. Thereafter, an analysis of key developments in the democratic era is made. The section examines employment trends in the labor market, the broader economy, and in key institutional settings such as higher education and in the political system. The assessment reveals that while the absolute numbers and employment share of women and black people have improved, it has coincided with higher unemployment for these groups. Occupational segregation continues as previously disadvantaged individuals are increasingly employed in low-skilled and low-paying jobs. Finally, the authors construct a simple discrimination index that will compare employment or participation trends in the highlighted segments across broader population or labor force shares. The analysis indicates that transformation has occurred and firms or institutions have undertaken the necessary shifts toward reducing discrimination along gender and racial lines. These findings are assessed against existing literature and present the prevailing evidence on the effects of affirmative action policies in South Africa. Keywords
Apartheid · Institutionalized discrimination · Discrimination index · Blacks · Whites · Black economic empowerment · Affirmative action
Introduction With a few notable exceptions, such as Malaysia, South Africa stands among countries whose economies have been shaped by discrimination, not only for its long history of colonial and apartheid-based discrimination but also by the fact that, unlike other countries, the majority of the population have suffered from discriminatory practices. To redress the past, the society has grappled with the contested issue of how affirmative action policies, aimed at the majority, can shape economic outcomes. We begin by outlining the institutionalization of racial discrimination in South Africa, outlining some of the features of the historical process that excluded black people1 from active participation in the economy. We move on to assess some key developments in the post-apartheid period, examining trends in the labor market, in the broader economy, in key institutional settings such as higher education, and in the political system. We assess our findings against existing literature and present the prevailing evidence on the effects of affirmative action policies in South Africa.
1
Black people is a generic term which includes the following population groups: Africans, Coloureds, and Indians.
20
The Economics of Discrimination and Affirmative Action in South Africa
483
The Institutionalization of Racial Discrimination in South Africa South Africa’s history of racial discrimination can be traced back to the arrival of European settlers in Cape Town in 1652. So began a long and torturous process of excluding black South Africans, the majority of the population, from participation in the economy. The discovery of diamonds in Kimberley in 1867, and the subsequent discovery of vast deposits of gold in the Johannesburg area in 1886, transformed South Africa’s economy and fundamentally shaped the formal and legal systems of racial discrimination. Through a host of laws, such as the Glen Gray Act of 1894 which annexed vast sections of African-held lands into the British colony and imposed a poll tax on African land ownership to force African labor into the nascent mining industry, the systematic exclusion of blacks was institutionalized. Following the end of the Anglo-Boer War between the Afrikaners and the British in 1902, the 1913 Native Land Act was an important marker of the systematic exclusion of blacks from access to economic resources. In terms of this act, approximately 7% of the total land area of South Africa was set aside as “native reserves” to accommodate Africans, who were restricted from acquiring land outside of these areas. This disenfranchisement process is excellently covered in Colin Bundy’s (1988) seminal book, The Rise and Fall of the African Peasantry. His work shows that in the nineteenth century, the African peasantry had responded with alacrity to market opportunities in agriculture and how this process was reversed by the institutionalization of policies and legislation to exclude the local African population from the economy to promote capitalist development of mining and white agriculture in South Africa. The election of the National Party to power in 1948 marked another important turning point in discrimination policies in South Africa through the institutionalization of the apartheid system. The Population Registration Act of 1950 provided for the classification of South Africans across four official race groups: Bantu (or African), whites, Coloured (persons of mixed race), and Indians (persons of Asian origin). The Group Areas Act of 1950 designated particular neighborhoods, or group areas, to particular race groups, restricted the ownership of property in these areas along racial lines, and imposed restrictions on interracial property transactions. The Pass Laws Act of 1952 required all Africans above the age of 16 to carry a pass at all times and restricted their access to most parts of the country, except for approved employment reasons. Various pieces of legislation in the early 1950s removed all groups, except whites, from the voters’ roll, thus politically disenfranchising Africans, Coloureds, and Indians. The Prohibition of Mixed Marriages Act of 1949 and the Immorality Act of 1950 prohibited marriage and sexual relations between the different races. Further institutionalizing the provisions of the 1913 Native Land Act, the 1959 Promotion of Bantu Self-government Act created eight distinct “homelands” in line with the demarcation of the 1913 Act, where Africans were given limited self-government rights under chieftaincies, thereby depriving all Africans of their citizenship of South Africa. In later years, some of these homelands were granted farcical “independence.” Through the implementation
484
I. Valodia and A. K. Ewinyu
of the Group Areas Act, black South Africans were ruthlessly dispossessed of their homes and belongings and dumped into areas restricted for blacks. Key elements of the apartheid architecture were designed to restrict and regulate blacks and especially African peoples’ participation in the labor market. The Bantu Education Act of 1953 consolidated the already racially demarcated education system and restricted education for Africans to developing skills for menial tasks. The act was premised on the principle that the role of Africans in the economy would be restricted to providing unskilled labor and the education system should not prepare Africans for any tasks beyond this. Through the act, various mission schools which had provided education to Africans were transferred to the new Department of Bantu Education. Though the data does not reveal the full perniciousness of the system, data on expenditure by race in education is revealing as funding resulted in pupil-to-teacher ratios of 1:18 in white schools and 1:39 in black schools (Ocampo 2004). The practice whereby all semiskilled and higher categories of job were reserved for white workers and excluded all blacks was institutionalized through the Mines and Works Act of 1957. These and other legal provisions of the apartheid system were insightfully captured by Harold Wolpe (1972). His work documents how apartheid served the interest of capitalist development in South Africa and the systematic exclusion of blacks, whose role was limited to the provision of cheap labor for agriculture, mining, and later, industrial development. The homelands, underdeveloped as they were, served as important sites for the reproduction of cheap labor. Doug Hindson’s (1987) Pass Controls and the Urban African Proletariat is an important study outlining how the pass system controlled the rights of African workers and limited their role to a source of cheap labor for the burgeoning industrial economy in the urban areas. Freund and Padayachee (2021) show some of the key economic and industrial policies that led to South Africa being among the fastest-growing economies in the post-World War II period and how this growth was based on apartheid policies that excluded blacks from the benefits of economic growth. Feinstein (2005) is a more comprehensive account of the period and the forces which shaped the enormous concentration of wealth among segments of the white population and the systematic exclusion of black South Africans. Two additional points are important to stress about the patterns of discrimination under apartheid. First, as highlighted by Wolpe (1972), the homeland system and the associated migrant labor system were key features of the exploitation and exclusionary discrimination. Migrant labor has been a key feature of the South African labor history which has had the effect of suppressing the wages of black workers to low levels and excluding groups, especially women, who were deemed to be surplus to the labor requirements of capitalist development in South Africa. Francis Wilson’s (1972) famous study of the effects of migrant labor in South Africa showed that over the period 1911 to 1966, the wages of black workers in the mining industry declined in real terms, while those of their white counterparts had significantly increased. Bank et al.’s (2020) study of migrant labor in post-apartheid South Africa shows how the historical antecedent of migrant labor continues to shape contemporary labor markets.
20
The Economics of Discrimination and Affirmative Action in South Africa
485
Second, we observe how black women, largely excluded from access to the urban industrial economy, came to bear the brunt of these exclusionary processes. Except for a small proportion who were allowed into the urban areas to work as domestic workers, most African women had to subsist on the margins of the economy in the homelands. Several insightful studies document these processes of exclusion, including Bozzoli and Nkotsoe (1991) and Walker (1990). Casale et al. (2021) detail how the processes of exclusion in the apartheid period continue to disadvantage women in post-apartheid South Africa. Of course, the disenfranchisement of Africans did not go unchallenged. The 1950s was a period of widespread social unrest and protests against policies such as the Pass Law, the Group Areas Act, and other key elements of the apartheid system. This period of resistance was finally crushed with the Sharpeville massacre of 1960; the banning of the African National Congress, the Pan African Congress, and the South African Communist Party, among others; and the imprisonment of Nelson Mandela and other African leaders. A new period of resistance began in 1973 with the Durban industrial strikes and later the 1976 Soweto uprisings. Following more than two decades of resistance, South Africa finally entered a new democratic era in 1994 (see Feinstein 2005, among others).
Post-1994 Reforms As South Africa entered a new democratic era, a key issue in the transition was how the society would deal with its history of racial exclusion and dispossession of black persons. At the political level, the Truth and Reconciliation Commission allowed the society to openly deal with aspects of its past. South Africa’s new constitution, under Section 9, provides for equality for all before the law. Under this provision, “equality” is defined to include “legislative and other measures designed to protect or advance persons, or categories of persons, disadvantaged by unfair discrimination. . .” (South Africa 1996). To address historic discrimination within the labor market, the Employment Equity Act of 1998 was promulgated. This act both prohibits discrimination in employment and provides for affirmative action programs to redress the consequences of apartheid-linked discrimination policies. The act provides for “designated groups” – that is black people, women, and those with disabilities, by requiring employers to establish programs that ensure that members of the designated groups are given preference in employment and training. Rather than imposing fixed targets, the act requires employers in consultation with workers to set suitable targets for their context, in line with the national and regional profile of the economically active population in which the enterprise is based. In the broader economic sphere, a range of black economic empowerment measures were introduced, culminating in the Broad-based Black Economic Empowerment Act of 2003, which establishes a framework of measures to promote ownership and management control by blacks and skills development and enterprise and supplier development programs aimed at the advancement of black persons in
486
I. Valodia and A. K. Ewinyu
South Africa. Ponte et al. (2007) provide a comprehensive review of the evolution of Black Economic Empowerment (BEE) policies in South Africa. Early policy concentrated on measures to support the development of small business, through finance and enterprise development programs. Frustration with the lack of progress led to the current president, Cyril Ramaphosa, being appointed to lead the Black Economic Empowerment Commission in 2001, consolidating the emerging political-linked black capitalist class’s shaping of BEE. The commission’s report, published 2 years later, argued for a more active role from the state in the implementation and monitoring of BEE. The report argued that economic growth in South Africa required dealing with the high levels of concentration in South Africa, which required the spreading of ownership and control of all aspects of the economy, including the so-called commanding heights. The latter phase of BEE, captured in the 2003 Act, was characterized by a set of sector-based charters and transformed ownership in many of the key sectors of the economy: mining, finance, and banking, among others. Critical in this phase was state involvement in defining a set of scores and codes. In a range of other areas – for example, access to schooling and higher education – a range of policies have been introduced to improve access by women and black persons. Branson and Lam (2021) provide an excellent review of the challenges and performance of the education system in post-apartheid South Africa, especially with respect to improved access. Key to this was the establishment of the South African Qualifications Act of 1995 that enabled the establishment of the National Qualification Framework that created a single system covering all aspects of education and training. Given the importance of land in the historical dispossession of African people in South Africa, matters of land redistribution and restitution were important elements of post-apartheid policies. The consensus in this area is that, with a few notable exceptions, for the most part, land redistribution and restitution policies have underachieved and land ownership continues to be a vexed political issue in South Africa. Cherryl Walker’s (2008) Landmarks: Land Claims and Land Restitution in South Africa is among the leading accounts of the policies, successes, and failures in the immediate post-apartheid period. Additional debates and disagreements are well covered by Sihlobo and Kirsten (2021) and Hall and Mtero (2021).
Analyzing Aspects of Employment and Empowerment Trends Following the discussion on policy and legislative changes to address historic discrimination in post-apartheid South Africa, we now shift our focus to analyze data to establish whether the legislative shifts have abolished or reduced discrimination. To test this, we will look at employment and earnings within the labor market. Next, we shall look at employment trends at specific institutions, namely: parliament, top management/board composition, and employment within higher education institutions. As is applicable for each of these, we will observe emerging trends by gender and across racial groups. We also include a discussion on the
20
The Economics of Discrimination and Affirmative Action in South Africa
487
prevalence of disability and characteristics of people with disabilities. Finally, we will construct a simple discrimination index that will compare employment or participation trends in the highlighted segments across broader population or labor force shares. Our expectation is that we would see similar employment and representation patterns at these institutions as are available in the broader population or labor force and that any deviations from this would be an indication of discrimination.
Employment Since 1995, the female share of the working age population (individuals aged 15–64) increased (Casale and Posel 2002). However, the share of male workers has been consistently rising, and the higher gender gap between the two groups has narrowed over time from 4.6% in 1995 to 0.9% in 2021. Over this period as well, there was an increase in the absolute number of people that were working or were available to work. The additional entrants into the labor force were overwhelmingly more likely to be male and black. Employment trends support this observation as fewer female and African workers were absorbed into the labor market (Fig. 1). Consequently, high and rising unemployment rates are observed for this group of workers. This is the case whether unemployment is defined narrowly or broadly2 as growth in the economy was too low to sufficiently absorb these additional workers (Oosthuizen and Bhorat 2005; Tregenna et al. 2021). It is worth noting that more women than men are classified as discouraged workers over this period. Employment shifts in post-apartheid South Africa indicate a bias toward higherskilled workers and shrinking demand for semiskilled and unskilled workers. These higher-skilled workers typically earn more and are further characterized as being mostly male and white. Occupational segregation means that African and female workers are overrepresented in lower-paying and low-skilled segments of the labor market (Tregenna et al. 2021).
Wages Female workers have consistently earned less than males (Table 1). This is the case at both the mean and median. Over time, the wage differential has fluctuated, but the overall trend has been toward a reduction as the gender pay gap has continuously
2
The narrow definition of unemployment, which is the official unemployment rate, is calculated by expressing the share of unemployed individuals as a proportion of total employed workers. Broad or expanded unemployment also includes the share of “discouraged” workers and will therefore be higher than the official measure.
488
I. Valodia and A. K. Ewinyu
80% 70% 60% 50% 40% 30% 20% 10% 0%
1995
2000
2005
Female
Male
African
2010
2015
2019
Coloured
Indian
White
Fig. 1 Employment trends by race and population group, 1995–2019. (Source: Own calculations from PALMS V3.3)
approached one. In 1995, women earned R77 for every R100 earned by male workers; by 2019, this had increased to R87. The median wage for African and Coloured workers was lower than the overall wage earned by all workers. White and Indian/Asian workers earned 2–3 times these overall wages, with the former group of workers earning the highest of all workers across the period under review. White workers also experienced the most growth in their wages (0.11% annually) compared to Africans who experienced a contraction in their wages estimated at 0.12% annually.
Employment Equity Patterns Among Top Management The Commission for Employment Equity (CEE) publishes an annual report on the status of employment equity in the previous fiscal year. Designated employees3 consult with their employees, analyze their workplace profiles, and submit these detailed data to the Department of Employment and Labour annually or biennially. This firm-level data is then collated by the commission that publishes aggregate trends across various metrics. The CEE reports on the top four occupational levels where black workers are most underrepresented. In this section of the report, we discuss only the top management. This is a senior level of workers at the organization that report directly to the chief executive officer and/or the board. They also oversee the daily
3
This refers to an employer who employs over 50 employees or one that has a stipulated annual turnover. Municipalities and various organs of state are also designated employees. Employers can also volunteer to become designated employers.
2005 2861 3,540 2,044 0.58 2,453 3,540 7,153 12,263 0.86 1.24 2.50 4.29
2000 3,045 3,836 2,082 0.54 2,342 3,123 6,576 11,711 0.77 1.03 2.16 3.85
0.73 0.95 2.35 3.33
2010 4,392 4,822 3,291 0.68 3,188 4,154 10,302 14,639 0.93 0.95 2.06 3.55
2015 3,581 4,263 3,001 0.70 3,348 3,386 7,386 12,707 0.92 0.97 1.95 3.42
2019 3,800 4,000 3,466 0.87 3,500 3,683 7,400 13,000
CAGR (%) 0.55 0.66 0.17 0.12 0.08 0.10 0.11
1995–2019 Absolute value 533 694 145 107 76 183 362
Source: Own calculations from PALMS V3.3., 2019 wage values are from StatsSA Labour Market Dynamics 2018 and 2019 Notes: 1. These are nominal wages
Median 1995 Overall 4,333 Male 4,694 Female 3,611 Gender pay gap (F/M) 0.77 African 3,607 Coloured 3,607 Indian/Asian 7,583 White 12,638 By race, share of overall median wage African 0.83 Coloured 0.83 Indian/Asian 1.75 White 2.92
Table 1 Median wages by gender and population group, 1995–2019
20 The Economics of Discrimination and Affirmative Action in South Africa 489
490
I. Valodia and A. K. Ewinyu
100%
90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2000
2002
2004
2006
African
2008
Coloured
2010
2012
Indian
2014
White
2015
2016
2018
2019
2020
Foreign Naonal
Fig. 2 Distribution of top management – by population group, 2000–2020. (Source: Commission for Employment Equity (CEE) Annual Report, 2000, 2010, and 2020–2021)
management of the firm, have some financial oversight or contribute to overall strategy. Some examples of these workers include chief operating officers, chief financial officers, and in some cases heads of sales and marketing. In the two decades since 2000, top management has remained predominantly white accounting for over two-thirds of workers at this occupational level (Commission of Employment Equipment, 2021) (Fig. 2). In keeping with the declining employment share of all white workers shown in Fig. 1 above, we note that the employment share of this category of workers has in fact declined by over 20 percentage points with the greatest decline being experienced in the 2000–2010 period. By contrast, African and Indian workers have increased their employment share at this occupational level by a total of 10 and 7 percentage points, respectively. Of all the previously disadvantaged groups, these two have seen the largest shifts in employment at this occupational level, relative to Coloured workers whose employment share has increased by 3 percentage points to peak at 6% in 2021. Prior to 2006, the commission did not collect and report on data pertaining to foreign nationals. Since then, foreign nationals have consistently accounted for less than 3.5% of total top management. By gender, we note that employment at this occupation level is overwhelmingly male – three in four workers are male. However, between 2010 and 2020, the share of male workers has declined as more and more female workers are hired at this level. However, these prevailing trends differ once we intersect race and gender and consider the characteristics of the employer. These more descriptive data are available from 2012 until 2020 – the discussion in this section will reference that period onward. For example, there are almost five times more foreign males at top management level than there are females. Also, when we compare employment across all
20
The Economics of Discrimination and Affirmative Action in South Africa
491
levels of government and parastatals, we observe that racial representation for this category of employer is representative of the broader population structure. The converse is true within the private sector, where white and Indian workers continue to be overrepresented. Finally, at educational institutions, we observe that the share of employment of white workers grew driven by the increase in white female workers. We pick up on this aspect in the following discussion that analyses employment trends at higher education institutions.
Higher Education Institutions In 2000, there were 21 public higher educational institutions in South Africa that employed 33,300 individuals. By 2019, there were 26 public universities with 64,100 staff. The Department of Higher Education and Training publishes data for permanent staff 4 across three major categories: instruction and research (academic), administration, and service staff. In addition to the permanent staff complement, universities also employ workers on fixed or short-term contracts. Due to data constraints, we exclude this latter category of workers from our analysis although their inclusion would certainly have enriched the discussion. Finally, before 2020, the department did not provide detailed data across each of the four main population groups; instead they provided the share of black workers which we take to include all three non-white racial groups. In terms of gender, the data show that slight transformation has occurred in the last two decades as female workers have consistently increased their employment share among all categories (Fig. 3). In fact, in 2019, there were 5,400 more female workers than male individuals. This expansion in employment has occurred among academic and administrative staff, with the greatest gains being observed in the latter category. Gender trends among the service staff are less obvious and fluctuated across this period. Regarding race, we note that between 2000 and 2019 the share of black workers employed at universities has increased. This is the case in absolute numbers and by employment share. However, for black workers, this expansion has mostly been in the services segment that was almost 100% black in 2019 – a growth of 30 percentage points since 2000. The employment share of all black employees has also consistently increased among academic and administrative staff, with the larger growth being observed in the latter category of workers. From their 2010 report, the CEE highlighted that educational institutions have an overrepresentation of white male and female workers at the top and senior management occupational levels. Statistics on workforce movements show that this subgroup of workers was most likely to be recruited, skilled, and promoted than any other group.
4
These are classified as individuals that contribute to an institutional pension or retirement fund.
492
I. Valodia and A. K. Ewinyu 70%
100% 90%
60%
80% 50%
70% 60%
40%
50% 30%
40% 30%
20%
20%
2000
2005
2010
2011
2012
Instruction and Research Staff
2013
2014
Academic Staff
2015
2016
2017
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
0%
Female
2017
Female
2015
Female
2014 Service Staff
Female
2013
Academic Staff
Female
2012
Female
2011
Female
2010
Instruction and Research Staff
Female
2005
Female
2000
Male
All White All White All White All White All White All White All White All White All White Black Black Black Black Black Black Black Black Black
Female
0%
Female
10%
10%
2019
Service Staff
Fig. 3 Permanent staff in public universities by population group and gender, 2000–2019. (Source: Department of Higher Education and Training 2010 and 2019). Notes: 1. Administration staff: all staff that spend more than 50% of their official time on administrative functions. 2. Instruction and research/academic staff: these are individuals appointed to teach or undertake research at a public higher education institution. 3. Service staff: all staff who are not engaged in supervisory or administrative functions that are undertaken at an office. 4. Not all data provided in 2000 disaggregated by population group and gender across all three categories of employment. We therefore assigned the same share of employment to academic and administrative staff
Executive and Legislative Representation South Africa has a 5-year electoral cycle. Parliament through elected officials exists to represent the electorate and to hold government accountable to deliver on promises and other undertakings made to the public. At the national level, the African National Congress (ANC) and the Economic Freedom Fighters (EFF) are the only two parties that practice voluntary party quotas. The ruling party, the ANC, first instituted a 30% quota in 1994, leading up to the first democratic election. In 2006, the ANC adopted a 50% gender quota that was later extended to national elections in 2009 (Myakayaka-Manzini 2003; ANC Constitution 2017). In the 2019 election, both the EFF and the ANC attained gender parity of 50 and 49%, respectively. However, gender representation of both parties’ top five officials fell woefully short of gender parity as they each had only one female official. These voluntary party quotas have contributed significantly to advancing the representation of women at parliament from 25% of the legislature being female in 1994 to half of them in 2019 (Fig. 4). The number of full ministers and deputy ministers in cabinet has also increased in democratic South Africa.
Disability The 2011 census provided statistical evidence on the prevalence of disability and the characteristics of persons with disabilities. However, the questions posed were dissimilar to those asked in the 1996 and 2001 census as well as the Community Survey of 2007, which constrains the overall comparability of these findings over time. Further limitations also arise because certain data on psychosocial and neurological disabilities were excluded. Thus, while limited, the findings are somewhat
20
The Economics of Discrimination and Affirmative Action in South Africa
493
60% Women in Ministerial Positions (%)
Women in Parliament (%)
50%
49.0… 41.40%
40%
40.00%
41.70%
38.10%
41.50%
42.30% 32.80%
30%
29.80% 23.08%
20%
46.40%
1995
25.00% 2000
2005
2010
2015
2020
Fig. 4 Parliamentary and ministerial representation by gender, 1995 - 2020. Source: Parline IPU 2010, 2017, 2021, Myakayaka-Manzini (2003)
informative and indicative of the existence of limitations on the full participation of individuals with disabilities in the South African economy and the society at large. Subject to the limitations discussed above, findings from the 2011 census reveal that among individuals 5 years and older, the national disability prevalence rate was 7.5%. Noticeable sex differences are observed as disability was more predominant among females than it was in males.5 African and Coloured individuals had the highest rates of disability. Individuals in rural areas reported higher disability rates compared to those residing in urban areas. In terms of income and educational attainment, people with disabilities had lower income and comparably worse educational outcomes.
Discrimination Index Our discussion so far has highlighted the fact that in absolute terms, the number of women and other previously disadvantaged population groups have increased their overall share. This increase together with lower absorption rates has resulted in higher unemployment among this category of workers. Furthermore, we have seen that these shifts have not occurred uniformly across all institutions. Racial and gender occupational segregation is still at play within the higher educational institutions where research and instructional roles are predominantly white and male compared to service staff who are mostly female and black. Relative
5
Among individuals of working age, those aged between 15 and 64, the overall disability rate is 5.5%. Again, we observe similar differences by sex as disability was more prevalent among females than males.
494
I. Valodia and A. K. Ewinyu
Table 2 Discrimination index for selected employment trends, 1995–2020 1995 Female 0.83 African 0.90 Employment – top management Female African Coloured Female academics Female administrative staff Black academics Black administrative staff Executive and legislative representation Female ministers 0.46 Females in parliament 0.49
2000 0.96 0.91
2005 0.93 0.91
2010 0.88 0.90
2015 0.88 0.93
2019/2020 0.85 0.92
0.28 0.08 0.31 0.84 0.84 1.28 1.28
0.39 0.15 0.44 0.88 1.26 1.22 0.84
0.39 0.16 0.49 0.90 1.27 0.90 0.78
0.43 0.18 0.51 0.93 1.24 1.10 0.79
0.48 0.19 0.63 0.95 1.01 1.01 0.73
0.75 0.59
0.82 0.65
0.79 0.83
0.82 0.82
0.97 0.92
to all levels of government, we observe that the private sector has remained mostly white and male. To make sense of these shifts and analyze whether there have been any significant gains toward reducing discrimination, we now construct a discrimination index. This is a simple analysis that will compare employment trends at these institutions to the corresponding aggregate employment shares for women and black individuals – subgroups that have historically faced discrimination. Cabinet and legislature representation will be compared to the population share of females. This is summarized in Table 2 where a ratio of less than 1 shows that there is variation from the selected metric and that there is continued discrimination faced by that group. Our analysis indicates that transformation has occurred, and firms or institutions have undertaken the necessary shifts toward reducing discrimination along gender and racial lines. At parliament and cabinet, we see how voluntary party quotas have resulted in greater female representation, and we can reasonably expect that gender parity will be attained at the next national election. Female employment across the entire labor force gradually increased hitting a peak around 2005 and declining thereafter. At top management and within higher educational intuitions, we see that the female share of employment is continuing to increase as the combined effects of national policy and legislation and firm-wide policies come into effect.
Evaluating the Evidence South Africa’s history of discrimination and dispossession has a long history and is especially complex. It is hardly surprising then that the post-apartheid policies enacted in the post-1994 period remain a contentious and politically charged set of issues. Unlike most other jurisdictions with laws to promote affirmative action policies aimed at promoting discrimination against minority groups, in the
20
The Economics of Discrimination and Affirmative Action in South Africa
495
South African case, it is the historical exclusion of the majority that needs to be redressed. In this section, we provide a broad-ranging discussion of the literature and evidence to date and outline some of the complexities for further investigation. Hwok-Aun Lee’s (2022) detailed study of affirmative action policies in Malaysia and South Africa is by far the most careful and considered evaluation of affirmative action policies in post-apartheid South Africa. Consistent with some of the data we present above, notwithstanding high dropout rates for black students, much has been achieved across the education system, including higher education, through policies to promote access and redress historic inequities. On the employment side, trends in the labor market parallel those in education, with the proportion of tertiary qualified workers increasing among all race groups. However, white workers, notwithstanding the fall in their share of graduations from the higher education system, continue to be the most highly educated and continue to dominate, proportionally, the higher echelons of the labor market. From a gender perspective, Espi et al. (2019) show that, notwithstanding women’s progress in education and progress especially for white women, in general, women continue to be higher skilled and work in management occupations in South Africa. They find that substantial gender pay gaps continue to exist, as we have outlined above. Francis and Valodia (2021) delve into the trends in the labor market over the 30 years since the Employment Equity Act concluding that it has shaped patterns of employment and attempted to address historical discrimination. They conclude that notwithstanding the objectives of the act, occupation and earning segmentation on the basis of race and gender continue to exist. Of course, one of the key reasons for this is that the labor market has performed poorly over the period with a sustained increase in levels of unemployment, with the latest data showing levels above 35% and 45%, respectively, using the strict and expanded definitions of unemployment (Statistics South Africa, 2022 – data for the fourth quarter of 2021). Macroeconomic conditions have also not been conducive to employment growth. This cost has of course been borne primarily by black South Africans and women – thus, in some respects, South Africa may have gone backward in its attempts to redress the historical discrimination in its labor market (Ponte et al. 2007). In terms of the broader ownership structure of the economy, while the structure of corporate ownership has changed dramatically in the post-apartheid period, the ownership by black groups has not increased significantly. Mondliwa and Roberts (2021) show that in terms of control of capitalization on the Johannesburg Stock Exchange, ownership by black groups has remained in single-digit figures and declined for much of the second phase of BEE. This they argue is because BEE has largely been about sharing of rents, rather than changing patterns of ownership (see also Ponte et al. 2007; Vilakazi and Bosiu 2021; and Vilakazi 2021). Several recent studies have measured inequality in income and wealth in South Africa over the post-apartheid period. Chatterjee et al. (2021) deal, among others, directly with the question of how income and wealth shares in South Africa have changed in the democratic era. They show a significant decline in the average income gap between black and white household income, from 7 to 4 times over the
496
I. Valodia and A. K. Ewinyu
period of approximately 30 years comparing 1993–1994 and 2015–2019. However, this decline is almost entirely explained by the income progression of a small group of black income earners in the top 10% of the income distribution. Excluding this group, the income gap increases from 10.5 to 11.5. These broad trends hold also for the gap in wealth between race groups. This finding is in line with many similar studies, which find growing levels of inequality in South Africa, but a small group of black households, linked to the political elite, have been able to capture a significant share of income and wealth. Studies by, among others, Orthofer (2016), Leibbrandt et al. (2010), and Chatterjee et al. (2022) confirm this general pattern of income and wealth distribution in post-apartheid South Africa.
Conclusion South Africa’s economy has been shaped by over 300 years of discrimination based on race and gender, among others. These historical processes of discrimination have given rise to an economy that has the highest levels of inequality in the world (at least for counties that have such data). In the post-apartheid period, the country has had to deal with the controversial matter of how to address this history of discrimination and exclusion. An ambitious set of policies in the labor market and more widely in the economy have attempted to reshape the patterns of employment and earnings and the ownership of assets and corporations. From the data it is clear that a small, politically connected elite have been able to increase their share of income and wealth. However, the overall patterns of income and wealth distribution have not been reshaped in the post-apartheid period. What might be some of the reasons why this is so? It may well be the case that 30 years is too short a period to fundamentally change patterns of distribution and reshape economic and social forces that have shaped the country’s economy for over 300 years. Furthermore, South Africa has attempted to redress its past during a time when the tide of global economic forces has been unfavorable. Across the globe, patterns of inequality have increased, notwithstanding the fact that in many countries the economic lives of the poorest have improved. These global forces shaping growth process since the 1970s have also impacted South Africa. Moreover, in the South African case, its dismal performance in the labor market, where structural unemployment has grown significantly is a key force shaping how policies aimed at addressing discrimination have had muted outcomes.
References African National Congress (ANC) Constitution, as amended and adopted at the 54th National Conference, 2017. Available via https://www.anc1912.org.za/wp-content/uploads/2021/01/ ANC-Constitution-2017.pdf. Accessed 4 March 2022 Bank L, Posel D, Wilson F (2020) Migrant labour after apartheid: the inside story. Cape Town: HSRC Press
20
The Economics of Discrimination and Affirmative Action in South Africa
497
Bozzoli B, Nkotsoe M (1991) Women of Phokeng: Consciousness Life Strategy and Migrancy in South Africa, 1900-1983. London/ Portsmouth, NH: Heinemann/ James Currey Branson, N, Lam D (2021) The Economics of Education in South Africa. Black economic empowerment in South Africa. In: Oqubay A, Tregenna F,Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Bundy C (1988) The rise and fall of the African peasantry. David Philip, Cape Town Casale D, Posel D (2002) The continued feminisation of the labour force in South Africa: an analysis of recent data and trends. South African J Econ 70(1):156–184 Casale D, Posel D, Mosomi J (2021) Women and work in South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Chatterjee A, Cjaska L, Gethin A (2021) Can redistribution keep up with inequality? Evidence from South Africa. World inequality lab working paper 2021/20, World Inequality Lab Chatterjee A, Cjaska L, Gethin A (2022) Wealth inequality in South Africa. World Bank Econ Rev 36(1):19–36 (2021) Commission of Employment Equity. In: 21st Commission for Employment Equity (CEE) annual report. Department of Labour, Pretoria Espi G, Francis D, Valodia I (2019) Gender inequality in the South African labour market: insights the employment equity act data. Agenda 33(4):44–61 Feinstein CH (2005) An economic history of South Africa: conquest, discrimination and development. Cambridge University Press, Cambridge Francis D, & Valodia I (2021) Black economic empowerment: a review of the literature. southern centre for inequality studies, University of Witwatersrand Freund W, Padayachee V (2021) The economic history of South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Hall R, Mtero F (2021) Land and agrarian development in South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Hindson D (1987) Pass controls and the urban African proletariat in South Africa. Ravan Press, Johannesburg Kerr A, Lam D, Wittenberg M (2019) Post-apartheid labour market series 1993–2019 (Version 3.3) [Dataset]. DataFirst [producer and distributor] Lee H (2022) Affirmative action in Malaysia and South Africa. ) London: Routledge Leibbrandt M Woolard I, Finn A, Argent J (2010) Trends in south African income distribution and poverty since the fall of apartheid. OECD social, employment and migration working paper no 101, OECD Mondliwa P, Roberts S (2021) Corporate structure, industrial development and structural change in South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Myakayaka-Manzini M (2003, November) Political party quotas in South Africa. In International Institute for Democracy and Electoral Assistance (IDEA)/Electoral Institute of Southern Africa (EISA)/Southern African Development Community (SADC) Parliamentary Forum Conference, “The Implementation of Quotas: African Experiences,” Pretoria, November (pp 11–12) Ocampo ML (2004) A brief history of educational inequality from apartheid to the present. Retrieved from: https://web.stanford.edu/~jbaugh/saw/Lizet_Education_Inequity.html Oosthuizen M, Bhorat H (2005) The post-apartheid South African labour market. University of Cape Town, Cape Town Orthofer A (2016) Wealth inequality in South Africa: evidence from survey and tax data. ReDI 3x3 working paper, University of Stellenbosch Parline IPU (2020) Monthly ranking of women in national parliaments. Available via https://data. ipu.org/women-ranking?month¼5&year¼2022. Accessed 28 Feb 2022 Ponte S, Roberts S, van Sittert L (2007) “Black economic empowerment”, business and the state in South Africa. Dev Chang 38(5):933–955. https://doi.org/10.1111/j.1467-7660.2007.00440.x
498
I. Valodia and A. K. Ewinyu
Sihlobo W, Kirsten J (2021) Agriculture in South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford South Africa (1996) Constitution of the Republic of South Africa, 1996. As adopted. Available here: https://www.justice.gov.za/legislation/constitution/saconstitution-web-eng.pdf. Accessed 28 February 2022 Tregenna F, Ewinyu AK, Oqubay A, Valodia I (2021) Challenges and complexities of the south African economy. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Vilakazi T (2021) Black economic empowerment in South Africa. In: Oqubay A, Tregenna F, Valodia I (eds) The Oxford handbook of the South African economy. Oxford University Press, Oxford Vilakazi T, Bosiu T (2021) Black economic empowerment, barriers to entry and economic transformation in South Africa. In: Andreoni A, Mondliwa P, Roberts S, Tregenna F (eds) Structural transformation in South Africa: the challenges of inclusive industrial development in a middle-income country. Oxford University Press, Oxford Walker C (1990) Women and gender in South Africa to 1945, Cape Town/London: David Philip/ James Currey Walker C (2008) Landmarked: land claims and land restitution in South Africa. Ohio University Press, Athens, Ohio Wilson F (1972) Migrant labour in South Africa. South African Council of Churches and SproCass, Johannesburg Wolpe H (1972) Capitalism and cheap labour-power in South Africa: from segregation to apartheid. Econ Soc 1(4):425–456
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
21
Timothy Amos
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Prewar Outcaste Emancipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Postwar Buraku Liberation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
500 501 507 513 514
Abstract
This chapter offers a general background discussion to the Buraku problem in modern Japan, focusing on the politics of financial redress and the ways in which state funding targeting the Buraku problem affected communities and understandings of the core issues. Despite a growing Anglophone scholarship on Burakumin in recent years, relatively little has been written about the modern politics of financial redress and the ways that worked to help alleviate problems facing Burakumin as well as enable the constitution of Buraku identity and shape understandings of pathways to liberation. Japanese state funding and government services particularly in the postwar period played a critical role in reducing inequality between Buraku and mainstream communities and lessening discrimination toward them. State funding of Buraku causes and strong controls exerted by the Buraku Liberation League (BLL) over distribution not only secured important material advances for targeted communities but also hastened the emergence of a new form of identification among Burakumin that emphasized their position as a minority group with separate interests requiring recognition, protection, and opportunities for civic engagement. In an important sense, the now dominant conceptualization of Burakumin as a fully fledged minority group arose in force only after the commencement of state funding and organizational T. Amos (*) School of Languages and Cultures, University of Sydney, Sydney, Australia e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_24
499
500
T. Amos
implementation of reforms which altered the very nature of the postwar debate about Burakumin, although such a view needs to be carefully distinguished from those put forward by recent historical revisionists. Keywords
Burakumin · Minorities · Japan · Affirmative action · Discrimination · Caste
Introduction Every society has “outcasts,” the socially and economically marginalized, politically disenfranchised and oppressed, and culturally stigmatized. However, not every society has “outcastes,” those who have experienced historic, systematic forms of sustained social exclusion, or the lingering effects of such segregation, rooted in complex factors such as a universalizing division based on birth or systems of endogamy (Habib 2002). Japan, the country at the heart of this study, has numerous communities that potentially qualify for the label “outcast,” whether the day laborer communities, indigenous Ainu, Zainichi Koreans, or Nikkei-Brazilian communities (Sugimoto 2014). Burakumin, however, comfortably fit the definition of “outcaste” offered by experts writing on the subcontinental context, and they are frequently compared to India’s Dalits (Amos 2020). Burakumin commonly claim descent from premodern outcaste communities. Yet while clearly a group physically linked through blood ties and geography to older outcaste communities, a distinctive Burakumin identity nonetheless really only emerged in the modern era, primarily through the experience of, and organized joint resistance to, state and societal discrimination. Burakumin identity has further been shaped by the material advances made in communities through local efforts and state funding and the promises and problems associated with those labors. Like many other forms of identity formation, notions of what it means to be a Burakumin and to face discrimination are both strong and tenuous, shaped by trends, movements, and practices of contestation linked to both internal and external forces. The modern outcaste emancipation movement is marked by various competing ideological claims about the nature of the Buraku problem and how best to envisage a path forward to liberation. Despite a growing Anglophone scholarship on Burakumin in recent years (Neary 2010; Amos 2011; Hankins 2014; Bondy 2015; Amos 2020), and with the exception of a recent work by Neary (2022), remarkably little has been written about the modern politics of financial redress and the ways that has worked to help constitute Buraku identity and understandings of pathways to liberation, particularly in the postwar period. Japanese state funding and services played a critical role in reducing inequality among and lessening discrimination toward Burakumin in the postwar period; government funding coupled with a strong degree of local autonomy in determining priorities and allocations substantially weakened the structural underpinnings of the historic traces of status-based discrimination in Japan as well as
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
501
community-based discriminatory consciousness toward Buraku communities. Yet even as structural equality and a basic respect for the human rights of all Japanese subjects began to take root in postwar Japanese society through socioeconomic reform and school- and community-based human rights education programs, discrimination against Burakumin was not eliminated. With state funding of Buraku causes, and strong controls exerted by Japan’s largest Buraku liberation organization – the Buraku Liberation League (BLL) – over distribution, there emerged over time a new form of identification among Burakumin that emphasized their position as a minority group with separate interests requiring recognition, protection, and opportunities for civic engagement. In an important sense, the now dominant conceptualization of Burakumin as a fully fledged minority group arose in force only after the commencement of state funding altered the very nature of the postwar debate about Burakumin and ultimately the state’s approach to addressing the problem, although such a view needs to be carefully distinguished from those held by recent historical revisionists (Amos et al. 2021). Such a shift in perspective, from people affected by a historic system of status which wrongly excluded them to a group that was wrongly discriminated against despite their distinct difference as Burakumin, was reinforced and perpetuated after a serious split in the postwar Buraku liberation movement led to the emergence of one clear leader among the liberation organizations to oversee distribution, the effects of which can still be felt today. This chapter offers a general background discussion to the Buraku problem in both the pre- and postwar periods in Japan but is layered with discussions of the politics of financial redress and local developments in some Buraku communities scattered throughout the main Japanese island of Honshū. It concludes with a brief discussion of the significance of this history in light of more recent trends in the Buraku liberation movement post-2002. Through such a discussion, it is hoped that the reader will gain a useful general overview of the “Buraku problem” (Buraku mondai) in Japan with a particular focus on the economics of redress and a better understanding of the issues facing groups of self-identifying Burakumin at the start of the third decade of the twenty-first century.
Prewar Outcaste Emancipation Given the remarkable public invisibility of discussions about the lingering effects of “caste” in modern Japan, one might be excused for thinking that the history surrounding Burakumin has largely been settled (Amos 2020; while “status” (mibun) is primarily used to refer to Japan’s historic system of social stratification, it can basically be considered a type of caste system for comparative purposes). After all, Japan’s modernization drive began more than 150 years, and it was widely recognized even at the time that the existence of outcaste communities on the archipelago was a major stumbling block for the country’s development as a modern nation state. Groups of outcastes, commonly referred to pejoratively in the Tokugawa period (1600–1868) as eta (literally, “greatly polluted”) and hinin (literally,
502
T. Amos
“not human”), along with other lesser-known groups, were elevated to “new commoner” (shinheimin) status with an 1871 statute that scholars have labeled in various ways: the “Emancipation Edict,” the “Edict Abolishing Outcast Statuses,” and so on (Watanabe 1961; Uesugi 1990; Osatake 1999). Regarding the estimated size of the population this law purportedly targeted, government statistics prepared in relation to the 1872 household registration system listed the total number at 382,886 (280,311 eta, 23,480 hinin, and 79,095 “leatherworkers and assorted others”) or a little under 1 percent of the total population (Motohama 2009). As many commentators have noted, the so-called Emancipation Edict came with little or no supporting legislation, and it was enforced to varying degrees and with differing effects across the country. The net result was that in most areas where “former outcaste” (moto eta, moto hinin, etc.) communities were found, discrimination against them continued, even becoming actively perpetuated by some local governments and communities. While individuals or smaller communities could and did successfully adapt their lives to mainstream commoner societal expectations in certain localities, this did not essentially remove the root cause of discrimination. Neither was it a realistic strategy for many communities to achieve, especially ones subjected to sustained forms of severe discrimination in the early modern era. Notwithstanding, “former outcaste” communities did sometimes attempt to move out of the traditional occupations and industries that previously sustained them or work toward social mobility through capital accumulation and educational advancement, trying to blend into society through the adoption of “civilized practices.” Such efforts were often spearheaded by wealthier members of the community – those in leadership positions who had in some cases inherited power due to earlier status positions (Amos 2015a). Despite these efforts, “new commoner” communities, as they soon came to be called, as well as their resident populations, continued to face stigmatization. Indeed, it was not long before the central and local governments identified such communities as potential hotbeds of contagion, criminality, and civil disruption. Racialization of such communities had earlier precedents, but Social Darwinism gave new life to older theories which speculated that former outcaste communities were racially distinct (Fujino 1994; Amos 2017a). “Former outcastes” came to be treated as “tribally” (zoku) distinct in many quarters, a sentiment reinforced by growing nationalism, expanding imperialist state ambitions, and capitalist growth that exacerbated class distinctions. As numerous authors have pointed out, the Chinese ideographs “tokushu buraku” also came into use, emphasizing the ambiguity of “outcaste community” racial backgrounds (Amos 2011; McCormack 2013). One of the most famous areas associated with Buraku history in Japan is a residential section of the area of Asakusa “Newtown” in Tokyo. Formerly a residential area of Tokyo housing a 300-household-strong community exclusively made up of outcastes under the leadership of Danzaemon in the early modern era, “Newtown” came to be referred to as Kameoka in the early Meiji period, before becoming integrated into the larger Imado area in the twentieth century (Amos 2011, 2015a, 2020). Early Meiji government policies certainly worked to help “normalize” this area, and the community’s leadership stratum also achieved a degree of success
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
503
in their commercial dealings, assisting in blurring the lines between “former outcastes” and “commoners” (Amos 2020). Such entrepreneurial endeavors also produced positive flow-on effects for members of the wider community, although they were unevenly experienced, sometimes even serving to alienate residents who came to wrestle with a compounded form of stigmatization rooted in both older and more modern discriminatory logics (Amos 2015b). Further west in Osaka, a community integrated into Yosami Village in the late nineteenth century, referred to in the early modern period as Sugimoto Shinden and today as Asaka, was one of the numerous “former outcaste” areas in the city that wrestled with Japan’s rapid modern transformation. The origins of the village are unclear, but apparently they date back to the early eighteenth century when a series of public works rendered a particular section along the Yamato River unusable and members of neighboring outcaste communities in the region requested permission to open up new rice lands in the area. The community appears to have been comprised of residents that largely made their living from tenant farming, production of leathersoled sandals, goods transport, and cattle trading (Maki 1979; Satogami 2003). It is not difficult to imagine how emerging conceptions of modern civilized life modeled on perceived Western norms would work to further compound the effects of discrimination for such a community. In the neighboring prefecture of Hyōgo, the village of Furonodani also became inextricably linked to the development of the city of Kōbe as a treaty port, with modern facilities such as prisons, hospitals, and workhouses also coming to be located near and associated with the community in the Meiji period. This is, as Daniel Botsman intimates, a formative development clearly connected to the community’s earlier experiences of an “older geography of social inequality and hierarchy” (Botsman 2016). Official modern investigations into the demographics of “poor Buraku [communities]” (saimin buraku) can only really be dated at a national level to the early 1900s, arising within the context of a society struggling with social and economic problems in the aftermath of the Russo-Japanese War. Before that, the primary interest in “former outcastes” tended to be expressed in newspapers, scandal sheets, literary works, slum reportage, and the occasional legal case (Amos 2011; McCormack 2013). In the 1907 survey conducted by the Home Ministry, “Special Buraku” (tokushu buraku) areas were numbered at 4324 (the conflicting number of 5470 was also given by one Ministry official) with a total population of 779,434 people (Motohama 2009). “Special Buraku” were clearly targets of suspicion at this time, particularly as some questionable links could be drawn between one Buraku area in Wakayama prefecture and the alleged perpetrators of prewar Japan’s most infamous incident of lese majesty: the 1910 plot to kill the Meiji Emperor (Shields 2017). Former outcaste areas, however, were not alone as targets of official angst and popular suspicion – modern slums, often not connected to these communities, also generated similar kinds of controversy and interest. Nonetheless, older “former outcaste” communities were frequently targeted for local improvements to “harmonize” (Yūwa) their position within mainstream Japanese society. The official hope was that social conciliation with these communities would be possible if efforts were made to eliminate discrimination against them originating from inherited social
504
T. Amos
structures, deleterious customs and practices, and various environmental issues. Some local governments, such as Osaka municipality, set up local offices to deal with problems pertaining to “Special Buraku” communities (Totten and Wagatsuma 1966). Government officials, industrial stalwarts, and religious and community leaders tended to push the hardest for harmony in the early twentieth century. Members of Buraku communities also contributed to these efforts both individually and at an organizational level, through groups such as the Yamato dōshikai (Yamato Brotherhood Society) formed in Nara, which attempted to unify the nation through humanistic efforts and strategies. Actions were also taken at the national level to unify local movements and encourage social harmonization, most notably with the formation of the Teikoku kōdōkai (Imperial Way Society) in 1914, an organization formed to allay the possible negative ramifications for Japan if radical thought permeated deeply through these communities (Neary 1989; Bayliss 2013). Over time, however, the earnestness for self-reform within Buraku communities tended to be overrun by skepticism. Younger community members were increasingly unsure of the reliability of a story that framed the reasons for their continued discrimination in internal factors. They were also less confident that local attempts at “social industry” (an apparent euphemism in these areas for self-help improvement measures) were simply benevolent attempts by elites to help the disadvantaged help themselves. Inspired by radical thought rapidly making its way into Japanese translation and onto factory floors and into dormitory rooms and study groups, members of Buraku communities began to seek an answer to their problems and ill-treatment through political and social mobilization. They increasingly chose to see oppressive social, political, and ultimately economic structures coupled with popular ignorance as the decisive determinants of their continuing daily struggles. By the late 1910s, a considerable number of people, galvanized by their dislike of an increasingly purist image of nation and society embodied in a “national essence” (kokutai), chose the path of collective identification and protest. Many “Special Buraku people,” as they increasingly came to be referred to by mainstream society, began to unite around a shared history of oppression. Terms such as “brethren” and “comrades” were used to link those who identified with these areas, greatly exacerbating the fears of officials and elite stakeholders who were also closely monitoring Russia and the international labor movement (Kubota 1979; Suzuki 1985). The grassroots Suiheisha (“National Leveler’s Association”) was formed in 1922, and branches of this liberation organization quickly sprang up around the archipelago. This movement aimed to “organize a new group movement” through which emancipation would be realized through the promotion of “respect for human dignity” (Neary 1989). One of the distinctive struggles this organization began to engage in was “denunciation” (kyūdan), essentially an attempt by members to protest discrimination against them by openly engaging in confrontational protest, an action that jarred in a society where hierarchy, obedience, and civic duty were so strongly valorized in political and social discourse. Such a strategy was both meaningful and widespread; only a few months after the Suiheisha was established, seven members of the Nara branch of the Suiheisha, for example, were sentenced for public disturbance after protesting a local high school principal’s condonement of
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
505
discrimination (Upham 1987). This was, however, not an isolated incident – one scholar has suggested the total number of denunciation campaigns conducted between 1922 and 1925 alone was in the thousands (Shimahara 1971). Given the global and domestic movements and fractures that abounded in the late 1910s, particularly the 1918 Rice Riots in which Burakumin were clearly participants but also unfairly targeted, it is no surprise that the Japanese national government had already commenced setting aside a budget for “local improvements” in 1920 (Asada 1979). The local improvements budget that year stood at 50,000 yen and was intended to create medical consultation facilities in Buraku areas, improve hygiene standards and school enrolment figures, as well as prevent delays in the payment of taxes (Teraki and Kurokawa 2016). The 1921 budget rapidly climbed to 145,860 yen, but national expenditures thereafter continued to rise, reaching a high of 2,374,484 yen in 1933, a figure that came to about 0.6% of the total annual national budget for that year (Neary 1989, 2022; Bayliss 2013). The total amount of national funds poured into “local improvements” (thereafter renamed “harmony projects”) between 1920 and 1941 came to more than 22,000,000 yen (Neary 1989, 104). The 10-year plan for assimilation works published in 1935 budgeted 50,000,000 yen to the period ending in 1945, but spending was curtailed as Japan converted to a wartime economy. As the above figures indicate, official attempts were made to address the economic problems within and surrounding prewar Buraku communities, albeit primarily because of apprehension about what these communities’ activities might spell for national cohesion and social stability if left unchecked. It is difficult to ascertain what such expenditure actually meant for Buraku communities throughout the archipelago in real terms, as well as the extent to which the money was meaningfully and evenly distributed across various communities, although it is reasonably clear that from the beginning of government expenditure, prefectures with the largest Buraku communities were directly targeted with the funds (Bayliss 2013; Neary 2022). Some communities, such as Kōzuki in Hyōgo prefecture, engaging in local activism in the 1930s through the auspices of a local Yūwa group known as the “Purity and Harmony Association” (Seiwakai), appear to have seen little state funding (Nakamura 1988). Other communities, however, did benefit through the injection of Yūwa funds: community centers (rinpokan) were built in various Buraku areas in Aichi and Shiga prefectures, among other places (Kubota 1980). Equally importantly, however, national funding for Buraku areas often did little to combat actual discrimination, for as the prevalence of “denunciation” incidents in the prewar era suggests, it was often deemed more effective to take matters into one’s own hands. Indeed as the 1933 Takamatsu court case where a Burakumin who married a woman without “revealing his background” was successfully sentenced for abduction signifies, official government institutions could not be relied upon by members of Buraku communities to secure justice against discriminatory attitudes and actions (Neary 2010). In Asakusa “Newtown” in Tokyo, while the area retained some links to its outcaste past in the post-Meiji era, subsequent reforms, as well as natural disasters such as the Great Earthquake and Fire of 1923, significantly affected the composition
506
T. Amos
of the community. No mention was actually made of Imado in the 1935 National Buraku Survey, suggesting that former associations between the area and early modern outcaste communities had largely come undone by that stage, at least in the bureaucratic imagination (Arai 1967, 86). In the Osaka area, the same 1935 survey listed Asaka (Sugimoto Shinden/Yosami) as having 265 households and 1193 residents (Chūō Yūwa Jigyō Kyōkai 1935). While it is not clear that residents immediately identified with the Suiheisha movement, they were involved in large protests in the 1920s, including against the amalgamation of the area into the city of Osaka. Within the same city of Osaka, however, other Buraku in the city, such as the one in Nishihama, were closely connected with the Suiheisha, establishing a branch of the organization less than a half a year after the movement’s initial founding (Naniwa-ku 2015). The extent to which these areas were the beneficiaries of significant state funding in the prewar period requires further research. Approaches to Buraku liberation during the period 1922–1945 deviated from one organization to the next, but as can be seen above, they could also differ between and even within communities. Differences emerged through competing ideological convictions, proximity to the Suiheisha and Yūwa movements, varying assessments of the current social and political situation, the place one envisioned Burakumin to have in the social order, the degree to which one ceded that compromise was important, and the texture of one’s ultimate vision for a “levelled” future. For those who imagined a self-emancipated future as being distinctly impossible within the strictures of prewar Japanese society, the struggle tended to be more intense, singularly focused, occasionally violent, and decidedly inflexible. For those more willing to privilege the long-term survival of the liberation movement over personal preferences, the struggle tended to be incremental, strategic, broadly conceived, compromising, and tenacious. Regardless of one’s approach to social and political action, however, most Burakumin connected with the Suiheisha recognized the evershrinking latitude with which one could go about activism that aggressively targeted the state or high-profile members of society and the reality of the ever-growing necessity to work with authorities and leaders at all levels of society to achieve even modest outcomes. At the same time, progress toward Buraku liberation in the prewar period continued to be hampered by numerous factors. Society, whether through marriage, education, or other key institutions, tended to preserve discrimination against Burakumin in ways that perpetuated pain and misery and demanded immediate redress. The state, moreover, proved incapable of and largely unwilling to work together for the cause of Buraku emancipation through its institutional arms. The 1925 Peace Preservation Law criminalized vocal forms of political opposition and social dissent; educational institutions engendered exceptionalism and fostered fascism; the judiciary proved reluctant to punish people and institutions that engaged in discrimination; the military failed to stamp out discrimination against Buraku conscripts; and local government instituted modern forms of capitalist discrimination and racialized difference in ways that preserved older discriminatory structures and practices albeit in modified forms. The economic distresses that came with the Great Depression, as well an increasing economic divide between urban and rural
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
507
populations, and the steady growth in Japanese militarism and domestic fascist controls all worked together to compound the nature of the problem, pushing resolution even further from reach. The suspension of financial support for local improvement projects in 1941 was also followed soon after by the suspension of the Suiheisha organization itself. The Buraku problem in many ways went underground during the war years, in the face of ideological trends and wartime policies that demanded internal unity and conformity within nation and empire.
Postwar Buraku Liberation The postwar Buraku liberation movement recommenced in 1946 with the establishment of the National Committee for Buraku Liberation (Buraku kaihō zenkoku i’inkai, hereafter NCBL). The NCBL was established for the purpose of “the complete liberation of the Buraku masses,” welcoming all prewar Buraku organizational members regardless of political affiliation and offering full membership to Buraku and non-Buraku people alike (Amos 2017b). Kyoto, untouched by air raids and home to some of Japan’s oldest premodern outcaste communities, became the new base for a resurgence in postwar activism; indeed, some of the chief architects and influencers of Japan’s postwar Buraku liberation movement were either born in Kyoto or came to work there during the Occupation years (Amos 2011). Most of the significant leaders within the NCBL had been heavily involved in the Suiheisha movement and were closely aligned with Marxist ideology, although not exclusively or always consistently, and some important prewar Yūwa movement figures were also quite closely involved in the reestablishment of the early postwar Buraku liberation movement. In the early postwar years, some local governments actually began funding initiatives in Buraku areas in advance of the national government, usually in response to local pleas for assistance, hinting at the possibility of some positive lingering effects of prewar state funding policies (Mizuuchi 1998). While some of this funding was not linked to NCBL activism, it seems likely that the scale and consistency of such funding would have remained negligible without a strong push from the organization or individuals active in the prewar movement. The NCBL drew upon the prewar Suiheisha in terms of membership, outlook, and strategies, and the relatively antagonistic approaches adopted toward the Japanese state and local governments that characterized the prewar movement were quickly mirrored in the NCBL’s postwar movement policies. The prewar “harmonization” strand within government and business circles, however, also continued into the postwar period, with the formation of the National Assimilation Policy Committee in 1951 and the establishment of National Assimilation Society in 1960 (the forerunner to the National Free Assimilation Society formed in 1986), although some important figures within this “harmonization” strand of the movement initially tended to work with rather than against the NCBL. And while the early postwar liberation movement continued to struggle with the question of how to conceive of and balance the ideal relationship between self-support and state provision, on the ground, the relationship between government officials and activists remained
508
T. Amos
complicated, as witnessed in the politics surrounding the purging of elected politicians and administrators from Buraku backgrounds who had held positions during the war (Neary 2010, 2022). The first significant inroad made by the NCBL in the postwar period was through a censuring of the Kyoto municipal government in 1951 over a discriminatory literary piece written for the magazine All Romance by an employee working at a local public health center belonging to the Environmental Hygiene Division. The NCBL, led by Zennosuke Asada who had fostered close wartime contacts within the municipal government, used the employment background of the author to launch a sustained and eventually successful attack on the mayor and his administration (Hankins 2014). Leveraging on this campaign, as well as another fought the following year in Wakayama prefecture, the Buraku Liberation League (BLL), the organizational name that replaced the NCBL in 1955, began to make national inroads with their activism by the late 1950s. In the face of sustained criticism, the national government entered into dialogue with key figures involved in the liberation movement, with a promise to support investigations into the social and economic conditions of Buraku conditions throughout Japan. In 1961, the Dōwa Taisaku Committee began meeting to research and write a report that was completed in 1965 and that contained a recommendation including four broad measures that encompassed environmental improvements, social welfare measures, industry and occupational initiatives, and educational policies. As McCormack has noted, however, national government action in relation to the Dōwa issue is probably attributable to a fear of Buraku radicalization and the development of a united progressive front; the absence of official participation in various key deliberations throughout the 1960s, the slowness in appointment of committees and acting on recommendations, and the decision to significantly increase public spending in Buraku areas well prior to the outcome of deliberations all suggest an underlying desire to tame the liberation movement (McCormack 2018). The national government, in response to the Dōwa Taisaku Committee report, as well as to considerable further agitation on the part of the BLL witnessed in events such as the “Sayama Struggle” (Ishikawa 2019), passed the Dōwa Special Measures Law (SML) in 1969, legislation that committed local governments with national budgetary support to target Buraku communities (referred to officially as Dōwa residential areas or “Residential Areas targeted for Assimilation”) within their respective jurisdictions with measures to address the aforementioned four key areas for at least an initial 10-year period. The SML had as its objectives “improvement of the living environment, promotion of social welfare, enhancement of industry, stabilization of employment, fulfillment of education, and strengthening of activities pertaining to human rights protections in order to eliminate the harmful causes that prevent the elevation of the social and economic position of residents within the targeted areas” (Shūgi’in 1969). Local governments, in line with the new legislation, began to distribute substantial amounts of funding from this time to a variety of bodies connected with both the BLL and Buraku areas. Government recognition of a “Dōwa problem” (i.e., the existence of areas that had failed to properly socially integrate) and its subsequent public commitment to
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
509
use taxpayer money to resolve conflicts arising out of such “integration issues” (dōwa mondai) was not universally recognized by Burakumin themselves as the best pathway forward to liberation. Conflict arose within the BLL about the desirability of accepting such government funding. While many Socialist Party stalwarts in the BLL fully endorsed the law and looked forward to the advances that could be made through the promised funding, Communist Party affiliates tended to be highly critical of the law, labeling it among other things a “poisoned cake” (dokumanju), anticipating and fearing the role the new legislation would have in preventing bluecollar unity and triggering state reliance within Buraku communities (Reber 1999). In-house fighting within the BLL, compounded by the death of the iconic leader Ji’ichirō Matsumoto and his replacement by the more controversial and confrontational figure of Zennosuke Asada, led to subsequent organizational rifts that triggered large numbers of membership annulments and the establishment of rival factions in the second half of the 1960s (Neary 2010). While the reasons for the rift within the BLL are complex, considerable disagreements existed between members of the movement that rested on both philosophical and tactical grounds, going beyond the issue of how a conservative government may (or may not) have been simply using public funds to compromise the strength of the postwar labor movement. While almost all stakeholders could and did acknowledge that “Buraku discrimination” (Buraku sabetsu) was a problem and that “Buraku residents” (Buraku jūmin) urgently required liberation (kaihō), considerable disagreement surfaced in relation to the details. Were Burakumin a group akin to an ethnic group or tribe, as had begun to be intimated by some in the prewar period, or were they people ascribed a certain status based on the false consciousness and faulty reasoning of a society that had stubbornly refused to leave the past behind? Was it advisable to receive government handouts to address the issues involved? How long should Buraku communities be targeted for economic assistance, and what should happen to those communities that did not wish to be targeted by the special measures? What practical form, moreover, should Buraku liberation politics adopt in the SML era? How should Burakumin engage with the rest of society as they struggled to assert a place for themselves in a world that was increasingly vocal about and accepting of civil rights and unique identities? The issues such questions pointed to were real, immediate, and complex, and varying answers to them brought sections of the Buraku liberation movement into serious conflict, leading to further fractures in the movement that still persist to the present. The most immediate consequence of the movement split was the formation of a separate organization that set as its goal the “normalization” of the BLL. This was eventually followed by the establishment of a rival liberation outfit in the 1970s affiliated with the JCP called Zenkairen (National Buraku Liberation Alliance) that thereafter became one of the BLL’s harshest critics (Buraku mondai kenkyūjo 1998). The BLL’s position, although itself complex given the very different leadership styles and foci of charismatic and powerful figures such as Zennosuke Asada and some of the subsequent BLL chairmen, was that the Japanese Communist Party (JCP) and Zenkairen were simply putting their own interests and agendas in front of the core issue of Buraku liberation. The non JCP-aligned group within the BLL did
510
T. Amos
not want the Buraku problem to be something that was sacrificed solely in the name of proletarian unity; and although the language of “minority” was not yet widely in currency, Burakumin within the BLL increasingly viewed themselves in such a light, spurred on by their landmark victory of concessions in the form of the SML in the face of “total societal discrimination” and “administrative stagnation,” and their observance of civil rights struggles in other places. For those who remained in the BLL, the path forward was not only to consolidate upon the successful campaign for the SML but to devise even more strategies to effectively capitalize upon these much needed gains. Two important areas that came to take prominence in BLL strategy from the late 1970s were public education and anti-discrimination campaigns. JCP-affiliated groups such as Zenkairen, however, were fundamentally opposed to the idea that Burakumin were indeed a minority in the sense that was beginning to appeal to members of the BLL (Asada 1979; the BLL’s second chairman Asada was clearly moving in that direction by the late 1970s). Those who had left the BLL argued that by simply accepting the SML and continuing along a path of government handouts, Buraku areas would effectively become overly reliant on government welfare, develop a clannish mentality that would only serve to perpetuate discrimination against them by mainstream society, and potentially engender an exaggerated victim consciousness that interpreted every single feature within society not agreeable to BLL members as fundamentally discriminatory. The way forward, they argued, was through “national integration”: rectifying the environmental, labor, and educational disparities between Buraku areas and neighboring communities; ensuring that “unscientific” understandings (i.e., prejudice) were eliminated from society; eliminating factors possibly contributing to discrimination arising from within Buraku communities themselves related to lifestyle attitudes or customs; and assimilating regions so that meaningful social interaction could take place (Tōjō 2018). Zenkairen critical rebuttals of BLL positions and activities were relentless and often incendiary, but they were also quite limited in their reach and scope, at least until around the time of the legal battle that originated over international recognition of a BLL-backed NGO that ensued between the two organizations during the years 1991–1994 (Ishikawa 1995; Tōjō 2018). The BLL clearly remained the largest Buraku liberation organization throughout the SML era by some margin, with a membership of 300,000 in 1989, eclipsing both the Zenkairen (80,000) and the Jiyū Dōwakai (the Liberal Assimilation Association) (40,000) (Asahi Shinbun 1989). As the largest liberation group with an unparalleled reach, the BLL was also often permitted to singularly legally administer the funds made available through the SML implemented within local Buraku communities, a process described as a “single window” policy (madoguchi ipponka) (Reber 1999). Strenuous Zenkairen objection to the BLL, of course, did not necessarily equate to refusal on their part to secure funds made available through the SML to address the needs of affected communities on behalf of its members. However, in effect, in many if not most cases, it appears that local branch managers of the BLL were given extraordinary levels of oversight in relation to the administration or coordination of local reforms (Ishikawa 1995).
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
511
The remarkable level of control permitted to the BLL (and to a lesser extent Zenkairen and Jiyū Dōwakai) to target areas of explicit need within local Buraku communities was in many cases a policy that worked to invest economic and social resources within communities in a targeted manner maximizing community benefits. The visibility of real socioeconomic differences combined with systematic institutionalized practices of discrimination against Burakumin was often targeted through a range of infrastructural and educational initiatives to great effect particularly at the regional level. The localized systems developed for the administration of state funds during the SML era, however, also opened the door for potential abuses, charges that were eventually levelled at a number of BLL branches or individuals within such branches, and to a lesser extent Zenkairen and other liberation group branch associations, in later periods (Terazono, Ichinomiya, and Gurūpu K21 2003). While many Burakumin leaders worked within their locales and with local governments to establish mechanisms to protect and develop their communities, entrenchment of privileges as well as abuse of the public fund distribution system certainly did sometimes take place, leading to a sense of entitlement and, in some more severe cases, the development of connections with the Japanese underworld. Contrary to some recent pseudo-scholarship that wishes to portray the BLL and liberation organizations at large as essentially criminal associations in their own right (Amos et al. 2021), the real picture (once such discriminatory views rooted in anecdotal evidence and environmental fallacies are put aside) is that tainting of the Buraku liberation movement took place through the actions of a handful of people who illicitly gained from their positions of trust and responsibility. And while Zenkairen and other critiques of the BLL were not without basis or merit, nonetheless SML-era distributive mechanisms proved successful in many if not most cases, lessening the discrimination faced by these communities and righting the historic, systemic disadvantages experienced by Burakumin in general (Neary 2021). The sheer magnitude of the amount of money poured into the SML between the years 1969 and 2002 is worthy of careful reflection. Under several different guises and appellations, and with different points of emphasis, the SML delivered somewhere in the vicinity of 15–16 trillion yen of public funds that were spent on various Buraku-related problems and initiatives (Amos 2011). Yet how these funds impacted particular Buraku communities on the ground is a story that requires further research and synthesis, beginning with the basic problem of how Buraku communities were defined. The Dōwa Response Committee that met in 1963 to discuss the problem and consider their recommendations for how the “Buraku problem” could be resolved initially planned to target 4160 districts with a population of 1,867,748 Buraku residents (Kadooka 2005). The eventual number of districts that became Dōwa (Buraku) areas, however, was, at least initially, considerably less, with only 3545 districts recognized in 1969. Somewhat perplexingly, that number rose to 4603 districts in 1993, a figure well above the 1963 number (Kadooka 2005). The disparity in community totals serves as an important reminder of the complexity of the “Buraku problem”: not every community identified in prewar statistical analysis as a Buraku area became recognized as a Dōwa area after 1969 and not everyone
512
T. Amos
identified as a Burakumin by a liberation organization claims such an identity for themselves. Natural disasters as well as tragic historical events such as the fire bombing raids by Allied Forces during the Asia Pacific War also left their mark on many Buraku communities in larger urban settings. This means that local conditions and community improvement were heavily impacted by a host of issues including public recognition and the vibrancy of local activism during the SML era. Examples are helpful in illustrating these points. In early postwar Tokyo, for example, although 2153 households were apparently marked down as potential recipients of government funding after the issuance of the Dōwa Response Committee report in 1965, no detailed surveys of areas deemed to be Dōwa (Buraku) areas were immediately carried out after the establishment of the SML (Harada 1975, 371). This effectively meant that for administrative purposes, although certainly not of course from the perspective of any self-identifying Burakumin living in Asakusa “Newtown” (or Imado) at the time, Buraku life in Tokyo became somewhat of a non-geographically based reality for government after 1969 (McCormack 2018). On the ground, however, the BLL remained committed to keeping a sense of the identity of the area of Imado as a Buraku area alive, although its activities in Tokyo also cannot be divorced from its overarching political strategy for furthering activism by working to extend its influence with important powerbrokers in the nation’s capital. In a series of articles written in 1975 for the BLL weekly newspaper Kaihō Shinbun (Kawamoto 1975a, b, c, d, e), for example, Yoshikazu Kawamoto engaged in interviews with people from various occupations closely connected to traditional outcaste industries such as leatherwork, butchery, and weaving. Kawamoto found discrimination in various forms, including stigma, wage differentials, and the maintenance of a deafening silence about one’s ancestry for fear of being outed as a Burakumin. Kawamoto also found a considerable level of antagonism between the BLL and JCP in some communities, going as far as to claim that the latter group was essentially waging an “anti-BLL campaign in some areas.” It is clear from Kawamoto’s reportage, however, that the traces between an outcaste past and a Buraku present were becoming exceptionally thin, despite the BLL’s best efforts to work through its Taitō and (newly established) Shinagawa ward branch offices. While it is clear in the coverage that BLL engagement was supported by the SML, the degree to which this translated to tangible benefits for local Buraku residents is unclear. In the city of Osaka, similar problems with counting and representation emerged in the postwar era, suggesting the need for caution when addressing issues of the efficacy of financial redress in that city. In 1935, Osaka city was listed as having in excess of 40 Buraku areas, but growth in the postwar Buraku liberation movement in Osaka took time, with BLL branch offices numbering only 7 in 1960 before growing to 47 in 1972 (Osaka no Burakushi iinkai 2009, 420). One of those areas that actually displays strong continuity as an outcaste-Buraku-Dōwa area is Asaka (discussed in the previous section). In the postwar period, the community again organized resistance to the building of a large railyard in the area, an action that was followed by a
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
513
movement to improve housing in the 1960s and an action plan to rebuild the community in the 1970s. A good deal of the activism in the area was conducted under the leadership of Yamamoto Yoshihiko, a popular activist who received localand national-level BLL official appointments. Under Yamamoto, Asaka experienced a complete transformation during the years of the SML (Nishimura 2010). While it is still difficult to get a true sense of the amount of public funds poured into the Asaka area in the 33 years the SML were active, the figures must have been substantial given the tremendous facelift the community received during that period.
Conclusion At present, with perhaps the exception of Neary (2022) and the Japanese language work he relies on in his important volume, precious little scholarship exists to help us gauge the full extent of the massive financial investment in postwar Japanese Buraku liberation, although the basic story line is well understood and recounted consistently enough. Whether or not this massive infusion of public funds into Buraku areas, industries, and education had the full desired effect in eliminating discrimination against Burakumin is one of the main bones of contention among the major streams of the liberation movement, linked as it is to the question of whether discrimination against Burakumin ostensibly remained in a meaningful form after the cessation of the SML in 2002. While there is no space to retell this story here, it can be pointed out that virtually every group in society apart from the JCP and its affiliated human rights group Zenkoku chi’iki jinken undō sōrengō (National Confederation of Human Rights Movements in the Community; Jinkenren, the new name given to JCP-backed activist group after the voluntary folding of the Zenkairen in 2004), now publicly affirms their belief that Buraku discrimination remains a serious problem in society, a view that was expressed particularly strongly at the time of the passing of the Act on the Promotion of the Elimination of Buraku Discrimination (APEBD) in 2016. How postwar government funds were distributed (or not distributed) to Japan’s postwar Buraku communities and the role they played in transforming and potentially alleviating a vast array of economic and social problems experienced by such communities in the modern era is a complex story that awaits a fuller recounting. Clearly it is a story that must be pieced together based on an examination of the experiences of thousands of disparate communities linked together by the presence of one or more liberation organizations that usually remained antagonistic to each other, had differing degrees of working relationships with local governments, and left their mark in various ways that have in turn reshaped localized visions of Burakumin and their place in society. A central part of any future history must also focus on the role local governments and third-party organizations (PILPS/NGO/ NPOs, etc.) played in administrating and allocating funds and implementing programs. The task of understanding how funding for Buraku liberation trickled down from numerous national and local government departments and bodies to localities is also a complex subject that is only now coming to be better comprehended, and it is clear that the lion’s share of funding between the years 1969–1991 came from the
514
T. Amos
Ministries of Agriculture, Forestry, and Fisheries, Construction, and Health and Welfare, as well as from various local governments (Mizuuchi 1998; Neary 2022). Also particularly difficult to track is affirmative action in the area of public sector employment among Burakumin in Japan. Tens of thousands of Burakumin were apparently recruited into public service in the postwar period. Kyoto municipality alone is said to have hired 6000 full-time Buraku public servants during the period 1969–2002, and in Osaka, the numbers are likely much higher (Amos 2011). The SML played an important role not only in addressing the “Buraku problem” but also in working to shape Buraku identity in distinctive ways in the postwar era. From the 1960s, influenced by the civil rights movement in the United States and elsewhere, Burakumin not only worked for equality with the rest of society in affected areas and among related peoples, but they also began to emphasize through their activism, particularly in the realms of education and community engagement, that the onus lay on society to develop a variety of systems that would ultimately recognize the successful retention of a unique Burakumin minority identity (Fukuyama 2019; see Fukuyama for more on the US dimensions of this global trend). The Buraku problem came to be regarded both inside and outside the liberation movement as a human rights issue affecting a minority group which required a variety of economic and other supports to enable their complete protection from society as well as the full expression of Buraku identity and difference. While significant sections of the Buraku liberation movement resisted this “redefinition,” it is clear that new forms of discrimination also came to accompany this emerging form of identity politics, something groups such as the Zenkairen and later the Jinkenren have been slow to recognize. This became particularly conspicuous as excesses within the liberation movement came to light after the cessation of the SML in 2002 and as society at large began to publicly baulk against public sector policies of affirmative action in an area of considerable economic decline and stagnation. Nonetheless, the Buraku liberation movement began to further reinvent itself in the twenty-first century and lobbied successfully for the establishment of the APEBD in 2016. Yet questions remain about the extent to which developments triggered by this new mode of identity politics and the promise of new economic supports that accompany them in Japan will actually permit a future reduction in the kinds of discrimination Burakumin face in the present.
References Amos TD (2011) Embodying difference: the making of Burakumin in modern Japan. University of Hawaiʻi Press, Honolulu Amos TD (2015a) Asakusa ‘Newtown’: the transformation of outcaste space in early modern Edo/modern Tokyo. Jpn Forum 27(2):213–234 Amos TD (2015b) Fighting the taboo cycle: Google map protests and Buraku human rights activism in historical perspective. Jpn Stud 35(3):331–353 Amos TD (2017a) The subaltern subject and early modern taxonomies: Indianisation and racialisation of the Japanese outcaste. Asian Stud Rev 41(4):577–593
21
Buraku Liberation and the Politics of Redress in Modern Japan, 1868–2002
515
Amos TD (2017b) Pathways to Buraku liberation: competing images of freedom in early postwar Japan. J Northeast Asian Hist 14(2):95–118 Amos TD (2020) Caste in early modern Japan: Danzaemon and the Edo outcaste order. Routledge, Abingdon Amos TD, Ehlers M, McKnight A et al (2021) Doing violence to buraku history: J. mark Ramseyer’s dangerous inventions. Asia Pacific J Jpn Focus 19(9). https://apjjf.org/2021/9/ Amos-Ehlers-McKnight-Ambaras-Neary.html Arai K (1967) Toshi senmin gyōseishi no kiso kōsatsu: tokyo no hisabetsu buraku to ‘gōmune’ buraku no baai. Tōyō hōgaku 11(4):83–115 Asada Z (1979) Sabetsu to tatakaitsuzukete. Asahi Shinbunsha, Tokyo Asahi Shinbun (1989) Dōwa sandantai ga hatsu no iken kōkan: sabetsu no genjō to kongo kataru, April 18 Bayliss JP (2013) On the margins of empire: Buraku and Korean identity in prewar and wartime Japan. Harvard University Asia Center, Cambridge, MA Bondy C (2015) Voice, silence, and self: negotiations of Buraku identity in contemporary Japan. Harvard University Asia Center, Cambridge, MA Botsman D (2016) The return of the outcast(e) map: Kobe, cartography and the problem of discrimination in modern Japan. Asia Pacific J Jpn Focus 14(3). https://apjjf.org/2016/18/ Botsman.html Buraku mondai kenkyūjo (1998) Buraku mondai kenkyūjo: 50 nen no ayumi. Buraku Mondai Kenkyūjo, Kyoto Chūō Yūwa Jigyō Kyōkai (1935) Zenkoku Buraku Chōsa. Chūō Yūwa Jigyō Kyōkai, Tokyo De Vos G, Wagatsuma H (eds) (1966) Japan’s invisible race: caste in culture and personality. University of California Press, Berkeley Fujino Y (1994) Hisabetsu buraku. In: Asao N, Amino Y, Ishii S, Kano M (eds) Iwanami kōza nihon tsūshi, 18. Iwanami Shoten, Tokyo, pp 135–167 Fukuyama F (2019) Identity: the demand for dignity and the politics of resentment. vol Book, Whole. Picador, New York Habib I (2002) Essays in Indian history: towards a Marxist perception; with the economic history of medieval India: a survey. Anthem Press, London Hankins JD (2014) Working skin: making leather, making a multicultural Japan. University of California Press Harada T (1975) Hisabetsu Buraku no rekishi. Asahi shinbunsha, Tokyo Ishikawa M (1995) “Kaidō bōryoku kyūmei saiban” shori no riyu. Buraku mondai kenkyūjo, Kyoto Ishikawa M (2019) Exclusionism and the Burakumin: literacy movement, legislative countermeasures and the Sayama incident. In: Shiobara Y, Kawabata K, Matthews J (eds) Cultural and social division in contemporary Japan: rethinking discourses of inclusion and exclusion. Routledge, London Kadooka N (2005) Hajimete no buraku mondai. Bunshun Shinsho, Tokyo Kawamoto Y (1975a) Buraku Tanbōki: Tokyo no hisabetsu buraku I. Kaihō Shinbun, June 23 Kawamoto Y (1975b) Buraku Tanbōki: Tokyo no hisabetsu buraku I. Kaihō Shinbun, June 30 Kawamoto Y (1975c) Buraku Tanbōki: Tokyo no hisabetsu buraku I. Kaihō Shinbun, July 7 Kawamoto Y (1975d) Buraku Tanbōki: Tokyo no hisabetsu buraku I. Kaihō Shinbun, July 14 Kawamoto Y (1975e) Buraku Tanbōki: Tokyo no hisabetsu buraku I. Kaihō Shinbun, July 21 Kubota K (1979) Senzen ni okeru dōwa chiku rinpo jigyō no rekishi (jō). Buraku kaihō kenkyū 19: 43–67 Kubota K (1980) Senzen ni okeru dōwa chiku rinpo jigyō no rekishi (ge). Buraku kaihō kenkyū 22: 116–135 Maki H (1979) Kinsei Osaka shinai ni okeru hisabetsu buraku no rekishi. Dōwa mondai kenkyū 3: 53–106 McCormack N (2013) Japan’s outcaste abolition: the struggle for national inclusion and the making of the modern state. Routledge, New York
516
T. Amos
McCormack N (2018) Affirmative action policies under the postwar Japanese constitution: on the effects of the dōwa special measures policy. Asia Pacific J Jpn Focus 16(4). https://apjjf.org/ 2018/5/McCormack.html Mizuuchi T (1998) Jūkankyō kaizen kara mita dōwa jigyō no rekishi to genjō. Chiri kagaku 53(3): 163–173 Motohama R (2009) Jinkō tōkei ni miru nihon no kindaika to hisabetsu buraku. Sōgō kenkyūjo shohō 17:41–54 Nakamura F (1988) Yūwa undōshi kenkyū. Buraku mondai kenkyūjo, Kyoto Naniwa-ku Osaka-shi (2015) Nishihama Suiheisha hasshō no chi” no hi: “Osaka no jinken kakuritsu no furusato. https://www.city.osaka.lg.jp/naniwa/page/0000000857.html. Accessed April 20 2020 Neary I (1989) Political protest and social control in pre-war Japan: the origins of Buraku liberation. Manchester University Press, Manchester Neary I (2010) The Buraku issue and modern Japan: the career of Matsumoto Jiichirō. Routledge, London Neary I (2021) Professor mark Ramseyer and the Buraku question: an introduction. Asia Pacific J Jpn Focus 19(9). https://apjjf.org/2021/9/Neary.html Neary I (2022) Dōwa policy and Japanese politics. Routledge, London Nishimura Y (2010) Civic engagement and community development among Japan’s Burakumin. In: Vinken H, Nishimura Y, White BLJ, Deguchi M (eds) Civic engagement in contemporary Japan: established and emerging repertoires. Springer, New York Osaka no Burakushi I’inkai (2009) Osaka no Burakushi, vol 10. Buraku kaihō/jinken kenkyūjo, Osaka Osatake T (1999) Meiji yonnen senshō haishi fukoku no kenkyū. Hihyōsha, Tokyo Reber E (1999) Buraku mondai in Japan: historical and modern perspectives and directions for the future. Harv Hum Rights J 12(Spring):297–359 Satogami R (2003) Rekishi bukai/gakushū hōkoku. https://blhrri.org/old/kenkyu/bukai/rekishi/ rekishi/rekishi_0027.html. Accessed March 30 2020 Shields JM (2017) Against harmony: progressive and radical Buddhism in modern Japan. Oxford University Press, New York Shimahara N (1971) Burakumin: a Japanese minority and education. Martinus Nijhoff, The Hague Shūgi’in (1969) Dōwa taisaku jigyō tokubetsu sochihō. http://www.shugiin.go.jp/internet/itdb_ housei.nsf/html/houritsu/06119690710060.htm. Accessed June 26 2020 Sugimoto Y (2014) An introduction to Japanese society. Cambridge University Press, Cambridge Suzuki R (1985) Kindai nihon Buraku mondai kenkyū jōsetsu. Hyōgo Buraku Mondai Kenkyūjo, Hyōgo Teraki N, Kurokawa M (2016) Nyūmon hisabetsu buraku no rekishi. Kaihō Shuppansha, Ōsaka Terazono A, Ichinomiya Y, Gurūpu K21 (eds) (2003) Dōwa riken no shinsō: masumedia ga mokusatsu shitekita sengoshi saigo no tabū. Takarajimasha, Tokyo Tōjō T (2018) Buraku mondai kaiketsu katei no shōgen: kenkyūjo no 70 nen wo chūshin ni. Buraku mondai kenkyūjo, Kyoto Totten GO, Wagatsuma H (1966) Emancipation: growth and transformation of a political movement. In: De Vos G, Wagatsuma H (eds) Japan’s invisible race: caste in culture and personality. University of California Press, Berkeley Uesugi S (1990) Meiji ishin to senmin haishirei. Kaihō Shuppansha, Ōsaka Upham FK (1987) Law and social change in postwar Japan. Harvard University Press, Cambridge Watanabe M (1961) Meiji kaihōrei happu no jijō ni tsuite: tokuni nozokechi mondai wo chūshin ni. Nihon daigaku shigakkai kenkyū gohō 4:38–50
Caste and Socoieconomic Disparities in India: An Overview
22
Rajesh Ramachandran
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Social Reproduction of the Caste System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Human Capital and Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Economic Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consumption, Income, and Wages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occupation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Poverty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wealth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Education and Public Sector Employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Politics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Regional Patterns of Caste Disparities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
518 521 524 524 526 527 527 528 529 530 531 531 533 535 538 539 539
Abstract
The chapter provides an overview on the extent of caste disparities in contemporary India. The evidence shows that the social basis of reproduction of caste – endogamy and untouchability practices – remain prevalent and widespread. Caste-based affirmative action is seen to be instrumental in improving human capital outcomes and providing access to better jobs for the disadvantaged group though in spite of these the gaps in higher education and white collar occupation are seen to have widened in the post-independence era. The regional patterns suggest that the caste disadvantage are especially acute in the northern and central R. Ramachandran (*) Department of Economics, School of Business, Monash University Malaysia, Selangor, Malaysia e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_23
517
518
R. Ramachandran
plains of India. The evidence shows that caste remains a key mediator of socioeconomic status in today’s India. Keywords
Caste disparities · Affirmative action human capital · Health · Income · Occupation
Introduction In the last two decades there has been a rapid increase in the number of theoretical and empirical studies explicitly accounting for the role of caste in the Indian economy.1 This chapter attempts to provide a concise overview on the evolution of caste groups on various socioeconomic indicators, as well as the extent of caste inequality, after 75 years of independence. The caste or Varna system is one of the oldest systems of social stratification. The word first appears in the Rigveda.2 However, the explicit classification of society into four varnas is elaborated in the Manusmriti, with its “final form” dated to around the second century BCE. The four varnas are the Brahmins (the priests, scholars, and teachers), Kshatriyas (the rulers, warriors, and administrators), Vaishyas (the agriculturalists and merchants), and Shudras (the laborers and service providers). Those who fall out of this system because of their previous sins are ostracized as outcastes (untouchables), considered outside the varna system, and referred to as avarna (Olivelle 2005). The word casta in the modern sense of the English word “caste” as endogamous, hereditary Indian social groups was first used by the Portuguese to describe the social milieu they perceived upon their arrival in India in 1498. The other relevant unit of social organization in India are Jatis, that is, “the operative category that determines the contemporary social code” (Deshpande 2001, p. 131). Under the Jati system, a person is born into a Jati with ascribed social roles and endogamy, i.e., marriages take place only within that Jati. Each Jati typically has an association with an occupation, geography, or tribe. They also form the basis of economic and social networks providing not only insurance against risks and shocks, but also act as providers of credit and capital and enable access to labor and business networks (Munshi 2019). Jatis are often referred to as sub-castes (Morris 1965) though their explicit identification with a specific Varna or caste arises from 1901 onward when the British, for the purposes of the Decennial Census, classified all
1 For instance, see Deshpande (2011) and Thorat and Neuman (2012) for detailed analysis of the role of caste in economic outcomes and Mosse (2018) and Munshi (2019) for recent reviews of the literature. 2 The evidence indicates that the bulk of the Rigveda Samhita was composed in the northwestern region of the Indian subcontinent between c. 1500 and 1000 BCE (Witzel 2019).
22
Caste and Socoieconomic Disparities in India: An Overview
519
Jatis into one or the other of the four varna categories.3 In addition, the British created the fifth category for what they described as “a fifth division for the large and miscellaneous group of untouchables (asprishya Sudra)” (Gait 1913, Chapter XI, 365). Thus, the Jatis which acted as the basis of daily social life, and in certain cases, straddled two or more Varnas based on their occupations, was grouped into a particular Varna by census enumerators who decided their caste (Kothari 1997). The British government in 1932 granted separate electorates in India for Muslims, Buddhists, Sikhs, Christians, Anglo-Indians, Europeans, and “Depressed Classes” (the individuals subject to the practice of untouchability). The Poona Pact between Gandhi and Ambedkar resulted in the dilution of the grant and it was accepted that for the “depressed classes” there would be joint electorates along with reserved constituencies.4 In 1937 began the first listing (scheduling) of these Jatis. At Independence, the newly independent Indian state decided to use the classification arising from the creation of the category of “depressed classes” and carried forward with the scheduling of these Jatis as the basis of extending affirmative action in government jobs and higher education. This resulted in the creation of the category of the Scheduled Castes (SC), the individuals previously subject to the practice of untouchability, who were to be afforded affirmative action in the newly created state. The seven explicit criteria that were employed for selection of groups as SC and listed in the Constitution of 1950 are summarized in Table 1; as can be seen they all pertain to social exclusion in various spheres arising from being subject to the practice of untouchability. Most of the tribal population of India was not included among the depressed classes and they were not earmarked as a group to be afforded affirmative action in the first report in 1947 on minority rights. The scheduling of these tribes and the creation of an inclusive schedule for these tribes happened in 1950. The new 1950 Constitution of India came into force and listed the criteria for selection of groups to be classified as tribal and are summarized in Table 1. The Constitution provides the same preferential treatment to the group that came to be classified as the Scheduled Tribes (ST). While the nomenclature of SC and ST are the official administrative categories, Dalit, meaning “oppressed,” and Adivasi, meaning “indigenous people,” is often used to describe the SC and ST communities. The Mandal Commission, or the Socially and Educationally Backward Classes Commission (SEBC) established in 1979 with a mandate to “identify the socially or educationally backward classes” of India resulted in the creation of the category of the Other Backward Classes (OBCs). The OBCs who were extended affirmative action in government jobs in 1993 and higher education in 2006. Consequently, 3
The 1872 and 1881 censuses already attempted to classify people by the Varna system though it was not uniformly applied and implemented throughout the country. 4 This implied that in the reserved constituencies only candidates from the “depressed classes” could be a candidate; however, the electorate would consist of both the “depressed classes” as well as the rest of the individuals living in the constituency, instead of just the former, as Ambedkar desired and had originally demanded.
520
R. Ramachandran
Table 1 Criteria used for the identification of scheduled castes and scheduled tribes Selection criteria for scheduled castes 1. Cannot be served by clean Brahmans 2. Cannot be served by the barbers, water-carriers, tailors, etc. who serve the caste Hindus 3. Pollutes a high-caste Hindu by contact or by proximity 4. Is one from whose hands a caste Hindu cannot take water 5. Is debarred from using public amenities such as roads, ferries, wells, or schools 6. Will not be treated as an equal by high-caste men of the same educational qualification in ordinary social intercourse 7. Is depressed on account of the occupation followed and, but for that, occupation would be subject to no social disability Selection criteria for scheduled tribes 1. Tribal origin 2. Primitive ways of life and habitation in remote and less accessible areas 3. General backwardness in all respects Source: Pande (2003) Notes: The above criteria were the required basis for the selection of scheduled caste and scheduled tribe communities, as stated in the Constitutional (scheduled caste and scheduled tribe) orders of 1950
India’s affirmative action programs stands as one of the most expansive ones in the world today (Deshpande 2013). The administrative categories of SC, ST, and OBCs have heightened import in today’s India as they form the basis of which individuals are afforded preferential access to higher education, prized government jobs, and political power. This raises a number of important questions about caste in today’s Indian society: is caste purely today a reflection of material differences across caste groups based on historically accumulated disadvantage? Were the practices of social exclusion the stigmatized groups subject to, and used in the identification of beneficiaries, relics of the past or is the social basis for the reproduction of the caste system still in operation today? What has been the efficiency and equity implications of the affirmative action program? Is it the case that the past 70 years of progressive policy has meant that caste is not a relevant cleavage to understanding inequality in 21st century India?5 I review the most recent evidence to shed some light on these questions. Section “Social Reproduction of the Caste System” provides evidence on the social basis for the reproduction of caste in India and its continued relevance in propagating ritual practices that reinforce social hierarchies among groups. Section “Human Capital and Health” examines the empirical evidence on two key capabilities
5
On the other hand, understanding how groups organize life and decisions regarding marriage, risk sharing and insurance, access to credit and capital, as well as to business and economic networks may require us to understand the interactions at the Jati or sub-caste level. This is however not the focus of this review chapter.
22
Caste and Socoieconomic Disparities in India: An Overview
521
required for the achievement of social equality, namely, human capital and health. In section “Economic Outcomes,” the evidence on the extent of caste disparities in economic outcomes, in particular consumption, income, poverty, wages, and wealth is summarized. Section “Affirmative Action” then turns to the existing evidence on the impact of affirmative action in the fields of higher education, government jobs, and politics on mediating the extent of inter-group disparities, as well as its implications for “efficiency.” Finally, section “Regional Patterns of Caste Disparities” presents the spatial variations in the human capital achievements of the caste groups, as well as how gaps in human capital across groups differ by districts in the country and section “Conclusion” concludes.
Social Reproduction of the Caste System Two of the core features of the caste system are its hereditary nature (Gupta 2000) and the notion of hierarchy or graded status of the various caste groups (Dumont 1970). The hereditary dimension of caste comes from caste as endogamous kin groups or Jatis (Mosse 2018). The practice of endogamy aims to maintain the “purity” of hereditary lines and to enclose affinal alliances and exchanges within group boundaries. Ambedkar in his undelivered speech, later self-published with the title Annihilation of Caste wrote “I am convinced that the real remedy is intermarriage. Fusion of blood can alone create the feeling of being kith and kin, and unless this feeling of kinship, of being kindred, becomes paramount, the separatist feeling – the feeling of being aliens – created by caste will not vanish. [. . .] The real remedy for breaking Caste is inter-marriage. Nothing else will serve as the solvent of Caste” (Ambedkar 2014, 20.5). However, after more than 80 years since the publication of Annihilation of Caste, marriage in India remains endogamous (within Jatis). Though information on Jati affiliation has not been collected by the Indian census since 1931, recent rounds of nationally representative surveys such as the Rural Economic and Demographic Survey and the India Human Development Survey (IHDS) show that around 95% of Indians continue to marry within their caste (Munshi 2019; Ray et al. 2020). Moreover, marriage remains principally arranged by families with one of the critical aspects being the kinship group of the prospective match. Figure 1 uses retrospective information from women respondents from the IHDS-2011–2012 to construct the rates of inter-caste marriage for India for the period 1970–2012. The rates of intercaste marriage remain extremely stable with an average of 5.38% over the entire period. The same data also shows that 68.32% of the women in the sample report meeting the husband only on the day of the marriage. Only 13.79% reporting having known the husband for less than a month and only 4.55% have known their husband for more than a year. In sum, arranged marriages organized along caste lines remain the norm in India.
522
R. Ramachandran
.2 .1
2
4
20
05
-1
-0 00 20
19
95
-9
9
-9 4 19 90
9 19
85
-8
4 19
80
-8
9 -7 75 19
19
70
-7
4
0
Proportion of women reporting intercaste marriage
.3
Panel B Men
Year of marriage
Fig. 1 Share of women in intercaste marriages by year of marriage. (Notes: The data is based on a sample of 26,598 women from the IHDS-II. It presents the results of regressing a dummy which takes value one in case women reports husband’s family caste is different from her natal family on the year of marriage period dummies, and then predicting the values of the dependent variable arising from the regression. The gray shaded area indicates the 95% confidence intervals)
The hierarchical nature of the caste system in the view of scholars such as Dumont (1970) follow from religious notions of purity and pollution.6 One of the principle demonstrations of the belief in ritual purity is the practice of untouchability, which is one of the central cornerstones defining the caste hierarchy. Castes whose traditional occupations were considered the most polluting (for instance, scavenging, sweeping, association with dead animals, e.g., in the leather industry) were ostracized and completely segregated such that even their sight was considered polluting. Despite decades of legal or formal equality between caste groups, discriminatory practices against Dalits such as residential segregation, violent hate and sexual crimes especially against Dalit women, denial of entry into temples, prohibitions on intercaste marriages, forms of bonded labor, segregation in classrooms and discrimination by teachers and health care providers, discriminatory access to
6 Bidner and Eswaran (2015) provide a theoretical model that shows that specialization of labor combined with wife’s complementary contribution to the occupation of her husband results in giving rise to endogamy to counter externalities imposed by out-caste marriages. This in turn they argue gave rise to “the purity–pollution dichotomy, the rituals associated with it, and the notions of status that figure so prominently in discussions of caste,” thus they suggest that the rituals of untouchability were the means employed to establish the caste system, and not its cause.
Caste and Socoieconomic Disparities in India: An Overview
523
.5
Panel B - Gaps in HAZ-score
Rajasthan
.4
MadhyaPradesh Jharkhand
.3
Orissa Bihar UttarPradesh Chhattisgarh Gujarat
.2
Maharashtra
Uttarakhand
WestBengal
TamilNadu Delhi
Himachal Pradesh
Haryana Punjab
.1 Kerala
Karnataka
AndhraPradesh Jammu&Kashmir Assam Tripura
0
Proportion of SC HHs reporting experiencing untouchability
22
0
.1
.2
.3
.4
.5
Proportion of HHs reporting practicing untouchability 95% CI
Fitted values
HH members experienced untouchability
Fig. 2 The prevalence and experience of untouchability. (Notes: The prevalence and experience of untouchability at the state level is calculated from the second round of the Indian Human Development Survey (IHDS) conducted in 2011–2012 and is based on self-reports by households on practice and experience. The sample is restricted to states where there are at least 100 households belonging to the category of scheduled castes. The shaded area shows the 95% confidence interval)
water and irrigation facilities, unequal treatment under the justice system, and discrimination in public streets and market places among others remain widespread and rampant (Barbour et al. 2007; Shah et al. 2006; Acharya 2010; Anderson 2011; Singh 2015; Sharma 2015; Girard 2018). This high prevalence can also be seen in nationally representative data sets; using recent data from the IHDS, Thorat and Joshi (2015) show that almost 27% of households in India self-report engaging in the highly stigmatizing practice of untouchability. Given that untouchability is legally abolished and its practice punishable by law, such high rates of self-reported adherence leads one to suspect that the actual rates of prevalence might be even higher. Using IHDS-2011–2012, Fig. 2 plots the scatter plot and the fitted line showing the self-reported practice of untouchability by households on the x-axis and the proportion of SC households reporting experiencing untouchability on the y-axis. The figure shows that the experience of untouchability is increasing in its practice suggesting that these practices are not confined to the private sphere but result in actual stigmatization of Dalits. Two, the prevalence is extremely high in the northern and central plains of India; the reported self-adherence to the practice exceeds 40% in Bihar, Chattisgarh, Odisha, Rajasthan, Uttrakhand, and Uttar Pradesh and 50% in Himachal Pradesh and Madhya Pradesh.
524
R. Ramachandran
Thus, both endogamy and the enforcement of the caste hierarchy through the rituals of untouchability remain highly prevalent allowing for the social reproduction of caste. Looked at from this perspective, the foundations of the caste system remain deeply implanted in contemporary Indian society.
Human Capital and Health Provision of education and skills remains one of the important paths to equalizing opportunity and consequently enabling socioeconomic mobility for the various caste groups. In what follows, evidence on the extent of educational gaps between groups, and then turn to summarizing evidence on group disparities in health is reviewed.
Education In one of the first comprehensive quantitative examination of caste gaps, Hnatkovska et al. (2012) employ data from five successive rounds of the National Sample Survey (NSS) of India spanning the years of 1983–1984 to 2004–2005. They show that there is convergence between the caste groups – SC-ST and non-SC-ST – in both the average years of schooling, as well as in the education category of secondary or above; in fact, the authors note that the “the sharpest convergence has been in category 5 (secondary or higher).” However, as Deshpande and Ramachandran (2019b) show that the conclusions reached regarding convergence depend on, one, the notion of convergence employed, that is, absolute vs relative. In particular, Hnatkovska et al. (2012) define convergence as the ratio of the indicator of educational attainment under consideration for the two groups; for instance, for the indicator of years of schooling they calculate the relative gap in the years of education as the ratio of non-SC/STs to SC/STs education years. Two, the study by Hnatkovska et al. (2012) undertakes a two-way comparison between SC-STs and non-SC- ST. In other words, it clubs the upper caste Hindus (UC-Hindus) with OBCs and Muslims; as seen in Table 2, the OBCs lie between the UC-Hindus and SC-ST and including them with the former understates the extent of gaps. Table 2 using data from the most recent round of the National Family Health Survey from 2015–16 (NFHS-IV) with a sample size of 467,569 observations shows in Panel A the average years of schooling, and in Panel B the share having at least entered higher education, for the three caste groups for the cohorts aged 45–49, 40–44, 35–39, 30–34, and 25–29 in 2015–2016. We see for the oldest cohort, the average years of schooling for the UC-Hindus, OBCs, and SC-ST was 6.51, 3.84, and 2.89, respectively. On the other hand, for the cohort aged 25–29, the average years of schooling, for the UC-Hindus, OBCs, and SC-ST is 10.19, 7.77, and 6.82, respectively. Columns (4) and (5) of Table 2 show the gaps in years of schooling between the UC-Hindus and OBCs and UC-Hindus and SC-ST, respectively. Looking at the gaps between the UC-Hindus and SC-ST for
22
Caste and Socoieconomic Disparities in India: An Overview
525
Table 2 Years of schooling and share with higher education by caste group and cohort Age in 2016
(1) UCHindus
(2)
(3)
OBCs
SC-ST
Years of Schooling 25–29 30–34 35–39 40–44 45–49
10.19 (0.0702) 8.946 (0.0647) 8.085 (0.0716) 7.235 (0.0755) 6.513 (0.0822)
7.77 (0.0429) 6.483 (0.0429) 5.433 (0.044) 4.722 (0.0431) 3.849 (0.043)
6.829 (0.0471) 5.45 (0.0498) 4.446 (0.0478) 3.616 (0.0488) 2.89 (0.045)
Share Higher Education 25–29 30–34 35–39 40–44 45–49
0.319 (0.00666) 0.226 (0.00574) 0.188 (0.0061) 0.153 (0.00585) 0.128 (0.00592)
0.178 (0.00307) 0.117 (0.00273) 0.0791 (0.00254) 0.0639 (0.00218) 0.0434 (0.00199)
0.132 (0.00312) 0.0821 (0.00306) 0.0629 (0.00266) 0.0493 (0.00242) 0.0412 (0.00203)
(4) (5) UC-HUC-HOBCs SC-ST Panel A Diff in Years of Schooling 2.42 3.361 (0.0802) (0.0823) 2.463 3.496 (0.0738) (0.0785) 2.652 3.639 (0.0839) (0.0829) 2.513 3.619 (0.0846) (0.0903) 2.664 3.623 (0.0919) (0.0919) Total Observations – 467,569 Panel B Diff in Share Higher Education 0.141 0.187 (0.007) (0.007) 0.109 0.1439 (0.0062) (0.0064) 0.1089 0.1251 (0.0067) (0.006) 0.0891 0.1037 (0.0059) (0.0062) 0.0846 0.0868 (0.0062) (0.0061) Total Observations – 467,569
(6) UC-H/ OBCs
(7) UC-H/ SC-ST
Ratio of Years of Schooling 1.31 1.49 1.38
1.64
1.49
1.82
1.53 2 1.69
2.25
Ratio Share Higher Education 1.79 2.42 1.93
2.75
2.38
2.99
2.39
3.1
2.95
3.11
Notes: The tables show the average years of schooling and the share having attained higher education by cohort and social group for the three caste groups – UC-Hindus (UC-H), the OBCs and SC-ST from NFHS-IV. It arises from estimating a regression of the dependent variables: (i) years of schooling; and (ii) a dummy that takes value 1 for having attained higher education and 0 otherwise; on a set of group and cohort dummies and its interaction. It additionally accounts for a rural and gender dummy and clusters the standard errors at the level of primary sampling unit. The standard errors are shown in the parenthesis
the cohorts aged 45–49, 40–44, 35–39, 30–34, and 25–29 shows the gaps are 3.36, 3.50, 3.64, 3.61, and 3.62, respectively. The difference-in-differences estimator comparing the oldest (45–49) to the youngest (25–29) cohort would be equivalent to 0.26 and statistically not different from zero (Duflo 2001). Thus, the standard evaluation methodology based on a difference-in-differences estimator would imply that the gaps remain constant across the UC-Hindus and SC-ST.
526
R. Ramachandran
However, using the notion of relative gaps employed by Hnatkovska et al. (2012), and shown in columns (5) and (6), would imply that the relative gaps between the UC-Hindus and SC-ST and UC-Hindus and OBCs for the cohort aged 45–49 is 2.25 and 1.69, respectively. On the other hand, the relative gaps between the UC-Hindus and SC-ST and UC-Hindus and OBCs for the cohort aged 25–29 is 1.49 and 1.31, respectively. Based on this evidence, Hnatkovska et al. (2012) conclude that there is convergence across caste groups. However, as Deshpande and Ramachandran (2019b) note the predominant mode of measuring returns to education is using the Mincerian wage equation that calculates the return to every additional year of schooling. This would in turn imply that the correct empirical methodology to evaluate convergence is a difference-in-differences estimator, and would invalidate the result of convergence across caste groups (also see Hnatkovska et al. 2012, Table 2, p. 208 which shows exactly the same pattern; further divergence between the SC-ST and non-SC-ST based on the standard difference-in-differences estimator but convergence based on the notion of relative gaps). Comparing the oldest (45–49) and youngest (25–29) cohorts of the OBCs and UC-Hindus implies a DID estimate of 0.24 years; in other words, the OBCs aged 25–29 closed the gap by 0.24 years though the absolute gap remains as large as 2.4 years of schooling. Thus, the picture in regard to education shows one where the gaps have remained largely constant across caste groups. Panel B of Table 2 shows for the category of higher education, the gaps based on a difference-in-differences estimator show more than a doubling of the gaps between UC-Hindus and SC-ST. More specifically, 12.8% of UC-Hindus and 4.12% of SC-ST for the cohort aged 45–49 in 2016 have entered higher education. This increases to 32 and 13.2% for the cohorts aged 25–29 in 2016 for the UC-Hindus and SC-ST, respectively. Thus, the absolute gap increases from 8.7 percentage points to 18.7 percentage points, or in other words, a doubling of gaps. However, looking at the ratios would suggest a decline from 3.11 to 2.42.
Health Recent research documents the existence of disparities in access, as well as outcomes, for the stigmatized caste groups (Nayar et al. 2007; Acharya 2010, 2018); for instance, Acharya (2018) shows that the proportion of SC and upper castes mother’s whose pregnancies were registered are equivalent (≈86%): however, 77% and 55% of SC mothers receive antenatal care from a skilled service provider and doctors, respectively, the corresponding figures for the upper castes are 86% and 70%, respectively. In terms of place of delivery, around 45% of SC women still give birth to children at home, whereas only 23% of upper caste Hindus mothers do so. These differences in access are reflected in a variety of outcomes for both children as well as mothers. SC-ST and OBC children are seen to suffer from higher rates of under-5, infant, neonatal, child, and post neonatal mortality (Acharya 2018; Bora et al. 2019). Ramachandran and Deshpande (2021) in recent work show whereas
22
Caste and Socoieconomic Disparities in India: An Overview
527
26% of UC-Hindu children under-5 were chronically malnourished in 2015–2016, the corresponding figures for SC-ST and OBC children is 40% and 36%, respectively. Moreover, more than 60% of SC children under 5 are seen to be anemic (Acharya 2018, p. 124). SC-ST mothers are also seen to suffer from worse anthropometric outcomes, greater prevalence of anemia, and higher maternal mortality rates (Balarajan et al. 2013; Horwood et al. 2020; Ramachandran and Deshpande 2021).
Economic Outcomes In this section, evidence on the gaps between groups on various material indicators of well-being is reviewed, namely, consumption, income, occupation, poverty, wages, and wealth.
Consumption, Income, and Wages The IHDS provides data on both consumption and income from a panel of ≈42,000 households interviewed in 2004–2005 and 2011–2012. Panel A and C of Fig. 3 shows that the per capita consumption expenditure, and per capita income, for the year 2011–2012 for households belonging to UC-Hindus, OBCs, and SC-ST is Rs. 33,044; 24,180; and 18,615 and Rs. 37,645; 22,508; and 19,275, respectively. Thus, the gaps in both consumption and income across the caste groups remain sizeable; for instance, the consumption and income of the UC-Hindus is 77% and 95% higher that of the SC-ST households. Moreover, as the IHDS is a house-hold level panel, it allows us to examine how gaps in consumption and income evolve across the caste groups within the same households, in other words, accounting for household fixed effects. Panel B and Panel D which show the difference-in-differences (DID) estimates for the indicator of per capita household consumption expenditure shows that the OBCs and SC-ST fall behind by Rs. 9210 and 14,330, respectively, over the 7 years the two rounds span. Panel D, in turn, shows the DID estimates for the indicator of per capita household income; again the OBCs and SC-ST fall behind by Rs. 4962 and Rs. 5333 over the two rounds. In relation to wages, Hnatkovska et al. (2012) show that the median wage premium of non-SC/STs relative to SC/STs has declined from around 36% in 1983 to 21% in 20,042,005. However, as noted before, they do not distinguish between the UC-Hindus and OBCs. Accounting for this distinction, Deshpande and Ramachandran (2019b) show that wages between SC-ST (OBCs) and upper caste Hindus converge for only the below median wage earners, whereas for wage earner above the log median daily wage they find neither convergence nor divergence between the OBCs and UC-Hindus and a widening of wage gaps between SC-ST and UC-Hindus.
528
R. Ramachandran
Panel A - Per Capita Consumption Expenditure
Panel B - Difference-in-Differences Per Capita Consumption Expenditure -8000.00 -10000.00 -12000.00 -14000.00
-14330.34
15000.00
18615.57
-9210.48
-16000.00
30000.00 20000.00
25000.00
24180.66
Mean per capita HH expenditure
35000.00
Mean per capita HH expenditure
33044.21
UC-Hindus
OBCs
SC-ST
OBCs # year=2012 SC-ST # year=2012
Panel D - Difference-in-Differences Per Capita Income
-4000.00
-2000.00
0.00
19275.80
-5333.64
-6000.00
-4962.83
-10000.00
-8000.00
22508.70
Mean per capita HH income
25000.00
30000.00
35000.00
37645.03
20000.00
Mean per capita HH income
40000.00
Panel C - Per Capita Income
UC-Hindus
OBCs
SC-ST
OBCs # year=2012 SC-ST # year=2012
Fig. 3 Per capita household consumption and income and their evolution by caste groups. (Notes: The figure in Panels A and C plots the predicted per capita consumption expenditure and per capita income, and its 95% confident intervals, arising from a regression of the same on a set of group dummies with standard errors clustered at the primary sampling unit level from the IHDS-II. Panels B and B plot the difference-in-differences estimates arising from regressing per capita consumption expenditure and per capita income, respectively, on interaction of the caste dummy with the round dummy, with household fixed effects and standard errors clustered at the primary sampling unit level from using a panel of households from the two rounds of IHDS)
Occupation Deshpande and Ramachandran (2019b) provide an overview of the occupational distribution and its changes for individuals aged 16–65 in 2011–2012. Table 3 drawn from the NSS reproduces the distribution across three broad occupation categories: agriculture, blue-collar jobs, and white-collar jobs.7 As Table 3 shows, and Deshpande and Ramachandran (2019b) discuss, the percentage of people employed in agricultural, blue-collar, and white collar jobs changes from 59.15 to 44.68, 27.79
7
This is based on a subjective classification following Hnatkovska et al. (2012). The category of agricultural jobs includes workers such as farmers, fishermen, loggers, hunters, etc.; blue collar jobs includes workers such as sales workers, service workers, and production workers; and white collar jobs includes people employed as administrators, executives, managers, professionals, and technical and clerical workers.
22
Caste and Socoieconomic Disparities in India: An Overview
529
Table 3 Percentage share in different occupational categories by cohort and social group Social Group
Agricultural jobs
SC-ST OBCs Others Total
63.85 62.04 48.2 59.15
SC-ST OBCs Others Total
57.2 54.26 38.1 50.88
SC-ST OBCs Others Total
53.81 50.68 36.42 48.11
SC-ST OBCs Others Total
50.07 47.44 32.46 44.68
Blue collar jobs Aged 56–65 in 2012 30.35 27.5 25.27 27.79 Aged 46–55 in 2012 35 7 32.36 33.1 33.34 Aged 36–45 in 2012 38.4 36.39 34.53 36.55 Aged 26–35 in 2012 41.6 39.06 38.84 39.83
White collar jobs 5.8 10.46 26.53 13.06
100 100 100 100
81 13.38 28.79 15.78
100 100 100 100
7.8 12.93 29.05 15.35
100 100 100 100
8.33 13.5 28.69 15.5
100 100 100 100
Notes: The tables show the percentage of people from each caste and cohort employed in different occupations from the National Sample Survey 2011–12(NSS-66). This is based on a subjective classification following Hnatkovska et al. (2012). The category of agricultural jobs includes workers such as farmers, fishermen, loggers, hunters etc.; blue collar jobs includes workers such as sales workers, service workers and production workers; and white collar jobs includes people employed as administrators, executives, managers, professionals, technical and clerical workers
to 39.83, and 13.06 to 15.5, respectively, for the cohort aged 55–65 to those aged 26–35 in 2012. Overall group differences in occupational distribution across the three categories have remained largely static. In fact, for the most prestigious occupational category, white collar jobs, the traditional hierarchies in occupational structures remaining largely unchanged since independence, irrespective of whether the notion of absolute or relative convergence is employed. The results are in sharp contrast with the work of Hnatkovska et al. (2012) who report convergence in the category of white-collar jobs; the difference arises as Hnatkovska et al. (2012) club the OBCs with the UC-Hindus and thereby understate the true extent of differences across the caste groups.
Poverty Though poverty remains a critical challenge, official estimates show a decline in the incidence from a high of 37% in 200,405 to 22% in 201,112, that is, a decline of
530
R. Ramachandran
15 percentage points.8 The decline notwithstanding there remain large differences in the incidence of poverty across caste groups in India. Using a four-way classification of caste groups, Thorat et al. (2017) employing data from the IHDS show that the incidence of poverty among the forward castes, OBCs, SC, and ST in 2004–2005 was 26, 38, 47, and 65%, respectively. This declines in 2011–2012 to 14, 20, 27, and 42% among the forward castes, OBCs, SC, and ST, respectively. The IHDS is a panel data set and also allows for the estimation of the dynamics of poverty. Thorat et al. (2017) find that the SC and ST are only 64 and 40% as likely as the forward castes to escape from poverty, and 2.2 and 2.6 times more likely to fall into poverty. Thus, the data suggest that not only is the incidence of poverty more than twice (thrice) as high among the SC (ST) compared to the forward castes, but in addition they are also more vulnerable to fall into poverty and less likely to be able to emerge from it. Moreover, as Deshpande and Ramachandran (2019a) show that even among the individuals classified as poor, the income of the poor SCs is only 73% of the poor upper castes. The disparities in terms of land ownership or cultivation are even starker: in 2012, of all those who report being involved in agriculture, 75% of the upper castes owned or cultivated land whereas the comparable figure for the SCs is 48%.
Wealth There are few estimates of the extent of disparities in wealth across caste groups. The two most important categories of wealth are land and buildings, where land and buildings together form 90% of the total household wealth, with the share of land being 56.36% declining from a high of 60.06% in 1981 (Bharti 2018). Zacharias and Vakulabharanam (2011) using the All India Debt and Investment Survey (AIDIS), conducted in 1991–1992 and 2002–2003, show that the SC-ST have substantially lower wealth than the UC-Hindus, while the OBCs occupy a position in the middle. The most comprehensive estimates from Bharti (2018) show at an all-India level, top 10% of the population had 45% of the total wealth in 1981 which increases to 58% in 2011, an increase of 15 percentage points in 30 years. In relation to caste disparities he estimates the population share of a given caste group in a given wealth decile compared to its share in overall population, also known as representational inequality (Jayadev and Reddy 2019). There is seen to be significant inequality in wealth across caste groups with the upper castes relatively more present in top 10% and middle 40% of the population where almost 90% of the wealth is concentrated. SC and ST have relatively more presence in the bottom 50% of the wealth distribution, whereas the OBC population is more evenly distributed across the wealth deciles. Looking at the relative wealth shares by caste groups, Bharti (2018) finds that in 2012 in the rural areas the upper castes in the top decile of wealth have a share of 32% whereas their population share is only 17.6%. The wealth share of OBCs is
8
These are based on the NSS estimates from the rounds collected in 2004–2005 and 2011–2012.
22
Caste and Socoieconomic Disparities in India: An Overview
531
45.9% and is slightly more than their population share in the survey (which is 44.2%). The SC in contrast are over represented in the bottom 50% of the wealth distribution and in the top 50% their share is around 40% points lower than their population share. In the urban areas there is even greater concentration of wealth among the upper castes. For the 2012 survey, in the top 10% of the wealth distribution, 54% are from upper castes and 29% are from OBCs. On the other hand, for the SC and ST, their share of wealth in the top 5 deciles is 24.2 and 4.8% points, respectively, lower than their population share. In conclusion, the SCs have the lowest wealth ownership in both urban and rural areas whereas the upper castes are over represented in the top decile or the top five deciles of the wealth distribution.
Affirmative Action Given the large disparities between caste groups at independence, Article 46 of the Indian Constitution put into place one of the most expansive affirmative action programs in the world. In particular, in the sphere of higher education and jobs, commensurate to their share in the population, 15% and 7.5% of the seats in government universities and public sector jobs are reserved for the SCs and the STs, respectively. In 1993, as a result of the acceptance of the Mandal Commission recommendations, affirmative action was extended to the OBCs reserving 27% of public sector jobs at the central government level. In 2006, this was further extended into the realm of education with 27% of seats in government universities being reserved for the OBCs (Deshpande 2013). Affirmative action extends beyond the realm of education and public sector jobs and extends into the political domain through the use of electoral quotas. Since 1950, SC-STs have been entitled to reserved seats in the Indian Parliament and state assemblies in proportion to their share of the population in each Indian state. In addition, since the early 1990s, SCs have also had reserved positions in village councils across India. I below summarize the key findings on the impact of affirmative action in the various realms.
Education and Public Sector Employment The implementation of affirmative action in higher education for all SC-ST has meant that estimating the causal impacts of affirmative action programs on the average individual has typically not been possible. Most studies involve estimating the impact on the marginal recipient. Bertrand et al. (2010) examine the impact of the affirmative action program for SC-ST and OBCs in engineering colleges in one state in India. Employing a regression continuity design that exploits the minimum score required for admission and its variation across groups they find that, one, caste-based targeting results in the redistribution of education resources to those who are more economically disadvantaged. Two, beneficiaries of affirmative action economically
532
R. Ramachandran
gain from attending engineering college. However, they find that affirmative action programs come at an absolute cost; the negative income level effect experienced by displaced upper-caste applicants is larger than the positive income level effect experienced by displacing lower-caste students.9 Bagde et al. (2016) also examine the effects of affirmative action in more than 200 private nonprofit engineering colleges from one large state in India. However, they evaluate effects not only on the marginal candidate but on admitted students over the entire ability distribution. They find that affirmative action increases college attendance of targeted students, particularly at relatively higher-quality institutions. Moreover, they find no evidence of mismatch: higher college quality has a positive effect on achievement for affirmative action beneficiaries on a comprehensive test taken after the first year of college. Moreover, the program does not adversely affect average graduation rates for the disadvantaged-caste students. Cassan (2019) exploits within Jati variation in access to affirmative action arising from the linguistic reorganization of state borders to estimate the effect of affirmative action on educational attainment. He finds positive affect on schooling attainment, in particular literacy and secondary schooling but not in the realm of higher education. However, there is a stark gender divide with only males benefiting from the policy of affirmative action, whereas females do not shown any gains. Deshpande and Ramachandran (2019b) examine the impact of reserving seats in central government jobs from 1993 onward for the OBCs. They employ a doubledifference estimator that leverages variation on two fronts – age and social group – and estimates the average effects of quotas. They find that the provision of quotas increases the share of OBCs in government jobs by around four percentage points. They also examine the incentive effects of job quotas on educational attainment. Government jobs require candidates to have at least completed secondary school education (Class X) in order to be eligible for even the lowest paying jobs. Consistent with this reasoning of incentives presented by job quotas, they find that job quotas resulted in an increase in the probability of completing secondary education by 5 percentage points for the youngest cohort of the OBCs. Examining the same policy change, Khanna (2020) confirms the findings of Deshpande and Ramachandran (2019b). He finds that affirmative action incentivizes about 0.8 additional years of education for the average OBC. Khanna (2020) also additionally examines the effect of the policy on the marginal recipient, as well as on how the impacts are moderated by the intensity of reservation. He finds that the marginal OBC student gains 1.2 years of schooling and that the beneficial effects are concave in the intensity of reservation.
It is important to note that the authors find that students from the reserved category are more likely to take up public sector jobs. Public sector jobs have lower incomes but often come up with a host of other benefits – subsidized housing, health care and pensions. Monetizing these benefits might suggest that the absolute cost is overestimated. Moreover, due to small sample sizes and their evidence being from engineering colleges in one state in India, their validity to other fields and parts of India should be extended with caution.
9
22
Caste and Socoieconomic Disparities in India: An Overview
533
One of the contentious issues surrounding reservation of public sector jobs is the criticism that such policies conflict with considerations of merit. As less qualified candidates are selected in place of more qualified candidates, it is often contended that poorer quality of work on the job results from such policies (Guha 1990; Shah 1991) though empirical evidence has been largely missing on this issue. Deshpande and Weisskopf (2014) examine the impact of reservation in Indian Railways on total factor productivity (TFP) using panel data. They find that increase in the proportion of jobs filled by SC-STs are not associated with lower factor productivity either in levels or in changes.10 In recent work, Bhavnani and Lee (2019) examine the bureaucratic performance of affirmative action recruits to the civil service measured using three outcomes: the number of households that received the guaranteed 100 or more days of employment under the national rural employment guarantee scheme; the log number of villages connected under the governments premier road building program; and the time taken by bureaucrats to approve projects proposed by national legislators. Their empirical strategy is based on a rich district level panel data that accounts for district and state-year level fixed effects. They combine this with an instrumental variables estimator that leverages the fact that bureaucrats early in their careers are quasi-randomly assigned to districts. They find disadvantaged group members recruited via affirmative action do not perform any worse than the bureaucrats who obtain their position on merit. These two studies suggest that improvements in diversity can be obtained without efficiency losses and that the critics fear that affirmative action inevitably worsen on the job performance might be overstated.
Politics One of the first studies to examine the effects of mandated political representation in providing disadvantaged groups influence over policy-making is the work of Pande (2003). She leverages the fact that the Indian constitution requires that the extent of state-level political reservation enjoyed by a group reflect the group’s population share in the state. However, as the extent of political reservation is only revised after a new census whereas a group’s population share varies continuously, the proportion of jurisdictions reserved for it changes with a lag. Exploiting this institutional feature, and examining outcomes at the state level, she finds that increase in SC and ST reservation do not affect the likelihood of land reform and has a significant negative impact on education spending. Interestingly, whereas increasing SC reservation increases job quotas but does not affect the total spending in a state, she documents the reverse effects for ST reservations. Using the same empirical strategy, Chin and Prakash (2011) examine the effect of mandated political reservation on state level poverty. They find that increasing the share of seats reserved for
In fact, under some specifications, the authors find that higher proportions of SC-ST employees in high-level positions are positively associated with higher TFP.
10
534
R. Ramachandran
Scheduled Tribes significantly reduces poverty while increasing the share of seats reserved for Scheduled Castes has no impact on poverty. Jensenius (2015) also examines the effects of mandated state-level political reservation. However, instead of examining the effects at the state level she examines the effects at the constituency level; the constituency is the geographical unit at which the seat is reserved. She compares constituencies that were reserved for 30 years, with those that were not reserved, on a wide variety of development outcomes – literacy, employment rate, share of agricultural laborers, as well as four village-level variables: whether village has electricity, school, medical facility, or a communication channel in village – using a matched-pair design. She finds no detectable constituency-level effect of reservation on overall development or redistribution to SCs, neither on the literacy rates or employment patterns of SCs or non-SCs, nor on village amenities in reserved constituencies. Since the 73rd amendment of the Indian constitution there is also reservation of seats at the lowest level of government, that is the head (pradhan) of the elected village councils, also known as the Gram Panchayat. The mandated political reservation in favor of SC/ST for the Pradhan position should reflect the SC/ST population share in that state. Moreover, the amendment also required that no GP be reserved for the same group for two consecutive elections Besley et al. (2004) exploiting this institutional feature look at its implications in three South Indian states. They find that SC-ST households in reserved GPs are 6% more likely to receive targeted public goods with low spillovers. Dunning and Nilekani (2013) also examine the reservation of seats for the head of village councils employing a variant of the regression-discontinuity (RD) to create two sets of councils that plausibly differ only in the presence or absence of quotas mandating village council presidents from marginalized castes or tribes. They, however, find that SC-ST based quotas for village council presidents have quite weak policy and distributive effects with no discernible effects on council spending on programs targeted toward those groups. Examining reasons for the failure of political reservations to affect developmental outcomes, Dunning and Nilekani (2013) and Jensenius (2015) highlight the role of political parties that remain gatekeepers for choice of which candidates to field. The parties are typically controlled by non-SC leaders, and SC politicians who do not toe the party line are not afforded opportunity to run for office (Chhibber 2001). Two, the design of the quotas is such that in SC reserved constituencies, SCs are almost always in minority and need to win non-SC votes to win elections. The non-SC forming the majority of the electorate might imply that party gatekeepers to win elections instruct SC politicians to give the majority group precedence over their own group with regards to policy formulation. Finally, it is important to note that beyond the material aspects of political representation, recent work shows that political representation for subaltern groups at the village level increases the perception that members of dominant groups are more accepting towards stigmatized caste groups and that hostile behaviors against members of the SCs are more likely to be sanctioned (Chauchard 2014). They are also seen to result in more favorable appraisals of SCs as politicians (Jensenius 2013).
22
Caste and Socoieconomic Disparities in India: An Overview
535
Regional Patterns of Caste Disparities The evidence on human capital and health as well as economic outcomes show there remain sharp disparities across the caste groups which have remained largely static or even increased in some instances (Deshpande and Ramachandran 2019b). However, it does not shed light on the extent of regional disparities along caste lines. Surprisingly there have not been many studies that examine the regional variations in caste gaps and the reasons underlying the same. One of the early comparisons is by Deshpande (2001) who compares SC-ST to the non-SC-ST using data from 1992 and finds that the northern states of Punjab and Haryana display the highest levels of caste disparity, whereas the southern states of Kerala and Karnataka display the lowest levels of disparities. In this section, an updated picture of the extent of regional disparities among caste groups based on the indicator of the average years of schooling is presented. Two of the primary motivations for looking at educational attainment are, one, it is one of the key capabilities that can enable individuals to improve their socioeconomic standing in society (Sen 1987, 1990), and, two, as Hnatkovska et al. (2012) highlight that improvements in education as underlying a major part of the wage and consumption convergence they find when comparing SC-ST to non-SC-ST. With district level averages of the years of schooling by caste group – UC-Hindus, OBCs, and SC-ST – six categories of average years of schooling are constructed, namely: (i) 0–2; (ii) 2–4; (iii) 4–6; (iv) 6–8; (v) 8–10; and (vi) greater than 10.11 Before turning to regional differences, the distribution for the three groups is shown in Table 4. It shows that only in 5 districts (0.87%) do the SC-ST average more than 10 years of schooling. The commensurate figure for the UC-Hindus and OBCs is 123 (or 27.09%) and 27 (or 5.11%) districts, respectively. On the other hand, in 38% of the districts the SC-ST have less than equal to 6 years of schooling. The commensurate figure for the UC-Hindus and OBCs is 2.86 and 29.36%, respectively. Turning to the regional patterns, Fig. 4 plots a color coded map showing the spatial distribution of the average years of schooling by district and caste group, where the thin black lines depict the district boundaries and the thick black lines the state boundaries. In terms of regional patterns, the graph shows a very clear pattern with the areas of low educational attainment for SC-ST concentrated in the states of Bihar, Chhattisgarh, Jharkhand, Madhya Pradesh, Rajasthan, and Uttar Pradesh, as can be seen by the northern and central plains being largely orange and red in color. These states are also referred to as BIMARU, an acronym formed from the first letters of the names of the Indian states of Bihar, Madhya Pradesh, Rajasthan, and Uttar Pradesh. This literally translates into “sick” in Hindi and is the part of the
11 Only districts where there are a minimum of 50 observations. On an average 177, 430 and 467 observations per district for the UC-Hindus, OBCs and SC-ST, respectively.
536
R. Ramachandran
Table 4 Prevalence rates of stunting by category and districts of India Yrs. of Education Less than equal to 2 Greater than 2 and less than equal to 4 Greater than 4 and less than equal to 6 Greater than 6 and less than equal to 8 Greater than 8 and less than equal to 10 Greater than 10 Total Less than equal to 2 Greater than 2 and less than equal to 4 Greater than 4 and less than equal to 6 Greater than 6 and less than equal to 8 Greater than 8 and less than equal to 10 Greater than 10 Total Less than equal to 2 Greater than 2 and less than equal to 4 Greater than 4 and less than equal to 6 Greater than 6 and less than equal to 8 Greater than 8 and less than equal to 10 Greater than 10 Total
No. of Districts Panel A – UC-Hindu 0 0 13 120 198 123 454 Panel B – OBCs
Percent
30 155 204 112 27 528 Panel C – SC-ST 5 101 218 185 61 5 575
5.68 29.36 38.64 21.21 5.11 100
0 0 2.86 26.43 43.61 27.09 100
0.87 17.57 37.91 32.17 10.61 0.87 100
Notes: The tables show the percentage of people from each caste and cohort employed in different occupations from the Nation Sample Survey 2011–12 (NSS-66). This is based on a subjective classification following Hnatkovska et al. (2012). The category of agricultural jobs includes workers such as farmers, fishermen, loggers, hunters etc.; blue collar jobs includes workers such as sales workers, service workers and production workers; and white collar jobs includes people employed as administrators, executives, managers, professionals, technical and clerical workers
country that has been traditionally the most socioeconomically backward (Bose 2000).12 In the state of Bihar the average years of schooling for the SC-ST in the median district is 3.06, and for Uttar Pradesh, Madhya Pradesh, and Rajasthan it is 5.19, 4.26, and 4.14, respectively. In contrast, for the UC-Hindus (OBCs), the average years of schooling in the median district of Bihar, Uttar Pradesh, Madhya Pradesh, and Rajasthan are 8.11 (4.47), 9.54 (6.06), 8.74 (5.48), and 8.05 (5.03), respectively. The state of Kerala presents a scenario at the other end of the spectrum where the
12
The present-day states of Chhattisgarh, Jharkhand, and Uttarakhand were part of Madhya Pradesh, Bihar, and Uttar Pradesh, respectively, at the time the BIMARU acronym was coined and we thus include them in the category of BIMARU.
22
Caste and Socoieconomic Disparities in India: An Overview
537
Fig. 4 The average years of schooling by caste and district. (Notes: The figure plots the average years of schooling by district for each caste group based on the NFHS-IV. The thin black lines depict the district boundaries and the thick black lines the state boundaries. Averages are shown only for districts where there are minimum of 50 respondents from the caste group)
average years of schooling for the SC-ST, OBCs, and UC-Hindus in the median district is 10.11, 11.06, and 12.37, respectively. In fact in no district of the BIMARU states do the SC-ST average more than 10 years of schooling and in 85% of the districts have less than equal to 6 years of schooling. On the other hand, in the southern13 and northern regions14 of the country in only 31% and 27% of the districts, respectively, do they have less than equal to 6 years of schooling. In contrast, the UC-Hindus average less than equal to 6 years of schooling in only 4.24% of the BIMARU districts, whereas the commensurate figure for the OBCs is 59%. Figure 5 plots a color code map that shows the gaps on the average years of schooling for the OBCs and the SC-ST compared to the UC-Hindus. The lightest shaded areas exhibit the lowest gaps and the darker shaded areas the highest gaps. It shows that the SC-ST and OBC have higher educational attainment than UC-Hindus in only 2 and 11.54% of the districts, respectively. Moreover, the gaps are largest in the same parts of the country as in which the SC-ST have the lowest levels of human capital. In other words, the north and central plains of India show both the highest levels of deprivation, as well as extent of caste-based disparities in the country. In none of the districts in the BIMARU states do the SC-ST have higher years of schooling than the UC-Hindus, and in fact, 158 (74%) and 17 districts (8%) have a gap greater than 3 and 6 years of schooling, respectively, as can be seen by orangered being the dominant color in this region. The regional picture thus indicates that public policy specifically targeted at the BIMARU region with special considerations for caste identity might be necessary to tackle the extreme level of caste disparities and gaps present in this region.
13
The states included in the southern region are Andhra Pradesh, Karnataka, Kerala, Tamil Nadu, and Telangana. 14 The states included in the northern region include Chandigarh, Delhi, Haryana, Himachal Pradesh, Jammu & Kashmir, and Punjab.
538
R. Ramachandran
Fig. 5 Gaps in the average years of schooling by caste and district (Notes: The figure plots the gaps in the average years of schooling by district when comparing the UC-Hindus with OBCs and SC-ST and is based on data from the NFHS-IV. The thin black lines depict the district boundaries and the thick black lines the state boundaries. Averages are shown only for districts where there are minimum of 50 respondents from both caste groups being compared)
Conclusion Caste based affirmative action was implemented in India since 1950 to help overcome the multitude of arenas in which the lower ranked caste groups were socially excluded and stigmatized (see Table 1). However, the evidence presented in section “Social Reproduction of the Caste System” shows that discriminatory practices, and in particular the stigmatizing practice of untouchability and exclusion from accessing key public amenities, as well as the practice of endogamy to maintain group boundaries, remain widespread in contemporary India. This in turn suggest that analyzing caste not purely as accumulated historical disadvantage but as a current socioeconomic system that actively marginalizes certain sections of society might help better address the extent of caste based disparities present in society today. On the one hand, the results on the impacts of affirmative action policies reviewed in section “Economic Outcomes” show that they have been instrumental in improving human capital outcomes, provided access to better jobs and also improved perceptions of Dalits as public office holders. On the other, the evidence on gaps in the domains of human capital, health, consumption, income, occupation, poverty, and wealth show that not only do substantial gaps remain but more worryingly
22
Caste and Socoieconomic Disparities in India: An Overview
539
they have remained largely static over the last 70 years. The evidence suggests that interventions that promote acquisition of human capital and health from early childhood with a special emphasis on certain parts of the country might be essential to address the problem of caste-based inequities in India.
Cross-References ▶ Caste and Gender ▶ Caste Quotas in India ▶ Does Political Affirmative Action Work, and for Whom? Theory and Evidence on India’s Scheduled Areas ▶ Experimental Evidence on Affirmative Action ▶ Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities? ▶ Stratification Economics
References Acharya S (2010) Caste and patterns of discrimination in rural public health care services. In: Blocked by caste: economic discrimination in modern India. Oxford University Press, New Delhi, pp 208–229 Acharya SS (2018) Health equity in India: an examination through the lens of social exclusion. J Soc Incl Stud 4(1):104–130 Ambedkar BR (2014) Annihilation of caste: the annotated critical edition. Verso Books, New Delhi, India Anderson S (2011) Caste as an impediment to trade. Am Econ J Appl Econ 3(1):239–263 Bagde S, Epple D, Taylor L (2016) Does affirmative action work? Caste, gender, college quality, and academic success in India. Am Econ Rev 106(6):1495–1521 Balarajan YS, Fawzi WW, Subramanian S (2013) Changing patterns of social inequalities in anaemia among women in India: cross-sectional study using nationally representative data. BMJ Open 3(3):e002233 Barbour S, Palikovic T, Shah J, Narula S (2007) Hidden apartheid: caste discrimination against India’s “untouchables”. Shadow report to the UN committee on the elimination of racial discrimination. The Center for Human Rights and Global Justice and the New York. University School of Law, New York Bertrand M, Hanna R, Mullainathan S (2010) Affirmative action in education: evidence from engineering college admissions in India. J Public Econ 94(1):16–29 Besley T, Pande R, Rahman L, Rao V (2004) The politics of public good provision: evidence from Indian local governments. J Eur Econ Assoc 2(2–3):416–426 Bharti NK (2018) Wealth inequality, class and caste in India, 1951–2012. PhD thesis, Masters thesis, Paris School of Economics Bhavnani RR, Lee A (2019) Does affirmative action worsen bureaucratic performance? Evidence from the Indian administrative service. Am J Polit Sci 65(1):5–20 Bidner C, Eswaran M (2015) A gender-based theory of the origin of the caste system of India. J Dev Econ 114:142–158 Bora JK, Raushan R, Lutz W (2019) The persistent influence of caste on under-five mortality: factors that explain the caste-based gap in high focus Indian states. PLoS One 14(8):e0211086
540
R. Ramachandran
Bose A (2000) North-south divide in India’s demographic scene. Econ Polit Wkly 35(20): 1698–1700 Cassan G (2019) Affirmative action, education and gender: evidence from India. J Dev Econ 136: 51–70 Chauchard S (2014) Can descriptive representation change beliefs about a stigmatized group? Evidence from rural India. Am Polit Sci Rev 108(2):403–422 Chhibber PK (2001) Democracy without associations: transformation of the party system and social cleavages in India. University of Michigan Press, Ann Arbor, Michigan Chin A, Prakash N (2011) The redistributive effects of political reservation for minorities: evidence from India. J Dev Econ 96(2):265–277 Deshpande A (2001) Caste at birth? Redefining disparity in India. Rev Dev Econ 5(1):130–144 Deshpande A (2011) The grammar of caste: economic discrimination in contemporary India. Oxford University Press, Oxford, England Deshpande A (2013) Affirmative action in India. Oxford University Press, Oxford, England Deshpande A, Ramachandran R (2019a) The 10% quota: is caste still an indicator of backwardness? Econ Polit Wkly 30:27–31 Deshpande A, Ramachandran R (2019b) Traditional hierarchies and affirmative action in a globalizing economy: evidence from India. World Dev 118:63–78 Deshpande A, Weisskopf TE (2014) Does affirmative action reduce productivity? A case study of the Indian railways. World Dev 64:169–180 Duflo E (2001) Schooling and labor market consequences of school construction in Indonesia: evidence from an unusual policy experiment. Am Econ Rev 91(4):795–813 Dumont L (1970) Homo Hierachius: the caste system and its implications. Weidenfeld & Nicolson, London, England Dunning T, Nilekani J (2013) Ethnic quotas and political mobilization: caste, parties, and distribution in Indian village councils. Am Polit Sci Rev 107(1):35–56 Gait E (1913) Census of India, 1911. Vol. I: India. Part I: report. Superintendent Government Printing, Calcutta Girard V (2018) Dont touch my road. Evidence from India on affirmative action and everyday discrimination. World Dev 103:1–13 Guha A (1990) Reservations in myth and reality. Econ Polit Wkly 25(50):2716–2718 Gupta D (2000) Interrogating caste: understanding hierarchy and difference in Indian society. Penguin Books India, Grurgram, India Hnatkovska V, Lahiri A, Paul S (2012) Castes and labor mobility. Am Econ J Appl Econ 4(2): 274–307 Horwood G, Opondo C, Choudhury SS, Rani A, Nair M (2020) Risk factors for maternal mortality among 1.9 million women in nine empowered action group states in India: secondary analysis of annual health survey data. BMJ Open 10(8):e038910 Jayadev A, Reddy SG (2019) Inequalities and identities. In: Dasgupta I, Mitra M (eds) Deprivation, inequality and polarization. Springer, New York, USA pp 109–145 Jensenius FR (2013) Power, performance and bias: evaluating the electoral quotas for scheduled castes in India. PhD thesis, UC Berkeley Jensenius FR (2015) Development from representation? A study of quotas for the scheduled castes in India. Am Econ J Appl Econ 7(3):196–220 Khanna G (2020) Does affirmative action incentivize schooling? Evidence from India. Rev Econ Stat 102(2):219–233 Kothari R (1997) Caste and modern politics. In: Kaviraj S (ed) Politics in India. Oxford University Press, New Delhi, pp 57–70 Morris MD (1965) The emergence of an industrial labor force in India: a study of the Bombay cotton mills, 1854–1947. University of California Press, Berkeley, USA Mosse D (2018) Caste and development: contemporary perspectives on a structure of discrimination and advantage. World Dev 110:422–436 Munshi K (2019) Caste and the Indian economy. J Econ Lit 57(4):781–834
22
Caste and Socoieconomic Disparities in India: An Overview
541
Nayar K et al (2007) Social exclusion, caste & health: a review based on the social determinants framework. Indian J Med Res 126(4):355 Olivelle, P (1998) Caste and purity: A study in the language of the Dharma literature. Contributions to Indian Sociology 32(2):189–216 Olivelle P (2005) Ix. Caste and purity: a study in the language of the dharma literature. IX. Caste and purity, 1000–1029 Pande R (2003) Can mandated political representation increase policy influence for disadvantaged minorities? Theory and evidence from India. Am Econ Rev 93(4):1132–1151 Ramachandran R, Deshpande A (2021) The impact of caste: a missing link in the literature on stunting in India. IZA DP No. 14173 Ray T, Chaudhuri AR, Sahai K (2020) Whose education matters? An analysis of inter caste marriages in India. J Econ Behav Organ 176:619–633 Sen AK (1987) Commodities and capabilities: Professor Dr. P. Hennipman lectures in economics, 1982 delivered at the University of Amsterdam. Oxford University Press, Oxford, England Sen A (1990) Development as capability expansion. In: The community development reader. pp 41–58 Sen A (2003) Development as Capability Expansion. In: Fukuda-Parr S, et al Readings in Human Development. New Delhi and New York: Oxford University Press Shah A (1991) Job reservations and efficiency. Econ Polit Wkly 26(29):1732–1734 Shah G, Mander H, Thorat S, Baviskar A, Deshpande S (2006) Untouchability in rural India. Sage, New Delhi Sharma S (2015) Caste-based crimes and economic status: evidence from India. J Comp Econ 43(1):204–226 Singh S (2015) Residential segregation and access to basic amenities: a village-level case study. Agrarian Stud 5(2):126–146. (2369-2021-033) Thorat A, Joshi O (2015) The continuing practice of untouchability in India: patterns and mitigating influences. India Human Development Study working paper 3 Thorat S, Neuman KS (2012) Blocked by caste: economic discrimination in modern India. Oxford University Press, New Delhi Thorat A, Vanneman R, Desai S, Dubey A (2017) Escaping and falling into poverty in India today. World Dev 93:413–426 Witzel M (2019) Which of us are Aryans?: rethinking the concept of our origins. Aleph, Delhi, India Zacharias A, Vakulabharanam V (2011) Caste stratification and wealth inequality in India. World Dev 39(10):1820–1833
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
23
Christophe Jaffrelot and Kalaiyarasan A.
Contents Trajectories of Socioeconomic and Educational Marginalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income: When Muslims Come Last . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Education: The Widening Gap Between Muslims and Hindus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jobs: When the Informal Sector Prevails . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Youth in Higher Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muslims in Government and Public Sector Jobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Politics and Policies of Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Timidity and Bias at the Center . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muslims in Dravidian Land . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Northern and Eastern India: Failed Attempts and the Delegitimization of Pro-Muslim Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muslims as Collateral Casualties of the Modi Government’s Style of Empowerment . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
545 545 549 550 552 554 554 555 557 560 561 561 562 563
Abstract
Social inequalities are found to have risen in India. Religion is one such axis of inequality. Muslims – the largest minority in India – is its first victim. The Sachar Committee Report (SCR) submitted in 2006 showed that Muslims were on the margins in terms of political, economic, and social indicators and that their The authors are grateful to the Henry Luce Foundation whose funding has made this research possible, as part of the “Indian Muslims Today” project. C. Jaffrelot (*) CERI-Sciences Po/CNRS, Paris, France King’s India Institute, King’s College, London, UK e-mail: [email protected] K. A. Brown University, Providence, RI, USA © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_26
543
544
C. Jaffrelot and K. A.
average condition was comparable to, or even worse than, the country’s backward communities including Dalits in certain indicators. The condition of Muslims since then has only worsened in the context of Hindu majoritarianism. However, Muslims are differentiated by region. They still fare better in development indicators in South India vis-à-vis the rest of India. This is partly due to the implementation of affirmative action schemes for Muslims in the South: Muslims do better where they are given reservation and lost to others including SCs and OBCs where they are excluded from positive discrimination framework. In South India, policies of affirmative action work better where they are supported by complementary politics and policies. The case of Indian Muslims enables us to engage with the prevalent contention around religion being the basis for affirmative action, whereas class, caste, and gender have assumed to be the only legitimate claimants. Keywords
Affirmative action · Muslims · Education and Inequality
In India, inequalities have been traditionally analyzed from the point of view of caste and class, in particular since the 1990s because of the liberalization policies.1 But caste and class are not the only relevant criteria for understanding social inequalities: religion is another one. The rise of Hindu majoritarianism has, indeed, found expression in new forms of discrimination – or the accentuation of old ones. On the job market, these discriminations are evident from most of the studies available. Among other things, scholars have demonstrated that Muslim applicants to jobs in the private sector were even more victims of prejudice than Dalits (Thorat and Attewell 2007). These new trends have been precipitated by older ones, including the decline of Urdu in North India since the 1980s at least (Jaffrelot 2011) and processes of ghettoization (often related to communal riots) (Gayer and Jaffrelot 2012). As a result, Muslims have experienced a significant decline in socioeconomic and educational terms, compared to other social categories, including Dalits. Part I provides details on the relative position of Muslims vis-à-vis other social groups in the most important of the socioeconomic indicators, i.e., income, education, and jobs. Contrary to the BJP’s narrative, evidence shows the worsening position of Muslims vis-à-vis other social groups between 2005 and 2012, before the BJP rose to power – a clear indication that they have not been “pampered” by the UPA Congress-led governments. Muslims further lost in educational outcomes since then. But this is the time when the Manmohan Singh government, after appointing the Sachar Committee, thought about correcting the Muslims’ decline by resorting to positive discrimination. Part II maps the trajectory of affirmative action policies initiated in the 2000s and their – often older – regional variations within India. 1
For a brief overview of the rise of inequalities in post-1991 India, see Jaffrelot and Kalaiyarasan (2021).
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
545
Trajectories of Socioeconomic and Educational Marginalization Our analysis covers 15 states. In the North, we include Jammu and Kashmir (where Muslims were 68.3% of the population in the 2011 census), Uttar Pradesh (19.3%), Rajasthan (9.1%), Madhya Pradesh (6.6%), Haryana (17%), and Delhi (12.9%); in the East, Bihar (16.9%), Assam (34.2%), and West Bengal (27%); in the West, Gujarat (9.7%) and Maharashtra (11.5%); and in the South, Karnataka (12.9%), Andhra Pradesh (9.6%), Tamil Nadu (5.9%), and Kerala (26.6%). These selected states account for 95% of the 170 million Muslims enumerated in 2011. To map socioeconomic situation of Muslim, we use both the rounds (2004–2005 and 2011–2012) of the Indian Human Development Survey (IHDS), done by the National Council of Applied Economic Research (NCAER) in collaboration with the University of Maryland.2 The data provides a panel for 2004–2005 and 2011–2012, and one can thus trace a mobility of the identical households. Our analysis relies on a sample of 4,562 Muslim households surveyed by the IHDS in 2011–2012 (see Table 1). The same sample is available for 2004–2005 except the households3, which got split in 2011–2012.
Income: When Muslims Come Last In 2011–2012, the average annual per capita income of Muslims was Rs. 20,062, which was the lowest in the country in terms of broad categories, followed by Rs. 20,472 for SCs, Rs. 23,841 for OBCs, and Rs. 26,037 for all Hindus. The Brahmin tops the income order with Rs. 40,569 followed by the category “Other Hindu upper castes” (excluding Brahmins) of Rs. 39,133. In terms of level and rate of change, Muslims’ condition has worsened vis-à-vis all the other groups. In 2004–2005, they earned 81% of what Hindus did, but in 2011–2012, the percentage declined to 77%. Similarly, Muslims earned 108% of what SCs made in 2004–2005, but in 7 years, the ratio came down to 98%. Muslims earned 89% of that of OBCs in 2004–2005 but only 84% in 2011–2012. We examine Muslims as a block because, from a socioeconomic point of view, differences between them are remarkably limited, in spite of the fact that they are differentiated along caste lines. The National Sample Survey in its 68th round (2011–2012) estimated that OBC Muslims constitute about 51% of total Muslims
2
The Indian Human Development Survey (IHDS) was done by the National Council of Applied Economic Research (NCAER) in collaboration with the University of Maryland. This paper uses both the rounds, i.e., the IHDS I done in 2004–2005 and IHDS II done in 2011–2012. It is a nationally representative, multi-topic survey of 42,152 households done in 1,503 villages and 971 urban neighborhoods across India. IHDS II reinterviewed about 83% of the IHDS I households plus any split households that resided in the same community. 3 The sample is identical for 2004–2005. However, a small percentage of households which got split (e.g., a family becoming into two after marriage, fragmented for other reasons) in 2011–2012 are not available for 2004–2005. Yet, the parental households are retained. The actual sample of HHs available for all India in 2004–2005 is 34,643.
546
C. Jaffrelot and K. A.
Table 1 Average annual per capita income (in Rs.) Regions 2004–2005 North J&K UP Haryana Delhi MP Rajasthan West Gujarat Maharashtra East Bihar West Bengal Assam South Karnataka Kerala AP Tamil Nadu All India 2011–2012 J&K North UP Haryana Delhi MP Rajasthan West Gujarat Maharashtra East Bihar West Bengal Assam South Karnataka Kerala AP Tamil Nadu All India
Brahmins
Upper castes
OBCs
SCs
All Hindus
All Muslims
15,611 10,643 15,072 18,785 10,683 13,917 20,528 19,492 9,747 21,054
16,226 13,052 18,720 21,576 9,886 13,948 14,653 15,344 7,548 17,653
5,991 12,909 16,183 6,256 9,891 7,751 10,434 4,944 9,540
12,841 4,915 7,489 17,122 4,560 6,366 9,964 9,445 4,243 7,722
14,637 6,969 13,596 18,166 6,381 9,497 9,862 11,566 5,271 11,681
11,179 6,104 6,391 13,179 7,107 8,454 7,546 8,515 5,672 6,242
15,377 23,641 11,800 25,425 17,003 15,316
11,816 12,298 18,348 11,708 7,588 14,711
11,742 9,127 13,906 8,260 11,520 8,667
11,213 7,698 10,991 6,796 6,991 7,142
11,587 9,814 14,454 8,423 9,758 9,559
10,905 8,274 15,007 9,622 9,101 7,750
38,548 25,774 60,509 58,807 32,879 41,906 44,144 44,638 28,093 50,113
52,511 28,430 79,506 72,633 33,025 37,512 46,406 35,264 25,304 37,995
15,539 35,260 37,093 17,837 24,131 26,282 28,448 14,439 20,963
31,393 11,762 20,566 45,285 14,929 17,240 27,348 26,172 10,767 22,276
40,554 17,004 42,206 51,273 18,986 24,583 30,142 30,917 15,125 28,092
36,865 15,505 13,952 36,266 17,997 26,496 18,864 22,847 12,418 17,606
38,060 73,505 53,461 50,242 63,749 40,569
30,781 37,129 53,770 26,937 64,611 39,133
46,913 27,387 45,609 21,118 35,382 23,841
31,006 20,183 41,658 18,345 28,109 20,472
31,491 27,491 47,098 21,395 35,078 26,037
18,636 20,473 34,305 18,933 24,737 20,062
in India (Note that Dalit Muslims are included in the OBC category). The IHDS that we use here also reports that 53% of total Muslims in India are OBCs. But the difference between OBC Muslims and non-OBC Muslims is lesser compared to the gap that remains between Hindu OBCs and upper caste or “General” Hindus. The average per capita income of OBC Muslims was Rs. 19,263 as against Rs. 20,490 in
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
547
2011–2012 for upper caste Muslims – a marginal 4% difference. Whereas the per capita income of upper caste Hindus was Rs. 39,133 in 2011–2012, the OBCs were much below this level at Rs. 23,841 – a 64% difference. OBCs, therefore, stayed behind the average per capita income of “All Hindus” by more than Rs. 10,000 a year. By contrast, the per capita income of Muslim OBCs was hardly different from that of the average per capita income of their community (Rs. 19,263 against Rs. 20,062). If disaggregated by states, Muslims in southern and western India are doing better compared to their counterparts in northern and eastern India except in small states such as J&K and Delhi (where annual per capita income of Muslims are Rs. 36,865 and Rs. 36,266, respectively). In 2011–2012, the annual per capita income of Muslims in Kerala was Rs. 34,305 which is more than twice that of the UP Muslims (Rs. 15,505) and almost thrice of Bihar (Rs. 12,418). The average income of Muslims in Tamil Nadu is Rs. 24,737 followed by Maharashtra (Rs. 22,847), Karnataka (Rs. 20,473), AP (Rs. 18,933), and Gujarat Muslims (Rs. 18,864). Other lower-performing states are Haryana (Rs. 13,952), MP (Rs. 17,997), and West Bengal (Rs. 17,606). The only exception to this pattern is Muslims in Rajasthan where they earn Rs. 26,496 quite close to their counterparts in South India. Muslims earn, on average, 77% of what Hindus do in India (see Table 2). But the gap is smaller in poor states (Muslims earn 95% of what the Hindus earn in MP, 91% in UP and J&K, and 82% in Bihar) than below the Vindhyas (where the percentages are, respectively, 70%, 73%, 74%, and 75% for Tamil Nadu, Kerala, Maharashtra, and Karnataka). In AP, they earn about 89% of what Hindus do. The gap is the largest in West Bengal as well as Gujarat (where Muslims earn only 63% of what Hindus earn in both states) and in Haryana (where Muslims earn only 33% of what Hindus earn, partly because of the poor condition of Muslim-dominated districts like Mewat and, in contrast, the affluence of the adjacent Hindu-dominated district of Gurgaon). Muslims earn only about 71% of what Hindus earn even in Delhi. In sum, Muslims earn less than Hindus in all the states except Rajasthan where they earn 108% of what Hindus earn. In no state are Muslims better off than Hindu OBCs except in Rajasthan where they earn 110% of what Hindu OBCs earned in 2011–2012. In states such as Haryana and Assam, Muslims earn less than half of what Hindu OBCs do. Muslims’ per capita income represents about 40% of that of Hindu OBCs in these two states, while it is 70% in Tamil Nadu, 72% in Gujarat, 80% in Maharashtra, 75% in Kerala and Karnataka, 84% in West Bengal, 86% in Bihar, and 90% in AP. In UP, Muslims earn exactly the same thing as Hindu OBCs and put up equal competition (98%) to the latter in Delhi. In India as a whole, Muslim earns 84% of the revenue of Hindu OBCs in 2011–2012. However, this was not the case 7 years before. Muslims were better off than Hindu OBCs in 5 of 15 states. The per capita income of Muslims was 117% of that of Hindu OBCs in AP, 115% in Bihar, 114% in MP, and 108% in Kerala in 2004–2005 and deteriorated, respectively, to 90%, 86%, 100%, and 75% in 2011–2012. Of the 15 states, in 11 states, Muslims’ relative position to OBCs deteriorated during 2004–2005 to 2011–2012. The states that witnessed the sharpest decline are
548
C. Jaffrelot and K. A.
Table 2 Jobs (percentage of salaried) in 2004–2005 and 2011–2012 Region
States
North
J&K MP Rajasthan UP Haryana Delhi Gujarat Maharashtra Bihar West Bengal Assam Karnataka Kerala AP Tamil Nadu
West East
South
All India North
West East
South
All India
J&K MP Rajasthan UP Haryana Delhi Gujarat Maharashtra Bihar West Bengal Assam Karnataka Kerala AP Tamil Nadu
Brahmins 2004–2005 32 21.4 33.5 19.5 40.5 74.2 40.3 42.9 13.4 39.7
Upper castes
OBCs
SCs
All Hindus
All Muslims
41.1 12.2 25 21.1 22.1 69.1 20.3 41.2 14.6 30.3
Nil 6.2 13.9 11.8 19.1 48.5 14.4 12.5 7.5 21.4
41.2 7.5 10 8.8 16.1 67.3 29.7 27 10.2 12.1
35.9 8.3 14.8 12.8 20.4 62.3 18.4 25.3 9 21.7
27.3 13.7 19.2 8.8 11.8 34.4 11.6 23.9 12.9 9.0
20.7 26 11.6 52.7 31 31.4
20 8.8 19.7 12.7 11.9 24.5
27.2 8.5 11 11.7 12.1 13
17.2 9.4 16.8 10.3 7.8 13.5
19.1 8.7 14.5 11.8 10.8 16
8.1 8.0 11.0 11.9 13.4 13.3
2011–2012 26.9 17.1 38.9 22.4 45.8 87.2 29.4 54.1 15.2 36.8
31.8 16.7 22.3 17.1 25.3 62.7 25.2 40.8 13.1 26.4
Nil 10.7 12.4 9.9 20.7 43.3 14.9 9.9 6 14.5
27.5 11.5 11.7 9.6 18.5 63.3 26.8 27.5 7.7 12.2
28.2 10.8 14.5 11.9 23 60.1 18 25.7 7.5 19.2
29.7 17.6 20.1 11.1 9.1 37.3 17.7 24.4 7.4 6.4
25.1 24.6 39.7 24.2 31.3 30.1
29.2 18.4 30.9 12.3 23.6 25
35.3 11.8 22 11.9 14 13.7
21.8 12.6 17.6 13.4 10.6 14.7
24.7 11.9 23.8 12.5 13.6 16.7
8.2 11.9 21.7 9.1 10.8 12.9
Assam (53% points) followed by Kerala (33% points), Bihar (29% points), and AP (27% points). Of the 15 states presented here, Muslims earn less than Dalits in 9 states in 2011–2012. In Assam, Muslims earn 60% of what Dalits do, followed by 68% in Haryana, 69% in Gujarat, 79% in West Bengal, and 80% in Delhi. The states where
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
549
Muslims earn more than Dalits are Rajasthan (153%) followed by UP (132%), MP (121%), J&K (117%), and Bihar where they earn 115% of what Dalits do. The states where Muslims compete very closely with Dalits are AP where they earn 103% of what Dalits do, followed by 101% in Karnataka and 88% in Tamil Nadu. As a whole of India, the average income of Muslims is 88% of what Dalits made in 2011–2012. This was, however, not the case 7 years before. Muslims were better off compared to Dalits in most of the states in 2004–2005. Muslims earned more than Dalits in 8 of 15 states. For instance, Muslims earned, as against Dalits, 156% in MP, 142% in AP, 137% in Kerala, 134% in Bihar, 133% in Rajasthan, 130% in Tamil Nadu, 124% in UP, and 108% in Karnataka. Seven years later, Muslims slipped down to Dalits in Kerala and Tamil Nadu. In Karnataka, they were neck to neck to Dalits in 2011–2012 at 101%. And the states where Muslims earned less than Dalits in 2004–2005 have seen a further decline in their position vis-à-vis Dalits except in J&K, where they have gained from 87% to 117%. In Haryana, Muslims earned 85% of what Dalits did in 2004–2005, and 7 years later, it slipped into 68% in 2011–2012. The corresponding figures for Assam are 97% and 60%, respectively. The respective figures for Gujarat are 76% and 69%. West Bengal has seen a marginal decline from 81% in 2004–2005 to 79% in 2011–2012. On all-India average, Muslims earned 109% of what Dalits did in 2004–2005, and 7 years later, they lost to Dalits, where the figure became 88% in 2011–2012. This trend only shows the secular decline of Muslims vis-à-vis not just Hindus but vis-à-vis Hindu OBCs and Dalits.
Education: The Widening Gap Between Muslims and Hindus Muslims do better in education in South India than in the rest of the country. Tamil Nadu tops in the education of Muslim, partly because of century-old reservation policy of the state where Muslims are accommodated within the scheme of reservation if we exclude small states like Delhi. In 2011–2012, the percentage of graduates among Muslims in Tamil Nadu is 10.7%, followed by J&K (7.6%), Assam (5.5%) and Kerala (4.3%), AP (3.8%), Maharashtra (3.6%), and UP (3.5%). The least performing states are Haryana (0.6%), Gujarat (1.6%), and West Bengal (2%). States such as Karnataka (2.8%), Rajasthan (2.4%), and MP (2.8%) are placed between these two extremes. These results are better than those of 2004–2005, but they do not necessarily reflect a significant educational improvement among Muslims when one compares them with Hindus. In all the states, the percentage of graduates among Muslims is lower than that of Hindus, except in Tamil Nadu where they have 10.7% as against 8.9% for Hindus, in J&K (7.6% against 7.1%), and in Delhi (12.2% against 10.7%). The educational distance between Muslims and Hindus is higher in rich states as compared to the poorer ones. For instance, the percentage of graduates among Muslims in Haryana is 0.6% as against 7.1% for Hindus, while they are, respectively, 4.3% and 10.7% for Kerala, 1.6% and 5.6% for Gujarat, and 3.6% and 7.6% for Maharashtra. The gap is narrowed in states such as Assam (5.5% among Muslims as against 6.6% for Hindus) and UP (3.5% as against 5.1% for Hindus), while the level of education is the same for both Hindus and Muslims in Bihar. In MP, the trend was
550
C. Jaffrelot and K. A.
reversed in 2011–2012 as 3.4% of graduates for Muslims as against 2.8% for Hindus. However, in the last 7 years, as in income, Hindus gained more than Muslims in education. The educational gap measured in terms of percentage of graduates between Hindus and Muslims in 2004–2005 was 1.7% points; it became 2.3% points in 2011–2012. Of the 15 states, in 8 states, the educational position of Muslims vis-à-vis Hindus further deteriorated. While Muslims’ position in education marginally improved since 2004–2005, their situation in relation to Hindu OBCs further worsened during 2004–2005 to 2011–2012 in 8 of 15 states presented here. States such as Assam, Maharashtra, Haryana, and Karnataka have seen the worst decline in the relative position of Muslims with respect to OBCs. This could be because of the upward educational mobility of OBCs thanks to the reservation in education introduced by the Mandal II (that resulted in the introduction of a 27% quota for OBCs in the Universities in 2006). Muslims lag behind Hindu OBCs in the percentage of graduates in all states except in MP (where they have 2.8% against 2.5% among Hindu OBCs) and in Bihar where they are at 3.1% against 2.7% among Hindu OBCs, in Tamil Nadu at 10.7% against 9.4% among Hindu OBCs, and in Delhi at 12.2% against 8.6% among Hindu OBCs. The educational distance between Hindu OBCs and Muslims in the percentage of graduates is much sharper in states such as Assam (14% against 5.5% among Muslims), Haryana (5.4% against 0.6% among Muslims), Kerala (9.2% against 4.3% among Muslims), and Maharashtra (7.8% against 3.6 among Muslims). The gap is narrower in Rajasthan, UP, and Gujarat. Muslims lagged behind Dalits in education in 7 of the 15 states presented here in 2011–2012 (see Table 3). In the rest of the states, they were very close to Dalits in the percentage of graduates except in Tamil Nadu (where they are at 10.7% against 5.5% among Hindu Dalits), in UP (3.5% against 2.3% among Hindu Dalits), in Bihar (3.1% against 0.9% among Hindu Dalits), and in Assam (5.5% against 4% among Dalits). Of the seven states where Muslims lag behind Dalits in percentage of graduates, the gap is much wider in states such as Kerala (4.3% as against 6.7% among Dalits), Gujarat (1.6% as against 4.8% among Dalits), Maharashtra (3.6% as against 5.1% among Dalits), and Haryana (0.6% as against 2.1% among Dalits). This was not the case 7 years before. The Sachar’s report was correct that Muslims were indeed better than Dalits in 8 of the 15 states analyzed here. Not only Muslim lost ground in two more states (Karnataka and AP) vis-à-vis Dalits during 2004–2005 to 2011–2012, but also their condition in relation to Dalits further worsened in the rest of the states. Here, the fact that Muslims are lagging behind Dalits in terms of education does not only reflect the lack of progress among the former but also of the relatively faster rate of progress among the latter.
Jobs: When the Informal Sector Prevails The socioeconomic condition of Muslims, who are overrepresented among artisans (including mechanics) and petty shopkeepers, is directly related to their very large participation in the informal sector.
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
551
Table 3 Education (percentage of graduates) in 2004–2005 and 2011–2012 Regions
States
North
J&K MP Rajasthan UP Haryana Delhi Gujarat Maharashtra Bihar West Bengal Assam Karnataka Kerala AP Tamil Nadu
West East
South
All India North
West East
South
All India
J&K MP Rajasthan UP Haryana Delhi Gujarat Maharashtra Bihar West Bengal Assam Kerala Karnataka AP Tamil Nadu
Brahmins 2004–2005 5.9 9.6 7.8 6.2 11.6 9.7 17 16 4.6 17.4
Upper castes
OBCs
SCs
All Hindus
All Muslims
10.3 6.5 6.7 8.4 9.2 10.4 5.9 8.3 5.5 11.5
Nil 1.7 2.1 2.2 3.1 3.4 1 3.5 1.9 4.3
1.1 0.8 1.1 1.2 0.8 5.4 2.7 1.9 0.5 1.8
5.9 2.4 2.7 3 4 6.5 3.2 4.2 2.1 5.8
3.3 1.2 1.8 1.9 0.5 7.0 0.8 2.3 1.6 1.3
13.4 21.9 12.8 7.1 19.7 10.5
4.2 4.7 11.8 4.9 6.9 7.4
6.4 2.9 6.4 1.9 6.2 2.9
3 1.4 3.2 1.4 2.1 1.5
3.8 3.2 7.2 2.2 4.9 3.6
1.9 1.6 1.9 1.5 4.4 1.8
2011–2012 7.1 10.8 12.9 10.1 15.7 21.9 19.1 26.1 7.5 22
11.0 9.1 11.6 13.9 14.4 16.6 11.7 11 8.1 14.2
Nil 2.5 3.4 3.6 5.4 5.9 2.1 7.8 2.7 6.5
2.0 1.6 2.1 2.3 2.1 8.6 4.8 5.1 0.9 2.7
7.1 3.4 4.7 5.1 7.1 10.7 5.6 7.6 3.1 7.6
7.6 2.8 2.4 3.5 0.6 12.2 1.6 3.6 3.1 2
17 18.5 24.4 Nil 21.1 14.4
7.2 17 6.9 9.9 21.1 11.5
14 9.2 4.8 4.9 9.4 4.9
4 6.7 3.4 4.4 5.5 3.1
6.6 10.7 5.3 5.5 8.9 5.9
5.5 4.3 2.8 3.8 10.7 3.7
On the whole, the percentage of salaried among Muslims is 12.9% – lower than Hindus (16.7%), Hindu OBCs (13.7%), and Hindu SCs (14.7%) in 2011–2012. They are employed in household industries and are by and large self-employed (as laborers, artisan, and shopkeepers). About 42% of Muslims work as daily
552
C. Jaffrelot and K. A.
laborers, which is one of the highest in the country next only to Dalits (54.8%). Not surprisingly, the next big sector where Muslims are employed is the petty trade. The percentage of Muslims who work in this sector is about 19% which is highest as compared to any other socio-religious groups in this country. Among the states, Delhi tops in the percentage of salaried Muslims where 37% of them are in salaried class followed by 29.7% in J&K, 24.4% in Maharashtra, 21.7% in Kerala, 20.1% in Rajasthan, and 17.7% in Gujarat. Less than 10% of them had a salaried job in the three states where they were among the poorest (Haryana, Bihar, and West Bengal), whereas more than 20% of them had a salaried job in two of the states where they were the richest (Kerala and Maharashtra). This achievement is partly due to reservation policies: many South Indian Muslims enjoy some form of positive discrimination there. Karnataka set aside 4% for Muslims as “more backward” within the OBC category, and in Kerala, Muslims are given 12% sub-quotas within the 40% reserved for OBCs in government jobs. The only exception to this trend is MP and Rajasthan where the percentage of salaried among Muslims is 18% and 20%, respectively. This enigma needs to be explored. If we compare with all Hindus, in no state are Muslims better off than Hindus in terms of salaried jobs, except J&K, MP, and Rajasthan. In Haryana, West Bengal, and Assam, the percentage of salaried among Muslims is less than half of that of Hindus. Similarly, in 9 of the 15 states analyzed here, the percentage of salaried among Muslims is less than that of Dalits. Kerala is one of the two states (the other is UP) where Hindu Dalits do not have a higher percentage of salaried jobs than Muslims (17.6% as against 21.6%).
Youth in Higher Education The NSS-PLFS data released by the Ministry of Statistics and Programme Implementation in June 2019 shows further deterioration of Muslims, particularly of its youth. We use three variables here: the percentage of Muslims’ educated youth vis-à-vis other social groups (21–35 years of age) having completed graduation and above, the percentage of youth (15 to 24 years of age) currently in educational institutions, and the percentage of youth who are in “neither in employment nor in education and training” (NEET) category, a broader measure of employment inadequacy for the age cohort 15–24 years of age. These variables together reflect pathways of educational mobility of youth in the country (Table 4). The youth having graduation and above among Muslims in 2017–2018 is 14% as against 18% among Dalits and 25% among Hindu OBCs and as high as 37% among Hindu upper castes. The gap between SCs and Muslims is 4 percentage points (ppts) in 2017–2018. Six years before (2011–2012), SC youth was just marginally above that of Muslims in educational attainment – just 1 percentage point. SCs clearly took over Muslims between 2011–2012 and 2017–2018. Similarly, the gap between Muslims and Hindu OBCs was 7 ppts in 2011–2012, and now it has gone up to 11 ppts. There are considerable geographical variations of such trend. For instance, the educational attainment gap between SCs and Muslims is highest in Uttarakhand (16 ppts) followed by Gujarat (14 ppts), Rajasthan (12 ppts), Haryana (12 ppts),
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
553
Table 4 Education (percentage of graduates for age 21–29 in 2018) Regions North
West East
South
All India
States MP Rajasthan UP Haryana Gujarat Maharashtra Bihar West Bengal Assam Karnataka Kerala AP Tamil Nadu
Hindu upper castes 41 41 42 36 39 35 38 23
Hindu OBCs 16 27 25 25 20 26 10 24
SCs 12 18 18 15 27 23 7 9
All Hindus 17 26 26 28 25 28 13 16
All Muslims 17 7 11 3 13 16 8 9
22 41 66 38
12 25 54 25 46 25
8 15 39 22 37 18
14 27 53 26 44 25
7 18 28 21 36 14
37
Kerala (11 ppts), and Maharashtra and UP (7 ppts). Six years before (2011–2012), in all of these states, SCs had taken over Muslims but stood just slightly above them. What we see now is convergence between Hindu SCs and OBCs who experience greater mobility, while Muslims are being sidelined. Such marginalization of Muslims becomes clear when we look at the percentage of youth currently in educational institutions because entry into educational system is a significant factor that sets the path for future educational attainment. Youth who are currently in educational institutions is the lowest among Muslims in the country. Only 39% of Muslims in the age cohort of 15–24 are in educational institution as against 44% in SCs, 51% in Hindu OBCs, and 59% among Hindu upper castes. States such as Haryana (22%), Gujarat (29%), Rajasthan (32%), UP (32%), and West Bengal (36%) perform lower than the all-India average, while Kerala (60%), Tamil Nadu (48%), Telangana (56%), Assam (44%), and Maharashtra (40%) do better than the national average. What is worrying is that a sizeable proportion of Muslim youth are making exit from the formal system, i.e., moving into the NEET category. Thirty-one percent of Muslim youth fall in this category – the highest in the country – followed by 26% among SCs and 23% among Hindu OBCs and 17% among Hindu upper castes. While sources of the marginalization began much before, what is disturbing is the acceleration of such marginalization in the recent times. A recent comprehensive study with a longer period only strengthens our argument that Muslims are biggest losers in the educational mobility, while others including SCs have gained (Asher et al. 2020).4
4 The recent study by Sam Asher et al. (2020) shows that upward educational mobility of SCs is exactly offset by the fall of Muslims in India, while Hindu upper castes and OBCs remain constant.
554
C. Jaffrelot and K. A.
Muslims in Government and Public Sector Jobs While Muslims in South India are doing better in education, jobs, and income as compared to their counterparts in North India, they still do poorly vis-à-vis other groups including Dalits in terms of government jobs – the jobs everybody are longing for at the bottom of the pyramid. In the ratio of the group shares in government to its population, Muslims lag behind the Hindus, OBCs, and Dalits.5 In 2018, the ratio was 0.56 for Muslims as against 1.04 among Hindus, 0.89 among OBCs, and 1.19 among SCs. Of the 15 states we analyzed, not even in a single state, representation of Muslims in government and public sector jobs was equal to their population share, whereas SCs have more representation than their population share in most of the states except MP, Rajasthan, Delhi, Assam, and West Bengal – OBCs’ representation too is close to their population share. As expected, upper castes do well and disproportionally claim higher share as compared to their population. Muslims’ representation was as low as 0.10 in Delhi to 0.98 in Jammu and Kashmir. However, what is surprising is that the southern states where Muslims enjoyed reservation do not do well as much as Dalits and OBCs (Table 5). Muslims’ representation in bureaucracy has not improved since the submission of SCR in 2006. Of 3,511 IAS intakes through the UPSC in 2016, only 96 – 2.7% – were Muslims as against 2.3% in 2006.6 Even those who graduated to IAS through the state cadre, only 68 of 1425 – 4.8% – were Muslims as against 5% in 2006. Similarly, Muslims’ representation in police force too has declined. Of 3,754 IPS officers, only 120 (3.2%) are Muslims in 2016 as against 128 of 3209 (4%) in 2006 – both decline in both absolute and relative terms (Shaikh 2016). Thus, the most recent statistics – some of them more than a decade old, unfortunately – show that Indian Muslims are losing ground vis-à-vis Hindus, including OBCs and even Dalits, in terms of socioeconomic criteria (revenue as well as quality of jobs) and education. This is why, as early as 2004, after the Congress-led UPA won the general elections against the BJP-led NDA, the first Manmohan Singh government decided to assess the Muslims’ conditions and started to think about policies of positive discrimination.
Politics and Policies of Affirmative Action The history of relative well-being of Muslims in South India is partly tied to the history of affirmative action in this region. By contrast, at the national level, the central government did very little.
5
We constructed this ratio using the NSSO PLFS data for those reported as employed in government and public sectors in 2017–2018. 6 In 1981, they were 2.98% among a total of 3883 IAS officers, while in 2000, they were marginally less at 2.83% (Hasan 2006).
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
555
Table 5 Ratio of share in government/public sector jobs to population share in 2018 Region North
West East
South
All India
States J&K MP Rajasthan UP Haryana Delhi Gujarat Maharashtra Bihar West Bengal Assam Karnataka Kerala AP Telangana Tamil Nadu
SCs 1.04 0.81 0.89 1.15 1.11 0.91 1.45 1.07 1.00 0.75 0.75 1.05 1.26 1.16 1.19 1.32 0.96
OBCs 0.96 0.75 0.90 0.74 0.83 0.67 0.87 0.95 0.97 0.92 1.17 0.88 0.84 0.97 0.89 0.97 0.86
Upper caste 0.98 1.90 1.38 1.32 1.05 1.19 1.02 1.08 1.27 1.14 0.89 1.26 1.27 1.01 1.68 0.96 1.22
All Hindus 1.06 0.99 1.05 1.10 1.09 1.15 1.04 1.04 1.05 1.17 1.18 1.06 1.34 1.01 1.04 1.03 1.04
All Muslims 0.98 0.58 0.50 0.43 0.39 0.10 0.68 0.51 0.67 0.60 0.64 0.51 0.31 0.69 0.72 0.35 0.56
Timidity and Bias at the Center At the national level, the constitutionality of religion as a marker in identifying backwardness or policy measure has been often delegitimized. As a result, even Dalit Muslims were excluded in availing benefits under the Scheduled Caste quota by a presidential order.7 One may argue that Dalit Muslims were thrice discriminated, i.e., by the society, own community elites, and the state. The first is driven by prejudice on religious terms, while the second is driven by caste bias within Muslims. By denying reservation within the SC quota, the state has only worsened their plight. Similarly, the inclusion of backward caste Muslims within the OBC list should have been much smoother than what has been the experience. In fact, the backward class commissions – Kaka Kalelkar in 1956 and Mandal Commission in 1978 – included Muslims in their OBC lists. Besides, there are 14 states which have identified as many as 100 Muslim castes in the OBC lists (Krishnan 2012). However, many state lists of Muslims have not been recognized by the central lists.8
Paragraph 3 of the Constitution (Scheduled Castes) Order, 1950, states “Notwithstanding anything contained in paragraph 2, no person who professes a religion different from the Hindu, the Sikh or the Buddhist religion shall be deemed to be a member of a Scheduled Caste.” While this order was amended to include Dalit Sikhs in 1956 and Dalit Buddhists in the 1990s, it continues to exclude Muslims and Christians. 8 For instance, All Bengal Minority Youth Federation made complaints to the state minority affairs in West Bengal that the state list of OBC Muslims was not accepted by the union government. For details, see https://www.hindustantimes.com/india/wb-muslim-group-unhappy-over-obc-quota/storyAZ8ZbdZXPaIe3IqHXNprsN.html. 7
556
C. Jaffrelot and K. A.
The union government (the UPA I) appointed a commission under Justice Ranganath Misra only in 2004 to identify sources of backwardness and remedial measures which submitted its report in May 2007. It made two significant recommendations among others: all caste groups which are part of the SCs and the OBCs should also be regarded as backward in the Muslim community, thereby deserving reservation under the present scheme of things.9 Moreover, the committee also said that denying reservation to Dalit Muslims under the SC quota was unconstitutional and was against fundamental rights.10 The commission highlighted the Supreme Court order of 1950, which denies reservation for Dalit Muslims and Christians, as inconsistent with Articles 14, 15, 16, and 25 of the Constitution.11 The commission also recommended 6% for Muslim OBCs within the OBC quota of 27%. The committee also recommended a 15% reservation in all government jobs for all minorities and, within this 15%, 10% for Muslims.12 Additionally, it also recommended that at least 15% seats in all non-minority educational institutions should be earmarked for the minorities and, within this 15%, 10% for Muslims.13 However, the report has not been tabled in parliament because the BJP opposed it and called it both anti-national and anti-Hindu (Rahman 2010). According to the Hindu nationalists, to consider religion as a criterion of positive discrimination was unconstitutional. Unlike the Misra report, the SCR did not advocate separate reservation for Muslims as a collective group but did endorse that Dalit Muslims and Muslim OBCs should be treated at par with their counterparts.14 In other words, both the commissions advocated delinking the faith to avail reservation in social terms.15 In addition, the SCR proposed other remedial measures such as scholarships, coaching centers, easier financial access, and non-discrimination/fairness provisions in selections. As a long-term measure, in the place of reservation, the Sachar Committee recommended a diversity index to ensure equal opportunity to “all socio-religious groups in the areas of education, government, private employment and housing”
9
See the Report p. 149 (16.1.6). The Sachar Committee has also argued that it would be most appropriate for them to be included under the SC category or at least clubbed with the most backward caste (MBC) category (pg. 198). 11 The order provided five percent reservations for STs and 12.5% for SCs but excluded Dalit Muslims and Dalit Christians in 1950. The reservation was also increased to 7.5% for STs and 15% for SCs in the 1970s. 12 p. 153 para. 16.2.16 of the report 13 p. 150 para. 16.2.7 of the report 14 There are contested opinions within Muslims on reservation question. One, as argued by Ashfaq Husain Ansari (2004), reservation for Muslims should not be conflated with reservation for OBC Muslims as the latter are legitimately entitled to. On the other hand, Syed Shahabuddin (2004), for instance, argues that Muslims in entirety are eligible for quota proportion to their population and the problems of social backwardness of certain Muslims can be dealt with internally on a preferential basis. 15 The SCR argued that either it would be most appropriate to include them under the SC category or at least clubbed with the most backward castes (p. 198). 10
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
557
(SCR, p. 242) and Equal Opportunity Commission (EOC) to look into “the grievances of the deprived groups” (ibid., p. 240). Following the recommendations, UPA I formed two bodies in August 2007, the committee to formulate diversity index under Amitabh Kundu (2008) and the Equal Opportunity Commission under N.R. Madhav Menon.16 While the Sachar Committee evaluated the socioeconomic condition of Muslims, the commission on diversity index was set out to measure diversity in education and employment to operationalize the Sachar Committee recommendations. Both entities submitted their reports in June 2008 and February 2008, respectively. Some argued that these three committees – the SCR, the diversity index by Kundu, and the EOC – marked a paradigmatic shift in addressing the question of equality in India as it aimed to move beyond the fixed quota-based reservation (Hasan 2009). However, both Misra report and SCR have not received as much attention in policy discourse as in politics – as evident from the aggressivity of Hindu nationalist critiques. Notably, one can safely argue that it marked a conceptual shift in addressing the community’s deprivation from a question of identity to one of substantial material concerns, i.e., educational backwardness and economic marginality of Muslims. But the action on both committees’ recommendations was dismal. In a debate, Syeda Saiyidain Hameed, a former member of the Planning Commission, admitted that the Planning Commission did not allocate the requisite funds for implementing the Sachar Committee’s recommendations.17 Another report (drafted by the Post-Sachar Evaluation Committee headed by Amitabh Kundu), in 2014 (p. 167), evaluated the implementation of the decisions taken on the SCR and concluded that the policy interventions were not as adequate, given the magnitude of deprivation. In sum, these reports – of Misra and the SCR – did not have as much impact on policy levels as they did in politics.
Muslims in Dravidian Land While Muslims were denied quota at national level, they continue to enjoy some forms of reservation in South India since the British colonial rule. Social movements in South, the Dravidian mobilization in particular, articulated concerns of Muslims better. Muslims in Kerala have better development indicators in education, jobs, and entrepreneurship. Besides their political alliance with lower castes which integrated them into mainstream politics, migration to West Asia thanks to oil boom also brought them prosperity. However, the affirmative action worked as a foundation for building educational capital among Muslims in Kerala. In fact, migration was
16
The committee was asked to examine and determine the structure of an equal opportunity commission to ensure the full equality of opportunities for SCs, STs, OBCs, and religious minorities. For more details, see Madhava Menon (2008). 17 For her interview: https://www.youtube.com/watch?v¼-V7Xeyrkq7M
558
C. Jaffrelot and K. A.
remunerative only owing to the educational empowerment through affirmative actions. Within Kerala, there were two trajectories of reservation – Malabar and Travancore. Malabar region was under Madras Presidency; hence, the reservation initiated in British colonial regime in 1920 continued till the formation of Kerala state in 1957. The affirmative action policies evolved on a different track in Travancore princely state. For instance, in Travancore region, Muslims were a part of broader coalition consisting of Ezhavas and a section of Christians who demanded a “proportional reservation” – quota proportion to population of each community – in government services. What began as a native struggle for representation in government against Tamil Brahmins’ domination in the Travancore administration in 1891 evolved into a full-fledged struggle for proportional representation for different communities in government services in 1932. Muslims were a part of this struggle. In 1936, the Travancore administration conceded the demand and introduced the quota. It was again revised in 1952 by fixing it to 45%, of which 10% was reserved for SCs and STs and 35% was reserved for OBCs which included a sub-quota of 5% for Muslims (Damodaran Committee p. 9). On a parallel trajectory, Muslims in Malabar region under the Madras Presidency enjoyed reservation since 1921. It continued till the formation of the state – Kerala – in 1957. Both systems of reservation – Travancore and Malabar regions – were merged by fixing the total quota of 50%, consisting of 35% for the OBCs (5% sub-quota for Muslims) and 10% for SCs and STs. The reservation was reworked again in 1970 on the basis of the Nettur Damodaran committee which introduced a sub-quota of 10% in services (ibid, p. 13) and 12% in certain services. After Ezhavas (14%), Muslims form the largest population who also claim a large share of subquota (12%) in the state’s OBC list. As in Kerala, reservation for Muslims in Tamil Nadu predates independence. It was in 1921 that the first communal G.O. was introduced for apportioning quotas for various communities in government services. Muslims were part of this history of struggles. The system continued post-independence with the first constitutional amendment (Kalaiyarasan and Vijayabaskar 2021). The list with Muslims being a part of the OBC was retained by First Backward Class Commission headed by A.N. Sattanathan in 1970. In Tamil Nadu, Muslims have been associated with Dravidian mobilization in the state (Anwar 2018).18 The Justice Party, which was in power in the Madras Presidency, not only ensured political representation for the Muslims in the legislature but also in various offices of the state. Not only that, Muhammad Ismail Sahib, a sole member in the Constituent Assembly, representing the Muslim League articulated on the need for continuing with the prevailing system of reservations introduced by the Justice Party in Madras Presidency (Abdul Khader Fakhri 2008).
For instance, Dr. Natesan of the Justice Party argued that the motion had to be reframed as “nonBrahmin Indians” and went on to claim that non-Brahmins include Muslims. There was even idea of Dravidian Islam among Tamil-speaking Muslims (Anwar 2018).
18
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
559
When the new constitution was promulgated in 1950, elite groups took the judicial route to scrap the reservation policies introduced by the Justice Party. The court scrapped the reservation policy in 1951, and following a massive agitation in the state, the first constitutional amendment in post-independent India was introduced to continue reservation in education on a new pattern with 25% reserved for OBCs which included Muslims.19 Besides continuing the reservation of Muslims within the OBC list, the DMK also introduced sub-quotas of 3.5% in higher education and government jobs in 2007.20 Hence, it is not surprising that Tamil Nadu, in fact, tops in educational attainment among Muslims in India. For instance, Of the total 1, 82,255 candidates admitted in engineering and technology courses in Tamil Nadu in 2013–2014, 7557 (4%) were Muslims. Karnataka too has a history of reservation for Muslims. In 1921, following the recommendations of the Justice Miller Committee, Muslims were given reservation in Mysore state (Miller Commission Report 1919, p. 3). This system continued after the formation of Karnataka in 1956. When the OBC list which included Muslims was challenged in the High Court, the state brought Muslims again under the list. Later, based on the L.G. Havanur Commission (1975) which had a Muslim member, the Devaraj Urs government passed an order in 1977 providing reservation to Muslims. The decision again was upheld by the state High Court in the 1979. Subsequent commissions including the Venkataswamy Commission, set up in 1984, retained reservation for Muslims. In 1995, on the basis of the Rahman Khan Committee (1995, the Veerappa Moily’s government granted Muslims a sub-quota – 4% within the OBCs. While Telugu-speaking backward caste Muslims enjoyed reservations under the OBC list for a long time, an entitlement inherited from Madras Presidency government introduced by the Justice Party, the state of Andhra Pradesh created in 1956 retained Muslims within the OBC list. In 2007, on the basis of the P.S. Krishnan Committee recommendations, the state not only enlarged the pool of Muslim OBCs eligible for reservations in jobs and education but also created a 4% sub-quota for Muslims. Following the BJP’s attack, this policy was put on hold. Telangana – a new state carved out from Andhra Pradesh in 2014 – which again inherited earlier 4% reservation for Muslims from the erstwhile AP state, passed a new bill in the state assembly to increase reservation from 4% to 12% in jobs and education for Muslims in 2017.
19
The decision in State of Madras vs Champakam Dorairajan is a landmark case in which the Supreme Court upheld a High Court judgment striking down a government order providing castebased reservation in government jobs and college seats (Kalaiyarasan and Vijayabaskar 2021). 20 In 2007, the Tamil Nadu Backward Class Christians and Backward Class Muslims (Reservation of Seats in Educational Institutions Including Private Educational Institutions and of Appointments or Posts in the Services Under the State) Act was legislated to create a subdivision of backward Muslims and Christians.
560
C. Jaffrelot and K. A.
Northern and Eastern India: Failed Attempts and the Delegitimization of Pro-Muslim Policies The question of reservations for Muslims came in North India along with OBC politics. In Bihar, Karpuri Thakur, who heralded backward caste mobilization in the state, initiated Muslim reservation as early as the 1970s. Based on the Mungeri Lal Commission report, he had recommended 3% reservation for Pasmanda Muslims in Bihar in 1977 (Gupta 2017), but it was never implemented. However, the Pasmanda communities such as Ansari, Mansoori, Idrisi, Dafali, Dhobi, Nalband, etc. were included in the OBC list in Bihar. Similarly, in UP, a committee was constituted by the Mulayam Singh Yadav government – supported by a coalition of the Samajwadi Party and the Bahujan Samaj Party – in 1995 in order to examine the status of Muslim OBCs. It recommended an 8.4% separate reservation for them (Alam and Kumar 2019). The rise of BJP and indictement of such policies with rhetoric of “Muslim appeasement” put the implementation of such recommendations on hold. The reservation debate came to West Bengal very late, much later than even in North India. But when the SCR exposed the alarming condition of Bengali Muslims, it was partly attributed to their exclusion from affirmative action policies. As in South India, most of the Muslims in West Bengal are lower caste converts who are often identified as the Chandal communities, Jolahas and Momins, etc. (Dasgupta 2009). Hence, they are entitled to these policies. Following the SCR, the Left Partyled governments added a few Muslim communities in the OBC list. Mamata Banerjee who came to power in 2011 not only increased the OBC reservation from 10% to 17% but also added as many as 99 Muslim communities in the OBC lists.21 However, given the 27% Muslim population in the state, the state is yet to create a sub-quota for Muslims within the OBCs like the southern states. Muslims in Assam too lag behind others, and following the West Bengal model, they were not included sufficiently in positive discrimination programs (Karmakar 2019). A meager 4.5% of seats are reserved for Muslims in Assam against 34% of their population share. In sum, since the SCR in 2006, there have been resurgent debates on Muslim reservations (Hasan 2006; Krishnan 2012). As a result, the proportion of Muslims who reported as OBCs has jumped from 39% in 2004 to 51% in 2011–2012 (as per NSSOEUS). However, in South India, the share of Muslims who reported as OBCs stands at 99% in Tamil Nadu and Kerala, 75% in Karnataka, and 34% in AP, while it is 68% in UP, 76% in Bihar, and 85% in Rajasthan. The OBCs are reported to be lowest in West Bengal and Assam – 8% and 1%, respectively. In the absence of religion as a basis for inclusion in affirmative action, this increased identification as the OBCs and their inclusion in the OBC list would still offer mobility for some Muslims.
21
The West Bengal State Higher Educational Institutions (Reservation in Admissions) Rules was enacted in 2013 mandating the reservation of 17% for OBCs including Muslims, adding to the SC 22% and ST 6%. For detail, see https://wbhed.gov.in/readwrite/uploads/Reservation_Rules.pdf (accessed on 12 January 2021).
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
561
Muslims as Collateral Casualties of the Modi Government’s Style of Empowerment Since 2014, policy interventions have taken a different or rather worrying turn (Fazal 2014). The group-specific interventions suggested by earlier expert commissions were discarded and dismissed in the name of “minority appeasement” and the BJP claimed to be a party that would usher in development sans appeasement.22 Following this trend, NITI Aayog, which was created by the BJP government to replace the Planning Commission, moved from the community-oriented preferential policies to area-based development. It has released a ranking of 101 most underdeveloped districts in the country. Eleven of the top 20 most underdeveloped districts in the ranking are Muslim-concentrated districts. Among them, Mewat, a Muslimdominated district, not far away from the national capital, is the country’s most backward district. The government of the day claims to improve the condition of Muslims by developing these districts. It also brought economic category as basis for reservation, thereby claiming Muslims would benefit from such reservation. In sum, Dalit Muslims were excluded from the SC quota, while they, along with backward Muslims, were included in the OBCs. But both are highly unlikely to compete with landowning castes such as Yadavs in the OBC list. Even Ashraf Muslims who are eligible for EWS quota could hardly compete with Brahmins.23 Muslims are being marginalized in all the three respective social categories besides their collective deprivation owing to their religious identity. In 2022, the Modi government discontinued the Maulana Azad National Fellowship that had been introduced by the UPA government after the release of the SCR in order to help financially students from the minorities. It had benefited to 6,722 members of the minorirtes, mostly Muslims, between 2014 and 2022.
Conclusion While the socioeconomic condition of Muslims vis-à-vis Hindus, OBCs, and SCs has worsened overtime, Muslims in South India still fare better in development indicators in relation to their counterparts in the rest of India. This is partly due to the early implementation of affirmative action for Muslims in this region. However, it needs to be qualified since the decline is in relative terms vis-à-vis other communities not in absolute terms. Hence, the relative decline is also a function of differential mobility of SCs and OBCs. The Minister of Minority Affairs of BJP termed the abovementioned interventions as “only appeasement of minorities and zero empowerment,” so he put forward a shift in policy toward Muslims as “empowerment without appeasement” (https://www.outlookindia.com/website/story/ there-was-only-appeasement-of-minorities-and-zero-empowerment-in-last-60-years-n/306060). 23 In 2019 the Modi government has introduced a 10% quota for the Economically Weaker Sections, the upper caste poor. 22
562
C. Jaffrelot and K. A.
As argued elsewhere in this volume and above in this chapter the OBCs have indeed gained in administrative jobs and educational outcomes thanks to the reservation implemented by V.P. Singh in jobs and Manmohan Singh-led UPA I in 2006. Dalits have been enjoying reservation for a long time. However, the caste-based empowerment mobilization of the OBCs through the SP and RJD and the BSP for SCs has ensured better implementation of these policies only in the 2000s (Jaffrelot and Kalaiyarasan 2020). However, Muslims are left out in the process. It is true that Muslims who fall under backward castes were listed in the OBC lists prepared by both the National Commission for Backward Classes (NCBC) and the committee set up in 1993 following the Mandal judgment (Indra Sawhney v. Union of India 1992). However, these Muslims were highly unlikely to be able to compete with the betteroff OBCs – particularly land-owning OBCs. The better performance of Muslims in South India therefore lies not only in the century of inclusion of Muslims within the affirmative action framework but also early creation of sub-quota for them. Another variable is be added for the well-being of Muslims in the South: the relatively lower stigma and animosity against Muslims. As some argue, demands of Muslims’ political parties differ in the South in relation to the North wherein the former articulates the concerns on socioeconomic terms, while the latter is on religious terms.24 However, both need to be probed deeper.
Appendix Basic statistics Muslim population as per Census Major states 2011 (in %) J&K 68.3 Assam 34.2 West 27.0 Bengal Kerala 26.6 Uttar 19.3 Pradesh Bihar 16.9 Delhi 12.9 Karnataka 12.9 Maharashtra 11.5
Share in total Muslims in India (%) 5.0 6.2 14.3
Muslim population as per Census 8,567,485 10,679,345 24,654,825
Total sampled persons 3,211 2,417 8,285
Total sampled HHs 696 756 2,391
Sampled Muslim persons 2,097 844 2,003
Sampled Muslim HHs 382 252 555
5.2 22.3
8,873,472 38,483,967
5,333 15,868
1,560 3,732
1,234 3,807
329 845
10.2 1.3 4.6 7.5
17,557,809 2,158,684 7,893,065 12,971,152
6,199 1,946 12,807 12,808
1,526 489 3,537 3,265
1,070 271 1,584 825
244 59 388 180 (continued)
24 For instance, Hilal Ahmed (2008) argues the politics of Muslims has moved beyond protection of Urdu, minority character of Aligarh Muslim University, protection of Muslim Personal Law, and the protection of Wakf, demanding socioeconomic rights including reservation in certain states. For details, see https://www.india-seminar.com/2008/586/586_hilal_ahmed.htm.
23
Indian Muslims: Varieties of Discriminations and What Affirmative Action Can Do
Gujarat Andhra Pradesh Rajasthan Haryana MP Tamil Nadu All India
9.7 9.6
3.4 4.7
5,846,761 8,082,412
6,704 6,467
1,724 1,975
610 494
158 135
9.1 7.0 6.6 5.9 14.2
3.6 1.0 2.8 2.5 100.0
6,215,377 1,781,342 4,774,695 4,229,479 172,245,158
10,560 7,072 11,451 5,970 150,988
2,668 1,755 3,093 1,888 40,018
1452 525 633 294 19,169
325 121 167 85 4,562
563
References Abdul Khader Fakhri (2008) Dravidian Sahibs and Brahmin Maulanas: the politics of the Muslims of Tamil Nadu, 1930-1967. Manohar Publishers, New Delhi Ahmed H (2008) Debating Muslim political representation, India-Seminar, No. 586. https://www. india-seminar.com/2008/586/586_hilal_ahmed.htm. Accessed 21 Nov 2020 Alam T, Kumar S (2019) Social and economic status of backward Muslims in Uttar Pradesh: need for an inclusive policy? Soc Change 49(Issue) Ansari AH (2004) Rejoinder to Syed Shahabuddin: reservation for Muslim backwards. http://www. milligazette.com/Archives/2004/16-30Nov04-Print-Edition/163011200463.htm. Accessed 20 Dec 2020 Anwar K (2018) Muslims and the Dravidian movement, No. 708, India-Seminar, New Delhi. https://www.india-seminar.com/2018/708/708_s_anwar.htm. Accessed 12 Dec 2020 Asher S, Novosad P, Rafkin C (2020) Intergenerational mobility in India: new methods and estimates across time, space, and communities. Dartmouth College. https://www.dartmouth. edu/~novosad/anr-india-mobility.pdf. Accessed 12 Jan 2021 Dasgupta A (2009) On the margins: Muslims in West Bengal. Econ Polit Wkly 44(16) Fazal T (2014) ‘Development’ leaves Muslims out, BusinessLine. https://www.thehindubusi nessline.com/opinion/development-leaves-muslims-out/article22995299.ece. Accessed 11 Nov 2020 Gayer L, Jaffrelot C (eds) (2012) Muslims in Indian cities: trajectories of marginalisation, 1st edn. Oxford University Press, New Delhi Government of Mysore (1919) Leslie C. Miller Committee Report, Bangalore, 1919 Gupta S (2017) Following the Karpoori Thakur model, OBC sub-categorisation could be icing on BJP’s cake. https://theprint.in/opinion/following-karpoori-thakur-model-obc-subcategorisation-icing-bjps-cake/11509/. Accessed on 11th Jan 2021 Hasan J (2006) Reservation for Muslims, India-Seminar, Issue no. 549. https://www.india-seminar. com/2005/549/549%20zoya%20hassan.htm. Accessed 11 Dec 2020 Hasan J (2009) Muslim deprivation and the debate on equality’, India-Seminar, No. 602. http:// www.india-seminar.com/2009/602/602_zoya_hasan.htm. Accessed 21 Dec 2020 Jaffrelot C (2011) The Muslims of India: towards marginalisation? In: Jaffrelot C (ed) India since 1950. Society, politics, economy and culture. Yatra Books, New Delhi, pp 564–580 Jaffrelot, Kalaiyarasan (2020) Social injustice, the Indian Express. https://indianexpress.com/ article/opinion/columns/reservation-in-india-privatisation-push-nirmals-sitharaman-backwardcastes-6494931/ Jaffrelot C, Kalaiyarasan A (2021) How India has become an unequal republic, livemint, January 27, 2021. https://www.livemint.com/news/india/how-india-has-become-an-unequal-republic11611673600547.html Kalaiyarasan A, Vijayabaskar M (2021) The Dravidian model: interpreting political economy of Tamil Nadu. Cambridge University Press, Cambridge, MA
564
C. Jaffrelot and K. A.
Karmakar S (2019) Assamese Muslims seek ethnic tag and reservation. https://www.deccanherald. com/national/assamese-muslims-seek-ethnic-716777.html. Accessed 22 Dec 2020 Krishnan PS (2012) A Step for inclusive development: reservations Kundu A (2008) Report of the expert group on diversity index (2008). Ministry of Minority Affairs, Government of India, New Delhi Kundu A (2014) Report of the ‘Post Sachar Evaluation Committee’ submitted to the Ministry of Minority Affairs. Government of India, New Delhi L.G. Havanur Commission (1975) Karnataka backward classes commission. Government of Karnataka, Bangalore Menon M (2008) Equal opportunity commission: what, why and how? Report submitted to the Ministry of Minority Affairs, Government of India, New Delhi Rahman SA (2010) Hindu fundamentalists to oppose job quota proposals. The National, 21 January. https://www.thenational.ae/world/asia/hindu-fundamentalists-to-oppose-job-quota-proposals-1. 543198?videoId¼5587173110001. Accessed 11 Jan 2021 Rahman Khan K (1995) Report of the high-power committee on socio-economic and educational survey-1994 of religious minorities in Karnataka, Karnataka State Minorities Commission, Government of Karnataka, Bangalore Report of the National Commission for Religious and Linguistic Minorities, Vols. I, Ministry of Minority Affairs, Government of India. http://www.minorityaffairs.gov.in/sites/upload_files/ moma/files/pdfs/volume-1.pdf. Accessed 12 Dec 2020 Shahabuddin S (2004) Reservation for muslims is constitutional and socially necessary in national context, Milli Gazette. https://www.milligazette.com/Archives/2004/01-15Oct04-Print-Edition/ 011510200471.htm. Accessed 20 Dec 2020 Shaikh Z (2016) Ten years since Sachar report, Muslims still 3 per cent in IAS and IPS. https:// indianexpress.com/article/india/india-news-india/ten-years-since-sachar-report-muslims-still-3in-ias-ips-2982199/. Accessed 12 Aug 2020 The Constitution (Scheduled Castes) Order (1950). http://socialjustice.nic.in/writereaddata/ UploadFile/scorder1950636011777382153618.pdf. Accessed 11 Dec 2020 Thorat S, Attewell P (2007) The legacy of social exclusion: a correspondence study of job discrimination in India. Econ Polit Wkly 42(41):4141–4145
Part V Dimensions of Discrimination
Stereotypes and the Administration of Justice
24
Brendan O’Flaherty and Rajiv Sethi
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Police Stops and Searches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conflict, Cooperation, and Clearance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preemption and Murder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lethal Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discrimination in Other Venues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
568 569 571 575 578 580 585 587 588
Abstract
Stereotypes play a pervasive role in the administration of justice, influencing police stops and searches, the use of force, bail and sentencing decisions, juror behavior and selection, felony disenfranchisement, and attitudes toward mass incarceration. We provide a selective survey of the literature on stereotypes and criminal justice and discuss implications for witness cooperation, homicide clearance rates, and rates of murder victimization. Prepared for the Handbook of Economics of Discrimination and Affirmative Action, edited by Ashwini Deshpande. We thank Ashwini Deshpande, Pumla Gobodo-Madikizela, and Morgan Williams Jr. for extremely helpful comments and suggestions, and Julie Seager, Emily Cai, and Olivia Bobrownicki for research assistance. Sethi thanks the Radcliffe Institute for Advanced Study at Harvard University for fellowship support. B. O’Flaherty (*) Department of Economics, Columbia University, New York, NY, USA e-mail: [email protected] R. Sethi (*) Department of Economics, Barnard College, Columbia University, and the Santa Fe Institute, New York, NY, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_10
567
568
B. O’Flaherty and R. Sethi
Keywords
Stereotypes · Crime · Policing · Law
Introduction The State of Mississippi adopted a new constitution in 1890 that identified the conditions under which a citizen would be permitted to vote (emphasis added): Every male inhabitant of this State, except idiots, insane persons and Indians not taxed, who is a citizen of the United States, twenty-one years old and upwards, who has resided in this State two years. . . and who has never been convicted of bribery, burglary, theft, arson, obtaining money or goods under false pretenses, perjury, forgery, embezzlement or bigamy, and who has paid. . . all taxes which may have been legally required of him. . . is declared to be a qualified elector.
The list of crimes here is striking for what it excludes. Murder, rape, robbery, and aggravated assault, all violent felonies that today carry the harshest sentences, were not considered disqualifying. The reasons for this were subsequently explained by the Mississippi Supreme Court in the 1896 case Ratliff versus Beale as follows (emphasis added): It is in the highest degree improbable that there was not a consistent, controlling directing purpose governing the convention by which these schemes were elaborated and fixed in the constitution. Within the field of permissible action under the limitations imposed by the federal constitution, the convention swept the circle of expedients to obstruct the exercise of the franchise by the negro race. By reason of its previous condition of servitude and dependence, this race had acquired or accentuated certain peculiarities of habit, of temperament, and of character, which clearly distinguished it as a race from that of the whites,—a patient, docile people, but careless, landless, and migratory within narrow limits, without forethought, and its criminal members given rather to furtive offenses than to the robust crimes of the whites. Restrained by the federal constitution from discriminating against the negro race, the convention discriminated against its characteristics and the offenses to which its weaker members were prone.
That is, the rationale for stripping rights from those who had committed “furtive” crimes while declining to do so for those convicted of more serious and violent offenses was to selectively disenfranchise African American men, who had won the right to vote by 1869 under the 14th and 15th amendments to the United States constitution. Unable to exclude this group by name, the framers of the Mississippi constitution chose instead to rely on stereotypes that were prevalent at the time. By careful curation of exclusionary criteria, they “swept the circle of expedients to obstruct the exercise of the franchise.” And this was stated boldly and approvingly by the Mississippi Supreme Court itself. Stereotypes of this kind continue to be pervasive in the administration of justice in the United States, though seldom in such blatant form. Often their effects are
24
Stereotypes and the Administration of Justice
569
implicit, operating at a subconscious level, affecting the degree of suspicion, or threat perception, or the credibility of a witness, or the sense that a defendant is deserving of harsh punishment. This chapter provides a selective survey of the literature on stereotypes, policing, and justice, starting with an overview of insights from psychology on the nature and ubiquity of categorization and generalization. For a more expansive discussion of many issues raised here, see O’Flaherty and Sethi (2019).
Stereotypes In navigating the modern world, people are confronted with novelty on a daily basis. The vehicle they are driving behind, the vegetables displayed in a supermarket aisle, and the rectangular object on the hotel room wall are all new to them, in the sense that they have never been encountered before. And yet they are also familiar. People assign them to categories – such as trucks and carrots and television sets – and treat objects within a category as equivalent in essential respects. People get categorized in this way, too. The couple at the adjacent table in a restaurant is elderly, interracial, or foreign. The woman in a white coat is a doctor, or perhaps a scientist. The man wearing a yarmulke is Jewish, the woman in a headscarf a Muslim, and both are probably observant, and so on. This tendency to categorize and generalize has been recognized by psychologists as an indispensable cognitive tool for escaping paralysis in the face of novelty. As Jerome Bruner (1957) observed: “If we were to respond to each event as unique and to learn anew what to do about it or even what to call it, we would soon be swamped by the complexity of our environment.” The term “stereotype” was coined by Walter Lippmann (1922), who observed that “we do not first see, and then define, we define first and then see. . . when a system of stereotypes is well fixed, our attention is called to those facts which support it, and diverted from those which contradict.” Along similar lines, Gordon Allport (1954) argued that a prejudice “is actively resistant to all evidence that would unseat it.” That is, reliance on categorization and generalization need not lead to statistically accurate beliefs. Stereotype-confirming evidence tends to be more readily accepted than evidence that contradicts existing beliefs, causing such beliefs to become entrenched over time. Allport believed that from some stereotypes, “even a kernel of truth may be lacking, for they can be composed wholly of hearsay evidence, emotional projections, and fantasy.” In such cases, one might expect stereotypes to change quite sharply across generations, and indeed there is some evidence for this. For instance, a classic study published in 1933 found that among a group of Princeton students, all white and male at the time, Jews were thought to be shrewd and mercenary, the Chinese superstitious and sly, and African Americans lazy (Katz and Braly 1933). Future waves of the study tracked changes in beliefs in this population, which were significant. By the late 1960s, African Americans were primarily stereotyped as
570
B. O’Flaherty and R. Sethi
being musical and happy-go-lucky, and by 2001 aggressive and quick-tempered (Karlins et al. 1969; Madon et al. 2001). More recent work by psychologists has distinguished between explicit or conscious beliefs, which can potentially be elicited using surveys, and implicit biases that are concealed even from those who hold them. The latter can be uncovered by measuring reaction times on carefully constructed classification tasks known as implicit association tests. For instance, subjects may be asked to distinguish between Black and white faces, and between pleasant and unpleasant words, using a keyboard. In one treatment, Black faces are paired with pleasant words in the sense that the same keystroke is used when either one of these is depicted on a screen (a different keystroke being used when a white face or unpleasant word is shown). In another treatment, the pairing is reversed – the same keystroke used for Black faces is also used for unpleasant words. If subjects perform the task faster and with less error in the latter treatment, this is interpreted as an implicit association of Black faces with unpleasant characteristics and white with pleasant ones, relatively speaking. The divergence between explicit and implicit beliefs can be substantial. As Banaji and Greenwald (2013, p. 46) observe, implicit association tests hold up “a mirror in which many see a reflection that they do not recognize.” About four-fifths of white subjects exhibit prowhite bias in a standard version of the test, but so do about a third of Black subjects. Responses of police officers on the implicit association test mirror those of the general public, although there is no clear evidence to date that such biases are connected to discrimination in police behavior, or that implicit bias training is effective in reducing disparities in enforcement actions (Correll et al. 2007; James et al. 2016; Worden et al. 2020). Stereotypes as defined here refer to a universal process of categorization and inference that is devoid of normative significance. In popular discourse, however, as well as in some academic work, the term is used to refer to racist caricatures associated with beliefs about inferiority and diminished moral worth. This chapter uses the term stigma to describe such attitudes (Goffman 1963). They can give rise to malice and bias in the treatment of stigmatized groups, which is also relevant for policing and punishment (Loury 2002). While much remains unknown about the precise connection between stereotypes and behavior, it is clear that racial attitudes affect routine decisions when strangers interact, as they do with great frequency in the criminal justice system. Offenders confront victims, officers engage suspects, prosecutors interrogate witnesses, and judges and jurors decide the fate of defendants, often with no prior knowledge of the people with whom they are interacting. Stereotypes invariably condition behavior under such circumstances, and this includes stereotypes about the stereotypes held by others.
24
Stereotypes and the Administration of Justice
571
Police Stops and Searches Police-initiated contacts with civilians in the United States are extremely common. In 2018, for instance, there were more than 24 million contacts arising from traffic stops and another 3.5 million from street stops (Harrell and Davis 2020). Such stops can be made under a standard of “reasonable suspicion” that is less demanding than the “probable cause” standard required for arrests or more invasive searches. The use of this weaker standard was affirmed by the Supreme Court in Terry v. Ohio (1967), and these stops are sometimes called Terry stops. Even under this lower threshold of suspicion, however, an officer must be able to point to “specific and articulable facts” suggesting that a crime has been, or is about to be, committed. Markers of race, ethnicity, or gender cannot legally be used as independent proxies for these facts. Consider a law enforcement agency that seeks to maximize the recovery of contraband, such as illicit drugs or unlicensed firearms. (We shall argue below that the pursuit of such narrow objectives can compromise broader and more important goals such as public safety or community flourishing, but there is an influential empirical literature that asks whether or not agencies act as if contraband recovery is indeed their primary goal when making decisions to stop and search.) Suppose that officers believe that recovery rates could be increased by profiling drivers or pedestrians, in the sense of applying different standards of suspicion to members of different groups. This belief may be statistically accurate, or it may be mistaken, being influenced by stigma. Acting on the belief would be impermissible in either case, but an examination of recovery rates across different groups can allow us to determine which of these mechanisms is in operation. The use of differential contraband recovery rates to uncover bias in police stops and searches is known as the hit-rate test. The idea dates back to Becker (1957), who argued that those who discriminate in market transactions must “either pay or forfeit income for this privilege.” Applying this idea to discrimination in police stops, consider a biased officer who applies a lower standard of suspicion when detaining Black (relative to white) civilians. Among those stopped by this officer will be a disproportionate share of innocent Black civilians, from whom no illicit items will be recovered. That is, the contraband recovery rates will be lower among those who are subject to biased treatment. The logic here is explained by Ian Ayres (2002, pp. 133–134) as follows: “The ex post probability that a police search will uncover contraband or evidence of illegality is strong evidence of the average level of probable cause that police require before undertaking a search. . . Any finding that the police searches of individuals with a particular characteristic (such as minority status) induce a systematically lower probability of uncovering illegality suggests that police search criteria unjustifiably subject that class of individuals to the disability of being searched.” Notice that the hit-rate test cannot tell us whether or not profiling is occurring; it can only tell us whether or not the police are profiling in such a manner as to lower performance. Profiling in the service of a legitimate goal such as the recovery of illicit drugs or guns is referred to as strategic discrimination in legal scholarship, while profiling based on racial animus or exaggerated stereotypes (which
572
B. O’Flaherty and R. Sethi
compromises performance) is referred to as prejudiced discrimination (Kennedy 1998, p. 227). Prejudice encompasses both malice in preferences and bias in beliefs; these two channels are observationally equivalent in most of the data sets used to test for discrimination in the administration of justice (Arnold et al. 2020). If profiling were entirely strategic, then the same threshold of suspicion would be applied to all individuals, regardless of their group membership or identity. Some would exceed this threshold by a substantial amount, while others would barely meet it. A proper test for prejudiced discrimination involves comparisons of contraband recovery rates only for marginal individuals, who are on the cusp of meeting the threshold of suspicion. However, since degrees of suspicion are much harder to observe than markers of race, ethnicity, or gender, early applications of the test often simply compared contraband recovery rates averaged across all individuals who were stopped, and not just those on the threshold of suspicion. This is the so-called problem of inframarginality (Ayres 2002). In an early and influential application of the hit-rate test to police stops and searches, Knowles et al. (2001) examined data on Maryland State Police activity on the I-95 highway. The authors used an average (rather than marginal) hit-rate test for their analysis and did not find any statistically significant evidence of hit-rate inequality between Black and white motorists. They justified the use of average rather than marginal hit-rates by observing that optimizing behavior by police and citizens leads to equality between the two rates when there is no individual-specific information that could point to innocence or guilt (in which case profiling is the only basis for the search). Note that such searches – without reasonable articulable suspicion at the individual level – would not be permissible under the Terry versus Ohio standard (Bjerk 2007). The inference that Knowles et al. (2001) drew from this analysis is that profiling by troopers was motivated by maximizing contraband recovery, and not by bias or malice. That is, the behavior of police in this sample could be explained by strategic discrimination alone. The authors did, however, find that Latino drivers were stopped with greater frequency than the maximization of contraband recovery would imply. Recognizing not only the limitations of an analysis based on average hit-rates, but also the difficulty of computing marginal hit-rates from the available data, Anwar and Fang (2006) proposed a different test. Their model allows for variation across individuals in the degree of suspicious behavior, and hence inequality between average hit-rates across groups even when contraband recovery is the only goal. In the absence of bias or malice, however, the rank order of average hit-rates across groups ought not to vary with the social identity of the trooper making the decision to stop and search. This is what the authors find, using data from the Florida State Highway Patrol. That is, they cannot reject the hypothesis that racial profiling in stops and searches is motivated by maximizing rates of contraband recovery. Using a similar approach but data from a different period and agency, the Boston Police Department during 2001-2002 (Antonovics and Knight 2009) reached a different conclusion – they found that white officers were more likely than Black officers to search the vehicles of stopped Black motorists, even after controlling for
24
Stereotypes and the Administration of Justice
573
contextual factors such as driver age and gender, the infraction alleged to have led to the stop, the time of day, the location of the stop, and vehicle characteristics. More recent work has exploited the availability of detailed data on a national scale. Pierson et al. (2020) examined about a hundred million stops nationwide and implemented a “veil of darkness” test initially proposed by Grogger and Ridgeway (2006). That is, they compared the rate at which Black and white drivers were stopped in darkness and in daylight, using changes in daylight savings time to control for hour-of-day effects. They found that Black drivers were considerably more likely to be stopped than white drivers in daylight – evidence of prejudiced discrimination. Pierson et al. (2020) also estimated the likelihood of contraband possession at the level of individual drivers, based on the location and other characteristics of the stop, and looked at the probability that a motorist would be searched contingent on being stopped. They concluded that the threshold for searching Black and Hispanic drivers was generally lower than that for searching white drivers across the states and municipalities in their data. That is, officers were not simply engaged in strategic profiling in order to maximize contraband recovery; their behavior did appear to be motivated by bias or malice. Black drivers were doubly penalized – by being stopped more often in daylight, and by being searched with greater frequency, given other observable characteristics. Street stops are considerably less common nationwide than traffic stops but nevertheless very frequent in some neighborhoods of large cities. At its peak in 2011, there were close to 700,000 stops under New York City’s “stop-question-andfrisk” program, largely concentrated in a few precincts. Officers making such stops must provide a written rationale for them on a standard form, and entries on these forms can be used to compute hit-rates by comparing what was suspected with what was recovered. Goel et al. (2016) examined three million New York City stops over the period 2008–2012 and focused specifically on those in which the civilian was suspected of criminal possession of a weapon. Only about 3% of such stops resulted in actual weapon recovery, most of which were knives, and (average) hit-rates varied widely across groups: 2.5% for Blacks, 3.6% for Latinos, and 11% for whites. That is, searches of Black and Latino pedestrians were significantly less productive than those of whites, suggesting that the former were being stopped at much lower levels of suspicion. For reasons discussed above, however, average hit-rates need to be interpreted with caution on account of the problem of inframarginality. But the authors go beyond a simple hit-rate test and estimate the likelihood of contraband recovery at the level of individual stops, based on observables such as suspect characteristics and the timing and location of the stop. They find that for most stops the likelihood of weapon recovery is predictably low, and that if stops were better targeted, 90% of the weapons could have been recovered with less than 60% of stops. They also find that a disproportionate share of predictably unproductive stops involve Black civilians, so that “optimizing for weapons recovery would simultaneously bring more racial balance to stop-and-frisk.” This is clear evidence that discrimination is not simply
574
B. O’Flaherty and R. Sethi
strategic, consistent with earlier findings by Gelman et al. (2007). As it happens, the scope of the program has since been drastically reduced. In 2019, there were about 13,500 stops in all, and this decline has been accompanied by reduced racial disparities in exposure to stops and searches (MacDonald and Braga 2019). Racial profiling in police stops and searches, even if entirely strategic, has incentive effects. Those targeted with greater frequency will take evasive action, while others may respond by increasing criminal involvement. If the former effect is weak and the latter strong, then overall levels of crime may rise (Bjerk 2007). In addition, public perceptions of criminality are shaped by arrest and incarceration rather than offending, which is observable only if detected. As a result, racial stereotypes that are damaging to the targeted group can be reinforced by profiling (Harcourt and Ludwig 2006). While much of the economics literature on profiling in police stops and searches has focused on the issue of whether it is consistent with the maximization of contraband recovery or some other legitimate law enforcement goal, there is also evidence that in some municipalities, enforcement strategies are shaped by a desire to raise revenues for local expenditures. For instance, an investigation of the Ferguson Police Department conducted by the Department of Justice in 2015 uncovered a pattern of “unnecessarily aggressive and at times unlawful policing” involving “stops without reasonable suspicion and arrests without probable cause. . . retaliation for protected expression. . . and excessive force,” driven in part by a “focus on revenue rather than by public safety needs,” pressured by a city that “budgets for sizeable increases in municipal fines and fees each year, exhorts police and court staff to deliver those revenue increases, and closely monitors whether those increases are achieved” (United States Department of Justice 2015b). These revenue-motivated enforcement efforts involved highly selective targeting. The investigation found that certain offences were brought “almost exclusively against African Americans,” who were subjected to “routinely disrespectful treatment.” As a consequence, “police and municipal court practices both reflect and exacerbate existing racial bias, including racial stereotypes,” resulting in “deep mistrust between parts of the community and the police department, undermining law enforcement legitimacy” and ensuring that “the partnerships necessary for public safety are, in some areas, entirely absent.” This is evidence for the pervasive presence of stigma. In an entirely different context, Goncalves and Mello (2021) find evidence of discrimination in the speed reported on tickets issued by officers with the Florida Highway Patrol. Since there are discontinuous jumps in the punishments associated with different speeds, officers can exhibit leniency by reporting a speed slightly below the threshold that would trigger a mandated court appearance or points toward the suspension of a license. This “bunching” below thresholds reveals a selective and discriminatory exercise of discretion, favoring white relative to Black drivers, and is again revealing of stigma. Aggressive and frequent stops and searches that disproportionately target specific communities – whether motivated by the strategic pursuit of legitimate law
24
Stereotypes and the Administration of Justice
575
enforcement objectives, or by bias, malice, and illegitimate goals such as the raising of revenue – have a corrosive effect on police-community relations and public trust. This can inhibit witness cooperation, lead to more unsolved crimes, more unpunished violence, and more preemptive and retaliatory killing. These effects are considered next.
Conflict, Cooperation, and Clearance Just 3% of those stopped under suspicion of criminal possession of a weapon under New York City’s stop-and-frisk policy actually had an illicit weapon in their possession. Most of the recovered weapons were knives. There were over five million stops in total over the period 2002–2016, with firearms recovered in just 0.2% of cases. Only one-tenth of stopped individuals were whites, the remainder being predominantly Black or Latino. There is evidence that the targeting involved prejudiced discrimination, in the sense that contraband recovery rates could have been increased while reducing racial disparities in exposure to searches. But even if the profiling had been largely strategic, it would have led to the detention and discomfort of a lot of innocents, who would have been aware that they had been negatively stereotyped. Those stopped in this manner were mostly in high crime neighborhoods, and hence more likely than other city dwellers to be witnesses to serious crimes. But given their prior interactions with officers of the law, such individuals would understandably be reluctant to come forward as witnesses. And the absence of witness cooperation can leave the most serious crimes unsolved, which means that such crimes can be committed with impunity. There is considerable evidence that the rates at which murders are solved – the homicide clearance rate – vary sharply by neighborhood and victim characteristics. An examination by the Washington Post of over 50,000 murders over a 10-year period ending in 2018 found large variations in clearance rates across cities, and across neighborhoods within cities (Lowery et al. 2018). Areas with the lowest clearance rates had concentrated poverty and large minority populations, with the consequence that murders with Black or Latino victims had a much lower likelihood of being solved (46% and 48%, respectively) than murders with white victims (63%). The situation in Baltimore and Chicago was especially dire. Earlier local studies found similar effects. An examination of over 800 criminal homicides in the city of Columbus, Ohio, over the period 1984–1992 found that the lowest clearance rates were in predominantly Black neighborhoods of the city; in fact, differences in clearance rates by victim race could be attributed almost entirely to the location where the killing took place (Puckett and Lundman 2003). Similarly, an analysis by the Los Angeles Times of all 9442 “willful homicides” in Los Angeles County over 1990–1994 concluded that killings with white victims were 40% more likely to be solved than those with Black or Latino victims, even after accounting for such factors as whether the crime was robbery or gang related,
576
B. O’Flaherty and R. Sethi
the relationship between victim and offender (if known), and the age and gender of the victim (Rohrlich and Tulsky 1996; Lee 2005). There are two different narratives commonly advanced to account for these gaps in clearance rates, one focused on a lack of concern, the other on a lack of cooperation. The first view points to the fact that African Americans have been killed with impunity for much of American history. Prior to abolition, slave-owners were shielded from criminal liability for killing a slave “if death resulted from violence administered for the purpose of subduing resistance or imposing discipline” (Kennedy 1998, p. 30). The legacy of these norms and laws persisted after abolition, while “many whites still needed to learn that killing a black person amounted to murder” (Litwack 1980, p. 286). Mild sentences for criminal homicides persisted for decades, even when both victim and offender were Black. In Mississippi during the 1930s, such killings led to convictions in just 30% of cases, which the anthropologist Hortense Powdermaker attributed to a system that places less value on Black life and hence “exacts less punishment for destroying it” (Powdermaker 1939, pp. 173–174). Along similar lines, a report from Atlanta in the 1930s lamented the fact that murderers with Black victims “have been known to get off with two and three years, and in some cases with six months” (Forman Jr. 2017, p. 83). Thirty years later, the 1968 Kerner Commission quoted testimony from journalist David Hardy to the effect that if “a black man kills another black man, the law is generally enforced at its minimum” (United States Kerner Commission 1968, p. 161). And later still, in his ethnographic study of a neighborhood in Philadelphia, Elijah Anderson (2000, p. 321) encountered “a generalized belief that the police simply do not care about black people. . . If a black man shoots another black man the incident will not be thoroughly investigated.” This apparent lack of concern with the loss of Black life is also revealed in studies of sentencing, which have consistently found that the death penalty is more likely to be imposed in capital murder cases with white victims, even after taking other characteristics of the killing into account (Baldus et al. 1997). A 1990 meta-analysis conducted by the United States Government Accountability Office surveyed 28 prior studies, and found that in four-fifths of these, “those who murdered whites were. . . more likely to be sentenced to death than those who murdered blacks,” and that this finding “was remarkably consistent across data sets, states, data collection methods, and analytic techniques” (United States General Accounting Office 1990). Given this history, it is not surprising that the low homicide clearance rates in neighborhoods with large Black populations are attributed to a lack of concern among police and prosecutors. But law enforcement officials take a different view, arguing that incentives to solve murder cases are strong, but hampered by a lack of witness cooperation. For instance, responding to the clearance rate disparity in Los Angeles, law enforcement sources maintained that “whites tend to be killed in middle- and higher-income neighborhoods where witnesses are more inclined to cooperate with police” (Rohrlich and Tulsky 1996).
24
Stereotypes and the Administration of Justice
577
The authors of the study of clearance rate disparities in Columbus, Ohio, echoed this view: “When citizens do not trust police enough to tell them what they saw, what they know, and what they suspect, the odds of a clearance decrease. . . citizens in African American neighborhoods do not trust police because police have long brought a far more heavy-handed and intrusive style of policing to Black as compared to white communities. . . It thus is possible that citizens in African American neighborhoods provide detectives with less information, which necessarily translates into lower clearance rates” (Puckett and Lundman 2003). Very similar sentiments were expressed in response to the nationwide disparities uncovered by the Washington Post. In this case, officials are quoted as blaming low clearance rates on “frayed relationships with residents and on witnesses who are unwilling to cooperate,” making it “almost impossible to close cases in areas where residents already distrust police.” In fact, there is a feedback effect of low clearance rates on police-community relations, as “distrust deepens and killers remain on the street with no deterrent” (Lowery et al. 2018). The building of trust between the police and the communities they serve can be extremely hard to accomplish, in part because the absence of trust becomes crystallized in norms that condemn and punish witness cooperation. Messages scrawled on public-facing walls or embodied in song lyrics exhort people to “stop snitching” or warn that “snitches get stitches.” These messages are effective. Several brazen and public killings of well-known artists and their associates have never been solved, in large measure because no witness was willing to cooperate (Jacobs 2006). Such cases attract media attention but are relatively rare; most fatal shootings with reluctant witnesses are seen only in aggregate statistics. The fear of reprisal or community disapproval in these instances may be understandable but is a major impediment to homicide clearance. Sometimes the reluctance to testify can be motivated by a concern that the testimony itself may exculpate a police officer. A Department of Justice report on the 2014 killing of Michael Brown by Darren Wilson in Ferguson found many instances of witnesses unwilling to come forward publicly for this reason. Among these was a man (Witness 103) whose son had previously been shot by police, and who harbored “antipathy toward law enforcement.” This witness demanded confidentiality and specifically referenced antisnitching norms but offered an account of events that was broadly consistent with that of the officer. Another (Witness 108) refused to appear before the grand jury, declaring that “he would rather go to jail than testify,” for fear of “reprisal should the. . . neighborhood find out that his account corroborated Wilson” (United States Department of Justice 2015a). Witness cooperation is especially difficult to secure when violence involves gangs or organized crime, and the threat of retaliation is very real. This was true in the Little Sicily neighborhood of Chicago a century ago, where victims, offenders, and witnesses were all drawn from the same Italian community. Brazen, public, ritualistic killings believed to be linked to a secretive syndicate known as the Black Hand were almost impossible to solve (Adler 2006). The same is true of gang-related violence today. In a particularly dramatic case, 7-year-old Tajahnique Lee was struck in the face by a stray bullet while riding her bicycle in the presence of more than
578
B. O’Flaherty and R. Sethi
20 witnesses, including her own grandmother, in Trenton, NJ, in 2007. None were willing to speak with police about the incident, let alone identify the gang members involved in the exchange of fire (Kocieniewski 2007). Testimony often requires corroboration in order to be credible and effective, especially in criminal matters where a standard of reasonable doubt is in place. As a consequence, the decision to testify is characterized by strategic complementarity – a potential witness who feels that no others would come forward to offer corroboration would be especially unwilling to incur the risks and costs of cooperation with police. This is the case both because the testimony is less likely to be effective, and also because retaliation is more likely if the person implicated remains at large. Collective silence can thus become entrenched as an equilibrium phenomenon, difficult to dislodge even if police-community relations improve (O’Flaherty and Sethi 2010d). Certain crimes typically occur in private settings where the victim is also the only witness. In such cases, the victim may not even be aware of the existence of others who could corroborate testimony, and serial offenders can operate with impunity for decades. Larry Nassar, described at his trial as “possibly the most prolific serial child sex abuser in history,” had more than a hundred and fifty victims over two decades. They included four Olympic gold medalists, and children as young as 6 at the time of the assault (Hobson 2018). Only after one of his victims, Rachel Denhollander, made her allegations against him in public did others feel safe in coming forward. At Nassar’s sentencing in 2018, Judge Rosemarie Aquilina described Rachael Denhollander as “the bravest person I’ve ever had in my courtroom.” Being the first to come forward publicly with an accusation, especially against individuals in positions of power and prestige, carries enormous costs and risks. But once a credible accusation is made, several others often follow. Each witness willing to speak on the record lowers the risks and costs to others for doing the same. Witnesses serve a crucially important but largely uncompensated function in the administration of justice, and their cooperation can only be relied upon if the risks and costs they face are tolerable, and they have faith in the fundamental fairness of the system (Akerlof and Yellen 1994; Tyler 2006). The risks are especially great for those who witness violence tied to gangs or organized crime. And even when they do not expect physical retaliation, they may fear repudiation and ostracism. A climate of collective silence in which witness recalcitrance is expected and entrenched is not conducive to high rates of clearance for murder and other serious crimes. Under these conditions, murder rates can remain stubbornly high through the logic of preemption and retaliation. This logic is explored next.
Preemption and Murder Most serious personal crimes are motivated either by the desire to acquire property (robbery, burglary, and larceny, for example) or the intent to cause grievous bodily harm (rape and aggravated assault). Willful homicide is different. It can certainly be motivated by acquisitive concerns or the desire to inflict harm, but it can also be an
24
Stereotypes and the Administration of Justice
579
act of self-preservation, if one fears being killed by another. And as Schelling (1960) recognized, this makes the meaning of self-defense ambiguous. For instance, I may fear someone simply because I believe them to be fearful of me, and worry that they may act in self-defense. This effect makes fearful people dangerous and gives fearsome people reason to be afraid. As Ta-Nehisi Coates (2015) put it in his memoir, many of his peers as a teenager in Baltimore were “powerfully, adamantly, dangerously afraid.” And Adler (2006) discusses several striking examples from Chicago around the beginning of the twentieth century of women killing their husbands while in fear for their lives. Such fears can be greatly amplified when people feel that they can be killed with impunity. Many homicides originate in disputes over relatively trivial debts or insults that escalate if not quickly resolved. If firearm prevalence is known to be high, and it is believed that killers are unlikely to face legal consequences, the preemptive motive for killing becomes very powerful. And the knowledge that it is powerful for others gives it further impetus. High rates of homicide are sustained by – and contribute to – a climate of fear in the presence of impunity. Such conditions also lead to the proliferation of firearms, as greater gun possession by some induces others to seek guns. Like witness cooperation, preemptive killing and firearm acquisition are characterized by strategic complementarity: An increased willingness to arm oneself or engage in preemptive violence raises the incentives for others to do the same (O’Flaherty and Sethi 2010a, b). This logic can be seen in the consequences of “stand-your-ground” laws, which permit individuals to use deadly violence when they feel threatened in public places, even when they have an opportunity to retreat safely. The staggered introduction of such laws across states has allowed researchers to estimate their causal effects (Cheng and Hoekstra 2013; McClellan and Tekin 2017). This work has found that such laws result in increased rates of criminal homicide, even though they lead to the classification of some killings as justifiable (and hence not criminal). One plausible explanation for this is that such laws intensify the preemptive motive for killing. While low rates of homicide clearance make it more likely that disputes will escalate and result in fatal violence, this is especially the case for disputes involving illicit trades. Drug sales, prostitution, and gambling all involve prohibited transactions. Since parties cannot turn to the courts when a dispute arises, violence becomes the primary means of conflict resolution. As Miron (2004) puts it, “market participants substitute guns for lawyers in the resolution of disputes.” And these guns are often acquired through theft or unregulated secondary markets rather than through legal purchases from licensed dealers (Cook et al. 2007). Economists have looked to the prohibition era in the United States for evidence that outlawing transactions for which there is significant demand can lead to violence. A surge of killings in Chicago during this period has been attributed to “broken contracts that did not lend themselves to polite resolution” (Okrent 2010). But the nationwide evidence is mixed – by raising the price and lowering the availability of alcohol, some violent crimes related to intoxication may have been averted. The net effect of prohibition appears to have led to less exposure to violence at some points of the age distribution while increasing exposure at others; “banning
580
B. O’Flaherty and R. Sethi
the commercial sale of alcohol appears to have protected children and teens from homicide, but at the cost of exposing young adults to more violence” (Owens 2014). In a cleanly identified analysis of the effect of drug enforcement on homicide, Melissa Dell (2015) examined towns in Mexico that had close mayoral elections in 2006, with some ending up with mayors aligned with the Partido Accin Nacional (National Action Party) of President Felipe Caldern, while others were in opposition. The former towns received greater federal funding to combat trafficking and experienced a sharper increase in homicide relative to the (otherwise similar) latter towns. Along similar lines, Lindo and Padilla-Romo (2018) showed that the capture of a drug kingpin led to more murders, and Chimeli and Soares (2017) found that increased violence followed on the heels of prohibitions related to the extraction and sale of big-leaf mahogany in parts of Brazil. Although the demand for narcotics and other illicit goods is diffuse and spreads quite evenly across different groups in the population, the retail sale of such goods has historically been tightly concentrated in predominantly Black neighborhoods (Washington 1915; Myrdal 1944; United States Kerner Commission 1968; O’Flaherty and Sethi 2010c). Murder is similarly concentrated in physical and social space (Papachristos 2009; Papachristos et al. 2015; O’Flaherty and Sethi 2015). Low homicide clearance rates, in combination with concentrated street vice, create a combustible environment in which minor disputes can escalate quickly through the logic of fear and preemption. The very same logic of fear and preemption comes into play when armed officers of the law confront suspects they believe to be dangerous. This is considered next.
Lethal Force Under a federal standard established in the 1985 case Tennessee versus Garner, the use of deadly force by police is justified only if “the officer has probable cause to believe that the suspect poses a significant threat of death or serious physical injury to the officer or others.” Individual states have different standards for officers, some of which are considerably more permissive (Flanders and Welling 2015). But preemption is the only basis for a police homicide to be deemed justified under federal law. The degree to which fear of imminent harm is warranted is therefore a central concern in evaluating the use of deadly force from a legal perspective. Philando Castile was fatally shot by Officer Jeronimo Yanez following a traffic stop in Falcon Heights, Minnesota, in July 2016. This incident was remarkable in that the immediate aftermath of the shooting was broadcast live on social media by Castile’s girlfriend, Diamond Reynolds, who was a passenger in the vehicle. In the back seat of the vehicle was Reynolds’ four-year-old daughter. Castile had informed the officer that he was in possession of a licensed firearm, and was reaching for his wallet to produce his documents when Yanez panicked and fired multiple shots. According to the county attorney John Choi, the officer’s use of deadly force was not “objectively reasonable and necessary, given the totality of the circumstances” (Choi 2016). The “totality of the circumstances” in this case included
24
Stereotypes and the Administration of Justice
581
a four-year-old child in the rear of the vehicle, a young woman in the passenger seat, and a driver who had calmly reported the fact that he was in legal possession of a firearm. Had the passengers not been Black, would the level of fear have risen to the point of inducing the use of lethal force? Many observers, including Minnesota’s governor Mark Dayton, did not think so (Smith et al. 2016). Peter Moskos, a former Baltimore police officer and current member of the faculty at the John Jay College of Criminal Justice wrote on his blog: “Honestly, in this shooting, with this cop, in this locale, I don’t think there’s a chance in hell Castile would have been shot had he been white” (Moskos 2016). Whether or not the officer’s fear was objectively reasonable, it does appear to have been genuine. In an interview on the day following the shooting, Yanez repeatedly confessed to having been terrified: “I, was scared and I was, in fear for my life. . . I thought I was gonna die. . . I was scared. . . I don’t remember the first couple shots” (Choi 2016). Yanez was charged with second-degree manslaughter but was acquitted at trial. There have been many cases such as this, where the evidence is highly suggestive but by its very nature can never be decisive. In 2014, within months of each other, John Crawford III was fatally shot by police in an Ohio Walmart after having picked up an unpackaged pellet gun from the sporting goods section, 12-year-old Tamir Rice was shot and killed on a Cleveland playground while in possession of a replica pistol, and Levar Jones was shot in the hip at a South Carolina gas station while reaching into his vehicle to retrieve documents requested by an officer. The Jones shooting was captured on dashcam video, and the officer in question – Sean Groubert – was terminated by the South Carolina Department of Public Safety on the grounds that he had “reacted to a perceived threat where there was none” (South Carolina Department of Public Safety 2014). In fact, this characterization applies to all of these cases, and numerous others that have animated the Black Lives Matter movement. This does not mean that the officers in all cases were guilty of crimes beyond a reasonable doubt but does suggest that their subjective perceptions were shaped by stereotypes and stigma that led them to ignore other salient factors. That is, the anecdotal evidence suggests that Black civilians are subjected to disparate treatment at the hands of law enforcement officers. If this is indeed the case, then the phenomenon should reveal itself in aggregate data on police homicides. For instance, one might expect that Black civilians constitute a greater share of police homicide victims than they do of police-civilian encounters. An immediate difficulty faced by researchers investigating the presence of disparate treatment is the quality of the available data. There are over 16,000 active law enforcement agencies in the country, each with its own data collection and reporting protocols. Official counts of total lethal force incidents are based on voluntary reports and involve substantial undercounts. More complete data is available from media and online sources, including the Guardian, the Washington Post, Fatal Encounters, and Mapping Police Violence. Each source has its limitations, and all rely on public media reports for verification so
582
B. O’Flaherty and R. Sethi
Fig. 1 Black and white exposure to deadly force in states with at least 50,000 residents from each group. Marker colors indicate census regions, with the Northeast in green, Midwest in yellow, South in brown, and West in red
miss incidents that remain unreported even in local media. Fatal Encounters is the most inclusive and largely encompasses the other sources; this has been used in a number of recent studies examining variations in exposure across social groups and locations (Finch et al. 2019; Edwards et al. 2019; Schwartz and Jahn 2020). Based on these data sources, there are about 1100 on-duty police homicides annually in the United States, with little change from 1 year to the next during the past decade. This is vastly greater as a share of the population than one finds in other democracies with high levels of per capita income (Zimring 2017). About fifty officers a year are feloniously killed in the line of duty, a rate that is also significantly higher than in comparable countries. Furthermore, the killing of an officer by a Black suspect gives rise to substantial increases in the use of force against Black civilians for several days, while no such pattern is detected when the suspect is white or Latino (Legewie 2016). Nationwide, African Americans are more than twice as likely to face deadly force than whites. However, disparities in exposure in most states and cities are much greater than this. Figure 1 shows Black and white exposure to deadly force, measured in annual fatalities per million population, for all states with at least 50,000 residents in each group. With the exception of a handful of Southern states, Black exposure is more than twice the level of white exposure, and in about half the states the disparity exceeds four-to-one. Figure 1 reveals that the process of aggregation diminishes the observed disparity in the component states. In fact, it is possible for the aggregate disparity to be lower than that in any component location, and this is in fact true of certain combinations of
24
Stereotypes and the Administration of Justice
583
Fig. 2 Exposure to deadly force in the nine census divisions, relative to population, number of officers, and reported violent crime
states such as California and New York. This happens because the geographic distribution of exposure to deadly force is such that the highest levels of exposure arise in regions where the share of the Black population is relatively small. This can be seen in Fig. 2, which confirms that the highest rates of exposure to deadly force are in the Mountain and Pacific census divisions, and the lowest in the Middle Atlantic and New England divisions. Whether one measures exposure relative to total population, total number of officers, or total reported violent crime, the pattern is the same. Figures 1 and 2 show that inferences based on highly aggregated data can be profoundly misleading. Consider, for example, the following argument made by Mullainathan (2015): “For the entire country, 28.9% of arrestees were African American. This number is not very different from the 31.8% of police-shooting victims who were African Americans. If police discrimination were a big factor in the actual killings (and every place were roughly the same), we would have expected a larger gap between the arrest rate and the police-killing rate. This, in turn, suggests that removing police racial bias will have little effect on the killing rate.” This argument relies on and assumption that geographic variations in rates of policecivilian contact mirror those in exposure to lethal force. While we do not have reliable data on contacts at the necessary level of geographic detail, the incidence of reported violent crime is a possible proxy. And we can see from Fig. 2 that variations
584
B. O’Flaherty and R. Sethi
across regions in exposure to deadly force remain large even when relative to the local incidence of reported violent crime. Mullainathan’s argument also relies on the assumption that arrests of Black and white civilians involve encounters that are about equally threatening to officers on average. This, too, is a highly questionable assumption. As long as there is racial profiling in police stops and searches, there will be differences by race in the qualitative characteristics of different arrest pools. There are several examples of arrests in which race was a factor in the manner in which the interaction between officer and civilian evolved – the arrests of Henry Louis Gates on his front porch, that of Sandra Bland following a vehicle stop, and those of Rashon Nelson and Donte Robinson at a Philadelphia Starbucks. In none of these cases was there an objective threat to officers or others. Their addition to the pool of Black arrestees not only swelled the size of this pool, but also made it less dangerous to officers on average (O’Flaherty and Sethi 2019). This reasoning suggests that any test for disparate treatment must pay close attention to details of individual encounter characteristics, including those that did not lead to a weapon discharge. Fryer (2019) has attempted to do this, based on a large number of incident reports provided by the Houston Police Department. He uses these reports to construct two sets of observations – one group consisting of arrests for selected offenses (attempted capital murder of a police officer, aggravated assault on a public safety officer, resisting arrest, evading arrest, and interfering in an arrest) and the other consisting of incidents where an officer discharged a weapon. Fryer finds that African Americans constitute 58% of the arrest pool, and 52% of the shooting pool, so the raw data is not indicative of racial bias. But he also finds that the Black arrestees are different from white arrestees on the whole, and less threatening to officers. In particular, they “attacked or drew weapon” in a significantly smaller proportion of cases. He adds this information along with a wide range of other characteristics as controls and still finds no evidence of racial bias. This study has been challenged for not taking adequate account of bias in the data-generating process itself (Knox et al. 2020; Durlauf and Heckman 2020). As noted above, if the likelihood that any given interaction results in arrest depends directly on civilian race, arrest pools corresponding to different groups will not be equally threatening to officers on average. Unless objective threat levels can be directly observed, controlling for contextual factors will not be enough to result in valid inferences about discrimination conditional on contact. An additional concern is that in many egregious officer-involved shootings, including those discussed above, the victim was not engaged in any unlawful activity, and would not have been included in Fryer’s arrest pool. It is entirely possible that there is an absence of bias in genuinely threatening situations, but some objectively nonthreatening encounters involve disparate treatment. And finally, given the staggering geographical variation in exposure to deadly force, results based on a single agency, or even a handful of jurisdictions in a small set of states, cannot be assumed to have external validity.
24
Stereotypes and the Administration of Justice
585
Discrimination in Other Venues To this point, the focus has been on the incidence and consequences of discrimination in policing. There is a vast and growing literature on discrimination in other venues, largely beyond the scope of this chapter, which are mentioned briefly here. Arnold et al. (2018) have applied a version of the hit-rate test to uncover discrimination in bail decisions, identifying marginal Black and white defendants by exploiting quasirandom assignment of cases to judges of varying leniency. Using pretrial misconduct (rearrest while awaiting trial or failure to appear for a scheduled hearing) as the outcome measure, they find evidence of discrimination against Black defendants, especially by inexperienced judges. That is, the marginal released white defendant is more likely to be arrested for pretrial misconduct than the marginal released Black defendant. This approach tests for disparate treatment arising from malice or bias; Arnold et al. (2020) find that statistical discrimination is also a significant contributor to racial disparities in bail decisions in New York City. McIntyre and Baradaran (2013) argue that disparities in bail decisions and amounts can be largely attributed to statistical discrimination, based on the perceived probability of arrest while awaiting trial. Kleinberg et al. (2018) show that the decisions of judges regarding the release of defendants have low predictive accuracy regarding pretrial misconduct in the following sense: Release based on an algorithm using only data from case files could achieve significantly lower rates of pretrial misconduct at current rates of release, or significantly higher rates of release at current levels of pretrial misconduct. Importantly, release decisions based on the algorithm would result in reduced racial disparities, suggesting that judges are not just poor predictors of the likelihood of pretrial misconduct, but that their errors are systematically biased against Black defendants. Bail decisions have significant long-term effects, since pretrial detention increases the likelihood of eventual conviction and the length of the resulting sentence (Leslie and Pope 2017; Dobbie et al. 2018; Aizer and Doyle 2015). In addition, there are large differences across judges in the rate at which they incarcerate Black defendants relative to white (Abrams et al. 2012). Given random assignment of cases to judges, this variation is evidence of disparate treatment, although no inference can be made from the variation alone about the direction of the bias. Stereotypes have long affected the behavior and selection of jurors. As Kennedy (1998, pp. 218–9) has observed, “many attorneys, prosecutors as well as defense counsel, racially discriminate in their deployment of peremptory challenges because they reasonably believe that doing so redounds to the benefit of the side they represent. . . racially discriminatory conduct might well reflect intelligent strategic decision-making.” All-white juries continued to be the norm in the American South for decades after the passage of the reconstruction amendments made racial discrimination in jury selection formally unlawful. This allowed prosecutors to obtain quick convictions against Black defendants, especially for alleged crimes against whites, even in the absence of meaningful evidence. In effect, they were exploiting and leveraging racial stigma within the general population.
586
B. O’Flaherty and R. Sethi
Discrimination in jury selection is especially tempting for prosecutors when a case hinges on testimony by a police officer, or an officer is the defendant. While race cannot be used explicitly to strike a juror from the pool, proxies such as prior incarceration can serve the purpose. And given continuing racial segregation in residential patterns, venue selection has direct and predictable effects on the composition of a jury. Significant disparities have been found to exist in the enforcement of drug laws at both state and federal levels. African Americans are heavily overrepresented among those arrested for marijuana possession, despite negligible differences across groups in marijuana use (Bunting et al. 2013). Similarly, penalties for trafficking in crack cocaine, imposed almost exclusively on Black defendants, have been significantly harsher than those for trafficking in powder cocaine (Sklansky 1995). The “war on drugs” is at least partially responsible for staggering racial differences in incarceration rates (Alexander 2010), although sentencing disparities across a broad range of violations are implicated (Rehavi and Starr 2014). More generally, prosecutorial discretion, truth-in-sentencing laws, and mandatory minimum sentences have all contributed to mass incarceration, as well as to disparate impact (Pfaff 2017). Disparities in incarceration rates have been tied directly to strategic discrimination in the labor market, driven by aversion among employers to hire individuals with criminal records (Holzer et al. 2006). This raises the possibility that regulations preventing employers from obtaining timely information about prior incarceration – such as “ban-the-box” policies – can hurt those who are incorrectly stereotyped as having a criminal record. There is evidence that such policies do indeed have this effect (Agan and Starr 2018; Craigie 2020; Doleac and Hansen 2020). Racial disparities in incarceration have repercussions that touch every aspect of social life. Charles and Luoh (2010) identify effects on marriage, finding that high male incarceration affects both the likelihood of marriage for women and the bargaining power that women have in their relationships with men. Since marriage remains largely endogamous in the United States, racial disparities in incarceration rates are reflected in disparities in marital rates and conditions. Perhaps the most insidious of disparities exists in the reaction of the general public to high rates of imprisonment. Loury (2002) has argued that the widespread acceptance of mass incarceration has been possible precisely because of its racial character; were there parity across groups at current levels of confinement, it would be treated as a national emergency and a broad indictment of society writ large. That is, the roots of mass incarceration lie in racial stigma. In fact, recent modest efforts at decarceration have come at a time when the share of whites in prison has been rising, especially among women (Sethi 2020). This is part of a more general deterioration of health and life outcomes among white Americans, including an increase in midlife mortality (Case and Deaton 2015). Public perceptions about who is deserving of sympathy rather than harsh punishment also reveal cleavages along racial lines. As the opioid epidemic started to ravage a number of white communities, drug users began to be viewed increasingly as “patients in need of our help and understanding, rather than criminals deserving scorn and incarceration” (Hart 2017). The very same activities that are viewed as
24
Stereotypes and the Administration of Justice
587
criminal or malignant when engaged in by African Americans are seen as benign or recreational when adopted by whites, resulting in large and sustained drug enforcement disparities.
Discussion With few exceptions, economists tend to view racial profiling in the service of goals such as crime reduction of contraband recovery as legitimate, at least when compared with discrimination based on bias or malice. But the law does not make such a clear distinction between strategic and prejudiced discrimination, and neither do most individuals subjected to profiling. The harm imposed on such persons needs to be accounted for when evaluating policy, under a broad conception of human welfare. As Claude Steele (2010) has observed, using the example of de jure segregation at swimming pools, the inconvenience caused by restrictions is much less upsetting than the reason for the inconvenience. Or as Justice Harlan argued in his lone dissent in Plessy versus Ferguson, “arbitrary separation of citizens, on the basis of race, while they are on a public highway, is a badge of servitude wholly inconsistent with the civil freedom and the equality before the law established by the Constitution.” Police stops and searches in the absence of reasonable articulable suspicion at the individual level are not just unlawful, they impose a genuine psychological burden that a proper welfare analysis ought to recognize. In addition, serious policy analysis must take account of general equilibrium effects. Aggressive policing involving large-scale stops and searches based on low thresholds for suspicion entraps many innocents, damages public trust, makes witnesses reluctant to cooperate and criminal homicides harder to solve, and raises incentives for preemptive and retaliatory violence. Underprotection is a direct consequence of overpolicing (Leovy 2015; O’Flaherty and Sethi 2019; Chalfin et al. 2020). Finally, policies aimed at reducing the incidence of lethal force in policing need to look beyond the culpability of the officer and examine “the complex organizational processes that recruited, hired, trained, supervised, disciplined, assigned, and dispatched the shooter before anyone faced a split second decision to shoot” (Sherman 2018, p. 434). Selection, training, leadership, and organizational culture are key factors in policing, and large geographic variations in performance suggest that major changes are possible through the proliferation of better practices. Whether this can be done through incremental reform or only by dismantling and reassembling organizations remains an open question. Examples of large-scale transformation, as in the city of Camden, deserve close examination (Goldstein and Armstrong 2020). The future of the American justice system is at a crossroads. Mass actions following the killing of George Floyd were among the largest in recent history, with more than a half million protestors in over 500 towns and cities at peak in June 2020. The time for change is ripe, but the path ahead remains unclear, and there is
588
B. O’Flaherty and R. Sethi
much that careful, evidence-based, and historically informed scholarship on these issues can contribute at this critical time.
References Abrams DS, Bertrand M, Mullainathan S (2012) Do judges vary in their treatment of race? J Leg Stud 41(2):347–383 Adler JS (2006) First in violence, deepest in dirt: homicide in Chicago, 1875–1920. Harvard University Press Agan A, Starr S (2018) Ban the box, criminal records, and racial discrimination: a field experiment. Q J Econ 133(1):191–235 Aizer A, Doyle JJ (2015) Juvenile incarceration, human capital, and future crime: evidence from randomly assigned judges. Q J Econ 130(2):759–803 Akerlof G, Yellen JL (1994) Gang behavior, law enforcement, and community values. In: Aaron HJ, Mann TE, Taylor T (eds) Values and public policy. Brookings Institution, Washington, DC Alexander M (2010) The New Jim Crow: mass incarceration in the age of colorblindness. New Press, New York Allport GW (1954) The nature of prejudice. Perseus, Cambridge Anderson E (2000) Code of the street: decency, violence, and the moral life of the inner city. WW Norton & Company Antonovics K, Knight BG (2009) A new look at racial profiling: evidence from the Boston police department. Rev Econ Stat 91(1):163–177 Anwar S, Fang H (2006) An alternative test of racial prejudice in motor vehicle searches: theory and evidence. Am Econ Rev 96(1):127–151 Arnold D, Dobbie W, Yang CS (2018) Racial bias in bail decisions. Q J Econ 133(4):1885–1932 Arnold D, Dobbie WS, Hull P (2020) Measuring racial discrimination in bail decisions. Technical report. National Bureau of Economic Research Ayres I (2002) Outcome tests of racial disparities in police practices. Justice Res Policy 4(1-2):131–142 Baldus DC, Woodworth G, Zuckerman D, Weiner NA (1997) Racial discrimination and the death penalty in the post-Furman era: an empirical and legal overview with recent findings from Philadelphia. Cornell Law Rev 83:1638 Banaji MR, Greenwald AG (2013) Blindspot: hidden biases of good people. Bantam Becker GS (1957) The economics of discrimination. University of Chicago Press Bjerk D (2007) Racial profiling, statistical discrimination, and the effect of a colorblind policy on the crime rate. J Public Econ Theory 9(3):521–545 Bruner JS (1957) Going beyond the information given. Contemp Approach Cognit 1(1):119–160 Bunting WC, Garcia L, Edwards E (2013) The war on marijuana in black and white. American Civil Liberties Union Case A, Deaton A (2015) Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century. Proc Natl Acad Sci 112(49):15078–15083 Chalfin A, Hansen B, Weisburst EK, Williams MC (2020) Police force size and civilian race. National Bureau of Economic Research Working Paper 28202 Charles KK, Luoh MC (2010) Male incarceration, the marriage market, and female outcomes. Rev Econ Stat 92(3):614–627 Cheng C, Hoekstra M (2013) Does strengthening self-defense law deter crime or escalate violence? Evidence from expansions to castle doctrine. J Hum Resour 48(3):821–854 Chimeli AB, Soares RR (2017) The use of violence in illegal markets: evidence from mahogany trade in the Brazilian Amazon. Am Econ J Appl Econ 9(4):30–57 Choi JJ (2016) State of Minnesota v. Jeronimo Yanez: Press conference remarks as prepared for delivery, November 16, 2016
24
Stereotypes and the Administration of Justice
589
Coates T-N (2015) Between the world and me. Spiegel & Grau Cook PJ, Ludwig J, Venkatesh S, Braga AA (2007) Underground gun markets. Econ J 117(524): F588–F618 Correll J, Park B, Judd CM, Wittenbrink B, Sadler MS, Keesee T (2007) Across the thin blue line: police officers and racial bias in the decision to shoot. J Pers Soc Psychol 92(6):1006 Craigie T-A (2020) Ban the box, convictions, and public employment. Econ Inq 58(1):425–445 Dell M (2015) Trafficking networks and the Mexican drug war. Am Econ Rev 105(6):1738–1779 Dobbie W, Goldin J, Yang CS (2018) The effects of pretrial detention on conviction, future crime, and employment: evidence from randomly assigned judges. Am Econ Rev 108(2):201–240 Doleac JL, Hansen B (2020) The unintended consequences of “ban the box”: statistical discrimination and employment outcomes when criminal histories are hidden. J Labor Econ 38(2): 321–374 Durlauf SN, Heckman JJ (2020) An empirical analysis of racial differences in police use of force: a comment. J Polit Econ 128(10):3998–4002 Edwards F, Lee H, Esposito M (2019) Risk of being killed by police use of force in the United States by age, race–ethnicity, and sex. Proc Natl Acad Sci 116(34):16793–16798 Finch BK, Audrey B, Brian Burghart D, Johnson R, Klinger D, Thomas K (2019) Using crowdsourced data to explore police-related-deaths in the United States (2000–2017): the case of fatal encounters. Open Health Data 6(1) Flanders C, Welling J (2015) Police use of deadly force: state statutes 30 years after Garner. Saint Louis Univ Public Law Rev 35:109 Forman J Jr (2017) Locking up our own: crime and punishment in Black America. Farrar, Straus and Giroux Fryer RG (2019) An empirical analysis of racial differences in police use of force. J Polit Econ 127(3):1210–1261 Gelman A, Fagan J, Kiss A (2007) An analysis of the New York City police department’s “stopand-frisk” policy in the context of claims of racial bias. J Am Stat Assoc 102(479):813–823 Goel S, Rao JM, Shroff R (2016) Precinct or prejudice? Understanding racial disparities in New York City stop-and-frisk policy. Ann Appl Stat 10(1):365–394 Goffman E (1963) Stigma: notes on the management of spoiled identity. Simon and Schuster Goldstein J, Armstrong K (2020) Could this city hold the key to the future of policing in America? New York Times, July 12, 2020 Goncalves F, Mello S (2021) A few bad apples? Racial bias in policing. Am Econ Rev 111(5): 1406–1441 Grogger J, Ridgeway G (2006) Testing for racial profiling in traffic stops from behind a veil of darkness. J Am Stat Assoc 101(475):878–887 Harcourt BE, Ludwig J (2006) Broken windows: new evidence from New York City and a five-city social experiment. Univ Chic Law Rev 73:271 Harrell E, Davis E (2020) Contacts between police and the public, 2018–statistical tables. Bureau of Justice Statics Report, NCJ, 255730 Hart CL (2017) The real opioid emergency. The New York Times Hobson W (2018) Twenty years of failure: many groups missed chances to stop Larry Nassar. Washington Post, 2018 Holzer HJ, Raphael S, Stoll MA (2006) Perceived criminality, criminal background checks, and the racial hiring practices of employers. J Law Econ 49(2):451–480 Jacobs A (2006) When rappers keep their mouths shut tight. New York Times, 2006 James L, James SM, Vila BJ (2016) The reverse racism effect: are cops more hesitant to shoot black than white suspects? Criminol Public Policy 15(2):457–479 Karlins M, Coffman TL, Walters G (1969) On the fading of social stereotypes: Studies in three generations of college students. J Pers Soc Psychol 13(1):1–16 Katz D, Braly K (1933) Racial stereotypes of one hundred college students. J Abnorm Soc Psychol 28(3):280 Kennedy R (1998) Race, Crime, and the Law. Vintage
590
B. O’Flaherty and R. Sethi
Kleinberg J, Lakkaraju H, Leskovec J, Ludwig J, Mullainathan S (2018) Human decisions and machine predictions. Q J Econ 133(1):237–293 Knowles J, Persico N, Todd P (2001) Racial bias in motor vehicle searches: theory and evidence. J Polit Econ 109(1):203–229 Knox D, Lowe W, Mummolo J (2020) Administrative records mask racially biased policing. Am Polit Sci Rev 114(3):619–637 Kocieniewski D (2007) A little girl shot, and a crowd that didn’t see. New York Times, July 9, 2007 Lee C (2005) The value of life in death: multiple regression and event history analyses of homicide clearance in Los Angeles County. J Crim Just 33(6):527–534 Legewie J (2016) Racial profiling and use of force in police stops: how local events trigger periods of increased discrimination. Am J Sociol 122(2):379–424 Leovy J (2015) Ghettoside: a true story of murder in America. One World/Ballantine Leslie E, Pope NG (2017) The unintended impact of pretrial detention on case outcomes: evidence from New York City arraignments. J Law Econ 60(3):529–557 Lindo JM, Padilla-Romo M (2018) Kingpin approaches to fighting crime and community violence: evidence from Mexico’s drug war. J Health Econ 58:253–268 Lippmann W (1922) Public Opinion. Harcourt, Brace and Company Litwack LF (1980) Been in the storm so long: the aftermath of slavery. Vintage Loury GC (2002) The anatomy of racial inequality. Harvard University Press Lowery W, Kelly K, Mellnik T, Rich S (2018) Where killings go unsolved. The Washington Post, June 6, 2018 MacDonald J, Braga AA (2019) Did post-Floyd et al. reforms reduce racial disparities in NYPD stop, question, and frisk practices? An exploratory analysis using external and internal benchmarks. Justice Q 36(5):954–983 Madon S, Guyll M, Aboufadel K, Montiel E, Smith A, Palumbo P, Lee J (2001) Ethnic and national stereotypes: the Princeton trilogy revisited and revised. Personal Soc Psychol Bull 27(8): 996–1010 McClellan C, Tekin E (2017) Stand your ground laws, homicides, and injuries. J Hum Resour 52(3): 621–653 McIntyre F, Baradaran S (2013) Race, prediction, and pretrial detention. J Empir Leg Stud 10(4): 741–770 Miron JA (2004) Drug war crimes: the consequences of prohibition. Independent Institute Moskos P (2016) Philando Castile. Cop in the Hood, July 7, 2016 Mullainathan S (2015) Police killings of blacks: here is what the data say. New York Times, October 18, 2015 Myrdal G (1944) An American Dilemma: the Negro problem and modern democracy. Harper O’Flaherty B, Sethi R (2010a) Homicide in black and white. J Urban Econ 68(3):215–230 O’Flaherty B, Sethi R (2010b) Peaceable kingdoms and war zones: preemption, ballistics and murder in Newark. In: Di Tella R, Edwards S, Schargrodsky E (eds) The economics of crime: lessons for and from Latin America. University of Chicago Press, pp 305–353 O’Flaherty B, Sethi R (2010c) The racial geography of street vice. J Urban Econ 67(3):270–286 O’Flaherty B, Sethi R (2010d) Witness intimidation. J Leg Stud 39(2):399–432 O’Flaherty B, Sethi R (2015) Urban crime. Handbook of Regional and Urban Economics:1519–1621 O’Flaherty B, Sethi R (2019) Shadows of doubt: stereotypes, crime, and the pursuit of justice. Harvard University Press Okrent D (2010) Last call: the rise and fall of prohibition. Simon and Schuster Owens EG (2014) The American temperance movement and market-based violence. Am Law Econ Rev 16(2):433–472 Papachristos AV (2009) Murder by structure: dominance relations and the social structure of gang homicide. Am J Sociol 115(1):74–128 Papachristos AV, Wildeman C, Roberto E (2015) Tragic, but not random: the social contagion of nonfatal gunshot injuries. Soc Sci Med 125:139–150
24
Stereotypes and the Administration of Justice
591
Pfaff J (2017) Locked. In: The true causes of mass incarceration–and how to achieve real reform. Basic Books Pierson E, Simoiu C, Overgoor J, Corbett-Davies S, Jenson D, Shoemaker A, Ramachandran V, Barghouty P, Phillips C, Shroff R, Goel S (2020) A large-scale analysis of racial disparities in police stops across the United States. Nat Hum Behav 4(7):736–745 Powdermaker H (1939) After freedom. Viking Press Puckett JL, Lundman RJ (2003) Factors affecting homicide clearances: multivariate analysis of a more complete conceptual framework. J Res Crime Delinq 40(2):171–193 Rehavi MM, Starr SB (2014) Racial disparity in federal criminal sentences. J Polit Econ 122(6): 1320–1354 Rohrlich T, Tulsky F (1996) Not all L.A. murder cases are equal. Los Angeles Times, December 3, 1996 Schelling TC (1960) The strategy of conflict. Harvard University Press Schwartz GL, Jahn JL (2020) Mapping fatal police violence across US metropolitan areas: overall rates and racial/ethnic inequities, 2013–2017. PLoS One 15(6):e0229686 Sethi R (2020) Crime and punishment in a divided society. In: Allen D, Somanathan R (eds) Difference without domination: pursuing justice in diverse democracies. University of Chicago Press, pp 93–114 Sherman LW (2018) Reducing fatal police shootings as system crashes: research, theory, and practice. Annu Rev Criminol 1:421–449 Sklansky DA (1995) Cocaine, race, and equal protection. Stanford Law Rev:1283–1322 Smith M, Capecchi C, Furber M (2016) Peaceful protests follow Minnesota governors call for calm. New York Times, 2016 South Carolina Department of Public Safety (2014) Second follow up: trooper involved shooting in Richland County on September 4, 19, 2014 Steele CM (2010) Whistling Vivaldi: how stereotypes affect us and what we can do. WW Norton & Company Tyler TR (2006) Why people obey the law. Princeton University Press United States Department of Justice (2015a) Report regarding the shooting death of Michael Brown by Ferguson, Missouri Police Officer Darren Wilson, Washington, DC United States Department of Justice (2015b) Investigation of the Ferguson Police Department, Washington, DC United States General Accounting Office (1990) Death penalty sentencing: research indicates pattern of racial disparities, Washington, DC United States Kerner Commission (1968) Report of the National Advisory Commission on Civil Disorders. US Government Printing Office Washington BT (1915) My view of segregation laws. The New Republic Worden RE, McLean SJ, Engel RS, Cochran H, Corsaro N, Reynolds D, Najdowski CJ, Isaza GT (2020) The impacts of implicit bias awareness training in the NYPD. The John F. Finn Institute for Public Safety/University of Cincinnati Center for Police Research and Policy Zimring FE (2017) When police kill. Harvard University Press
Evidence of Covariation Between Regional Implicit Bias and Socially Significant Outcomes in Healthcare, Education, and Law Enforcement
25
Tessa E. S. Charlesworth and Mahzarin R. Banaji
Contents A Brief Definition of Implicit and Explicit Attitudes and Beliefs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Relationship of Implicit Bias and Individual Discriminatory Behavior . . . . . . . . . . . . . . . . . . . The Relationship of Implicit Bias and Systemic Discriminatory Behaviors: An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disparities in Education and Opportunity: Standardized Testing, School Discipline, and Economic Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disparities in Healthcare: Medicaid Spending, Death Rates, and Infant Health Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disparities in Policing: Lethal Force and Traffic Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Examining Implicit Bias as the Outcome Explained by Systemic Predictors . . . . . . . . . . . . . . . . . Implications for Understanding Implicit Bias and Systemic Behaviors . . . . . . . . . . . . . . . . . . . . . . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
596 597 597 601 604 605 606 608 608 609 609
Abstract
When viewing the state of the world today, social and behavioral scientists face a puzzling inconsistency: how can it be that evidence of discrimination persists in all significant aspects of life – from housing and jobs to healthcare and law enforcement – even though individuals and institutions adamantly stand for equality? Over the past two decades, research has demonstrated that at least
This research was supported by the Mind Brain Behavior Interfaculty Initiative award, the Hao Family Inequality in America award, and the Foundations of Human Behavior Initiative award to M. R. Banaji and T. E. S. Charlesworth. The authors declare no competing financial interests. Correspondence concerning this chapter should be addressed to Tessa Charlesworth, Department of Psychology, Harvard University, Cambridge, MA 02138, at [email protected] T. E. S. Charlesworth (*) · M. R. Banaji (*) Harvard University, Cambridge, MA, USA e-mail: [email protected]; [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_7
593
594
T. E. S. Charlesworth and M. R. Banaji
part of the answer to this puzzle can be attributed to the implicit nature of biases – attitudes, beliefs, and identities that are less conscious and controllable but that nevertheless exist and shape behavior. Today, it is taken as a given that evidence is strong and substantial for the presence of implicit bias in the minds and behaviors of individuals. This chapter, however, reviews an emerging body of research that uses large-scale, aggregated data across millions of tests of implicit attitudes and beliefs to understand outcomes of socially significant systemic behaviors ranging from the police use of lethal force to infant healthcare to school suspensions and discipline. Methodologically, the studies quantify social and psychological processes acting in the real world and introduce data of unprecedented scope across geography and time. Theoretically, both the approach and findings of this research underscore a new meaning of the term systemic discrimination that recognizes how implicit bias both shapes and is shaped by broad structural systems and outcomes. Keywords
Education · Discrimination · Healthcare · Implicit attitudes · Implicit bias · Implicit stereotypes · Policing · Systemic discrimination For students of the social sciences, the term “bias” is commonly used to capture at least two distinct meanings. In one sense, a bias refers to a behavior or cognition that deviates from accuracy. If two individuals of differing ethnicity (e.g., Asian and White) are both US citizens, both born in the United States, and both live in the United States, then it is accurate to consider them both to be American. Yet measures of implicit beliefs indicate that one group is perceived to be more American: White Americans are more “American” than Asian Americans (Devos and Banaji 2005; Devos and Mohamed 2014). Such a belief can therefore be said to be biased in the sense that it deviates from accuracy. A second meaning of bias refers to thoughts or behavior that deviates from one’s own consciously stated values. If a company’s stated, conscious values dictate that the best candidate should be hired and yet the evidence shows that male candidates are repeatedly hired even when female candidates are equally qualified (Moss-Racusin et al. 2012), then that behavior is taken to be biased because it is inconsistent with stated ideals. Both forms of bias – deviations from accuracy or from ideals – are known to be prevalent in shaping cognitions and behaviors (Pager 2007). First, evidence of bias can be found in experimental studies, in which researchers systematically manipulate some expected causal effect on behavior (e.g., the gender of a candidate) and then assign participants to one of several experimental conditions. As an example, participants (drawn from a population of scientists) were randomly assigned to see either a male or female applicant profile for a science lab manager position (MossRacusin et al. 2012). Systematically, participants were more likely to recommend that the male applicant be hired, given a higher salary, and given more mentorship compared to the identical and equally qualified female applicant. Similarly,
25
Evidence of Covariation Between Regional Implicit Bias and. . .
595
participants in another study were more likely to hire a male candidate to complete an arithmetic task over an identical female candidate and, most surprisingly, even continued to select the male candidate after receiving evidence that women outperform men on the task (Reuben et al. 2014). A complementary method – audit studies, also called “experimental field studies,” in which real-world respondents do not know they are in a study and may therefore be less prone to experimenter demand effects (Pager 2007) – has also been used to reveal biased behavior toward targets differing on race, sexual orientation, criminal history, and more. For instance, counselling professionals were more likely to initiate further counselling treatment for a potential client if the hypothetical client was White rather than Black (Shin et al. 2016), and hiring managers were more likely to call back an applicant if the applicant was White rather than Black (Bertrand and Mullainathan 2004). Applications from ostensibly heterosexual job candidates sent to real hiring managers across 1,769 jobs were more successful than applications from identically qualified but openly gay candidates (Tilcsik 2011). And hypothetical job candidates that listed a previous criminal record were less likely to receive a call back than identical candidates without a criminal record, especially if the candidate with a criminal record was Black (Pager 2003). Because experiments and audit studies control for all features of the hypothetical candidates except the candidate’s salient identity (e.g., their race, gender, or sexual orientation), the evidence conclusively shows the existence of biased behaviors. That is, despite respondents stated ideals for equality or for accuracy, respondents in all studies were more likely to treat a candidate favorably (e.g., hire them, offer them medical treatment) when the candidate came from a typically preferred, high-status, or dominant social group (e.g., male, White, straight), relative to a typically dispreferred, low-status, or minority social group (e.g., female, Black, gay). Crucially, this evidence for biased behavior and discrimination is not an isolated phenomenon among a few “bad apples” – a common explanation that suggests bias would be solved if the few “bad apples” were rooted out. Although individual differences no doubt exist in the frequency or severity of discriminatory behavior, the bulk of the evidence shows that such behavior is far more pervasive than explained by the “few bad apples” account. Both men and women favored the male over female candidate for a STEM job (Moss-Racusin et al. 2012); hiring managers across most US states favored a straight over gay job candidate (Tilcsik 2011); and hiring managers across most occupations favored a White over Black candidate, generally regardless of the job requirements (Bertrand and Mullainathan 2004). The widespread pervasiveness of discriminatory behaviors becomes difficult to square with the finding (often from these very same studies) that few people explicitly endorse biased beliefs. For example, the very same individuals whose behavior reveals unequal medical decisions to Black and White patients most often express equitable explicit beliefs about Black and White patients (Green et al. 2007), and despite aforementioned evidence of gender discrimination in hiring, most people explicitly endorse the belief that women are more intelligent than men (Storage et al. 2020). Such discordant results prompt the question: if discriminatory behaviors
596
T. E. S. Charlesworth and M. R. Banaji
cannot be entirely attributed to explicit attitudes and beliefs, what is the source of such discrimination? What explains this inconsistency between expressed beliefs and revealed biases? Over the past two decades, research and theory on implicit attitudes and beliefs – attitudes and beliefs that are less accessible to conscious introspection and deliberate control (Greenwald and Banaji 1995) – have provided a compelling answer to the inconsistency between pervasive discrimination and low explicit biases. These discriminatory behaviors are being shaped and maintained, at least in part, by underlying implicit biases that exist and persist in the minds of individuals and in the culture. Indeed, as reviewed below, the evidence that implicit attitudes and beliefs predict individual behaviors (e.g., hiring decisions, seating distance) is substantial, today stemming from hundreds of studies reviewed in depth across major meta-analyses (Greenwald et al. 2009; Kurdi et al. 2019; Oswald et al. 2013). But the role of implicit bias in behaviors does not stop with the individual decision-maker. In line with evidence showing the pervasiveness of discriminatory behaviors across most people, the social sciences have long recognized that bias and discrimination are widely embedded in the broader systems of society – systems of healthcare, policing, education, and more. Recently, using massive data of implicit attitudes and beliefs aggregated across millions of respondents, an emerging body of studies has begun to quantify the relationship between implicit bias and such socially significant systemic outcomes, whether racial gaps in infant health outcomes (Orchard and Price 2017) or the gender gap in science and math achievement (Nosek et al. 2009). These studies have not yet been summarized together. Thus, the goal of the current chapter is to provide an introduction to this new and growing empirical evidence that reveals the coupling between implicit cognition and systemic discriminatory outcomes.
A Brief Definition of Implicit and Explicit Attitudes and Beliefs In 1995, building from a wealth of literature showing that humans’ conscious, introspective minds are not the whole story of cognition (Nisbett and Wilson 1977), Greenwald and Banaji (1995) proposed a distinction between two forms of social attitudes and beliefs. On the one hand, there are explicit attitudes and beliefs – thoughts and feelings about social groups that are relatively more controlled, deliberate, and reflective of conscious, personal values. On the other hand, there are implicit attitudes and beliefs – thoughts and feelings about social groups that are relatively automatic, uncontrolled, and less accessible to introspective access. In short, one cannot easily look into one’s own mind to understand or control these kinds of implicit cognitions. The distinction in format between implicit and explicit attitudes and beliefs also requires a distinction in measurement. Explicit attitudes and beliefs, being controlled and accessible to cognitive awareness, can be self-reported through the typical tools of social surveys or Likert items. One can ask respondents “do you prefer elderly people or younger people?” or “to what extent do you think younger people are
25
Evidence of Covariation Between Regional Implicit Bias and. . .
597
smarter than elderly people?” and receive answers that reflect explicit cognitions. Implicit attitudes and beliefs, however, being less accessible to conscious awareness and control, inherently cannot be measured in such direct ways. Instead, implicit attitudes and beliefs are measured through indirect measures, including response time tasks such as the widely used Implicit Association Test (Greenwald et al. 1998); for a recent review, see (Greenwald et al. 2020). Today, in the more than two decades that have followed Greenwald and Banaji’s (1995) initial conceptualization, the evidence continues to be compelling for both explicit and implicit bias (Greenwald and Banaji 2017), revealing that they are related but distinct mental constructs (Bar-Anan and Vianello 2018; Cunningham et al. 2001; Nosek and Smyth 2007).
The Relationship of Implicit Bias and Individual Discriminatory Behavior Implicit and explicit attitudes and beliefs don’t sit idle in the mind but, rather, also reveal themselves in behaviors. When studies of the relationship between implicit bias and behaviors are well-powered and performed with precision, the correlation between an individual’s IAT score and their discriminatory behavior is of moderate to large magnitude, r ~ 0.40 (Kurdi et al. 2019), a correlation above the majority of effect sizes in psychology (Funder and Ozer 2019). Moreover, though both implicit and explicit cognitions often relate to individual’s behaviors, the two cognitions show incremental predictive validity, meaning that each explains variance above and beyond the other (Kurdi et al. 2019), lending confidence to the unique and complementary role played by implicit bias. Notably, the majority of evidence for such relationships between implicit bias and behavior comes from studies of individuals. As an example of such a study, Green et al. (2007) assessed individuals’ implicit pro-White/anti-Black attitudes and found that stronger implicit pro-White/anti-Black attitudes correlated with less treatment for hypothetical Black patients with cardiovascular disease (B ¼ 0.19) but more treatment for White patients (B ¼ 0.17). Over hundreds of such studies, reviewed across three meta-analyses (Greenwald et al. 2009; Kurdi et al. 2019; Oswald et al. 2013), implicit attitudes and beliefs help explain why some people act in more or less discriminatory ways. Recently, however, investigations have turned to a new type of discriminatory behavior that is revealed through socially significant behaviors aggregated across millions of people.
The Relationship of Implicit Bias and Systemic Discriminatory Behaviors: An Overview Prompted, in part, by the new availability of big data documenting implicit bias across millions of people around the globe (through the Project Implicit demonstration website, https://implicit.harvard.edu), the past few years have seen more than a dozen studies testing the role of aggregated implicit bias in both shaping and being
598
T. E. S. Charlesworth and M. R. Banaji
shaped by systemic outcomes. In addition to the theoretical implications elaborated throughout this chapter, the new focus on aggregated systems or behaviors offers methodological advantages to identify the role of implicit cognition in discrimination. Specifically, aggregation allows for greater precision in the estimates of both implicit cognition and systemic behaviors and can help reveal the underlying strong relationships that may previously have been obscured by imprecise and noisy estimates taken from single individuals (Payne et al. 2017, 2022). Additionally, big data allows for investigations of implicit cognition in real-world behaviors measured in vivo and at scale, greatly improving ecological validity and enabling more generalizable conclusions across broader samples and geographic locations. Nearly all of the studies examining aggregate relationships between implicit cognition and behavior have adopted correlational designs: implicit attitudes or stereotypes are aggregated across geographic locations (counties, states, countries, etc.) and correlated with aggregated systemic outcomes (e.g., rates of lethal force by police, gender gaps in math tests) across those same geographic regions, while controlling for a variety of structural factors (e.g., demographic representation, GDP, etc.). Because such correlations are generally noncommittal regarding the direction of the relationship, the studies can be grouped into two complementary approaches in thinking about the meaning of the relationship. First, there is a set of studies (see Table 1) that take discriminatory systemic behavior as the dependent variable to be explained by measures of implicit bias. For instance, this sort of study may look at how country-level differences in implicit gender stereotypes help explain country-level differences in the outcome of gender gaps in standardized tests (Nosek et al. 2009). Because the goal in this chapter is to show how implicit cognition can illuminate systemic discrimination, this first set of studies constitutes the primary focus. Below, the studies are organized and reviewed according to the outcome across socially significant domains of (1) education, (2) healthcare, and (3) policing. A second type of study in this area considers the relationship between implicit bias and discriminatory systemic behaviors by flipping the equation and identifying the sources of implicit bias itself as the outcome (i.e., IAT scores become the dependent variable). For instance, studies of this type may look at how countrylevel differences in the representation of fat people help explain the country-level differences in the implicit anti-fat/pro-thin attitudes (Marini et al. 2013) or how demographic and physical features of the environment (e.g., the number of mental healthcare providers or the number of lakes) predict the magnitude of implicit biases (Hehman et al. 2020). Although these studies primarily seek to understand the sources of implicit bias (rather than the contribution of implicit bias to behaviors), the correlational nature of the studies means that they can still provide complementary insight into the coupling of implicit cognition and specific discriminatory behaviors. This group of studies is reviewed in the final section.
25
Evidence of Covariation Between Regional Implicit Bias and. . .
599
Table 1 Selected studies using implicit cognition to explain systemic discriminatory behaviors
Authors Nosek et al. (2009)
Geographic aggregation level Countries
IAT topic Gender-science IAT (male-science/ female-arts)
Societal outcome Eighth grade standardized math/ science testing
Black-White gap in school discipline (i.e., out-of-school suspension, in-school suspension, law enforcement referrals, expulsions, and in-school arrests) Black-White gap in standardized testing scores (3rd–8th grade for math and English) and school discipline
Riddle and Sinclair (2019)
US counties
Race attitude IAT (Black-bad/Whitegood)
Chin et al. (2020)
US counties
Race attitude IAT (Black-bad/Whitegood) among teachers
Chetty et al. (2020)
US counties
Race attitude IAT (Black-bad/Whitegood)
Upward mobility among Black and White boys and girls (roughly defined as making more money than their parents)
Hehman et al. (2018)
US communitybased statistical area (CBSA)
Race attitude IAT (Black-bad/Whitegood) and stereotype IAT (Black-weapon/ White-object)
Police lethal force toward Black Americans versus White Americans
Key result Stronger malescience/female-arts associations correlated with larger gender gaps on 8th grade testing, r ¼ 0.60 Stronger pro-White/ anti-Black implicit attitudes correlated with larger BlackWhite gaps in school discipline
Stronger pro-White/ anti-black implicit attitudes among teachers correlated with larger BlackWhite gaps in standardized test scores. Counties with +1 SD attitudes had +0.037 SD in the Black/White gap or 6.7% of the test gap Stronger pro-White/ anti-Black implicit attitudes correlated with less upward mobility among black boys and girls. Black boys and girls living in counties with +1 SD in implicit bias earn ~0.8 percentiles less income when they grow up Stronger pro-White/ anti-black implicit attitudes correlated with greater disproportionate use of police lethal force, B ¼ 0.35 (continued)
600
T. E. S. Charlesworth and M. R. Banaji
Table 1 (continued)
Authors Leitner et al. (2016)
Geographic aggregation level US counties
IAT topic Race attitude IAT (Black-bad/Whitegood)
Societal outcome Black American and White American death rates from circulatory diseases
Leitner et al. (2018)
US states
Race attitude IAT (Black-bad/Whitegood)
State spending on Medicaid disability programs (more likely to assist Black than White Americans)
Orchard and Price (2017)
US counties
Race attitude IAT (Black-bad/Whitegood)
Black-White gap in infant low birth weight and preterm births
Giasson and Chopik (2020)
US states
Age attitude IAT (old-bad/younggood)
Stelter et al. (2022)
US counties
Race attitude IAT (Black-bad/Whitegood) and stereotype IAT (Black-weapon/ White-object) among White respondents
Health behaviors (e.g., smoking, diet) and health (e.g., self-reported physical and mental health) among adults aged 65+ Police traffic stop rates of Black drivers relative to Black population. Specifically, percentage of Black drivers stopped (from all drivers stopped) minus
Key result Among black Americans, stronger pro-Black/antiWhite implicit attitudes correlated with higher death rates from circulatory diseases, B ¼ 0.11 Among White Americans, stronger pro-White/antiBlack implicit attitudes correlated with less spending on Medicaid, B ¼ 0.33 Stronger pro-White/ anti-black implicit attitudes correlated with larger BlackWhite gaps in infant birth weight and preterm births. Counties with +1 SD attitudes had 14% larger gap in low birth weight and 29% larger gap in preterm births Stronger pro-young/ anti-old implicit attitudes correlated with worse health outcomes among elderly, B ¼ 0.29 Stronger pro-White/ anti-Black implicit attitudes among White respondents correlated with greater racial disparities in the proportion of traffic stops, r ¼ 0.30 (for (continued)
25
Evidence of Covariation Between Regional Implicit Bias and. . .
601
Table 1 (continued)
Authors
Geographic aggregation level
IAT topic
Ekstrom et al. (2022)
US counties
Race attitude IAT (Black-bad/Whitegood)
Johnson and Chopik (2019)
US states
Race stereotype IAT (Blackweapon/Whiteobject)
Societal outcome
Key result
percentage of Black people in population
counties with >150 respondents) to r ¼ 0.07 (for all counties). No correlations between implicit race stereotypes and racial disparities in police traffic stops Stronger pro-White/ anti-Black implicit attitudes correlated with greater racial disparities in the proportion of traffic stops, r ¼ 0.31
Police traffic stop rates of Black versus White drivers. Difference in percentage of Black drivers stopped (from Black drivingaged population) versus percentage of White drivers stopped (from White driving-aged population) Centers for Disease Control statistics on weapon-related deaths among Black and White Americans
Stronger Blackweapon/Whiteobject implicit stereotypes were associated with greater rates of Black people dying by weapon violence, B ¼ 0.26, and lower rates of White people dying by weapon violence, B ¼ 0.89
Disparities in Education and Opportunity: Standardized Testing, School Discipline, and Economic Mobility In the first paper to use aggregated Project Implicit data obtained from individual minds, Nosek et al. (2009) investigated whether country-level differences in implicit male-science/female-arts stereotypes correlated with gender gaps on 8th grade standardized mathematics achievement tests. The authors found that, across hundreds of thousands of respondents aggregated across 34 countries, those countries
602
T. E. S. Charlesworth and M. R. Banaji
with higher implicit associations between male-science/female-arts also showed larger gender gaps (where boys outperformed girls) on standardized science and math tests (r ¼ 0.6, or R2 ¼ 0.36). Additionally, this relationship persisted after controlling for country-level explicit associations, underscoring the incremental predictive validity of implicit cognitions in explaining a socially significant outcome of gender differences in performance. A few years later, a similar analysis examined the correlation between countrylevel implicit male-science/female-arts stereotypes and the representation of women in the STEM workforce and STEM tertiary education (Miller et al. 2015). Although this study technically used the implicit stereotypes as the outcome rather than the predictor, we interpret the results alongside the study by Nosek and colleagues because of the correlational nature of the study and the fact that gender representation in STEM could theoretically be both a predictor and an outcome of bias (Charlesworth and Banaji 2019). Indeed, Miller and colleagues found that countries with higher representation of women in STEM tertiary education had weaker implicit male-science/female-arts stereotypes (R2 ¼ 0.26), with similar effects observed for explicit stereotypes (R2 ¼ 0.21). Additionally, women’s representation in STEM appeared to be a key variable in the relationship between achievement and implicit stereotypes: in fact, when women’s representation in STEM was controlled for, the relationship between achievement and implicit stereotypes was eliminated. Although there are individual bivariate correlations between country-level implicit bias and the outcomes of gender gaps in STEM achievement, the role of representation or prevalence of a minority group appears to be a mediating, explanatory variable. Following these early studies on gender gaps in STEM achievement and representation, recent studies have looked to also explain race gaps in educational achievement and school discipline. Specifically, Chin et al. (2020) found that higher county-level pro-White/anti-Black implicit bias among teachers (a sufficient subsample of the Project Implicit data) correlated with larger Black/White gaps in standardized math and English testing for 3rd–8th grade students (see also Pearman 2021). Furthermore, county-level relationships also emerged between teacher’s implicit bias and Black/White gaps in student discipline, such that high-bias counties had greater disproportionate discipline of Black students across K-12 (Chin et al. 2020; Riddle and Sinclair 2019). How do these regression estimates (see Table 1) translate into real-word numbers of harm done to young people? Using the predicted probabilities from Chin et al. (2020), we see that in counties where teachers’ IAT D scores are at the mean (D ¼ 0.36), approximately 13% of Black students receive in-school suspensions, compared to only 5% of White students; in counties with lower IAT D scores (D ¼ 0.15; a threshold typically used for denoting “no bias”), the race difference shrinks such that approximately 8% of Black students and 4% of White students receive in-school suspensions. Cook county in Illinois (one of the states that best represents the American electorate) has a population of 5.15 million, of which 22% are schoolaged (1.13 million), 65% are White, and 24% are Black. If such a county had average IAT bias (D ¼ 0.36), this means that approximately 35,100 Black school-aged
25
Evidence of Covariation Between Regional Implicit Bias and. . .
603
children would receive in-school suspensions; if, on the other hand, such a county had lower IAT bias (D ¼ 0.15), we would expect 21,600 Black children to receive in-school suspensions. In other words, nearly 13,500 Black children are likely to receive in-school discipline in a high IAT bias county than an average IAT bias county. Ultimately, such consequences of large racial disparities in school suspension and discipline among counties with high implicit race bias appear particularly poignant because early experiences of school discipline interrupt future opportunity and increase interactions with policing and prisons, creating the so-called “school-toprison” pipeline (Smith 2009). Finally, implicit Black/White race attitudes also have effects for future opportunity in income earning more broadly. That is, when implicit race attitudes are aggregated within US census tracts, greater aggregated county-level implicit biases are found to correlate with less upward social mobility (i.e., they are less likely to make more money than their parents) among historically disadvantaged groups (Chetty et al. 2020). Black American boys or girls living in a high implicit bias neighborhood have a lower chance of upward mobility and income earning than if they had, by luck or coincidence, grown up in a relatively low implicit bias neighborhood. In other words, the so-called American Dream of upward mobility appears to only be true for some types of children living in relatively unbiased contexts. Why are these sorts of results interpreted as “discrimination”? Of course, an outcome such as the aggregated math testing gap between boys and girls in an entire country does not arise from any single actor producing discriminatory behavior (e.g., a specific biased examiner or teacher). Rather, the outcomes arise from the systemic presence of beliefs that are so pervasive that we refer to them as being “in the air.” We use this analogy in the same way that Claude Steele referred to “a threat in the air” when explaining women’s underperformance on mathematics tests after stereotypes were evoked (Steele 1997). This air reveals itself, for example, through ambient cues that signal who belongs in a certain field (Cheryan et al. 2009) or in the stereotypes revealed in the word associations across billions of words constituting humans’ collective language (Caliskan et al. 2016; Charlesworth et al. 2021). In turn, such systemic cues may evoke stereotype threat among girls (Spencer et al. 1999) or stereotype lift among boys (Walton and Cohen 2003). These studies thus build an understanding that stereotypes are widely embedded in society and therefore have the potential to shape the discriminatory behaviors of all those interacting in that society. This spread may be especially true for implicit stereotypes (Payne et al. 2017) because such hidden, indirect beliefs can pervade and persist more easily in the face of conscious values and ideals against them. In this way, the results summarized above, and in the remainder of the chapter, give new meaning to the idea of systemic discrimination: they quantify how discriminatory outcomes arise from the “air” of implicit bias aggregated across thousands of people in a region.
604
T. E. S. Charlesworth and M. R. Banaji
Disparities in Healthcare: Medicaid Spending, Death Rates, and Infant Health Outcomes A second group of outcomes in studies of aggregated implicit bias centers on the domain of health, including healthcare spending, rates of disease, and infant health. For instance, Leitner et al. (2018) used health data on US state-level Medicaid spending from across the United States to examine whether states that are lower in implicit anti-Black attitudes also had greater Medicaid spending. Because Medicaid spending is more likely to preferentially affect Black than White Americans due to persistent and significant racial gaps in income inequality and other health insurance coverage (Smedley et al. 2003), state investment in programs like Medicaid is, in part, a reflection of their intention to reduce racial inequality. In line with this perspective, states with higher implicit anti-Black attitudes were also found to be lower Medicaid spending states (Leitner et al. 2018), with a likely consequence of limiting equitable access to healthcare and thus compromising overall health. Indeed, implicit biases have recently been linked directly to health outcomes: Giasson and Chopik (2020) found that US states with higher implicit age bias (anti-old/pro-young) revealed worse health outcomes among elderly adults aged 65+. That is, states with higher implicit age biases were also those states in which elderly adults reported, among other indicators, having physical activity limitations, feeling physically unhealthy, experiencing mental distress, or engaging in unhealthy behaviors (e.g., insufficient sleep, smoking, binge drinking). Perhaps even more consequential, implicit race bias has been shown to correlate directly with death rates, not only of the targets of bias (e.g., death rates among Black Americans in places where White Americans hold high bias) but also the holders of those biases (Leitner et al. 2016a, b; Zestcott et al. 2021). For instance, counties in which White respondents held stronger pro-White/anti-Black attitudes were also counties with higher death rates from cardiovascular disease among Black respondents (B ¼ 0.074) and among White respondents (B ¼ 0.081; Zestcott et al. 2021). The additional stressors that arise from living in areas “polluted” by bias thus appear to have consequences for health throughout the population. Finally, looking at health outcomes earlier in the life span, researchers have investigated the relationship between county-level pro-White/anti-Black implicit attitudes and the persistent racial gap in infant health outcomes in the United States (Orchard and Price 2017). Because of a complex set of compounded stressors including lower prenatal healthcare and nutrition and repeated experiences of discrimination, Black American mothers in the United States are 1.6 times more likely to have preterm births and twice as likely to give birth to infants with low birth weight. Using aggregated county-level implicit race attitudes, Orchard and Price quantified bias that may shape these racial gaps in health outcomes. The authors showed that, in low-bias counties (those at 1 standard deviation below the mean), approximately 15% of births to Black mothers were preterm, while in high-bias counties (+1 SD above the mean) approximately 17% of births to Black mothers were preterm. For a representative county like Cook County, IL, with about 5,000 births to Black mothers in a given year, even this 2 percentage point difference
25
Evidence of Covariation Between Regional Implicit Bias and. . .
605
would translate into 100 more preterm Black babies born if that county had high versus low bias. Given the social and healthcare costs associated with preterm birth in the United States (Beam et al. 2020), the cost difference would be approximately 7.6 million dollars between a high- and low-bias county. Moreover, considering that early health consequences from aggregated race bias have also been demonstrated on outcomes ranging from the likelihood of Black foster children being adopted (Bell et al. 2021) to Black and Latinx youth’s brain development (Hatzenbuehler et al. 2021), it becomes clear that region-level implicit biases play a powerful role in shaping and maintaining health outcomes from birth to death.
Disparities in Policing: Lethal Force and Traffic Stops Perhaps one of the starkest displays of life-or-death consequences emerges in the relationship between pro-White/anti-Black implicit attitudes and the disproportionate use of police lethal force toward Black Americans (Hehman et al. 2018). First, using fact-checked data from the Guardian’s reports of US individuals killed by police, Hehman and colleagues observed that Black Americans constituted approximately 23% of all deaths from police lethal force but only 12% of the population, indicating disproportionate overrepresentation in deaths at the hands of police. The new result is that disproportionate overrepresentation in lethal force can be explained by the strength of implicit bias in a region: the greater the implicit antiBlack/pro-White attitudes and implicit Black-weapon stereotypes in a region, the greater the likelihood of lethal use of force by police toward Black Americans. For instance, a Black American who happened to be living in a region with relatively high implicit Black-weapon/White-object stereotypes was more likely to be killed by police than if they resided in a region with relatively lower implicit stereotypes (see also Correll et al. 2007; Johnson and Chopik 2019). The possibility of police use of lethal force begins at the moment of police interaction; disproportionate rates of police and civilian encounters, such as racial differences in the rates of traffic stops, can give rise to later disparities in lethal force (such as in the cases of Dante Wright or Philando Castile). Indeed, two recent papers examined the role of aggregated implicit bias in accounting for the fact that Black drivers are stopped by police more often than White drivers in the United States (Ekstrom et al. 2022; Stelter et al. 2022). Both studies found an association such that regions with higher implicit pro-White/anti-Black attitudes also had higher rates of disproportionate traffic stops for Black drivers. For instance, inspecting the data reported by Ekstrom and colleagues, in a given county with high implicit race bias (at the maximum IAT D score), the predicted model estimates show a racial difference in stop rate scores of 79, which would translate, for example, as police stopping 179 Black drivers (per 100 Black driving-aged residents, i.e., many drivers are stopped more than once) but 100 White drivers (per 100 White driving-aged residents, i.e., drivers are stopped based on their population rate). In contrast, a representative low-bias county (at the minimum IAT D score) had a stop difference score that of 29, which can translate into 71 Black drivers being stopped (per
606
T. E. S. Charlesworth and M. R. Banaji
100 Black driving-aged residents, i.e., some drivers are never stopped) and 100 White drivers (per 100 White driving-aged residents). Ultimately, despite many analytical and theoretical differences between the two papers (e.g., how they operationalized traffic stop rates, how they controlled for demographics), the consistency of the link between regional racial attitudes and consequential policecivilian interactions highlights the robustness of the aggregate approach to understanding human behavior (Payne and Rucker 2022).
Examining Implicit Bias as the Outcome Explained by Systemic Predictors Implicit bias not only shapes systemic outcomes across aforementioned domains of education, healthcare, and policing but is also shaped by the demographics and structural features of the environment. Indeed, knowing that the level of implicit biases varies meaningfully in magnitude across geography, the question naturally arises: where does such variation come from in the first place? A handful of recent studies have begun to shed light on the answer by using aggregated implicit attitudes and beliefs as the dependent variables predicted from a range of systemic predictors. The majority of these studies has focused on the role of regional demographic representation – the frequency or diversity of specific groups such as Black Americans or fat people – to explain geographic variation in implicit bias. In some cases, the relationship is intuitive and aligns with expectations of intergroup contact theories, in which greater intergroup contact will correlate with less bias (Allport 1954; Pettigrew and Tropp 2006). For instance, higher frequency of Asian Americans in US metropolitan areas was found to correlate with lower implicit stereotypes in the association of Asian American with “foreign” and European American with “American” (Devos and Sadler 2019); and, most compelling, temporal fluctuations in Asian American representation and diversity were found to correspond with fluctuations in implicit Asian American-foreign biases (Devos et al. 2021). Similarly, for implicit anti-gay/pro-straight attitudes, higher self-reported personal contact, as well as higher county-level frequencies of sexual minorities, correlated with lower implicit bias (MacInnis et al. 2017). And, finally, as noted above, higher countrylevel representations of women in STEM correlated with lower implicit malescience/female-arts stereotypes (Miller et al. 2015). However, the relationship between demographic representation and implicit bias is not always straightforward. In fact, a higher representation of fat people in a country (presumably allowing for more outgroup exposure) was found to correlate with higher country-level implicit anti-fat/pro-thin attitudes (Marini et al. 2013). Additionally, greater frequency of Black Americans in a US state was found to correlate with higher state-level implicit pro-White/anti-Black attitudes among White Americans (Rae et al. 2015, 2022; but see O’Shea et al. 2019). Perhaps these counterintuitive findings could be explained by considering not just the quantity of intergroup contact but also the quality of that contact (Paluck et al. 2018). Indeed, areas with higher representations of a marginalized group may
25
Evidence of Covariation Between Regional Implicit Bias and. . .
607
counterintuitively result in more negative intergroup interactions, thereby perpetuating negative implicit bias. More nuanced insights into the relationship of representation and aggregate implicit bias may therefore come from considering not only raw frequencies of minority representation but also the integration and variety of minority groups. For instance, when looking at multiple indicators of context racial diversity (e.g., integration, prevalence, and variety of groups), implicit stereotypes associating Black Americans with weapons were indeed found to be weaker in US metropolitan areas with greater integration and a larger variety of minority groups (Sadler and Devos 2020). It is also notable that the relationships between demographic representation and implicit bias have deep and complex roots in the historical patterns of representation across the country. Recent data show that the greater the proportion of enslaved to free people in the southern states in the 1860s, the greater the implicit anti-Black/proWhite bias among White Americans in those areas today, 160 years later (Payne et al. 2019). In fact, the correlation between historical rates of slavery from 1860 and state-level implicit race bias today (r ¼ 0.87) was nearly three times larger than the relationship between contemporary Black American representation and implicit race bias (r ¼ 0.32). Studies like this suggest that, when one thinks about group-based discrimination, one must also think about their effects as extending forward in time, translated through the continuous presence of social structures and reminders of inequality (e.g., the presence of confederate monuments; Payne et al. 2019). The result also suggests that today’s Americans who live in regions of greater historical legacies of slavery must be acquiring the particles embedded in the biased air. Systemic discrimination is a useful term in this case as it helps capture the pervasiveness of discriminatory treatment extending across space and time. Indeed, time has featured as an important variable in several recent studies examining the temporal relationships between implicit bias and macrolevel societal variables. For instance, although the implicit Asian American ¼ foreign stereotype had been slowly decreasing over the past decade, racial slurs tweeted during the beginning of the Covid-19 pandemic were found to have reversed the trend and coincided with sharp spikes in implicit bias (Darling-Hammond et al. 2020). Other data similarly suggest that levels of implicit bias fluctuate in response to the actions and events in the world – whether legislation (Ofosu et al. 2019), social movements (Sawyer and Gampa 2017), pathogens such as Ebola (Inbar et al. 2016), fat-shaming tweets (Ravary et al. 2019), or other group-targeting rhetoric (Charlesworth and Banaji 2022). These studies contribute to an understanding of the link between implicit bias and societal structures by showing that macrolevel events can shape the thoughts and feelings of respondents which, in turn, reshape the state of society.
608
T. E. S. Charlesworth and M. R. Banaji
Implications for Understanding Implicit Bias and Systemic Behaviors The studies reviewed in this chapter belong to a new generation of research that holds the potential to develop a more robust understanding of the sources and consequences of systemic discrimination. Across more than a dozen studies, the evidence reveals relationships between aggregated implicit cognition across millions of individuals and socially significant outcomes ranging from academic performance to upward mobility to health and mortality. In the process, the findings also encourage an emerging perspective on implicit cognition as perhaps best understood as social representations that paradoxically are hidden from conscious awareness, yet pervasively embedded in the structures that surround us (Charlesworth and Banaji 2021; Payne et al. 2017). Recently, Payne et al. (2017) have summarized this new perspective in the “bias of crowds” model to suggest that individual measures of implicit bias are a noisy indicator of the true signal of bias embedded in the broader culture, and by aggregating across people, one can better capture that signal. Such a structural perspective of implicit cognition stands in contrast to the early notions that implicit measures can reveal a bona fide pipeline (Fazio et al. 1995) to an individual’s trait-like personality or permanent individual inclinations (Greenwald et al. 1998). Instead, the mounting evidence reviewed here points more clearly toward the notion that implicit cognition reflects the thumbprint of the culture on the mind. While theorizing about implicit attitudes and beliefs as reflecting culture is not altogether unprecedented, what is new is the availability of large-scale data and analytic methods to quantify evidence for the operation of implicit bias in socially significant outcomes.
Concluding Remarks The concept and measurement of implicit bias began in psychology but has since permeated many disciplines, from medicine (Green et al. 2007) to computer science (Caliskan et al. 2016) to business (Banaji et al. 2003) to law (Kang and Banaji 2006). Today, a particularly fruitful interdisciplinary approach has arisen from the intersection of psychology and those social sciences such as economics that are typically focused on larger units of society (Carlana 2019; Chetty et al. 2020). At this intersection, new combinations of variables are being studied that interweave measures of the thoughts and feelings inside individual minds with broader outcomes such as the opportunity of upward mobility, the likelihood of police shootings, or the health of infants. Evidence has also accumulated for the reverse relationships, in which structural variables, such as the frequency and diversity of demographic representations (e.g., of women, Black Americans), help shed light on the magnitude of implicit bias as an outcome itself. Together, this emerging body of work shows the tight coupling between implicit bias at the level of the individual and socially significant outcomes at the level of society. The strength of relationships being what it is – with at least small-to-moderate effect sizes
25
Evidence of Covariation Between Regional Implicit Bias and. . .
609
compounded over millions of people that experience hundreds of such interactions (Greenwald et al. 2015) – this evidence cannot be set aside. We, as researchers and as citizens, must take seriously the clear link between implicit bias and systemic discrimination. Moreover, these new studies remind us that discrimination is not always a simple person-to-person act but arises throughout the systems that shape the lives that inhabit these systems such as education, healthcare, or policing. Tackling discrimination therefore requires that one not only address individual decisions but also societal practices, systems, policies, and laws. With such goals, it is clearly an exciting and pressing time for social and behavioral scientists to collaborate in the study of implicit bias and discrimination. Such collaborations will consolidate the joint expertise in theory and methods to better understand the nature of the mind, socpety, and the way that each reflects and reinforces the other.
Cross-References ▶ Gender-Based Discrimination in Health: Evidence from Cross-Country ▶ Insights from Social Psychology: Racial Norms, Stereotypes, and Discrimination
References Allport GW (1954) The nature of prejudice. Addison-Wesley, Boston, MA Banaji MR, Bazerman MH, Chugh D (2003) How (un)ethical are you? Harv Bus Rev 81(12) https:// hbr.org/2003/12/how-unethical-are-you Bar-Anan Y, Vianello M (2018) A multi-method multi-trait test of the dual-attitude perspective. J Exp Psychol Gen 147(8):1264–1272. https://doi.org/10.1037/XGE0000383 Beam AL, Fried I, Palmer N, Agniel D, Brat G, Fox K, Kohane I, Sinaiko A, Zupancic JAF, Armstrong J (2020) Estimates of healthcare spending for preterm and low-birthweight infants in a commercially insured population: 2008–2016. J Perinatol 40(7):1091–1099. https://doi.org/ 10.1038/s41372-020-0635-z Bell SB, Farr R, Ofosu E, Hehman E, DeWall CN (2021) Implicit bias predicts less willingness and less frequent adoption of Black children more than explicit bias. J Soc Psychol. https://doi.org/ 10.1080/00224545.2021.1975619 Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal?: a field experiment on labor market discrimination. Am Econ Rev 94(4). https://doi.org/10.4324/ 9780429499821-53 Caliskan A, Bryson JJ, Narayanan A (2016) Semantics derived automatically from language corpora necessarily contain human biases. Science 356(6334):183–186. https://doi.org/10. 1126/science.aal4230 Carlana M (2019) Implicit stereotypes: evidence from teachers’ gender bias*. Q J Econ 134(3): 1163–1224. https://doi.org/10.1093/qje/qjz008 Charlesworth TES, Banaji MR (2019) Gender in science, technology, engineering, and mathematics: issues, causes, solutions. J Neurosci 39(37):7228–7243). Society for Neuroscience. https:// doi.org/10.1523/JNEUROSCI.0475-18.2019 Charlesworth TES, Banaji MR (2021) Patterns of implicit and explicit attitudes II. Long-term change and stability, regardless of group membership. Am Psychol 76(6):851–869. https://doi. org/10.1037/amp0000810
610
T. E. S. Charlesworth and M. R. Banaji
Charlesworth T, Banaji MR (2022) Patterns of implicit and explicit attitudes: IV. Change and stability from 2007 to 2020. Psychol Sci 9567976221084257. https://doi.org/10.1177/ 09567976221084257 Charlesworth TES, Yang V, Mann TC, Kurdi B, Banaji MR (2021) Gender stereotypes in natural language: word embeddings show robust consistency across child and adult language corpora of more than 65 million words. Psychol Sci 32(2):218–240. https://doi.org/10.1177/ 0956797620963619 Cheryan S, Plaut VC, Davies PG, Steele CM (2009) Ambient belonging: how stereotypical cues impact gender participation in computer science. J Pers Soc Psychol 97(6):1045–1060. https:// doi.org/10.1037/a0016239 Chetty R, Hendren N, Jones MR, Porter SR (2020) Race and economic opportunity in the United States: an intergenerational perspective*. Q J Econ 135(2):711–783. https://doi.org/10.1093/qje/ qjz042 Chin MJ, Quinn DM, Dhaliwal TK, Lovison VS (2020) Bias in the air: a Nationwide exploration of teachers’ implicit racial attitudes, aggregate bias, and student outcomes. Educ Res 0013189X2093724. https://doi.org/10.3102/0013189X20937240 Correll J, Park B, Judd CM, Wittenbrink B, Sadler MS, Keesee T (2007) Across the thin blue line: police officers and racial bias in the decision to shoot. J Pers Soc Psychol 92(6):1006–1023. https://doi.org/10.1037/0022-3514.92.6.1006 Cunningham WA, Preacher KJ, Banaji MR (2001) Implicit attitude measures: consistency, stability, and convergent validity. Psychol Sci 12(2):163–170. https://doi.org/10.1111/1467-9280.00328 Darling-Hammond S, Michaels EK, Allen AM, Chae DH, Thomas MD, Nguyen TT, Mujahid MM, Johnson RC (2020) After “the China virus” went viral: racially charged coronavirus coverage and trends in bias against Asian Americans. Health Educ Behav 47(6):870–879. https://doi.org/ 10.1177/1090198120957949 Devos T, Banaji MR (2005) American ¼ white? J Pers Soc Psychol 88(3):447–466. https://doi.org/ 10.1037/0022-3514.88.3.447 Devos T, Mohamed H (2014) Shades of American identity: implicit relations between ethnic and National Identities. Soc Personal Psychol Compass 8(12):739–754. https://doi.org/10.1111/ spc3.12149 Devos T, Sadler M (2019) Context diversity predicts the extent to which the American identity is implicitly associated with Asian Americans and European Americans. Asian Am J Psychol 10(2):182–193. https://doi.org/10.1037/aap0000149 Devos T, Sadler M, Perry D, Yogeeswaran K (2021) Temporal fluctuations in context ethnic diversity over three decades predict implicit national inclusion of Asian Americans. Group Processes and Intergroup Relations 24(1):3–25. https://doi.org/10.1177/1368430219887440 Ekstrom PD, Le Forestier JM, Lai CK (2022) Racial demographics explain the link between racial disparities in traffic stops and county-level racial attitudes. Psychol Sci, 095679762110535. https://doi.org/10.1177/09567976211053573 Fazio RH, Jackson JR, Dunton BC, Williams CJ (1995) Variability in automatic activation as an unobtrusive measure of racial attitudes: a bona fide pipeline? J Pers Soc Psychol 69(6): 1013–1027. https://doi.org/10.1037/0022-3514.69.6.1013 Funder DC, Ozer DJ (2019) Corresponding authors: evaluating effect size in psychological research: sense and nonsense. Adv Methods Pract Psychol Sci 2(2):156–168. https://doi.org/ 10.1177/2515245919847202 Giasson HL, Chopik WJ (2020) Geographic patterns of implicit age bias and associations with state-level health outcomes across the United States. Eur J Soc Psychol 50(6):1173–1190. https://doi.org/10.1002/EJSP.2707 Green AR, Carney DR, Pallin DJ, Ngo LH, Raymond KL, Iezzoni LI, Banaji MR (2007) Implicit bias among physicians and its prediction of thrombolysis decisions for black and white patients. J Gen Intern Med 22(9):1231–1238. https://doi.org/10.1007/s11606-007-0258-5 Greenwald AG, Banaji MR (1995) Implicit social cognition: attitudes, self-esteem, and stereotypes. Psychol Rev 102(1):4–27. https://doi.org/10.1037/0033-295X.102.1.4
25
Evidence of Covariation Between Regional Implicit Bias and. . .
611
Greenwald AG, Banaji MR (2017) The implicit revolution: reconceiving the relation between conscious and unconscious. Am Psychol 72(9):861–871. https://doi.org/10.1037/amp0000238 Greenwald AG, McGhee DE, Schwartz JLK (1998) Measuring individual differences in implicit cognition: the implicit association test. J Pers Soc Psychol 74(6):1464–1480. https://doi.org/10. 1037/0022-3514.74.6.1464 Greenwald AG, Banaji MR, Nosek BA (2015) Statistically small effects of the implicit association test can have societally large effects. J Pers Soc Psychol 108(4):553–561 Greenwald AG, Brendl M, Cai H, Cvencek D, Dovidio J, Friese M, Hahn A, Hehman E, Hofmann W, Hughes S, Hussey I, Jordan C, Jost J, Kirby TA, Lai C, Lang J, Lindgren K, Maison D, Ostafin B, . . . Wiers R (2020) The Implicit Association Test at age 20: what is known and what is not known about implicit bias. https://doi.org/10.31234/osf.io/bf97c Greenwald AG, Poehlman TA, Uhlmann EL, Banaji MR (2009) Understanding and using the implicit association test: III. Meta-analysis of predictive validity. J Pers Soc Psychol 97(1): 17–41. https://doi.org/10.1037/a0015575 Hatzenbuehler ML, Weissman DG, McKetta S, Lattanner MR, Ford JV, Barch DM, McLaughlin KA (2021) Smaller hippocampal volume among black and latinx youth living in high-stigma contexts. J Am Acad Child Adolesc Psychiatry. https://doi.org/10.1016/j.jaac.2021.08.017 Hehman E, Flake JK, Calanchini J (2018) Disproportionate use of lethal force in policing is associated with regional racial biases of residents. Soc Psychol Personal Sci 9(4):393–401. https://doi.org/10.1177/1948550617711229 Hehman E, Ofosu EK, Calanchini J (2020) Using environmental features to maximize prediction of regional intergroup bias. Social Psychological and Personality Science Inbar Y, Westgate EC, Pizarro DA, Nosek BA (2016) Can a naturally occurring pathogen threat change social attitudes? Evaluations of gay men and lesbians during the 2014 Ebola epidemic. Soc Psychol Personal Sci 7(5):420–427. https://doi.org/10.1177/1948550616639651 Johnson DJ, Chopik WJ (2019) Geographic variation in the black-violence stereotype. Soc Psychol Personal Sci 10(3):287–294. https://doi.org/10.1177/1948550617753522 Kang, J., & Banaji, M. R. (2006). Fair measures: A behavioral realist revision of “affirmative action.” Calif Law Rev, 94(4), 1063–1118. https://doi.org/10.2307/20439059 Kurdi B, Seitchik AE, Axt JR, Carroll TJ, Karapetyan A, Kaushik N, Tomezsko D, Greenwald AG, Banaji MR (2019) Relationship between the implicit association test and intergroup behavior: a meta-analysis. Am Psychol 74(5):569–586. https://doi.org/10.1037/amp0000364 Leitner JB, Hehman E, Ayduk O, Mendoza-Denton R (2016a) Blacks’ death rate due to circulatory diseases is positively related to whites’ explicit racial bias. Psychol Sci 27(10):1299–1311. https://doi.org/10.1177/0956797616658450 Leitner JB, Hehman E, Ayduk O, Mendoza-Denton R (2016b) Racial bias is associated with ingroup death rate for blacks and whites: insights from project implicit. Soc Sci Med 170: 220–227. https://doi.org/10.1016/j.socscimed.2016.10.007 Leitner JB, Hehman E, Snowden LR (2018) States higher in racial bias spend less on disabled medicaid enrollees. Soc Sci Med 208:150–157. https://doi.org/10.1016/j.socscimed.2018. 01.013 MacInnis CC, Page-Gould E, Hodson G (2017) Multilevel intergroup contact and antigay prejudice (explicit and implicit). Soc Psychol Personal Sci 8(3):243–251. https://doi.org/10.1177/ 1948550616671405 Marini M, Sriram, N., Schnabel, K., Maliszewski N, Devos, T., Ekehammar, B., Wiers R, Cai, H., Somogyi, M., Shiomura, K., Schnall, S., Neto, F., Bar-Anan, Y., Vianello M, Ayala A, Dorantes G, Park J, Kesebir S, Pereirax A, . . . Nosek BA (2013) Overweight people have low levels of implicit weight bias, but overweight nations have high levels of implicit weight bias. PLoS One, 8(12):1–9. https://doi.org/10.1371/journal.pone.0083543 Miller DI, Eagly AH, Linn MC (2015) Women’s representation in science predicts National Gender-Science Stereotypes: evidence from 66 nations. J Educ Psychol 107(3):631–644. https://doi.org/10.1037/edu0000005
612
T. E. S. Charlesworth and M. R. Banaji
Moss-Racusin CA, Dovidio JF, Brescoll VL, Graham MJ, Handelsman J (2012) Science faculty’s subtle gender biases favor male students. Proc Natl Acad Sci 109(41):16474–16479. https://doi. org/10.1073/pnas.1211286109 Nisbett RE, Wilson TD (1977) Telling more than we can know: verbal reports on mental processes. Psychol Rev 84(3):231–259. https://doi.org/10.1037/0033-295X.84.3.231 Nosek BA, Smyth FL (2007) A multitrait-multimethod validation of the implicit association test: implicit and explicit attitudes are related but distinct constructs. Exp Psychol 54(1):14–29. https://doi.org/10.1027/1618-3169.54.1.14 Nosek BA, Smyth FL, Sriram N, Lindner NM, Devos T, Ayala A, Bar-Anan Y, Bergh R, Cai H, Gonsalkorale K, Kesebir S, Maliszewski N, Neto F, Olli E, Park J, Schnabel K, Shiomura K, Tulbure BT, Wiers RW, . . . Greenwald AG (2009) National differences in gender-science stereotypes predict national sex differences in science and math achievement. Proc Natl Acad Sci 106(26):10593–10597. https://doi.org/10.1073/pnas.0809921106 O’Shea BA, Watson DG, Brown GDA, Fincher CL (2019) Infectious disease prevalence, not race exposure, predicts both implicit and explicit racial prejudice across the United States. Soc Psychol Personal Sci 194855061986231. https://doi.org/10.1177/1948550619862319 Ofosu EK, Chambers MK, Chen JM, Hehman E (2019) Same-sex marriage legalization associated with reduced implicit and explicit antigay bias. Proc Natl Acad Sci U S A 116(18):8846–8851. https://doi.org/10.1073/pnas.1806000116 Orchard J, Price J (2017) County-level racial prejudice and the black-white gap in infant health outcomes. Soc Sci Med 181:191–198. https://doi.org/10.1016/j.socscimed.2017.03.036 Oswald FL, Mitchell G, Blanton H, Jaccard J, Tetlock PE (2013) Predicting ethnic and racial discrimination: a meta-analysis of IAT criterion studies. J Pers Soc Psychol 105(2):171–192. https://doi.org/10.1037/a0032734 Pager D (2003) The mark of a criminal record. Am J Sociol 108(5):937–975. https://doi.org/10. 1086/374403 Pager D (2007) The use of field experiments for studies of employment discrimination: contributions, critiques, and directions for the future. Annals of the American Academy of Political and Social Science 609(1):104–133. https://doi.org/10.1177/0002716206294796 Paluck EL, Green SA, Green DP (2018) The contact hypothesis revisited. Behav Public Policy 1–30. https://doi.org/10.2139/ssrn.2973474 Payne BK, Rucker JM (2022) Explaining the spatial patterning of racial disparities in traffic stops requires a structural perspective: further reflections on Stelter et al. (2022) and Ekstrom et al. (2022). Psychol Sci:095679762110566. https://doi.org/10.1177/09567976211056641 Payne BK, Vuletich HA, Brown-Iannuzzi JL (2019) Historical roots of implicit bias in slavery. Proc Natl Acad Sci U S A 116(24):11693–11698. https://doi.org/10.1073/pnas.1818816116 Payne BK, Vuletich HA, Lundberg KB (2017) The bias of crowds: how implicit bias bridges personal and systemic prejudice. Psychol Inq 28(4):233–248. https://doi.org/10.1080/ 1047840X.2017.1335568 Payne BK, Vuletich H, Lundberg KB (2022) Critique of the bias-of-crowds model simply restates the model: reply to Connor and Evers (2020). Perspect Psychol Sci 17(2):606–610. https://doi. org/10.1177/1745691621997394 Pearman FA (2021) Collective racial bias and the black-white test score gap. Race Soc Probl. https://doi.org/10.1007/s12552-021-09347-y Pettigrew TF, Tropp LR (2006) A meta-analytic test of intergroup contact theory. J Pers Soc Psychol 90(5):751–783. https://doi.org/10.1037/0022-3514.90.5.751 Rae JR, Newheiser A-K, Olson KR (2015) Exposure to racial out-groups and implicit race bias in the United States. Soc Psychol Personal Sci 6(5):535–543. https://doi.org/10.1177/ 1948550614567357 Rae JR, Skinner-Dorkenoo AL, Reiman AK, Schmid K, Hewstone M (2022) Mixed evidence for interactive effects of outgroup proportions and intergroup contact on racial bias in the United States. Soc Psychol Personal Sci 13(2):476–489. https://doi.org/10.1177/19485506211027756
25
Evidence of Covariation Between Regional Implicit Bias and. . .
613
Ravary A, Baldwin MW, Bartz JA (2019) Shaping the body politic: mass media fat-shaming affects implicit anti-fat attitudes. Personal Soc Psychol Bull 014616721983855. https://doi.org/10. 1177/0146167219838550 Reuben E, Sapienza P, Zingales L (2014) How stereotypes impair women’s careers in science. Proc Natl Acad Sci U S A 111(12):4403–4408. https://doi.org/10.1073/pnas.1314788111 Riddle T, Sinclair S (2019) Racial disparities in school-based disciplinary actions are associated with county-level rates of racial bias. Proc Natl Acad Sci U S A 116(17):8255–8260. https://doi. org/10.1073/pnas.1808307116 Sadler M, Devos T (2020) Ethnic diversity matters: putting implicit associations between weapons and ethnicity in context. Group Processes and Intergroup Relations 23(2):285–300. https://doi. org/10.1177/1368430218796933 Sawyer J, Gampa A (2017) Implicit and explicit racial attitudes changed during black lives matter. Personal Soc Psychol Bull 44(7):1–21 Shin RQ, Smith LC, Welch JC, Ezeofor I (2016) Is Allison more likely than Lakisha to receive a callback from counseling professionals? A racism audit study. Couns Psychol 44(8):1187–1211. https://doi.org/10.1177/0011000016668814 Smedley, B. D., Stith, A. Y., & Nelson, A. R. (2003). Unequal treatment: confronting racial and ethnic disparities in health care. National Academic Press. https://doi.org/10.17226/12875 Smith CD (2009) Deconstructing the pipeline: evaluating school-to-prison pipeline equal protection cases through a structural racism framework. Fordham Urban Law Journal 36. https://doi.org/ 10.2139/ssrn.2267133 Spencer SJ, Steele CM, Quinn DM, Crocker J, Eron L, Markus H, Myers D, Nisbett R, Von Hippel W (1999) Stereotype threat and women’s math performance. J Exp Soc Psychol 35:4–28. https://doi.org/10.1006/jesp.1998.1373 Steele CM (1997) A threat in the air: how stereotypes shape intellectual identity and performance. Am Psychol 52(6):613–629. https://doi.org/10.1037/0003-066X.52.6.613 Stelter M, Essien I, Sander C, Degner J (2022) Racial bias in police traffic stops: white residents’ county-level prejudice and stereotypes are related to disproportionate stopping of black drivers. Psychol Sci 095679762110512. https://doi.org/10.1177/09567976211051272 Storage D, Charlesworth TES, Banaji MR, Cimpian A (2020) Adults and children implicitly associate brilliance with men more than women. J Exp Soc Psychol 90. https://doi.org/10. 1016/j.jesp.2020.104020 Tilcsik A (2011) Pride and prejudice: employment discrimination against openly gay men in the United States. Am J Sociol 117(2):586–626. https://doi.org/10.1086/661653 Walton GM, Cohen GL (2003) Stereotype lift. J Exp Soc Psychol 39:456–467. https://doi.org/10. 1016/S0022-1031(03)00019-2 Zestcott CA, Ruiz JM, Tietje KR, Stone J (2021) The relationship between racial prejudice and cardiovascular disease mortality risk at the state and county level. Ann Behav Med. https://doi. org/10.1093/abm/kaab103
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
26
Smriti Sharma
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What Do the Numbers Show? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Understanding the Sources of Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Demographic Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Credit Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Customer Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bridging the Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Affirmative Action in Government Contracting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reducing Unconscious Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Training Programs and Role Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
616 617 618 618 620 622 624 624 626 627 628 628
Abstract
Ethnic and racial minorities are underrepresented in business ownership across the world. This chapter focuses on racial and ethnic disadvantage and discrimination in rates of entrepreneurship and self-employment, and entrepreneurial success. It discusses the literature on the sources of gaps and ways to potentially bridge the racial and ethnic gaps in self-employment and entrepreneurial performance.
Revised version: July 2022. Prepared for the Handbook on Economics of Discrimination and Affirmative Action, edited by Ashwini Deshpande. I thank Saurabh Singhal for comments on an earlier draft. All errors are mine. S. Sharma (*) Newcastle University Business School, Newcastle, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_11
615
616
S. Sharma
Keywords
Discrimination · Self-employment · Entrepreneurship · Race · Ethnicity
Introduction The view that self-employment is a means for marginalized groups to escape poverty is not a new one. The idea is that self-employment and entrepreneurship serve as a route to financial independence, economic advancement, and social mobility for disadvantaged ethnic and racial minorities and provide an alternative to wage employment in the labor market where discrimination is prevalent (e.g., Clark and Drinkwater 2000; Dunn and Holtz-Eakin 2000; Fairlie and Woodruff 2010). Furthermore, to the extent that minority business owners are more likely to hire workers from their own group, it has been argued that promoting business growth among disadvantaged groups is a useful avenue for improving employment opportunities for those groups. Several governments across the world provide loans, preferential procurement programs, tax exemptions to small businesses, and entrepreneurial training and additional programs to boost small business ownership and performance among minorities and women. This emphasis on self-employment for disadvantaged minorities is also underpinned by the assumption that discriminatory attitudes that are prevalent among employers in labor markets are virtually non-existent in other markets – credit, land, and product – that are intricately linked to the success of entrepreneurial ventures. However, research shows that minority groups remain underrepresented in this sector and that labor market discrimination can spill over into self-employment and related markets, creating apparently discriminatory outcomes in these markets too (Coate and Tennyson 1992). This chapter focuses on racial and ethnic disadvantage and discrimination in rates of entrepreneurship and self-employment and measures of entrepreneurial success. In doing so, it will discuss evidence from developed and developing countries. Clearly, work using data from the United States dominates the existing literature. However, the extent and nature of self-employment varies across countries. Gindling and Newhouse (2014) estimate that a third of workers in low- and middle-income countries are self-employed as compared to less than 10% of the labor force in highincome countries. This chapter proceeds as follows. Section “What Do the Numbers Show?” provides a brief overview of the incidence of self-employment across racial and ethnic groups in some countries. Section “Understanding the Sources of Gaps” discusses the literature that focuses on the sources of the gaps. Section “Bridging the Gaps” discusses the research on ways to potentially bridge these racial and ethnic gaps in self-employment. Section “Conclusion” concludes.
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
617
What Do the Numbers Show? In the United States, whites have historically had higher self-employment rates than blacks and Hispanics (e.g., Blanchflower 2009; Fairlie and Meyer 2000; Fairlie and Woodruff 2010). For instance, as per the US Bureau of the Census in 1993, approximately 11.6% of white workers were self-employed. On the other hand, only 3.8% of blacks were self-employed. While there has been some convergence over time, differences persist. Hipple and Hammond (2016) report that according to the US Bureau of Labor Statistics in 2015, approximately 11% of white and 9.6% of Asian Americans were self-employed, relative to 5.2% of blacks and 8.3% of Hispanics. Research has also shown that businesses owned by whites and Asian-Americans have higher sales, profits, number of employees, and higher survival rates than those owned by blacks and Hispanics (e.g., Fairlie and Robb 2007; Fairlie and Woodruff 2010). Fairlie and Robb (2007) document that “. . . the entire distribution of business net profits before taxes for black-owned firms is to the left of the distribution for white-owned firms. . .” (p. 293). Fairlie (2020) finds that African American business owners were hardest hit in the first 3 months after the onset of the COVID-19 pandemic in March 2020, followed by Latino and Asian business owners. He goes on to suggest racial differences in the scale of business as well as distribution across industrial sectors as the key explanations for these groups being adversely affected. In India too, the historically marginalized lower castes and tribes are significantly underrepresented in the ownership of enterprises, also relative to their population shares. Deshpande and Sharma (2013), using data from the Micro, Small, and Medium Enterprises (MSME) Census that covers the registered manufacturing sector, find that while in 2001–2002, 7.7% and 3.5% of firms were owned by Scheduled Castes (SCs) and Scheduled Tribes (STs) respectively, these figures were somewhat smaller at 6.3% and 3% in 2006–2007 Census. Iyer et al. (2013), using data from the Economic Census of 2005 calculate that SCs and STs owned 10% and 4% of non-farm enterprises, respectively, and these numbers were fairly similar in 1990 and 1998 Censuses. There is a small improvement over time such that in the 2013–2014 Economic Census, these numbers are 11% and 5% for SCs and STs, respectively. Further, enterprises owned by SCs and STs are less profitable, smaller in terms of number of employees, use mainly family labor or are owneroperated, rely less on external finance and operate mostly in the unregistered sector as compared to enterprises owned by non-SCSTs (Deshpande and Sharma 2016; Iyer et al. 2013; Thorat and Sadana 2009; Jodhka 2010). Deshpande and Sharma (2016) also note that the caste gaps in enterprise incomes are larger at the lower end of the earnings distribution. For South Africa, Leibbrandt et al. (2010) document racial gaps in selfemployment over time. Using data from National Income Dynamics Survey of 2008, they estimate a self-employment rate of 10.9% among Africans, 5.9% among the Colored, 17.7% among Whites, and 20% among Indians. The self-
618
S. Sharma
employment rate among Africans represents a substantial increase from just under 2% in 1993. Colored workers started out at roughly the same self-employment rates as that of Africans in 1993 but only increased by about half the rate of Africans by 2008.
Understanding the Sources of Gaps A persistent question is why racial and ethnic gaps exist in participation in selfemployment and entrepreneurial outcomes. This section will discuss the literature on the various sources of the racial gaps in entry and exit from self-employment and performance in self-employment. It will also discuss the sources of discrimination. Discrimination is usually categorized as one of two types. The first, “taste-based discrimination,” postulates that one is prejudiced against a particular group, and therefore chooses to discriminate against its members because one receives disutility from interacting with them. Second, “statistical discrimination” occurs against a particular group because productivity is imperfectly observed, and the group is judged as having lower productivity on average either due to exogenous differences or as part of a self-fulfilling equilibrium. Alternatively, the group’s productivity may be perceived to have a different variance thereby making the signal of its productivity less informative. While statistical discrimination assumes that beliefs are correct or that behavior is driven by rational expectations, in recent work, Bohren et al. (2019) focus on “inaccurate statistical discrimination” wherein they argue that an individual’s beliefs about productivity of different groups may be incorrect.
Demographic Characteristics A lot of the existing work has focused on disparities in human capital and financial capital as the key factors explaining racial, ethnic, and caste gaps in entry into selfemployment, business success, and exit from self-employment (e.g., Deshpande and Sharma 2016; Fairlie and Meyer 2000; Fairlie and Woodruff 2010; Fairlie 2018; Lofstrom and Wang 2009). Higher wealth in the form of personal net worth, home ownership, and home equity leads to favorable credit histories and can be used as collateral enabling greater access to start-up finance. It also provides a safety net in the early stages of business formation. In the United States, the low mean and median wealth among blacks and Hispanics are a key barrier limiting their ability to launch new businesses. Fairlie (1999) estimates that, “. . . quadrupling current black asset levels would result in only a 15% reduction in the racial gap in entry rate.” (p. 103). Differences in family background partially explain why whites display a greater propensity to start their own businesses and why they are larger and more profitable than black-owned firms (e.g., Fairlie 1999; Fairlie and Robb 2007; Hout and Rosen 2000). The influence of parents can occur through the intergenerational transmission of general and business-specific human capital, opportunities for relevant business
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
619
experience, access to financial resources, and parents acting as role models, increasing children’s interest in self-employment (Dunn and Holtz-Eakin 2000). Using regression-based decompositions, Fairlie (2018) finds that low level of wealth is the key factor associated with lower business formation rates among blacks and Hispanics in the United States explaining 21% and 49% of the white-black and white-Hispanic business formation gap, respectively. Further, education emerges as the key factor explaining racial differences in business performance. Deshpande and Sharma (2016) find that completed years of education contributes around 40% of the explained component of the business income gap between the upper and lower castes in India. Networks are a valuable form of social capital and can enhance self-employment performance by enabling information acquisition, providing access to credit and social insurance, and helping to seek better market opportunities, particularly in the presence of market imperfections. In developing countries in particular, social networks, defined mostly along kinship lines, are a key determinant of business formation and firm performance, and business activity tends to revolve heavily around these community networks (e.g., Damodaran 2008; Fafchamps 2000; Munshi 2014). Using historical data from late nineteenth and early twentieth century on Indian entrepreneurship, Gupta et al. (2020) find that community connections, primarily based upon caste and religion, were important predictors of early entry into business and that the presence of communities in earlier stages determined their presence in later decades too. Similarly, clans have played a large role in the growth of private enterprise in China by allowing for substitution of formal contract enforcement with social trust and long-term relationships (Dai et al. 2018). However, excessive reliance on community networks also can have some negative consequences. Focusing on a knitting garment production cluster in southern India, Banerjee and Munshi (2004) show that differential access to capital across two business communities on account of strong social ties results in substantial misallocation of capital with capable members of the newer business community losing out. Further, Alby et al. (2020) find that social redistributive norms prevalent in sub-Saharan Africa whereby the wealthy are expected to share their resources with their extended family results in local Africans either not entering entrepreneurship or having poorly performing enterprises compared to non-African entrepreneurs (such as the Indians in East Africa and the Lebanese in West Africa) who are not subject to such norms. Spatial factors have also been shown to play an important role in minority selfemployment. It has been suggested that self-employment of minority groups may be more prevalent in geographical clusters or enclaves of co-ethnics. Such communities may have higher demand for ethnic-specific goods and services, provide easy access to information, credit, and a pool of workers (Parker 2004). Conversely, ethnic enclaves might discourage self-employment because of increased competition from other co-ethnics and because ethnic enclaves might be economically disadvantaged. The empirical evidence on the relationship between self-employment and ethnic enclaves is mixed with some studies finding a negative association (e.g., Clark
620
S. Sharma
and Drinkwater 2000, 2002; Bates and Robb 2014) while others document a positive relationship (e.g., Borjas 1986; Evans 1989). Finally, an emerging strand of research has examined the role of differences in inherent preferences and aspirations for entrepreneurship among minority groups. For example, Koellinger and Minniti (2006) do not find blacks to have less preference for business ownership as compared to whites in the United States. Goel and Deshpande (2020) find that earnings considered as remunerative from selfemployment vary across castes such that relative to the upper castes, SCs and STs perceive 5–19% lower amounts as being remunerative. This could be due to internalized expectations of discrimination, lower self-valuation, or a self-imposed ceiling on what they think they can earn based on the experience of other members of their caste group.
Credit Access Another important aspect is that of access to credit and the terms of credit offered to the self-employed that can affect business outcomes. Coate and Tennyson (1992) study credit discrimination assuming that lenders are unable to observe entrepreneurial ability. Individuals from a group discriminated against in the labor market will receive less favorable terms in the credit market since lenders know that for such individuals, the opportunity cost of entering self-employment is lower, and, thus, they are willing to take more risks. Such groups will be charged higher interest rates, thereby reducing the expected returns from self-employment. Empirical evidence has been documented on the presence of racial and ethnic differences in lending to small businesses by formal banks and financial organizations which manifests in the form of loan denials and rates of interest charged on approved loans being higher for black-owned businesses than whites in the United States (e.g., Cavalluzzo et al. 2002; Cavalluzzo and Wolken 2005; Blanchard et al. 2008; Asiedu et al. 2012). Goraya (2019) finds that correcting for caste-based misallocation of capital on account of differential access to credit in India would result in substantial increases in total factor productivity of small and medium enterprises and reduce the productivity gap between low caste and high caste owned firms. Most early empirical studies used observational survey data to estimate the coefficient of race or ethnicity in a regression with bank lending decisions as outcome variables after controlling for an increasingly rich set of characteristics. For example, Blanchflower et al. (2003) found that black-owned businesses in the United States are more than twice as likely to have a loan application denied and that interest rates charged on approved loans is higher for them as compared to whiteowned firms, even after accounting for personal wealth and creditworthiness, etc. The probability of existing loan renewals is lower for black- and Hispanic-owned businesses (Asiedu et al. 2012). Cavalluzzo and Wolken (2005) found that differences in endowments explain only about a third of the difference in denial rates across groups. For the case of India, Kumar (2013) finds that public sector banks
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
621
operating in areas with more upper castes tend to discriminate more against low-caste loan applicants. However, there are important concerns with the interpretation of a significant and negative coefficient of race or ethnicity as reflecting discrimination in studies using observational data. A key issue is that of omitted variables bias wherein the researcher is not able to identify all the information that a bank or lending organization observes when making lending decisions. Moreover, how loan applications are allocated across loan officers is usually not observed in datasets. Therefore, any differences could be driven by omitted variable bias associated with the unobserved characteristics of borrowers and loan officers. Another concern is that of selection bias in the form of ex-ante racial differences in loan application rates wherein minority groups may have a lower demand for credit or be less likely to apply on account on expected discrimination and differential treatment by banks. For instance, Fairlie et al. (2020) find that black entrepreneurs are about thrice as likely to not apply for credit when needed due to fear of loan denial. Kumar and Venkatachalam (2019) find that there are caste-based differences in loan application rates in India with the SCs and STs being about 16–20% less likely to apply than the upper castes. One potential solution is to implement a correspondence study where loan requests are randomly assigned to loan officers while randomizing the race or ethnicity of the applicant. While correspondence studies have been used to study discrimination in labor markets by sending fictitious resumes and estimating identity-specific call-back rates for job interviews (e.g., Bertrand and Mullainathan 2004), such experiments in formal credit markets are largely unfeasible since credit scores from credit rating agencies are used early in the lending process to check the applicants’ background. Recent studies have been able to overcome some of these concerns by focusing their attention on alternative forms of lending such as peer-to-peer lending and crowdfunding platforms that allow private individuals to transact with one another without involving banks. In these online platforms, as the information visible to lenders when making funding decisions is what has been posted by the borrowers/ entrepreneurs, researchers observe the same information as lenders about the borrower. Pope and Sydnor (2011) examine racial discrimination in a peer-to-peer lending platform that pools funds from individual lenders to create loans for borrowers that do not require collateral. Borrowers create loan listings and post information about their credit profile. They can also post other information and photos (which allows race to be inferred) which are not usually asked in applications by traditional lending organizations. They find that compared to whites, listings with blacks in photos are almost 30% less likely to get funding. Furthermore, conditional on getting a loan, blacks pay 60–80 basis points more than whites with similar creditworthiness. Younkin and Kuppuswamy (2018) use data from projects on Kickstarter – a crowdfunding site where entrepreneurs solicit funding from online communities and offer products and services in exchange – to find that controlling for a range of project characteristics, African American entrepreneurs’ projects are less likely to be funded than those of white entrepreneurs. Moreover, black
622
S. Sharma
entrepreneurs’ projects receive lesser funding and have fewer backers. In an experiment that holds project type constant and varies the details of the entrepreneur, they find that the perceived quality of black entrepreneurs’ projects is lower. Some studies have also used exogenous policy changes to highlight the role of credit access. Chatterji and Seamans (2012) examine the effect of deregulation of credit card markets in the United States in the late 1970s on new business formation and entrepreneurship by race. Blacks are more likely to finance their businesses using credit card debt than whites due to differential access to traditional loans. Using a difference-in-differences approach, they find that residing in a state that increased the credit card interest rate ceiling led to a significant increase in likelihood of self-employment, especially for black entrepreneurs and for blacks in states with a history of discrimination (proxied by former slave state, anti-miscegenation laws, and absence of fair housing laws). There is also some literature examining the calculation of credit scores that are now used almost universally by lenders to determine creditworthiness, and whether these credit scores themselves may be racialized. Using data from new start-ups, Henderson et al. (2015) examine group-based differences in credit scores and find that whites are treated more favorably in credit score determination than African Americans are with the same firm characteristics and owner characteristics. On the other hand, Robb and Robinson (2018) find that black borrowers do not face worse credit scores than whites do. They summarize, “If anything, minority business owners experience actual repayment histories that are slightly worse than would be predicted by their forward-looking credit scores.” (p. 442).
Customer Preferences A final factor that constitutes an important source of racial and ethnic gaps in entrepreneurship and that poses significant challenges for performance of minority-owned businesses is customer preferences and consumer discrimination. In fact, discrimination by consumers can also explain discrimination by other actors. For instance, employers may be reluctant to hire minorities if consumers do not like engaging with minority salespersons (e.g., Laouenan 2017; Bar and Zussman 2017). Borjas and Bronars (1989) study consumer discrimination and find that relative gains of entering self-employment are reduced for ethnic minorities because they have to compensate white consumers by lowering prices. Laouenan and Rathelot (2022) find that Airbnb (an online home rental marketplace) hosts from minority groups (African Americans in North America and Arabs, Muslims, and Sub-Saharan Africans in North America and Europe) charge 3.2% less for similar listings in the same neighborhoods. They estimate that the entire price gap can be attributed to statistical discrimination. In a similar vein, Tjaden et al. (2018) discover ethnic discrimination in online carpooling marketplaces in Germany against drivers with Arab/Turkish/Persian sounding names as compared to drivers with typically German names.
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
623
As discussed above, while correspondence studies are not feasible in the market for formal credit, the availability of online marketplaces, in particular, has provided ample opportunities for researchers to conduct experiments where they can randomize the race of the seller (with race being signaled by race-typical names or skin color in photos), holding product features constant, to test whether buyers avoid transacting with minority sellers and if they treat minority sellers differently. However, names to signal race is not without its problems. Certain names may also be perceived as being more prevalent in some socioeconomic classes apart from race, providing a potential confound. Doleac and Stein (2013) study racial discrimination in the product market by posting online advertisements to sell used iPods on Craigslist (a classified advertisements website) in the United States. They experimentally varied the race of the seller by including a photograph of a white or black hand holding the product. They find that black sellers receive 13% fewer responses, less trusting responses, and 18% fewer offers, while holding constant other advertisement and market-related characteristics, and this results in less valuable offers as compared to those made to white sellers. They find black sellers’ outcomes to be particularly poor in markets with the most racial isolation and property crime, suggesting the importance of statistical discrimination. However, the authors also find that black sellers do better in markets with predominantly black populations, suggesting that disparities in outcomes may be driven partially by in-group bias. Nunley et al. (2011) examine racial discrimination in online product auctions on eBay. They sell brand new unopened identical goods under different seller names where names as used as an indicator for race. They document the presence of in-group bias. More specifically, they find that white-named sellers receive higher prices than black-named sellers for products geared toward white buyers and that black-named sellers receive higher prices than white-named sellers for products that are targeted toward black buyers. However, the in-group bias in the form of price differences dissipates as sellers develop credible reputations by way of higher number of positive reviews, suggesting that statistical discrimination drives the price differences. In an online experiment similar to a crowdfunding scenario, Younkin and Kuppuswamy (2019) vary the race of the potential founder, and for the same product, they find respondents (similar to backers of a project) to recommend a lower price for the black founder than for the white founder, suggesting that consumers use a seller’s race to assess the appropriate price for a product. They find some evidence that this reflects an assumption about differences in the relative effort level of white and minority founders, consistent with statistical discrimination. Zussman (2013) conducted an experiment in Israel to establish whether Arab sellers are discriminated against relative to Jewish sellers. They posted fictitious advertisements on a classifieds website for used cars and randomly assigned distinctly ethnic and representative Arab and Jewish first names to signal the ethnic identity of the sellers. They find that advertisements posted by Arab sellers receive only half as many phone calls as advertisements posted by Jewish sellers do. Further, both Arab and Jewish buyers discriminate against Arab sellers. Using survey data,
624
S. Sharma
they show that discrimination by Jewish buyers is associated with their belief that Arabs are more likely than Jews to cheat, but not with other beliefs, in particular those capturing prejudicial tastes. This leads them to conclude that discrimination against Arab sellers is statistical rather than taste-based. Building on the previous study, Zussman (2016) analyses how the Israeli-Palestinian conflict affects interethnic transactions in the Israeli market for used cars. They find that when violence intensifies, the number of transactions between Arab sellers and Jewish buyers drops while the number of transactions between Arab sellers and Arab buyers increases. On the other hand, they find no effect on the volume of transactions between Jewish sellers and Jewish and Arab buyers. For developing countries, there are very few empirical studies on consumer discrimination even though there exists anecdotal evidence which shows that customer discrimination is prevalent against minority-owned businesses, and this provides a promising avenue for new research. A recent paper by Siddique et al. (2020) is an important contribution to this literature, utilizing features of local product markets in a developing country. They examine whether professional rice buyers discriminate against ethnic minority rice farmers in Bangladesh. In a field experiment, they ask ethnic majority professional rice buyers to evaluate the quality of rice and quote a price where they randomly assign typically ethnic majority or minority sounding names to the rice samples. They find no ethnic differences in buyers’ assessment of rice quality based on the assigned farmer ethnicity. However, they do find that ethnic majority local buyers who have local monopsony power quoted a lower price for rice associated with ethnic minority sounding names as compared to quotation for rice produced by farmers with ethnic majority sounding names. On the other hand, for buyers operating in a competitive environment, there was no ethnic difference in prices quoted. The evidence of ethnic differences in price and not in assessed quality is therefore consistent with taste-based discrimination.
Bridging the Gaps Affirmative Action in Government Contracting To counteract the effects of the barriers faced by minority-owned firms, a policy is that of affirmative action in government contracting. The rationale is that such policies help minority firms to overcome problems related to consumer and credit discrimination, alleviate liquidity constraints and network limitations, and help to increase employment of minority group members. The current empirical literature evaluating the performance of such procurement programs is based primarily on data from United States. In the United States, this has operated in the form of “contracting set-asides” wherein a certain percentage of the number or value of government contracts should be awarded to minority-owned enterprises. In India, a public procurement policy introduced in 2012 requires all government ministries, departments and public sector units to procure 25% of the annual value of goods and services from registered micro and small enterprises
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
625
(MSEs) with 4% being from MSEs owned by Scheduled Castes and Scheduled Tribes and 3% from women-owned MSEs. Similar policies have been enacted in Malaysia as part of its New Economic Policy (NEP) in the 1970s which requires 30% of government contracts to be reserved for the Malay majority. Despite a number of these other countries also having such supplier diversity programs in place, their effectiveness has not yet been evaluated in a rigorous manner. Affirmative action in procurement programs remains an intensely debated issue on grounds of fairness and efficiency. Such set-aside programs have been challenged judicially and their legality called into question across many parts of the United States resulting in changes in application over time. Blanchflower and Wainwright (2005) provide a detailed description of the history of affirmative action in the United States construction industry. Chatterji et al. (2014) use the variation induced by staggered timing of the contracting set-aside programs across US cities to estimate impacts on selfemployment rates of African Americans during 1979–1989. They find that black business ownership rates increased after the introduction of this program and resulted in a decline in the black-white self-employment gap by 35–40%. The gains in self-employment among blacks were concentrated in industries heavily affected by set-asides (including sectors such as construction, transportcommunication, retail trade and repair services) and were larger for the better educated. They also find that program appears to have led to a reallocation of selfemployment from whites to blacks, as there is little change in the overall selfemployment rates. As the self-employment rate of blacks was declining in the 1970s, their estimates suggest that introduction of such programs helped to stem such declines. Blanchflower and Wainwright (2005) and Marion (2009a) document that removal or curtailment of such set-aside programs with race-neutral programs results in a sharp drop in participation of minority-owned firms. Going a step further and examining the intensity of the program by exploiting the variation in the states’ affirmative action goals over time, Marion (2011) finds that more intensively used affirmative action in the highway construction industry increased purchases from minority-owned firms, particularly in states where the rules were better enforced. However, there was little impact on purchases from women-owned firms. These policies are however not without their drawbacks. For instance, Bates and Williams (1996), using data on firms during 1987–1991, found that business survival rates were lower among younger firms that relied excessively on government contracts through affirmative action, possibly indicating that such firms are “fronts” for non-minority business owners seeking to benefit from a government contract. These programs can also be costly – Marion (2009b) estimates that the elimination of the affirmative action program for state-funded contracts in California resulted in significant financial cost savings for the government. Of course, this needs to be traded off against the societal benefits of increasing representation of minority groups.
626
S. Sharma
Reducing Unconscious Bias This subsection will discuss some ways aimed at reducing unconscious and implicit bias by decision-makers. A large literature has established the presence of implicit bias and how stereotyping works to the detriment of minority groups. In the context of online marketplaces, studies tend to find that a potential response by discriminated against groups is to manipulate their identity to avoid adverse outcomes. Zussman (2013) tests whether this form of identity manipulation improves outcomes for Arab sellers in Israel. They run an experiment in which fictitious advertisements are posted for used cars but randomly leave a third of the advertisements with the name field blank (while the rest of the advertisements feature sellers with either Arab or Jewish names). Their results indicate that identity manipulation works such that the outcomes for advertisements without a seller name fare better than for advertisements posted with an Arab name. There is also anecdotal and descriptive evidence showing the minority individuals conceal obvious indicators associated with their identity to improve their chances of success. In an experiment, Younkin and Kuppuswamy (2018) find that hiding indicators of the business founders’ race (“whitewashing”) by removing photos and replacing the name with initials on a crowdfunding site improves the perceived project quality to be at par with that of a white founder. Such approaches of suppressing group attributes are also validated by recent research that sheds light on differences in costly attention allocated by decisionmakers toward minority and majority candidates. In experiments using a correspondence study design, Bartos et al. (2016) enhanced candidates’ fictitious resumes by providing hyperlinks to applicants’ websites as a way to assess the share of employers allocating additional attention to applicants by opening the links to garner further information. On the whole, they find lower callback rates for minority candidates. Further, they found evidence of “attention discrimination” with more employers taking time to open the links of the majority group applicants compared to those of minority candidates. Their findings also have implications for timing the point at which identity-related information is revealed to minimize the asymmetry in attention. This has been adopted in some European countries on a pilot basis with selected companies wherein for the first screening step of the job application process, resumes were anonymized with any identifying information hidden. Improving the representation of minority groups in decision-making roles can also enable a reduction in bias, for instance, in the allocation of credit to minority business owners. While not focused on lending to entrepreneurs per se, Fisman et al. (2017) use data from a large public sector bank in India that follows an official policy of exogenous geographical rotation of loan officers across the bank’s branches leading to variation in the matches between loan officers’ and borrowers’ religion and caste. They find that probability of loan approval and total amount of new loans approved for a group is higher when the loan officer is also from the same religion or caste in India. As loan officers are predominantly high caste to begin with, this leads to lower amounts lent to low caste applicants. They argue that this is due to in-group bias rather than animus toward the out-group. Using the same setting as the
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
627
aforementioned paper, Fisman et al. (2020) find that Hindu loan officers who were exposed to communal (Hindu-Muslim) riots in their hometown prior to joining the bank lent less to Muslim borrowers and more to Hindu borrowers.
Training Programs and Role Models To the extent that low business ownership and worse firm performance among minorities is due to a lower stock of business-specific and managerial human capital and lack of relevant sector experience, business and skill training and exposure to successful role models can help with reducing the racial gap in firm ownership and performance. While there are papers focusing on improving the performance of female-owned firms and reducing the gender gap in firm performance (e.g., Field et al. 2010; Field et al. 2016), there is a dearth of literature focusing on firm owners belonging to historically disadvantaged groups. A recent meta-analysis of traditional training programs by McKenzie (2021) concludes that the effects of training on sales and profits are modest but significant. A few studies have found simple trainings focusing on a few heuristics or financial rules-of-thumb to be potentially effective for subsistence enterprises and for less educated firm owners (e.g., Beaman et al. 2014; Drexler et al. 2014). Another recent set of studies has focused on developing an entrepreneurial mindset or managerial capital by offering personal initiative or soft skills training or combining it with traditional training. These studies have found some promising results, at least in the short run but the quality of trainers is a key ingredient for program performance (e.g., Campos et al. 2017; Ubfal et al. 2022). Given that lower castes and ethnic minorities in developing countries tend to be concentrated mostly in the lower tail of the business income distribution and have lower levels of education, such simpler trainings might offer potential solutions to bridge some of the business-specific human capital gaps. Another aspect is that of role models and learning from peers. Minority individuals typically have few business-relevant role models to emulate, and this may affect their entrepreneurial spirit and ambitions. Role models also might serve an informational role by enhancing knowledge on the business and market access and help develop relevant networks. For instance, within a micro-entrepreneurship training program, Lafortune et al. (2018) introduce an intervention that exposes microenterprise owners to successful entrepreneur role models in Chile. They find increased business participation and income resulting from this exposure, driven by confidence rather than improved business knowledge. In a randomized evaluation in Kenya, Brooks et al. (2018) find that a one-to-one mentoring by someone from the same community and in the same industry is more effective at improving business profits in the short run among inexperienced microenterprise owners as compared to a formal business training. Narciso et al. (2018) investigate the impact of a set of role model interventions among ethnic minorities in Vietnam who lag behind the majority in enterprise ownership, with the objective of stimulating behavioral change. The intervention
628
S. Sharma
uses videos of successful entrepreneurs who had managed to escape poverty through entrepreneurial activities. By portraying role models belonging to the ethnic minority or the ethnic majority, the authors attempt to disentangle the role of information versus aspirations that may be shaped by seeing examples of successful others from the in-group. However, they do not find any discernible effect of this intervention on self-employment. This suggests that aspirational and informational constraints are potentially of second-order importance relative to harder constraints such as credit access.
Conclusion This chapter has provided an overview of the existing gaps in entry and exit from and performance in self-employment and entrepreneurship between racial and ethnic majority and minority groups. While there exists a large literature on identifying such gaps in the self-employment and entrepreneurship domain, there has clearly been much more research and innovation in the context of labor market discrimination. Further, while papers using correspondence studies do attempt to make a distinction between taste-based and statistical discrimination on the assumption of accurate beliefs, there is a need to account for inaccurate statistical discrimination where beliefs underpinning statistical discrimination may be inaccurate, as shown in recent work (Bohren et al. 2019; Laouenan and Rathelot 2022; Bordalo et al. 2016). This chapter also provides an overview of possible ways to bridge the gaps. As also outlined by Bertrand and Duflo (2017), there is a substantial need for new research identifying ways in which performance gaps between enterprises owned by different social groups and biases toward minority business owners can be mitigated. Finally, as evidenced in the chapter, most of the economics literature on ethnic and racial gaps in self-employment is based on data from the United States. Given the survivalist nature of self-employment and high rate of firm failure in developing countries, more research is needed on how to improve entrepreneurial performance in developing countries, with an additional focus on reducing ethnic and caste gaps in these settings.
References Alby P, Auriol E, Nguimkeu P (2020) Does social pressure hinder entrepreneurship in Africa? The forced mutual help hypothesis. Economica 87:299–327 Asiedu E, Freeman JA, Nti-Addae A (2012) Access to credit by small businesses: how relevant are race, ethnicity, and gender? American Economic Review: Papers & Proceedings 102(3): 532–537 Banerjee A, Munshi K (2004) How efficiently is capital allocated? Evidence from the knitted garment industry in Tirupur. Rev Econ Stud 71:19–42 Bar R, Zussman A (2017) Customer discrimination: evidence from Israel. J Labor Econ 35(4): 1031–1069
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
629
Bartos V, Bauer M, Chytilova J, Matejke F (2016) Attention discrimination: theory and Field experiments with monitoring information acquisition. Am Econ Rev 106(6):1437–1475 Bates T, Robb A (2014) Small business viability in America’s urban minority communities. Urban Stud 51(13):2844–2862 Bates T, Williams D (1996) Do preferential procurement programs benefit minority business? Am Econ Rev Pap Proc 86(2):294–297 Beaman L, Magruder J, Robinson J (2014) Minding small change among small firms in Kenya. J Dev Econ 108:69–86 Bertrand M, Duflo E (2017) Field experiments on discrimination. In: Banerjee A, Duflo E (eds) Handbook of Field experiments, vol 1, Elsevier Bertrand M, Mullainathan S (2004) Are Emily and Greg more employable than Lakisha and Jamal? A Field experiment on labor market discrimination. Am Econ Rev 94:991–1013 Blanchard L, Zhao B, Yinger J (2008) Do lenders discriminate against minority and women entrepreneurs? J Urban Econ 63(2):467–497 Blanchflower DG (2009) Minority self-employment in the United States and the impact of affirmative action programs. Ann Finance 5:361–396 Blanchflower DG, Wainwright J (2005). An analysis of the impact of affirmative action programs on self-employment in the construction industry. NBER working paper 6840 Blanchflower D, Levine P, Zimmerman D (2003) Discrimination in the small-business credit market. Rev Econ Stat 85(4):930–945 Bohren JA, Haggag K, Imas A, Pope D (2019) Inaccurate statistical discrimination: An identification problem. NBER working paper 25935 Bordalo P, Coffman K, Gennaioli N, Shleifer A (2016) Stereotypes. Q J Econ 131:1753–1794 Borjas G (1986) The self-employment experience of immigrants. J Hum Resour 21:485–506 Borjas G, Bronars (1989) Consumer discrimination and self-employment. J Polit Econ 97(3): 581–605 Brooks W, Donovan K, Johnson TR (2018) Mentors or teachers? Microenterprise training in Kenya. Am Econ J Appl Econ 10(4):196–221 Campos F, Frese M, Goldstein M, Iacovone L, Johnson HC, McKenzie D, Mensmann M (2017) Teaching personal initiative beats traditional training in boosting small business in West Africa. Science 357(6357) Cavalluzzo KS, Wolken JD (2005) Small business loan turndowns, personal wealth, and discrimination. J Bus 78(6):2153–2178 Cavalluzzo K, Cavalluzzo L, Wolken J (2002) Competition, small business financing, and discrimination: evidence from a new survey. J Bus 75(4):641–679 Chatterji AK, Seamans RC (2012) Entrepreneurial finance, credit cards, and race. J Financ Econ 106(1):182–195 Chatterji AK, Chay KY, Fairlie RW (2014) The impact of city contracting set-asides on black selfemployment and employment. J Labor Econ 32(3):507–561 Clark K, Drinkwater S (2000) Pushed out or pulled in? Self-employment among ethnic minorities in England and Wales. Labour Econ 7:603–628 Clark K, Drinkwater S (2002) Enclaves, neighbourhood effects and employment outcomes: ethnic minorities in England and Wales. J Popul Econ 15:5–29 Coate S, Tennyson S (1992) Labor market discrimination, imperfect information and selfemployment. Oxf Econ Pap 44:272–288 Dai R, Mookherjee D, Munshi K, Zhang X (2018) Community networks and the growth of private enterprise in China. Cambridge-INET working paper series no: 2018/11 Damodaran H (2008) India’s new capitalists: caste, business, and industry in modern India. Palgrave Macmillan, New York Deshpande A, Sharma S (2013) Entrepreneurship or survival? Caste and gender of small business in India. Economic and Political Weekly, pp 38–49 Deshpande A, Sharma S (2016) Disadvantage and discrimination in self-employment: caste gaps in earnings in Indian small businesses. Small Bus Econ 46(2):325–346
630
S. Sharma
Doleac JL, Stein LCD (2013) The visible hand: race and online market outcomes. Econ J 123(572): F469–F492 Drexler A, Fischer G, Schoar A (2014) Keeping it simple: financial literacy and rules of thumb. Am Econ J Appl Econ 6(2):1–31 Dunn TA, Holtz-Eakin DJ (2000) Financial capital, human capital, and the transition to selfemployment: evidence from intergenerational links. J Labor Econ 18(2):282–305 Evans MDR (1989) Immigrant entrepreneurship: effects of ethnic market size and isolated labor pool. Am Sociol Rev 54:950–962 Fafchamps M (2000) Ethnicity and credit in African manufacturing. J Dev Econ 61:205–235 Fairlie RW (1999) The absence of the African-American owned business: an analysis of the dynamics of self-employment. J Labor Econ 17(1):80–108 Fairlie RW (2018) Racial inequality in business ownership and income. Oxf Rev Econ Policy 34(4): 597–614 Fairlie RW (2020) The impact of COVID-19 on small business owners: evidence from the first 3 months after widespread social-distancing restrictions. Journal of Economic Management and Strategy 29(4):727–740 Fairlie RW, Meyer BD (2000) Trends in self-employment among black and white men, 1910–1990. J Hum Resour 35(4):643–669 Fairlie RW, Robb A (2007) Why are black-owned businesses less successful than white-owned businesses? The role of families, inheritances, and business human capital. J Labor Econ 25(2): 289–323 Fairlie RW, Woodruff CM (2010) Mexican-American entrepreneurship. The BE Journal ofEconomic Analysis & Policy 10(1 (Contributions)):Article 10 Fairlie RW, Robb A, Robinson DT (2020) Black and white: access to capital among minorityowned startups. Management Science, forthcoming Field E, Jayachandran S, Pande R (2010) Do traditional institutions constrain female entrepreneurship? A field experiment on business training in India. American Economic Review Papers & Proceedings 100:125–129 Field E, Jayachandran S, Pande R, Rigol N (2016) Friendship at work: can peer effects catalyze female entrepreneurship? Am Econ J Econ Pol 8:125–153 Fisman R, Paravisini D, Vig V (2017) Cultural proximity and loan outcomes. Am Econ Rev 107(2): 457–492 Fisman R, Sarkar A, Skrastins J, Vig V (2020) Experience of communal conflicts and intergroup lending. J Polit Econ 128(9):3346–3375 Gindling TH, Newhouse D (2014) Self-employment in the developing world. World Dev 56: 313–331 Goel D, Deshpande A (2020) Social identity and perceived income adequacy. Rev Dev Econ 24(2): 339–361 Goraya SS (2019) How does caste affect entrepreneurship? Birth versus worth. Barcelona GSE working paper 1104 Gupta B, Mookherjee D, Munshi K, Sanclemente M (2020) Community origins of industrial entrepreneurship in colonial India. Working paper Henderson L, Herring C, Horton HD, Thomas M (2015) Credit where credit is due? Race, gender, and discrimination in the credit scores of business startups. Rev Black Polit Econ 42:459–479 Hipple SF, Hammond LA (2016) Self-employment in the United States, Spotlight on statistics, march 2016. US Bureau of Labor Statistics, Washington DC Hout M, Rosen H (2000) Self-employment, family background, and race. J Hum Resour 35(4): 670–692 Iyer L, Khanna T, Varshney A (2013) Caste and entrepreneurship in India. Economic and Political Weekly, pp 52–60 Jodhka SS (2010) Dalits in business: self-employed scheduled castes in north-West India. Economic and Political Weekly, pp 41–48
26
Disadvantage and Discrimination in Self-Employment and Entrepreneurship
631
Koellinger P, Minniti M (2006) Not for lack of trying: American entrepreneurship in black and white. Small Bus Econ 27:59–79 Kumar SM (2013) Does access to formal agricultural credit depend on caste? World Dev 43: 315–328 Kumar SM, Venkatachalam R (2019) Caste and credit: a woeful tale? J Dev Stud 55(8):1816–1833 Lafortune J, Riutort J, Tessada J (2018) Role models or individual consulting: the impact of personalizing micro-entrepreneurship training. Am Econ J Appl Econ 10(4):222–245 Laouenan M (2017) Hate at first sight: evidence of consumer discrimination against African Americans in the US. Labour Econ 46:94–109 Laouenan M, Rathelot R (2022) Can information reduce ethnic discrimination? Evidence from Airbnb. Am Econ J Appl Econ 14(1):107–132 Leibbrandt M, Woolard I, McEwen H, Koep C (2010) Employment and inequality outcomes in South Africa, University of Cape Town: Southern Africa Labour and Development Research Unit Lofstrom M, Wang C (2009) Mexican-American self-employment: a dynamic analysis of business ownership. Res Labor Econ 29:197–227 Marion J (2009a) Firm racial segregation and affirmative action in the highway construction industry. Small Bus Econ 33:441–453 Marion (2009b) How costly is affirmative action? Government contracting and California’s proposition 2009. Rev Econ Stat 91(3):503–522 Marion J (2011) Affirmative action and the utilization of minority and women-owned businesses in highway procurement. Econ Inq 49(3):899–915 McKenzie D (2021) Small business training to improve management practices in developing countries: re-assessing the evidence for “training doesn’t work”. Oxf Rev Econ Policy 37(2): 276–301 Munshi K (2014) Community networks and the process of development. J Econ Perspect 28(4): 49–76 Narciso G, Newman C, Tarp F (2018) Information, identification or neither? Experimental evidence on role models in Vietnam. UNU-WIDER working paper 2018/185 Nunley JM, Owens MF, Howard RS (2011) The effects of information and competition on racial discrimination: evidence from a field experiment. J Econ Behav Organ 80(3):670–679 Parker S (2004) The economics of self-employment and entrepreneurship. Cambridge University Press, Cambridge Pope DG, Sydnor JR (2011) What’s in a picture? Evidence of discrimination from prosper.com. J Hum Resour, 46(1):53–92 Robb A, Robinson DT (2018) Testing for racial bias in business credit scores. Small Bus Econ 50: 429–443 Siddique A, Vlassopoulos M, Zenou Y (2020) Market competition and discrimination. IZA DP No. 13269 Thorat S, Sadana N (2009) Caste and ownership of private enterprises. Econ Polit Wkly XLIV(23): 13–16 Tjaden JD, Schwemmer C, Khadjavi M (2018) Ride with me – ethnic discrimination, social markets, and the sharing economy. Eur Sociol Rev 34(4):418–432 Ubfal D, Arraiz I, Beuermann DW, Frese M, Maffioli A, Verch D (2022) The impact of soft-skills training for entrepreneurs in Jamaica. World Dev 152:105787 Younkin P, Kuppuswamy V (2018) The colorblind crowd? Founder race and performance in crowdfunding. Manag Sci 64(7):3269–3287 Younkin P, Kuppuswamy V (2019) Discounted: the effect of founder race on the price of new products. J Bus Ventur 34:389–412 Zussman A (2013) Ethnic discrimination: lessons from the Israeli online market for used cars. Econ J 123:F433–F468 Zussman A (2016) Conflict and the ethnic structure of the marketplace: evidence from Israel. Eur Econ Rev 90:134–145
Discrimination in Credit
27
Sunil Mitra Kumar
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laws, Rules, and Institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mechanisms and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
634 635 638 641 642 644 645
Abstract
This chapter provides a thematic discussion of discrimination in credit. Through a selective review of the literature, I illustrate that caste, gender, and race are all persistent axes of discrimination in credit, and that discrimination has been shown to exist across diverse contexts. I examine the main conceptual tools used in this literature to shed light on the causal mechanisms that lead to discrimination, and in particular attempt to delineate the role of individuals, institutions, and formal regulation. By briefly exploring the links between discrimination in credit and group-based inequality in other arenas of economy and society, I argue why the implications of the former extend far beyond credit alone, and are a powerful force in shaping inequality more generally including across generations. Keywords
Discrimination · Credit · Formal and Informal Lending
S. M. Kumar (*) King’s India Institute and Department of International Development, King’s College London, London, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_14
633
634
S. M. Kumar
Introduction In the simplest understanding, discrimination signifies unequal treatment on the basis of group-belonging such as gender, caste, or race. This inequality must usually be unjustified for it to qualify as discrimination. In the realm of credit, inequality generally takes two forms: access to credit and terms including the price of credit. To offset the risk of default, lenders usually require collateral and information on past borrowing, both of which might be obtained through formal structures in the case of a bank, or situated hierarchical relationships in the case of informal lenders. In either context, these requirements result in unequal access or access at unequal terms, reflecting the existing unequal distributions of wealth and past borrowing. Discrimination can strengthen this inequality, again reflecting existing patterns of discrimination in other realms of economic and social life. Equally, laws, rules, and institutions can also play a role, either by actively perpetuating discrimination or else indirectly supporting existing discriminatory mechanisms. Writing as an economist, discrimination is defined in terms of two mechanisms that can operate simultaneously. Taste discrimination (Becker 1957) reflects animus or prejudice against members of some group, whereas statistical discrimination (Arrow 1973; Phelps 1972) is a rational response to limited information in situations where the decision-maker cannot observe a certain relevant attribute, and uses group-belonging as a proxy for this missing information. Intuitively, since lending can involve asymmetric information – for instance about the quality of investment for which a loan is being sought – situations involving lending decisions could a priori involve both taste and statistical discrimination. However, these definitions focus on the role of individual agents as decisionmakers and they demand certain extensions. Borrowers choose to seek loans, and might not do so or might apply for smaller amounts if they expect to be turned down by the lender. This implies that any assessment of discrimination needs to take into account borrowers’ expectations. If those expectations reflect patterns analogous to existing discrimination – the group more hesitant to seek credit is also the one more likely to be denied credit – then the historical experiences that explain expectations also ought to be accounted for in any analysis of discrimination. This aspect can also apply to understanding group-wise differences in the sources from which credit is sought, particularly in contexts where informal and formal borrowing exist side by side. A more nuanced understanding of discrimination additionally involves consideration for contextual factors that can support inequality. Even if discrimination might not be involved, caste, race, or gender inequalities can be cemented both over time as intergenerational processes or across arenas where several interlinked processes operate in tandem. Lending is a powerful example that highlights both aspects. Criteria for obtaining a loan such as the ability to provide collateral or show credit history are not only entirely objective but also very relevant – after all, they both exist to counter the risk taken on by the lender. Yet the ability to demonstrate credit history and to offer collateral depends on existing wealth and past access to credit, thereby perpetuating inequality including across generations. Equally, access to
27
Discrimination in Credit
635
credit can influence and be influenced by access to education, capital, and networks, thereby reflecting and strengthening inequalities or discrimination in these arenas as well. In addition, discrimination might be implicitly perpetuated by rules or laws that favor the members of one group over another, even if the decision-maker or institution is not per se guilty of discriminating (Pager and Shepherd 2008). Algorithms that determine credit scoring and loan decisions are one example, in that they represent part of the rules by which lending decisions take place. As we discuss below, there are substantive concerns about their role in perpetuating inequality and historical discrimination as well as suggestions for how to tackle this. A closely related concept is that of institutional discrimination. Individual loan officers operate within banks and their actions are shaped by formal rules, regulations, as well as institutional norms, all of which can give rise to discrimination even if individual decision-makers are not guilty of discrimination themselves. Small and Pager (2020, 52) define institutional discrimination as “differential treatment by race that is either perpetrated by organizations or codified into law.” Translating this definition into measurable metrics is challenging, not only because it is difficult to find appropriate data but also because there will always be ambiguity in distinguishing between institutional discrimination and the actions of individuals within those institutions. As a result, apart from some stark examples of discrimination codified into law, institutional discrimination has received relatively less attention within economics – the discipline which otherwise has yielded significant empirical insights into taste and statistical discrimination. Taking these aspects into account, in this chapter, I have attempted to provide a thematic overview of discrimination in credit. Through a selective review of the literature, I illustrate that caste, gender, and race are all persistent axes of discrimination in credit, and that discrimination has been shown to exist across diverse contexts and through different mechanisms. I begin below by reviewing the evidence for discrimination in section “Evidence.” Section “Implications” briefly explores the implications of discrimination in credit, and why even seemingly limited effects so far as access to credit goes can have disproportionate importance so far as its wider effects. Since the aim is to not only recognize the extent and forms of discrimination but to also gain insights into the mechanisms through which it manifests, we return to the question of underlying causal mechanisms in section “Mechanisms and Methods,” and conclude in section “Conclusion.”
Evidence Prejudice or taste discrimination in lending has been documented across different contexts. It can result in lower chances of loan approval, higher costs of credit, or smaller loans. However, taste discrimination likely coexists alongside statistical discrimination, and a priori neither can be ruled out. The majority of literature focuses on taste discrimination in credit outcomes, and infers this by adjusting for group-wise differences in creditworthiness, credit histories, and other indicators relevant to lending such as the ability to offer collateral. In other words, this
636
S. M. Kumar
approach attempts to adjust for potential sources of statistical discrimination. As we will see, a small subset of the literature focuses directly on statistical discrimination, investigating how beliefs about creditworthiness shape lending decisions and vary across time or circumstances. Racial discrimination in the USA has been widely studied. While discrimination is illegal in several countries in view of generally application legislation – e.g., in India, where the Constitution prohibits discrimination on the basis of caste, race, sex, and religion (Article 15), guarantees equality before the law (Article 14) and equality of opportunity (Article 16) – the USA is one of the few countries which has law that specifically seeks to abolish discrimination in lending through the Equal Credit Opportunity Act of 1974. However, a large body of literature continues to document racial discrimination in mortgage and business lending. Ross and Yinger (2002) provide comprehensive evidence on discrimination in mortgage approvals and cost, and Bayer et al. (2018) and Clarke and Rothenberg (2018) similarly find that African American borrowers face higher mortgage costs even after conditioning on relevant variables. In similar vein, Ross et al. (2008) and Hanson et al. (2016) find evidence that African American mortgage applicants face higher rates of nonresponse and receive less information than their White counterparts. Ross et al.’s (2008) study involved sending Black-White or Hispanic-White pairs of individuals to visit a lender, while Hanson et al. (2016) also send paired requests for information about loans but via email, where client’s names are selected to convey race. More generally, Ladd (1998) and Dymski (2006) offer reviews of the literature on racial discrimination in lending in the USA. As we discuss below, discrimination in this context involves more than just lending decisions as it extends to the actions of estate agents, but also and most egregiously reflects the legacy of “redlining” where areas overwhelmingly inhabited by non-whites were designated to carry higher credit risk. Loans for businesses are a second arena where discrimination can operate, particularly in the case of small businesses where the owner’s identity can play a direct role loan approval. In the USA, Blanchflower et al. (2003) find that Blackowned businesses are twice as likely to be denied credit even after adjusting for creditworthiness. Their findings are based on a combination of qualitative interviews with business owners and econometric analysis of two rounds of surveys with firms that collected detailed information about loan applications, credit histories, and subsequent approvals. Cavalluzzo and Wolken (2005) and Blanchard et al. (2008) find that non-white business owners face higher loan costs. Henderson et al. (2015) find that male as well as White business start-up owners have better access to lines of credit even after adjusting for various characteristics including credit scores, and Cavalluzzo et al. (2002) similarly find that female and non-white owners face higher loan denial rates and are less likely to apply for loans than their respective counterparts. Storey (2004) also finds evidence of racial discrimination in small business loan approvals in Trinidad and Tobago. There is likewise evidence of discrimination against women entrepreneurs across a range of contexts. Using census data on micro, small, and medium enterprises in India, Chaudhuri et al. (2020) find that women-led businesses face lower rates of loan approval from banks, a finding echoed more generally by Muravyev et al.
27
Discrimination in Credit
637
(2009) using data from 34 countries who also find that female entrepreneurs pay higher interest rates. Bellucci et al. (2010) find that women entrepreneurs in Italy face higher collateral requirements and lower credit availability compared to their male counterparts, and explain this difference at least partly in terms of taste discrimination. Bardasi et al. (2011) use data from three regions – Eastern Europe and Central Asia (ECA), Latin America (LA), and sub-Saharan Africa (SSA) – and find no evidence that women entrepreneurs face discrimination in accessing formal loans, but do find that women in the ECA are less likely to seek formal finance which could be explained by the higher costs for collateral faced by female-owned firms. De Andrés et al. (2020) find that female entrepreneurs in Spain are both less likely to seek credit and also less likely to have loan applications approved. Ongena and Popov (2016) go one step further to link discrimination with gender bias. They elicit gender bias in attitudes via a survey of US-based descendants of citizens from 17 European countries. Analyzing firm-level data from these countries, they then show that for countries with high gender bias, female entrepreneurs are less likely to complete loan applications and more likely to seek informal finance even though there is no evidence of actual discrimination in banks’ lending decisions. Beck et al. (2017) take a different approach, and examine the effects of borrower-loan officer gender combinations on loan outcomes for a large Albanian lender. Borrowers matched with an officer of the opposite gender pay higher interest rates and are less likely to apply for a second loan, but they also find that bias against the opposite gender decreases with officers’ experience and that competition reduces the scope for biased decision-making. Exceptions in this literature include Pham and Talavera (2018) who find that female entrepreneurs in Vietnam are more likely to get a formal loan and at lower cost than their male counterparts, and Hansen and Rand (2014) who find that female small firm entrepreneurs in 16 sub-Saharan countries are less likely to be credit constrained than their male counterparts with the reverse pattern for medium-sized enterprises. Corsi and De Angelis (2017) likewise find no evidence of discrimination against women in a Ugandan microfinance program. There is a large literature from India on caste-based discrimination and inequality in access to credit. This includes small business owners, farmers, and households. Examples of qualitative studies include Prakash (2010) who surveyed 90 SC-owned businesses across 6 states in India and reports the challenges faced in obtaining bank loans to set up a business to set up an enterprise, leading in many cases to borrowing from the informal sector at higher cost, and Jodhka (2010) who documents the difficulties faced by SC entrepreneurs in Northwest India in accessing bank credit. Kumar (2013) uses nationally representative survey data on farmers’ access to credit and argues that scheduled caste (SC) and scheduled tribe (ST) farmers likely face discrimination in bank lending, particularly when they live in areas where higher castes are economically dominant. This conclusion is based on patterns of outstanding bank loans, and therefore potentially conflates any differences in loan-seeking behavior with loan approvals by lenders. Kumar and Venkatachalam (2019) tackle this problem, by studying loan applications and approvals separately, and find that SCs and STs are less likely to apply for credit and that STs are less likely to have
638
S. M. Kumar
loans approved. They examine sensitivity to potentially unobserved credit histories through a Monte-Carlo simulation approach, and show that ST credit histories would have to be extremely poor to explain the latter finding as statistical discrimination. Kochar (1997) finds that farmers are (formal) credit constrained using data from Uttar Pradesh by accounting for a hierarchy of loan sources in terms of preference where informal sources are ranked lower than banks. Fisman et al. (2017) analyze data from a large Indian public sector bank and find that loans officers are more likely to approve, and borrowers are more likely to repay, when both parties belong to the same religion, or within Hindus the same caste group. They interpret this finding – somewhat problematically perhaps – not as discrimination but as preferential treatment based on cultural proximity and better information, and show how this phenomenon might give rise to statistical discrimination against minorities simply because loan officers of those groups are fewer in number. In contexts where informal sources of credit coexist with formal sources such as banks, studying discrimination can also be linked to the sources from which loans are sought and obtained. For example, Pal (2002) analyzes data from three South Indian villages and finds that higher-caste households are more likely to have loans from formal sources. Guérin et al. (2013) combine qualitative fieldwork with quantitative surveys in a study of rural Tamil Nadu households to document how the intersection of gender and caste determines both the sources of credit available to households as well as the terms on which loans are obtained – broadly speaking, to the disadvantage of Dalits (SCs). They examine debt more generally, as a social transaction and not simply loans borrowed, to explain how social hierarchies, class, trust, and patronage all combine to shape the nature and sources of debt. In a similar vein, Guérin (2014) argues that microcredit also operates via complex social relationships, and can both strengthen existing hierarchies as well as reduce dependence on unwanted sources of debt. Fafchamps (2000) studies borrowing by manufacturing firms in Kenya and Zimbabwe and finds no evidence of inferior access to bank loans among female or Black owners, but does find that network effects play a significant role in shaping access to credit from suppliers due to which female and Black owners have reduced access to trade credit, potentially also because of discrimination. Pope and Sydnor’s (2011) study of peer-to-peer lending via an online platform in the USA shows that discrimination in informal lending can also exist in economies where lending otherwise takes place overwhelming within formal structures. However, while they find that Blacks are less likely to receive funding, they also find that these rates are in fact higher than those predicted by default risk – in other words, it is possible that taste discrimination in favor of Blacks coexists with statistical discrimination against them.
Laws, Rules, and Institutions The majority of studies referred to above study credit transactions from the perspective of individuals. In most cases, these are the individuals who do or do not apply for credit. Any discrimination thus inferred is implicitly or otherwise attributed to the
27
Discrimination in Credit
639
institutions from which loans were sought, and this could result from institutional practices or the decisions of individual bank officers or a combination. A smaller number of studies (e.g., Beck et al. 2017; Fisman et al. 2017) jointly analyze data on the individuals seeking loans and the bank officers who decide on those loans. Fisman et al. (2020) present a somewhat related finding, and show that exposure to religion-based violence makes Hindu bank officers less willing to lend to Muslim borrowers, but do not relate this to institutional norms and rules or the lack thereof. While institutional norms might influence how the latter make decisions, the fact that these vary according to officers’ gender or caste suggests that emerging patterns at least in part reflect the officers’ individual preferences. The challenge is to delineate these twin factors of influence. To this end, this section provides a selective discussion to explicitly consider the role of institutions, rules, and laws in shaping discrimination, recognizing that these can play an important role in shaping the decisions of said individuals. One approach to distinguish between the extent to which the actions of individuals within institutions reflect their personal preferences or prevailing norms and rules is to examine loan outcomes before and after a change in rules, since individuals’ preferences are likely to remain unchanged in the short term. Cozarenco and Szafarz (2018) provide an example of this by studying the effects of a rule change in a French microfinance institution (MFI). The rule imposed a ceiling on microcredit loans, and it was brought about by the requirements imposed by cofinancing via a bank. This led to the MFI switching from lending biased in favor of females to lending biased against, which the authors argue is most likely because the bank’s own lending norms started to be reflected in the MFI’s lending practices. In the absence of a change or natural experiment, patterns of correlation can also provide insight into institutions. Kumar (2013) argues that discriminatory patterns of lending that are unfavorable to lower castes are likely driven by institutional capture of cooperative banks by members of higher castes. This cannot be directly observed in the data, but is supported by the correlation between caste dominance at district level and caste-wise differences in access to loans. This correlation is present for cooperative banks that have decentralized structures, but not for commercial banks that are managed centrally. This interpretation is also supported by Drèze et al.’s (1997) long-term study of Palanpur village in Uttar Pradesh, India, which demonstrates how institutional capture of banks might take place by caste groups that are locally dominant, besides more generally painting a detailed picture of the lending sources available in the village and how access to credit is shaped by politics and hierarchy. It is also usually difficult to identify the effects of the regulations or laws separately of institutional norms and individuals’ behavior. In India, while the Constitution outlaws discrimination on the basis of caste, religion, race, and sex, there is no specific application of this for credit. There are also long-standing regulatory norms set by the central bank, viz Reserve Bank of India, that encourage banks to lend to disadvantaged groups, one of which is scheduled castes and scheduled tribes (e.g., section 16 in Reserve Bank of India 2020). Priority sector lending is however fungible across categories of
640
S. M. Kumar
disadvantage, and moreover these norms are not binding, nor do they rule out discrimination or inequality in credit access of the sorts we have discussed. In contrast, the USA provides at least one stark historical example of how formal regulation shapes discrimination. Interventions aimed at stabilizing and supporting lending for housing and the housing market following the Great Depression directly contributed to discrimination against Blacks. As Small and Pager (2020) explain, purpose-created government agencies worked to make the property appraisal process more systematic. One of these, the Home Owners Loan Corporation, created maps that divided major cities into zones which played a significant role in determining the creditworthiness of a property. Neighborhood demographics contributed far more to the appraisal than the condition of the property itself (Woods 2012). Zones ranged from A–D, with “D” demarcated in red on maps – hence redlining – and socioeconomic status including race was directly used as a metric to demarcate zones. Specifically, the proportion of Blacks in a given neighborhood was directly and inversely proportional to the zone grade. This resulted in Blacks finding it very difficult to buy a property in a zone with a good grade – existing homeowners would resist this since Black families moving in would reduce that grade, and finding loans for properties located in predominantly Black areas was likewise difficult as well. Aaronson et al. (2021) use data from 1940–2010 and show that redline designation causally effected the racial composition and subsequent development of neighborhoods, property values, and rents, most likely by reducing access to credit and increasing associated borrowing costs. This led to a cyclical, long-run pattern of reduced investment in erstwhile redlined areas, the effects of which are visible even in 2010 even though redlining had long been abolished, and racial segregation has been declining following the Equal Credit Opportunity Act of 1974 and similar policy actions during 1960–1970. Finally, the algorithms used to create credit scores and to guide loan approval and pricing decisions are a fast-evolving example of the rules that govern lending processes. Unlike laws or institutional rules, however, algorithms are far less transparent in terms of how they use data to make decisions, and in the case of machine learning, the types of data on which the algorithms have been trained. Even if an algorithm purposely avoids, say, race as an input, it is difficult to rule out that certain other variables have not been used as a proxy. This problem can be compounded by the more general characteristic of machine learning algorithms: It is difficult to determine how different variables combine to produce output. Add to this the proprietary nature of algorithms, and it is easy to see that even laws that prohibit the use of group-belonging as a characteristic in making decisions might not be sufficient to curtail discrimination. In common with other applications of prediction through machine learning, there is therefore the risk of direct discrimination against minorities as well as indirect discrimination – where an algorithm relies on groupbelonging as a proxy for other relevant but missing indicators thereby giving rise to statistical discrimination. In either case, if the result is that certain groups are less likely to receive loans, this leads to vicious circle wherein data on credit histories for these groups becomes relatively sparse over time, perpetuating the problem of limited information and thereby statistical discrimination.
27
Discrimination in Credit
641
Cowgill and Tucker (2019) discuss these and related concerns about discrimination and algorithmic fairness, most of which apply to but are not unique to lending (see also Munoz et al. (2016) on the challenges of fairness in using big data). There is also a growing literature on algorithmic fairness specifically in the context of lending. Hurley and Adebayo (2016) and Gillis (2020) discuss the problems of fairness posed by credit-scoring algorithms based on big data, including the extent to which existing laws do or do not provide suitable protection against these problems, and how better transparency and accuracy can be implemented in aid of fairness and avoiding discrimination. Lee and Floridi (2021) offer a normative assessment of credit-scoring algorithms in terms of different definitions of fairness, focusing on the ethical trade-offs involved in prioritizing multiple objectives. Analyzing a dataset for start-ups in the USA, Robb and Robinson (2018) do not find evidence of racial bias in business credit scores, whereas Henderson et al. (2015) find the opposite: Blackowned business start-up receive lower than expected credit scores and receive less favorable treatment in accessing credit than do White-owned firms, with similar patterns of inequality against female-owned firms. Fuster et al. (2022) conduct an exercise to compare traditional logit with machine learning predictions of credit default using data on US mortgages, and find that the latter are better at modeling default risk but do not make progress in addressing missing data problems. This results in more granular predictions of default risk, leading to gains for White and Asian borrowers but losses for Blacks and Hispanics. Bartlett et al. (2019) provide comparative evidence of human versus algorithmic decision-making again by analyzing US mortgage data, and conclude that fintech algorithms also discriminate against minority borrowers but to a far lesser extent than do face-to-face lenders, and that too only in loan pricing, not loan approvals.
Implications The research discussed provides evidence of discriminatory gaps in loan approvals or the cost of credit between men and women, Whites and non-whites, and scheduled and higher castes. To appreciate the wider economic and social implications of these gaps, we must also recognize the role of lending in acquiring and growing capital more generally, particularly across generations. Depending on the context, small business entrepreneurship and home ownership are both crucial determinants of wealth accumulation. Discrimination in lending then has significant knock-on implications for inequality and disadvantage beyond loans alone, effects that can be compounded if there is discrimination in allied areas too. For instance, there are persistent caste- and gender-based disparities in small business entrepreneurship encompassing profitability, size, location, and sector (Deshpande and Sharma 2013, 2016). This combined with evidence on caste and gender discrimination in access to credit helps build a bigger picture where discrimination and inequality shape access to capital more generally. Indeed, entrepreneurship is often considered an attractive alternative to salaried employment in emerging economies, and particularly so in the context of India. The share of formal sector
642
S. M. Kumar
employment has remained low and members of the lower castes face stubborn barriers in entering these jobs thanks to a combination of cultural capital, discrimination, and network effects. With this view, the economic significance of small business ownership cannot be overstated. Yet, as the studies we have discussed above illustrate, the same disadvantages and discrimination which prevent members of the lower castes from breaking into well-paid formal sector employment are also in operation in the arena of entrepreneurship and employment (see also Deshpande (2011), Mosse (2018), and Das (2008)). Add to this the phenomenon of “selfcensorship” – wherein members of lower castes are less likely to seek bank loans (Kumar and Venkatachalam 2019), or self-selection toward informal sources of credit depending on gender and caste (Guérin et al. 2013) – and the resulting inequalities are cemented further, potentially across generations (e.g., Iyer et al. (2013)). Similar can be said of mortgage lending in the USA. Home ownership is a crucial determinant of wealth and racial gaps therein (Shapiro 2006). The lower likelihood of mortgage approval, lower wealth, and hesitancy to apply for a mortgage all combine with the result that Blacks are less likely to transition to home ownership (Charles and Hurst 2002). Home buyers and sellers also receive differential treatment from other market players depending on their race. For instance, Ondrich et al. (2003) show that the marketing efforts of estate agents as a function of house price vary by customers’ race, as does the portfolio of properties that they show to potential buyers. Set against a backdrop of ingrained spatial segregation by race (Hwang et al. 2014) – supported by erstwhile redlining – these sorts of patterns combined with discrimination in loan approvals contribute to a bigger picture where home ownership is deeply unequal across race, and inequality, discrimination, and homophilic preferences can all combine in cyclical ways to perpetuate this inequality. Race also played a role in the effects on the housing market in the wake of the 2008 crisis, with racial patterns of unemployment, mortgage pricing, and credit profiles all shaping racial differences in foreclosure rates to the disadvantage of minorities (Reid and Laderman 2009).
Mechanisms and Methods So far as borrowing from formal sources goes, it is clear that discrimination can arise from a combination of the actions of loan officers and the institutional environment in which they operate including norms, rules, algorithms, and laws. In reality, even though difficult to estimate empirically, these factors likely operate jointly to shape patterns of access. Given that research in this field aims to measure inequality of access and to pinpoint where discrimination is involved, it is necessary also to explain the mechanisms or causal processes through which discrimination operates. In this regard we offer three observations. First, as we have discussed, institutional norms and rules also demand examination alongside analysis of loan approval, size, and pricing decisions. Without this dimension, research into discrimination provides loan officer–level averages –
27
Discrimination in Credit
643
implicitly if using borrower-level data, explicitly using borrower-officer paired data – but it is unclear whether such behavior reflects their preferences solely as individual members of society, or whether institutional factors additionally shape their decision-making. There are relatively few examples of this sort of analysis in the literature, but when possible, they add significantly to our understanding of the mechanisms of discrimination. Second, the rising use of algorithms in credit decisions makes it easier to study discrimination. Inferring discriminatory decision-making on the part of a human being demands data on actual transactions coupled with a strong set of assumptions regarding the information they have taken into account to make that decision. Specifically, assumptions must be made about the decision-maker’s beliefs about creditworthiness, their personal past experience, stereotypes, or knowledge acquired through disparate means, all of which are difficult to observe and measure, but all of which might guide their decisions alongside any prejudice. Algorithms avoid this complexity by making known the universe of input variables. The problem of proprietary access aside, algorithmic decision-making can be studied via real data or simulated synthetic data to understand their decision-making process, as to which objectives or variables play a significant role and whether discrimination results. Compared to human decision-making, in theory at least algorithms are also easier to change, so that the policy implications of researching discrimination in algorithmic decision-making could be more tangible. Third, inferring discrimination in credit using observational data can draw on insights from research on discrimination in other arenas, and specifically with an eye toward examining beliefs. Data on beliefs are crucial for distinguishing between statistical and taste discrimination, and as outlined above, any analysis of decisionmaking must also make assumptions about the decision-maker’s beliefs regarding creditworthiness. This is difficult to do. While data on actual creditworthiness such as credit histories or repayment behavior are frequently available, these do not equate to data on beliefs, because the latter can be shaped by personal experience and stereotypes and by definition vary across individuals. The role of information and beliefs has been explicitly examined in other arenas of discrimination. This includes natural experiments such as “ban the box” legislation (Agan and Starr 2018; Doleac and Hansen 2020), laboratory experiments such as Castillo and Petrie (2010) who find that race-based statistical discrimination disappears with better information, and field experiments such as Bohren et al. (2019) who find that biased beliefs initially lead to gender discrimination which later dissipates with better information. Fewer papers in this area use observational data, but some examples include Laouénan and Rathelot (2020) on Airbnb rentals and Tjaden et al. (2018) who study online carpooling platforms, both focusing minority-ethnic discrimination, and Altonji and Pierret (2001) on how racial discrimination decreases as firms acquire more information about workers’ productivity. In forthcoming work Kumar and Venkatachalam (2021) suggest an approach to incorporate beliefs to estimate taste and statistical discrimination using observational data. The approach draws on causal mediation frameworks (see, e.g., Pearl (2014)) that specify the causal effect of treatment both directly as well as indirectly via
644
S. M. Kumar
specific mediators, viz. variables themselves causally effected by the treatment which then also causally effect the outcome. Beliefs are a mediator, since they are effected by group-belonging and in turn shape outcomes. Specifically, in a lending situation we might posit that group-belonging causally effects loan officers’ beliefs about creditworthiness and that these beliefs causally affect their decision-making, thereby potentially giving rise to statistical discrimination. In this framework, taste discrimination is the Natural Direct Effect of treatment on outcome which operates independent of all mediators (Pearl 2001), while statistical discrimination is a specific type of indirect effect which operates via beliefs (see Kumar and Venkatachalam (2021)). Implementing these ideas to analyze discrimination in lending would help distinguish more clearly between prejudice and statistical discrimination, thereby helping pinpoint how the provision of better information can reduce statistical discrimination.
Conclusion Discrimination in credit is an involved example of discrimination more generally. The specific characteristics of lending situations can give rise to taste, statistical, and institutional discrimination. Depending on the context, the resulting mechanisms can be placed into three broad categories: the beliefs and prejudice of individual loan officers potentially shaped by the institutional context; differences in the quality and existence of credit histories across borrowers depending on the group to which they belong; and algorithms that guide decision-making about loans. For informal lenders, the list of factors arguably broadens, incorporating social relations, trust, and hierarchy. These are supply-side factors. To fully appreciate the implications of discrimination we must also recognize patterns of inequality in access to capital – including but not limited to collateral – that mirror group-based disadvantage more broadly, as well as the likely existence of discrimination in other arenas of economy, society, and polity that both feed into as well as reflect unequal access to credit. These arenas include education, networks, and market transactions. Not least, discrimination or the perceived risk thereof on the supply side might manifest as the hesitation to seek credit on the demand side, further perpetuating unequal access to credit. Finally, it is clear that any appraisal of discrimination in credit covers a wide territory indeed, spanning fintech algorithms in richer economies to situated hierarchical power relations in rural contexts in emerging ones. Beyond evidencing the diverse causal mechanisms that shape discrimination, this range demonstrates that discrimination operates across societies and markets with starkly different characteristics, degrees of formality, and regulation. Markets are not self-correcting so far as weeding out discrimination in lending goes, and for emerging economies in the midst of structural and sectoral transitions, it is difficult to overstate the significance of links between inequality more generally and discrimination in credit.
27
Discrimination in Credit
645
References Aaronson D, Hartley D, Mazumder B (2021) “The Effects of the 1930s HOLC” Redlining “Maps.” American Economic Journal: Economic Policy 13(4):355–92 Agan A, Starr S (2018) Ban the box, criminal records, and racial discrimination: a field experiment. Q J Econ 133(1):191–235 Altonji JG, Pierret CR (2001) Employer learning and statistical discrimination. Q J Econ 116(1): 313–350 Arrow K (1973) The theory of discrimination. Discrim Labor Mark 3(10):3–33 Bardasi E, Sabarwal S, Terrell K (2011) How do female entrepreneurs perform? Evidence from three developing regions. Small Bus Econ 37(4):417 Bartlett R, Morse A, Stanton R, Wallace N (2019) Consumer-lending discrimination in the FinTech era. Technical report working paper 25943. National Bureau of Economic Research, Cambridge Bayer P, Ferreira F, Ross SL (2018) What drives racial and ethnic differences in high-cost mortgages? The role of high-risk lenders. Rev Financ Stud 31(1):175–205 Beck T, Behr P, Madestam A (2017) Sex and credit: is there a gender bias in lending? SSRN scholarly paper ID 2126489. Social Science Research Network, Rochester Becker G (1957) The economics of discrimination. University of Chicago Press, London Bellucci A, Borisov A, Zazzaro A (2010) Does gender matter in bank–firm relationships? Evidence from small business lending. J Bank Financ 34(12):2968–2984 Blanchard L, Zhao B, Yinger J (2008) Do lenders discriminate against minority and woman entrepreneurs? J Urban Econ 63(2):467–497 Blanchflower DG, Levine PB, Zimmerman DJ (2003) Discrimination in the small-business credit market. Rev Econ Stat 85(4):930–943 Bohren JA, Imas A, Rosenberg M (2019) The dynamics of discrimination: theory and evidence. Am Econ Rev 109(10):3395–3436 Castillo M, Petrie R (2010) Discrimination in the lab: does information trump appearance? Games Econ Behav 68(1):50–59 Cavalluzzo K, Wolken J (2005) Small business loan turndowns, personal wealth, and discrimination. J Bus 78(6):2153–2178 Cavalluzzo KS, Cavalluzzo LC, Wolken JD (2002) Competition, small business financing, and discrimination: evidence from a new survey. J Bus 75(4):641–679 Charles KK, Hurst E (2002) The transition to home ownership and the black-white wealth gap. Rev Econ Stat 84(2):281–297 Chaudhuri K, Sasidharan S, Raj RSN (2020) Gender, small firm ownership, and credit access: some insights from India. Small Bus Econ 54(4):1165–1181 Clarke KA, Rothenberg LS (2018) Mortgage pricing and race: evidence from the northeast. Am Law Econ Rev 20(1):138–167 Corsi M, De Angelis M (2017) Gender discrimination in microfinance? Some evidence from Uganda. J Dev Stud 53(5):723–740 Cowgill B, Tucker CE (2019) Economics, fairness and algorithmic bias. https://doi.org/10.2139/ ssrn.3361280 Cozarenco A, Szafarz A (2018) Gender biases in bank lending: lessons from microcredit in France. J Bus Ethics 147(3):631–650 Das MB (2008) Minority status and labor market outcomes: does India have minority enclaves? Policy research working papers. The World Bank de Andrés P, Gimeno R, Mateos de Cabo R (2020) The gender gap in bank credit access. J Corp Finan 101782:101782 Deshpande A (2011) The grammar of caste: economic discrimination in contemporary India. Oxford University Press, New Delhi Deshpande A, Sharma S (2013) Entrepreneurship or survival? Caste and gender of small business in India. Econ Polit Wkly 48:38–49
646
S. M. Kumar
Deshpande A, Sharma S (2016) Disadvantage and discrimination in self-employment: caste gaps in earnings in Indian small businesses. Small Bus Econ 46(2):325–346 Doleac JL, Hansen B (2020) The unintended consequences of “ban the box”: statistical discrimination and employment outcomes when criminal histories are hidden. J Labor Econ 38(2): 321–374 Drèze J, Lanjouw P, Sharma N (1997) Credit in rural India: a case study. Development Economics Research Programme 6. London School of Economics, London Dymski GA (2006) Discrimination in the credit and housing markets: findings and challenges. In: Rodgers III WM (ed) Handbook on the economics of discrimination. Edward Elgar Publishing, Cheltenham, UK, pp 215–259 Fafchamps M (2000) Ethnicity and credit in African manufacturing. J Dev Econ 61(1):205–235 Fisman R, Paravisini D, Vig V (2017) Cultural proximity and loan outcomes. Am Econ Rev 107(2): 457–492 Fisman R, Sarkar A, Skrastins J, Vig V (2020) Experience of communal conflicts and intergroup lending. J Polit Econ 128(9):3346–3375 Fuster A, Goldsmith-Pinkham P, Ramadorai T, Walther A (2022) Predictably Unequal? The Effects of Machine Learning on Credit Markets. The Journal of Finance 77(1):5–47 Gillis TB (2020) False dreams of algorithmic fairness: the case of credit pricing. SSRN Scholarly Paper ID 3571266, Social Science Research Network, Rochester Guérin I (2014) Juggling with debt, social ties, and values: the everyday use of microcredit in rural South India. Curr Anthropol 55(S9):S40–S50 Guérin I, d’Espallier B, Venkatasubramanian G (2013) Debt in rural South India: fragmentation, social regulation and discrimination. J Dev Stud 49(9):1155–1171 Hansen H, Rand J (2014) The myth of female credit discrimination in African manufacturing. J Dev Stud 50(1):81–96 Hanson A, Hawley Z, Martin H, Liu B (2016) Discrimination in mortgage lending: evidence from a correspondence experiment. J Urban Econ 92:48–65 Henderson L, Herring C, Horton HD, Thomas M (2015) Credit where credit is due?: race, gender, and discrimination in the credit scores of business startups. Rev Black Polit Econ 42(4):459–479 Hurley M, Adebayo J (2016) Credit scoring in the era of big data. Yale J Law Tech 18:148 Hwang J, Hankinson M, Brown KS (2014) Racial and spatial targeting: segregation and subprime lending within and across metropolitan areas. Soc Forces 93:1081–1108 Iyer L, Khanna T, Varshney A (2013) Caste and entrepreneurship in India. Econ Polit Wkly 48: 52–60 Jodhka SS (2010) Dalits in business: self-employed scheduled castes in North-West India. Econ Polit Wkly 45:41–48 Kochar A (1997) An empirical investigation of rationing constraints in rural credit markets in India. J Dev Econ 53(2):339–372 Kumar SM (2013) Does access to formal agricultural credit depend on caste? World Dev 43: 315–328 Kumar SM, Venkatachalam R (2019) Caste and credit: a woeful tale? J Dev Stud 55(8):1816–1833 Kumar SM, Venkatachalam R (2021). Causal inference for discrimination: an analysis of NYPD stop-and-frisk data. Unpublished Working Paper Ladd HF (1998) Evidence on discrimination in mortgage lending. J Econ Perspect 12(2):41–62 Laouénan M, Rathelot R (2020) Can information reduce ethnic discrimination? Evidence from Airbnb. Am Econ J Appl Econ 14(1):107–132 Lee MSA, Floridi L (2021) Algorithmic fairness in mortgage lending: from absolute conditions to relational trade-offs. Mind Mach 31(1):165–191 Mosse D (2018) Caste and development: contemporary perspectives on a structure of discrimination and advantage. World Dev 110:422–436 Munoz C, Smith M, Patil D (2016) Big data: a report on algorithmic systems, opportunity, and civil rights. Executive Office of the President. The White House
27
Discrimination in Credit
647
Muravyev A, Talavera O, Schafer D (2009) Entrepreneurs’ gender and financial constraints: evidence from international data. J Comp Econ 37(2):270–286 Ondrich J, Ross S, Yinger J (2003) Now you see it, now you don’t: why do real estate agents withhold available houses from black customers? Rev Econ Stat 85(4):854–873 Ongena S, Popov A (2016) Gender bias and credit access. J Money Credit Bank 48(8):1691–1724 Pager D, Shepherd H (2008) The sociology of discrimination: racial discrimination in employment, housing, credit, and consumer markets. Annu Rev Sociol 34:181 Pal S (2002) Household sectoral choice and effective demand for rural credit in India. Appl Econ 34(14):1743–1755 Pearl J (2001) Direct and indirect effects. In: Proceedings of the seventeenth conference on uncertainty in artificial intelligence. Morgan Kaufmann Publishers Inc., San Francisco, pp 411–420 Pearl J (2014) Interpretation and identification of causal mediation. Psychol Methods 19(4): 459–481 Pham T, Talavera O (2018) Discrimination, social capital, and financial constraints: the case of Viet Nam. World Dev 102:228–242 Phelps ES (1972) The statistical theory of racism and sexism. Am Econ Rev 62(4):659–661 Pope DG, Sydnor JR (2011) What’s in a picture? Evidence of discrimination from Prosper. com. J Hum Resour 46(1):53–92 Prakash A (2010) Dalit entrepreneurs in middle India. In: Harriss-White B, Heyer J (eds) The comparative political economy of development Africa and South Asia. Routledge, London, p 26 Reid C, Laderman E (2009) The untold costs of subprime lending: examining the links among higher-priced lending, foreclosures, and race in California. Federal Reserve Bank of San Francisco, San Francisco Reserve Bank of India (2020) Master direction – priority sector lending (PSL) – targets and classification. https://www.rbi.org.in/Scripts/BS_ViewMasDirections.aspx?id¼11959#Weaker_Sections Robb A, Robinson DT (2018) Testing for racial bias in business credit scores. Small Bus Econ 50(3):429–443 Ross SL, Yinger J (2002) The color of credit: mortgage discrimination, research methodology, and fair-lending enforcement. MIT Press, Cambridge/London Ross SL, Turner MA, Godfrey E, Smith RR (2008) Mortgage lending in Chicago and Los Angeles: a paired testing study of the pre-application process. J Urban Econ 63(3):902–919 Shapiro TM (2006) Race, homeownership and wealth poverty, justice, and community lawyering: interdisciplinary and clinical perspectives. Wash Univ J Law Policy 20:53–74 Small ML, Pager D (2020) Sociological perspectives on racial discrimination. J Econ Perspect 34(2):49–67 Storey DJ (2004) Racial and gender discrimination in the micro firms credit market?: evidence from Trinidad and Tobago. Small Bus Econ 23(5):401–422 Tjaden JD, Schwemmer C, Khadjavi M (2018) Ride with me—ethnic discrimination, social markets, and the sharing economy. Eur Sociol Rev 34(4):418–432 Woods LL (2012) The Federal Home Loan Bank Board, redlining, and the national proliferation of racial lending discrimination, 1921–1950. J Urban Hist 38(6):1036–1059
Gender-Based Discrimination in Health: Evidence from Cross-Country
28
Aparajita Dasgupta
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Early-Life Discrimination Within Households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Excess Female Mortality and Missing Women . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gendered Differences in Utilization of Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discrimination by Health Providers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measurement Issues Around Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Micro-foundations of Gendered Institutions and Economic Development . . . . . . . . . . . . . . . . . . . . Economic Cost of Gender-wise Discrimination in Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Policy Insights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
649 650 650 653 654 655 656 659 660 661
Abstract
This review presents an analytical framework to discuss the existence and forms of gender discrimination in health. We survey the evidence of gender discrimination in health at the various life stages of an individual. We discuss the underlying drivers of this discrimination and its relationship with economic growth and development. Lastly, the review outlines the welfare implications of gender discrimination in health and offers policy insights. Keywords
Gender discrimination · Health outcomes · Health utilization · Demand and supply of health · Cross-country evidence
A. Dasgupta (*) Ashoka University, Sonepat, Rai, Haryana, India e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_12
649
650
A. Dasgupta
Introduction Gender-based discrimination has an important bearing on health and health care, which varies across levels of development in countries. On a micro level, gender discrimination within households can start as early as during in-utero period affecting sex ratio at birth and child mortality outcomes. Discrimination within a household can take the form of systematic underinvestments in girls such as manifested by lower levels of breastfeeding and nutrition, immunization, and health visits during the first few years of life. Lack of such investments in early years of life can have far-reaching consequences that affect cognitive, schooling, adult health, and labor market outcomes. Additionally, early-life discrimination by gender can shape the preferences and expectations around health and can manifest in terms of not just reduced human capital as adults but lower health utilization and health access across the life stages. In addition to affecting demand-side preferences, gender biased social norms can induce and magnify supply-side provider-level discrimination in health care. In this review we cover the evidence on both these fronts. We include an analytical framework to discuss the existence and forms of health discrimination and its underlying drivers and its implications on long-run development. Next, we discuss the welfare consequences of discrimination and outline its implications for long-run economic growth and development. We also explore the evidence base on the converse question, which answers how does gender gap in health outcomes respond to changes in income and exposure to shocks. We use the following road map: in the next section we survey the literature on how gender-based discrimination manifests within households, as measured by health and nutritional investments, access, and utilization of health services, at various life stages. In section “Discrimination by Health Providers,” we discuss health discrimination at the level of health providers. In section “Measurement Issues Around Discrimination,” we outline an analytical framework to measure discrimination in health. In section “Micro-foundations of Gendered Institutions and Economic Development,” we discuss the underlying drivers of discrimination and how these relate to institutions of gender and the level of development in the economy. Section “Economic Cost of Gender-wise Discrimination in Health” outlines the economic costs and welfare implications of gender-based discrimination in health. Section “Policy insights” concludes the discussion by offering policy insights and outlining areas for new research.
Early-Life Discrimination Within Households Excess Female Mortality and Missing Women Gender discrimination in early life can manifest in various forms starting from selective abortion of female fetus to systematic underinvestments in female children in the form of inadequate nutrition and lower health care. This leads to excess female mortality (Visaria 2008) and also affects their long-term human capital formation.
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
651
This is in spite of the fact that if given similar level of care, women have better survival chances at all ages, including in utero. Studies have shown that it is boys who are more vulnerable to illness and disease in early life, with infant mortality for boys being higher than girls due to biological and genetic factors (Pongou 2013). The fact that men outnumber women in certain parts of the world, such as in Asian and North African countries, speaks to systematic neglect and care(Sen 1992). This phenomenon of “missing women” has received wide attention since it was first documented by Sen (1992) which estimated that over a 100 million women were missing worldwide, as a result of systematic neglect and unequal access to resources. With an equal amount of care to both genders, developed countries such as America and European nations have lower female mortality rates which allow the sex ratio (biased toward women) to rise even further with age. This is in contrast to the developing world of Asia and north Africa where low life expectancy and higher fertility translate into dwindling sex ratios. Sahn and Younger (2009) find about half of the total health inequality at the country level is within households, more than total country inequality between households. Studying “missing women” by age, Anderson and Ray (2010) argue that most of the missing women belong to adult age in India and Sub-Saharan Africa as compared to the prenatal stage as in China, at the start of the twentieth century. Klasen (1994) reexamines trends in excess female mortality and finds a rise in the absolute number of missing women, but a decline in the share of the number of women alive. Klasen and Wink (2003) note that while China continued to fare poorly and contributed considerably to the absolute number of missing women owing to strict family planning policies, the state interventions in Sri Lanka and pre-reform China helped in reducing gender discrimination. Notably, Vlassoff (2007) find that gender is closely related to the determinants and consequences of poor health in both developed and developing countries. The gap in life expectancy of females versus males across countries is narrowing in most European countries as noted in Barford et al. (2006). In WHO’s world health report of 2002, only six countries (Nepal, Botswana, Zimbabwe, Lesotho, Bangladesh, and Swaziland) saw higher male life expectancy, which a year later reduced to only two countries- Qatar and Maldives. Although the female disadvantage in mortality had reduced in the 1990s, a new disadvantage in the form of sex selective abortions came up at that time (Sen 2003; Visaria 2008). Studies from India, China, and Bangladesh document discrimination is manifested against females in terms of prenatal care when the sex is known in advance (Bharadwaj and Lakdawala 2013; Anand and Chen 2018). Importantly, sex selective prenatal care in tetanus explains a large proportion of excess female neonatal mortality in India. Further, Baird et al.’s (2011) cross-country analysis suggests gender-wise heterogeneity in infant mortality amidst negative economic shocks – notably, female infant mortality was more sensitive to negative economic shocks than male infant mortality. Female disadvantage in breastfeeding, food allocation, and vaccination has been widely documented across countries. A large body of evidence, including both theoretical and empirical studies, documents selective neglect of children by gender and sex composition of sibling, with disproportional costs for girls with older sisters
652
A. Dasgupta
(Behrman 1988; Pande 2003; Haddad et al. 1996). Girls in India are breastfed for a shorter duration, and have lower consumption of fresh (by 14%) and breast milk (by 21%) than boys (Fledderjohann et al. 2014). The shorter duration of breastfeeding for girls is associated with higher under 5 mortality. This accounts for 50% of the survival disadvantage faced by girls in India, as compared to other low-income countries. Using data from the Pakistan Integrated Household Survey, Hazarika (2000) finds while boys have higher immunization rates and more access to medical professionals girls seem to be no less nourished than boys. Jayachandran and Kuziemko (2011) find that girls are breastfed lesser than boys, and children with older brothers are breastfed more in India. Strikingly, parent’s higher valuation of sons’ health explains just one-third of this gender gap, and the remaining is attributed to parents’ valuation of future sons that guides their breastfeeding behavior. They find the gender gap in breastfeeding drives 14% of excess female child mortality in India, which is about 22,000 “missing girls” each year. A similar finding is reported in Ghana, where Garg and Morduch (1998) find that children who had all and only sisters had better health outcomes than children who had all and only brothers. Even in southern part of India, which is known for a weaker son preference compared to its northern counterpart, gender discrimination was equally prevalent Mishra et al. (2004). Though the findings vary by the child’s birth order, female children received shorter periods of breastfeeding, lower rates of immunization, and lesser attention for treatment seeking for respiratory infections and infections. They also find excess female childhood mortality to be greater in families with older female sibling. Kishore and Spears (2014) use the subsequent wave of the NFHS of 2005–2006 and study a relatively passive/implicit form of discrimination in households. They show that urban households with a male first child were more likely to use cleaner cooking fuel, as opposed to a households with a female first child. A similar result is found by Rosenblum (2013) who finds son-preferring fertilitystopping rules result in first-born boys having fewer siblings than a first-born daughter, and this results in discrimination and excess child mortality among females. Interestingly, as a result, boys are better off with sisters and girls are better off with brothers. Importantly, one needs to take into account implicit form discrimination along with explicit discrimination. Son-biased fertility stopping rules can result in an implicit discrimination where girls end up with more siblings. Thus, even without the existence of an explicit discrimination between boys and girls, the fertility stopping behavior, with girls having more siblings than boys, leads to fewer resources and worse health outcomes for girls. Kashyap and Behrman (2020) find that explicit discrimination against girls weakened post mid-1990s, and showed staggering improvements after the mid-2000s in northern India. Ability to realize son preference at lower parities due to sex-selective abortions is likely to be one of the factors responsible for weakened postnatal discrimination after the mid-1990s.
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
653
Gendered Differences in Utilization of Health Differences in health utilization among adult men and women can stem from differences in access, expectation, affordability of individuals, and insurance coverage toward healthcare and quality of healthcare, among other factors. Importantly, inequality in health can arise from demand side constraint, i.e., individual’s healthseeking behavior as well as from supply side restrictions on healthcare provision. A growing number of studies from developing countries show that household’s monetary and time resources are skewed in favor of male members (Dupas and Jain 2021; Shaikh et al. 2018; Ahmed et al. 2003; Saikia and Bora 2016; Kapoor et al. 2019; Batra et al. 2014; Adamson et al. 2003). The evidence seems to suggest that under sex neutral conditions, across a majority of diseases, women’s share of hospitalizations and bed-days is disproportionately low. Additionally, there exists gap in diagnosis and follow-up systematic to gender, such as found in the case of detection of severe anemia in women in India (George et al. 2005). Using the Oaxaca-Blinder decomposition method, Saikia and Bora (2016) find that healthcare expenditure for females is lesser than that of males because of the belief that male health is more important. Saikia et al. (2019) corroborate the finding that average healthcare expenditure to be lower for adult women, irrespective of disease type and duration of stay in the hospital in India. Moreover, gender discrimination can worsen during health shocks when healthcare is financed by borrowing and sale of assets (Kumar et al. 2020). While women’s share of health expenditure has been found disproportionately low, the evidence on the treatment-seeking behavior by gender is mixed. Peltzer et al. (2014) find that compared to men, women are less likely to use inpatient services and more likely to use outpatient services, using data from China, Ghana, India, Mexico, the Russian Federation, and South Africa. Similarly, a study in Kenya (Mwabu et al. 1993) find men to be less constrained in seeking healthcare, both in terms of distance and medical fees compared to women. Women are also more likely to use primary health care, have a lower frequency of hospitalization, as found by a study of the elderly population in Serbia (Gajovic et al. 2019). Now, differences in prevalence of infectious diseases by gender can stem from differences in the prevalence of the risk factors, such as higher smoking behavior, among men (Marcoa 2018). While earlier, the higher incidence of the contagious disease tuberculosis (TB) among men was inferred as a result of men’s higher participation in social and public interactions, studies from Vietnam and Bangladesh revealed that this might have been a result of underdiagnosis, underreporting, and inequity in access to TB services among women (Govender and Penn-Kekana 2007). While men in India are more likely to have tuberculosis than women, they are less likely to be diagnosed at the same time (Rao 2009). Using data from standardized patient visits in India, Daniels et al. (2019) find no significant gender bias in the quality of care. Burger (2019) points out that this result may be due to factors, such as behavioral differences (if men are more critical, it is likely to have an impact on their health-seeking behavior), preferences, service expectation,
654
A. Dasgupta
self-assessed value of their time, and their willingness to acknowledge and express negative experiences, and thus requires further investigation. Alderman (1997) studies the effect of income on the demand for medical care for children in Pakistan, where healthcare expenditure on girls is more elastic. Interestingly, there is heterogeneity in gender bias by the age group. In the setting of rural China, Gao and Yao (2006) find that intra-household allocation of healthcare favors women in their childbearing and maternal periods during which a women’s healthcare is considered crucial for the quality of health of the next generation. In general, health expenditures for women are more elastic to income, while men’s expenditures depend on factors such as treatment time. We have seen how gender gaps at the structural level can induce gaps in health. These are only exacerbated amidst health shocks. For example, the gender gap in Covid-19 vaccination coverage is widening in India. Even after taking into consideration the fact that India has around 6% more men than women, the gap between vaccinated men and women has grown from 2% on April 10 to 15% as of June 15 (Deccan Herald 2021). The National Family Health Survey of 2019–2020 shows that in rural India, only 30% of the women use the internet as opposed to 50% of the men. The disadvantage in terms of access to technology and limited mobility is likely to be translated to deficits in health usage, such as what we witness in the vaccination gap.
Discrimination by Health Providers Statistical discrimination in health has been widely documented in the literature in the context of developed countries, by patient characteristics Balsa et al. (2005), Van Ryn and Fu (2003), and Angerer et al. (2019). Statistical discrimination at the level of health providers can broadly follow three mechanisms – doctors being less willing to interact with certain groups, giving differential interpretations of symptoms by these groups, and prior beliefs and stereotypes doctors may hold against certain groups (Balsa and McGuire 2003). Thus, even in the absence of taste-based discriminatory behavior, the quality of healthcare may get affected, which in turn lowers the demand for health care from among the discriminated group of individuals. Across developed and developing countries, it has been found how individuals from lower class are often subjected to discrimination when it comes to waiting time in hospitals (Johar et al. 2013; Shaikh 2018; Sabarwal et al. 2010). Arguably, this kind of disadvantage could be higher for women, where identities interact to shape choices and expectations of quality of health services. Shim (2010) proposes the concept of cultural health capital as that which involves cultural skills, verbal and nonverbal competencies, attitudes and behaviors, and interaction styles, from the point of view of both patient and healthcare provider. At the intersection of gender and caste-based discrimination, Sabarwal et al. (2010), through interviews of 500 Dalit women across 17 districts in Andhra Pradesh, Bihar, Tamil Nadu, and Uttar Pradesh, note that women had to wait longer for delivery under the orders of
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
655
the medical staff. On similar note, Acharya (2010) studies discrimination experienced by Dalit children and find that a majority of Dalit children faced discrimination in receiving medication, facing untouchability, and allocation of inadequate interaction time from doctors, anganwadi workers as well as lab technicians. Now, coverage of health insurance schemes at the household level may not address the gender gap in healthcare access and can in theory raise the intrahousehold gender-wise inequality at the beginning, where men are more likely to avail the benefits of health insurance programs. Examining the impact of the Rashtriya Swasthya Bima Yojana (RSBY) in India, Cerceau (2012) finds men formed a higher proportion of the enrollment than women. Furthermore, even among the women who were enrolled, there was little awareness about the details of the scheme, like that of the medical services/facilities it covered. Moreover, women, especially those with lower levels of literacy, also reported having experienced unresponsive and rude behaviors from the healthcare providers and the hospital staff. The dearth of data on the quality of healthcare and provider level behavior, in the context of developing countries, poses one of the major challenges to understand gender discrimination on these lines. There is a small but growing body of literature using Vignettes based survey instruments, randomized control trials, audits, and lab-based field experiments to study provider level discrimination in the context of low-income countries (Das and Hammer 2005, 2007) with mixed evidence on provider level taste-based discrimination (Singh and Mitra 2013; Dasgupta et al. 2016).
Measurement Issues Around Discrimination How much of the gender gaps in morbidity outcomes (that are often self-reported) can be ascribed to gender-wise discrimination? For the same level of latent health, gender gaps in health can emerge from biological differences in disease-specific vulnerabilities or from differences in reporting behavior apart from the discrimination faced by gender. Additionally, the relative gender gap in morbidity outcomes is a function of mortality selection that can also vary by demographic subgroups. For example, in the presence of sex selective abortion technology, there would be a gross underestimation of gender discrimination if one were to just look at postnatal differences in health outcomes, as much of the discrimination might be manifested at the prenatal stage. The issue is even more complicated as gender differences in health outcomes can evolve in response to sex ratios. If the biological theory of sexual dimorphism holds, implying that males are more likely to be taller in places where sex ratios are biased against women, this creates further grounds for discrimination against females (Deaton 2008). Apart from mortality bias, differences in reporting of self-assessed morbidity can be yet another factor that influences the measurement of health disparity. Perceptions of one’s health as well as healthcare seeking behaviors vary considerably by demographic subgroups. Sen (1993) uses the term “positional objectivity” to help
656
A. Dasgupta
illustrate and reconcile an apparent puzzle regarding measurement of health in India: a state like Kerala that has highest life expectancy at birth has consistently had very high morbidity rates. On the other hand, states like Uttar Pradesh and Bihar that are doing poorly in terms of life expectancy and health transition are the states with lowest morbidity rates. On similar note, Banerjee et al. (2004) document a discord between the quality of public health services and self-reported health in rural Rajasthan in India. It is likely that individuals use different thresholds to evaluate their own health, which is a function of their access to and expectations from the health service and the quality of health in their peer group. Hence, it is important to take note of systematic measurement issues around subjective health outcomes when measuring gaps in morbidity outcomes by population subgroups, such as by gender. While self-reported health is highly correlated with mortality indicator and is the most frequently used indicator for health research, it is a function of expectations, health access, and peer group’s preferences among other factors that are likely to vary by gender. Anchoring vignettes has been an increasing popular tool (more so in the developed country context) to account and correct for the systematic patterns of reporting heterogeneity and improving comparability by demographic subgroups (Grol-Prokopczyk et al. 2011; Bago d’Uva et al. 2008). The gendered pattern of reporting behavior in self-reported health is somewhat mixed, and accounting for reporting differences has varying impacts on gender-wise health disparity in the study setting (Sadana et al. 2002; Molina 2016). Grol-Prokopczyk et al. (2011) show that in the Wisconsin Longitudinal Study, the female advantage in self-reported health disappeared after accounting for reporting heterogeneity. Bago d’Uva et al. (2008) find while there is a clear male advantage in the domains of cognition and pain, purging reporting differences reduces the disparity in India and China and raises it in Indonesia. In the context of Europe, using the Survey of Health, Ageing and Retirement in Europe (SHARE), Oksuzyan et al. (2019) challenge the gender stereotypes in the context of reporting health problems. Overall, after controlling for the reporting heterogeneity, the proportion of women with poor health was higher than that of men across all ages, and, especially at older ages. Dasgupta (2018) using data from 2007 to 2009 World Health Survey of India find women were more likely to report poor health after accounting for innate differences in objective health. The study validates the response consistency assumption in the use of the vignettes technique in the context of developing countries and supports its use to allow better comparability in self-assessed health outcomes.
Micro-foundations of Gendered Institutions and Economic Development Culture and patriarchal norms are well-documented reasons for the neglect of the girl child. Forms of agriculture, kinship norms, and inheritance laws have a deep-rooted influence on setting gendered institutions. Sen (1992) documents the vast difference in sex ratios across the world, addressing why poor countries like India have consistently seen low sex ratios and high rates of female infant mortality. The
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
657
scope for old age security policies to address gender bias has been highlighted in several studies (Chung and Das Gupta 2007; Jayachandran 2015; Ebenstein 2014; Ebenstein and Leung 2010). Jayachandran (2015) recognizes that gender disparities in poor countries are not simply a result of a lower GDP. Rather, the cultural context of these countries renders a better understanding of these disparities. Social practices such as patrilocality, patrilineality systematically skew education, health, and other human capital investments away from females. Restrictions on female labor force participation and autonomy further impede women to escape these structural/institutional barriers. In India, gender gaps have persisted despite rapid economic growth. Can economic development reduce the gender gap in health outcomes? While economic development is associated with lowering of fertility, in the presence of stark son bias, this problem of skewed sex ratio becomes even worse Jayachandran and Pande (2017). The role of son-biased fertility stopping rules and favoritism toward eldest sons has been found to play a major role in driving India’s height deficit (with fifth highest stunting rate among 81 low-income and low-middleincome countries) relative to poorer regions like that of African countries. Strikingly, this is in spite of the fact that India has performed better in most health and development indicators such as life expectancy, food security, poverty incidence, maternal mortality, and educational attainment. Notably, the birth order gradient is steeper in richer families compared to poorer ones. Their study puts forth a stubborn social challenge in overcoming gender inequity which is aided little by rising incomes. While India performed remarkably in declining fertility in the years following independence, Das Gupta and Mari Bhat (1997) note that even during the period of 1981–1991, which was characterized by economic growth, high sex ratios persisted as well as increased. The decline in fertility in India has come at the cost of 5% of excess mortality of girls compared to that of boys owing to sex-selective abortion. Even in South Korea, which performed better than India on the economic front, witnessed an increasing sex ratio at birth until as late as the mid-1990s due to sex-selective technologies (Chung and Das Gupta (2007)). Although income and health are correlated, Deaton 2002 argues against targeting health or wealth in isolation. The study finds the relationship between increasing income and health is a gradient and not a threshold, in that health improves at all levels of income. Moreover, he emphasizes that health and wealth are mutually determined, suggesting the need for policy to be directed toward income redistribution rather than narrowly targeting health inequities. This is in contrast with the policy suggestion of Witter et al. (2017) which recommends targeting of health policies, specifically at women and further marginalized groups, such as lower caste women, since gender-neutral policies disproportionately favor men. Evidence from Gambin et al. (2005) suggests a strong link between health and wealth, with larger premium of good health for men and in contrast larger penalty for poor health for women, using data from European Community Household Panel. While the impact of self-assessed health on wages was greater for men than women, the impact of chronic illness on the other hand is more significant for women.
658
A. Dasgupta
Parental and especially mothers’ inputs have been shown to have considerable impact on children’s cognitive development, especially for boys compared to girls (Bertrand and Pan (2013)). The evidence on the effect of maternal employment on child’s health outcomes is mixed that varies by the level of development, including both positive and negative effects (Dehejia and Lleras-Muney 2004; Vikram et al. 2018; Jensen 2010). The gendered impact of economic shock on infant mortality can vary by the response in women’s labor supply and selective fertility during shocks, which can again affect mortality gap by gender. Depending on the relative dominance of the income and substitution effect, economic shocks can have both a positive and a negative impact on human capital (Ferreira and Schady (2009)). The effect of improvement in technology on health outcomes and gender bias therein is mixed. In the context of the USA, Albanesi and Olivetti (2016) find that medical progress has led to an improvement in maternal health which along with an increase in fertility has allowed more married women to enter the labor force. Glied and Lleras-Muney (2008), who studies the role of education in accruing the benefits of healthcare innovations in the USA, showed that education advantage in decreased mortality was smaller for women. In fact, health disparities can go up with technological innovation where educational gradients for health are likely to increase with an increase in health-related innovation (Deaton (2002)). In the context of a developing country setting, with stark son preference norms in the society, access to ultrasound technology can lead to increase prenatal discrimination and decrease in postnatal discrimination by gender. Anukriti et al. (2020) study the impact of the introduction of pre-natal sex detection technologies such as the ultrasound on female under 5 mortality rates and find reduced gender gaps in outcomes such as mortality, health investments for children of second order or higher births orders, in firstborn girl families as opposed to firstborn boy families. Fertility declined post the ultrasound, and more so in firstborn girl families. Conversely, Dasgupta and Sharma (2021) find that with ban on ultrasound technology for sex selective abortion in India, the relative mortality for girls go up, more so for firstborn girl families, along with an increase in fertility outcomes. There is mixed evidence on the income elasticity of health outcomes by gender ranging from beneficial effects of positive income shocks on women’s health in Indonesia (Maccini and Yang 2009) to disproportional negative effects of early rainfall shock affecting men’s health in South Africa (Dinkelman 2017). However, the benefits of agricultural innovation, such as the adoption of High Yielding Varieties (HYV) crops, have been higher for boys with higher reduction in infant mortality across several developing countries (Goltz et al. 2020). In India, this impact of HYV adoption on infant mortality for girls was only about half of that for boys (Bharadwaj et al. 2020). However, there also exists biological factors that suggest higher survival chances of female fetuses when mothers are faced with a nutritional shock. Using data from fertility survey Song (2012) finds that the deterioration of nutrition levels of mothers in China after the Great Leap Forward Famine of 1959–1961 led to a decline in sex ratios. They find that the famine induced a variation in mothers’ nutrition levels,
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
659
which became the primary factor behind the decrease in male births during the famine.
Economic Cost of Gender-wise Discrimination in Health Gendered disparity in health has far-reaching macroeconomic implications for the growth trajectory and economic development given the dynamic nature of human capital formation (Bloom et al. 2020; Klasen 2000; Dollar and Gatti 1999). Using a dynamic general equilibrium model Bloom et al. (2020) argue contrasting implications for improvements in health by gender. While female health improvements accelerate demographic transition and thereby economic growth, improvements in male health delay the transition through the channel of increased fertility. Thus, when households prefer male health improvements that provide higher static utility gain there is a conflict between short-run household interests and long-run development goals. Klasen (2000) also find evidence of a robust association between gender equality and economic growth and development, using a cross-country panel. There is substantive evidence from across countries on the association between health and parents’ socioeconomic status highlighting intergenerational persistence in health (Cutler et al. 2008; Edwards 2017). Bhalotra and Rawlings (2013) find the intergenerational transmission of health weakens with mother’s education, immunization rates, and economic upturns, in the years of and before birth. Maternal deprivation or poor nutrition affects their fetus and is likely to lead to long-term health problems (Osmani and Sen 2003). Additionally the impact of ill-health in childhood affects health in adulthood through a feedback mechanism, which has labor market consequences such as low participation and lower earnings. Low social status may also cause poor health, an association that is more likely to hold in adulthood. All these structural relationships are stacked against women that reinforce future discrimination. Cunha and Heckman (2007) construct a model formalizing the critical and sensitive periods in child development and production process behind skill formation that includes self productivity, i.e., skill development in one stage of life improve skills attained thereafter and dynamic complementarity, i.e., skills produced in one stage raise the productivity of investment in future stages. Heckman (2012) stresses on policy measures and a developmental approach to remediate health disparities, focusing on prevention rather than cure in early life, through early childhood interventions. Using these theoretical relationships, we can hypothesize that discrimination in early years is likely to be magnified down the line affecting the gender gap across the various life stages. Understanding at what stage investments shape capabilities is imperative from a policy perspective to achieve equity and economic efficiency at the same time.
660
A. Dasgupta
Policy Insights It is clear from the existing body of evidence that gender-based discrimination within household is a big source of inequality in life outcomes. Discrimination in early life has negative consequences from both equity and efficiency point of view. The channels of self-productivity and dynamic complementarity for human capital formation imply that these costs would multiply over time. Furthermore, it feeds into the vicious cycle of low investments for females that reduces their productivity and innate valuation in the labor market and sustains the son-biased investments over time. Thus, policies such as old-age pension and financial support systems that weaken the link between economic dependence on sons during old age can help reduce the bias. Higher female literacy, employment, urbanization, and older age at marriage can also help address son bias in the population (Das Gupta et al. 2003; Chung and Das Gupta 2007). We see that gender gap in health outcomes magnifies during exposure to health shocks, such as when healthcare is financed through distress sale of assets. Thus, social security programs and universal basic insurance can help in this regard. However, it is also important to recognize that social norms such as kinship patterns (such as patrilocality) are difficult to change in the short-run. Hence, developing a framework of universal healthcare provision, in a rights-based framework can help. Additionally, sensitization of the community, including both men and women, about their right to healthcare; sensitizing parents about equal treatment of boys and girls can help. Policies targeting equality of opportunity in terms of health investments, such as through conditional cash transfers, can help mitigate gender-biased investments. Kashyap and Behrman (2020) and Sahn and Younger (2009) point out the need for targeting health investments in girls, specifically along with indirect measures to weaken son preference and improve women’s outcomes. Additionally, policymakers need to account for time cost or opportunity cost in addition to monetary costs of health investment and utilization decisions at the household level. Interventions making employment opportunities for women more accessible and addressing gender-pay gap can help improve outcomes for girls, making them more likely to be in school and have better health (Jensen 2010; Craigie and Dasgupta 2017). Further, as we realize, health utilization is a function of both demand and supply/ access, improving one without addressing the problem in the other will be of limited use. Importantly, there has to be a commensurate supply-side improvement along with the demand-side intervention and vice versa. A major problem in healthcare systems across countries is that though demand-side simulations like vouchers and conditional cash transfers have helped mitigate women’s financial barriers to maternal and child healthcare access, they have not been sufficiently met by supply-side improvements (Witter et al. 2017; Sen and Govender 2015). On similar lines, supply-side, top-down policies may be of limited use and can even backfire in the presence of a demand-side constraint (Anukriti 2018; Dasgupta and Sharma 2021, 2022).
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
661
There is an urgent need to address the paucity of data on quality of healthcare, more so in developing countries. Vignettes-based survey instruments can be handy in detecting and assessing provider knowledge and quality of care (Das and Hammer 2005, 2007; Das et al. 2008). A rigorous analytical framework is needed in order to assess and address gender-based discrimination in health along the various points in the life cycle. Importantly, the framework needs to account for features of the data generation process, taking account of potential biases from gender biased selective attrition, mortality or reporting differences. There is ample scope for innovation and improvement on collection, harmonization, and usage of microdata that are genderdisaggregated, which can feed into research.
References Acharya S (2010) Caste and patterns of discrimination in rural public health care services. In: Blocked by caste: economic discrimination in modern India. Oxford University Press, New Delhi, pp 208–229 Adamson J, Ben-Shlomo Y, Chaturvedi N, Donovan J (2003) Ethnicity, socio-economic position and gender—do they affect reported health—care seeking behaviour? Soc Sci Med 57(5): 895–904 Ahmed SM, Adams AM, Chowdhury M, Bhuiya A (2003) Changing health-seeking behaviour in Matlab, Bangladesh: do development interventions matter? Health Policy Plan 18(3):306–315 Albanesi S, Olivetti C (2016) Gender roles and medical progress. J Politic Econ 124(3):650–695 Alderman G (1997) Intrahousehold Resource Allocation in Developing Countries. Johns Hopkins University Press Anand N, Chen X (2018) Investing in the womb: Identifying gender discrimination through the lens of prenatal ultrasounds. Technical report, Mimeo, Yale University Anderson S, Ray D (2010) Missing women: age and disease. Rev Econ Stud 77(4):1262–1300 Angerer S, Waibel C, Stummer H (2019) Discrimination in health care: A field experiment on the impact of patients’ socioeconomic status on access to care. Am J Health Econ 5(4):407–427 Anukriti S (2018) Financial incentives and the fertility-sex ratio trade-off. Am Econ J Appl Econ 10(2):27–57 Anukriti S, Bhalotra S, Tam EH (2020) On the quantity and quality of girls. Policy Research Working Paper Bago d’Uva T, Van Doorslaer E, Lindeboom M, O’Donnell O (2008) Does reporting heterogeneity bias the measurement of health disparities? Health Econ 17(3):351–375 Baird S, Friedman J, Schady N (2011) Aggregate income shocks and infant mortality in the developing world. Rev Econ Stat 93(3):847–856 Balsa AI, McGuire TG (2003) Prejudice, clinical uncertainty and stereotyping as sources of health disparities. J Health Econ 22(1):89–116 Balsa AI, McGuire TG, Meredith LS (2005) Testing for statistical discrimination in health care. Health Serv Res 40(1):227–252 Banerjee A, Deaton A, Duflo E (2004) Wealth, health, and health services in rural Rajasthan. Am Econ Rev 94(2):326–330 Barford A, Dorling D, Smith GD, Shaw M (2006) Life expectancy: women now on top everywhere. Br Med J 332:808 Batra A, Gupta I, Mukhopadhyay A et al (2014) Does discrimination drive gender differences in health expenditure on adults: evidence from cancer patients in rural India. Technical Report Behrman JR (1988) Intrahousehold allocation of nutrients in rural India: Are boys favored? do parents exhibit inequality aversion? Oxf Econ Pap 40(1):32–54
662
A. Dasgupta
Bertrand M, Pan J (2013) The trouble with boys: Social influences and the gender gap in disruptive behavior. Am Econ J Appl Econ 5(1):32–64 Bhalotra S, Rawlings S (2013) Gradients of the intergenerational transmission of health in developing countries. Rev Econ Stat 95(02):660–672 Bharadwaj P, Lakdawala LK (2013) Discrimination begins in the womb: evidence of sex-selective prenatal investments. J Hum Resour 48(1):71–113 Bharadwaj P, Fenske J, Kala N, Mirza RA (2020) The green revolution and infant mortality in India. J Health Econ 71:102314 Bloom DE, Kuhn M, Prettner K (2020) The contribution of female health to economic development. Econ J 130(630):1650–1677 Burger R (2019) A promising approach to identifying health system discrimination and bias. Lancet Glob Health 7(5):e546–e547 Cerceau S (2012) Gender equity in access to health care: the role of social health protection, Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH, Eschborn Chung W, Das Gupta M (2007) The decline of son preference in South Korea: The roles of development and public policy. Popul Dev Rev 33(4):757–783 Craigie T-A, Dasgupta S (2017) The gender pay gap and son preference: evidence from India. Oxf Dev Stud 45(4):479–498 Cunha F, Heckman J (2007) The technology of skill formation. Am Econ Rev 97(2):31–47 Cutler DM, Lleras-Muney A, Vogl T (2008) Socioeconomic status and health: dimensions and mechanisms. Oxford Daniels B, Kwan A, Satyanarayana S, Subbaraman R, Das RK, Das V, Das J, Pai M (2019) Use of standardised patients to assess gender differences in quality of tuberculosis care in urban India: a two-city, cross-sectional study. Lancet Glob Health 7(5):e633–e643 Das Gupta M, Mari Bhat P (1997) Fertility decline and increased manifestation of sex bias in India. Popul Stud 51(3):307–315 Das Gupta M, Zhenghua J, Bohua L, Zhenming X, Chung W, Hwa-Ok B (2003) Why is son preference so persistent in east and south Asia? a cross-country study of China, India and the Republic of Korea. J Dev Stud 40(2):153–187 Das J, Hammer J (2005) Which doctor? combining vignettes and item response to measure clinical competence. J Develop Econ 78(2):348–383 Das J, Hammer J (2007) Money for nothing: the dire straits of medical practice in Delhi, India. J Develop Econ 83(1):1–36 Das J, Hammer J, Leonard K (2008) The quality of medical advice in low-income countries. J Econ Perspect 22(2):93–114 Dasgupta A (2018) Systematic measurement error in self-reported health: is anchoring vignettes the way out? IZA J Dev Migrat 8(1):1–30 Dasgupta A, Sharma A (2021) Can legal bans on sex detection technology reduce gender discrimination? Technical report. Ashoka University, Department of Economics Dasgupta A, Sharma A (2022) Missing women: A review of underlying causes and policy responses, vol 06). Oxford Research Encyclopedia of Economics and Finance Dasgupta U, Mani S, Singh P (2016) Searching for religious discrimination among anganwadi workers in India: An experimental investigation Deaton A (2002) Policy implications of the gradient of health and wealth. Health Aff 21(2):13–30 Deaton A (2008) Height, health, and inequality: the distribution of adult heights in India. Am Econ Rev 98(2):468–474 Deccan Herald (2021) Worrying gender gap in vaccination. Dehejia R, Lleras-Muney A (2004) Booms, busts, and babies’ health. Quart J Econ 119(3): 1091–1130 Dinkelman T (2017) Long-run health repercussions of drought shocks: evidence from South African homelands. Econ J 127(604):1906–1939 Dollar D, Gatti R (1999) Gender inequality, income, and growth: are good times good for women? vol 1. Development Research Group, The World Bank Washington, DC
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
663
Dupas P, Jain R (2021) Women left behind: Gender disparities in utilization of government health insurance in India. Technical Report, Working Paper Ebenstein A (2014) Patrilocality and missing women. Available at SSRN 2422090. Ebenstein A, Leung S (2010) Son preference and access to social insurance: evidence from china’s rural pension program. Popul Develop Rev 36(1):47–70 Edwards M (2017) The barker hypothesis. Handbook of Famine, Starvation, and Nutrient Deprivation: From Biology to Policy, pp 191–211 Ferreira FHG, Schady N (2009) Aggregate Economic Shocks, Child Schooling, and Child Health. World Bank Res Obs 24(2):147–181 Fledderjohann J, Agrawal S, Vellakkal S, Basu S, Campbell O, Doyle P, Ebrahim S, Stuckler D (2014) Do girls have a nutritional disadvantage compared with boys? statistical models of breastfeeding and food consumption inequalities among Indian siblings. PLoS One 9(9): e107172 Gajovic G, Janicijevic K, Andric D, Djurovic O, Radevic S (2019) Gender differences in health care utilization among the elderly. Serb J Exp Clin Res 1. (ahead-of-print) Gambin LM et al (2005) The impact of health on wages in Europe-does gender matter. HEDG WP 5(03) Gao M, Yao Y (2006) Gender gaps in access to health care in rural china. Econ Dev Cul Change 55(1):87–107 Garg A, Morduch J (1998) Sibling rivalry and the gender gap: Evidence from child health outcomes in Ghana. J Popul Econ 11(4):471–493 George A, Iyer A, Sen G (2005) Gendered health systems biased against maternal survival: preliminary findings from Koppal, Karnataka, India. Glied S, Lleras-Muney A (2008) Technological innovation and inequality in health. Demography 45(3):741–761 Goltz JVD, Dar A, Fishman R, Mueller ND, Barnwal P, McCord GC (2020) Health impacts of the green revolution: Evidence from 600,000 births across the developing world. J Health Econ 74:102373 Govender V, Penn-Kekana L (2007) Gender biases and discrimination: a review of health care interpersonal interactions. Grol-Prokopczyk H, Freese J, Hauser RM (2011) Using anchoring vignettes to assess group differences in general self-rated health. J Health Soc Behav 52(2):246–261 Haddad LJ, Peña C, Nishida C, Quisumbing AR, Slack A (1996) Food security and nutrition implications of intrahousehold bias: a review of literature. International Food Policy Hazarika G (2000) Gender differences in children’s nutrition and access to health care in Pakistan. J Dev Stud 37(1):73–92 Heckman JJ (2012) The developmental origins of health. Health Econ 21(1):24 Jayachandran S (2015) The roots of gender inequality in developing countries. Dent Econ 7(1): 63–88 Jayachandran S, Kuziemko I (2011) Why do mothers breastfeed girls less than boys? evidence and implications for child health in India. Quart J Econo 126(3):1485–1538 Jayachandran S, Pande R (2017) Why are Indian children so short? the role of birth order and son preference. Am Econ Rev 107(9):2600–2629 Jensen RT (2010) Economic opportunities and gender differences in human capital: Experimental evidence for India. Working Paper 16021, National Bureau of Economic Research Johar M, Jones G, Keane MP, Savage E, Stavrunova O (2013) Discrimination in a universal health system: Explaining socioeconomic waiting time gaps. J Health Econ 32(1):181–194 Kapoor M, Agrawal D, Ravi S, Roy A, Subramanian S, Guleria R (2019) Missing female patients: an observational analysis of sex ratio among outpatients in a referral tertiary care public hospital in India. BMJ Open 9(8):e026850 Kashyap R, Behrman J (2020) Gender discrimination and excess female under- 5 mortality in India: A new perspective using mixed-sex twins. Demography 57(6):2143–2167
664
A. Dasgupta
Kishore A, Spears D (2014) Having a son promotes clean cooking fuel use in urban India: Women’s status and son preference. Econ Dev Cul Change 62(4):673–699 Klasen S (1994) “Missing women” reconsidered. World Dev 22(7):1061–1071 Klasen S (2000) Does gender inequality reduce growth and development? evidence from crosscountry regressions. Klasen S, Wink C (2003) “Missing women”: Revisiting the debate. Feminist Econ 9(2–3):263–299 Kumar K, Singh A, James K, McDougal L, Raj A (2020) Gender bias in hospitalization financing from borrowings, selling of assets, contribution from relatives or friends in India. Soc Sci Med 260:113222 Maccini S, Yang D (2009) Under the weather: Health, schooling, and economic consequences of early-life rainfall. Am Econ Rev 99(3):1006–1026 Marcoa R, Z. D. (2018) Tuberculosis and gender – factors influencing the risk of tuberculosis among men and women by age group. Pulmonol J 24(3):199–202 Mishra V, Roy TK, Retherford RD (2004) Sex differentials in childhood feeding, health care, and nutritional status in India. Popul Dev Rev 30(2):269–295 Molina T (2016) Reporting heterogeneity and health disparities across gender and education levels: Evidence from four countries. Demography 53(2):295–323 Mwabu G, Ainsworth M, Nyamete A (1993) Quality of medical care and choice of medical treatment in Kenya: an empirical analysis. J Hum Resour:838–862 Oksuzyan A, Danko MJ, Caputo J, Jasilionis D, Shkolnikov VM (2019) Is the story about sensitive women and stoical men true? Gender differences in health after adjustment for reporting behavior. Soc Sci Med 228:41–50 Osmani S, Sen A (2003) The hidden penalties of gender inequality: fetal origins of ill-health. Econ Hum Biol 1(1):105–121 Pande RP (2003) Selective gender differences in childhood nutrition and immunization in rural India: the role of siblings. Demography 40(3):395–418 Peltzer K, Williams JS, Kowal P, Negin J, Snodgrass JJ, Yawson A, Minicuci N, Thiele L, Phaswana-Mafuya N, Biritwum RB et al (2014) Universal health coverage in emerging economies: findings on health care utilization by older adults in China, Ghana, India, Mexico, the Russian Federation, and South Africa. Glob Health Action 7(1):25314 Pongou R (2013) Why is infant mortality higher in boys than in girls? a new hypothesis based on preconception environment and evidence from a large sample of twins. Demography 50(2): 421–444 Rao S (2009) Tuberculosis and patient gender: An analysis and its implications in tuberculosis control. Lung India 26(2):46–47 Rosenblum D (2013) The effect of fertility decisions on excess female mortality in India. J Popul Econ 26(1):147–180 Sabarwal S, Sinha N, Buvinic M (2010) How do women weather economic shocks? a review of the evidence. A Review of the Evidence (December 1, 2010). World Bank Policy Research Working Paper (5496) Sadana R, Tandon A, Murray CJ, Serdobova I, Cao Y, Xie W, Chatterji S, Ustun BL (2002) Describing population health in six domains: comparable results from 66 household surveys. World Health Organization, Geneva Sahn DE, Younger SD (2009) Measuring intra-household health inequality: explorations using the body mass index. Health Econ 18(S1):S13–S36 Saikia N, Bora JK (2016) Gender difference in health-care expenditure: evidence from India human development survey. PLoS One 11(7):e0158332 Saikia N et al (2019) Gender disparities in health care expenditures and financing strategies (hcfs) for inpatient care in India. SSM Popul Health 9:100372 Sen A (1992) Missing women. BMJ 304(6827):587 Sen A (1993) Positional objectivity. Philos Public Aff:126–145 Sen A (2003) Missing women-revisited. BMJ 327(7427):1297–1298
28
Gender-Based Discrimination in Health: Evidence from Cross-Country
665
Sen G, Govender V (2015) Sexual and reproductive health and rights in changing health systems. Glob Public Health 10(2):228–242 Shaikh MR (2018) Waiting time at health facilities and social class: Evidence from the Indian caste system. Shaikh M, Peters SA, Woodward M, Norton R, Jha V (2018) Sex differences in utilisation of hospital care in a state-sponsored health insurance programme providing access to free services in south India. BMJ Glob Health 3(3):e000859 Shim JK (2010) Cultural health capital: a theoretical approach to understanding health care interactions and the dynamics of unequal treatment. J Health Soc Behav 51(1):1–15 Singh P, Mitra S (2013) Absolute versus relative performance pay: Evidence from an experiment targeting child malnutrition in West Bengal. Available at SSRN 2332896. Song S (2012) Does famine influence sex ratio at birth? evidence from the 1959–1961 great leap forward famine in china. Proc R Soc Lond B Biol Sci 279(1739):2883–2890 Van Ryn M, Fu SS (2003) Paved with good intentions: do public health and human service providers contribute to racial/ethnic disparities in health? Am J Public Health 93(2):248–255 Vikram K, Chen F, Desai S (2018) Mothers’ work patterns and children’s cognitive achievement: evidence from the India human development survey. Soc Sci Res 72:207–224 Visaria L (2008) Improving the child sex ratio: Role of policy and advocacy. Econ Pol Wkly:34–37 Vlassoff C (2007) Gender differences in determinants and consequences of health and illness. J Health Popul Nutr 25(1):47 Witter S, Govender V, Ravindran TS, Yates R (2017) Minding the gaps: health financing, universal health coverage and gender. Health Policy Plan 52(suppl_5):v4–v12
Dimensions of Discrimination: Discrimination in Housing
29
Saugato Datta and Vikram Pathania
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature Review: Housing Discrimination Globally . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Housing Discrimination in India: Theoretical Considerations, Historical Experience . . . . . . . . Existing Literature on Housing Discrimination in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Location and Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Names and Contact Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tracing Callers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Descriptive Statistics: Listings, Landlords, and Applicants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Listing and Landlord Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Applicant and Application Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result 1: Landlords Are Significantly Less Likely to Respond to Muslims Applicants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result 2: Relative to UC Applicants, Muslim Applicants Receive Fewer Callbacks But Landlords Who Do Respond Make Similar Numbers of Attempts to Contact UC and Muslim Applicants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result 3: Differences in the Probability of Response and Count of Responses Between UC and SC/OBC Are Not Statistically Significant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result 4: There Is Suggestive Evidence that Landlords Who Respond to Both UC and Muslim (or SC) Applicants Are More Likely to Call UCs Sooner as Compared to Muslims (SC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result 5: There Is Heterogeneity By Gender and Religion of Landlord and By Size and Rental Price of the Listed Property in the Likelihood of Response to the Various Applicant Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
668 672 674 676 677 677 678 679 680 681 681 682 683 685
685 687
689
691
S. Datta (*) ideas42, Boston, MA, USA e-mail: [email protected] V. Pathania University of Sussex, Falmer, UK © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_13
667
668
S. Datta and V. Pathania
Discussion, Interpretation, and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Abstract
People from religious or ethnic minorities or other historically disadvantaged groups frequently face discrimination when they seek rental housing. This chapter surveys the literature on such housing discrimination, focusing on the experimental literature, including in-person housing audits (as first pioneered by the US Department of Housing and Urban Development) and the more recent development of remote impersonal housing audits in the United States and Europe. It then describes an audit experiment carried out on India’s largest real estate websites, which document striking variations between landlords’ treatment of upper-caste Hindus, Other Backward Classes, Scheduled Castes, and Muslims. Strong evidence is found of discrimination against Muslim applicants, both in terms of probability of being contacted and the number of contacts, relative to upper-caste Hindu (UC) applicants, in the rental housing market in Delhi and its largest suburbs. While the probability that a landlord responds to an upper-caste applicant is 0.35, this is only 0.22 for a Muslim applicant. There is also suggestive evidence that when landlords respond to both UC and Muslim applicants, they call back the UC applicant sooner. Muslim applicants are especially disadvantaged when applying to rent one-bedroom houses; there is an additional 20 percentage point reduction in the probability of a callback. In contrast, there is no clear evidence that landlords are less likely to respond to Scheduled Castes and other backward classes. However, our estimates may understate the true differentials in callback ratios as a result of the failure to perfectly link all callbacks to a listing as well as Scheduled Caste/Other Backward Class names acting as only imperfect signals of caste background. Keywords
Housing · India · Discrimination
Introduction The persistence of disparate educational, economic, financial, and health outcomes between social groups defined along racial, ethnic, gender, or religious lines has been a long-standing concern for public policy in many countries across the world, and especially so in countries which have a high degree of social heterogeneity. Scholars in fields ranging from anthropology to sociology to criminal justice have consequently employed a variety of disciplinary perspectives to attempt to understand the forces – whether legal, social, institutional, behavioral, or political – that generate and sustain such gaps in outcomes between members of different social groups (Dohan 2003; Hirsch 1983; Lamb 2005; Pager and Shepherd 2008).
29
Dimensions of Discrimination: Discrimination in Housing
669
While acknowledging the possible role of innate or learned differences in ability or preferences between groups, discrimination against marginalized or historically disadvantaged groups by members of more powerful groups is widely held to be a plausible source of at least some (and sometimes a major) part of observed intergroup differences. Indeed, at least since Becker (1957), economists have sought to understand, establish, and quantify the possible role of discrimination to explain worse outcomes (including those related to employment, wages, access to credit, job performance, housing patterns, and occupational choice) among members of marginalized or disadvantaged groups, racial or ethnic minorities, and women than the corresponding outcomes among historically dominant or privileged groups, e.g., racial or ethnic majorities, nonimmigrants, men, etc. (Aigner and Cain 1977; Altonji and Blank 1999; Arrow 1973; Bertrand and Mullainathan 2003; Blau and Kahn 2006; Blau and Ferber 1987). In most settings where such disparities have been carefully studied, researchers have found that some of the gap between groups can be attributed to discrimination. For example, Bertrand and Mullainathan (2003) find evidence of a large racial gap in callbacks in response to job applications in the United States. Pager et al. (2009) carry out a hiring audit focused on low-wage jobs in New York City and find that black applicants are half as likely as equally qualified whites to receive a callback or job offer, with Black and Latino applicants with “clean backgrounds” faring no better on average than white applicants just released from prison. Within this broad area of research, a growing literature focuses on the existence, drivers, and magnitude of discrimination in housing markets against members of racial, ethnic, caste, or religious minorities, who are often observed to have unequal access to high-quality housing in desirable locations or to be clustered in locations with lower-quality housing, poorer services, etc. (Choi et al. 2005; Galster 1991; Turner et al. 2002). The theoretical justification for this literature arises in the first instance from considerations of fairness and the right of individuals and households to live where they choose (Danziger and Lin 2000; Massey and Denton 1993). Beyond this, research into this question is motivated by the recognition that inequities in access to housing lead to worsening residential segregation, which is often recognized as a social ill or policy “bad” on its own terms (Denton 1999; South and Crowder 1998). Further reasons to explore housing discrimination stem from the recognition of the effects of discriminatory housing practices on a variety of downstream policy outcomes. Here, the broad argument is that given existing spatial inequalities in the quality of public services in most countries, housing discrimination limits the ability of individuals and households from disadvantaged groups to access quality schooling and healthcare, which in turn affects their schooling, labor, and credit market outcomes and could therefore lead to the persistence of measured intergroup inequality (Galster 1991; Yinger 1995, Kain and Quigley 1972). As with other forms of discrimination (e.g., discrimination in labor and credit markets), it is difficult for researchers to infer unequal treatment in the housing market from aggregate data alone, since some relevant characteristics of applicants may be visible to potential employers, lenders, or landlords but unknown to the researcher. As a result, studies that decompose observed outcomes by observable
670
S. Datta and V. Pathania
applicant characteristics risk finding biased results, where a portion of the variation that is in fact due to some characteristic of the applicant unobservable to the researcher but observable to the landlord or interviewer is erroneously attributed to discrimination, a classic instance of omitted variable bias (Altonji and Blank 1999). Efforts to mitigate this bias have a variety of econometric techniques, such as the use of instrumental variables. For example, Hubbard et al. (2011) use an IV strategy to show that minorities applying for mortgages are more likely to be rejected than whites in the prime housing market but less likely to be rejected than whites in the subprime market, results supportive of the information-based theory of discrimination. Beyond this, researchers studying discrimination have used quasi-experimental approaches. An influential quasi-experiment in the field of workplace discrimination is Goldin and Rouse (2000), which uses the effect of the introduction of blind auditions into orchestra hiring on women musicians’ hiring outcomes to measure gender discrimination. More recently, researchers on discrimination – whether in labor, credit, or housing markets – to the problems with observational studies have turned to experimentation. Experimental approaches in the field of discrimination have centered on audit studies. In the area of labor market discrimination, researchers have used hiring audits, where comparable minority and nonminority candidates are sent to actual interviews to measure the existence and extent of differential treatment (Altonji and Blank 1999). These studies are closely related to “mystery shopping” studies of various kinds (e.g., Mullainathan et al. (2012) on the market for financial advice) where trained actors are enlisted to model various characteristics or needs for a service provider. Rubineau and Kang (2012) use repeated quasi-audit studies to test for within-cohort changes in disparities in medical student behaviors as they interact with white and black patient actors, finding that disparate behaviors increase between the first and second years of medical school. Fix and Struyk (1993) provide a comprehensive overview of audit studies into various facets of discrimination in the United States. The counterpart of these studies for housing discrimination is “housing audits,” where researchers send actors from different groups of interest who are trained to otherwise present as identical and follow preset scripts and have been used in order to measure differences in apartments or houses shown, rental prices quoted, and of course ability to rent. Such systematic audit has played an important role in enabling researchers to measure the extent of discrimination against minorities in housing markets in the United States more credibly than would be possible using aggregate data and has been institutionalized by the Department of Housing and Urban Development in the form of the Housing Market Practices Survey (HMPS) in 1977 and three subsequent housing discrimination studies (1989, 2000, and 2012), whose results are summarized in Turner et al. (1991, 2002, 2013). Audit studies clearly offer many advantages over cross-sectional analyses. However, as Bertrand and Mullainathan (2003) point out in the context of labor market discrimination, they have at least one major shortcoming. The fact that actors are aware of the purpose of the experiment (i.e., that the experiment is not and cannot be double blind) increases the inherent difficulty of ensuring the absence of
29
Dimensions of Discrimination: Discrimination in Housing
671
observational differences between mock job candidates in a pair and could therefore introduce error into the estimates of discrimination. These shortcomings have led researchers to devise what could be called “impersonal” or “automated” audit studies, with resume audits a la Bertrand and Mullainathan (2003) being a prime example. In this class of studies, researchers vary the perceived group membership of a fictitious applicant (for a job or an apartment) without enlisting actors to interact with the potential employer, landlord, or other service provider by relying on the ability to apply remotely for jobs or apartments using mail or Internet-based applications. These studies provide the cleanest possible evidence of differential treatment, but with the significant limitation that they can only capture differential treatment for a much earlier stage of the hiring process than might be possible with an in-person audit (Bertrand and Mullainathan 2003; Banerjee et al. 2009). In the labor market case, this means that they can only measure whether a potential employer called an applicant back, and not whether the potential employee was in fact offered a job, or what salary was offered. Nonetheless, as Bertrand and Mullainathan (2003) note, “to the extent that the search process has even moderate frictions, one would expect that reduced interview rates would translate into reduced job offers.” As with labor markets, such “remote” audits have also been carried out by researchers studying housing discrimination, with the bulk being in developed countries (Ahmed and Hammarstedt 2008; Andersson et al. 2012; Choi et al. 2005; Yinger 1995; Saltman 1979). Oh and Yinger (2015) provide an overview of the evidence generated by these correspondence studies in housing in North America and Europe. Very little evidence exists on non-developed country settings. The remainder of this chapter describes a recent attempt to apply these methods to the case of caste and religious discrimination in housing markets in urban India. In India, discrimination is a salient issue for policy. There are persistent differences in key outcomes between various social groups differentiated by religion as well as caste, which historically has been the primary axis of social differentiation in South Asia (Beteille 1992). A growing body of literature has begun to quantify caste-based discrimination in labor markets (Banerjee et al. 2009; Deshpande 2011; Siddique 2011; Thorat and Neuman 2012). In the area of housing, here are several historical and anthropological accounts of enforced housing segregation along caste lines (Beteille 1969; Dumont 1980; Ghurye 1961; Srinivas 1957). There is also considerable anecdotal evidence, some of which has received widespread media coverage, about the existence of discrimination against Muslims, the country’s largest religious minority (see Wire Staff 2015). However, until recently, there has been little experimental evidence on the extent to which religious and caste minorities in India experience housing discrimination. A literature search suggests that there is only one published experimental study of housing discrimination in India. Thorat et al. (2015) use a variety of audit techniques to document the differential treatment of Muslims and Scheduled Castes in the housing market of India’s national capital region; their results are discussed in section “Existing Literature on Housing Discrimination in India.” A web-based audit of the market for rental properties offered directly by owners/landlords using a sample of 170 rental properties in the Delhi region is thus conducted. The findings
672
S. Datta and V. Pathania
complement and extend the results in Thorat et al. (2015) by adopting a different experimental strategy to generate what are likely cleaner measures of differential treatment attributable only to perceived differences in caste and religion. The central findings are as follows. There is strong evidence of discrimination against Muslim applicants, both in terms of probability of being contacted and the number of contacts. Where the probability that a landlord contacts an upper-caste Hindu (henceforth, “UC”) tenant is 0.35, this is only 0.22 for a Muslim seeking to rent the same property. A prospective Muslim tenant must respond to 45.5 listings to receive 10 landlord callbacks, while a UC tenant must respond to only 28.6 listings to receive the same number. In other words, a Muslim tenant must contact approximately 60% more landlords to receive an equivalent number of callbacks as a UC tenant. This points to a significant disadvantage faced by Muslim applicants, who must expend significantly more effort to find housing, relative to upper-caste Hindus. In contrast, the study fails to find statistically significant evidence of bias against Scheduled Castes (SCs) or other backward classes (OBCs). However, these estimates may understate the true differentials in callback ratios as a result of the failure to perfectly link all callbacks to a listing. In other notable results, there is some suggestive evidence that landlords wait longer to call Muslim (and to a lesser extent Scheduled Caste) applicants back after receiving a query from them than they do to call back upper-caste Hindus. The results also indicate some heterogeneity in landlord responses to applicant types. Notably, landlords offering one-bedroom properties are 20% points less likely to respond to Muslims. As a rule, applicants to one-bedroom properties tend to be single men or women. Since all applicants in the study are male, this implies that the housing rental market is especially hostile to single Muslim men. Also, Muslim landlords are no more likely to respond to Muslim applicants.
Literature Review: Housing Discrimination Globally A growing strand within the broader literature on discrimination against underprivileged or marginalized groups focuses on the existence, drivers, and magnitude of discrimination in housing markets against members of racial, ethnic, caste, or religious minorities. In comparison to other “majority” groups, members of these groups are often observed to have unequal access to high-quality housing in desirable locations or to be clustered in locations with lower-quality housing, poorer services, etc. (Choi et al. 2005; Galster 1991; Turner et al. 2002). As with other dimensions of discrimination, the well-understood econometric issues that arise when attempting to infer discrimination from observational studies have led researchers on housing discrimination to espouse a variety of experimental techniques to arrive at “cleaner” measures of discrimination. A major strand of research has therefore used “housing audits,” where researchers send actors from different groups of interest who are trained to otherwise present as identical and follow preset scripts, in order to measure differences in apartments or houses shown, rental prices quoted, and of course ability to rent. These studies are also closely related to
29
Dimensions of Discrimination: Discrimination in Housing
673
“mystery shopping” studies of various kinds (e.g., Mullainathan et al. (2012) on the market for financial advice) where trained actors are enlisted to model various characteristics or needs for a service provider, including hiring audits in labor markets, where comparable minority and nonminority candidates are sent to actual interviews to measure the existence and extent of differential treatment (Altonji and Blank 1999). Such systematic audit or correspondence studies have played an important role in enabling researchers to measure the extent of discrimination against minorities in housing markets in the United States more credibly than would be possible using aggregate data (see Saltman 1979; Yinger 1995). In the United States in particular, these studies have been institutionalized by the Department of Housing and Urban Development (HUD) which sponsors regular housing audits in many large metropolitan areas, including the Housing Market Practices Survey (HMPS) in 1977 and three subsequent housing discrimination studies (Turner et al. 1991, 2002, 2013). The scope of these studies, which ranged from 3264 pairwise tests focusing on discrimination against black renters in 40 metropolitan areas in the United States in 1977 to 8047 pairwise tests of discrimination against black, Latino, and Asian tenants in 28 metropolitan areas in 2012, as well as a variety of facets along which black, Latino, and Asian tenants are disadvantaged relative to white ones, is summarized in Oh and Yinger (2015). These studies also allow researchers to track temporal and spatial changes in minorities’ outcomes in the housing market. For example, Choi et al. (2005) use national audit data from the 2000 housing discrimination study and show that rental housing discrimination has declined since 1989 but that it persists in several types of housing agent behavior and is driven by agents’ own prejudice and by their response to the prejudice of their white clients. Each study also found evidence of “racial steering,” where non-white applicants are “steered” toward neighborhoods with a lower percentage of white residents (Oh and Yinger 2015). A more recent strand in this literature exploits the growth of classifieds and Internet-based housing portals, which make it possible for a potential tenant to express interest in an apartment or house offered for rent without interacting with the putative landlord either in person or over the telephone. This has made it possible to implement “remote” audits in the housing market. Oh and Yinger (2015) provide a comprehensive overview of the scope and findings from these studies, some of which are discussed below. For the United States, Ewens, Tomlin, and Wang (2014) find a 8–9 percentage point lower rate of positive responses for AfricanAmerican renters compared to white ones, and that revealing positive information about socioeconomic deficit does not reduce this gap. Hanson and Hawley (2011) find that landlords reply faster, more “politely,” “formally,” and at greater length to white renters than African-American ones. Hanson and Santas (2014) find some evidence of discrimination against recent Hispanic immigrants. Elsewhere in North America, Hogan and Berry (2011) find significant discrimination against Asian men, black applicants, Arab women, and Arab men (who face the highest measured level of discrimination) in an email audit carried out in Toronto, Canada. For Europe, Ahmed and Hammarstedt (2008) find that Arab males have 21–26 percentage points
674
S. Datta and V. Pathania
lower probability of being invited to further contacts and a 7 percentage point lower probability of being invited to see a flat than Swedish males. Ahmed, Andersson, and Hammarstedt (2010) similarly find that Arab males have a 15–23 percentage point lower callback rate than Swedish males and that this gap is not reduced by the provision of more information about the prospective tenant. Baldini and Federici (2011) study discrimination in Italy and find that Arab and Eastern European renters are both less likely (22 and 16 ppt, respectively) to receive a positive response to an emailed query than native Italian applicants and that males face more discrimination than women. Carlsson and Erickson (2013) find that ethnic minority applicants in the United Kingdom have a 13 percentage point lower probability of being invited to see an apartment than white British applicants, with Arab applicants facing the most discrimination. As in the case of labor market discrimination studies, such audits in the case of housing markets must restrict themselves to measuring an early-stage outcome (i.e., whether the landlord applied to responded to a potential tenant’s expression of interest) without being able to measure differences in downstream outcomes such as what apartments were offered or the terms on which apartments were made available. Nonetheless, as in the case of resume audits, impersonal housing audits do provide the cleanest possible evidence of differential treatment of different categories of people seeking housing of all study methods. And, building upon the argument in Bertrand and Mullainathan (2003), the existence of search frictions should mean that those from groups that receive fewer landlord callbacks are also likely to be offered fewer houses or to take longer to find a house that meets their requirements.
Housing Discrimination in India: Theoretical Considerations, Historical Experience There are several reasons that suggest, a priori, that discrimination might impinge upon the results of certain categories of individuals’ or families’ quests to rent or buy housing in India. Further, it appears ex ante likely that such discrimination may be faced both by individuals who occupy a historically subordinate position within the hierarchy of caste, the complex system of hierarchical social relations that long governed and in important ways still governs some aspects of Indian society (notably marriage, see Banerjee et al. (2013)), and by Muslims (India’s largest religious minority, which by no means socioeconomically homogeneous, does on average have poorer socioeconomic indicators on a host of dimensions than upper-caste Hindus). First, the strong correlations between caste and religion and socioeconomic status, occupational choice, housing outcomes, etc. seen in the data suggest a possible role for discrimination against members of these groups along some or all of these dimensions (see Banerjee et al. 2009 for a related discussion). In general, upper-caste Hindus have better economic outcomes than both non-upper-caste Hindus (including but not limited to Scheduled Castes and the other backward classes or OBCs) and Muslims.
29
Dimensions of Discrimination: Discrimination in Housing
675
Using data from India’s National Sample Survey, a nationally representative repeat cross-sectional household survey, Deshpande (2005) provides a detailed account of levels and patterns of consumption expenditure (used as a proxy for income in the absence of reliable income data) by social group over a 20-year period. She finds that Scheduled Castes/Tribes (SCs/STs) have lower per capita consumption expenditure than other groups, and while each social group’s consumption expenditure has risen over the 20-year period studied, the increase is lower for SC/ST than the rest of the population. Deshpande (2001) uses five indicators of standard of living (land holding, occupation, education, ownership of consumer durables, and of livestock to construct a “Caste Development Index” (CDI) and finds that in the early 1990s, there was no state where the CDI of SC/ST populations was higher than that of non-SC/ST populations. Desai and Dubey (2012) analyze data from a nationally representative survey of 41, 554 households conducted in 2005 to argue that there continues to be persistent disparities in education, income, and social networks by caste. There are also persistent disparities in housing quality. For instance, according to the 2011 census, only 34% of SC households have a latrine on the premises, compared with 46.7% of all households in the country (Office of the Registrar General and Census Commissioner 2012). The existence of affirmative action for members of the Scheduled Castes, Tribes, and other backward classes necessitates official data collection about these groups. Since there is no such national policy for Muslims, data on Hindu-Muslim differentials are sparser. Nevertheless, what data there are is suggestive. For example, while the widespread prevalence of disguised unemployment or underemployment in developing countries limits the utility of official measures of unemployment, it is striking that both Scheduled Castes and Muslims are overrepresented among those whom the Indian state classifies as “marginal workers” – those employed for less than 6 months of the prior year. This figure is 10.9% for Scheduled Castes and 6.5% for Muslims, while it is 3.5% for the country as a whole (Banerjee et al. 2009). According to the Sachar Committee, a government committee set up to probe the socioeconomic status of Muslims in India, Muslims had the highest unemployment rate of any socioreligious group in India (Sachar 2006). Shariff (1995), one of the few analyses of socioeconomic differentials between Hindus and Muslims in India, finds that 22% of Muslims had a monthly household per capita expenditure of less than Rs.110 in 1987–1988, compared with 13.1% of Hindus. It is worth noting that given well-documented caste differentials within the Hindu population, the difference between Muslims and the non-Scheduled Caste/Tribe Hindu population would likely be even higher. Secondly, the nature of caste and religion in India lends strong a priori plausibility to the idea that members of lower-caste and minority groups are likely to face discrimination in the housing market in particular. Along with restrictions on commensality, intermarriage, education, and occupational choice, housing segregation is central to the logic of caste (Beteille 1969; Dumont 1980; Ghurye 1961; Srinivas 1957). Just as there were restrictions on inter-dining and intermarriage between members of “higher” and “lower” castes (particularly those belonging to the groups formerly referred to as untouchable and since independence as Scheduled
676
S. Datta and V. Pathania
Castes), housing in most Indian villages and towns was organized along caste and community lines to conform with strongly held notions of purity and pollution where the presence of certain groups within a certain distance of higher-status groups was held to be “polluting.” For example, most Indian villages had hamlets occupied by members of specific caste groups (such as the “Agraharam” or the “Brahmin” quarter of traditional Tamil villages and towns) with those arbitrarily assigned lower rungs in the social hierarchy often prohibited from entering the parts of the village occupied by those of ostensibly “higher” status and from using the same places of worship, sources of drinking water, or other public facilities as the latter. Thirdly, while modernization and urbanization have loosened the ties of caste and clan so that large Indian cities contain many neighborhoods that are, at least in theory, available to anyone who can afford to live there, there is considerable anecdotal evidence that members of India’s Muslim community (and, to some extent, other religious minorities), some linguistic groups, and those belonging to the Scheduled Caste or other backward classes (historically disadvantaged albeit “non-untouchable” groups whose members are eligible for some forms of positive discrimination in education and public sector employment) continue to face difficulties in accessing the housing of their choice. For example, there have been a series of media reports that present cases where even upwardly mobile, middle-class, professional or elite Muslims face discrimination when they look for housing, even in India’s biggest cities. And while relatively little media attention focuses on housing discrimination against SCs and OBCs – both groups whose members have historically been considered “unclean” or “inferior” by dominant caste groups and who were traditionally barred from living in the same areas as those from more powerful groups – it seems plausible that they would face similar hurdles. For example, preliminary results from a recent nationally representative household survey of over 40,000 households show that 52% of Brahmins and 24% of non-Brahmin “upper-caste” households practiced untouchability either directly or in that they were hesitant to admit a member of the Scheduled Castes into the kitchen (NCAER 2014). Indeed, Vithayathil and Singh (2012) find that segregation by caste is greater than that by class in all seven Indian megacities that they study, suggesting that urbanization is by no means a sufficient condition for the elimination of caste-related practices.
Existing Literature on Housing Discrimination in India As in other countries, spatial inequalities in the provision of public services (e.g., schools, hospitals, roads, etc.) and the signaling role of an individual’s address mean that being unable to access the housing and area of one’s choice is (quite apart from being unjust) a cause of other persistent gaps (such as in educational attainment, health status, and employment status). However, although Vithayathil and Singh (2012) argue for the relevance of audit studies to the research agenda on urban housing and patterns of residential segregation in India, there is a surprising paucity of rigorous empirical evidence on this issue, with Thorat et al. (2015), which
29
Dimensions of Discrimination: Discrimination in Housing
677
experimentally measures the extent of discrimination against non-upper-caste Hindus and Muslims in the city of Delhi and its suburbs using face-to-face and telephonic audits, being a notable exception. The present study builds on this study but differs along some important dimensions, which are discussed in detail later but which have to do both with the nature of the experiment implemented and the measures of landlord interest and effort that are captured and analyzed. Briefly, the audit methodology manipulates perceived applicant identity through the names used on web-based applications for rental housing to ensure that inadvertent and unconscious experimenter effects do not bias the results of the audit, thus sidestepping some lingering concerns with audit studies that rely on live subjects and/or telephonic conversations with landlords. Secondly, this study exploits the fact that landlords can and do call back potential tenants multiple times and that tracking the number and timing of their calls can be used to develop measures of landlord interest and effort. For instance, the study can measure not just whether landlords were more likely to call back upper-caste tenants than Muslims or Scheduled Castes at all but also whether they were more likely to call back the former more frequently, or sooner, or more persistently. The nature of the experiment thus leads to a rich set of observable and quantifiable landlord behaviors which are exploited to deepen the analysis beyond a simple analysis of callback rates. Since there is information on features of advertised houses and we can infer some limited information about landlords from their names, it is also possible to test for interactions between landlord and apartment characteristics and callbacks.
The Experiment This section describes the experimental strategy in greater detail.
Location and Sample This experiment, which was carried out entirely remotely, exploited one of India’s most popular online housing search platforms. Over the course of a roughly 2-month period in the summer of 2015, the most recently posted rental listings seeking tenants for apartments or houses in India’s second-largest city and capital, Delhi, and its two largest contiguous suburbs (Gurgaon/Gurugram in Haryana and NOIDA in Uttar Pradesh) on this website were scanned regularly. This allowed the identification of a convenience sample of 171 listings posted directly by landlords (i.e., not ones posted by an agent from a rental or property agency), taking care to avoid sending more than one set of applications to a given landlord in order to avoid arousing any suspicions that would result if a landlord received an application in two rounds of sending and noticed that the same number was now attached to a different name. The three administrative units in this study form the core of what is known as the national capital region (or NCR), which is envisaged by urban planners as eventually constituting a single commuter zone centered on the national capital, Delhi. While
678
S. Datta and V. Pathania
the entire NCR extends in a wide arc around the city and is not yet a fully realized vision, the three areas in this study are in many ways a single economic entity since they are connected by recently built expressways and the same rail-based mass transit system, the Delhi Metro. The development of the Metro and the growth of many service and manufacturing sector industries in NOIDA and Gurgaon have led to a substantial population, particularly of middle-class white-collar workers, who live in one of the three and work in another. For the purposes of this study, it is thus reasonable to think of these three administrative units as constituting a single housing market with several subregions (south, east, west, and north Delhi, Gurgaon and NOIDA to a first approximation). Although the behavior of agents is interesting in its own right, this study chose to focus on landlords for this initial study since doing so avoids any potential issues with principal-agent problems arising, for example, from agents having imperfect information about landlords’ preferences. While no explicit attempt was made at representativeness, the final counts of apartments applied to in each part of the city should be broadly indicative of rental housing flows in those areas, since the most recent landlord-posted advertisements in each region were sampled. The study purposely sought to over-sample Muslim landlords, leading to their being somewhat overrepresented in the sample.
Names and Contact Strategy As discussed above, past research suggests that both Scheduled Castes and Muslims, as well as individuals belonging to the type loosely known as other backward classes (OBCs), may face discrimination in many aspects of life in India (see Deshpande 2005). The choice was therefore made to send applications from fictitious tenants belonging to four social categories – upper-caste Hindu (UC), Muslim, Scheduled Caste (SC), and OBC. Four queries were sent to each landlord contacted. Queries were sent using the online web form (see Fig. 1 for a sample of the form in which the landlord received the web query) with one fictitious candidate from each of these four categories applying to each landlord. Care was taken not to send a flurry of applications but rather to apply sequentially with time gaps between applications. The order in which the applications were sent was randomized so that each social type was equally likely to be first, second, third, or fourth to apply to a given landlord. Two names for each type were used, with both having common male first names beginning with the letter “A” and last names that denoted caste and religion (with Muslim applicants having both first and last names that denoted their religion, since Muslims have distinctive first and last names). The Hindu first names were Anil, Arun, and Amit; the Muslim first names were Abbas and Arif. The OBC last names were Yadav and Ahir; the UC last names were Gupta and Sharma; the SC last names were Paswan and Manjhi; and the Muslim last names were Khan and Ahmed. All applicants were male, abstracting for this study from the difficulties faced by women from all social groups seeking housing in urban India. For the Scheduled Caste and
29
Dimensions of Discrimination: Discrimination in Housing
679
Fig. 1 Online rental application form
other backward class names, government lists of the castes included in these categories in Delhi and its surrounding Hindi-speaking states were used as a source for maximum signaling value. Previous qualitative work carried out by Banerjee et al. (2009) was also used; the authors field-tested possible SC, Muslim, and OBC names to identify those most widely recognized by middle-class residents of Delhi. Queries were identical except for the name and email of the applicant, which reflected the assigned social type, the associated cellphone number (which was mated to type to aid clear assignment to social type), and the time and intra-ad order at which the query was sent which, as discussed above, was randomized to avoid order effects. In all, the study reports results from landlord responses to 681 unique applicants to 171 apartments. While the design was fully blocked, there was one instance where the listing was deleted in the midst of sending the four applications. As a result, OBC queries were sent to 171 listings but UC, SC, and Muslim queries to 170. Because of this almost fully blocked design, there is no worry about balance across social categories when it comes to listing and landlord characteristics.
Data and Analysis Apart from the details on the listings (e.g., square footage, monthly rent, location, etc.), the key data come from whether, how often, and at what intervals the landlords to whom we sent queries called the fictitious tenants. Calls were received on cellphones carrying Indian SIM cards procured for this experiment. Each SIM (i.e., each number) was mated permanently to a particular type (e.g., SC or Muslim). Call log data was downloaded into Excel and coded into
680
S. Datta and V. Pathania
STATA to enable the analysis. Keeping count of the number of callbacks and number of unique callers to each type of applicant was thus trivial. Phone calls were not answered, so there is no direct way of gauging what landlords were calling to say (except in cases where landlords also sent either a text or email, which was only in a small minority of cases). In many cases landlords called back multiple times, allowing us to measure not just whether they called but how many times and how soon after the receipt of an online application expressing in an interest in their apartment landlords called the fictitious tenants. However, it is not enough for the purposes of this study to track simply the total number of unique numbers that called back different categories of tenants. There are two reasons for this. First, there is a need to link callbacks to listings in order to analyze the effects of landlord and listing characteristics. Secondly, there is a distinct possibility (borne out by the subsequent findings) that not all calls received were in response to the queries sent out as part of our study or at least not directly. To elaborate, there was a strong possibility that some calls were spam, telemarketing calls, or calls from brokers or agents whom had not been contacted but who had somehow got hold of the fictitious tenants’ numbers and were calling to offer them their services. While including such “spurious” calls in raw counts is not entirely uninformative (after all, it would be striking if spammers, too, were more interested in some social categories than others), there is a need to focus the estimates on genuine calls from landlords who had been applied to in order to make reliable inferences about landlord behavior. In addition, calls that cannot be linked to a listing cannot be used in regression estimates.
Tracing Callers The difficulty of linking callbacks to listings was exacerbated by the fact that the majority of listings do not list the landlord’s actual cellphone or landline number. Rather, the housing portal assigns them a masking number, presumably to protect landlords from spam and to protect their privacy. The portal’s practice is similar to that employed by online classified services in the United States, such as Craigslist, which provide users with a masking email so as to protect their privacy. Potential tenants (and researchers) thus mostly see only a specially assigned number, dialing which connects them to the advertiser’s actual number, which they do not see. In practice, this meant that except for a small subset of those applied to, it was not possible to immediately link a callback to a specific listing, since the callback came from the landlord’s actual number and not the one on the online listing, which is essentially a call-forwarding service. However, the use of web-based call-tracking resources such as Truecaller, publicly available information, and some supplemental calling back allowed us to solve this problem to a great extent. Callbacks were received from 118 unique numbers, whether landlines or cellphones. Of these, 22 were conclusively identified as either telemarketing calls, misdials or wrong numbers, calls from property agents who appear to have gained access to the numbers used, or (in one case) to a homeowner
29
Dimensions of Discrimination: Discrimination in Housing
681
other than one in the sample. Such spam calls (and the associated calling numbers) are dropped from further analysis since none of these categories are informative about the preferences and effort of the landlords in the sample. Of the remaining 96 unique (potentially legitimate) numbers, 89 (or 92.7%) correspond to a listing that was applied to. This confirmation is done in four ways. First, some landlords do list their actual numbers on the listing, making it trivial to link their callbacks to a listing. Second, some landlords also emailed and identified themselves (and provided a number). Third, Truecaller and other web-based search engines, together with the call logs that showed when a call from a given number was first received, enabled a large number of the remainder to be traced. Finally an intensive period of calling hitherto unidentified numbers back about 10 days after the conclusion of the experiment (and about a month since most numbers had first called) where attempts were made to determine whether these numbers were in fact landlords contacted as part of the study, and if this was indeed the case, which listing each number corresponded to, was carried out. The success of these tracing attempts makes us confident in asserting that any results seen based only on traced calls (which are the only ones that are used in regressions) are not driven by differential success in tracing callers to different categories of applicants. It is worth noting that of the seven numbers that could not either be traced to a landlord or tagged as spam, four called only once. Only three numbers that called the experimental numbers more than once (a mere 2.5% of the 118 numbers that called us at any point) are thus ones that remain untraceable and unclassifiable. Nonetheless, to ensure that the results, which are based on traced calls only, are not driven by the call patterns of these seven untraceable numbers, a bounding exercise on the callback ratios to see how they would change under assumptions chosen to go against the hypothesis was also carried out.
Descriptive Statistics: Listings, Landlords, and Applicants Listing and Landlord Characteristics Table 1 provides an overview of the features of the properties in the sample. A large majority of listings (71%) were for two- or three-bedroom apartments, 20% were for one-bedroom properties, and 9% had four bedrooms. The distribution of properties differed somewhat between the city and suburbs, with fewer one-bedroom flats in the suburbs. As expected, city flats were about one-and-a-half times more expensive per square foot (Rs. 28.5 psf compared with Rs. 18.1 psf in Gurgaon or NOIDA) and were smaller on average (at a little over 1100 sf, compared with an ample 1600+ sf in the suburbs). Those landlord characteristics that could be discerned from names with a high degree of certainty, which were religion and gender, are also reported. About 12% of the landlords applied to were Muslim, the lion’s share of these in the city. This is slightly higher than Muslims’ share in the city’s population, which is estimated at around 10%.14 About 13% of the landlord sample was female.
682
S. Datta and V. Pathania
Table 1 Listing and landlord summary statistics
Number of bedrooms:b 1 2 or 3 4+ Rent (Rs) Floor area (Sq ft) Rent/sq. ft
% Femalec % Muslimd N
Delhi city Suburbsa House characteristics
All
0.26 0.10 0.67 0.78 0.07 0.12 36100.90 29333.32 (53348.86) (24292.45) 1138.92 1616.93 (654.86) (702.55) 28.47 18.12 (20.74) (10.37) Landlord characteristics 0.14 0.10 0.16 0.03 111 60
0.20 0.71 0.09 33726.31 (45353.01) 1306.64 (707.91) 24.84 (18.45) 0.13 0.12 171
Standard errors in parenthesis Suburbs – Gurgaon and NOIDA b 1–1.5 bedrooms coded as 1 bedroom, 2–3.5 as 2–3 bedrooms c We were unable to code gender for13 of the 171 landlords (due to missing first name, e.g., only initial). The reported female % is computed over all 171 landlords d We were unable to code religion for 1 of the 171 landlords. The reported Muslim % is computed over all 171 landlords a
Table 2 Application summary statistics
UC OBC SC M
Order 2.49 (1.11) 2.50 (1.12) 2.47 (1.12) 2.51 (1.13)
Daytime 0.84 (0.37) 0.74 (0.44) 0.72 (0.45) 0.84 (0.37)
Weekday 0.59 (0.49) 0.51 (0.50) 0.51 (0.50) 0.59 (0.49)
Gap (days) 4.05 (7.59) 4.09 (7.58) 4.08 (7.61) 4.03 (7.60)
N 170 171 170 170
Standard errors in parenthesis Order is the chronological position of the applicant type within the set of 4 applications sent to each landlord Daytime: 6:01 AM–18:59 PM; Weekday: Mon–Fri Gap is number of days between date of applying and date the ad was posted
Applicant and Application Characteristics Table 2 provides an overview of application characteristics. Since the design was fully blocked, with each listing receiving one application of each type (UC, SC, OBC, and M), there can be no differences in the proportion of each type applying to
29
Dimensions of Discrimination: Discrimination in Housing
683
any kind of apartment or landlord. As discussed earlier, the position of each type within the applicant pool for each listing was randomized to avoid order effects, since it is possible that an applicant who applies before others may have an advantage. As seen in column (1), there are negligible differences in mean position of types within a listing. As columns (2) and (3) show, the staggered nature of the application procedure employed does introduce some variation when within the day or week a particular kind of applicant applied to rent a property. OBC and SC applications were more likely to have been sent over the weekend and outside of daytime hours in India. The likely effects of these differences are unclear. Perhaps the most likely effect is on time to first response: it seems reasonable that a query sent within office hours or on a weekday would be acted on immediately – although the opposite is also possible, since this is personal work for landlords and may in fact be harder to attend to during work hours or on a workday. Indeed, queries received during work hours are more likely to be missed due to inattention. Thus, while it is not clear whether and how these differences in sending time matter, some specifications control for them. These findings will be informative for future audit studies, which could seek to maximize response rates by optimizing sending times. Finally, applications to older listings should, ceteris paribus, elicit less response. Again, due to the fully blocked design, the average gap between the application date and the listing date is about the same across the four categories.
Results The patterns in the raw counts of calls, callers, and all responses (inclusive of emails and texts) are discussed below. Table 3 displays raw counts of responders and responses by applicant type. Column 1 of panel A presents the count of all unique calling numbers by type and appears to suggest that there is a large callback differential between UC applicants and all other types. However, the tracing exercise tempers these conclusions somewhat. As can be seen from the move between column 1 and column 2, part of the UC versus Others differential in column 1 is due to the fact that UC applicants were spammed more. This is, of course, interesting in its own right, particularly because some of the spam callers were in fact property agents who were not contacted for the study but who contacted the fictitious tenants independently, while others presumably wanted to offer them goods or services. This issue is not explored further here, although it is flagged for future research. For the present purposes, those numbers identified as belonging to unsolicited callers (uncontacted brokers or landlords, telemarketers, etc.) are excluded from the analysis. This substantially narrows what initially seems to be a large callback differential between UC and SC, for example. After excluding the spam callers, most of the remaining calling numbers can be matched to landlords in the sample, as discussed above. Column (3) contains
684
S. Datta and V. Pathania
Table 3 Counts of responders and responses A.
Counts of responders Unique callers Total Excl. spam (1) (2) UC 80 66 OBC 67 59 SC 63 58 M 52 40 B. Counts of responses
UC OBC SC M
All calls (1) 192 165 126 101
Excl. spam (2) 157 149 112 80
Landlords who Calleda Texted (3) (4) 56 5 50 6 55 4 35 4
Emailed (5) 2 2 3 3
Respondedb (6) 59 51 58 38
Traced to landlords: Calls Texts (3) (4) 132 5 142 6 111 4 75 4
Emails (5) 2 2 3 3
Total (6) 139 150 118 82
a
In Panel A, Column (3) differs from (2) because some numbers cannot be traced, and because some landlords called from more than one number b In Panel A, Column (6) is not the sum of (3)–(5) since some landlords both called and texted or emailed
the counts of the landlords who could be traced among the callers. Untraceable numbers are not the only source of the differences between columns 2 and 3 in panel A of Table 3. It is worth noting that 13 landlords called from 2 numbers each, explaining why the numbers in column 3 are so much smaller than those in column 2. Once column 3 is supplemented by information about landlords who texted or emailed, the total number of landlords who responded is obtained, as recorded in column (6). These numbers are the fundamental raw data for the results presented in the remainder of this section. As column (6) shows, roughly similar numbers of landlords respond to upper-castes and SC applicants. Somewhat fewer respond to OBCs and substantially fewer to Muslims. Panel B of Table 3 displays analogous counts for the number of calls, emails, and texts. Since calls were not answered, landlords (or others) could persist in calling applicants. In many cases they did, leading to a much larger number of calls than callers. The relevant numbers are, again, in column 6, which measures the total number of non-spam contacts received by each type. Here, there continues to be a big difference between UC and Muslim, with UC applicants receiving 60% more calls than Muslim applicants. We also see a substantial difference between UC and SCs here, despite there being no difference at all in the number of callers: UCs receive 18% more calls than SCs, despite being called by almost exactly the same number of landlords. Finally, it is worth noting that while OBCs receive more calls than even UCs, this is driven entirely by one landlord who called only the OBC applicant 33 times.
29
Dimensions of Discrimination: Discrimination in Housing
685
Result 1: Landlords Are Significantly Less Likely to Respond to Muslims Applicants The central result can be seen from Table 4, which presents differences in mean response rates (defined as the percentage of landlords who either called, texted, or emailed an applicant) between UC applicant and others. While the probability of a landlord responding to the UC applicant is 0.35, the corresponding probability is 0.22 for Muslim applicants. The difference of 0.13 is statistically significant at conventional levels of significance. A simple way to scale this difference is by calculating the number of landlords each type must contact in order to have a pool of 10 apartments to consider. Whereas a UC applicant needs to send out just under 29 queries to hear from 10 landlords, a Muslim applicant needs to send out nearly 45 queries to achieve the same degree of interest. Muslims must therefore expend considerably greater time and effort, including search time, to have access to a similar-sized pool of potential rental properties as upper castes. The regression counterpart of these results can be found in Table 5, which presents OLS regressions for the probability of being called back at the applicant level. The coefficients of interest are those on the dummy variables for each applicant type. Once again, we see that a Muslim candidate is about 12.4% points less likely to be contacted by a landlord than a UC candidate and that this coefficient is highly statistically significant. This result survives the addition of controls for sending patterns and is essentially unchanged by a specification that uses landlord fixed effects.
Result 2: Relative to UC Applicants, Muslim Applicants Receive Fewer Callbacks But Landlords Who Do Respond Make Similar Numbers of Attempts to Contact UC and Muslim Applicants As discussed earlier, the fact that calls were not answered means that landlords make multiple attempts to contact those applicants they are interested in pursuing. The number of calls and other forms of contact (and not just the number of callers, which has been the key measure so far) can thus be used as an additional measure of landlord interest. Table 4 presents the mean count of landlord responses per application by each applicant type. This number is 0.82 for UCs and 0.48 for Muslims. The difference of 0.34 per listing is strongly statistically significant. Put differently, a Muslim applicant would need to send about 21 expressions of interest to get 10 callbacks, whereas an UC candidate would only need to send just over 12. Muslims must expend significantly greater time and effort to elicit a comparable number of calls. However, conditional on getting a response, UC applicants receive about the same number of callbacks as Muslim applicants. Table 6 implements regressions using the count of responses rather than merely the probability of being called back. The coefficient on the Muslim dummy is negative and statistically significant, indicating that Muslims get 0.58 fewer responses.
0.05 (0.05) 0.01 (0.05) 0.13** (0.05)
Diff. versus UC –
Mean responses (all landlords)a 0.82 (1.52) 0.88 (2.90) 0.69 (1.50) 0.48 (1.12) 0.05 (0.25) 0.13 (0.16) 0.34** (0.14)
Diff. versus UC –
a
Responses include calls, emails, and texts Total traced responses divided by number of landlords contacted b Total traced responses divided by number of landlords who responded to that type
M
SC
OBC
UC
Fraction responding 0.35 (0.48) 0.3 (0.46) 0.34 (0.48) 0.22 (0.42)
Table 4 Fraction landlords responding and mean responses per landlord Mean responses (responding landlords)b 2.36 (1.74) 2.94 (4.74) 2.03 (1.96) 2.16 (1.42) 0.58 (0.70) 0.33 (0.34) 0.2 (0.32)
Diff. versus UC
0–6
0–10
0–33
Range responses (min-max) 0–9
686 S. Datta and V. Pathania
29
Dimensions of Discrimination: Discrimination in Housing
687
Table 5 Probability of response by landlord Muslim OBC SC
(1) 0.124*** [0.034] 0.049 [0.032] 0.006 [0.034]
(2) 0.124*** [0.034] 0.04 [0.032] 0.005 [0.034] 0.018 [0.030] 0.021 [0.033] 0.017 [0.034] 0.009*** [0.003] 0.081 [0.056] 0.032 [0.061] 0.014 [0.066] 0.101 [0.065] 0.381*** [0.115] 0.003** [0.001]
0.347*** [0.037] 681 0.012
0.269** [0.106] 681 0.079
Order ¼ 2a Order ¼ 3 Order ¼ 4 Gap (days)b Weekday (Mon–Fri) Daytime (6 AM–6:59 PM) Suburbs 2–3 beds 4+ beds Rent (Rs/sqft) Landlord FE Constant Observations R-squared
(3) 0.124*** [0.039] 0.039 [0.037] 0.003 [0.040] 0.018 [0.034] 0.007 [0.047] 0.028 [0.049] 0.021 [0.034] 0.129* [0.065] 0.031 [0.065]
x 0.369*** [0.120] 681 0.671
OLS regression coefficients (linear probability models); the dependant variable is a dummy for any response from the landlord Robust standard errors in brackets, clustered on landlord ***p < 0.01, **p < 0.05, *p < 0.1 a Order is the chronological position of the applicant within the set of 4 applications to a landlord b Gap in days between date application sent and date ad posted
Result 3: Differences in the Probability of Response and Count of Responses Between UC and SC/OBC Are Not Statistically Significant As the rows for OBCs and SCs in Tables 4, 5, and 6 show, there is no statistically significant evidence of discrimination against these two types as compared to UCs.
688
S. Datta and V. Pathania
Table 6 Count of responses by landlords Muslim OBC SC
(1) 0.528*** [0.146] 0.07 [0.263] 0.164 [0.142]
(2) 0.531*** [0.146] 0.175 [0.273] 0.064 [0.161] 0.268 [0.220] 0.513* [0.295] 0.503* [0.296] 0.058*** [0.017] 0.28 [0.309] 0.453 [0.280] 0.269 [0.265] 0.036 [0.472] 0.694 [0.522] 0.017** [0.008]
0.201 [0.142] 681
0.106 [0.618] 681
Order ¼ 2a Order ¼ 3 Order ¼ 4 Gap (days)b Weekday (Mon–Fri) Daytime (6 AM–6:59 PM) Suburbs 2–3 beds 4+ beds Rent (Rs/sqft) Ad fixed effect Constant Observations
(3) 0.531*** [0.147] 0.092 [0.220] 0.136 [0.134] 0.237 [0.202] 0.352 [0.277] 0.412 [0.298] 0.236 [0.182] 0.409 [0.299] 0.212 [0.281]
x 19.081*** [1.290] 681
Poisson regression coefficients (dep. variable is count of responses to an applicant) Robust standard errors in brackets, clustered on landlord ***p < 0.01, **p < 0.05, *p < 0.1 a Order is the chronological position of the applicant within the set of 4 applications to a landlord b Gap in days between date application sent and date ad posted
The probability that a landlord contacts an OBC applicant is 0.30 (see Table 4), which is lower than the 0.35 for an UC applicant. The difference of 0.05 is not, however, statistically significant at conventional levels. The corresponding difference between UC and SC is a trivial 0.01. Almost exactly as many landlords call SC applicants back as callback upper-caste applicants. This pattern is replicated in the corresponding columns for the mean number of callbacks. OBCs have a higher number of callbacks in aggregate (and therefore per listing), but the difference is not
29
Dimensions of Discrimination: Discrimination in Housing
689
statistically significant (recall that one landlord called an OBC applicant back 33 times). The point estimate of the difference between mean number of callbacks for SCs and UCs points to a disadvantage for SCs but is once again statistically indistinguishable from zero. The regressions in Table 5 confirm this finding. The coefficient on the OBC dummy for the probability of being called back is negative but not significant. The one for the SC dummy is positive but statistically and economically insignificant. There are also no significant results for OBCs and SCs when it comes to the response count (Table 6). A caveat to these three results is in order. Recall that the regressions and callback ratios are constrained to use only data from the 89 numbers (and emails and texts) which called the experimental numbers that were neither spam/unrelated to the experiment, nor untraceable. Twenty-two numbers that called one or more of the applicant categories had to be dropped from the analysis because they were not from landlords who had been contacted as part of the study. While some of these were pure spam or telemarketing, 12 of these numbers were identified as belonging either to property agents or a potential landlord offering a different apartment or house than the ones applied for during the audit. There is of course no way to map these unsolicited callers to listings or indeed to know whether these individuals received information about all four of the applicants or only a subset. But it is suggestive that while six of such agents/landlords called an UC applicant, only one of them called an SC applicant.
Result 4: There Is Suggestive Evidence that Landlords Who Respond to Both UC and Muslim (or SC) Applicants Are More Likely to Call UCs Sooner as Compared to Muslims (SC) Table 7 looks at the length of time that elapsed between the fictitious applicant sending an online query to a landlord and the first time the landlord contacted the applicant. Our results, while not statistically significant (in part due to the pairwise regressions only being able to utilize those listings where a landlord responded to both types in the pair), are suggestive. The point estimates suggest that landlords wait about 6.5 h longer before calling a Muslim applicant than they do for an uppercaste candidate. The results for SC candidates are smaller in magnitude but of the same sign. Both groups, therefore, likely would need to search longer for housing before being able to find a place to rent. Table 8 displays the findings from a related analysis that investigates whether landlords who respond to two types are more likely to respond first to one type over the other. One hypothesis is that to whom the landlord first responds is independent of the order in which the applications were received. In this case, the null (assuming no bias) is that the proportion of landlords who first respond to a given type in a pair is 0.5 (in half the cases, the landlords should respond to one type and in the other half to the other type). A more plausible hypothesis (again assuming no bias) is that
690
S. Datta and V. Pathania
Table 7 Pairwise comparison: time to first response (hours) Hours between time application sent and first response received UC v. OBC UC v. SC UC v. M OBC v. SC OBC v. M 0.386 [5.939] 2.767 0.331 [4.735] [2.400] 6.75 0.244 [7.036] [4.776] 49.021 29.782** 26.800*** 50.924*** 21.524*** [33.069] [11.327] [1.539] [8.374] [6.875] X X X X X 78 84 62 88 46 0.947 0.966 0.987 0.993 0.688
OBC SC M Constant Landlord FE Observations R-squared
SC v. M
0.977 [6.719] 31.483** [11.944] X 58 0.907
Robust standard errors in brackets, clustered on landlord OLS regressions at the applicant level Each regression only includes landlords who replied to both types in the relevant pair Controls include the rank order of the type with the application set, weekday dummy, daytime dummy, gap in days between date ad posted and date application sent, and landlord FE ***p < 0.01, **p < 0.05, *p < 0.1
Table 8 Pairwise comparison: who receives the first response?
Type 1 v 2 UC v OBC UC v SC UC v M OBC v SC OBC v M SC v M
Na 39 42 31 44 23 29
Type 1 1st appliedb 0.46 0.52 0.61 0.45 0.43 0.41
Type 1 1st responsec 0.54 0.64 0.65 0.61 0.57 0.52
Null hypothesis 1d 0.5 0.5 0.5 0.5 0.5 0.5
pvalue 0.63 0.07 0.11 0.13 0.53 0.85
Null hypothesis 2e 0.46 0.52 0.61 0.45 0.43 0.41
pvalue 0.33 0.12 0.71 0.03 0.21 0.26
a
Number of landlords who responded to both types (includes those who responded to others as well) The fraction of landlords to whom Type 1 applied to before Type 2 c The fraction of landlords that responded to Type 1 before Type 2 d The null is that landlords first respond to either type independent of the order in which the types apply e The null is that landlords first respond to the two types in the same order in which the types apply b
landlords simply respond in the same order in which they receive applications. The pairwise response patterns are tested against the two nulls in Table 8. As before, being restricted to landlords who responded to both types in a pair diminishes statistical power. Nevertheless, there is some suggestion that landlords who respond to both UCs and SCs first call UCs (almost significant at 10%).
29
Dimensions of Discrimination: Discrimination in Housing
691
Result 5: There Is Heterogeneity By Gender and Religion of Landlord and By Size and Rental Price of the Listed Property in the Likelihood of Response to the Various Applicant Types Table 9 displays the results of regressing interactions of landlord and property features with applicant type, on the likelihood of response. Although the small proportion of female or Muslim landlords points to limited statistical power, there are still some interesting patterns. First, female landlords are about 13.4% points more likely to respond to an SC applicant. Second, Muslim landlords are 22% points less likely to respond at all (the corresponding point estimate for females is about 16% but borderline insignificant at the 10% level.) Third, landlords offering one-bedroom properties are 20% points less likely to respond to Muslims. As a rule, applicants to one-bedroom properties tend to be single men or women. Since all applicants are male, this implies that the housing rental market is especially hostile to single Muslim men.
Table 9 Probability of response: interactions of applicant type with landlord/property features
Z OBC OBC*Z SC SC*Z Muslim Muslim*Z Constant Observations R-squared
Interacting characteristic of landlord/property (Z): Female Muslim One bed 0.157 0.224* 0.139 [0.095] [0.130] [0.085] 0.043 0.043 0.06 [0.034] [0.035] [0.040] 0.048 0.015 0.039 [0.088] [0.081] [0.031] 0.018 0.007 0.002 [0.038] [0.037] [0.041] 0.134* 0.091 0.015 [0.078] [0.105] [0.060] 0.140*** 0.140*** 0.105*** [0.037] [0.038] [0.039] 0.036 0.005 0.200*** [0.077] [0.008] [0.070] 0.289*** 0.276*** 0.305*** [0.083] [0.080] [0.083] 630 677 681 0.041 0.049 0.055
High price 0.146** [0.073] 0.110** [0.043] 0.044 [0.048] 0.049 [0.046] 0.069 [0.052] 0.184*** [0.046] 0.052 [0.049] 0.338*** [0.090] 681 0.044
Robust standard errors in brackets, clustered on landlord OLS regressions at the applicant level (linear probability model) Controls include the rank order of the applicant with the application set, weekday dummy, daytime dummy, gap in days between date ad posted, and date application sent ***p < 0.01, **p < 0.05, *p < 0.1
692
S. Datta and V. Pathania
Discussion, Interpretation, and Conclusion The results indicate that Muslims in particular face serious disadvantages in the search for rental housing. To get an expression of interest from 10 landlords, a UC applicant has to apply to about 29 listings, while Muslim applicant must apply to almost 45 listings – about 60% more. Although lacking statistical power, point estimates suggest discrimination on other dimensions as well. Landlords who do respond to Muslim applicants as well as UC applicants tend to respond sooner to UCs and call more frequently. The findings are unlikely to be biased by the inability to trace the identity of some callers. A simple bounding exercise confirms that even with conservative assumptions about the untraced callers, about 50% more landlords respond to UCs as compared to Muslims. Landlords offering one-bedroom properties also appear particularly reluctant to respond to Muslim applicants. Since male applicants for one-bedroom properties are commonly perceived to be single men, this suggests that single Muslim men may be finding it especially challenging to find suitable housing in Delhi and its suburbs. This study complements and extends the findings of Thorat et al. (2015). They too find significant discrimination against Muslims. But they also find evidence of discrimination against SCs, while here no statistically significant difference in the ratio of landlords who respond to UCs versus SCs is found. However, the two studies use very different audit techniques and the findings are not easily comparable. Thorat et al. (2015) employ telephonic or face-to-face audits, and landlords may be loath to overtly discriminate in such settings. Note, for instance, that 99.7% of the UC applicants in their studies received a positive response while only 33% of the UC applicants in this study received a response. In a sense, the two studies are looking at discrimination at different points in the search process. For instance, one could imagine Thorat et al. (2015) as studying the interaction between landlords and the 33% of the UC sample who were contacted by a landlord, once the contacted applicants return the landlords’ calls. Alternatively, the Thorat et al. (2015) landlord sample may not be using the online market and thus could differ from ours. Why then do their findings match ours for Muslim applicants? Interviewer effects may also come in play (see Bertrand and Mullainathan (2003)). Absent further information on the details of how caste and religion were signaled in the other study, it should be noted that there may have been differences in the extent to which landlords were able to discern the caste/religion of the potential tenant in that study and in this one. Whereas this study relies entirely on landlords picking up on the import of a set of last names, Thorat et al. (2015) had more flexibility in how affiliations were signaled. Another reason could be that in our study landlords were clearly able to identify Muslims but may not have been as certain about the caste identity of the SC names. Given that this study relies on landlords associating particular surnames with caste groups, inattentiveness toward last names could have reduced caste recognition among landlords in this study, while leaving Muslims, who had distinctive first and last names, unaffected. Online housing markets offer anonymity and flexibility, making them convenient platforms to conduct “clean” discrimination audits. Understanding
29
Dimensions of Discrimination: Discrimination in Housing
693
discrimination in such settings is also increasingly policy relevant, as more markets and transactions move online, even in urban settings in the developing world, such as the one studied here. There are important questions about how discrimination manifests itself in online settings that facilitate anonymity. In our context, landlords may be more comfortable in discriminating online than they would be in person. In turn, disadvantaged groups may evolve different coping strategies. Specialized markets or agents may emerge who assist the disadvantaged in finding housing. Alternatively, disadvantaged applicants may seek to “disguise” their identity in order to at least get the proverbial “foot through the door.” Yet another question is how much of the observed discrimination is taste based versus statistical. If the latter plays a major role, signaling strategies must adapt to the online setting. In addition, there are quirks that may be idiosyncratic to the Indian setting. For instance, dietary preferences are often cited a major reason to discriminate across tenants – many upper-caste landlords are vegetarian and prefer vegetarian tenants. These remain open questions for further research into housing discrimination in India.
References Ahmed AM, Hammarstedt M (2008) Discrimination in the rental housing market: a field experiment on the internet. J Urban Econ 64(2):362–372 Ahmed AM, Andersson L, Hammarstedt M (2010) Can discrimination in the housing market be reduced by increasing the information about the applicants? Land Econ 86(1):79–90 Aigner DJ, Cain GG (1977) Statistical theories of discrimination in labor markets. Ind Labor Relat Rev 30:175–187 Altonji JG, Blank RM (1999) Race and gender in the labor market. In: Handbook of labor economics, vol 3. North-Holland, Amsterdam, pp 3143–3259 Andersson L, Jakobsson N, Kotsadam A (2012) A field experiment of discrimination in the Norwegian housing market: gender, class, and ethnicity. Land Econ 88(2):233–240 Arrow K (1973) The theory of discrimination. In: Ashenfelter O, Rees A (eds) Discrimination in labor markets. Princeton University Press, Princeton, pp 3–33 Baldini M, Federici M (2011) Ethnic discrimination in the Italian housing market. J Hous Econ 20(1):1–14 Banerjee AV, Bertrand M, Datta S, Mullainathan S (2009) Labor market discrimination in Delhi: evidence from a field experiment. J Comp Econ 37(1):14–27 Banerjee AV, Duflo E, Ghatak M, Lafortune J (2013) Marry for what? Caste and mate selection in modern India. Am Econ J: Microecon 5(2):33–72 Becker GS (1957) The economics of discrimination. University of Chicago Press, Chicago Bertrand M, Mullainathan S (2003) Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. Am Econ Rev 94(4):991–1013 Beteille A (1969) Castes old and new. Essays in social structure and social stratification. Asia Publishing House, Bombay Beteille A (1992) The backward classes in contemporary India. Oxford University Press, Bombay Blau FD, Ferber MA (1987) Discrimination: empirical evidence from the United States. Am Econ Rev 77(2):316–320 Blau FD, Kahn LM (2006) The US gender pay gap in the 1990s: slowing convergence. Ind Labor Relat Rev 60(1):45–66 Carlsson M, Ericksson S (2013) Ethnic discrimination in the market for shared housing. Ninnaeus University Centre for Labour Market and Discrimination Studies, Kalmar
694
S. Datta and V. Pathania
Choi SJ, Ondrich J, Yinger (2005) Do rental agents discriminate against minority customers? Evidence from the 2000 Housing Discrimination Study. J Hous Econ 14(1):1–26 Danziger S, Lin AC (2000) Coping with poverty: the social contexts of neighborhood, work, and family in the African-American community. University of Michigan Press, Ann Arbor Denton NA (1999) Half empty or half full: segregation and segregated neighborhoods 30 years after the Fair Housing Act. Cityscape 4:107–122 Desai S, Dubey A (2012) Caste in 21st century India: competing narratives. Econ Polit Wkly 46(11):40 Deshpande A (2001) Caste at birth? Redefining disparity in India. Rev Dev Econ 5(1):130–144 Deshpande A (2005) Levels and patterns of consumption expenditure among scheduled castes and tribes. Technical report. Indian Institute for Dalit Studies, New Delhi Deshpande A (2011) The grammar of caste: economic discrimination in contemporary India. Oxford University Press, New Delhi Dohan D (2003) The price of poverty: money, work, and culture in the Mexican American Barrio. University of California Press, Berkeley Dumont L (1980) Homo hierarchicus: the caste system and its implications. University of Chicago Press, Chicago Ewens M, Tomlin B, Wang LC (2014) Statistical discrimination or prejudice? A large sample field experiment. Rev Econ Stat 96(1):119–134 Fix M, Struyk R (1993) Clear and convincing evidence: measurement of discrimination in America. The Urban Institute, Washington, DC Galster GC (1991) Housing discrimination and urban poverty of African-Americans. J Hous Res 2(2):87–122 Ghurye GS (1961) Caste, class and occupation. Popular Book Depot, Bombay Goldin C, Rouse C (2000) Orchestrating impartiality: the impact of “blind” auditions on female musicians. Am Econ Rev 90(4):715–741 Hanson A, Hawley Z (2011) Do landlords discriminate in the rental housing market? Evidence from an internet field experiment in U.S. cities. J Urban Econ 70(2):99–114 Hanson A, Santas M (2014) Field experiment tests for discrimination against Hispanics in the U.S. rental housing market. South Econ J 81(1):135–167 Hirsch AR (1983) Making the second ghetto: race and housing in Chicago 1940–1960. University of Chicago Press, Chicago Hogan B, Berry B (2011) Racial and ethnic biases in rental housing: an audit study of online apartment listings. City Community 10(4):351–372 Hubbard RG, Palia D, Yu W (2011) Analysis of discrimination in prime and subprime mortgage markets. Available at SSRN: https://ssrn.com/abstract=1975789 or https://doi.org/10.2139/ssrn. 1975789 Kain JF, Quigley JM (1972) Housing market discrimination, home-ownership, and savings behavior. Am Econ Rev 62:263–277 Lamb CM (2005) Housing segregation in suburban America since 1960: presidential and judicial politics. Cambridge University Press, Cambridge Massey DS, Denton NA (1993) American apartheid: segregation and the making of the underclass. Harvard University Press, Cambridge, MA Mullainathan S, Noeth S, Schoar A (2012) The market for financial advice: an audit study. Technical report. National Bureau of Economic Research, Cambridge, MA National Council of Applied Economic Research (2014) Biggest caste survey: one in four Indians admit to practicing untouchability, November 2014 Office of the Registrar General and Census Commissioner (2012) Census of India. Technical report. Ministry of Home Affairs Oh SJ, Yinger J (2015) What have we learnt from paired testing in housing markets? Cityscape 17(3):15 Pager D, Shepherd H (2008) The sociology of discrimination: racial discrimination in employment, housing, credit, and consumer markets. Annu Rev Sociol 34:181
29
Dimensions of Discrimination: Discrimination in Housing
695
Pager D, Western B, Bonikowski B (2009) Discrimination in a low-wage labor market: a field experiment. Am Sociol Rev 74(October):777–799 Rubineau B, Kang Y (2012) Bias in white: a longitudinal natural experiment measuring changes in discrimination. Manag Sci 58(4):660–677 Sachar R (2006) Sachar Committee report. Technical report. Government of India Saltman J (1979) Housing discrimination: policy research, methods and results. Ann Am Acad Polit Soc Sci 441(1):186–196 Shariff A (1995) Socio-economic and demographic differentials between Hindus and Muslims in India. Econ Polit Wkly 30(46):2947–2953 Siddique Z (2011) Evidence on caste based discrimination. Labour Econ 18:S146–S159 South SJ, Crowder KD (1998) Housing discrimination and residential mobility: impacts for blacks and whites. Popul Res Policy Rev 17(4):369–387 Srinivas MN (1957) Caste in modern India. J Asian Stud 16:529–548 Thorat S, Neuman KS (2012) Blocked by caste: economic discrimination in modern India. Oxford University Press, Oxford Thorat S, Banerjee A, Mishra VK, Rizvi F (2015) Urban rental housing market caste and religion matters in access. Econ Polit Wkly 50:26–27 Turner MA, Fix M, Struyk RJ (1991) Opportunities denied, opportunities diminished: racial discrimination in hiring. U.S. Department of Housing and Urban Development, Office of Policy Development and Research, Washington, DC Turner MA, Ross SL, Galster GC, Yinger J (2002) Discrimination in metropolitan housing markets: national results from phase 1 HDS 2000. U.S. Department of Housing and Urban Development, Office of Policy Development and Research, Washington, DC Turner MA, Santos R, Levy DK, Wissoker D, Aranda C, Pitingolo R (2013) Housing discrimination against racial and ethnic minorities 2012. U.S. Department of Housing and Urban Development, Office of Policy Development and Research, Washington, DC Vithayathil T, Singh G (2012) Spaces of discrimination. Econ Polit Wkly 47(37):60–66 Wire Staff (2015) Video: denied flat because she’s Muslim, Delhi academic asks Kejriwal for help. https://thewire.in/2015/07/24/video-denied-flat-because-shes-muslim-delhi-academic-askskejriwal-for-help-7165/ Yinger J (1995) Closed doors, opportunities lost: the continuing costs of housing discrimination. Russell Sage Foundation, New York
Part VI Affirmative Action
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
30
Thomas E. Weisskopf
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alternatives to Ethnicity-Based Preferential Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why Not Use Other Policies to Aid Disadvantaged Ethnic Communities? . . . . . . . . . . . . . . . . Why Not Base Preferential Selection Policies on Characteristics Other Than Ethnicity? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Optimal Structuring of Ethnicity-Based Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sphere of Applicability of Positive Discrimination Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Choice of Beneficiary Communities Eligible for Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Configuration of Positive Discrimination Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
700 701 701 703 705 705 708 709 715 716
Abstract
This chapter discusses whether or not policies of positive discrimination, such as reservations, should be deployed in order to reduce the social and economic marginalization of disadvantaged racial, caste, or other ethnic communities. PD policies giving preferences to members of such disadvantaged communities are likely to have significant negative as well as positive consequences. Are there alternative policies – such as class-based preferences – that could deliver similar benefits without significant costs? After reviewing the diversity of ways in which a race-, caste-, or ethnicity-based PD policy can be structured, this chapter concludes that the case for deploying such PD policies is strong if the policies are carefully designed to maximize the benefits and minimize the costs.
T. E. Weisskopf (*) University of Michigan, Ann Arbor, MI, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_44
699
700
T. E. Weisskopf
Keywords
Discrimination · Reservation(s) · Preference(s) · Beneficiary(ies) · Community(ies) · Ethnic(ity) · Selection · Policy(ies) · Characteristic(s) · Diversity · Disadvantage(d) · Identifiability
Introduction In every country in South Asia, there are racial, caste, or other ethnic communities whose average welfare – by many social and economic indicators – is significantly below that of the population as a whole. In several of these countries, policies of positive discrimination (PD) have been introduced in an effort to reduce historically persistent lags in the social and economic welfare of relatively poor communities. By “positive discrimination,” I mean preferential selection of members of underrepresented ethnic communities to desirable positions in society. Where such policies have been implemented, they have most often proven highly controversial. Indeed, they have given rise not only to heated debate but also at times to major public demonstrations, pitting actual or prospective beneficiaries of PD policies against those who are – or believe they are – disadvantaged by the policies. Should PD policies be deployed to combat the marginalization of members of ethnic communities that find themselves significantly underrepresented in society’s upper strata? Proponents of such policies have suggested many ways in which they may help to bring about a more equal and/or a more vital society: – PD policies compensate for explicit negative discrimination as well as implicit biases that would otherwise unfairly penalize members of underrepresented ethnic communities. – PD policies redistribute resources and opportunities from relatively well-off ethnic communities to relatively poorly off communities. – PD policies improve the motivation of members of underrepresented ethnic communities to aspire to and work toward more desirable positions in society. – PD policies lead to better performance by institutions or organizations where greater diversity of personnel in key positions (resulting from greater representation of underrepresented ethnic communities) contributes to productive efficiency. – PD policies enable members of underrepresented ethnic communities to gain better access to social capital – i.e., useful contacts and networks that improve one’s career opportunities – which are currently available mainly to members of other communities. – PD policies serve to integrate underrepresented ethnic communities into society’s elite, thereby fostering a more legitimate and vital democratic order. Critics of PD policies, on the other hand, have pointed to many reasons why such policies may prove undesirable and possibly counterproductive:
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
701
– PD policies clash with the liberal principle of evaluating people on the basis of their individual abilities rather than on ascribed group characteristics. – PD policies benefit mainly the best-off members of underrepresented communities (the “creamy layer”) rather than those members who most need help. – PD policies arbitrarily penalize those members of nonbeneficiary communities who are displaced by preferential access for members of beneficiary communities. – PD policies lead to poorer performance on the part of beneficiaries than displaced members of non-beneficiary communities. – PD policies lead to the successes of beneficiary communities being attributed to PD policies rather than their own abilities. – PD policies, as a consequence of all of the above, have a strong potential for exacerbating intercommunity tensions/divisions. Given these potential drawbacks, one is led to ask: aren’t there better alternatives to PD policies that can help achieve the prospective benefits, without giving rise to as many possible costs?
Alternatives to Ethnicity-Based Preferential Selection Why Not Use Other Policies to Aid Disadvantaged Ethnic Communities? One obvious set of alternatives to PD policies, based on preferential selection of members of underrepresented ethnic communities, consists of policies that directly protect members of such communities from negative discrimination. The second set of alternatives consists of policies in which resources are redistributed to members of such communities. In practice, both these types of policies often accompany policies of preferential selection. However, given the potential negative consequences of PD policies, one might ask if it is not better to pursue only antidiscrimination policies or only resource-transfer programs as a means of aiding such communities. The United States’ experience with “affirmative action” policies in the 1960s displays the limitations of a purely antidiscrimination strategy. The initial thrust of these policies was to neutralize negative discrimination through new antidiscrimination measures. The landmark 1964 Civil Rights Act was based on the equal protection clause of the US Constitution; it formally outlawed discriminatory practices in almost all public spheres of American life. This was followed by a series of executive orders to promote equal opportunity in employment and education. It was expected that the assertion of formal, legal equality of all citizens, the removal of overtly discriminatory barriers, and a much wider diffusion of relevant information would lead to significant increases in the representation of African Americans in desirable jobs and educational institutions. In this context, the term “affirmative action” was used to describe active outreach efforts made by organizations to assure
702
T. E. Weisskopf
equal access to jobs and educational opportunities for underrepresented communities and the more advantaged citizens – without any regard for ethnic identity. It soon became apparent, however, that affirmative action of this kind would not have an immediate and significant impact on the numbers of African Americans in most professions and educational institutions in which they had always been greatly underrepresented. The US labor department began to measure progress in ending discrimination in terms of quantitative increases in the percentage of African Americans in various fields – this was done on the premise that a major cause of underrepresentation was the subtler form of discrimination against African Americans. Public organizations, as well as nonprofit institutions and private companies supported by or otherwise linked to the federal government, came under pressure to increase the representation of African Americans in order to demonstrate that they were operating in a nondiscriminatory manner. Such pressure increased as courts began to accept statistical information on the low percentage of African Americans employed, in relation to their proportion of the population presumably qualified for a position, as evidence of racial discrimination. By the late 1960s, these developments had led many government agencies, as well as some private organizations, to discriminate in favor of African American candidates for jobs or contracts. Moreover, many educational institutions – both public and private – and some nonprofit organizations sought to increase the representation of African Americans by applying preferences in selection. In this context, the term “affirmative action” took on a stronger new meaning – which it has retained ever since – of positive discrimination via preferential selection in favor of underrepresented communities. Why not, then, pursue policies involving resource transfers instead of preferential selection? Such resource transfers could be directed to members of underrepresented communities in a way that helps them develop the skills needed to qualify for better jobs or to acquire the capital needed to launch business enterprises, thus contributing directly to economic development as well as social uplift. Preferential selection policies are problematic because, among other things, they concentrate losses on applicants displaced by the admission of beneficiaries; this leads, not only the displaced applicants but also other rejected applicants, to attribute their rejection to the unfairness of preferential policies. Developmental aid policies for underrepresented communities, on the other hand, involve costs that are spread much more broadly among the general public – and, if the revenue-raising system is reasonably progressive, the cost burden falls largely on those who can most afford to pay. However, these policies tend to be significantly more expensive than preferential selection policies, since they involve substantial resource transfers from government and private institutions to beneficiaries – such organizations have limited budgets to help members of particular communities and many demands on their largesse. For these organizations, it is far less costly to mandate preferential selection – and thereby to deflect most of the costs of the aid away from the organization itself – than it is to transfer resources. In any case, preferential policies and resourceredistributional policies are not mutually exclusive ways of reducing group disparities. Firstly, a significant element of preference is involved in directing resource transfers to members of a specific group. Secondly, as I will argue below, a certain
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
703
amount of resource transfer is likely to be very helpful, if not indispensable, to the success of preferential policies.
Why Not Base Preferential Selection Policies on Characteristics Other Than Ethnicity? Critics often allude that preferential policies tend to favor the best-off members of underrepresented beneficiary communities rather than the worst-off, since it is the former who are best placed to qualify for selection to top educational institutions, influential jobs, etc. To assure that PD policies help those who are most disadvantaged, many observers propose that preferential selection should be geared not to members of underrepresented ethnic communities but to individuals of low socioeconomic status (SES). Class-based preferential selection policies, as compared with ethnicity-based preferential selection policies, would lead to the selection of (1) many more poorly off and far fewer better-off applicants from underrepresented ethnic communities and (2) many poorly off applicants from well-represented communities. This would result in much greater socioeconomic diversity and much less ethnic diversity. There are several reasons for which this would be an undesirable outcome. Firstly, for any given number of preferential selection beneficiaries, the number of beneficiaries from underrepresented ethnic communities would obviously be smaller, except in the (unlikely) event that there were relatively few members of other ethnic communities in the lower socioeconomic strata of society. Yet, one of the central purposes of PD policies is to reduce disparities specifically between ethnic communities, independently of any reduction of socioeconomic disparities. This is because disparities among ethnic communities are especially likely to be attributed to past or present mistreatment of underrepresented communities by other communities – a notable source of continuing tension in heterogeneous societies. Moreover, members of ethnic communities are particularly vulnerable to damaging negative stereotyping, to stigmatization, and to residential segregation. These kinds of phenomena tend to adversely affect even the best-off members of underrepresented ethnic communities. Secondly, class-based preferential selection is likely to lead to significantly poorer performance on the part of its beneficiaries than is ethnicity-based preferential selection. Many of the benefits of PD policies depend on successful performance by its beneficiaries; likewise, many of the costs are correlated with rates of beneficiary failure. A good rate of beneficiary success is therefore essential to the overall success of a PD policy. In all societies, however, there is a high correlation between SES and preparedness to function well in educational institutions and job settings. Class-based preferential selection substitutes low-SES members of all ethnic communities for higher-SES members of underrepresented communities – this brings into demanding environments many more people who find it challenging to perform well, as compared to beneficiaries of ethnicity-based preferences.
704
T. E. Weisskopf
This means that a class-based PD policy will either simply result in a much higher rate of failure or require many more resources to assist challenged beneficiaries in being successful. Additionally, low-SES beneficiaries (and those from low-SES neighborhoods) also need greater financial support as compared to higher-SES beneficiaries for simply persisting in educational institutions or job settings. It follows that, even if other things (including success rates) are equal, class-based preferences will be costlier than ethnicity-based preferences and thus be implemented only on a smaller scale. This suggests that socioeconomic disadvantage cannot and should not replace the identity of underrepresented ethnic communities as a basis for identifying appropriate beneficiaries of social policy. Also, the implementation of policies benefitting underrepresented communities does not preclude the simultaneous implementation of policies targeting the lower socioeconomic classes. To the contrary, in multicultural societies, it is generally desirable to adopt policies that target both –ethnicitygroup inequalities and socioeconomic class inequalities. Some observers have proposed that while PD policies should remain focused on underrepresented ethnic communities, a means test should be introduced for channeling the benefits of preferential selection to relatively low-SES members of such communities – such as the “creamy layer” test designed to exclude members of affluent OBC families from public sector job reservations; see Sivaramayya (1996). On its face, this proposal is attractive because it addresses the criticism that PD policies primarily benefit the better-off members of beneficiary communities. The wisdom of such a test, however, is highly questionable, as it is subject to some of the same drawbacks as outright class-based preferences. First, it would remove precisely the best prepared members of underrepresented ethnic communities from becoming PD beneficiaries. Second, it would remove those members from becoming PD beneficiaries who are likely to require less resources for persisting and succeeding in their new environments. An additional weakness of class-based as compared with ethnicity-based preferential selection is that SES is more difficult to ascertain than ethnic identity. SES can be and has been measured in a variety of ways; indicators such as income, wealth, and educational achievement are all arguably relevant. However, one must not only decide which indicators to take into account and with what relative weights; one must also decide whether these indicators should be measured for the individual in question, for their parents, or for their grandparents. Having made these decisions, one is then faced with the daunting task of accurately measuring the extent of people’s income and wealth. The determination of a person’s ethnicity involves only one categorical variable that is generally unambiguous and applicable to the whole family. For some of the same reasons, moreover, SES is easier to alter – or to falsify – than ethnic identity. In recognition of these weaknesses of class-based preferential policies, some critics of ethnicity-based preferential policies have suggested the alternative of geography-based PD policies – providing preferences in selection to individuals living in disadvantaged localities (neighborhoods, villages, towns, or regions) or to
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
705
individuals attending disadvantaged schools. The most appealing feature of such plans is that, while steering benefits to groups consisting disproportionately of the underrepresented communities to whom ethnicity-based PD policies are oriented, they eschew direct reference to race, caste, or any other kind of ethnic characteristic. Geography-based PD policies are therefore less likely than ethnicity-based PD policies to exacerbate intercommunity tensions. On closer examination, however, this alternative also has some very significant weaknesses – some of them much the same as class-based PD policies. First of all, policies favoring people with particular geographic characteristics can obviously not do as good a job of increasing opportunities for underrepresented ethnic communities, or low-SES communities, as policies directly targeting disadvantaged communities. Second, the direct beneficiaries of geography-based preferences are likely to be less well prepared and, hence, less capable of performing well in their new and more challenging environments than prospective beneficiaries of ethnicity-based preference policy – this is because most of them come from relatively impoverished families who generally cannot provide their children with high-quality education. As a result, the rate of failure of the beneficiaries of geography-based PD policies will be relatively high. Finally, geographic location and/or school attendance is relatively easy to alter, as compared to ethnicity. Thus, geography-based preferences are likely to encourage people who are in no way disadvantaged to change their geographic location in order to become eligible for preferential selection – thus undermining the objective of PD policies to increase opportunities for members of underrepresented groups. In sum, class- or geography-based preferences are inadequate substitutes for ethnicity-based preferential selection. Concerns about socioeconomic inequalities can best be addressed by policies oriented directly to the disadvantaged, rather than by changing the scope of eligibility for positive discrimination. Thus, there is a strong case for adopting simultaneously ethnicity-based policies of preferential selection and class-based policies of resource transfer.
Optimal Structuring of Ethnicity-Based Preferences Sphere of Applicability of Positive Discrimination Policies The primary spheres in which ethnicity-based policies of positive discrimination have been applied in practice are political representation (e.g., reserved parliamentary seats), educational institutions (mainly at the level of higher education), employment (more often in public organizations than in private firms), and government contracting with private firms. Of these four spheres, political representation is the sphere for which the “prima facie” case for positive discrimination is strongest. As Andre Beteille (1981) has emphasized, a central function of politicians is to represent their constituencies, and a political system operates more effectively if all relevant constituencies are indeed represented in political bodies. Thus, the potentially worrisome trade-off between adequate representation of disadvantaged groups,
706
T. E. Weisskopf
on the one hand, and efficient performance, on the other, is greatly muted in the political arena. Turning to the other three primary spheres of PD policies, the trade-off between representation and performance can potentially be quite salient. In this context, there are a number of reasons which make it more desirable to focus PD policies on access to educational opportunities than on access to jobs or government contracts. Firstly, the difficulties that PD beneficiaries face in achieving good performance in a challenging setting tend to be less daunting when the beneficiaries are younger and have not yet suffered as much from the cumulative effects of group stigmatization and socioeconomic disadvantage. The earlier in the life cycle that PD beneficiaries are asked to catch up with their less disadvantaged peers, the more important are their potential capabilities as opposed to their realized capabilities, and the greater their chances of success are likely to be. Secondly, the negative consequences of poor performance in educational institutions are likely to be confined largely to the individual PD beneficiary, whereas in job or contracting settings, poor performance can hurt other parties as well. To the extent that PD policies generate a conflict between justice/equity and merit/efficiency, this conflict is likely to be smaller in magnitude and easier to manage in an educational setting than in job or contracting environments. Indeed, PD policies in higher education provide a way of defusing the tension between representation and expertise in responsible positions, because attending more selective universities enables members of disadvantaged communities to gain the skills necessary to qualify for – and fulfill the responsibilities of – such high-status positions. In this way, the advantages flowing from a more integrated professional elite can be gained without much loss in the competence of those filling the positions. Another reason to prefer PD policies in the educational sphere to policies in the employment sphere is that passions and resentments about the actual and/or perceived unfairness of the selection processes involved are likely to run higher in the latter than the former. This is because the stakes for the individuals involved – whether they are preferred or displaced by PD – tend to be greater when it comes to a job as compared to admission to an educational institution (unless the latter serves as a perfectly straightforward channel to a good job). Thus, the costs of positive discrimination associated with intergroup tension and conflict may be less serious in the case of educational admissions. Everyone can agree that it would be desirable to find ways to enable underrepresented community members to increase their numbers in high-quality educational institutions and in esteemed occupations by becoming more successful in a fully competitive selection process, rather than by receiving PD preferences. This has led some observers to propose that policies of “preferential positive discrimination” be replaced by policies of “developmental positive discrimination.” Developmental PD would encompass programs to enable applicants from disadvantaged communities to perform better on standardized tests, whether for admission to educational institutions or for selection to jobs, and, more generally, programs to improve the physical and social environments of the relatively poor regions in which members of such communities typically live, so as to make these environments more conducive to
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
707
their advancement. Certainly, developmental PD, as defined in this way, is much to be desired. It cannot, however, be considered an alternative to preferential PD. On the one hand, developmental PD involves the channeling of resources disproportionately – if not exclusively – to members of certain underrepresented communities, so it entails a significant element of preference. On the other hand, preferential PD often enables its beneficiaries to attain positions where they are likely to be better able to develop their skills and abilities, so it becomes partly developmental in character. Indeed, preferential PD in the sphere of higher education serves as a form of developmental PD with respect to the sphere of employment, for it enables community members to gain skills that make them more competitive in the job market. Furthermore, a developmental program typically calls for a greater commitment of resources than does a purely preferential PD policy. This is why it has proven much more difficult to get developmental programs going on an adequate scale than to enact preferential policies. In general, it is desirable to ensure that PD policies serve a developmental function, not merely one of preferential selection. Except in circumstances where the selection process identifies under-credentialed applicants who are actually better prepared to do the work, the PD beneficiary will face the challenge of catching up with peers selected without any preference. The chances of success will depend significantly on the extent to which the new position that PD has enabled the beneficiary to attain has a development component – one that will help the beneficiary to develop capabilities and overcome initial disadvantage. This is likely to be the case when PD is applied to the selection of applicants for an educational program than for employment. But within the latter sphere, PD policies can be expected to work better, the more significant the role of on-the-job training is. Another relevant consideration in the choice of sphere for positive discrimination is the degree to which the institution involved is elite as opposed to run-of-the-mill in character. Since integration of the societal elite is a principal objective of PD policies, the net benefits from a PD policy are likely to be especially high in the case of elite institutions – whether these are schools from which graduates move directly into high-status professions or major enterprises with a high degree of economic or political power. Moreover, such institutions are bound to be relatively well-endowed, so they will be better able to mount programs that increase the prospects of PD beneficiary success. A final important question with respect to positive discrimination is whether or not there is reason to differentiate between the public and the private sector in applying PD policies. In principle, there is no reason to expect a difference in the benefits and costs of positive discrimination (ceteris paribus), no matter which sector the institutions, organizations, or enterprises involved happen to be in. The same arguments and considerations affecting the net benefits of a PD policy apply irrespective of the sector. Thus, the case for focusing PD policies on educational institutions in preference to job settings, on programs with a developmental emphasis, and on relatively elite institutions remains strong – whether the institutions are public or private.
708
T. E. Weisskopf
In practice, one must recognize that society-specific historical, constitutional, and legal traditions will strongly affect whether PD policies emanate from governmental or nongovernmental decision-makers. Such factors will also influence also how well PD policies can be implemented and enforced and what role government should be expected to play in this regard. One can therefore adequately address the issue of public and/or private sector PD policies only in the context of particular societies and historical situations.
Choice of Beneficiary Communities Eligible for Preferences The desirability of pursuing a PD policy will vary from one underrepresented ethnic community to another. Indeed, the identification of communities to be favored by preferential selection processes is the most critical decision to be made in adopting policies of positive discrimination. Since integration of the societal elite is arguably the most important objective of PD policies, the eligibility of any particular ethnic community should be positively related to the degree to which the community is underrepresented in society’s most desirable positions. This is likely to be highly correlated with the extent to which the average socioeconomic status of community members is below the societal average. But relatively low average socioeconomic status does not alone justify positive discrimination on the basis of ethnic identity. An ethnicity-based PD policy is warranted to the extent that members of a disadvantaged community have been – and continue to be – mistreated and stigmatized on the basis of their group identity. In principle, one should try to estimate the overall net benefits associated with each distinct community for whom a case for PD seems plausible on the above grounds, in order to determine eligibility. In practice, it would make sense to limit such an effort to communities whose members have a strong sense of common ethnic identity, so that many will be able to share – materially as well as psychically – in the benefits that will go initially to a relatively advantaged few. Unless there is considerable solidarity among community members, the immediate beneficiaries of PD policies will not be inclined to use their enhanced positions in ways that expand opportunities for many less favored members of the community, nor will the former serve as compelling role models for the latter. Deciding eligibility for PD benefits is likely to be an especially difficult task when a prospectively eligible community contains distinct subgroups, for this gives rise to the question: should all, or only some of them, be eligible? A related question is whether different communities eligible for PD policies in any given country should be accorded the same amount and kind of benefits. The optimal magnitude of preference is likely to differ from one community to another, so one should, in principle, establish different preference magnitudes for different eligible communities. I will discuss the issue of multiple preference magnitudes further on. Determining on a rational basis which underrepresented ethnic communities should be eligible for positive discrimination is, to be sure, a great challenge. In principle, it should be done on the basis of meticulous and dispassionate social
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
709
scientific research into the extent of net benefits to potentially eligible communities. Given the political stakes in the outcome of such research, it is bound to be very difficult to keep politics from interfering with the process; see, for example, Radhakrishnan (1996). In advocating reconsideration of the eligibility of communities currently favored by PD, one must recognize that there is an asymmetry between the adoption and removal of a PD policy. The removal of PD eligibility from a community is a political blow of considerably greater magnitude than the failure to grant such eligibility in the first place. Thus, the removal of eligibility from some communities or community subgroups, while others continue to be eligible, must be limited to situations where the case for doing so is strongly supported by widely respected research findings. Given all the difficulties and complications associated with the determination of which underrepresented ethnic communities should be eligible for PD, there is a very good case for simplicity in several respects: (1) simplicity in limiting the number of eligible communities to the most compelling cases; (2) simplicity in defining communities by simple criteria that clearly distinguish those subgroups and individuals who are members of the community from those who are not; and (3) simplicity in grouping together into one community subgroups that might have an argument for being considered separate communities – unless that argument is extremely compelling.
Configuration of Positive Discrimination Policies As noted earlier, the overall net benefits of positive discrimination policies are significantly and positively correlated with the (average) quality of performance by beneficiaries in the institutions and organizations to which they gain preferential access. One can therefore view the detailed configuration of a PD policy as confronting a type I/type II error problem. In selecting formally underqualified but potentially successful applicants from underrepresented beneficiary communities, one hopes to minimize errors of two types: Type I: failure to select someone who would have been successful (a true null hypothesis is incorrectly rejected) Type II: selecting someone who will be unsuccessful (a false null hypothesis fails to be rejected) Among the key elements of choice in configuring PD policies are the magnitude of the preference given to beneficiary groups, the sensitivity of the selection process, the identifiability of the PD beneficiaries, and the extent of support for PD beneficiaries. After addressing, briefly, the issue of quotas versus preferences, I will examine each of the above elements to determine how a PD policy can be best designed to minimize errors of type I and II and thus to maximize net benefits.
710
T. E. Weisskopf
Quotas Versus Preferences Whether a PD policy is based on a system of quotas or a system of preferences is often assumed to be its most important defining characteristic. A quota system involves the establishment of a certain number of positions reserved for applicants from beneficiary communities, thus dividing the overall competition for selection into two or more separate competitions: one for members of each beneficiary community and the other open to all other applicants. Whatever the criteria for ranking applicants, they are applied separately to each group of applicants; and the highest-ranking applicants in each group are selected until the available positions are filled. The principal alternative to such a quota system is one in which there is a single general competition for selection, but PD beneficiary communities are given preferences – more precisely, preferential boosts – in the form of more favorable consideration in the determination of the ranking of candidates. In the case of quantitative selection procedures, in which an applicant’s qualifications are summarized in an overall point score to determine his/her position in the rank order, the preferential boost could take the form of a certain number of additional points credited to a beneficiary community’s applicants. In the case of qualitative selection procedures, in which a variety of applicant qualifications are taken into account but not formally aggregated into a single overall point score, the preferential boost would take a less precise form – for example, applications from members of relevant communities could be viewed in a rosier light or given extra credit for signs of unrealized potential. The difference between a quota system and a preferential-boost system is not as great as it may first appear. Corresponding to a quota system that selects any given number of beneficiary community applicants for a particular position, there is bound to be some amount of preferential boost that leads to the same outcome. In the case of a selection process in which applicants’ qualifications are summarized in a single point score, the amount of preferential boost that would do so is the number of points needed to bring the marginal beneficiary community applicant’s score up to the level that would make him/her the last applicant admitted in the general competition. In the case of a qualitative selection process, the amount of the preferential boost would be more difficult to calculate; but the same principle holds. In practice, quota systems are most often constrained by specification of minimum conventional qualifications (e.g., a minimum qualifying score), below which beneficiary community applicants will be rejected, even if their quota is not filled. Whenever such a requirement serves to keep the number of selected applicants below or equal to the quota, this kind of constrained-quota system has the same effect as a preferential-boost system, in which the size of the preferential boost is equal to the gap between the minimum conventional qualifications required of a successful applicant in the limited competition and the conventional qualifications of the last applicant admitted in the general competition. An unconstrained-quota system, by focusing on a target number of beneficiary community applicants to be selected, ignores problems likely to arise if the conventional qualifications gap between PD beneficiaries and other applicants selected is
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
711
substantial. To try to fill quotas of reserved seats with applicants from disadvantaged communities, irrespective of their conventional qualifications, is to invite poor performance by those selectees whose conventional qualifications are well below the qualifications required of applicants in a general competition. This is especially likely to be the case if the size of the quota is determined by the community fraction of the overall population, as opposed to that of a smaller population consisting of plausible recruits to the positions in question. An unconstrained-quota system thus makes little sense. Since a constrained-quota system does not differ in its essence from a preferential-boost system, the choice between the two is immaterial for the success or failure of a PD policy. Several other characteristics of the selection system are likely to have much more significant effects on the consequences of the policy.
Magnitude of the Preference An obviously important dimension of any PD policy is the extent to which it intervenes into and perturbs the allocation of available positions between beneficiaries and other applicants, as compared with the allocation that would obtain in the absence of the policy. A useful way to measure the impact of a PD policy focuses on the magnitude of the preference extended to PD beneficiaries. This magnitude is easy to measure for selection procedures that are quantitative, in the sense that applicants’ qualifications are summarized in an overall point score. In the case of a quantitative preferential-boost system, it is simply the number of additional points granted to beneficiary community members. In the case of a quantitative constrained-quota system, it is the difference between the point score of the marginal applicant selected in the general competition and that of the marginal applicant selected in the quota competition. When a PD selection procedure is qualitative, involving consideration of a variety of qualifications that are not scored and aggregated, the average amount of preference extended to PD beneficiaries is implicit in the process and much harder – and perhaps impossible – to measure. The preference magnitude is an especially important characteristic of a PD policy because of its impact on type I and II errors and hence on the net benefits of the policy. The smaller the magnitude, the fewer will be the number of beneficiary community applicants selected (thus more type I errors), but the greater will be the proportion of them who are likely to perform as well as their non-beneficiary peers (thus fewer type II errors). The larger the magnitude, the greater will be the number of PD beneficiaries (thus fewer type I errors), but the proportion who perform as well as their peers will be smaller (thus more type II errors). There is bound to be some level of the preference magnitude at which the costs of marginal unsuccessful candidates selected will start to exceed the benefits of marginal successful ones. Where the optimal preference magnitude lies depends on the relative frequency of each type of error at each level and on the costs associated with each type of error; see Chatterjee (1983). In principle, with full information about the consequences of different preference magnitudes, and with a common standard for evaluating the net benefits from a good and a bad selection decision, one could estimate both the number of PD beneficiaries and the overall net benefits associated with each preference magnitude. For any given underrepresented community, the net benefits (ceteris paribus) from PD would
712
T. E. Weisskopf
presumably rise initially, as the magnitude of the preference was raised from zero, because at low magnitudes PD beneficiaries could be expected to perform almost as well as other applicants. After a certain point, however, the additional net benefits from a higher preference magnitude would turn negative because, at ever higher magnitudes, an ever-smaller proportion of additional PD beneficiaries selected would be able to perform well. The magnitude of the preference at that turning point could therefore be identified as the optimal one, which maximizes the expected net benefits from a PD policy favoring the given community. PD policies would then be judged worthy of adoption in favor of each community for which the maximum expected net benefits are in fact positive, with the magnitude of the PD preference set at the optimal level for each eligible community. In reality, of course, decision-makers will never have access to sufficient information to determine optimal preference magnitudes and the corresponding expected net benefits in such a precise manner, for each community under consideration. Instead, they will have to mix available information with educated guesses and rely on their best judgment to determine which communities should be made PD eligible and at what level the preference magnitude for each eligible community should be set. In the interests of simplicity (as advocated earlier), the preference magnitudes should be set at the same level unless there is a very compelling reason to do otherwise for a particular community.
Sensitivity of the Selection Process PD policies vary greatly with respect to the sensitivity – as opposed to the rigidity – of the process, whereby applicants are preferentially selected. The most rigid, mechanical type of process involves a quantitative procedure in which all applicants take some kind of standardized test and are ranked simply by scores on that test. In a quota system, the test rankings are used to select applicants in separate competitions; in a preferential-boost system, a certain number of points are added to beneficiary applicant scores before all candidates are ranked for selection in a single competition. A somewhat less rigid selection process would take account of several different qualification criteria, not just a test score. This kind of process would assign scores on each criterion to every applicant and aggregate every applicant’s scores into a composite quantitative index for purposes of ranking, prior to the selection of beneficiary applicants via a quota or preferential-boost system. A selection process becomes progressively more sensitive when there is greater variety and number of criteria involved in ranking applicants. Even more important, a selection process becomes more sensitive the more the process of evaluating an applicant’s standing with respect to relevant criteria is a qualitative rather than a quantitative one, involving considered judgment by selection personnel, rather than mechanically determined scores fed into a composite index. A highly sensitive and nuanced PD selection process would not only include qualitative evaluation of the extent to which a beneficiary community applicant satisfies various relevant criteria but also treat disadvantaged community status as a signal to look especially hard for evidence of additional applicant characteristics suggesting a strong potential for good performance.
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
713
It is clearly desirable to make the PD selection processes as sensitive and nuanced as possible, so as to maximize the potential for determining which community applicants – who are underqualified in terms of conventional indicators – have the greatest potential to be successful and thereby to minimize both type I and type II errors. This implies, first of all, that one should not try to fill quotas of reserved seats with underrepresented community applicants, irrespective of their conventional qualifications; rather, one should give each such applicant the same magnitude of preference in competition with noncommunity applicants. This implies, further, that one should make use of a variety of relevant qualification criteria and that evaluations of the extent to which an individual applicant is qualified should be based as much as possible on the exercise of qualitative judgment by selection decisionmakers rather than on the input scores into a quantitative index. Sensitive selection processes are, to be sure, harder and costlier to administer than rigid ones, since they require that more information of various kinds be gathered from applicants and that more people be employed to implement the selection process. The extent to which a PD selection process can be made sensitive, therefore, depends on the availability of resources to finance the process; and the costs of raising such resources must be weighed against the benefits expected from a more sensitive process. There is a difficult decision to be made about how far to invest limited resources into processes of selection; but it is a safe guess that such investment is usually not carried as far as it should be, since some of the benefits of better selection will not accrue to the organization doing the selection. Government subsidies may therefore be warranted to achieve the desired end.
Identifiability of the Beneficiaries The existence of a PD policy means, of course, that some individuals selected in a competition will owe their selection to that policy. Depending on how the PD policy is implemented, however, the identity of the beneficiaries – as well as their total number – may or may not be known to anyone other than those administering the selection process (and sometimes not even to them). This is by no means a trivial matter, because both the self-esteem of a beneficiary and his/her treatment by others are likely to be adversely affected by knowledge that he/she would not have been selected in the absence of the PD policy. Thus, one characteristic of a PD policy relevant for its consequences is the extent to which beneficiaries are made identifiable. One might at first presume that a quota system necessarily makes PD beneficiaries identifiable, and a preferential-boost system does not. The fact that a quota system channels members of certain communities into separate competitions for reserved positions, however, does not require that those chosen in each competition be separately identified. And even when the number of positions reserved for a particular community is known, it does not mean that an equivalent number of community members owe their positions to PD. It is always possible that some of the successful applicants did not need PD to be selected and/or that some of the reserved positions went unfilled. Conversely, the fact that a PD policy is based upon a preferential boost does not always preclude the selection administrators from revealing who required such a boost in order to be selected. Under a preferential-boost
714
T. E. Weisskopf
system, it is not always clear precisely which selected beneficiaries owe their selection to the PD policy, so identification of PD beneficiaries is less likely to occur under such a system than under a quota system. But whether or not it occurs under any selection process is really a policy variable. It is clearly best that a PD policy not render its beneficiaries easily identifiable as such – and hence stigmatized – by others; this reduces the prospect of type II errors. One of the most common ways in which PD beneficiaries are, in practice, rendered identifiable is by including them in remedial and developmental programs designed to facilitate their adjustment to the demands of the competitive environment to which they have gained access. The objective of such programs is certainly a worthy one (as I discuss just below). But to avoid the negative fallout of making PD beneficiaries easily identifiable, it is desirable – wherever possible – not to confine these programs exclusively to PD beneficiaries and to include non-PD beneficiaries who are likely to benefit from them as well.
Extent of Support for Underprepared Beneficiaries Whether a PD beneficiary is able to meet the challenges of the position to which PD has provided access is likely to depend significantly and positively on the extent to which support is made available after the person is selected, thus reducing type II errors. Policies of positive discrimination may be confined to the selection process itself (costing no more than the administrative expenses of that process) or involve various forms of subsequent support for PD beneficiaries (costing a good deal more). There are several kinds of support at the organizational or institutional level that can be helpful for PD beneficiaries, whether in an educational or a workplace setting. These include both human resources, such as favorable attitudes on the part of supervisors and mentoring on the part of colleagues, and financial resources made available for programs and activities that help PD beneficiaries adjust to their new settings and work productively in them. Resources of these kinds will no doubt be most plentifully available in well-endowed elite institutions. The provision of such support to PD beneficiaries requires either that they be publicly identified as beneficiaries or that the support be extended to a broader group, including non-beneficiaries as well, which is correspondingly more costly. Just as it takes resources to improve the sensitivity of selection processes, it takes resources to provide developmental support for underprepared beneficiaries too. Decisions about how much to invest in support of PD beneficiaries will therefore have a significant impact on the overall net benefits of the PD policies. Resources are always limited, and the payoffs to this kind of investment are not always obvious – nor are they all captured by the investors. Here too, there will probably be a need for subsidies to encourage organizations practicing PD to make sufficient investments of this kind. I have argued above that the prospective success of a PD policy can be significantly enhanced by financial resource commitments to selection programs, which increase the ability of select promising beneficiaries, and to developmental programs, which improve the performance of beneficiaries. There is one further kind of financial resource commitment that can make a significant contribution to the
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
715
success of a PD policy: direct financial aid to needy beneficiaries. Student PD beneficiaries may not be able to afford all of the (unsubsidized) expenses associated with attending an educational institution, so they may require financial aid just to enroll. Moreover, for some PD beneficiaries, economic insecurity is a potentially significant source of poor performance. Students from socioeconomically disadvantaged families are much more likely to have dropped out of an educational program temporarily or permanently – in order to help their families by contributing their labor within the family or by earning additional income. Employees from socioeconomically disadvantaged families are more likely to absent themselves from work – if not to quit their job – in order to address family crises. Financial aid to PD beneficiaries who are students, and loans to those who are employees, can help to prevent such problems from compromising the success of PD policies. The latter context is one where a means test is perfectly appropriate and wholly desirable. As noted above, excluding well-to-do PD beneficiaries from access to PD preferences would mean foregoing some important benefits of PD policies. It would be perfectly appropriate, however, to exclude them from financial aid programs designed to enable socioeconomically disadvantaged PD beneficiaries to remain active in the positions to which PD has given them access.
Conclusion Are policies of positive discrimination a good way to enhance the well-being of disadvantaged and underrepresented ethnic communities? PD policies have the potential to generate significant benefits – perhaps most importantly in facilitating greater integration of beneficiary communities into society’s elite – which can contribute to a more vital democracy and more effective societal institutions, as well as greater equity across communities. On the other hand, PD policies also have the potential to give rise to significant costs – such as poorer performance by individuals in key positions and heightened tensions between ethnic communities. In this chapter, I have argued that the likelihood that a PD policy will generate net benefits depends on a host of factors, including the nature and circumstances of the prospective beneficiary community, the sphere(s) to which the policy is applied, and the way in which the policy is configured. Key elements of PD policy configuration include the magnitude of the preference afforded to members of a beneficiary community and the extent of support provided for underprepared beneficiaries. A critical determinant of the success of a PD policy is the ability of its beneficiaries to perform well, once selected preferentially to a desired position. An inherent obstacle to the success of PD policies is that the likelihood of good performance by PD beneficiaries tends to vary inversely with the strength of the need for giving preference to a particular underrepresented ethnic group – since most members of groups with a strong case for positive discrimination will ipso facto be poorly prepared to compete with and perform as well as members of more advantaged groups (except perhaps in the arena of representative politics). This puts a premium on the capacity of PD policies to identify the most promising applicants among
716
T. E. Weisskopf
members of underrepresented ethnic communities and to ensure support for PD beneficiaries in the positions that they are enabled to attain. A PD policy is more likely to be successful in generating net benefits when it conforms with the following conditions: – The beneficiary community is fairly homogeneous; its members have been and continue to be subject to mistreatment and stigmatization by other communities; and they are significantly underrepresented in esteemed positions in the society. – The PD policy is applied to a sphere of activity that has a significant developmental component, so that beneficiaries can acquire human and social capital that will significantly enhance their abilities and opportunities. – The PD policy applies to high-quality and relatively well-endowed institutions and organizations that provide access to the upper strata of the society. – The PD policy is applied to a sphere in which the quality of performance of a beneficiary affects mainly the individual beneficiary, and not other parties. – The magnitude of the PD preference granted to beneficiaries is not so large as to make it difficult for them to succeed in the environment to which they gain preferential access. – The process of selection of PD beneficiaries is sensitive, nuanced, and capable – at least to some extent – of identifying formally underqualified applicants who are most likely to succeed, if given the opportunity. – The PD policy does not render its individual beneficiaries clearly identifiable as different from all of their peers. – The PD beneficiaries are afforded significant human and financial support after being preferentially selected. The greater the extent to which a PD policy is structured to meet the above desiderata, the more likely it will prove to be a good way of enhancing the well-being of disadvantaged and underrepresented ethnic communities. Finally, a good policy of positive discrimination is one that is ultimately time-limited. To the extent that it is successful, a PD policy will reduce the degree of a disadvantaged community’s underrepresentation that provided the rationale for preferential selection of its members in the first place. The conditions giving rise to the need for a PD policy, however, are enduring and deep-seated. One cannot, therefore, expect that the need for such a policy will disappear within just a generation or two; it is more likely a matter of at least a century. While it remains in effect, there is every reason to revisit a PD policy periodically, to adjust it so that it functions effectively under everchanging circumstances.
References Anderson EA (2002) Integration, affirmative action and strict scrutiny. NYU Law Rev 77: 1195–1271 Beteille A (1981) The problem. Seminar 268:10–13
30
Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?
717
Chatterjee BB (1983) Social costs of reservation in higher education: a decision theoretic view. J High Educ 9:77–89 Cunningham CD, Loury GC, Skrentny JD (2002) Passing strict scrutiny: using social science to design affirmative action programmes. Georgetown Law J 90:835–882 Fryer RG, Loury G (2005) Affirmative action and its mythology. J Econ Perspect 19-3:147–162 Loury GC (2002) The anatomy of racial inequality. Harvard University Press, Cambridge, MA Montejano D (1998) Maintaining diversity at the University of Texas. In: Post R, Rogin M (eds) Race and representation: affirmative action. Zone Books, New York, pp 359–369 Patwardhan V, Palshikar V (1992) Reserved seats in medical education: a study. J Educ Soc Change 5:1–117 Radhakrishnan P (1996) Mandal commission report: a sociological critique. In: Srinivas MN (ed) Caste: its twentieth century avatar. Viking Press, New Delhi Sivaramayya B (1996) The Mandal judgment: a brief description and critique. In: Srinivas MN (ed) Caste: its twentieth century avatar. Viking Press, New Delhi Weisskopf TE (2004) Affirmative action in the United States and India: a comparative perspective. Routledge, London, printed and distributed in South Asia by Foundation Books
Experimental Evidence on Affirmative Action
31
Ve´ronique Gille
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How Is Affirmative Action Implemented? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contextual Background to Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Can Affirmative Action Increase Representation Without Harming Efficiency? . . . . . . . . . . . . . . Affirmative Action Increases Participation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Is There an Efficiency Loss? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Are Affirmative Action Beneficiaries Penalized by Others? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evaluation of Beneficiaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cooperation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dishonest Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
720 720 720 721 722 722 723 726 726 727 727 727 728
Abstract
Although beliefs about the effect of affirmative action are widespread, behavioral responses to affirmative actions (from beneficiaries and non-beneficiaries) cannot easily be understood without experimental evidence. In this chapter I review the findings from the experimental literature on two key questions: can affirmative action increase representation without harming efficiency? Are affirmative action beneficiaries penalized by others? The findings highlight that beneficiaries respond to affirmative action as it changes their probability of participating in competitions and it changes their effort level. But the efficiency cost of affirmative action to the society, if any, is small, as affirmative action enhances the participation of highly performing individuals, who would have not participated otherwise. Moreover, beneficiaries do not suffer from backlash from V. Gille (*) Université Paris-Dauphine, Université PSL, LEDA, CNRS, IRD, Paris, France e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_37
719
720
V. Gille
non-beneficiaries. The negative societal impact of affirmative action is therefore likely to be limited. Keywords
Affirmative action · Experiment · Behavior · Participation · Efficiency loss
Introduction Affirmative action policies are highly controversial. As highlighted by Fryer and Loury (2005), lots of misconceptions surround the debates on affirmative action. While part of the believes are just wrong (such as the belief that most people have that they personally suffer from affirmative action), part of them relate to questions that have theoretically ambiguous answers. This is, for example, the case of the belief that affirmative action beneficiaries will decrease their effort level (Coate and Loury 1993). Empirical evidence is therefore needed to clarify the effects of affirmative action. The empirical literature that studies the consequences of affirmative action policies using observational data has boomed in the last 20 years. This is due to the large scale expansion of affirmative action programs in the world, such as affirmative action policies in electoral positions in India or quotas for women in corporate board seats in Norway. The features of these policies enable researchers to convincingly evaluate some of the effects of affirmative action policies. However, behavioral changes due to affirmative action are rarely observed. The experimental literature aims at exploring the mechanisms underlying the aggregate impacts measured with observational data, while providing some answers to the most common beliefs on affirmative action policies. In this chapter I review the findings from the experimental literature on affirmative action. In section “Experimental Designs,” I first present how affirmative action is implemented in experiments. I then review the findings on the impacts of affirmative action on representation and performance in section “Can Affirmative Action Increase Representation Without Harming Efficiency?” In section “Are Affirmative Action Beneficiaries Penalized by Others?” I present the experiments that explore whether the beneficiaries are penalized because of their affirmative action status. In section “Concluding Remarks,” I provide concluding remarks.
Experimental Designs How Is Affirmative Action Implemented? Affirmative action is a general term that covers different policies. In practice, the experimental literature mostly studies affirmative action by implementing quotas in games with real effort tasks, where participants compete for a prize. Typically, the
31
Experimental Evidence on Affirmative Action
721
outcomes (effort, scores, cooperation, etc...) of a round without affirmative action are compared to the outcomes of a round where at least one of the winner is picked from a group with a given characteristic (sex, race, caste, etc...). One of the exception is the paper by Calsamiglia et al. (2013), who implement preferential treatment in a competition across children in solving sudokus. The children with no previous experience in solving sudokus begin with extra points. However, the type of affirmative action does not seem to matter a lot. In an experiment designed by Balafoutas and Sutter (2012), four types of affirmative action policies are compared: quotas (one of the two winners must be a woman), two variants of preferential treatment where a fixed increment is added to women’s performance, and repetition of the competition. They are interested in understanding whether the type of affirmative action differently impacts the participation of women in competition, and how it affects performance. Except for the treatment where the competition is repeated when no women are selected, the other three interventions increase participation. The highest increase comes from the preferential treatment with a large increment. Average performance of the winners is not affected by any of these interventions. Similarly, Banerjee et al. (2018) find no difference in the competitiveness of affirmative action beneficiaries when comparing quotas and preferential treatment.
Contextual Background to Affirmative Action Affirmative action is meant to help groups of people that are lagging behind in terms of socio-economic outcomes. The experimental literature tries to reproduce this context by generating inequality across players at baseline, or by giving affirmative action to groups of players with a specific identity.
Baseline Inequality Several papers introduce inequality across groups at baseline, by using games where the affirmative action group performs less well. The participants in the experiment of Calsamiglia et al. (2013), for example, are children from two different schools that have different training in the real effort task (Sudoku). Other papers also use a real effort task where the performance at baseline is lower for groups that receive affirmative action later in the game. This is the case in Balafoutas and Sutter (2012), Maggian and Montinari (2017), and Niederle et al. (2013), who use a math game (adding sets of 2 digits numbers) where women perform less well on average. Another way of introducing inequality is by reducing the time that people have to do the task as in Petters and Schroeder (2020) or by paying less one group for the same task such as in Kölle (2017) or Dorrough et al. (2016). Most of these papers do not examine how inequality at baseline impacts the way people react to affirmative action. Among the exceptions are the experiments by Ip et al. (2020) and Petters and Schroeder (2020), who find contradictory results. Ip et al. (2020) highlight that when women are disadvantaged in the hiring process, quotas for women managers improve hierarchical relations, as measured by the level
722
V. Gille
of effort provided by workers, and the level of wages decided by managers in a giftexchange game. However, when women are not disadvantaged in the hiring process, effort and wages decrease. On the contrary, Petters and Schroeder (2020) show that initial inequality does not change the effect of quotas on peer-reviews distortions. More research is needed on this issue.
Salient Identity With respect to identity, two different strategies are followed. In most experiments, the identity of the beneficiaries of affirmative action is explicit, and just meant to reflect the real world: it is most of the time women (Balafoutas and Sutter 2012; Heilman et al. 1997; Ip et al. 2020; Kölle 2017; Maggian and Montinari 2017; Niederle et al. 2013) and scheduled castes in the paper by Banerjee et al. (2018). However, when identity is salient, the reaction of others to affirmative action may be confused with a reaction to the identity of the beneficiaries. To get rid of this concern, Petters and Schroeder (2020) give colors to groups, which do not reflect real characteristics of the participants. Balafoutas et al. (2016) and Dorrough et al. (2016) follow the same strategy and compare the impact of affirmative action when the beneficiaries are women to when the beneficiaries are people with a specific color attributed at the beginning of the game. While Dorrough et al. (2016) find no difference (but overall almost no effect of affirmative action on cooperation), Balafoutas et al. (2016) finds that the performance of beneficiaries decreases when quotas are given to a color rather than to women. Again, more research is needed to understand how identity interacts with the impacts of affirmative action.
Can Affirmative Action Increase Representation Without Harming Efficiency? The first and main question asked in the literature on the impacts of affirmative action is: does it work? and if yes, at what cost? In the experimental literature, these questions are rephrased as: is affirmative action effective at increasing the participation of the targeted group in competitive settings? And how does it impact the average performance of selected winners?
Affirmative Action Increases Participation The answer to the question of whether affirmative action increases participation is unambiguous. Yes, affirmative action works, it increases the participation of targeted groups to competitive games. As described in section “How Is Affirmative Action Implemented?,” this question is studied in the context of games with a real effort task. To understand how affirmative action affects participation, the players have to choose between compensation schemes where only the best performers get a reward, or a piece-rate payment, and choices are compared in rounds with quotas and rounds
31
Experimental Evidence on Affirmative Action
723
without quotas. The reward is higher in the compensation scheme where only the best performers get a reward (Balafoutas and Sutter 2012; Niederle et al. 2013; Banerjee et al. 2018). The results highlight that in the presence of quotas, the share of individuals among the targeted population that choose the compensation scheme where only the best performers to get a reward increase. This is true when the targeted group is women (Balafoutas and Sutter 2012; Niederle et al. 2013) or low-castes (Banerjee et al. 2018). Niederle et al. (2013) go a step further and look at which women respond to the implementation of quotas. Interestingly, they find that women tend to over-react to quotas, as quotas increase the participation of the best performing women, but also of women that have a higher expected gain with the piece-rate payment. Similarly, Balafoutas and Sutter (2012) find that affirmative action increases the participation of strong and weak performers, but not of intermediate performers. Affirmative action is designed to encourage the participation of a given group, but it also mechanically lowers the expected gains of non-beneficiaries. What is the effect of affirmative action on the participation of non-beneficiaries? The effect is negative in all the three papers that study this question, but significant in only one of the paper: in Balafoutas and Sutter (2012) and Banerjee et al. (2018), the effect is not significant, respectively, for men and general categories (with respect to caste), but affirmative action significantly decreases the participation of men in Niederle et al. (2013). Interestingly, Niederle et al. (2013) find that participation decreases for men at all performance levels. The difference in findings between Balafoutas and Sutter (2012) and Niederle et al. (2013) is hard to explain as the settings of the experiment are almost identical. While quotas make sure that the target group gets represented, whether participation translates into representation with other forms of affirmative action is an open question. Calsamiglia et al. (2013) finds that a small preferential treatment is enough to increase the representation of children that are less trained at baseline, as they increase their effort level (see section “Does Affirmative Action Impact Effort?” below). However, this result cannot be generalized as the answer obviously depends on the distribution of skills among the beneficiaries and the non-beneficiaries as well as how affirmative action impacts effort and performance.
Is There an Efficiency Loss? Given that the participation and the representation of individuals benefitting from affirmative action increase, the follow-up question is whether affirmative action harms the average performance of selected winners. This is a key question, as the opponents of affirmative action often argue that the cost of affirmative is too high to the society (Deshpande and Weisskopf 2014). And it seems at first sight intuitive that affirmative action would be at the cost of efficiency, as it is usually targeted to groups that perform less well on average before the implementation of the policy. However, this is not a so clear-cut question. First, affirmative action changes the composition of the pool of players as explained above in section “Affirmative Action Increases
724
V. Gille
Participation.” As it enhances the participation of some and discourages the participation of others, the final effect on performance depends on the characteristics of those who react to the policy. Second, affirmative action may change the effort level of the participants, positively or negatively. We explore below the findings of the experimental literature on these two aspects.
What Is the Profile of Affirmed Winners? The change in the composition of the pool of participants following affirmative action is studied in Balafoutas and Sutter (2012) and in details in Niederle et al. (2013). Despite the fact that they measure the competence on the task of participants by their performance in the game, they are able to disentangle the two mechanisms – change in effort and change in the competence level of participants – thanks to their design: the choice that the players have to make between a piece-rate payment and a compensation scheme where only the best performers get a reward (explained in section “Affirmative Action Increases Participation”) is made for a round that they already played and that was rewarded with a piece-rate payment. The behavior and performance on the task is therefore not impacted by the implementation of quotas. Both papers find that quotas enhance the participation of highly performing women. This increase in the number of highly performing women is enough to compensate for the significant loss in the number of highly performing men that choose to compete in Niederle et al. (2013), such that the average performance of winners does not change with affirmative action in both papers. Whether this is more generally true is unlikely, as it depends on the distribution of competence with respect to the task in the beneficiary and non-beneficiary population. However, these two papers highlight that it is not necessarily the case that affirmative action comes at the cost of efficiency. Does Affirmative Action Impact Effort? Another way the average performance of winners could be changed is if the presence of quotas changes the effort level of participants. Effort is complicated to measure, as final performance, which is often used as proxy for effort, may also capture other psychological effects induced by affirmative action, such as the stigmatization effect of making the identity of a group salient. Researchers have tried to deal with this in various ways. Calsamiglia et al. (2013), in one arm of the experiment, reveal to the players the training disadvantage of one of the group without implementing affirmative action. This allows taking into account in later analysis the fact that revealing this information could impact performance. Dulleck et al. (2017) measure effort in two ways: by the number of “hints” the players ask (i.e., ask for some kind of help to continue further in the game) and by the final performance level. In Schotter and Weigelt (1992) and Ip et al. (2020), effort is measured through a monetary cost that players choose to spend. Keeping in mind the difficulty of measuring effort, what is the impact of affirmative action? The results seem to vary a lot with the context, and average effects may hide heterogeneous impacts. I first review the impact of affirmative action on the effort level of beneficiaries and then on the effort level of non-beneficiaries.
31
Experimental Evidence on Affirmative Action
725
Effort of Beneficiaries: While Calsamiglia et al. (2013) find that the effort of beneficiaries of affirmative action increases when affirmative action is implemented, papers in which heterogeneous effects are explored have more ambiguous findings. Schotter and Weigelt (1992) find that the impact of affirmative action on effort varies with the level of disadvantage of the beneficiaries. When the disadvantage is small, affirmative action decreases their effort. When it is large, affirmative action increases effort through a decrease in dropout (those who dropout are counted as making 0 effort). Balafoutas et al. (2016) explore the heterogeneity of the effect of affirmative action on effort with respect to the identity of the group that benefit from affirmative action. They find that when affirmative action beneficiaries are women, there is no impact on effort. However, when affirmative action is targeted to a group that has a specific color attributed randomly at the beginning of the game, effort of the beneficiaries increases. Finally, Dulleck et al. (2017) find that affirmative action reduces the effort of beneficiaries, except when there is a negative stereotype associated with their performance in the game. Effort of Non-beneficiaries: The effects of affirmative action on the effort level of non-beneficiaries is also highly heterogeneous. Calsamiglia et al. (2013) explore the heterogeneity of the impact with respect to participants’ ability. They find that affirmative action enhanced effort for most subjects. But the effect declines with ability such that it is negative for high ability non-beneficiaries. Balafoutas et al. (2016) and Ip et al. (2020) find heterogeneous effects with respect to the context of implementation of affirmative action. Balafoutas et al. (2016) find no effect when the group of beneficiaries is composed of women, but a negative impact on effort when the affirmative action status is determined by a random characteristic. Ip et al. (2020) find a positive impact on effort when women are discriminated against in the recruitment process, but negative otherwise. Affirmative action therefore seems to impact effort in various ways, and there is to date no consensus on the direction of the effect both on the beneficiaries and non-beneficiaries. More research is needed on this issue to get a clear picture of this question and to understand the behavioral mechanisms at play.
Aggregate Impacts Given the results that we discussed above, it is hard to draw a conclusion on the efficiency effects of affirmative action. On the one hand, affirmative action tends to enhance the participation of highly skilled individuals among beneficiaries. On the other hand, non-beneficiaries are less likely (although not always significantly) to compete. Moreover, effort levels of both beneficiaries and non-beneficiaries vary in both directions, depending on the context. The aggregate impact on the performance level of winners is therefore not clear. However, what seems to be clear, is that in all the experiments the change in the average performance of winners is small. Overall, the efficiency cost of affirmative action is likely to be small.
726
V. Gille
Are Affirmative Action Beneficiaries Penalized by Others? Affirmative action therefore does not seem to harm society as the experiments show that the effect on efficiency is, if anything, small. However, affirmative action may actually harm beneficiaries. This may be the case if beneficiaries fail because the level is too high, for example in universities. This may also be the case if beneficiaries are penalized because of their status. This is this latter situation that the experimental literature has explored. This question has been studied following three angles that I detail below.
Evaluation of Beneficiaries The first angle is evaluation. Does affirmative action status biases the evaluations that other people make of the performance of beneficiaries? It may happen if people have stereotypes on affirmative action beneficiaries, or if they deliberately want to punish them. Two papers examine the impact of affirmative action on the evaluation of affirmative action beneficiaries. Heilman et al. (1997) are interested in the evaluation given by external people, who have no particular interest in the outcome of the procedure. They ask real managers to evaluate the profile of fake employees. The profile includes information about how good they have been on their job and whether they got the job through an affirmative action program. They vary the affirmative action status, and the level of ambiguity in the information on performance. When the information on the performance of the affirmative action beneficiary is ambiguous, managers tend to decrease their own evaluation of the employees. It is not the case for non-affirmative action beneficiaries, and it is not the case when the information on the performance of the affirmative action beneficiary is clear. As the evaluator is external, the most likely mechanism is stereotype: in the presence of ambiguity, people tend to give more weight to the information on affirmative action. Petters and Schroeder (2020) look at a totally different setting to study the impact of affirmative action on peer-review evaluation. Participants are in groups of four, and are asked to make a creative task. They then have to evaluate the creative task of their team members. Without affirmative action, the two best participants get a prize. With affirmative action, the best participant gets a prize as well as the best participant from the affirmative action group. The affirmative action status is given by a color (yellow or green), which is randomly attributed at the beginning of the game. They find that affirmative action negatively impacts the evaluation of the work of affirmative action beneficiaries. However, these distortions in peer reviews only come from affirmed raters, so they argue that the mechanism is enhanced competition within the affirmative action group. Both papers find that the performance evaluation of affirmative action beneficiaries are distorted, but the mechanisms are totally different: in one case, the observed biases are likely to be driven by stereotypes toward affirmative action beneficiaries. In the other case, distortions arise because of a strategic behavior of competitors.
31
Experimental Evidence on Affirmative Action
727
Importantly, non-beneficiaries do not distort their reviews, such that it is clear that there is no retaliation mechanism involved.
Cooperation The second set of papers examines the impact of affirmative action on cooperation. The three papers that study this question find no impact of quotas on cooperation. Dorrough et al. (2016) study the behavior of individuals in a real effort task where the performance of all the team members is pooled and the price equally divided across team members. They find little change in the effort level of people in the presence of affirmative action. In a similar setup, Kölle (2017) finds no effect of affirmative action on cooperation within teams, as measured by performance in the team task where an affirmative action does not impact the willingness of people to work in teams. Finally, Balafoutas and Sutter (2012) also find no effects of four different types of affirmative action programs in a coordination game where the two players have to both pick a high number to maximize their payoff, but can penalize the other player by choosing a low number.
Dishonest Behavior Finally, the third outcome researchers have looked at is dishonesty. Banerjee et al. (2018) and Maggian and Montinari (2017) look at whether affirmative action enhances dishonest behavior using a dice game. People roll a die and have to say out loud what the number is. Some numbers are winning numbers. As the die is not observed by the organizers, people have an incentive to declare a winning number even if they got another one. Both papers find no impact of quotas in a previous game on the probability of people to declare another number as the one they actually got.
Concluding Remarks What do we learn from this experimental literature on affirmative action? People react to affirmative action: it changes their probability of participating in competitions and it changes their effort level. But the negative impact on the performance level of winners (and therefore the efficiency cost of affirmative action to the society), if any, is small, as affirmative action enhances the participation of highly performing individuals, who would have not participated without. Moreover, there is no backlash effect: the papers that study the impact of affirmative action on cooperation or dishonest behavior do not find any change in the behavior of beneficiaries or of non-beneficiaries. The negative societal impact of affirmative action is therefore limited.
728
V. Gille
However, the findings from the literature looking at the impact on evaluation, and in particular on how affirmative action beneficiaries are perceived, paint a gloomier picture. The fact that beneficiaries are less well rated for a given performance when their status is known should not be overlooked as this could have important impacts on their economic prospects. This literature also brings new questions. The understanding of the mechanisms at play, in the behavioral changes that we observe, is still at an early stage. For example, the heterogeneity of results in the literature that looks at the impact on affirmative action on effort should be further explored, theoretically and empirically. The role of the initial setup, which makes affirmative action seem more or less fair, is also little understood. Further research on this could help designing affirmative policies with minimal negative spillovers.
References Balafoutas L, Sutter M (2012) Affirmative action policies promote women and do not harm efficiency in the laboratory. Science 335(6068):579–582 Balafoutas L, Davis BJ, Sutter M (2016) Affirmative action or just discrimination? A study on the endogenous emergence of quotas. J Econ Behav Organ 127:87–98 Banerjee R, Gupta ND, Villeval MC (2018) The spillover effects of affirmative action on competitiveness and unethical behavior. Eur Econ Rev 101:567–604 Calsamiglia C, Franke J, Rey-Biel P (2013) The incentive effects of affirmative action in a realeffort tournament. J Public Econ 98:15–31 Coate S, Loury G (1993) Antidiscrimination enforcement and the problem of patronization. Am Econ Rev 83(2):92–98 Deshpande A, Weisskopf TE (2014) Does affirmative action reduce productivity? A case study of the Indian railways. World Dev 64:169–180 Dorrough AR, Leszczy’nska M, Barreto M, Glöckner A (2016) Revealing side effects of quota rules on group cooperation. J Econ Psychol 57:136–152 Dulleck U, He Y, Kidd MP, Silva-Goncalves J (2017) The impact of affirmative action: evidence from a cross-country laboratory experiment. Econ Lett 155:67–71 Fryer RG, Loury GC (2005) Affirmative action and its mythology. J Econ Perspect 19(3):147–162 Heilman ME, Block CJ, Stathatos P (1997) The affirmative action stigma of incompetence: effects of performance information ambiguity. Acad Manag J 40(3):603–625 Ip E, Leibbrandt A, Vecci J (2020) How do gender quotas affect workplace relationships? Complementary evidence from a representative survey and labor market experiments. Manag Sci 66(2):805–822 Kölle F (2017) Affirmative action, cooperation, and the willingness to work in teams. J Econ Psychol 62:50–62 Maggian V, Montinari N (2017) The spillover effects of gender quotas on dishonesty. Econ Lett 159:33–36 Niederle M, Segal C, Vesterlund L (2013) How costly is diversity? Affirmative action in light of gender differences in competitiveness. Manag Sci 59(1):1–16 Petters LM, Schroeder M (2020) Negative side effects of affirmative action: how quotas lead to distortions in performance evaluation. Eur Econ Rev 130:103500 Schotter A, Weigelt K (1992) Asymmetric tournaments, equal opportunity laws, and affirmative action: some experimental results. Q J Econ 107(2):511–539
Does Political Affirmative Action Work, and for Whom? Theory and Evidence on India’s Scheduled Areas
32
Saad Gulzar, Nicholas Haas, and Benjamin Pasquale
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theory and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Extensive Margin (Size of the Pie) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Intensive Margin (Distribution of the Pie) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Context: Identity, Quotas, and Development in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scheduled Areas in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Panchayat Extension to Scheduled Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quotas and Political Conflict: A Case Study of Jharkhand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comparisons Across Indian Identity Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Local Government and Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The National Rural Employment Guarantee Scheme (NREGS) . . . . . . . . . . . . . . . . . . . . . . . . . . . Beyond NREGS: Rural Roads and Other Public Goods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geographic Regression Discontinuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analysis of Balance with Census Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
730 733 733 734 736 736 737 737 738 739 739 741 742 744 744 745
This chapter was originally published in 2020 as an article by the American Political Science Review, in its Volume 114, Issue 4, pp. 1230–1246, and can be seen along with supplementary material at https://doi.org/10.1017/S0003055420000532. Re-published here with permission. S. Gulzar (*) Princeton University, Princeton, NJ, USA e-mail: [email protected] N. Haas Aarhus University, Aarhus, Denmark e-mail: [email protected] B. Pasquale New York, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_51
729
730
S. Gulzar et al.
The Impact of Scheduled Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Impacts on the National Rural Employment Guarantee Scheme (NREGS) . . . . . . . . . . . . . . . . Impacts on the Rural Roads Program (PMGSY) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Impacts on Public Goods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion: Bringing the Results Together . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Investigating the Electoral Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scheduled Areas Prior to PESA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Local Elections in Scheduled Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Targeted Minority Electoral Influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quota Overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
746 746 748 749 751 752 752 753 753 754 755 757 758
Abstract
Does political affirmative action undermine or promote development? We present the first systematic analysis of Scheduled Areas in India, home to 100 million, where local political office is reserved for the historically disadvantaged Scheduled Tribes. A newly constructed dataset of 217,000 villages allows us to probe conflicting hypotheses on the implementation of the world’s largest workfare program, the National Rural Employment Guarantee Scheme. We find that reservations deliver no worse overall outcomes, that there are large gains for targeted minorities, and that these gains come at the cost of the relatively privileged, not other minorities. We also find improvements in other pro-poor programs, including a rural roads program and general public goods. Reservations more closely align benefits to each group’s population share, allaying concerns of overcompensation for inequalities. Contrary to the expectations of skeptics, results indicate that affirmative action can redistribute both political and economic power without hindering overall development. Keywords
Affirmative action · Electoral quota · Scheduled Areas · Scheduled Tribes · Scheduled Castes · Employment · Public goods · India
Introduction Many countries have adopted political affirmative action with the express aim of raising the voice of marginalized communities in how governments function. This chapter asks how improvements in descriptive representation might impact economic welfare. Studying this question is of particular importance where poor populations rely on large-scale government welfare programs such as in the case of the Benazir Income Support Program in Pakistan that provides 5.4 million poor women income supplements as a safety net, the Supplemental Nutrition Assistance Program in the USA that helps 46 million low-income individuals purchase
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
731
groceries every month, and the National Rural Employment Guarantee Scheme (NREGS) in India, the world’s largest employment program that we examine in this study. Does descriptive representation achieved through affirmative action deliver improved welfare for marginalized communities, or does restricting representation prove self-defeating in that it damages the economic prospects of the populations it was designed to politically empower? We study electoral quotas, an affirmative action policy that directly yields descriptive representation and is implemented in over 100 countries,1 and ask two related, yet underexplored questions: Do electoral quotas improve or hinder development? And how are the benefits (and costs) of electoral quotas distributed in society? Prior evidence is mixed and does not offer clear theoretical expectations. Focusing on minorities explicitly targeted under an electoral quota, some studies, which we review below, show strong positive welfare effects, while others report no improvements. We organize and extend hypotheses from previous work in a novel theoretical framework that enables an explicit accounting of how electoral quotas affect the extensive margin of program implementation (that is, the overall size of the pie) and the intensive margin (the distribution of the pie) for targeted disadvantaged groups, nontargeted disadvantaged groups, and for the comparatively privileged groups under the status quo. This exercise allows a fuller understanding of the trade-offs involved in the implementation of affirmative action policies. A solidarity hypothesis predicts that shared interests and experiences between minority groups should lead to positive program spillovers from quota-targeted to nontargeted minorities. A crowding-out hypothesis predicts that gains for a quota targeted minority will come at the cost of other groups, particularly nontargeted minorities. And, a performance hypothesis predicts better outcomes for targeted minorities and unchanged outcomes for others, or, at the very least, negative outcomes for others that do not outweigh gains for targeted minorities. This chapter presents the first systematic evidence on a large electoral quota in India that brought increased descriptive representation to well over 100 million citizens. Shortly after Independence from the British in 1947, the Indian parliament declared certain regions in the country as Scheduled Areas (SA), a designation linked to the protection of a historically disadvantaged category of minority groups, the Scheduled Tribes (ST). From 2000, under the Panchayat Extension to Scheduled Areas (PESA) Act, India’s national parliament implemented a dramatic electoral quota in Scheduled Areas requiring that all chairperson positions in three tiers of local government councils, as well as at least half the seats on each of those councils, be reserved for individuals from the Scheduled Tribes. Why does understanding the impact of this electoral quota matter? First, the quota has received no systematic quantitative analysis despite the fact that it is present in half of India’s states and covers nearly half of the territory within those states. Second, the quota targets ST, who are considered to be among the most economically vulnerable and politically excluded groups in India. Third, the permanence of
1
See Krook and Zetterberg (2014).
732
S. Gulzar et al.
the Scheduled Areas quota is qualitatively different from population-based quotas that rotate over time. Scholars have argued that rotation is an impediment to longterm quota success (Dunning and Nilekani 2013; Bhavnani 2009). Isolating the causal effect of Scheduled Areas is not straightforward. Indeed, comparing SA to non-SA using data from the 2001 Indian Census shows that they differ on a number of dimensions. By employing a geographic regression discontinuity (RD) design similar to Dell (2010), we absorb variation that correlates with geographic space, allowing for a comparison of villages lying just on one or the other side of the border between non-Scheduled and Scheduled Areas. We first examine the impacts of Scheduled Areas using data from NREGS, a flagship federal program in India with an annual cost of approximately US$6 billion. Each year, the social protection scheme officially guarantees 100 days of minimumwage employment for every rural household in India. We study program delivery to rural populations in 2013, up to 12 years after the first implementation of PESA. We do this by creating a new dataset with 217,144 villages that combines official NREGS implementation data with an original spatial dataset of Scheduled Area status. The scale and depth of these data, which permit us to evaluate both the extensive and intensive margins of program delivery, are a substantial advance on existing work on affirmative action and economic development.2 Results show that NREGS delivery improves substantially for the targeted minorities (ST), who receive 24.1% more workdays in Scheduled Areas. Improvement appears to come primarily at the cost of work for nonminorities (non-SC/ST), who receive 12.5% fewer workdays. We find no evidence that the quota causes a change in employment for the nontargeted, historically disadvantaged minorities (SC). Our evidence thus offers support for the crowding-out and performance hypotheses, but not for the solidarity hypothesis. Overall, the results indicate that the delivery of government programs in Scheduled Areas are no worse than in non-Scheduled Areas. Contrary to concerns that quotas might overcompensate for historical inequalities, we find that the Scheduled Areas quota more closely aligns NREGS work, by identity category, to each group’s share of the local population. Are these effects specific to NREGS? We evaluate broader impacts of Scheduled Areas by examining a second large-scale development scheme as well as outcomes from the 2011 Census. Data show improved provision of public goods that is likely to benefit disadvantaged communities. We also observe increased rural road connectivity from the Pradhan Mantri Gram Sadak Yojana (PMGSY) village roads program. These improvements are consistent with the results from NREGS, insofar as they reflect a higher responsiveness to the needs of marginalized communities. To what extent are the results we observe the function of an electoral politics mechanism? We provide four pieces of evidence. First, qualitative evidence from
2 While Jensenius (2015), Pande (2003), and Das et al. (2017) conduct similar exercises, our detailed data help to disaggregate the nontargeted group into meaningful categories of SC and nonminorities, allowing us to study the causal effect of reservations on both efficiency and redistribution.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
733
Indian historical studies, as well as quantitative evidence from the PMGSY program and the Indian Census, shows that villages on opposite sides of the Scheduled and non-Scheduled border were very similar, and followed parallel trends, prior to the implementation of PESA. Second, there is evidence from two programs that ST politicians strategically target benefits to ST communities. Third, the effects of the quota are reduced in areas of overlap with quotas for state-level ST legislators. Fourth, the impact of the quota is largest when it constitutes the greatest shock to political representation: that is, following the first election. This chapter makes theoretical, empirical, and policy contributions. The theoretical contribution is to explicitly lay out hypotheses on the trade-offs of affirmative action on targeted and historically disadvantaged, nontargeted and disadvantaged, and nontargeted and nondisadvantaged identity-based groups, and combine them into a unified framework. Empirically, our unique data allow us to test these hypotheses in the context of three critical, village-level data sources – the largest rural employment scheme in the world, a national rural roads development program, and public goods and economic measures from the census of the world’s largest democracy, India. From a policy perspective, all too often policymakers and analysts treat parallel pro-poor economic and political efforts in isolation. By considering their interaction, we hope to advance our understanding of how politics can be made to work for inclusive development.
Theory and Hypotheses In this section, we review conflicting findings and draw hypotheses from existing work on the effects of political affirmative action on government functioning.
Extensive Margin (Size of the Pie) Given the same resources and institutional design, do electoral quotes positively or negatively affect the overall efficacy of government programs? Implementation of government programs would suffer if quota politicians are less competent than nonquota politicians (Jensenius 2017). Jensenius (2015) presents qualitative evidence that SC quota politicians are viewed as inexperienced and referred to as “weak,” “inefficient,” and “useless” (p. 202). Deshpande and Weisskopf (2014) document how some oppose affirmative action policies due to a belief that they result in less qualified individuals and worse performance. Bertrand et al. (2010) find that students admitted under quotas see an increase in income, but these gains are more than offset by losses in earnings for individuals displaced by the quota. Conversely, implementation could improve if quota politicians work harder for their constituents. Chin and Prakash (2011) report that ST quotas, but not SC quotas, result in lower levels of overall poverty. Deshpande and Weisskopf (2014) find that a greater proportion of high-level SC/ST employees in the Indian Railways is correlated with both increased productivity and growth. Evidence also suggests that
734
S. Gulzar et al.
women exert more effort and outperform men when positions of political influence are available to them (Beaman et al. 2010). Das et al. (2017) argue that in the presence of asymmetric group sizes, affirmative action can improve the efficiency of outcomes. Finally, government performance could remain unchanged if quota politicians perform no better or worse than nonquota politicians, as Bhavnani and Lee (2019) find for Indian bureaucrats.
Intensive Margin (Distribution of the Pie) We now turn to the impact of affirmative action on targeted and nontargeted historically disadvantaged communities, and more privileged communities. While theoretical examinations predict positive results for targeted minorities, existing research from India on electoral reservations has found mixed effects. Besley et al. (2007), Duflo and Chattopadhyay (2004), and Beaman et al. (2010) show that reservations for SC/ST and women improve the welfare of direct beneficiaries. Other work, such as Dunning and Nilekani (2013) and Jensenius (2015), find no overall effect of electoral quotas on targeted groups. Unlike our case, one explanation for weak effects in the literature is the rotating nature of quotas in these contexts, which limits politicians’ incentives to target benefits along ethnic lines. We expect targeted minorities to benefit under affirmative action. Less clear is what we should expect for nontargeted groups. We draw three hypotheses from existing literature for why gains for targeted minorities may alternatively result in positive, negative, or no spillovers to other groups. Solidarity Hypothesis Nontargeted minorities may experience positive spillovers from quotas targeting other minorities. Studies have found that minority politicians may carry intrinsic motivations – absent electoral motivations – to help individuals with whom they identify (Broockman 2013; Adida et al. 2016; Singh 2015). They may also share policy preferences with other minorities: Kaufmann (2003) writes that African Americans and Latinos in the USA “share objective circumstances [and] interests” (2003, p.199) and may have a “minority group consciousness.” Consistently, Adida et al. (2016) show that African Americans respond positively not only to co-ethnic but also to co-minority (Latino) political cues. Under this hypothesis, therefore, minority groups not targeted by the quota should also benefit from improved program implementation. Some evidence from India is consistent with this prediction: SC Reserved councillors increase village expenditures in a manner that benefits both SC and ST in their village (Palaniswamy and Krishnan 2012). These theories are largely silent on the expected effects on nonminorities. While one can extrapolate that this group will not benefit under this hypothesis because of a lack of solidarity with targeted minorities, it is unclear if they will be worse off or remain at the status quo. As a consequence, there are also no clear predictions on what happens to outcomes on the extensive margin.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
735
Crowding-Out Hypothesis Gains in descriptive representation for one minority group may come at the expense of benefits for nontargeted minorities, especially where targeted and nontargeted minorities are in competition. Meier et al. (2004) examine changes in representation among African Americans and Latinos and find that improvements in administrative and teaching positions for one group are associated with losses for the other. Expectations of intercaste competition and negative spillover effects are captured by Khosla (2011), who argues that as “different castes vie to capture NREGS benefits, they limit the access of other caste groups” (p.65). Under this hypothesis, quotas could leave outcomes for nonminorities unchanged especially if targeted minorities still live under social pressure from nonminorities, or where nonminorities are not in competition for the same goods. Alternatively, if competition for resources exists, and if nonminority groups do not retain full control over their distribution, in a weaker version of the hypothesis, nonminority groups could suffer losses.3 Extant evidence is limited: Jenkins and Manor (2017) note that there is no systematic evidence from India that asks if participation of “non-poor” in NREGS crowds out the “genuinely poor” (p. 168). Overall, critics of affirmative action cite concern that negative spillovers will outweigh any benefits to the targeted group. Performance Hypothesis Unlike the previous hypotheses that examined the relationship between various groups on the basis of solidarity or competition, the performance hypothesis simply states that improvements for a targeted minority may come without necessarily incurring costs on other groups, if, for instance, quota politicians exert more effort than nonquota politicians. Beaman et al. (2010) consider the effects of a quota for women on a nontargeted minority group, Muslims, and find that improved outcomes for women do not appear to crowd out benefits for Muslims. Iyer and Mani (2012) find that quotas for women increase reporting of crimes against women but do not appear to affect reporting for crimes against men. Since the spirit of this hypothesis is to make a claim about the net effects of the quota, a weaker version of the hypothesis would state that potentially positive effects on the targeted minority are greater than or equal to any negative spillovers to other groups. A confirmation of the weaker version still carries significant implications for arguments against affirmative action rooted in claims that quotas will result in a loss in overall efficiency. Table 1 summarizes the conflicting predictions offered by each of the three hypotheses.
3
Studies also indicate that individuals may be willing to forgo economic gains where they might come with social costs under an out-group leader, which could lead nonminority groups to opt out of competition (Akerlof and Kranton 2010; Moffitt 1983; Gille 2013). As we argue below, the design of NREGS makes this unlikely.
736
S. Gulzar et al.
Table 1 Summary of predictions Empirical implications: benefits for. . . Targeted minority Hypotheses Overall (ST) Solidarity ? " Crowding-out : " " Performance :# "
Nontargeted minority (SC) " # ?
Non-SC/ST :" :" ?
Context: Identity, Quotas, and Development in India The Indian government has instituted numerous forms of political quotas since Independence. In the political arena, the constitution provides dramatic guaranteed representation through quotas for individuals from the Scheduled Tribes (ST), Scheduled Castes (SC), Other Backward Classes (or Other Backward Castes, OBC), and/or women in the national parliament, state legislatures, and from 1993 in the country’s three-tier system of local government councils, called Panchayati Raj.4
Scheduled Areas in India We focus in this chapter on India’s Scheduled Areas, a government institution targeting tribal populations that has not yet been subject to systematic quantitative analysis. Scheduled Areas cover over 100 million citizens across nine Indian states – Andhra Pradesh, Chhattisgarh, Gujarat, Himachal Pradesh, Maharashtra, Madhya Pradesh, Jharkhand, Odisha, and Rajasthan. The demarcation of Scheduled Areas has changed little since the initial formulation during the pre-Independence period. British authorities first provided a list of “Aboriginal Tribes” and “Semi-Hinduised Aboriginal Tribes” in the Census of 1872 (Corbridge 2002, 64) and implemented special institutions targeting these tribes under the Scheduled Districts Act of 1874. Following Independence in 1947, the new Indian state identified Scheduled Areas in the Fifth Schedule of the Constitution, with minor differences from the British Scheduled Districts Act. The government justified Scheduled Areas specifically as a means to improve representation and welfare for Scheduled Tribes (ST) through special programs and institutions such as the state-level Tribes Advisory Council.5
4
While religion is an additional important identity category, since Independence the Muslim minority group has been excluded from political quotas. 5 We focus on the Fifth Schedule that governs the majority of Scheduled Areas in India. An additional Sixth Schedule of the Constitution details the administration of tribal areas in four northeastern states. For more information, see Appendix A.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
737
The Constitution assigns responsibility for adding, subtracting, or modifying Scheduled Areas to the president in consultation with the relevant state’s governor. In 1962, the Dhebar Commission proposed that an area should be eligible to become a Scheduled Area according to four, relatively vague, criteria: (i) preponderance of tribals in the population; (ii) compact and reasonable size; (iii) underdeveloped nature of the area; and (iv) marked disparity in economic standards of the people. In practice there has been no exact formula for updating or adjusting the previous notification or de-notification of Scheduled Areas in India, and these Areas have remained remarkably stable since their initial formulation (see Appendix A).
Panchayat Extension to Scheduled Areas Despite government commitments to promote ST interests in Scheduled Areas, villages on opposing sides of the Scheduled Areas border show few differences on observables or overtime trends prior to the implementation of the local-level political quotas that we study in this chapter. Indeed, additional legislation instituting political quotas were designed in large measure to give Scheduled Areas teeth. The Panchayats Extension to Scheduled Areas Act of 1996 (PESA) mandated that all chairperson positions at the three levels of local government, and at least 50% of all seats on these councils, be reserved for ST individuals. Hence, when local elections were next held – as early as 2000 for Rajasthan and as late as 2010 for Jharkhand – these reforms gave a tremendous positive shock to the local-level political representation of Scheduled Tribes in India. Unlike other quotas in India that rotate by constituency and over time, the quotas in Scheduled Areas introduced with PESA remain fixed. To summarize, in the research design to follow, we will be comparing villages in (non-Scheduled) areas where local elections were introduced from 1992 due to the Panchayati Raj Act, with (Scheduled) villages where elections were only rolled out beginning in 1996, with PESA.
Quotas and Political Conflict: A Case Study of Jharkhand By way of more detail, we provide a case analysis of the state of Jharkhand that has arguably the most politically charged and turbulent path to local elections with quotas in Scheduled Areas. Even in this politically fraught case, the actual boundaries of the Scheduled Areas have remained relatively unchanged. While Jharkhand passed an amendment in 2001 to allow for PESA-compliant panchayat elections, a legal challenge postponed elections. Only after a decision by the Indian Supreme
738
S. Gulzar et al.
Court in 2010, upholding the constitutional status of identity-based quotas in India, were local elections held in Jharkhand in 2010.6 Although the state of Jharkhand was created in part to better represent tribal populations in the state of Bihar, the actual Scheduled Areas within this region did not change. The Scheduled Areas assigned as part of the Indian Constitution’s Fifth Schedule remained almost entirely consistent through the Bihar Scheduled Areas Regulation of 1969 and re-notification again in 1977 and 2007. The only changes were the addition to the Scheduled Areas of a single block – Bhandaria of Garhwa district – in 1977, and the Scheduling of two village-clusters, both within Satbarwa block, in 2007.7
Comparisons Across Indian Identity Categories ST are not the only historically disadvantaged minority category in India, nor the only category targeted via special legislation. Others include the Scheduled Castes, Other Backward Classes (OBC), and women. While OBC also receive mandated representation in local government outside of Scheduled Areas in India, on average, and in taking India as a whole, SC and ST communities in existing literature are considered the most stigmatized, economically vulnerable, and politically excluded communities. The Indian government has acknowledged the vulnerable position of SC and ST communities and accordingly regularly groups SC and ST together for the purposes of special legislation.8 Outside of Scheduled Areas that privilege ST, since 1992 all local government councils across the country restrict local council leadership positions for SC and ST, using identical quotas in proportion to their local population that rotate every election cycle (see Duflo and Chattopadhyay 2004; Dunning and Nilekani 2013). Both popular and academic writing often describe SC and ST in tandem as examples of minority groups that are the poorest and most vulnerable throughout the country. The Indian government even studies the development of individuals from both groups together via the elite, national government–appointed, Planning
6
Union of India and Others v. Rakesh Kumar and Others. Supreme Court of India, January 12, 2010. 7 Appendix A provides further discussion on what constitutes a Scheduled Tribe and the Scheduled Areas in Jharkhand. 8 SC and ST categories first gained some preferential representation in the Government of India Act of 1935, officially sanctioned in the Constitution via Constitution (Scheduled Castes) Order, 1950, and the Constitution (Scheduled Tribes) Order, 1950. National Commissions for SC and ST were instituted via Articles 338 and 338A, respectively. Legislation was passed to protect individuals from both identity categories from violence in 1989 by means of the Scheduled Castes and Scheduled Tribe – Prevention of Atrocities Act.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
739
Commission.9 For these reasons, we consider outcomes for SC a useful comparison to ST outcomes – as both groups are similarly vulnerable, yet enjoy very different political opportunities in Scheduled Areas. Appendix A provides more details about political quotas in India and SC and ST identity categories.
Local Government and Development Local government panchayat institutions in India are responsible for two key aspects of development: welfare schemes and infrastructure, each of which provide local public goods (see Besley et al. 2007). Existing literature identifies roads, sanitation, electricity, water, telephones, school and health facilities, irrigation, and communication as important development sectors for measuring performance of panchayat institutions (Cassan and Vandewalle 2017; Munshi and Rosenzweig 2015). Our empirical goal is to measure how political reservations affect the implementation of government programs.
The National Rural Employment Guarantee Scheme (NREGS) As our key outcome, we chose NREGS, India’s largest development program and the largest employment program in the world. NREGS and rights-based policies in India build on prior legislation on decentralization and devolution of power to local government agencies. Kapur and Nangia (2015) classify this welfare scheme as part of the lowest tier of social protection in India that covers the vast majority of workers in the country (up to 94%) (pp. 76–77). Together with other programs like Public Distribution System and the National Social Assistant Program, NREGS is a riskcoping, instead of risk-mitigating, program that provides protection to those already at risk. The scheme officially guarantees 100 days of minimum-wage employment to every rural household in the country, with no eligibility requirements. Though increases in welfare spending in general might come at the expense of other spending priorities, NREGS funding comes primarily from federal and state budgets. Accordingly, local politicians who do not take full advantage of the NREGS program are effectively “leaving money on the table.”10 Jenkins and Manor (2017) document how NREGS has helped improve the lives of the poorest in India. Work given out under NREGS is a product of a large ecosystem that includes informal institutions, bureaucrats, and collective deliberation (Dutta et al. 2014; Khosla 2011; Marcesse 2018). However, recent research shows that village-level politics are likely to play an outsized role in the distribution of NREGS benefits
9
See for instance https://web.archive.org/web/20180218235157/ http://planningcommission.gov. in/aboutus/taskforce/inter/inter_sts.pdf 10 We discuss concerns related to leakage in Appendix G.
740
S. Gulzar et al.
(Marcesse 2017) (see Appendix A). Local-level council chairpersons – whose seats are reserved for ST under the Scheduled Areas quota – have both the capacity and discretion to significantly alter the quality of NREGS implementation and the distribution of NREGS benefits (Besley et al. 2007; Dasgupta and Kapur 2020; Dunning and Nilekani 2013; Dutta et al. 2014; Sukhtankar 2017; Marcesse 2018). NREGS has bolstered the legitimacy and efficacy of local governments by empowering local-level authorities. These authorities are responsible for: selecting projects through collective deliberation in village assemblies, selecting program beneficiaries, implementing at least 50% of all works, maintaining and transmitting records to higher authorities to process payments, and responding to citizens appeals for work (Sukhtankar 2017; Jenkins and Manor 2017; Dunning and Nilekani 2013; Marcesse 2018; Besley et al. 2007; Munshi and Rosenzweig 2015). While NREGS implementation remains uneven (see Fig. 1), the scheme’s implementation carries
Fig. 1 Variation in 2013 NREGS Workdays across India
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
741
political rewards. Good NREGS performance can also be an election winning device in local politics (Maiorano 2014, p. 95). Following changes in representation, and consistent with our theoretical conceptualization above, we can expect changes in both demand and supply to affect change in NREGS work outlay. On the demand side, at least two stories of change emerge. On the one hand, ST might feel more comfortable requesting work when an ST politician is elected (Gille 2018). On the other hand, non-SC/ST may opt out of demanding work from ST politicians due to associated social stigma (Akerlof and Kranton 2010; Moffitt 1983; Gille 2013). The latter possibility is unlikely because NREGS targets poor households and individuals in rural areas with work such as digging ditches and building wells. This is work that is “physically taxing, of uncertain duration, and provides no employment benefits” (Dutta et al. 2014, 14). NREGS was designed for those most in need of work, and as a last resort. Put differently, “By insisting that participants do physically demanding manual work at a low wage rate, workfare schemes such as MGNREGS aim to be self-targeted...nonpoor will not want to do such work, and poor people will readily turn away from the scheme when better opportunities arise” (Dutta et al. 2014, 540). More generally, we expect that changes in supply will have greater explanatory power than will changes in demand. Prior research indicates that the binding constraints on NREGS implementation are not demand-side but are driven almost entirely by supply-side factors (Khosla 2011). Dutta et al. (2014) write that “unmet demand for work is the single most important policy-relevant factor in accounting for this gap between actual performance and the scheme’s potential” (p. xxv). Jenkins and Manor (2017) write that while NREGS promises jobs on demand, “many, if not most, poor rural people have little or no experience of making direct demands on authority figures” (p. 69). Similarly, Marcesse (2018) argues that demand itself is affected by incentives of supply agents.
Beyond NREGS: Rural Roads and Other Public Goods In addition to welfare schemes, we take two approaches to evaluate broader impacts on public goods. First, we examine impacts of Scheduled Areas on PMGSY, the Prime Minister’s Village Road Program. This program was established in 2000 to connect rural villages to the all-weather road network by focusing on constructing and upgrading feeder roads that either did not exist or were unpaved (Asher and Novosad 2020). As of 2001, only about half of the 600,000 villages in India were connected to such roads. Importantly, “100 percent funding for construction [under this program was provided] by the Central Government” (ILO 2015). As with NREGS, local politicians are critical to PMGSY’s implementation, whereby a standardized planning process is in place that incorporates representatives from district, block, and village councils. In fact, the key role of local governments in helping carry out construction and maintenance of roads at the local level has been inspired by their success doing the same under NREGS (ILO 2015). Note, however,
742
S. Gulzar et al.
unlike NREGS, PMGSY is a not a rights-based program. While individuals and households apply for job cards and work under NREGS, PMGSY necessitates more centralized planning. Local community inputs are filtered through the gram panchayats, which play a key role in planning for road upgrades and maintenance and whose elected representatives sit on important PMGSY committees (ILO 2015, 24). Coordination for road links across gram sabhas begins at the block level, where committees comprising key officials, including those of the gram sabha, decide on allocations that are finalized at the district level. In addition to the PMGSY program, we also take a more systematic approach to studying effects on public goods outcomes by using data from the 2011 Indian Census – roughly 10 years following the implementation of PESA.
Data Construction To systematically assess how the Scheduled Areas political quota affects development outcomes, we construct a village-level dataset for the nine states that have Scheduled Areas. We begin by using the Socioeconomic High-resolution RuralUrban Geographic Dataset for India (SHRUG) (Asher and Novosad 2019). This dataset allows us to track the same villages over three different Census waves: 1991, 2001, and 2011. SHRUG includes limited Census data from these waves and data on the PMGSY roads program. While SHRUG provides information at the village level, NREGS outcomes are measured at the village-cluster (gram panchayat) level. We utilize a directory from the Indian Water Ministry as a matching directory for village and village-cluster data, and then apply fuzzy matching methods to combine SHRUG and NREGS into a single dataset. We next add information on reservations: both on whether a village falls within or outside of the Scheduled Areas, and whether a village falls within an Assembly Constituency that is reserved for ST, for SC, or not reserved. Our final step is to merge the combined dataset with spatial data, as well as with a more complete set of Census variables than was available from SHRUG, on villages from the 2001 and 2011 Indian Censuses. These additional Census data were procured from InfoMap India. Outcome Variables We use data on three central outcomes of interest from NREGS in 2012–2013: Jobcards are the total number of identification documents issued to prospective workers before they can request to be hired under the program; Worked are the number of households that received work under the program in the year; and Workdays measures the total number of days worked by individuals under the program. These measures were collected at the lowest level for which they are recorded, the village cluster, from the official NREGS portal.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
743
Critically, NREGS data provides all three outcomes for ST, for SC, for those who are neither SC nor ST, separately.11 Fig. 1 shows that there is considerable variation in program implementation across India. We supplement our main analysis by considering effects of the quota on road construction under the PMGSY program (from SHRUG), and on public goods provision (2011 Census, but only available for a subset of villages, called market villages – see Appendix D for details). Scheduled Areas Our key independent variable is an indicator for whether a village is or is not part of the Scheduled Areas. We obtained information on Scheduled Areas status from the Government of India’s Ministry of Tribal Affairs. See Appendix B for data sources. States release official documents either listing specific villages as Scheduled or, where all villages within a block or district are Scheduled, the names of those blocks and districts. While two states list individual village names (Andhra Pradesh and Rajasthan), the remaining states list block and district names. To remain consistent in our coding strategy across states, and to avoid human error that was more likely to occur had we manually coded each village as Scheduled or not in the two states that released information at this level, we elected to code an entire block as Scheduled if any village was designated as Scheduled within the block. Empirically, this approach is conservative because, while it accurately codes Scheduled Areas when all villages in a district and block are inside the treatment area, it codes some untreated villages within a block as treated – that is, the resulting bias will be in the direction of zero. Our coding is illustrated spatially in Fig. 2 and we present a validation exercise in Appendix B. Control variables: Our control variables, as well as the variables we use to evaluate sorting and overtime changes, are sourced from the Census (for 1991, from SHRUG, and for 2001, from the 2001 Census shape files). Summary statistics: We combine 1991, 2001, and 2011 census data with NREGS data, Scheduled Areas coding, and data on the PMGSY roads program. The dataset successfully matches approximately 217,000 of the 274,026 villages (79%) in the sample. Nineteen percent of the villages in our data are coded as belonging to a Scheduled Area. ST comprise about 28% of the population, while SC are only 13% of the population. Nonminorities form the remaining 59%. Appendix C presents summary statistics.
11 While we cannot further decompose the identit(ies) of these non-SC/ST individuals – we do expect that these individuals will be comparatively more likely to be members of the High/Forward Castes, and to have, on average, better economic opportunities. At the very least, in Appendix D we control for district-level Muslim population and find that results are unchanged.
744
S. Gulzar et al.
Fig. 2 Scheduled Areas in India – unit is the block. Outlined regions refer to district boundaries
Empirical Strategy Geographic Regression Discontinuity Consider two proximate villages lying on opposite sides of the Scheduled/nonScheduled boundary. If they are sufficiently similar on observable characteristics, we can say that the only difference between the two villages is that one village lies in a Scheduled Area, while the other is in a non-Scheduled Area. We approximate this thought experiment with a geographic regression discontinuity design that restricts
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
745
attention to villages geographically proximate to a boundary dividing Scheduled Areas and other areas within a state.12 We use the following specification: yvgs ¼ γ Scheduled Areavgs þ as þ f Xvgs , Y vgs þ Z 0vgs ϕ þ óvgs
ð1Þ
8 v s:t: Xvgs , Y vgs ðh, hÞ where yvgs refers to outcomes for village v in gram panchayat g and state s. The official NREGS portal only releases data at the gram panchayat level. In NREGS regressions, all villages in the same gram panchayat are assigned the same outcome value, whereas for Census 2011 and PMGSY, y varies at the village level. Although treatment is assigned at the village level, we cluster standard errors at the gram panchayat level throughout the paper. This has the benefit of correcting for outcome interdependence within the gram panchayat in the NREGS analysis. Scheduled Areavgs is the treatment variable that equals 1 if a village is coded as being in a Scheduled Area, and 0 otherwise. Outcomes that are left-skewed are logged such that γ can be interpreted in percentage terms. State fixed effects as account for any state-level shocks, including the different timing of PESA implementation. f(Xvgs, Yvgs) is a flexible smooth function in two dimensions, latitudes (X) and longitudes (Y ).13 Adding these geographic controls helps the regression absorb spatial trends that might be superfluously driving results. For each village, we calculate distance in kilometers h to a Scheduled Areas border within the same state so that we may compare villages that provide the closest approximation to random assignment. Based on bandwidth selection algorithms (see Appendix Table A3), we take a conservative bandwidth of 10 kilometers as our standard bandwidth (h). Ten kilometers is about one-fifth the size of the median distance (54.4 km), and about one-ninth the mean distance (91.3 km), from the border in the data (see Appendix Fig. A7). Last, we include a vector of all villagelevel Census 2011 indices as well as 1991 and 2001 SC and ST population shares, Z0vgs : Throughout the analysis, we conduct various robustness tests, including varying bandwidths and functional forms, and considering alternate transformations of outcomes. For our main NREGS outcomes, we also report Conley (1999) standard errors that account for spatial spillovers.
Analysis of Balance with Census Data With pretreatment census data at the village level from 2001 and population data from 1991, we analyze balance by evaluating if Scheduled Area predicts census variables. To manage the vast number of 2001 census variables, we collapse the
12 13
Appendix D presents the OLS results. Following Dell (2010), we use the functional form: x þ y þ x2 þ y2 þ xy þ x3 þ y3 þ x2y þ xy2.
746
S. Gulzar et al.
140 variables into 14 substantively meaningful indices by taking the simple mean of their standardized values (see Appendix H). Overall, we find that the geographic RD model yields good balance between Scheduled and non-Scheduled Areas. While we are able to tell the two groups apart in some cases statistically because of the large sample size, the substantive differences across Scheduled and non-Scheduled Areas are small: only three indices, water, urbanization, and banking, exceed 0.1 standard deviations. Even in these cases, the differences stay below 0.22 and two of the three larger differences are positive suggesting that development is higher in Scheduled Areas on these dimensions, which should bias any treatment effects toward zero. More substantively, the differences we do observe tend toward zero as the bandwidth of analysis shrinks; for all variables we can trace across the 1991 and 2001 census waves, there is little reason to believe that the baseline differences are trending differently over time in Scheduled versus non-Scheduled Areas, indicating that controlling for level differences between Scheduled and non-Scheduled Areas may be sufficient, and sensitivity analysis shows that the magnitude of correlations for confounders would need to be much larger than those observed for the three most imbalanced indices for omitted variables to be an important source of bias in treatment estimates. Nevertheless, in our analysis below, we control for population shares and all 14 2001 indices, both imbalanced and balanced, and conduct additional robustness tests that further ameliorate concerns that imbalance might drive our observed treatment effects. See Appendices C and D for details.
The Impact of Scheduled Areas Impacts on the National Rural Employment Guarantee Scheme (NREGS) Table 2 presents the main results on NREGS outcomes. The first column shows treatments effects at the extensive margin, while the remaining three columns decompose this effect across ST, SC, and Non-SC/ST categories. Our first finding is that NREGS outcomes improve substantially for STs. As shown in column 2, 20.7% (p < 0.01) more job cards are issued to STs in Scheduled Areas. This result carries forward to the number of households that receive work during the year through NREGS – the coefficient reflects a 20.4% (p < 0.01) increase. Overall, the number of workdays STs receive increases by 24.2% (p < 0.01), a jump of about 1040 more days of work.14 Second, there is strong evidence that non-SC/STs are the main losers, as shown in column 4. Not only does this group receive 9.4% (p < 0.01) fewer job cards in Scheduled Areas, they also suffer a reduction in the number of households employed (8.1%, p < 0.01) as well as 14
Appendix E shows how improved employment for ST might be particularly beneficial for ST women.
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
747
Table 2 The effect of Scheduled Areas on NREGS (10 km RD) (1) Total Panel A: Log Jobcards Scheduled areas
0.001 (0.014) [0.017] Control mean (unlogged) 652.979 # GPs 14,933 # Villages 32,641 Panel B: Log households worked Scheduled areas 0.011 (0.023) [0.034] Control mean (unlogged) 220.579 # GPs 14,933 # Villages 32,641 Panel C: Log workdays Scheduled areas 0.006 (0.036) [0.054] Control mean (unlogged) 9748.164 # GPs 14,933 # Villages 32,641
(2) STs
(3) SCs
(4) Non-SCs/STs
0.207*** (0.025) [0.037] 259.373 14,933 32,641
0.046 (0.031) [0.039] 92.768 14,933 32,641
0.094*** (0.024) [0.031] 300.838 14,933 32,641
0.204*** (0.029) [0.042] 98.339 14,933 32,641
0.023 (0.031) [0.040] 29.806 14,933 32,641
0.081*** (0.029) [0.039] 92.435 14,933 32,641
0.242*** (0.046) [0.065] 4306.585 14,933 32,641
0.000 (0.053) [0.066] 1259.986 14,933 32,641
0.115** (0.045) [0.062] 4181.593 14,933 32,641
Notes: *p < 0.1, **p < 0.05, ***p < 0.01. Standard errors clustered by GP in parantheses. Conley (1999) standard errors in brackets
the total number of workdays (11.5%, p < 0.05). Third, we find no evidence that SCs are worse off under Scheduled Areas: The point estimates on all variables are substantively small and are not statistically distinguishable from zero. Finally, putting these results together in column 1, there is no evidence that Scheduled Areas affect the extensive margin of program implementation – the total amount of work remains the same across Scheduled and non-Scheduled Areas, as the point estimates on outcomes are small, ranging from 1% on households worked to 0% on jobcards.15 Robustness In Appendix D we show that our results are robust to a number of tests, including various functional forms, bandwidths, spatial grid fixed effects, transformations of outcomes, restrictions to balanced subsamples, and controls for the number of matched villages. In addition, we take a few different approaches to allay concerns about omitted variables explaining results. First, standardized treatment
Appendix G shows that we find no evidence for two alternative explanations: discrepancies in reporting, and differences in reliance on centralized government-explained effects.
15
748
S. Gulzar et al.
effects are larger than, or at the very least equal to, the degree of standardized imbalance; further, we show that once population levels are controlled for, controlling for imbalanced 2001 Census barely affects point estimates, suggesting that Census indices explain minimal variation in the outcomes. Second, treatment effects are similar if we match villages on all census indices, population shares, a spatial function, as well as distance to the border.
Impacts on the Rural Roads Program (PMGSY) Are there implications of instituting electoral quotas beyond the effects we observe on NREGS? Finding evidence of broader impacts will improve our confidence that the institution of Scheduled Areas improved the lives of poor communities. It would also help allay the concern that changes in NREGS come at the cost of changes in other programs.16 We first consider impacts on the PMGSY roads program. Column 1 of Table 3 shows that villages in Scheduled Areas are about 4 percentage points more likely to have completed roads through the program using our geographic RD specification. An important feature of the PMGSY data is the time variation in road construction, which, along with state-by-state variation in the implementation of PESA elections, affords us the opportunity to study the impacts of Scheduled Areas on roads before and after the introduction of electoral quotas. Using village, year, and year since PESA elections fixed effects, a difference-in-differences strategy allows us to consider within-village changes in PMGSY implementation over time. In column 2, where we continue looking only at the 10 km geographic RD sample, we find that villages in Scheduled Areas are 1 percentage point more likely to have a PMGSY road after the introduction of PESA elections, an effect size of about 20% compared against the non-Scheduled Area mean of 0.05. This effect increases to nearly 5 percentage points (an increase of about 90%) using the full dataset of villages. Additionally, Fig. 3 shows that road sanctioning increased in Scheduled Areas soon after the introduction of elections but not before. Effects on the completion of roads followed after a few years. How are roads distributed across identity categories? Unlike NREGS, where distributional impacts can cleanly be estimated as the data are disaggregated by identity category, roads are a nonexcludable good and therefore are harder to match to the identity of beneficiaries. However, as shown in columns 4 and 5 of Table 3 which considers within gram panchayat variation, while villages with an ST Minority are more likely to receive a road in Scheduled Areas after PESA, ST plurality villages are even more likely to have a PMGSY road completed.17 Though these are heterogeneous effects, they indicate that ST politicians treat roads similarly to NREGS, channeling resources to ST.
16
For example, quota politicians may prefer NREGS relative to other priorities because NREGS allows them to perpetuate patronage through handout of state resources (Marcesse 2017). 17 We define ST Pluralityv ¼ 1 [ST popv max (SC popv, non SC/ST popv)].
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
749
Table 3 The effect of Scheduled Areas on rural roads (PMGSY) Outcome: Model: Sample: Scheduled areas Sch areas post PESA election
Road ¼ 1 Geo RD Diff-in-Diff Diff-in-Diff 10 km 10 km Full (1) (2) (3) 0.029*** (0.005) 0.010*** 0.049*** (0.003)
(0.002)
0.051 14,933 32,641 456,974 – Yes No Yes Yes
0.055 93,875 206,364 2,889,096 – Yes No Yes Yes
Sch areas post PESA ST plurality Non-scheduled mean # GPs # Villages # Observations Geo RD controls Village FE Gram panchayat FE Year FE Year of PESA FE
0.127 14,933 32,641 32,641 Yes – – – –
Diff-in-Diff Diff-in-Diff 10 km Full (4) (5)
0.002
0.036***
(0.004) 0.013**
(0.002) 0.017***
(0.006) 0.051 14,933 32,641 456,974 – No Yes Yes Yes
(0.003) 0.055 93,875 206,364 2,889,096 – No Yes Yes Yes
Notes: *p < 0.1, **p < 0.05, ***p < 0.01. Standard errors clustered by GP. Models include all constituent terms of interactions
Impacts on Public Goods Guided by the literature on the responsibilities of local governments in India detailed in the Background section, we also evaluate the effect of Scheduled Areas more broadly on public goods using data from the 2011 Census. We construct six mean indices that take the average of binary indicators on the presence of particular public goods in a village, such as whether there is a gravel road. These indices measure the average provision of roads, water, irrigation, electricity, communication, and education. Similarly, an overall public goods index averages all individual public goods indicator variables. Overall, the results presented in Fig. 4 show that public goods provision in Scheduled Areas improved by 2011, particularly in terms of road, water, communication, and education access. The results on roads are consistent with our earlier results: We see positive treatment effects on the most local kinds of roads, gravel roads, and projects that are targeted specifically by the NREGS program.18
18
Appendix F presents several robustness exercises.
-.1
-.05
Water Index
Tank, Pond, Lake (Y/N)
Fig. 4 The effect of Scheduled Areas on public goods (Census 2011) .05
.1
.15
Years Since PESA (residualized)
5 10
5 10
College (Y/N)
Senior Secondary School (Y/N)
Secondary School (Y/N)
Middle School (Y/N)
Primary School (Y/N)
Education Index
Mobile coverage (Y/N)
Telephone (Y/N)
0
Communications Index
0
Road Completed (residualized)
0
Electricity for Commercial Use (Y/N)
Non-Scheduled Areas
Electricity for Domestic Use (Y/N)
Electricity for Agriculture Use (Y/N)
Electricity Index
River/Canal (Y/N)
Tubewell/Borehole (Y/N)
Uncovered Well (Y/N)
Covered Well (Y/N)
.1 -5
Tap Water Treated (Y/N)
Handpump (Y/N)
Gravel Road (Y/N)
All weather road (Y/N)
Other District Road (Y/N)
Major District Road (Y/N)
State Highway (Y/N)
National Highway (Y/N)
Roads Index
Overall Public Goods Index
.05 -5
Irrigation Index
0
Treatment Effect 0
.05
.1
.15
.2
Road Sanctioned (residualized)
750 S. Gulzar et al.
Years Since PESA (residualized)
Scheduled Areas
Fig. 3 The effect of the introduction of PESA Election on PMGSY roads. (Notes: This figure plots binned means of PMGSY roads by Scheduled Area status on a dataset of all villages that are residualized for village fixed effects)
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
751
Discussion: Bringing the Results Together Evaluating Hypotheses In light of these results, we return to the competing hypotheses suggested by the literature that we summarized in Table 1. First, we consider the NREGS results which allow us to study distribution of resources across identity categories. We do not find any evidence for the solidarity hypothesis: Increased descriptive representation for one minority group, STs, does not appear to improve outcomes for a nontargeted minority group, SCs. We find support for the crowding-out hypothesis to the extent that there is negative substitution away from the residual non-SC/ST group, which is consistent with the aims of programs designed to redistribute economic and political power.19 Importantly, there is no evidence for crowding out occurring through worsening outcomes for SCs. Overall, we find no evidence of changes to NREGS implementation in Scheduled Areas at the extensive margin. However, evidence on PMGSY and public goods show improvements across the board. How might we square these contrasting results in light of the performance hypothesis? One interpretation consistent with results and the literature (for example, Duflo and Chattopadhyay [2004]) is that marginalized politicians empowered under Scheduled Areas invest more in policies prioritized by their communities. While NREGS allows for efficient targeting to individuals as shown in improved ST outcomes in Table 2, delivering benefits to in-group members under less excludable roads and public goods programs might require more effort and thus gains at the extensive margin. In that sense, effects across the programs potentially reflect greater investment in the welfare of marginalized communities. Policy Implications Importantly, the results run contrary to the expectations of affirmative action skeptics: While we do not find that politicians from underrepresented groups outperform other politicians on NREGS, they certainly do not perform worse, and they perform better on a program (PMGSY) where explicit targeting of benefits to marginalized communities is less possible. In addition, gains for the targeted group under NREGS do not come at the expense of similarly marginalized populations. One might still be concerned that Scheduled Areas will tip the scales too far in favor of ST. To put the policy takeaway in perspective, we compare NREGS benefits relative to identity group population shares in Fig. 5 (Girard 2018). First, we find that Scheduled Areas do not move outcomes much on the extensive margin, where non-Scheduled Area share is mechanically equal to 1. Second, despite being a large share of the population in our study area, ST receive far fewer benefits relative to their population share under the status quo. However, Scheduled Areas close this gap. Third, SC receive more benefits than their comparatively small population
19
Data on homicides against ST do not support the theory that crowding out could be driven by privileged groups opting out of NREGS work due to social stigma. See Appendix F.
S. Gulzar et al. 1
752
Ove Overall
Scheduled Tribes Tribes
Non-SC/ST
.6 .4 .2
erall Overall
Scheduled Tribes Tribes
Scheduled Castes
Non-SC/ST
Workdays
0
.2
.4
.6
.8
1 0
Share
Scheduled Castes
Households Worked
.8
1 0
.2
.4
.6
.8
Jobcards
Overall
Share of:
Scheduled Tribes Population
Scheduled Castes Non-Scheduled Area
Non-SC/ST Scheduled Area
Fig. 5 The effect of Scheduled Areas on NREGS relative to population shares. (Notes: This chart shows how the share of NREGS benefits correlates with population shares of ST, SC, and non-SC/ST. All data come from 10 km bandwidths around the cutoffs. Non-Scheduled Area bars represent simple means in the data, while Scheduled Area bars are calculated by using the estimated treatment effects in Table 2. Population shares are from the 2001 census)
share. Fourth, while Scheduled Areas redistribute work away from non-SC/ST, this change results in an improved alignment of benefits with population share. Importantly, it is worth noting that population shares are conservative benchmarks in this analysis because the share of NREGS eligible population is likely much higher for SC and ST groups than for the non-SC/ST. In that sense, while Scheduled Areas appear to be closing the gap, more work might remain to align benefits with need.
Investigating the Electoral Mechanism To what extent does an electoral mechanism explain how the Scheduled Areas have improved development outcomes for ST? We present four pieces of evidence in support of an electoral mechanism.
Scheduled Areas Prior to PESA As discussed above, prior to the implementation of electoral quotas, Scheduled Areas and non-Scheduled Areas looked very similar in our geographic RD analysis
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
753
of the 2001 Census. This analysis mirrors a critical 1995 report by the Indian Parliament–appointed Bhuria Commission, which found little to no devolution of governance and authority to tribal bodies in Scheduled Areas, and argued that tribal populations should enjoy greater self-governance and less governmental administrative interference. . . . since planned development has been an article of faith with us, it has to be ensured that implementation of the policies and programmes drawn up in tribal interest are implemented in tribal interest. Since, by and large, the politico-bureaucratic apparatus has failed in its endeavour, powers should be devolved on the people so that they can formulate programmes which suit them and implement them for their own benefits.
Policies following from these findings were made into law via PESA, passed in 1996 and going into effect with state panchayat elections from 2000. In this way, PESA gave the Scheduled Areas teeth that they had theretofore lacked.
Local Elections in Scheduled Areas The PMGSY program, for which we have a village panel dataset, provides the most direct evidence consistent with an electoral mechanism that is consistent with the historical discussion above. Figure 3 shows that while Scheduled and non-Scheduled Areas followed parallel trends in PMGSY implementation before PESA elections were held, it was only after Scheduled Areas got “teeth” through the introduction of PESA elections that outcomes in Scheduled Areas differentially improved. In contrast, because we only observe NREGS outcomes at a single point in time, we lack any within-state variation in PESA introduction when considering these outcomes. With this limitation in mind, in Appendix F we interact the Scheduled Areas indicator with the number of elections between 2000 and 2012 that have taken place in a state under PESA: either one, two, or three. We find that the main results hold up but the magnitude decreases over time. This may be consistent with quotas having the greatest marginal impact for the targeted minority in the first election, where quotas constitute a shock to political representation. In subsequent elections, as members of the targeted minority “catch up” to other groups, and as quota politicians learn the intricacies of the position, quota politicians may distribute more benefits to other groups and deliver overall gains.
Targeted Minority Electoral Influence Prior work suggests that quota effects are largest where the targeted minority group constitutes a large share of the population (Chin and Prakash 2011; Jensenius 2015; Pande 2003; Das et al. 2017). For instance, Jensenius (2015) reports that some SC politicians want to divert funds to SC constituents but do not do so “because they are scared of being branded as ‘too SC”’ (p. 203) by the majority of voters who are
754
S. Gulzar et al.
non-SC and on whose votes they depend. We expect the method of targeting to differ depending on whether individuals can be excluded from program benefits. When a vote-maximizing politician cannot exclude individuals from receiving a good such as under the PMGSY program, they may want to target gains to areas where in-group members constitute a larger portion of the population and thus where more group members will receive the good. Indeed, evidence in columns 4 and 5 of Table 3 is consistent with this interpretation. Alternatively, where benefits are individually targetable such as in programs like NREGS, it may be in a politician’s interest to deliver benefits to in-group individuals who stand to gain the most marginal benefit. We proxy for this by creating an indicator variable for whether ST are a nonplurality: ST Minorityv ¼ : ST Pluralityv. Table 4 presents heterogeneous effects with our standard RD specification. There are two findings. First, for each of the three main outcomes under NREGS, we find that Scheduled Areas have a larger positive effect for ST in places where ST comprised an electoral minority prior to the implementation of PESA. Second, as before, the negative spillover on the residual non-SC/ST category is also more pronounced in these areas.
Quota Overlap A certain proportion of State Assembly seats across India are reserved for minorities including ST and SC based on population (Jensenius 2012). Although the higherlevel quotas are not randomly assigned, we can use them to investigate quota overlap at different levels of government. On the one hand, multiple quota politicians should reinforce the effect of political quotas by improving potential coordination between politicians who share an identity. On the other hand, there could exist some diminishing returns to quota politician effort because of credit claiming difficulties and free riding problems (Gulzar and Pasquale 2017). While our main results are robust to controlling for the incidence of Assembly Constituency–level ST reservation (see Appendix F), in Table 5 we interact these higher-level reservations in the latest election before 2013 with the Scheduled Areas treatment indicator to study if overlapping Assembly Constituency reservations moderate effects on program implementation. The results show that Scheduled Areas reservations and Assembly Constituency reservations for ST, separately, improve NREGS program implementation tremendously for ST. However, when the two quotas overlap, the overall implementation of the program is less than the separate parts, suggesting that there exist some ceiling effects.20 There is also evidence that overlapping quotas may similarly affect implementation of PMGSY. Overall, the results are consistent with program implementation varying with political institutions.
Interestingly, we find that when the Assembly Constituency reservation is for SC, there are no negative quota overlap effects for ST (see Appendix F).
20
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
755
Table 4 Treatment effects by ST plurality versus minority (10 km RD)
Panel A: Jobcards Scheduled areas ST minority Scheduled ST minority Control mean (unlogged) Panel B: Worked HH Scheduled areas ST minority Scheduled ST minority Control mean (unlogged) Panel C: Workdays Scheduled areas ST minority Scheduled ST minority Control mean (unlogged) # GPs # Villages
(1) Total
(2) STs
(3) SCs
(4) Non-SCs/STs
0.011 (0.017) 0.039* (0.021) 0.024 (0.022) 652.979
0.035 (0.030) 0.139*** (0.038) 0.388*** (0.044) 259.373
0.057 (0.044) 0.131*** (0.042) 0.026 (0.048) 92.768
0.068* (0.035) 0.105*** (0.034) 0.058 (0.039) 300.838
0.014 (0.027) 0.012 (0.037) 0.008 (0.037) 220.579
0.063* (0.036) 0.053 (0.043) 0.318*** (0.049) 98.339
0.001 (0.042) 0.134*** (0.040) 0.053 (0.047) 29.806
0.058 (0.039) 0.183*** (0.042) 0.050 (0.045) 92.435
0.023 (0.043) 0.017 (0.060) 0.064 (0.059) 9748.164 14,933 32,641
0.057 (0.053) 0.174** (0.071) 0.420*** (0.078) 4306.585 14,933 32,641
0.051 (0.072) 0.246*** (0.072) 0.111 (0.081) 1259.986 14,933 32,641
0.070 (0.060) 0.279*** (0.068) 0.098 (0.072) 4181.593 14,933 32,641
Notes: *p < 0.1, **p < 0.05, ***p < 0.01. Standard errors clustered by GP
Conclusion Policymakers often treat economic and political efforts in isolation. We argue that political affirmative action and development programs may serve as complementary levers to deliver better outcomes for marginalized communities, at no cost to other minorities, nor to society overall. Our empirical setting is political affirmative action in India, where Scheduled Areas, as well as similar reservations more generally, are hotly debated and politically divisive. Protests and riots have broken out for a myriad of related affirmative action issues – out of fear of reductions in protections for SC and ST throughout India, in anticipation of the implementation of elections in Scheduled Areas, by groups agitating for inclusion in identity categories targeted by quotas, and in an
756
S. Gulzar et al.
Table 5 Additional quota in Assembly Constituency (10 km RD)
Panel A: Job cards Scheduled areas AC reserved, ST Scheduled AC reserved, ST Control mean (unlogged) # GPs # Villages Panel B: Households worked Scheduled areas AC reserved, ST Scheduled AC reserved, ST Control mean (unlogged) # GPs # Villages Panel C: Workdays Scheduled areas AC reserved, ST Scheduled AC reserved, ST Control mean (unlogged) # GPs # villages
(1) Total
(2) ST
(3) SC
(4) Non-SC/ST
0.011 (0.019) 0.203*** (0.017) 0.015 (0.027) 652.979 14,933 32,641
0.405*** (0.041) 0.883*** (0.035) 0.482*** (0.051) 259.373 14,933 32,641
0.098** (0.046) 0.098** (0.044) 0.104* (0.062) 92.768 14,933 32,641
0.125*** (0.031) 0.138*** (0.032) 0.075 (0.046) 300.838 14,933 32,641
0.021 (0.033) 0.457*** (0.030) 0.096** (0.044) 220.579 14,933 32,641
0.387*** (0.049) 0.920*** (0.040) 0.464*** (0.060) 98.339 14,933 32,641
0.108** (0.047) 0.019 (0.044) 0.138** (0.062) 29.806 14,933 32,641
0.107** (0.042) 0.055 (0.041) 0.034 (0.057) 92.435 14,933 32,641
0.051 (0.051) 0.597*** (0.047) 0.029 (0.068) 9748.164 14,933 32,641
0.419*** (0.076) 1.300*** (0.061) 0.519*** (0.094) 4306.585 14,933 32,641
0.176** (0.078) 0.001 (0.073) 0.293*** (0.105) 1259.986 14,933 32,641
0.186*** (0.060) 0.094 (0.061) 0.103 (0.086) 4181.593 14,933 32,641
Notes: *p < 0.1, **p < 0.05, ***p < 0.01. Standard errors clustered by GP
effort to extend Scheduled Areas into new jurisdictions (AlJazeera 2018; Iyengar 2015). Despite their importance, scale, and salience, Scheduled Areas remain understudied in political science and related disciplines. To our knowledge, this paper provides the first systematic evaluation of this institution. We propose a novel theoretical framework comprising solidarity, crowding-out, and performance hypotheses to understand the systematic effects of political affirmative action across groups. To test these, we build a new large-scale dataset combining administrative data on the largest employment program in the world, a rural roads program, as well as public goods from the Indian Census. We find that quotas deliver no worse outcomes overall and that gains for targeted minorities come
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
757
at the cost of the relatively privileged, rather than other historically disadvantaged groups. More broadly, improvements in other pro-poor policies, including a rural roads program and general public goods, further attest to the complementary impacts of political affirmative action and pro-poor economic development. Effects appear to operate through an electoral mechanism. They appear strongly (1) after the introduction of local elections with reservations for minorities, (2) in places where we would expect vote-maximizing politicians to target benefits, and (3) where there is no overlap with other quotas targeting the same minority. What are the implications of our results on debates surrounding affirmative action? Skeptics routinely argue that open competition in the political sphere brings the best politicians to the fore. However, our results show that quota politicians perform no worse than status quo politicians. This suggests that status quo institutions may prevent equally qualified individuals from marginalized communities from running for office and more effectively representing their communities. We hope other researchers add to these findings by further disaggregating identities of beneficiaries, as well as adding evidence on other links in the causal chain from affirmative action to development, from district- and block-level actors through village councils and to individual households. What are the long-term consequences of electoral affirmative action? Our study measures impact up to 12 years after implementation of the institution and finds large positive effects for the targeted minority, which results in a closer alignment of the distribution of benefits with identity category population share. However, one concern for the longer term is that fixed political affirmative action may develop its own unequal political structures by simply replacing which identity group is on top. Efforts that helpfully redistribute political power initially could create long-run political monopolies. We consider this to be an important open question to be explored in the future.
Cross-References ▶ Experimental Evidence on Affirmative Action ▶ Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need of a Rethink ▶ Stratification Economics ▶ The Economics of Discrimination and Affirmative Action in South Africa ▶ The Effectiveness of Affirmative Action Policies in South Africa Acknowledgments The authors would like to thank Ingo Rohlfing and three anonymous reviewers at the American Political Science Review, Ramnarayan Bhagat, Rachel Brulé, Lauren Davenport, Miriam Golden, Justin Grimmer, Maira Hayat, Clément Imbert, Hakeem Jefferson, Francesca Jensenius, David Laitin, Stephane Lavertu, Daniel Masterson, Durgesh Pathak, Vijayendra Rao,Cyrus Samii, David Stasavage, Milan Vaishnav, and seminar participants at Gothenburg, ISI Delhi, Lahore School of Economics, NYU, Wisconsin-Madison, MPSA, OSU, Oxford, and Uppsala for helpful comments.
758
S. Gulzar et al.
References Adida CL, Davenport LD, McClendon G (2016) Ethnic cueing across minorities a survey experiment on candidate evaluation in the United States. Public Opin Q 80(4):815–836 Akerlof GA, Kranton RE (2010) Identity economics: how our identities shape our work, wages, and well-being. Princeton University Press, Princeton AlJazeera (2018) Dalits in India hold protests against ‘dilution’ of SC/ST Act. AlJazeera. https://www. aljazeera.com/news/2018/04/dalits-india-hold-protests-dilution-scst-act-180402082213061.html Asher S, Novosad P (2019) Socioeconomic high-resolution rural-urban geographic dataset for India (SHRUG). https://doi.org/10.7910/DVN/DPESAK Asher S, Novosad P (2020) Rural roads and local economic development. Am Econ Rev (forthcoming) Beaman L, Duflo E, Pande R, Topalova P et al (2010) Political reservation and substantive representation: evidence from Indian village councils. In: India policy forum 2010/11 volume, vol 7. Brookings Institution Pres and NCAER, pp 159–191 Bertrand M, Hanna R, Mullainathan S (2010) Affirmative action in education: evidence from engineering college admissions in India. J Public Econ 94(1–2):16–29 Besley T, Pande R, Rao V (2007) Political economy of panchayats in South India. Econ Polit Wkly 42(8):661–666 Bhavnani RR (2009) Do electoral quotas work after they are withdrawn? Evidence from a natural experiment in India. Am Polit Sci Rev 103(01):23–35 Bhavnani RR, Lee A (2019) Does affirmative action worsen bureaucratic performance? Evidence from the Indian administrative service. Amn J Polit Sci (Early View) 65:5–20 Broockman DE (2013) Black politicians are more intrinsically motivated to advance blacks interests: a field experiment manipulating political incentives. Am J Polit Sci 57(3):521–536 Cassan G, Vandewalle L (2017) Identities and public policies: unintended effects of political reservations for women in India. IHEID working papers, 18-2017 Census (2011) Office of the Registrar General and Census Commissioner of India, Ministry of Home Affairs, Government of India, Available at: https://censusindia.gov.in/census.website/ Chin A, Prakash N (2011) The redistributive effects of political reservation for minorities: evidence from India. J Dev Econ 96(2):265–277 Conley TG (1999) GMM estimation with cross sectional dependence. J Econ 92(1):1–45 Corbridge S (2002) The continuing struggle for India’s Jharkhand: democracy, decentralisation and the politics of names and numbers. Commonwealth Comp Polit 40(3):55–71 Das S, Mukhopadhyay A, Saroy R (2017). Efficiency consequences of affirmative action in politics: evidence from India. IZA discussion paper No. 11093 Dasgupta A, Kapur D (2020) The political economy of bureaucratic effectiveness: evidence from local rural development officials in India. Am Polit Sci Rev 114(4):1–19 Dell M (2010) The persistent effects of Peru’s mining Mita. Econometrica 78(6):1863–1903 Deshpande A, Weisskopf TE (2014) Does affirmative action reduce productivity? A case study of the Indian railways. World Dev 64:169–180 Duflo E, Chattopadhyay R (2004) Women as policy makers: evidence from a randomized policy experiment in India. Econometrica 72(5):1409–1443 Dunning T, Nilekani J (2013) Ethnic quotas and political mobilization: caste, parties, and distribution in Indian Village councils. Am Polit Sci Rev 107(1):3556 Dutta P, Murgai R, Ravallion M, van de Walle D (2014) Right to work: assessing India’s employment guarantee scheme in Bihar. World Bank Publications, Washington, DC Gille V (2013) Stigma in positive discrimination application? Evidence from quotas in education in India. Conference paper, Centre d’Economie de la Sorbonne Gille V (2018) Applying for social programs in India: roles of local politics and caste networks in affirmative action. J Comp Econ 46(2):436–456 Girard V (2018) Don’t touch my road. Evidence from India on affirmative action and everyday discrimination. World Dev 103:1–13 Gulzar S, Pasquale BJ (2017) Politicians, bureaucrats, and development: evidence from India. Am Polit Sci Rev 111(1):162–183
32
Does Political Affirmative Action Work, and for Whom? Theory and. . .
759
ILO (2015) ILO-PMGSY rural road project strengthening capacity of Panchayati Raj Institutions: managing maintenance of rural roads. http://pmgsy.nic.in/RSCPR.pdf Iyengar R (2015) Riots break out in India over a dominant caste’s attempt to gain ‘backward’ status. Time. https://time.com/4011001/hardik-patel-/ Iyer L, Mani A (2012) Travelling agents: change and bureaucratic turnover in India. Rev Econ Stat 94(August):723–739 Jenkins R, Manor J (2017) Politics and the right to work: India’s National Rural Employment Guarantee Act. Oxford University Press, Oxford Jensenius FR (2012) Political quotas in India: perceptions of constituent political representation. Asian Surv 52(2):373–394 Jensenius FR (2015) Development from representation? A study of quotas for the scheduled castes in India. Am Econ J Appl Econ 7(3):196–220 Jensenius FR (2017) Social justice through inclusion: the consequences of electoral quotas in India. Oxford University Press, Oxford Kapur D, Nangia P (2015) Social protection in India: a welfare state sans public goods? India Rev 14(1):73–90 Kaufmann KM (2003) Cracks in the rainbow: group commonality as a basis for Latino and AfricanAmerican political coalitions. Polit Res Q 56(2):199–210 Khosla R (2011) Caste, politics and public good distribution in India: evidence from NREGS in Andhra Pradesh. Econ Polit Wkly 46(12):63–69 Krook ML, Zetterberg P (2014) Electoral quotas and political representation: comparative perspectives. Int Polit Sci Rev 35(1):3–11 Maiorano D (2014) The politics of the mahatma Gandhi national rural employment guarantee act in Andhra Pradesh. World Dev 58:95–105 Marcesse T (2017) Patronage guaranteed? Public policy reform, patronage and party–building in rural India. Working paper Marcesse T (2018) Public policy reform and informal institutions: the political articulation of the demand for work in rural India. World Dev 103:284–296 Meier KJ, McClain PD, Polinard JL, Wrinkle RD (2004) Divided or together? Conflict and cooperation between African Americans and Latinos. Polit Res Q 57(3):399–409 Moffitt R (1983) An economic model of welfare stigma. Am Econ Rev 73(5):1023–1035 Munshi K, Rosenzweig M (2015) Insiders and outsiders: local ethnic politics and public goods provision. National Bureau of Economic Research working paper 21720 Palaniswamy N, Krishnan N (2012) Local politics, political institutions, and public resource allocation. Econ Dev Cult Chang 60(3):449–473 Pande R (2003) Can mandated political representation increase policy Influence for disadvantaged minorities? Theory and evidence from India. Am Econ Rev 93(4):1132–1151 Singh P (2015) How solidarity works for welfare: subnationalism and social development in India. Cambridge University Press, Cambridge Sukhtankar S (2017) India’s National Rural Employment Guarantee Scheme: what do we really know about the World’s largest workfare program? Brookings-NCAER India Policy Forum 13: 231–286
Gender Quotas and Representation in Politics
33
Vidhu Verma
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Democracy, Representation, and Diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quotas As Fast Track to Equal Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Opportunities and Challenges to Gender Quotas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
762 763 766 771 774 776
Abstract
The controversy over affirmative action policies, quotas, or reservations for historical injustice constitutes the most salient current battlefront in the conflict over the status of social justice. No debate has been more protracted or more riddled with complex issues. This chapter addresses three themes about the significance of gender in implementing quotas: The first is gender quotas as a form of affirmative action that reverse discrimination and increase women’s participation in political democracy. The second relates to how female legislators become equal players in the policymaking process through gender quotas. The third is the challenges of intersectionality and the gradual withdrawal of the state from some of these policies. The chapter relies upon various significant findings regarding quotas and affirmative action by addressing these themes. It argues that to make this strategy successful, the context in which quotas are debated and implemented has to be explored. The review of research suggests that quotas for women are broadly effective at achieving political equality and opening new avenues for representation. However, political institutions are gendered; therefore, the adoption of gender quotas is linked to the implementation processes. Even while women have access to quota seats, informal norms dilute the impact of the implementation of gender quotas. Breaking through male-dominated V. Verma (*) Centre for Political Studies, Jawaharlal Nehru University, New Delhi, India © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_43
761
762
V. Verma
institutions often requires the presence of laws attacking gender discrimination. A closer look at explanatory accounts suggest that it might be helpful to encourage women’s participation in many forms of associations and grassroots organizations that can impact their role in decision-making in politics. Keywords
Affirmative action · Gender · Women · Representation · Democracy · Intersectionality · Participation · Quotas · Politics · Equality · Nondiscrimination · Reservations · Policy
Introduction Affirmative action encompasses a variety of policies to benefit members of disadvantaged groups, usually those who have suffered discrimination in the past. Such policies seek to address some form of group disadvantage grounded in the particularities of ethnic difference and inequality prevalent in society. In one sense or another then, these policies relate to any measure that allocates public goods to members of disadvantaged groups who experience a higher risk of discrimination, exclusion and poverty than the general population. Besides, affirmative action has been defended as a form of redress for past exclusion from public domains and is a kind of justice to make up for past wrongs. Most importantly, affirmative action and quotas aim to overcome the underrepresentation of diverse minorities, women, and discriminated groups in society. In recent writings, political theorists present an “integration” argument for affirmative action to overcome the structural exclusion of minorities to serve the ideal of equal opportunity (Anderson 2010). To cultivate leaders with legitimacy, the path to leadership at the national and international level, in arts, sciences, politics, and government, must be open to talented and qualified individuals of every social group in a heterogeneous society (Bowen and Bok 1998). In contrast to these views, affirmative action is associated with a vast array of often inconsistent practices and policies couched in identity politics and reverse “discrimination” (Young 1990). Since the liberal tradition argues for individual rights, the embrace of group rights in affirmative action policies questions the public institutions’ neutrality based on equal rights for all citizens. The character of these programs comes under constant scrutiny in countries like India and the USA, where implementation of affirmative action is viewed as an exception to anti-discrimination laws (Deshpande 2013). In certain countries like Malaysia and South Africa, affirmative action embodies the principles of restructuring and redistributing through preferential treatment to a historically disadvantaged majority community (Adam 1997; Verma 2002). This has led proponents to argue that such action is incompatible with equal opportunity; others have argued that affirmative action does not disturb the mechanisms that generate maldistribution or open up employment and educational opportunities without challenging class stratification under capitalism (Fraser 1998). These policies on preferential treatment are seen to work as a kind of
33
Gender Quotas and Representation in Politics
763
social good by departing from standard meritocratic selection criteria; they confound desert by severing reward from a person’s character, talents, choices, and abilities (Nagel 1973; Thomson 1973). Whatever skepticism may exist about group-based policies in affirmative action, they are here to stay to offset historical and ongoing racial and gender discrimination, caste segregation, and bias. In the background of these debates, this chapter will focus on gender quota laws in politics. A comparative perspective on gender quotas is attempted, given the region- and time-specific nature of most research on women in politics. What is common in the growing body of work on gender quotas is that it interprets affirmative action policies as mechanisms to address women’s interests and identities widely, referred to as substantive representation. It not only addresses claims about the effects of these policies but also aims to enhance understanding of how political identities are constructed and deployed during the policymaking process. The paper argues that gender quotas question a conventional definition of politics related to competing interests to convey the need to study the power dynamics that constitute the personal and the political. Quotas prevent future discrimination or exclusion by ensuring that political institutions do not erect barriers against the entry of women. Although constitutions might grant women citizenship with opportunities and rights, notably through education, welfare, and the right to vote, women face discrimination in public and private life. The impetus toward quotas is twofold: first to maximize opportunities for all, along with its presumed benefits, and second to redress perceived disadvantages due to overt, institutional, or involuntary discrimination. In a narrow sense, quotas are an attempt to promote equal opportunity that focuses on the recruitment of women in political institutions. Redressing disadvantages is to compensate for past discrimination and counteract current discrimination. These measures are historical correctives seeking to undermine and alter discriminatory structures by ensuring that opportunities are distributed equally. In a broader sense, quotas are viewed as affirmative action measures aiming at “equal results” (Bacchi 1999, 101). They raise many fundamental principles of gender justice, including the construction of “women as a politically relevant category,” equal rights of representation in political institutions, and equal share-holders with men in political power (Dahlerup 2007, 6).
Democracy, Representation, and Diversity Political representation is a cardinal feature of democracy, without which rights, liberties, and institutions cannot function. Although value of political equality is central to normative theories of democracy, it is the presence of multiparty system and competitive elections that enhance the descriptive aspect of representation in parliament (Tremblay 2007). In this respect, representation is linked to improving legislative processes to ensure that many groups play a part in debating and commenting on the policymaking process. More specifically, political representation is vital for the electorate as it determines both the way representatives are elected and
764
V. Verma
the relationship between the elected members and their constituents is sustained. Further, it assumes that the government considers a variety of views and interests in the policymaking processes, and the issue of representation acquires greater significance for the articulation of the interests of disadvantaged groups. In the past, women’s lack of political participation in democracies has often been linked to a lack of access to resources or informal networks. An emerging body of literature argues that institutions determine factors in women’s voting, participation in political parties and campaigns, and engagement in protests. Secondly, a democratic regime positively influences women’s political rights and civil liberties and enhances their participation (Viterna and Fallon 2008). However, studies show women can be pretty active against authoritarian regimes, as in Latin America in the anti-Allende and anti-Pinochet movements and during the suppression of minority rights in Egypt and India (Baldez 2002; Mustafa 2020). However, the strengthening of women’s participation in all spheres of life has been an important issue in the discourse on economic development and democracy. Political participation is more complex in representative democracy than in an authoritarian regime, which, in principle, should certify equal treatment of all groups in its institutions (Fallon et al. 2012). As argued above, since the value of political equality is central to normative theories of democracy, women must share equally with men in decision-making. From the 1970s, international norms and organizations supported quotas spread through transnational discourses and advocacy groups. As a result, international institutions implemented policies first in the women in development programs and later in the transformation to gender and development approaches to integrate women into economic and social processes. It was forums like the United Nations First World Conference (1975), the United Nations Convention for the Implementation of All Forms of Discrimination Against Women (CDEAW 1979), and the Beijing Platform for Action (1995) that issued recommendations that led to the awareness of quota policies in many countries. It was discovered that although women in most democracies vote nearly as often as men, they do not run for political offices or positions in government (Gilardi 2015). Even though women joined organizations and expressed themselves in public forums, they were invisible in institutions for making government policies (Krook and Childs 2015). Political systems faced criticism from those who argued that democracies did not represent various social groups such as religious minorities and women (Geisel and Hust 2005). Women who formed more than 50% of the total world population were not so visible in their elected political institutions. It seemed that universal franchises now widely accepted across democracies could not guarantee women’s presence in political bodies; despite their numerical strength, women were poorly represented in political institutions (Kittelson and Schwindt-Bayer 2012). The lack of women in the political arena was due to several factors. In most countries, structural barriers such as entrenched gender roles related to caregivers and negative gender stereotyping limit involving women. Women were also affected by multiple forms of discrimination and resource distribution that restricted their ability to get access to public and political roles. In developing countries of the south, lower educational achievements and the prevalence of social norms severely
33
Gender Quotas and Representation in Politics
765
restricted their participation in public life. Women’s participation was also influenced by several factors such as governmental, institutional, and electoral systems, culture, and strength of civil society. The few women who actively participated were often members of the social and economic elite and from political dynasties. Conceiving women as a discrete group with specific interests raised concerns about reducing gender identity to an essential biological difference from men. Moreover, in countries of Latin America, female politicians, when in power, universally followed gender roles and prioritized caregiving, morality, honesty, and homemaking (Chaney 1979; Craske 1999). Women were increasingly characterized as a politically marginalized group despite the regional and national differences. Even though women had enfranchisement, they were still vastly underrepresented in national assemblies, accounting for just 25% of all national parliamentarians. A steady consensus grew on increasing women’s representation, ranging from government requirements for political parties to support parity between men and women or government-mandated reservations of a certain percentage of seats in an elected body for women. Feminist scholars campaigned for reservations of seats for women in various legislative bodies. Previously feminist scholars focused on women’s lack of resources or will to participate in politics, but now institutional and cultural mechanisms of exclusion that prevent women from decision-making in political institutions are emphasized. Whether legislated or voluntarily imposed by political parties, quotas were now a familiar tool for those serious about improving women’s participation in diverse contexts (Sawer et al. 2006). Some studies also viewed the electoral system’s favorable effect, especially proportional systems’ entry of women into parliament and other variables (Tremblay 2008). In this way, a whole new discourse rapidly emerged, focusing on the mechanisms of exclusion through institutional practices. The problems associated with the traditional approach in politics as the study of the machinery of government and electoral politics led to the creation of stereotypes and homogenization of women as a group. It rendered women invisible in the public sphere despite their importance in the private sphere, a legitimate subject of politics. Moreover, the assumptions about men and women in public and private spheres affected how political actors were defined with decision-making roles. As a result, what counts as political became a narrowly defined activity. The worldwide adoption of gender quota laws to nominate or choose women for office directed scholars’ attention to a new set of questions: Do quotas that help more women get elected to legislatures seek to justify and defend the way power is distributed? Do they demand a new distribution to correspond with conceptions of justice, welfare, or the utility? If more women get elected to public institutions, do female legislators support policies that benefit women in society? Do gender quotas affect outcomes that promote women’s interests? Does changing representatives transform what is represented? Do male legislators change their preferences and behavior because of quotas? This chapter examines some of these questions in the gender and politics scholarship in the following sections. Although there is no clear pattern about the origins
766
V. Verma
and outcomes of gender quota policies, most research reveals the shifts in political representation for women globally. The broad objective is to promote a greater awareness of the challenges of implementing these quotas. Different groups of women experience mechanisms of asymmetrical power and gender discourses in different ways. The differences among women embody social relations that constitute the setting within which quota policies are more effective than others in facilitating women’s access to political office (Lazar 2005; Parpart et al. 2000). In addition, this chapter also points out the need to study how institutionalized power practices are intertwined with normative patterns of behavior and thoughts that help understand how power is created and recreated in the political domain. Subsequently by bringing up factors such as race, class, caste, ethnicity, social status, and age, the earlier perspectives of women that shaped the distinction between the public and private have seen a paradigm shift. Such policies are part of a more significant shift in conventional views on politics. Their implementation helps the diffusion and adoption of epistemological assumptions inherent in feminist scholarship, political theory, and policy analysis.
Quotas As Fast Track to Equal Representation Gender quotas are defined by Carol Bacchi as a “form of affirmative action” aimed at reversing discrimination in law and increasing women’s representation in elected legislative bodies (Bacchi 2006). Drude Dahlerup believes that electoral gender quotas represent the fast track to equal representation of women and men in politics in contrast to the “incremental track” (Dahlerup 2007, 6). She comprehensively defines quotas as not only “party nominations” but also different ways in which countries try to increase political representation of women through affirmative action measures. Gender quotas are numerical targets that propose the number of women that must be included in a list or the number of seats allocated to women in a legislature. More specifically, quotas represent “regulations” in public elections to find a “certain minimum in numbers or percentage of a specific group at one of these levels” (Dahlerup 2007: 19). Gender quotas, according to Dahlerup, regulate political parties and underscore their role as “gatekeepers” through which citizens pursue opportunities for political leadership. Three types of gender quotas in politics emerged as key to women’s presence in institutions: (1) legislated candidate quotas which regulate the gender composition in the political parties that compete for elections; (2) legislated reserved seats that regulate the gender composition of elected bodies either by reserving a percentage of seats or several seats for women; and (3) party quotas that are adopted by each party voluntarily for their parties and are enshrined in their party statutes and rules (Dahlerup et al. 2013). In 2003, as the debate about using gender quotas to increase political participation gained momentum, studies demonstrated that quotas effectively increase the number of women holding office in many countries (Darhour and Dahlerup 2013; Tripp and Kang 2007). Gender quotas in governments across 1185 countries have achieved
33
Gender Quotas and Representation in Politics
767
international recognition to further the political inclusion of women from the statistical evidence accumulated over the last decade (Bush 2011; Krook 2009; Tripp and Kang 2008). Quotas for women have proved a viable strategy to overcome structural obstacles to substantive representation (MDG 2013). But the extensive body of empirical research concluded that quotas had a differential impact in different electoral systems due to their historical experience with elections, the global context of transition to democratic freedoms, and the implementation of quotas (Nir and McClurg 2015). Legislated candidate quotas regulate the gender composition of the candidate lists and are binding by law for all particular parties. Legal candidate quotas apply to all parties in a system since individual parties can adopt the party quotas for women by choice. For example, in France, a legal candidate quota in 2004 required that candidate lists for each election include 50% of each gender. Despite this quota, elections to the National Assembly yielded only 12% of women members in 2002 and 19% of women in 2007. But when the Australian Labor Party (ALP) passed its affirmative action resolution in 1994, the party saw the number of women in the national parliamentary team go up to 50%. In Ireland and Belgium, legislated quotas, according to their constitutions, govern the minimum percentage of women candidates. Indonesia introduced a 30% quota for women candidates in the 2004 general election, and by 2009, only one party did not manage to meet the 30% quota for candidate selection. However, while the selected parties could successfully fulfill the 30% requirements, the election of female candidates at the top of party lists remained low. Only a few female candidates found their names appear at the top (Novi 2010). The legislated reserved seats by law measure the gender composition of elected bodies by reserving a certain number or percentage of seats for women members implemented through special electoral rules. These focus on the number of seats women are to get in parliament and are commonly used in almost 36 countries ranging from Tanzania to Rwanda and Pakistan (Devlin and Elgie 2008). Around 37 countries have reached the critical mass of 30% women in their lower houses of parliament (Dahlerup et al. 2013, 15). However, only four countries have 50% or more women in parliament in single or lower houses; Rwanda has 61%, Cuba and Bolivia with 53%, and the United Arab Emirates with 50% (Inter-Parliamentary Union 2020). The Nordic states have the highest regional average of female parliamentarians. In 2013, these states had 42 percent women in their single or lower houses. The Nordic states have the highest regional average of female parliamentarians. In 2013, these states had 42 percent women in their single or lower houses. But the Arab states, with 17.8% of women, and Asia, with 19.1%, are clearly at the bottom. The global nature of gender quota laws has led to the consolidation of a broad body of scholarship on many aspects of quota implementation. The empirical uncertainty concerns the conditions under which gender quotas encourage women’s participation. Contrary to earlier theoretical assumptions about the relationship between democracy and women’s rights, conservative regimes have been active in implementing quota policies for women (Htun 2003). Although the timing and reasons vary from country to country, creating political, educational, and
768
V. Verma
employment opportunities for women is justified as an essential notion of their roles. Women received public benefits in which their roles in the public economy were reinforced. As Molyneux explains, demands for women’s citizenship were played out through “idealized representations of motherhood and wifely duty. . .the sacred qualities of motherhood could be deployed in the service of society” (Molyneux 2001, 169). In the context of Brazil, Latin America’s largest, most vibrant, and the most diverse feminist movement, studies found that while policies advanced women’s capabilities and opportunities, they remained underrepresented in political office, unlike other states like Argentina and Costa Rica. The Brazilian Congress that approved an electoral quota for female candidates only provided incentive to party elites to support increase in a limited number of female leaders (Miguel 2008). Institutional features related to the weakness of women’s quota law, electoral rules, and clientelistic parties obstructed the path to greater participation in politics (Htun 2002). Still others viewed federalism affecting gender policy mainly through creation of feminist activist networks in the areas of education, foreign policy and health in the USA (Banaszak 2010). The evolution of women’s suffrage was also crucial for understanding contemporary women’s political behavior (Paxton et al. 2020). Despite variations in the historical struggle for women’s rights, the underrepresentation of women in political institutions continues due to gendered spaces and unequal opportunity structures. One of the first cross-national studies of the impact of gender quotas on women’s participation found that quotas had not a significant impact (Zettererg 2009). The global average for women in the lower houses of parliament was only 20% in 2011, with enormous variations between the Nordic countries and the Pacific region. Asia’s economy soared from 12% of global GDP in the 1960s to 31% in 2015. South Asian countries were going through a structural transformation where the young population grew. This growth allowed more women to get educated and have better health, and income. Yet despite these gains, women are poorly represented in the political arena. Women’s subordination in the South Asian region is acknowledged primarily due to structural reasons that include the “embedded system of patriarchy” (MUH 2004, 80). Women mainly belonging to the poor and powerless groups find themselves triply disadvantaged and vulnerable as, on a day-to-day basis, they still struggle to improve the conditions of their lives. The question arises: Do female legislators become equal players in the policymaking process? Those scholars who hold a robust causal connection between gender quotas and democratization of the polity found that the numbers of women do not always influence the outcome of legislation. Research shows that the increasing number of women as legislators is mediated by political variables and institutional rules, including party identification, legislative specialization, and committee structures in the legislature. Although politicians influence political institutions, the latter’s interests are shaped by their social and cultural origins and gender socialization. It is not surprising scholars claim that voluntary quotas do not increase the effectiveness of the number of women who finally gain access to some political office (Gray 2003). Studies support this claim that the number of females influences
33
Gender Quotas and Representation in Politics
769
women’s political engagement in public office rather than the number of women seeking office positions. A female in senior political positions is likely to promote political participation in governing institutions (Alexander and Jalazai 2020; Liu and Banaszak 2017). But women’s representation is likely to depend not merely on female leaders’ presence through quotas but on the institutional arrangements that enable or constrain female leaders’ effectiveness. Representation of women depends on the institutional environment, shaped by quota policies. Gender sensitivity in electoral laws, political parties, and legislative institutions is cited as a major support for making the female legislators’ role more effective. In the case of Argentina and Mexico, both federal systems with 30% quota laws, and high degrees of electoral competition, disciplined political parties played a role in women’s substantive representation for several reasons. In Argentina, a law was adopted in 1991, mandating a 30% gender quota which slowly increased women’s representation in both legislative chambers. Although women’s representation increased in 1994 in the national legislature due to the gender quotas, Mexico usually fails to fill its quota, while the mechanism of affirmative action is more potent in Argentina. In 2009, the Mexican Congress consisted of 23% women compared to 39% in Argentina. However, the policy impacts are more negligible in Argentina than in Mexico as the parties are more ideologically and programmatically coherent in the latter, affecting legislators’ strategies to promote gender policy. Mexico also has commissions that play a role in policymaking (Piscopo 2011). Studies also argued that voluntary quotas are effective when the party that adopts them is liberal and not conservative and where the area is urban and not rural (Bonomi et al. 2013; Fallon et al. 2012). In the context of Brazil, data shows that left-leaning parties promote feminine or feminist agendas more often than left-leaning women meaning that ideologies trump gender identity (Htun and Power 2006). A possible reason for associating gender quotas with leftist ideology is the egalitarian ethos of many socialist parties. This trend is noticeable in countries of Latin America, where it is common for leftist parties to adopt quotas. Several studies in Western democracies have analyzed the role of gender in predicting the level of participation in parliamentary debates (Tamerius 1995). While women legislators are more likely than men to raise issues related to gender equality, an important variable that affects the implementation of quotas is the policy areas where female legislators work. Female legislators are mainly allocated portfolios associated with women’s interests – culture, housing, welfare, children, and special needs. Women’s soft policy areas show how gender-role socialization still influences party leaders or their constituents (Schwindt-Bayer 2006). Traditional distributions in portfolios suggest that female legislators’ agenda-setting activity has a lot to do with how male legislators also transmit women’s interests through programs that encourage motherhood and protect children. In South Africa and Argentina, research has shown that policy outcomes arising out of gender quotas depend not merely on the legislators and their interests but also parties and legislative cultures (Geisler 2000). If the legislative environment is patriarchal, female legislators are silenced, and they cannot raise policy issues that men view as irrelevant. Sometimes female MPs supporting women’s rights are
770
V. Verma
verbally assaulted in male-dominated legislatures or dismissed as crazy or immature (Piscopo 2008). In contrast to the studies on Brazil, entrenched gender bias in the legislature trumps party ideologies and intentions. Female legislators are less successful in raising agendas or bringing their bills to floor votes due to the decreased priority given to their soft policy areas. Another set of studies has associated a negative approach toward quota policies claiming that they give an unfair advantage to minority groups like women. There is an assumption that such policies counter the trend toward a meritocratic society. Even though quotas have been implemented for more than a decade, studies show that men in the Flemish political system were strongly opposed to gender quotas. Men believed that the underrepresentation of women was linked to the fair rewards they received from the system for their efforts. The belief that the distribution of power and wealth is based on equal access to the current system leads many people to argue that quota policies for women were unfair and unmeritocratic (Meier 2008). Electoral quotas, or reservations as they are called in India, were introduced for women at the local government mandated by the 73rd and 74th constitutional amendments of 1993. At least one-third of the elected positions were reserved exclusively for female candidates in the local, rural panchayats and urban municipal councils. The Indian government encouraged women to compete inside the quota system at the local level but did not find it acceptable to have a similar quota in parliament. Although women continue to be underrepresented in the national and state assemblies, the women’s reservation bill for parliament encountered vehement opposition and has failed to pass since it was first introduced in 1996 (Verma 2012). But many scholars are astounded by the remarkable increase in female voter turnout rates in the last few general elections. In addition, many applaud how female executives in local governing institutions increased the overall women’s participation. As women representatives gain skills in local governance, their experience has been recorded by scholars. In India, voting districts mandated to have a female leader in the elections went ahead with voting for a female leader instead of districts with no quotas. Thus, quotas can be a vital instrument to instill long-term cultural change (Beamen et al. 2009). The increased number of women in leadership positions provides the availability of role models in political life that can inspire others to achieve similar outcomes. Some studies also examined the impact in districts in India where girls had higher aspirations and better educational outcomes than in districts with no quota policies (Beamen et al. 2012). However, this literature also indicates that female role models are unattainable in a context where intersecting minority identities such as race, class, and caste dominate. Therefore, it is likely that quotas might only benefit a small number of women. Quotas can be seen to increase the number of women from a particular class or caste and, therefore, less likely to benefit those groups who think that their identities are underrepresented. Many studies show that the effect of electoral quotas for women was to reduce the representation of women of lower caste groups or the representation of low caste groups in general in the local governance. In short, gender quotas tend to favor groups at the top of caste hierarchies (Turnbull 2019).
33
Gender Quotas and Representation in Politics
771
Despite these drawbacks in implementing quotas, it is argued that reservations can have positive benefits in inheritance rights and marriage negotiations. Scholars claim that quotas can create economic and social equality if they provide women with adequate economic resources. Reservations are helpful if elected female leaders could demand effective economic rights enforcement via political participation mobilization. To argue for decision-making at the local level for women is not sufficient unless accompanied by a critique of traditional entitlements like rights and resources for men (Brule 2020). A separate strand of research on women in Latin American politics has analyzed the role of social movements and state response to reform for women’s representation (Htun 2003). Such research claims that the critical variable is not the number of women in the parliament but how their proposals for change threaten or overturn existing gender roles, economic distribution, and religious-doctrinal orders (Htun and Weldon 2010). Suppose the policymaking role of executive bureaucracies is studied as opposed to parliaments. In that case, it is found that state agencies can become spaces wherein public officials can influence the policy networks and intervene in advocacy issues related to women (Weldon 2002). Transnational activist networks that are criticized for setting norms and influencing debates like the recent Me Too movement have focused on rights and voices for women. Thus, apart from bureaucrats and social activists, women, in parliament, could influence legislation and policymaking in a far more fundamental way than quotas for women. There is no causal connection between the number of women and the outcome of legislation. For these reasons, Dahlerup now pays more attention to crucial actors: those legislators, male or female, who initiate policy proposals on their own and embolden others to promote policies for women regardless of the number of female representatives.
Opportunities and Challenges to Gender Quotas Over time, the narrow stand on quotas as part of an equal opportunity policy is seen as inadequate compared to affirmative action policies that target structural discrimination and may result in substantive representation. Quotas mediate the relationship between the number of female legislators and the articulation of women’s interests in the legislative process. The former might result in higher levels of female representation but not a woman-friendly policy agenda or an agenda that promotes women’s interests (Mansbridge 1999). Implementing gender quotas to secure justice limits our attention to liberal institutions and stays within narrow conceptions of the political. A democratic argument for increased representation instead focuses on increasing women in power because they will engage in political activity differently and redefine the structures and procedures of democracy itself. A conceptual problem arises in balancing the concerns of those committed to gender neutrality and those who feel that, in the context of a patriarchal society, the pursuit of liberal equality results in assimilation to a dominant norm of masculinity. To adopt a different perspective means one affirms gender differentiation: It requires probing the existence of separate private and public spheres that define women in
772
V. Verma
specific fixed roles related to reproduction and nurturing (Lister 2003, 71). Sometimes this radical approach draws on the symbol of motherhood to emphasize the distinctiveness of women’s possible contribution. In the past, feminist scholars have looked at contexts and culture and their role in shaping values and norms in any society (Ruddick 1995). The issue here is not to argue for women to abandon symbols of motherhood but rather to propose a conception of democracy that might incorporate maternal thinking. However, this proposal fails to disrupt what Joan Tronto describes as a moral boundary between the family and the “political” that gives rise to the claim that women have different interests in politics because they emerge from the private sphere with a different set of experiences (Tronto 1993). Gender-neutral theorists are concerned that this project reinforces existing stereotypes of women and aims to introduce values and relationships that are not adequately political into the public arena. This gives rise to the need to address gender discrimination and also demand for the inclusion of women in shaping policies, programs, and legislation that affect their lives. Gender quotas have to be understood in the context of a diversity of forms that gender and sexism assume in different cultures and across time. The structure of gender, and the power asymmetry it entails, has been remarkably persistent over time and place. There are both overt forms of gender oppression and the subtler and seemingly innocuous form of power. The differences in entitlement, perceived capabilities, and social expectations of men and women, reflecting society’s norms, laws, and social values, have profound implications on how they participate in the market or non-market work and community as a whole. These differences embody social power relations that constitute the setting for the implementation of development programs, and these differences, therefore, influence program outcomes. Following the work of Hanna Pitkin, there is a well-known distinction between descriptive and substantive representation to prove the fairness and legitimacy of political order (Pitkin 1967). While the former refers to the action of legislators that stand for their constituent groups with whom they share ascriptive similarities in voices and opinions, substantive representation refers to the making of decisions for all subgroups. Underlying this distinction is the claim that women have distinct perspectives and interests which can only be promoted by women legislators. For some countries like the USA and Argentina, studies show that descriptive representation has great success in the way female legislators often introduce bills that favor women’s rights, children, and families (Jones 1997). Research from Latin American countries shows that female legislators are active on questions of domestic violence, sexual harassment, and reproductive rights. Similarly, as women gain access to parliament in Canada and New Zealand, they have become more articulate about child care, parental leave, pay equity, and domestic violence (Sawer 2002). It becomes increasingly clear that although women are different, aggregate trends could be noted in the identities and preferences of women without claiming that women are essentially alike. For example, especially on sexual health reform, there has always been disagreement on using contraceptives to defend individual autonomy or to control population growth (Piscopo 2011). The pattern that emerges is that
33
Gender Quotas and Representation in Politics
773
the content of female legislation has more concern for children, families, violence against women, sexual health, and reproductive rights. Quotas would reinforce the status quo as they generate mandates for women legislators while also reinforcing negative stereotypes about their role as politicians (Franceschet and Piscopo 2008). Nevertheless, once we acknowledge that women legislators share similar concerns, women’s representation becomes dependent on an essentialist vision wherein the contextual factors such as political institutions and policy norms that shape women’s identities and preferences are ignored. Related to this is the empirical question: What does solidarity mean in a political landscape where caste stratifications, religious frictions, and class interests are major social divisions among women? There has been an attack on assumed solidarity among women across social divisions. This has been followed by questioning the universalizing approach of a concept of gender adopted by the theorists proposing quotas to increase women’s presence in political institutions. Leading scholars have attempted to destabilize existing categories and oppositions common in existing feminist scholarship as they reproduce practices that exclude women (Butler 1992). These studies have led to more interest in unraveling the gendered structures of political institutions and checking how structures produce the very gendered subject of politics (Foucault 1980). The assumption that a group perspective is present entirely in any person from a group implies that including members from women is enough to represent the group perspective. However, the correlation between women’s increased presence in legislative bodies and advances in gender equality legislation does not always hold. Gender perspectives can reside in coalitions of organizations, newspapers, journals, and cultural productions by participating in groups that provide a more profound knowledge of the issues concerning women. But these perspectives are different from a policy position or recommendations. It is unlikely that all women could legitimately claim to speak for their group without having participated in interaction with other women and also because they lack the epistemological basis for doing so (Weldon 2010). Lastly, since the economic liberalization policies in many parts of the world, several states have been cutting down on welfare services and public sector jobs. At the same time, the increasing popularity of right-wing forces promoting family values and traditional understandings of gender roles has marginalized some of the earlier gains of gender equality. Since affirmative action in the form of gender quotas assumes a model of state intervention to combat discrimination and underrepresentation, these changes need to be interrogated. However, the main focus of the state is no longer to remedy inequality in people’s circumstances or to combat discrimination through reservation policies. The institutional commitments of the state are very limited in the current context in achieving substantive equality. This allows neoliberal governments today to present themselves as cutting back on interventionist policies. They present the state welfare programs as mostly unaffordable and unsustainable. The inability to follow this transition has led to endless bewilderment resulting in hasty judgments and faulty prognosis of affirmative action programs.
774
V. Verma
It is not surprising that the arguments and political rhetoric of affirmative action continue to baffle many despite the pioneering attempts by many scholars to explain the main features of gender discrimination. But opponents continue to argue that quotas are about the different treatment of individuals, and the move toward reservations is an exception within anti-discrimination statutes. For these reasons, any move toward quotas for women is subject to contestation and is viewed as a form of “special treatment” or “specific advantage” to compensate for past or present discrimination. In addition, institutions outside the state through interest groups, unions, and other associations can influence and advance gender parity.
Conclusion Over the past two decades, the introduction of quotas for women in legislatures and political parties has doubled the average share of female parliamentarians. However, quotas remain controversial even though the ability of electoral quotas to render the groups they represent and society overall better off is acceptable. This chapter presents a rich and nuanced analysis of the complex workings of power in political institutions. However, it limits its objective to examining the different social contexts surrounding the debates on gender quotas and the compelling rationales invoked for underscoring their role in fostering a democratic polity. The rationales behind invoking gender quotas diverge in the justifications given to affirmative action in the local context. For France, quotas focus on the unity of the state and assimilation of immigrants into French culture and society. India’s quota policy was to overcome the caste discrimination and systematic disadvantage faced by groups in educational institutions, political representation, and employment. South Africa argued for national unity and centered on giving affirmative action to the majority community. The US government relied on the notion of liberal democracy with a focus on equal opportunity and diversity. Given these differences, gender quotas are related to the broader context in which they have been introduced, especially the dynamics of representation flowing from the tensions and differences among the different groups in a society. The chapter examines the politics of underrepresented groups like women to recognize the importance of quota design, the impact of quotas on the dynamics inside the political parties, and the role of formal and informal institutions in shaping the application and outcomes of quota policies. Since the main research question was related to the significance of gender in implementing quota politics, the chapter focused on women’s political participation. Based on the theories and findings of existing work, it is argued that quotas for women are broadly effective at achieving political equality and opening new avenues for representation. The term political is expanded to include a wide range of activities in which women have participated and through which they impact institutions, organizations, and practices. Throughout the chapter, women’s political experiences in quota politics are placed in the context of development and democratization.
33
Gender Quotas and Representation in Politics
775
One of the significant obstacles to assessing the impact of quotas is the lack of robust data about the outcome of quotas which we urgently need if we are to modify or expand quota politics further. In addition, while most policymakers would like to increase the representation of a particular group as quickly and effectively, implementing quotas can be a considerable challenge. These problems suggest that specific points should be kept in mind while designing affirmative action policies. First, for questioning stereotypes and prejudices women must make up a substantial number in any given context. The positive effect of role models can only be realized if a diverse group of potential role models is available. This also means that quotas must be available at all positions and not concentrated in lower echelons of political parties. Second, the justification for these quotas should be specified so that men’s overrepresentation and the consequences of relying on a narrow pool of talent become clear to all. Thirdly, quotas need to be linked to ending structural discrimination so that they can be justified in challenging women’s discrimination. Finally, merit-based criteria should be exposed to avoid adverse outcomes like the stigmatization of affirmative action beneficiaries. Many challenges for quota politics arise from the issues of intersectionality. Much pioneering work has been questioned for not paying attention to race, class, caste, sexuality, and disability, in short, what Crenshaw has called political intersectionality (1991). By adopting the stance of intersectionality, the normative and methodological paradigm of quota politics for gender will have to shift to deal with particularism. Intersectionality of women needs different justificatory frameworks to make it more apparent to policymakers and those affected by this strategy that quotas need a democratic and transparent governance framework. In the future, research needs to focus on examining regimes and state institutions rather than specific policies to examine the bureaucratic structures and the broader process of knowledge production. The interest in state structure has generated an interest in examining historical legacies in the way policies and activities combine to create more significant effects on gender relations. The literature on gender regimes generates different models following the demise of the breadwinner regimes: the universal breadwinner, the caregiver parity, and the universal caregiver model. Thus, a study of welfare states and the impact of institutional structures and employment strategies in challenging or reproducing gender inequalities must be undertaken. The past several years have seen significant developments substantially altering the nature of affirmative action programs. In light of critiques of affirmative action policies, they may be embarking on a new path. The evolution of affirmative action from an anti-discriminatory policy into a policy of quotas for the poor has sparked a debate regarding the road map of this policy. While affirmative action policies have historically been contested within a paradigm of social justice, analysis of the recent cases suggests a new strategic emphasis on disputing the legitimacy of the state’s support of quota policies.
776
V. Verma
References Adam K (1997) The politics of redress: south African style affirmative action. J Mod Afr Stud 35(2):231–249 Alexander A, Jalazai F (2020) Symbolic empowerment and female heads of states and government: a global, multilevel analysis. Polit Groups Ident 8(1):1–20 Anderson E (2010) The imperative of integration. Princeton University Press, Princeton Bacchi CL (1999) Women, policy and politics. The construction of policy problems. Sage, London Bacchi CL (2006) Arguing for and against quotas: theoretical issues. In: Dahlerup D (ed) Women, quota and politics. Routledge, Abingdon, pp 32–51 Baldez L (2002) Why Women Protest: Women’s Movements in Chile. Cambridge, UK: CUP. Bowen WG, Bok D (1998) The shape of the river: long-term consequences of considering race in college and university admissions. Princeton Press, Princeton Banaszak LA (2010) The women’s movement inside and outside the state. Cambridge University Press, New York Beamen L, Duflo E, Pande R, Topalova P (2012) Female leadership raises aspirations and educational attainment for girls: a policy experiment in India. Science 335(6068):582–886 Beamen L, Chattopadhyay R, Duflo E, Pande R, Topalova P (2009) Power women: quotas and affirmative action. Does exposure reduce bias? Q J Econ 124:1497–1540 Bonomi G, Brosio G, Di Tommaso M (2013) The impact of gender quotas on votes for women candidates. Evidence from Italy. Fem Econ 19:48–75 Bush S (2011) International politics and the spread of quotas for women in legislatures. Int Organ 65(1):103–137 Butler J (1992) Feminists theorise the political. Routledge, London Brule R (2020) Women, power, and property: the paradox of gender equality laws in India. Cambridge University Press, Cambridge Chaney EM (1979) Supermadre: women in politics in Latin America. Institute of Latin American Studies, University of Texas, Austin Craske N (1999) Women and politics in Latin America. Polity, Cambridge Press Crenshaw K (1991) Mapping the margins: intersectionality, identity politics, and violence against women of color. Stanford Law Rev 43(6):1241–1299 Darhour H, Dahlerup D (2013) Sustainable representation of women through gender quotas. A decade’s experience in Morocco. Women’s Stud Int Forum. https://doi.org/10.1016/j.wsif.2013. 04.008 Dahlerup DZ Hilal NK, Kandawasvika-Nhundu R (2013) Atlas of electoral gendered quotas. International Institute for Democracy and Electoral Assistance, IDEA, Inter-Parliamentary Union, and Stockholm University. Downloaded 18 Feb 2021 Dahlerup D (ed) (2007) Women, quota and politics. Routledge, London Devlin C, Elgie R (2008) The effect of increased women’s participation in parliament, the case of Rwanda. Parliam Aff 61(2):237–254 Deshpande A (2013) Affirmative action in India. Oxford University Press, Oxford Fallon K, Swiss ML, Viterna J (2012) Resolving the democracy paradox: democratisation and women’s legislative representation in developing nations 1975–2009. Am Sociol Rev 77(3): 380–408 Foucault, Michel (1980) Power/knowledge: selected interviews and other writings 1972–77 (C. Gordon. ed). Harvester, London Franceschet S, Piscopo JM (2008) Gender quotas and women’s substantive representation: lessons from Argentina. Polit Gend 4(3):393–342 Fraser, Nancy (1998) Social justice in the age of identity politics: Redistribution, recognition, participation, WZB Discussion Paper, No. FS I 98–108, Wissenschaftszentrum Berlin für Sozialforschung (WZB), Berlin https://www.econstor.eu/bitstream/10419/44061/1/269802959. pdf
33
Gender Quotas and Representation in Politics
777
Geisel B, Hust E (2005) Democratic mobilisation through quotas: experiences in India and Germany. Commonw Comp Polit 43(2):222–244 Geisler G (2000) “Parliament is another terrain of struggle”: Women, men and politics in South Africa. Journal of Modern African Studies, 38(4):616–619 Gilardi F (2015) The temporary importance of role models for women’s political representation. Am J Polit Sci 59(4):957–970 Gray T (2003) Electoral gender quotas: lessons from Argentina and Chile. Bull Lat Am Res 22: 52–78 Htun M (2002) Puzzles of women’s rights in Brazil. Soc Res 69(3):733–751. https://www.muse.jhu. edu/article/558530 Htun M (2003) Women and democracy. In: Dominguez JI, Shifter M (eds) Constructing democratic governance in Latin America. John Hopkins University Press, Baltimore Htun M, Power TJ (2006) Gender, parties and support for equal rights in the Brazilian congress. Latin Am Polit Soc 48(4):83–104 Htun M, Weldon SL (2010) When do governments promote women’s rights? A framework for the comparative analysis of sex equality policy. Perspect Polit 8(1):207–216 Jones M (1997) Legislator gender and legislator policy priorities in the Argentine Chamber of Deputies and the United States House of Representatives. Policy Studies Journal 25(4):613–629 Kittelson MC, Schwindt-Bayer LA (eds) (2012) The gendered effects of electoral institutions: political engagement and participation. Oxford University Press, New York Krook ML (2009) Quotas for women in politics: gender and candidate selection reform worldwide. Oxford University Press, New York Krook ML, Childs S (eds) (2015) Women, gender and politics: a reader. Oxford University Press, Oxford Lazar M (2005) Politicizing gender in discourse: feminist critical discourse analysis as political perspective and praxis. In: Lazar M (ed) Feminist critical discourse analysis. Macmillan Palgrave, London. https://doi.org/10.1057/9780230599901_1 Lister R (2003) Citizenship: feminist perspectives. Macmillan, Palgrave Liu S-JS, Banaszak LA (2017) Do government positions held by women matter? Polit Gend 13(1): 132–162 Mansbridge J (1999) Should blacks represent blacks and women represent women? A contingent yes. J Polit 61:628–657 Miguel LG (2008) Political representation and gender in Brazil: quotas for women and their impact. Bull Lat Am Res 27:197–2014 Meier P (2008) A gender gap not closed by quotas. Int Fem J Polit 10:329–347 Molyneux M (2001) Women’s movements in international perspective: Latin America and beyond. Palgrave, New York Mustafa S (2020) Shaheen Bagh and the Idea of India. Writings on a movement for justice, liberty, and equality. Speaking Tiger Books LLP,125-A, Shahpur Jat, New Delhi Nagel T (1973) Equal treatment and compensatory discrimination. Philos Public Aff 2(4):348–363 Nir L, McClurg SD (2015) How institutions affect gender gaps in public opinion expression. Public Opin Q 79(2):544–567 Novi Rusnarty Usu. (2010). Affirmative action in Indonesia: the Gender Quota system in the 2004 and 2009 elections, ASIA ONLINE. Downloaded https://www.catalogue.nla.gov.au/Record/ 4979285 on 19 February 2021 Parpart J, Connelly P, Barriteau E (2000) Theoretical perspectives on gender and development. The International Development Research Centre, Ottawa Paxton P, Hughes MM, Barnes TD (2020) Women, politics and power: a global perspective. Rowman and Littlefield, Maryland, USA and London, UK Pitkin H (1967) The concept of representation. The University of California Press, Berkeley Piscopo JM (2011) Do women represent women? Gender and policy in Argentina and Mexico, Phd Thesis, University of California, San Diego, CA Ruddick S (1995) Maternal thinking. Towards a politics of peace. Beacon Press, Boston
778
V. Verma
Sawer M, Tremblay M, Trimble L (2006) Representing women in parliament. A comparative study. Routledge, London Sawer M (2002) The representation of women in Australia: meaning and make-believe. In: Ross K (ed) Women, politics and change. Oxford University Press, New York, 5–18 Schwindt-Bayer LA (2006) Still supermadres? Gender and policy priorities of Latin American legislators. Am J Polit Sci 50(3):570–585 Sharma K (1998) Power vs. representation: feminist dilemmas, ambivalent state and the debate on reservation for women in India. Occasional paper no 28. CWDS, Delhi, pp 1–22 Susan F, Jennifer M. Piscopo (2008) Gender Quotas and Women’s Substantive Representation: Lessons from Argentina. Politics and Gender, 4(03): 393–405. Tamerius K (1995) Sex, gender and leadership in the representation of women. In: GeorginaLahti D, Kelly RM (eds) Gender, power and leadership, and governance. Michigan Press, Ann Arbor Thomson JJ (1973) Preferential hiring. Philos Public Aff 2:364–384 Tremblay M (ed) (2008) Women and legislative representation: electoral systems, political parties, and sex quotas. Routledge, London Tremblay M (2007) Democracy, representation, and women: a comparative analysis. Democratization 14(4):533–553. https://doi.org/10.1080/13510340701398261 Tripp AM, Kang A (2008) The global impact of quotas on the fast track to increased female legislative representation. Comp Polit Stud 41(3):338–361. https://doi.org/10.1177/ 0010414006297342 Tronto J (1993) Moral boundaries. Routledge, Great Britain Turnbull B (2019) Quotas as opportunities and obstacles: revisiting gender quotas in India. Polit Gend 17:1–25 Viterna J, Fallon K (2008) Democratization, women’s movements, and the gender equitable states: a framework for comparison. Am Sociol Rev 73(4):668–689 Verma V (2012) Non-discrimination and equality. Contesting boundaries of social justice in India. Routledge, London. (2017 Paperback edition for South Asia Routledge) Verma, Vidhu (2002) Malaysia: State and Civil Society in Transition. Lynne Rienner Publishers, Colorado, USA Weldon SL (2002) Beyond bodies: institutional sources of representation for women in democratic policymaking. J Polit 64(4):1153–1174 Young I (1990) Justice and the politics of difference. Princeton University Press, Princeton Zettererg P (2009) Do gender quotas foster women’s political engagement? Lessons from Latin America. Polit Res Q 62(4):715–730
Reports The Mahbub al Haq Human Development Centre (MUH) (2004) Human development in South Asia 2003. The employment challenge. Oxford University Press, Oxford CDEAW (1979) United Nations Convention for the Implementation of All Forms of Discrimination Against Women MDGR (2013) The Millennium Development Goals report (MDGR), UNDP, New York Inter-Parliamentary Union (2020) Women in National Parliaments, 1 October 2020 https://www.unwomen.org/en/what-we-do/leadership-and-political-participation/facts-and-figures. Downloaded 18 February 2021
Caste Quotas in India
34
Amit Thorat
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Affirmative Action in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quotas in Public Sector Occupation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quotas in Public Sector Educational Institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quotas in the Legislature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quota for Economically Weaker Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Demands for Reservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contemporary Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Criticism of Quotas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
780 785 786 787 788 789 790 791 792 794 794
Abstract
The national policy of reservation for the historically discriminated and marginalized groups of India, namely, the scheduled castes, scheduled tribes, and the other backward classes, in the form of quotas, or a fixed share of earmarked seats, in state-run educational institutions, government employment, and the legislature, is the state’s chief instrument for providing representation to group members, in the nation’s economic life. This chapter examines the motivation behind the use of quotas as a state policy tool, its efficacy, and its limitations in advancing the cause of historically marginalized groups. Keywords
India · Caste · Reservations · Quotas
A. Thorat (*) Jawaharlal Nehru University, New Delhi, India © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_38
779
780
A. Thorat
Introduction The idea of group identities, based on divine origins, described and ascribed by religious doctrine, in the form of caste is unique to South-Asian, where many regions historically have been or continue to follow Hindu religion. Hindu religious texts are considered to be ancient and of divine origins or written by sages, to whom, it is believed, divine knowledge came(to), and they acting as vessels, wrote down among other, many different origin stories of humans and society. These origin stories are different from each other as well as similar in many ways. However, one among them, the “Purusha Shukta,” a heyem, in the Rig Veda1 (BC 1500–BC 1000), one of four Vedas or compilations, considered to be the oldest and some of the holiest Hindu texts, has come to define the divine origins story of society and dominate the belief and nearly every aspect of life, of Hindus all over the world. The Purusha Shukta reads: “For the prosperity of the world He (the creator) from his mouth, arms, thighs and feet created the Brahmin, Kshatriya, Vaishya and Shudra.” (Ambedkar 1987, pg. 112)
This (fourfold) social order called Chaturvarna has been reiterated and justified by later religious law givers. For instance, Apastamba Dharmasutra2 states: “There are four castes-Brahmana, Kshatriya, Vaishyas and Shudra. Among these, each preceding (caste) is superior by birth to the following”
This is repeated by Vasishtha Dharmasutra which says: “There are four castes (Varnas), Brahmins, Kshatriyas, Vaishyas and Shudra,” it adds “The four castes are distinguish by their origin and by particular sacraments, The Brahmin was his (creator) mouth, Kshatriya formed his arms, the Vaishya his thighs, the Shudra was born from his feet.”
The Hindu religious text offers its version of the origins story of society, dividing it into hierarchical, endogamous, hereditary groups, with members ascribed unequal rights, privileges, and social status. Let us understand what is the meaning of each of these characteristics. Hierarchical means that each varna is not equal in status or standing. The Brahmin(s) are given the highest status. This high status is not just notional, it translates into group members enjoying the highest level of social, economic, political, civil, and human rights. These rights diminish as one moves down the four-group (fold) hierarchy, with the least rights being accorded to the fourth varna or groups. In addition to these four groups, a fifth group came into existence
1
The Vedas are considered to be the most sacred of ancient Hindu religious texts. There are said to be four Vedas. Rig veda, Sama veda, Yajur veda, and Atharva veda. 2 Britannica, The Editors of Encyclopaedia. “Dharma-sutra”. Encyclopedia Britannica, 7 Dec. 2011, https://www.britannica.com/topic/Dharma-sutra. Accessed 10 October 2022.
34
Caste Quotas in India
781
over time, however outside of this fourfold classification, whose members are denied any of the rights or privileges enjoyed by the previous four groups. These are the Avranas (outside of the fourfold categorization) or the ex-untouchables or ex-enslaved. While many claim that the body parts from which these four groups are ostensibly believed to have emerged, represent in some way the nature of the vocation, that these groups were engaged with, or chose to follow and nothingmore. The choice is based on one’s own nature or inclination. This interpretation is offered to indicate that people drew toward occupation types of their own inclinations and personality type.While this might be true, we find this division of laborers eventually being enforced by religious law. The top to bottom (head to toe) nature of the emergence has come to signify a hierarchical status in practice over millennias and some believe that this and the hereditary nature were part of the original interpretation and practice of the doctrine. Each Varna is ascribed specific economic rights, which would include the right to own (or not) land, wealth, assets, what occupations are allowed to be undertaken, among other economic rights. With respect to occupations, for example, the Brahmins are accorded the highest status, and by religious law, supposed to be teachers and educators. However, their status being the highest, they are also free to take up any other vocation or occupation that have been prescribed to the remaining three groups or Varnas. Similarly, the next group in social hierarchy, namely, the Kshatriyas, too have similar privileges of not only following their religiously ordained occupation, of being soldiers and rulers, but also allowed to undertake the occupation or vocation of the fourth varna, namely, the Vaishyas, who are ascribed trade, business, and agriculture as their occupation. So any group that is higher in status is allowed to take up the occupations or work of those below them. The reverse, however, is not allowed, that is, for those from a lower status group to peruse the occupations, ascribed for those in a group above them. Therefore, the Shudras, the fourth varna members, are not theoretically allowed to take occupations of any other groups, but only follow their own ascribed occupations. In all the ancient texts, the Shudras are also considered to be the “untouchable.” They were considered to be physically, spiritually, and ritually defiling. Their shadow, touch, bodily fluids like sweat, spit, blood, urine, etc. were (and are still) considered as polluting for the other three upper varna or caste groups. Thus, the first three groups follow(ed) untouchability and segregated living with the Shudras. Since the Vaishyas were tasked with conducting business, trade, and agriculture, they over time passed on the burden of hard agricultural labor on to the untouchable Shudras to work on their fields. Since the Shudras had to work with producing food grain, vegetable, fruits, and the processes of sowing, harvesting, cleaning, sorting, and storage, which was not possible without physical touch, it was impossible to practice untouchability with them and also employ them as agrestic labor. It is understood that over time the Shudras who were agricultural laborers lost their “untouchable” status and many subsequently also came to possess and own land. Those Shudras who were engaged in certain peculiar occupations, that were considered as polluting and defiling, such as working with dead animals or their hide for leather, cleaning, and sweeping, then came to bear the burden of untouchability.
782
A. Thorat
Over time many other occupations were included into this fifth group, such as basket weaving, traditional healing, cloths washing, ironmongery, other skilled and semiskilled trades and crafts, etc. This fifth group that emerged out of the larger Shudra group were the Avarnas, the “untouchables” (Yamazaki 1997). These religiously sanctioned norms and related caste (varna/group) codes eventually came to be converted in law. The credit for this is given to Manu,3 the first man, according to Hindu mythology, who came into existence. The book of laws came to be known as the Manava-dharma-shastra or popularly known as the Manu smriti4 around BC 200 (Ambedkar 1947, pg. 9). Manu is therefore also referred as the law-giver. These laws are seen as sacrosanct and have come to define and justify all forms of social, economic, cultural, and even civil and human rights norms of living, for the Hindus. While these norms might have existed as part of Hindu theology and religious theory, written down in various Dharma-shastras5 as mentioned above, they did not affect day-to-day practical life and were neither sacrosanct nor binding in social life of people. People could take up any occupation, marry anyone, buy and sell land and property, accumulate wealth, and learn trades and craft. It is also believed that they could have been an indigenous aboriginal group or tribes that tried to assimilate into the Vedic Hindu culture and society and ended up being relegated to its margins, discriminated, excluded, exploited, and used as servile labor and as enslaved. Historically, even if they were assimilated into the Hindu socioreligious cultural life, eventually they ended up being denied human, social, civil, political, or economic rights. They could not own land, asset, or property, conduct business or trade, grow food, or learn any new vocation or skill. Their labor being servile and socially necessary that served the four groups placed above them but ritually defiling, eventually got religiously ordained and fixed. Both Max Muller6 and B.R. Ambedkar,7 among others, believe that this verse (Purusha shukta) was added much later into the Rig Veda, in order to justify a re-classification of the Hindu society into the four broad endogamous, hierarchical groups with unequal economic, social, political, civil, and human rights. These rights were bestowed in full on the top most group (Brahmins) and were denied progressively as one moved down the social order, with the out-castes being accorded no rights whatsoever, except to serve the castes above them in social ranking. These broad four groups have over time come to be further divided into innumerable
3
Britannica, T. Editors of Encyclopaedia (2010, June 7). Manu. Encyclopedia Britannica. https:// www.britannica.com/topic/Manu 4 Britannica, T. Editors of Encyclopaedia (2015, February 4). Manu-smriti. Encyclopedia Britannica. https://www.britannica.com/topic/Manu-smriti 5 Britannica, T. Editors of Encyclopaedia (2018, October 24). Dharma-shastra. Encyclopedia Britannica. https://www.britannica.com/topic/Dharma-shastra 6 https://www.britannica.com/biography/Max-Muller 7 https://en.wikipedia.org/wiki/B._R._Ambedkar
34
Caste Quotas in India
783
sub-groups, known as Jaties. These group and sub-group identities are hereditary and marriages are endogamous at sub-group level. South-Asia at present is home to a population of 201.6 million Hindus “considered as (ex)untouchables” socially, despite political and legal bans on this inhuman practice. Of these, 97% live in India, the remaining distributed across Nepal, Bangladesh, Pakistan, and Sri Lanka and also across the rest of the world. In India, Hindus as well as those who historically converted from Hinduism to Islam, Christianity, Sikhism, and Jainism still practice untouchability (Thorat and Joshi 2020). While the socially overt forms of the practice are not as widely and openly observable in urban areas, in rural regions of the country these practices are still the norms which dictate interpersonal and social interactions (Coffey et al. 2018). Untouchability is the social practice of avoiding eating, dining, sitting, sharing food, etc. (Shah et al. 2006, documents a wide range of such practices) with particular groups (ex-untouchables or Dalits – a self-professed empowering term, meaning oppressed) that are forced to live in segregated villages hamlets and even urban localities and slums (Gayatri et al. 2019), not allowed to marry out of their caste group, in addition to a slew of other market and non-market discriminatory practices (Shah et al. 2006; Thorat and Attewell 2007, Thorat and Joshi 2020). Group members even today do not have access to many common public resources, such as ponds, public water hand-pumps, temples, etc., or have separate ones designated for their usage in Indian villages and at times in urban centers. Even areas within cremation groups are segregated on broad cast (class) lines. Hindus can identify each other as belonging to these four broad groups, as well as at times the sub-groups, from the family name or the second name (surname) of a person. These names are hereditary and signify a person’s sub-group affiliations, unless a person choose to change his/her second name with a caste-neutral name officially. It is therefore not impossible, if difficult to ascertain the caste of individuals. Additionally, historic discrimination and exclusion of the untouchables and the caste-enslavement, (slavery was banned in 1843) that continued unabated till India’s independence, manifests itself as “down the stream” consequences, clearly visible as differences in well-being indicators, such as wealth, income and education etc. It is therefore not difficult to identify members of these groups from their economic, social and cultural life differences. The caste categorization of Hindus, therefore, is one of the four pillars of Hindu theology and is central to its belief system. It is believed that one is born into a particular caste and sub-caste as a result of ones deeds in the past life. Once born in a caste group, one must follow the caste-ordained occupation as well as the associated rules without question, if one hope to be born into a higher caste in the next life. This is broadly the “Karma theory of” Hinduism. Caste identities are therefore hereditary. In 2011, the India Human Development Survey found that nearly 95% of all marriages that happened in India were arranged within sub-caste categories (Hindu, newspaper 2016).8
8
https://www.thehindu.com/features/magazine
784
A. Thorat
In addition to untouchability, recent studies (Mohan; 2015, Vishwanath; 2015) show that many sub-castes within the broad caste group of the untouchables, such as the Pulayas and Pariahs in the south of India, were designated as enslaved castes and were largely used as agrestic enslaved labor (Saradamoni 1980). They were owned as property. Family members of enslaved caste could be separated, husbands from wives, sons and daughters from parents, they could be bought, sold, gifted, given as dowery for a daughter’s marriage or as interest and debt payments, etc. The owner had all these rights, as well as the right to punish and end their life. The law of Manu says “a Shudra, whether bought or un-bought, should be reduced to slavery because he is created by God for the service of a Brahman” (Bhuler1886, Law of Manu, p. 24). Manu is considered to be a sage from the ancient period, who compiled the rules of law and punishment from the various “Dharmashastras”9 into a compendium, called Manu’s Law (Manu Smriti10). Even today many Hindu castes consider the Law of Manu to be the law of the Hindus – rules and behaviors that are prescribed and considered as desirable in social and interpersonal conduct, if not decreed as law. These stand in direct contradiction to constitutional law of India. An example of this underlying belief is the presence of the statue of Manu in the premises of the High Court of Rajasthan (a state in North-Western India). This condition of the untouchables and the enslaved among them persisted more or less in varying degrees and manner right from the ancient period to the medieval, right through the Mughal occupation of the sub-continent as well as over the first 73 years of British rule in India (that followed the Mughal rule), till the British legally abolished caste slavery in 1843 as an extension of the Anglo-Indian law, constituted by them in British colonial India. While slavery was legally abolished as a socio-economic norm, in practice, identical landlord–labor relations continued in the form of bonded, attached, and other forms of labor that had all the features of slavery except legal ownership of the agrestic workers by their landlord masters. A reason for the continuation of caste-enslavement in different forms, was the failure of the British to announce or provide reparations, in the form of land and cattle, or any other means of earning a livelihood, that would help them start afresh and be free of the economic dependence of their land-owning enslavers. Historians such as Sanal Mohan contend that enslavement of particular members of the sub-caste of untouchables continued in practice till the third quarter of the twentieth century in India (Mohan 2015). At the same time while slavery was banned by the British, this did not mean untouchability was, which continued till the modern period, when India became an independent (1947) sovereign state with its own constitution, which gave back civil, social, political, and economic rights to untouchables. Given this long history (2200–2500 years) of exclusion and discrimination that lead to enslavement and untouchability that denied them all forms of rights and fair
9
https://www.britannica.com/topic/Dharma-shastra https://www.britannica.com/topic/Manu-smriti
10
34
Caste Quotas in India
785
remunerations for work and the intergenerational enrichment, accumulation of wealth and growth of the non out-caste groups, by systematic appropriation of the product of the labor of the enslaved, a strong case can be made for Reparations and Restitution for the ex-untouchables of India.11 However, there has been a lack of serious academic analysis, discussion, and debate around this particular form of redistributive justice. On the other hand, affirmative action (AA) policies have received more than its due share of attention. AA policies have been the main policy tool to try and bring about social justice in India historically.
Affirmative Action in India The India, that its people inherited, after nearly 200 years of British colonial rule was poor, drained of its many resources, and in need of a new development strategy and program of growth and development. The model that was adopted by its founding fathers was of a closed socialist economy, with the state assuming the role of the driver of economic, political, social, and human development. It was envisaged as a middle of the path strategy between Marxist Communism and Western Capitalist model of economic development. The constitution that the country adopted reflected its new ideals, as well. India was constituted as a democratic, socialist, secular, sovereign republic, to secure all its citizens Justice, Liberty, and Equality and to promote Fraternity. In order to do right by the historically oppressed groups, such as the untouchables (and the large indigenous tribal population), Article-17 of the constitution abolished “untouchability,” forbidding its social practice in any form. Following this in 1955, “The protection of Civil Rights Act” was passed, that prescribed punishment for the “preaching and practice of untouchability.” Despite these laws being enacted, it was observed that these were not enough to address the continuing crimes perpetrated against the ex-untouchables as well as the tribal population of India. This lead to the enactment of “The Scheduled Castes (SC) and Scheduled Tribes (ST) (Prevention of Atrocities) Act” in 1989, to prohibit discrimination, prevent atrocities, and hate crimes against the (by then them) ex-untouchables and tribal groups. In order to ascertain the individual identities of all the different ex-untouchable groups and sub-caste groups (or jaties), as well as all the numerous tribes of India, lists or schedules of such sub-caste and tribe names were compiled officially by the central government separately, for each of the states of India, for ease of identification of group members. As a result, all the ex-untouchables groups and the tribes are also known in legal parlance as the scheduled castes (SC) (ex-untouchables) and the scheduled tribes (ST) (indigenous/aboriginal tribal inhabitants). These lists are then used to identify group members for reservation and quota provisions. In addition to laws enacted to protect the scheduled castes (SC) and the scheduled tribes (ST), AA policies were included at the time of conception of the constitution. 11
Thorat & Thorat, Lancet Commission Report on Reparations
786
A. Thorat
Article-16 of the constitution (1949) allows for the state to make special provision for backward sections of the society, to give them adequate representation. This representation came to be provided in government or federal job appointments and promotions, government-run and -funded educational institutions, as well as in the form of reserved seats in the Central, State, and local level legislature. The reserved seats are allocated in proportion to their population shares, 16% for the SC and around 8% for the ST. The total population of SC and ST therefore combines to about 25% of India’s population, that is, 305 million people. The idea behind these measures was to provide representation to these historically discriminated and excluded groups and bring them into the main stream of economic and political life, as well as give them an opportunity to voice their concerns and issues.
Quotas in Public Sector Occupation The provision of Article-16 (4) of the Indian constitution states “Nothing in this article shall prevent the State from making any provision for the reservation of appointments or posts in favour of any backward class of citizens which, in the opinion of the State, is not adequately represented in the services under the State.” This allows the state to reserve seats in government jobs for the SC and ST in proportion to their share in the total population. Reservations are also provided in promotion of government employees. Government employment includes the civil services, public sector undertakings, statutory and semi-government bodies, voluntary agencies that are under the control of the state, or receive grants as aid. However, certain sectors of the state are excluded, such as the military, judiciary, and specialized jobs such as medical surgery, scientific research, etc. While these provisions for the scheduled castes and the scheduled tribes have been in existence from the very beginning, as part of constitutional provisions, subsequently a new category of “other backward castes” (OBC) was added to this list. In 1949, under the provision of Article-16(4) of the constitution. This group was extended similar provisions of reserved seats, in government jobs and educational institutions, but none for central and state legislature, except at lower levels such as urban ward and block levels and rural block and village levels. However, these provisions finally came into being in 1987. The OBC constitute a slew of castes (part of the shudra group) that lie just above the ex-untouchables in the Hindu hierarchical classification and while not as deprived as them in terms of loss of rights, practice of untouchability, etc., “suffered from economic and educational backwardness” due to their lower status as compared to the top three groups. However, a section of this group historically have managed to become economically and educationally empowered and upwardly mobile, and accordingly termed as the “creamy layer” and are exempted from receiving any form of reservations. This layer of people are those who belong to households that have an annual income above 8 lack rupees or around 10,000 US dollars. Given the long historic nature of marginalization of these groups, it is also recognized that group members, despite hard work and commitment, might not be able to compete at par with the groups that have been their oppressors, due to
34
Caste Quotas in India
787
centuries of intergenerational transfer of knowledge and privilege, and are way ahead in the race. To compensate for this gap, certain concessions are accorded to the SC, ST, and OBC groups. These are (i) minimum age criterion for entry into civil services does not apply to them, (ii) the minimum qualification criteria is marginally relaxed for them, (iii) job interviews from them are held separate from the rest, with representatives from their group, present in the interview board to ensure a fair evaluation.
Quotas in Public Sector Educational Institutions As mentioned earlier, the Indian constitution makes special provisions to provide representation to members of the SC, ST, and OBC groups in all educational institutes that are Central and State government run or funded and aided by them. This was instituted with the idea of providing assured access and therefore representation to group members that had historically been denied education under religious laws and social norms. Restrictions to education for the ex-untouchables go back in time to the ancient period (BC 600 onward) of India’s history and there on continued through the Medieval age (AD 700 to AD 1756) and further under the British rule from 1754 to 1855 (Sharma 1958; Jha 2018; Moosvi 2011; Ambedkar 1982). It is only in 1855 that the British opened education to the depressed castes (Ambedkar 1982). For the ancient period, Manu’s laws framed around BC 200 stated, “One may not teach him (Shudra) the law or enjoin upon him religious observances. For he who tells him (shudra) the law and he who enjoins upon him (religious) observances, he indeed together with that (Shudra) sinks into the darkness of heaven” (Manu Smriti, Buhler,1886). The Islamic regimes that ruled subsequently in the Middle Ages did not interfere much in the Hindu laws and customs including those related to education (Moosvi 2011). The British carried on with the existing Hindu laws and customs further, quite judiciously. The first Anglo-Indian law under the East India Company, enacted in 1772, stated, “In suits regarding succession, inheritance, marriage, and ‘caste’, and all religious usages and institutions, the Hindu laws are to be considered the general rules by which the judges are to form their decision.” It is only in 1913 that the British made a public declaration regarding education of the native people being a responsibility of the state. However, education was opened as early as 1855, but due to stiff opposition from the high castes, the depressed classes could not gain access to public schools, although the laws permitted the same. Responding to a complaint made by an untouchable boy, for refusal of admission to him in a public school, on May 29th, 1928, Governor General of Bombay Presidency had stated that “It would not be right for the sake of single individual, the only Mahar (an untouchable caste) who ever came forward to beg for admission into a school attended by the pupils of high castes and to force him into association with them, at the probable risk of making the institution practically useless for the great mass of natives” (Ambedkar 1982: pp. 419).
788
A. Thorat
Indeed, the opposition by high castes was so great that the admission of low-caste pupils had led to the closure of five or six large schools for 2 years (Ambedkar 1982). In the face of intense opposition from the high caste to the entry of depressed castes in public schools, the British government tried an alternative arrangement of creating separate schools and offering grants to Christian missionary schools for admitting children from depressed castes. These measures, however, did not result in much improvement, as the enrolment of depressed classes remained very low (Ambedkar 1982). It is after the adoption of Constitution in 1950 and positive measures mentioned above that some progress became visible, although with limited outcomes. The gap in the educational attainment between the scheduled castes and high caste persists. In 2017/16 the enrolment rate of the scheduled castes was 21% compared with 26% of National average and 41% for high castes. This gap is both due to poorer economic condition of the SC and their continued discrimination in higher education institutions. Decomposition exercise related to gaps in enrolment rate in higher education institutions between SC and high caste shows that about 64% of the gap is explained by the endowment effect and 36% due to caste discrimination (Thorat and Khalid 2022). Increasing number of studies reveal discrimination of SC in higher education institutions (Thorat 2008, Sukumar 2022). In Ambedkar’s words, the “past is again be made to live in the present.” Given the widespread and serious nature of discrimination, the University Grant Commission was forced to enact a regulation against discrimination in 2012. The issue of limited access to higher education of socially disadvantaged children remains also due to inadequate financial support in public education institutions and growing privatization of higher education. In 2019/ 20 close to 41% of university, 65% of colleges, and 67% of standalone institutions were in the private sector, where there is no reservation for the low castes. Unless these issues are addressed, the caste disparities in education will not only persist but mostly likely widen. Also education was perceived as the tool with which group members over generations change their lot and catch up intellectually and hence economically with the rest of the population. Certain concessions are also accorded in the minimum qualifying requirements for admission to educational institutes; education fees and other financial concessions, scholarships, financial aid, special earmarked accommodations, etc. are other policies in place for additional support.
Quotas in the Legislature Similarly even in the Legislatives assembly, seats are reserved for three groups in proportion to their population shares. According to the last national census of 2011, the were population shares for the SC and ST are 16.2% and 8.2%,12 respectively. While political representation exists, it however is seen as being not as effective as 12
https://censusindia.gov.in/Tables_Published/A-Series/A-Series_links/t_00_005.aspx
34
Caste Quotas in India
789
intended and does not translate into the voicing of concerns of these groups, by their elected representatives, to the extent hoped for. One reason often cited is that constituencies have majority non-SC/ST populations. These constitutional provisions along with the laws have come a long way in providing some level of social security and national representation to group members, as well as helped them progress educationally and economically, from their very dismal condition at the dawn of independence.
Quota for Economically Weaker Sections While the concept of reserving seats as quotas for marginalized social groups found favor, with most, if not all, since the very beginning of the establishment of a democratic India state, in the form of constitutional provisions, in recent years additional seat quotas have been allocated to the economically but not socially weaker segments of the society. In 2019 the Indian constitution amended Article 15 and 16, of the constitution, allowing the state to institute reservation of a maximum of up to 10% of seats in educational institutions, both state-aided and -unaided private institutions, except minority institutions. This 10% reservation is over and above the seat reservation provided for the SC, ST, and OBC and applies to those not belonging to these three groups, also called the general population. This 10% reservation for the economically weaker sections also applies to federal or government appointments. The following criteria are applied to ascertain if a person is eligible for this provision: (i) applicant’s annual family income is less than 8 lakh rupees (10779$) per annum, (ii) her/his family must not own more than 5 acres of agricultural land or (iii) a residential flat of an area of 1000 sq ft. or more or (iv) a residential plot in a notified municipal sector of 100 square yards or more or (v) a residential plot in a non-notified municipal sector of 200 square yards or more. Quotas for the economically weaker sections have been very controversial as they are seen as standing in direct contravention to the idea and motivation behind the social identity-based quotas. The argument against these new quotas are the following. When India attained independence, it was an impoverished nation with a large section of its population, nearly 60%, living below the poverty line. The nation then took up on itself to become self-reliant in food production, industrial, infrastructural, and overall growth. Poverty alleviation had been and continues to be a large part of the state’s financial allocation and concern. It is argued that all the policies and schemes instituted since independence, serve and work toward helping those below the poverty line to earn a basic minimum income, give accesses to sufficient amounts of food, education, housing, etc. among other schemes. These apply to all those who need them, irrespective of their social identities (caste, religious, ethnic, regional, gender, etc.), and help them come out of poverty. Reservation policy was seen as the only state mechanism that aspired to counter the effects of the continuing social practice of identity-based discrimination and exclusion over and above its goal to provide representation to these historically enslaved, segregated, marginalized, and exploited groups. A correspondence study
790
A. Thorat
(Thorat and Attewell 2007) for India shows that equally qualified job applicants from the scheduled caste and the upper caste groups did not face equal likelihood of being invited for an interview, with the former having significantly lower chances of being called for an interview than and upper caste candidate, based solely on the candidates name (revealing social identity), when qualifications were kept equal. Even when SC applicants had higher qualifications than their upper caste counterparts, the latter did not have significantly lower chances of a job interview call than the former. This and other similar studies show that prejudiced employers prefer candidates from their own or similar social background, even when equally or more qualified candidates from minority groups are available and willing to work and also belong to similar economic class. Decision to discriminate it therefore neutral to class and based on the social identity of an individual, which could in addition get compounded by race, ethnicity, sexuality etc. of the person. The private sector is still open to market and non-market discrimination with no provisions whatsoever. So the critics of the EWS reservation point that there are a large number of schemes and state measure to address economic disadvantage which applies to the entire populace and reservation policy, the only tool addressing identity-based discrimination, should not be extended and its purpose and meaning diluted by including the EWS to it.
Demands for Reservation In recent years, certain groups have been making vociferous demands to be included into the list of those OBC groups, who can avail reservations. The most vocal among these are the Marathas, from the state of Maharashtra, the Jats from Haryana, and the Patidars from Gujarat. These groups are part of the larger Hindu sub-group category of the Shudras. These groups are traditionally landed into agriculture or related activities or into small businesses. The reason for their demand for reservations, despite data showing that these groups are better-off, with respect to asset ownership, occupation type, and average household incomes, than the SC, ST, and OBC, is their inability to transition, intergenerationally from their traditional occupation to jobs in the post-liberalization markets of India (Jaffrelot and Kalaiyarasan 2019). They feel, they have been lift out in the new, liberal economy, while groups that have access to reservation have managed to attain higher education through quotas in public educational institutions and transitioned into the market space. Since the sole purpose of reservations was to be a policy tool to provide representation to groups that have historically been systematically excluded from accessing education and confined to only caste-ascribed jobs, these demands for reservation by the abovementioned three groups fall outside the original ideology or thought process behind this positive discrimination policy. While the Supreme Court of India has rejected these demands, the Jats have been accorded OBC status and reservation in the State of only Haryana, as on 29th March 2019. Supreme court in 2019 also struck down the reservation granted to Maratha
34
Caste Quotas in India
791
community in the state of Maharashtra. As mentioned above, in a recent judgment the Supreme court, in contravention of the sprit and principle of the idea behind reservation, upheld the central government’s decision to accord an additional grant of a 10% quota in public educational institutions and jobs, for the economically weaker section (excluding the SC/ST and OBC groups) for the upper cast population. This is also in contravention of Supreme court’s earlier judgments, with respect to granting of reservation to other backward caste groups, where it evoked the argument that total reservation or quotas should not accede 50% of all positions or seats. It however has made an exception and allowed an additional 10% of seats being reserved, however, this time not for the historically discriminated and excluded social group, but for the privileged and dominant group member that are economically weak. This exception by the Supreme court has come under criticism, for applying the 50% reservation cap in one instance and not another and is seen more as a politically motivated move rather than one based on constitutional norms and underlying principles.
Contemporary Reality Seventy-four years of constitutional provisions in independent India is a very short period of time, for group members, to, in any meaningful way, overcome the economic, social, and psychological disadvantages of systematic denial of rights, exclusion, discrimination, and violence, and its consequences passed on intergenerationally, over a period of anywhere between 2200 and 2500 years. This manifests itself in among other, a persisting gap in the development indicators, between them and the rest. If we compare a few broad well-being indicators for group members, with non-members we can see this gap (Table 1). Well-being measures, as indicated in the table above, are seen to be compromised for the SC (& ST), and even the OBC, in some cases, when compared with that of the Others (the so-called upper castes). While this gap has reduced significantly over time, it seems difficult, if not impossible, to bridge. This reduction in the gap can largely be credited to the provisions of quotas for these three groups (Vani 2007) and the supporting policy measures. Despite these provisions, the persisting gap signifies not just the burden of historic but continuing exclusion and discrimination, in many spheres of economic, social, and political life. Contemporary studies indicate that social, household, and individual mindsets rooted in religious and social belief persist and result in continuing discrimination and exclusion and might even be getting worse in recent years. Studies show that 30% of households reported that at least one family member practiced untouchability with members of the lower castes, in a pan-India face-to-face survey13 (Thorat 2020). When the same (identical) question was asked in a telephonic survey14 across 13 14
India Human Development Survey Social Attitudes Research, India
792
A. Thorat
Table 1 Well-being indicators Average land size owned (2013, acres)a Gross enrolment ratio (2014)b School drop-out rate (2017–18)c Neonatal morality rated Infant mortality rated Poverty (2011–12)e Share of combined value of assets* (2013)f Average value of assets owned/householdf
ST 1.61 15.1 18.8 33 45 43 n.a n.a
SC 0.67 20.2 14 31 44 29.4 7.6 6.2
OBC 1.48 28.6 11.1 30 42 20.7 n.a n.a
Others 2.03 43.4 9.6 24 33 12.5 41 27.7
a
Source: Land and Livestock survey of India, 2003 and 2013 (authors calculations) Source: National Sample Survey (NSS) data on social consumption 2014 c Source: 75th round NSS data on education, 2017–18 (authors calculations) d Source: National Family Health Survey (NFHS-4): 2015–16 (authors calculations) e Source: NSS Consumption Expenditure Data (authors calculations) f Source: Debt and Investment survey, NSS, 2013 b
five states and two metro cities of India, it was found that more than 62% of rural population of the state of Uttar Pradesh (state with the largest population in India, around 230 million) and 40% of the population of Delhi reported practicing untouchability (Coffee 2018).
Criticism of Quotas While reservation has been observed to be the only policy tool that has worked exceedingly well in bringing the historically marginalized and discriminated groups into the mainstream and increased their representation (Borooah et al. 2007), in education and public sector jobs, and eventual intergeneration social mobility, these measures have not been popular, even unwelcomed by the privileged groups, who fall outside the purview of these measures. Studies (Coffey et al. 2018) based on primary perception surveys, in seven states and two metro cities of India, show that anywhere between 30% and 70% of members of the Brahmin and the so-called “forward” caste communities expressed their opposition to reservation policies, as opposed to 60–90% support for the same from the SC and OBC communities. The recent 10% reservation provided to the economically backward classes (EWS), for the privileged caste groups, as mentioned above, is seen as a move to counter both the constitutional motivations that were the basis of providing fixed quotas, namely, caste-based historic discrimination and introduce class as a basis. Thus, this is a move away from caste as the criteria for reservations to class as a metric, as well as extending it to privileged caste groups. There has also been a narrative that has existed, largely a opinion, that those who use reservation to avail public sector jobs are less efficient and productive in terms of work output, as their minimum criteria for job application might be slightly lower than the rest of the population (e.g., 5% grace for minimum qualifying scores at
34
Caste Quotas in India
793
graduation level), to compensate them for their historic and continuing discrimination, as the same criteria for evaluation cannot be applicable to them, as the historically discriminating and privileged caste group members. The one study that has examined this precise claim (Deshpande and Weisskopf 2014) examined India’s and the world’s largest public sector undertaking, the Indian Railways, that employed 13,26,437 people in 2014/15.15 The railways, like all government public sector employers, hire workers in four categories. The top two categories, A and B consist of professionals and administrative officials, while category C and D workers, are clerical and semi-skilled staff. The study examines, whether higher proportion of affirmative action employment reduces efficiency in the railways system. It not only finds no evidence of reduced efficiency, but, in the case of A and B category employees, found them to be more efficient that the general category employees. Another criticism of affirmative action policies in terms of reservation of seats for student in educational institutions, that is levied, is that students from marginalized communities are under prepared to meet the academic rigors of the institutions they are enrolled into. This is exemplified according to some, by the fact that many students from marginalized communities drop out of courses enrolled into or score poorly, citing their inability or merit as a reason for not being able to cope with the academic content. Most of these students are either first-, second-, or at most third-generation learners lacking educational, social, and cultural capital, which students from non-reserve categories have received as intergenerational transfers over a millennia and more. It would be unfair and frankly a goss injustice to expect them to catch up and hit the ground running, at the time of their entry into educational institutions. While no one can doubt their desire to do well and their motivation to work hard and make good of the opportunity, to receive good education and transform their lives, they do need encouragement, mentoring, some level of handholding, and remedial classes to overcome their historic disadvantage. In fact what has been observed and documented systematically by numerous studies is the opposite. These students report facing active and subtle discriminatory and exclusionary behavior from their teachers as well as their peers at the educational institutes where they are enrolled (Thorat 2008, Sukumar 2022). A government appointed committee (s) (Thorat committee report 200716), under the aegis of the University Grants Commission, looking into allegations of gross discrimination, exclusion, and segregation of students at an elite Medical college, in the national capital region (All India Institute of Medical Sciences), found systematic and pervasive discrimination of category students at all level, by the teaching faculty, the administrative staff, as well as the students. Other studies have highlighted similar experiences at leading universities of India.
15 16
Number of people Employed by Indian Railways Thorat committee report, 2007
794
A. Thorat
A case in point was the social, political, and institutional discrimination and exclusion of Rohit Vemula, a Ph.D. scholar at the University of Hyderabad, India, who committed suicide on 17th January, 2016. His death sparked nation-wide protests and outrage and came to be recognized as a case of state-sponsored discrimination against a Dalit student in India. Various other media reports and systematic studies have also pointed out the overt, subtle, and micro aggressions that category students face in campuses of higher educational institutions on a daily basis, in addition to bullying and physical abuse and violence.
Conclusion It would be fair to surmise that the provisions of social identity-based quotas in India for the scheduled caste, the scheduled tribes, and the other backward caste groups have been the only effective policy intervention that has helped group member get representation and participation and therefore progress, despite continuing discriminatory and exclusionary behavior in public sector appointments and hiring in educational and government jobs. It has also been the only measure that intended to work as a compensatory mechanism, albite insufficient, to provide representation to group members for historic wrongs. It is seen that reservations have indeed been critical for improving the lives of group member. In the absence of these measures, one wonders what would be their fate today. Since the transition of the Indian economy from a closed socialist to a capitalist one, there has been privatization of a large segment of federal jobs, shrinking the space for reserved positions. In recent years, the growth of educational institutions has been largely in the private sector, where reservations apply, but are not always accompanied with lower fees and other concessions, as provided in public sector educational institutions, thus reducing access for group members. While there have been demands for reservations to be applied to the private sector, this has not been an issue taken up seriously either politically or has formed part of serious academic and civil society concerns. Industry too is not keen to implement reservations; they, however, have shown openness for voluntary affirmative action. However, observations and reviews indicate that industries have performed dismally in their approach and implementation of affirmative action policies. As more of the economy inevitably will get privatized, in the long run and as representation falls and access becomes difficult, serious affirmative action policies or quotas might have to be though as a future political solution, to be extended to the private market space.
Bibliography Borooah VK, Dubey A, Iyer S (2007) The effectiveness of jobs reservation: caste, religion and economic status in India. Dev Chang 38(3):423–445 Coffey D, Hathi P, Khurana N, Thorat A (2018) Explicit prejudice. Econ Polit Wkly 53(1):47
34
Caste Quotas in India
795
Deshpande A, Weisskopf TE (2014) Does affirmative action reduce productivity? A case study of the Indian railways. World Dev 64:169–180 Jha V (2018) Candala: untouchability and caste in early India. Primus Books, Delhi Jaffrelot C, Kalaiyarasan A (2019) The political economy of the jat agitation for other backward class status. Econ Polit Wkly 54(7):29–37 Mohan PS (2015) Modernity of slavery: struggles against caste inequality in colonial Kerala. Oxford University Press, India Shah G, Mander H, Thorat S, Deshpande S, Baviskar A (2006) Untouchability in rural India. Sage Publishing, India Thorat S, Attewell P (2007) The legacy of social exclusion: a correspondence study of job discrimination in India. Econ Polit Wkly 42:4141–4145 Thorat A, Joshi O (2020) The continuing practice of untouchability in India. Econ Polit Wkly 55(2):37 Yamazaki GI (1997) Introduction: social discrimination in ancient India and its transition to the medieval period. In: Caste System, Untouchability and the Depressed. Manohar, New Delhi
The Effectiveness of Affirmative Action Policies in South Africa
35
Rulof Burger, Rachel Jafta, and Dieter von Fintel
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The South African Labor Market and Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Before Democracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Since Democracy in 1994, But Before Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Affirmative Action Legislation in South Africa: From Employment Equity to Broad-Based Black Economic Empowerment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Redress and Access: Affirmative Action Legislation in the 1990s . . . . . . . . . . . . . . . . . . . . . . . . . Critique and Review: A Brief Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Multidimensional Measures: Broad-Based Black Economic Empowerment . . . . . . . . . . . . . . Another Road to B-BBEE: The Sector Transformation Charters and Codes . . . . . . . . . . . . . . Amendments Since 2012: Toward Simplifying Complexity and Fostering Flexibility? . . . Empirical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
798 800 800 801 803 804 805 806 808 809 812 814 814 814 816
Abstract
Affirmative action policies in South Africa are best described as cumulative and becoming increasingly complex over time. The first national affirmative action policy, the Employment Equity Act, was implemented in 1999. This policy was followed by the Broad-Based Black Economic Empowerment Act in 2003, which had a broader focus and stronger compliance incentives. In 2007, the Codes of Good Practice were introduced to support the B-BBEE legislation by prescribing explicit and measurable benchmarks to South African firms and the establishment of institutional structures that monitored its implementation. Since 2009, several industry-specific Sectoral Charters and Codes have also been adopted. R. Burger (*) · R. Jafta (*) · D. von Fintel (*) Stellenbosch University, Stellenbosch, South Africa e-mail: [email protected]; [email protected]; [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_39
797
798
R. Burger et al.
A comparison of racial and gender earnings gaps over this period indicates that it was only after the introduction of the Codes of Good Practice that racial and gender earnings gaps started narrowing. This suggests that the details of AA legislation matter: policies that lack clear and quantifiable goals that are poorly monitored, or that offer weak incentives to comply, are unlikely to be effective. Strengthened monitoring and implementation has, in this case, been achieved by clearer General Sector Codes that support the legal framework. This has decentralized implementation to specific economic sectors, moving legal requirements higher up the priorities of business decision-making. Although such legislation was partially successful in reducing between-group earnings gaps, this only occurred at the top of earnings distribution and remained substantial. Keywords
Affirmative action · Discrimination · Employment equity · Labor market policies · South Africa
Introduction The rationale for affirmative action (AA) legislation and practices in South Africa is firmly rooted in addressing the long-run effects of the country’s colonial and apartheid past. Both historical periods entrenched discrimination in many aspects of South Africans’ lives, but particularly in the labor market. Stark wealth inequalities between black and white groups characterized South African society at the onset of democracy in 1994. This racial discrepancy was a defining challenge for the first democratically elected government. The majority of South Africans were excluded from owning productive assets, and the economic participation of black people was limited. The story of South African inequality is rare in this aspect: redress for the injustices of the past had to be effected for a large majority, constituting more than 80% of the population. A simple redistribution strategy was not feasible, especially in the context of the imperative for reconstruction and reconciliation set by the first democratically elected President, Nelson Mandela (Cobb 2013). Gary Becker (1993), in his Nobel Prize acceptance speech, emphasized that discrimination by a minority of a majority in apartheid South Africa was too costly to sustain. Apartheid policies were ineffective, and by the 1970s, the racial wage gap started to narrow, even if it remained large. On the one hand, the high costs of discrimination should have supported the continuation of these trends. By implication, discrimination should have gradually declined and become negligible in size in response to economic forces in South Africa. On the other hand, racial inequality was large and vexing and entrenched in South African economic relations. Market forces would therefore not necessarily unbundle discrimination. This is because discrimination was the product of very long-run informal institutions and accepted norms, and not only of formal legislation. Therefore, removing discriminatory legislation (as was the case at South Africa’s transition to democracy) does not
35
The Effectiveness of Affirmative Action Policies in South Africa
799
remove the practice and impacts of discrimination in the workings of society (Gradín 2019). Labor market discrimination did not decline in the early years after democracy. Affirmative action legislation was one solution tabled to dismantle inequality inside the labor market and to eliminate racial earnings gaps. The Employment Equity Act in 1998, the 2003 Broad-Based Black Economic Empowerment (BBBEE) Act, and the 2007 General Sector Codes that flowed from this legislation are the main affirmative action policies that supported these objectives. Rather than removing historical obstacles to broad-based participation in the labor market, the policies intended active redress and to create new opportunities for designated groups that historically faced formal and informal discrimination. Though the evidence on the efficacy of affirmative action policy in South Africa is mixed, the discussion around “economic freedom” of black South Africans continues to dominate public discourse. After more than 25 years of democracy and 20 years of formal affirmative action legislation, racial differences in income and wealth remain large. The very early versions of affirmative action policies had only limited positive effects on small groups of designated South Africans. Ponte et al. (2007) highlight widespread perceptions that the dividends of affirmative action were limited to a minority of black elites. Robinson (2014) shows how politically connected black elites benefited from joining multiple boards of historically white companies, creating a relatively small network of real beneficiaries. By 2007, a highprofile government advisory panel consisting of local and Harvard economists concluded that there should be a change in emphasis from racial patterns in firm ownership, to developing skills of broader sections of society (Acemoglu et al. 2007) – the call was for broad-based change. Burger and Jafta (2006) showed that early affirmative action policies had a small effect on narrowing the discriminatory wage gaps at the very top of the earnings distribution. However, there has been no noticeable reduction in the wage gap at other parts of the earnings distribution (Burger and Jafta 2006, 2010). Similarly, Klasen and Minasyan (2020) show that early BBBEE legislation only improved the representation of black women in top management positions. Improved methodology points to small impacts after 2003 (Burger et al. 2016). The limited impact of the early phases of affirmative action therefore raises the question as to how the policy could be adapted to reach its intended goals. Less empirical evidence is available to assess the impact of affirmative action after policies continued to evolve in the post-BBEEE era. In 2007, the General Sector Code was developed to create a framework for sectors to implement affirmative action. Specific sectors implemented their own codes in later years. The BBBEE Act was amended to improve monitoring and to impose fines on firms that did not meet intended targets. These amendments addressed shortcomings of the early phases. Firstly, they intended to strength compliance. Secondly, they created a framework for decentralized monitoring. In this chapter, brief empirical evidence is presented to fill the gap for this period. Results emphasize that strengthening the policies started a new trajectory of narrowing racial wage disparities, though they remain high.
800
R. Burger et al.
The South African case is instructive for thinking about the progression of affirmative action legislation. Firstly, designing successful affirmative action policies must reckon with deeply entrenched historical norms and does not only serve the purpose of dismantling formal historical discriminatory legislation. Secondly, when the intended beneficiaries constitute a large majority of the population and if the initial inequalities are very large, the benefits of legislation may be limited to small sections of society. Thirdly, effective affirmative action requires implementation that is located closer to business decision-makers and may require decentralized monitoring to have success. The rest of this chapter starts by reviewing South Africa’s history with discrimination, both before and after the transition to democracy. Section “Affirmative Action Legislation in South Africa: From Employment Equity to Broad-Based Black Economic Empowerment” provides a detailed overview of affirmative action legislation, while section “Empirical Analysis” provides a brief empirical overview of racial wage gaps since the start of democracy up to 2012. Section “Conclusion” offers final remarks.
The South African Labor Market and Discrimination Before Democracy Labor market discrimination is central to understanding South Africa’s economic history. Starting with the arrival of European settlers in 1652, both slavery and indentured work created a racially segmented, class-based labor market under Dutch and British colonial dispensations (Delius and Trapido 1982; Ross 1993). Many of the earliest occurrences of discrimination were not formally legislated. Rather, informal workplace practices, the propagation of the slave trade, and conquest favored European settlers above the indigenous population and slaves, who were brought to South Africa from other parts of the world (including Asia). These historical events and norms created extreme levels of inequality between race groups, and overall inequality was high by historical and international standards (Fourie and von Fintel 2010, 2011). While informal practices characterized these inequalities, formal legislation progressively entrenched them. The “Colour Bar Act” was introduced as early as 1911, shortly after the unification of former colonies that created South Africa’s modernday borders. This legislation institutionalized the reservation of semiskilled and skilled work for white individuals. However, the most famous discriminatory legislation was the battery of laws introduced by the apartheid regime that came to power in 1948. The white government pursued policies of explicit racial “separate development,” which stretched far beyond the labor market. They regulated political participation, social interactions, the education system, and most other facets of life. Limitations on black African political participation extended into the workplace; opposition from the apartheid regime complicated the task of organizing labor, and black African unions struggled to advocate for fairness in the workplace. At the
35
The Effectiveness of Affirmative Action Policies in South Africa
801
height of occupational segregation in 1970, Knight and McGrath (1977) estimated that white men were paid as much as 5.7 times as much as what black men earned. The multiple was between 3.6 and 5.0 once adjusting for differences in education. South Africa’s Gini coefficient at this time was 0.68, with nearly two-thirds of income inequality accounted for by differences across race groups (Leibbrandt et al. 2009). During the 1970s, these patterns started to change somewhat. Unions could organize more effectively; and job reservation proved less effective with time. Mariotti (2012) argues that job reservation became progressively redundant and unraveled without legislative change. As levels of education among white individuals rose, many of them chose to enter highly skilled jobs. This left many semiskilled jobs vacant. These vacancies therefore gradually became accessible to black individuals, whose education levels were also on the increase. The racial gap in educational attainment was 7.2 years for individuals born in 1920 and halved to 3.6 years for the 1980 birth cohort (Van der Berg 2007). Consequently, wages slowly started to converge across race groups, though the large historical discrepancies did not disappear completely. This process continued, even if apartheid labor market legislation remained in force. Moll (2000) documented a continued decline in racial wage discrimination from 1980 to 1993. While differences between race groups narrowed, within-racial inequality increased, as educational attainment of some black workers increased. In effect, by the 1990s, the Gini coefficient had not declined from the high levels recorded in the 1970s and as far back as the seventeenth century. However, at the dawn of democracy, less than half of the income gap could be ascribed to between-race differences (Leibbrandt et al. 2009). Rising education levels, narrowing educational differences, and increased black worker bargaining power therefore contributed to narrowing racial wage gaps during apartheid and long before affirmative action was proposed.
Since Democracy in 1994, But Before Affirmative Action By 1996, when a new constitution was adopted, the democratically elected government was committed to implementing affirmative action policies to aid this process. However, the narrowing of the wage gap that started in the 1970s did not continue in the intervening period between democratization in 1994 and the first affirmative action legislation in 1998. Sherer (2000) estimated a racial wage gap for 1995 (after accounting for differences in human capital) that was larger than the estimates that Moll (2000) presented for 1993. Mwabu and Schultz (2000) show that more than half of the racial wage gap could be accounted for by differences in educational attainment. Erichsen and Wakeford (2001) also estimated an increase in wage discrimination toward black men between 1993 and 1995. The trend of a widening racial wage gap continued between 1995 and 1997 (Allanson et al. 2002). More broadly, Rospabé (2002) showed that racial employment discrimination reduced somewhat over the period 1993–1999, even if wage and occupational discrimination increased during the same time. Brookes and Hinks (2004) also provide evidence over a longer time horizon and find that the racial employment gap grew between
802
R. Burger et al.
1995 and 2002. Democratic political institutions that removed formal discriminatory practices did not guarantee that there would be convergence. A number of binding constraints that emerged before the transition placed limits on the economy to reverse the gaps without legislative redress. As the status quo started to shift in the 1970s, the economy simultaneously underwent a structural shift toward skills-intensive production, and other demographic trends also influenced the labor market. Regarding the latter, labor force participation increased among black Africans, and especially black women. The feminization of the labor force followed growth in educational attainment, lower marriage rates, and urbanization after the relaxation of apartheid-era restrictions on movement and residence (Casale and Posel 2002). Though black women have become more integrated into the potential labor force, they have not transitioned into jobs at the same rate, contributing to higher unemployment. Structural changes in the economy meant that the chances for low-skilled jobseekers to be matched with work were on the decline, despite growing in bargaining power. Since the 1970s, employment has become increasingly concentrated in highly skilled tertiary sector work (Bhorat and Hodge 1999; Banerjee et al. 2008). Excess demand for skilled workers and excess supply of unskilled jobseekers progressively entrenched structural unemployment. A new state of high equilibrium unemployment emerged and is now a defining characteristic of South Africa’s post-transition labor market (Burger and von Fintel 2014). In 1995, shortly after the transition, the official unemployment rate was 17.6%. This figure has never dropped below 20% after 1998. In the years since the 2008 financial crisis, unemployment has risen unabated from a post-2000 low of 21.5% to 29.1% in the first quarter of 2020, even before the fallout from the COVID-19 pandemic intensified this situation. The least skilled – many of whom are black – bore the brunt of skill-biased technical change. This long-run trajectory translated to a widening racial gap in unemployment: the unemployment rate was 18 percentage points higher for black compared to white individuals in 2008, and this gap has grown to 22 percentage points shortly before the onset of COVID-19. Nevertheless, solving overall between-race inequality relies more heavily on narrowing the racial wage gap than the differences in unemployment. Wage inequality explains more than two-thirds of total income inequality, with unemployment contributing to less than one-sixth (Leibbrandt et al. 2009). However, long-run labor demand shifts also reflect in highly convex returns to education. Individuals with low levels of education have almost no wage returns to schooling; by contrast, individuals with tertiary education have very high rates of return inside the labor market (Keswell and Poswell 2004). Similarly, the relationship between socioeconomic status (SES) and school performance is also highly convex. School quality differs vastly along the income distribution and serves to reinforce existing racial inequalities in the labor market (Van der Berg 2007). Schools that were historically reserved for white children under apartheid, and that still tend to serve high SES communities, continue to give learners an advantage that is crucial for transitions into tertiary education and into the job market. More pertinently, school quality is a binding constraint to obtaining well-paid, highly skilled work in a labor
35
The Effectiveness of Affirmative Action Policies in South Africa
803
market context where skill-biased technical change requires entrants to have advanced education. Indeed, more than one-third of unexplained racial wage gaps are accounted for by racial differences in school quality (Van der Berg et al. 2011). Another important constraint to reducing wage gaps is hiring within social networks. Schoer et al. (2014) estimate that two-thirds of jobseekers find work through personal contacts. Hofmeyr (2010) shows that manufacturing jobs are ethnically concentrated, with job referral inside social networks dominating where workers find jobs. Relying on referrals reduces jobseekers’ propensity to enter new occupations, industries, and local labor markets that are historically occupied by individuals from other ethnic groups. Gradín (2019) confirms that occupational segregation has remained highly stable since the onset of democracy up to 2015, regardless of the implementation of affirmative action legislation and the improvements in black human capital. In particular, black individuals remained more concentrated in low-paying occupations than white people. Other racial groups have fared better with time, and the occupational differences between them and whites have subsided. Networks therefore still tend to stream particularly black individuals into separate, low paid occupational clusters. The persistence of informal hiring practices therefore reduces the speed at which wage gaps narrow.
Affirmative Action Legislation in South Africa: From Employment Equity to Broad-Based Black Economic Empowerment Affirmative action measures in South Africa are best described as cumulative and becoming increasingly complex over the years. Although preparatory work1 to put in place the architecture for AA was already underway ahead of the first democratic elections in 1994, it was a facilitating mandate in the newly adopted Constitution of the Republic of South Africa Act (No. 108 of 1996) which gave impetus to the introduction of the first AA legislation. The Constitution makes provision for government to take measures that will effect i) redress and ii) equality. The first is backward looking in the sense that it seeks to counter the legacy of past discrimination, while the second is forward looking toward a future where equality would be more than an ideal (Burger and Jafta 2010: 5). The implication was that AA and other policies should aim for redress, but also put in place measures to enable the beneficiaries to gain access, and progress by improving their education and skills, among other things (Burger and Jafta 2006). The first two pieces of legislation to incorporate this task were the Employment Equity Act (No. 55 of 1998) and the Skills Development Act (No. 97 of 1998). We elaborate on these in the next section.
1
See Burger and Jafta (2010: 3–5) for an overview of this work.
804
R. Burger et al.
Redress and Access: Affirmative Action Legislation in the 1990s The Employment Equity Act (No. 55 of 1998) aimed to increase the participation of black South Africans in the labor market in terms of quantity and quality (Department of Trade and Industry 2000). The purpose of the Act is to eliminate unfair discrimination in the labor market and enact positive measures that result in the attraction, development, and retention of previously marginalized groups. These groups include blacks (African, Colored,2 and Indians), women,3 and people with disabilities. The Act requires that employers consult with employees and trade unions, identify employment equity (EE) gaps in their labor force, compose welldefined EE plans with targets and timelines, and submit these plans and report progress against these plans to the Department of Labor. In addition, income differentiation reports must be submitted. It is the responsibility of the inspectors at the Department of Labor and the Employment Equity Commission to monitor compliance (Burger et al. 2016). Since the intent was also to improve the mobility of previously disadvantaged groups, a set of legislation, the Skills Development Act of 1998, and the Skills Development Levies Act (1999) focused on skills formation and the means to fund it. The latter obliges all employers with a payroll of R500 000 and more to commit to contribute 1% of their payroll to the relevant Sectoral Education and Training Authority. The Skills Development Act (No. 97 of 1998) was intended to compel all employers to commit to investing in the human capital of designated groups through training and education. The Act explicitly focuses on skills development or industry-based learning and was accompanied by a significant overhaul of the education and skills formation landscape. The Skills Development Act replaced the Manpower Training Act of 1981, while the National Skills Authority (NSA) replaced the National Training Board established by the Manpower Training Act. The NSA is a statutory body instituted under the Act in 1999, which is tasked with advising the Minister of Labor (NSA 2020). Furthermore, Sectoral Education and Training Authorities (SETAs) replaced the Industry Training Boards, apprenticeships, and training centers. This is an important break from the past in that an entirely new institutional framework, which was integrated with the National Qualification Framework (NQF), was created. This framework administration is responsible to provide learnerships (instead of apprenticeships) and finance skills development using the National Skills Fund (NSF – a successor of the Manpower Development Fund) and lastly provide for and regulate employment services (Reddy et al. 2018).
2
This term is used in the official government racial classifications but is controversial in broader society. We use it here for consistency. 3 Including white women; although they have not suffered racial discrimination under apartheid, they did not have equal access under the prevailing patriarchal system (Klasen and Minasyan (2017:4)).
35
The Effectiveness of Affirmative Action Policies in South Africa
805
If all went according to plan, one would have expected to see improvements in: i. The elimination of discrimination, for example, reducing wage gaps across races for the same value of work. ii. Broadened access to employment for designated groups. iii. Improved upward mobility to higher occupation levels. iv. Improvements in education and skills formation to enable labor market progress. However, the AA implementation did not go smoothly, and, within the first few years of its existence, calls for review and broadening were heard with regularity.
Critique and Review: A Brief Summary Affirmative action was never universally accepted and drew criticism from various quarters, including those not in favor of race-based legislation and those in favor, but who believed that it did not go far enough. Review of the legislation and its implementation (see, e.g., Bezuidenhoud et al. 2008; Msimang 2000; Bhorat et al. 2014) highlighted several points of critique. The criticism boils down to the following: • Widespread noncompliance. • Weak sanctions for noncompliance, with only the possibility and not the prescription of a fine. • Capacity problems on the part of government, especially the Department of Labour, with respect to implementation and monitoring. • Data inconsistencies, e.g., at the Employment Equity Commission, making inference from the data problematic. • The creation of a black elite, which was contrary to the intent of broad access. • Slow progress on education and skills development. • Employment equity is too narrow a focus, whereas deeper and broader intervention is required to ensure meaningful participation of the majority in the economy. On the latter part, both the Black Management Forum and the Black Economic Empowerment Commission weighed in. Even before the enactment of the EE Legislation, the Black Management Forum asserted that efforts geared to correcting economic imbalances were too weak and success will remain elusive (Mparadzi 2014). The Black Economic Empowerment (BEE) Commission, under the leadership of Cyril Ramaphosa, presented its report to President Mbeki in 2001. The report set out the rationale for casting the net wider over the economy to achieve empowerment and defined BEE as follows: It is an integrated and coherent socio-economic process. It is located within the context of the country’s national transformation programme, namely the RDP (Reconstruction and Development Programme). It is aimed at redressing the imbalances of the past by seeking to
806
R. Burger et al.
substantially and equitably transfer and confer the ownership, management and control of South Africa’s financial and economic resources to the majority of its citizens. It seeks to ensure broader and meaningful participation in the economy by black people to achieve sustainable development and prosperity. — BEE Commission Report, (2001: 2).
The onus then shifted to the Department of Trade and Industry (now the Department of Trade, Industry, and Competition) to prepare and propose legislation to give effect to the intent of Black Economic Empowerment. The result was the BroadBased Black Economic Empowerment Act 53 of 2003 and the Codes of Good Practice to follow the Act.
Multidimensional Measures: Broad-Based Black Economic Empowerment The South African government developed the Broad-Based Black Economic Empowerment Act (B-BBEE) (2003) to promote economic empowerment and participation of all black people. The beneficiaries of this legislation are defined as “African, Colored, and Indians” with a specific focus on women,4 youth, workers, people with disabilities, and people living in rural areas, to ensure economic justice, through integrated socioeconomic strategies (Department of Trade and Industry 2003). This legislation goes beyond employment equity and skills development to include ownership and control, as well as provisions for social development and access to the economy for black businesses. This legislation also takes steps toward quantitative indicators against which compliance can be measured. Whereas companies previously set their own EE plans and targets, the Department of Trade and Industry’s Codes of Good Practice (known as the generic scorecard) now prescribes targets (in percentages) to be achieved by designated entities. The B-BBEE Act, which mandated the Minister of Trade and Industry to develop the Codes of Good Practice to give concrete content to the intent outlined therein, was assented to on the seventh of January 2004. The Codes of Good Practice were only gazetted much later, on the ninth of February 2007, and became effective in August 2008. The Codes serve to standardize the definition of B-BBEE to provide explicit and measurable benchmarks and institutional structures that can facilitate the implementation and evaluation of this expanded transformation undertaking. In short, the Codes are a tool to deliver deeper empowerment by unifying the system and providing content to the regulatory framework. The time elapsed between the Act and the Codes can provide an explanation for why B-BBEE was not effectively implemented in this period.
4
Note that in this Act, white women, who are included in the Employment Equity Act as previously disadvantaged on a gender basis, are now excluded.
35
The Effectiveness of Affirmative Action Policies in South Africa
807
The apparatus of monitoring an evaluation now also included verification agencies, specializing in the verification of compliance to obtain a BEE certificate. According to the Department of Trade and Industry (www.dtic.gov.za), the South African National Accreditation System must conduct accreditation of verification agencies. Compliance with these codes is mandatory for any company that wants to do business with the state or seek approval from the state, for instance, in obtaining licenses. A powerful instrument in the hands of the government is preferential procurement. BEE certificates are therefore required by the organs of the state when procurement is done. This has a permeating effect throughout the economy in that even if a firm is not directly doing business with government, it may have clients or suppliers that do, which means its BEE status is important to their own progress (Bowmans 2020: 1). The involvement of private sector verification agencies decentralized implementation and monitoring of B-BBEE away from government. The Department of Trade and Industry initially created the “balanced scorecard,” which contains percentage weightings. The latter formed the basis for a firms’ “BEE accreditation.” Moreover, the 2007 generic scorecard initially comprised of three core components and seven sub-elements of B-BBEE used to assess firms’ compliance with the B-BBEE legislation. The first component is that of “Direct Empowerment” which comprises of sub-elements that measure the degree of “Ownership” and “Managerial Control” of the firm by historically marginalized group, while the “Human Resource Development” component considers the “Skills Development” which entails the training of historically marginalized groups conducted by the firm and “Employment Equity” as outlined earlier in section “Redress and Access: Affirmative Action Legislation in the 1990s.” Finally, the “Indirect Empowerment” component consists of three sub-elements, namely, “Preferential Procurement” from black enterprises, “Enterprise Development,” and the “Residual Elements” which focuses on socioeconomic development (Department of Trade and Industry 2003). A company’s performance would be assessed for each of the elements, and all the scores added up to arrive at a point out of 100. It is possible to exceed 100 points when bonus points are earned, for example, procurement from a majority black woman-owned business. BEE levels range from 1 to 8, where 1 is the top contributor. Scoring less than 40 points amounts to being labeled a noncomplier. The scorecard composition and weightings are differentiated by business type, namely, large enterprises, small to medium-sized enterprises which are termed exempted microenterprises (EMEs) and qualifying small enterprises (QSEs) and start-up enterprises, foreign companies, and multinational corporations (Bowmans 2020: 12–13).5 BEE certificates last for 1 year. In subsequent amendments, the “balanced scorecard” would be discarded in favor of priority elements and more stringent measures introduced (see section “Amendments Since 2012: Toward Simplifying Complexity and Fostering Flexibility”). In addition to the Generic Codes, several industries and sectors in consultation with
5
The permutations are many and could get rather complex. See https://www.bowmanslaw.com/ insights-landing/guides/ for a full illustration.
808
R. Burger et al.
government and labor unions started negotiation of Sector and Industry Charters and Codes more appropriate to the particular circumstances of the industry or sector. These Charters and Codes were different from the top-down AA and B-BBEE legislation in that it involved a lengthy and robust negotiation process often involving business executives and consideration at board level. This signaled the possibility of more strategic “buy-in.” We take a closer look at these Charters and Codes in the next section.
Another Road to B-BBEE: The Sector Transformation Charters and Codes The Sectoral Charters and Codes seem more endogenous relative to the AA legislation, since they originated at the industry and sector level where the driving force was the need to design solutions that will promote BEE in an industry-specific way. The initiation of the Industry Transformation Charters predates the issuing of the Generic Codes, where the initial motivation was to attempt transformation on a sector’s own terms, i.e., to avoid government intervention and to gain some flexibility in the form of trade-offs between elements of the scorecard. For example, the first Transformation Charters in liquid fuels, mining, and finance all had different targets for ownership at 25, 26, and 10 percent, respectively. The idea was then to compensate by committing to a higher weighting in other categories to achieve the overall transformation level (Rumney 2010). When the Generic Codes were issued in 2007, a process began to adapt the Industry and Sector Charters and Codes to align to the Codes issued in terms of the B-BBEE Act. The result is that there are now three types of Charters and Codes. The Sector Charters gazetted in terms of Section 12 of the B-BBEE Act are those formulated by industries that simply reflect their commitment to BEE. They are not legally binding on the organs of state and public entities. On the other hand, the Sector Codes gazetted under Section 9(1) have been approved by the Minister of Trade and Industry and carry the Status of the Codes of Good Practice. The Sector Codes, like the Generic Codes, prescribe the benchmarks for BEE compliance that the entities in the different sectors must meet. They seek to level the playing field and maximize the contribution of all companies by considering the unique features of their industries (Strata-g 2018). At the point of writing, there have been nine Sector Codes6 gazetted under Section 9(1) capturing key industries in the South African economy. The Forest Sector Code and Integrated Transport Sector Codes came into effect in 2009. This was followed by the Chartered Accountancy Sector Code in 2011. Subsequently, in 2012, the Agri-BEE Charter, Financial Sector Charter, and the Property Sector Charter became effective. The Tourism Sector Code was adopted in 2015. Finally,
6
A lengthy discussion of the Codes does not serve the purpose of this chapter, but a detailed explanation can be found at www.dtic.gov.za and a user-friendly summary at https://www. bowmanslaw.com/insights-landing/guides/
35
The Effectiveness of Affirmative Action Policies in South Africa
809
the Minister of Trade and Industry gazetted the Information and Communication Technology (ICT) Charter and the Marketing Advertising and Communication (MAC) Sector Code in 2016 (Bowmans 2020). The BEE transformation apparatus, including the Generic and Sector Codes and Charters, have come in for criticism when their impact on the economy and society is considered. We incorporate the key points of criticism in the next section when we recount subsequent amendments to legislation and regulations, which, in part, attempt to address these points of critique.
Amendments Since 2012: Toward Simplifying Complexity and Fostering Flexibility7? Between 2012 and 2019, the government reviewed and amended several pieces of the affirmative action legislation, including the Employment Equity Act and the B-BBEE Act, as well as the Codes of Good Practice. This follows several years of less than desired progress and public criticism. The criticism falls mostly into three categories: i) Flawed design of the legislation and supporting infrastructure. This ranges from assertions that the proper balance between “carrot” and “stick” in ensuring compliance is not struck to a more structural concern that an overemphasis on the ownership component detracts from the elements such as skills development and enterprise development that is likely to benefit smaller enterprises and people at the lower end of the income distribution (Bhorat et al. 2014). It is further argued that the emphasis on measurement turns the process into a tick-box exercise with little real impact in changing mindsets to view BEE as a strategic issue. Indeed, in the annual BEE surveys conducted by KPMG (2013), 50% of respondents indicated that they do not consider the economic value of BEE for their business and effectively view it as a sunk cost. The Minister of Labor’s response to yet another report on noncompliance gives insight: It is this state of affairs that leaves us with no option, but to consider, drafting-in harsher consequences for non-compliance. It’s time to ‘up the ante’ and this may include promulgating the ‘stick’ sections of the Employment Equity Act because quite frankly, the ‘carrot’ sections have not delivered the desired results. (Workinfo 2017)
ii) Implementation-related critique includes the complexity of the number of elements and sub-elements in the Codes, to the cost associated with achieving some of the thresholds, such as ownership and skills development. The quality of
7
With recognition to Reddy et al. (2018), for the use of this apt phrase.
810
R. Burger et al.
education remains a lament, and so do concerns about loopholes that allow fronting (the pretense of empowerment, while power remains firmly in the hands of non-designated persons). iii) Lastly, monitoring remained problematic. While the decentralization of monitoring and introduction of independent verification agencies improved matters, the emergence of nonaccredited BEE verification agencies caused inconvenience and had cost implications for non-suspecting companies (DTI 2015). In popular debates, it is often said that legislation should have “more teeth.” In the amendments to the legislation governing BEE, the government sought to give more teeth to improve compliance but also aimed to simplify and address some of the other concerns discussed above. To this end, the sanctions in both EE and B-BBEE became more stringent. The equal pay for equal work principle is now entrenched as a legal requirement, and noncompliant employers can be referred directly to the Labor Court (previously there was an opportunity to respond or remedy the situation). Penalties in the form of fines have increased significantly.8 With respect to the BEE legislation, penalties for BEE fronting practices, misrepresentation, and related offences have been introduced. The new penalty for contravening this Act includes fines up to 10% of annual turnover, and individuals can be fined and/or imprisoned for up to 10 years. They can also be barred from contracting with state-owned entities for 10 years. Cancellation of authorization of existing contracts with the state granted on account of false information will result (DTIC 2020)9. On the enabling front, the designated employers are required to enhance enterprise development beyond financial assistance, by, for example, offering mentoring and training. To improve skills development, permanent employment for learners is required to get absorption points. Increased scope for SETA funding claims, i.e., stipends linked to bursaries, is now included as skills development cost (Bowmans 2020). Lastly, to decrease complexity, the elements in the BEE scorecard have been simplified, while priority elements have been introduced to align with government’s intention of broad-based empowerment. The 2013 B-BBEE amendment which became effective in 2014 reduced the seven B-BBEE sub-elements of the Codes to five. On the new scorecard, the “Employment Equity” sub-element no longer stands alone but was compressed into the “Management Control” sub-element. The “Indirect Empowerment” component is comprised of the newly established “New Enterprise and Supplier Development” which is made up of the “Preferential Procurement” and “Enterprise Development” sub-elements. The amendment further introduced new weightings for measuring B-BBEE status. The Ownership sub-element increased from 20 plus 3 bonus points to 25 points. The amended “Management Control” sub-element has a score of 15 plus 4 bonus points, and the
8
For details, please see https://www.bowmanslaw.com/insights-landing/guides/ All B-BBEE legislation, Codes, and regulations are available at http://www.thedtic.gov.za/ financial-and-non-financial-support/b-bbee/b-bbee-codes-b-bbee-acts-strategies-policies/
9
35
The Effectiveness of Affirmative Action Policies in South Africa
811
recently introduced “New Enterprise and Supplier Development” sub-element carries a weight of 40 plus 4 bonus points. This amendment encourages the support of other black enterprises in the economy. The “Socioeconomic Development” sub-element’s weighting remains unchanged. Lastly, the total number of available points has increased from 107 to 118 (Werkmans 2015). In November 2019, significant changes to the B-BBEE Codes of Good Practice came into effect. Key amendments introduced by the DTI include changes in recognition requirements for exempted small enterprises (EMEs) and qualifying small enterprises (QSEs) which clarifies that these entities are only eligible for enhanced B-BBEE recognition if they are 51 percent or 100 percent black-owned. The second key amendment revised the targets for “New Enterprise and Supplier Development” sub-element, increasing the procurement target from companies which are 51 percent black-owned from 40 percent to 50 percent of the total procurement budget. Moreover, the points awarded for the support of such enterprises increased from 9 to 11. This benefit is extended to large enterprises that are 51 percent black-owned. The third amendment of the Codes introduces the skills development expenditure on bursaries for black students. Finally, for the “Preferential Procurement” scorecard, a new term “designated group supplier” is defined as a supplier that is not only 51 percent black-owned but is owned in part by the currently identified disempowered black groups, specifically unemployed people, youth, people with disabilities, people living in rural areas, and/or black military veterans (Kassen 2019). This new term reflects the government’s intention to spread preferences even beyond the “new black elite.” As to the question of whether these amendments will deliver better results, the picture is nuanced. The KPMG Surveys (2012, 2013) reported increased commitment to BEE, signaling that BEE has become less of a tick-box exercise. Respondents reported that the strongest reasons for implementing BEE measures were that their customers required it, to avoid reputational damage, and that their board of directors were taking a meaningful interest. This is opposed to the view that the legislation drives commitment. On the other hand, the B-BBEE Commission, an entity established in terms of the B-BBEE Act to monitor compliance, reveals in its 2020 report that between 2018 and 2019, compliance worsened for public as well as private sectors. Of the listed entities required to submit compliance reports in terms of Section 13G of the Act, 42% were in legislative compliance in 2019, whereas it was 15% for state and public entities. Furthermore, 69% all reporting entities did not respect reporting submission deadlines. On the BEE indicators, it is striking that 53% (2018:60%) of listed entities achieved the priority elements, while 39% (2018:50%) of organs of state did (B-BBEE Commission 2020). Against this backdrop of institutional changes and mixed reported outcomes, the next section includes an empirical analysis of racial wage gaps. In particular, it assesses the effect of later AA legislation, which was associated with improved monitoring and guides for implementation.
812
R. Burger et al.
Empirical Analysis One of the key objectives of AA legislation is to reduce the wage gap between groups that have historically benefitted from discrimination – primarily white males – and those that have borne the cost of discrimination: black, Colored and Indian South Africans, and women. Other objectives, like greater upward mobility and improved skills formation for marginalized groups, should also have the effect of narrowing the earnings gap. The effectiveness of AA legislation in achieving its intended objectives can therefore be assessed by using survey data to investigate trends in earnings gaps between different demographic groups. In this section, we estimate such trends on a series of nationally representative household surveys between 1997 and 2012, a period during which several pieces of AA legislation were enacted and for which reliable South African survey data exists. If a specific law was effective, we would expect to see a downward trend in the earnings gap after its enactment. Earnings gaps are estimated at different points of the earnings distribution, to investigate whether those at the bottom, middle, or top of the earnings distribution benefitted from the legislation. Individual observations were reweighted to normalize the education and age distributions of the different demographic groups across different trends, so time trends reveal changes in discrimination rather than human capital. The methods and data used in the section are discussed in the Appendix. Figure 1 shows the evolution of the earnings gaps at the 25th and 90th percentiles between 1997 and 2012.10 Not surprisingly, earnings gaps between white men and all other demographic groups are exceptionally large at the beginning of sample period, shortly after the end of apartheid. These gaps are largest for black Africans, followed by Coloreds and Indians, and larger for women than men. This pattern is observed at all points in the earnings distribution. There is no systematic decrease in any of these gaps, or at any point of earnings distribution, following the enactment of either the EE legislation in 1998 or the B-BBEE Act in 2003. Since it is not possible to observe the counterfactual – what would have happened to earnings gaps in the absence of this legislation – it is not possible to conclude that the policies had no effect on labor market outcomes. Nevertheless, it does not appear that either of these policies was effective in their stated objective of improving the relative earnings of marginalized groups. The first signs of narrowing earnings gaps are observed around 2007, which coincides with the implementation of the Codes of Good Practice. Figure 1 reveals a substantial and significant decrease in the gaps for black, Colored, and Indian men (vis-a-vis white men) at the 90th percentile of the earnings distribution. However, no decrease is observed for men at the 25th percentile. Turning to women, we observe decreases for Colored and Indian women and an even larger decrease for black women after 2007 at the 90th percentile. As was the case for men, we see no downward trends at the 25th percentile. This evidence suggests that the
10
Similar comparisons were performed for the 10th, 50th, and 75th percentiles but are not reported.
35
The Effectiveness of Affirmative Action Policies in South Africa
813
Fig. 1 Estimated earnings gaps, relative to white men: 1997 to 2012. Source: own calculations
implementation of the Codes of Good Conduct, which brought about quantified affirmative action goals for South African firms, as well as improved monitoring, was somewhat successful in improving the labor market outcomes of designated workers at the top of the earnings distribution. The same comparison at other parts of the earnings distribution provides further support for this interpretation. The empirical analysis therefore suggests that more recent affirmative action policies were a qualified success: it appears to have reduced earnings gaps relative to white men at the top of the earnings distribution but had no effect at the middle and bottom of the earnings distribution. It is also worth pointing out that white men continued to earn significantly more than any other demographic group at the end of the period under consideration, even after the implementation of more stringent affirmative action legislation.
814
R. Burger et al.
Conclusion The first national affirmative action policy, the Employment Equity Act, was implemented in South Africa in 1999. This policy was followed by the B-BBEE Act in 2003. Whereas the Employment Equity Act included relatively weak sanctions for noncompliance – the possibility of a fine and public “naming and shaming” – B-BBEE provided stronger compliance incentives such as preferential procurement to compliant firms. In 2007, the Codes of Good Practice were introduced to support the B-BBEE legislation by prescribing explicit and measurable benchmarks to South African firms and the establishment of institutional structures that monitored its implementation. A comparison of racial and gender earnings gaps over this period indicates that it was only after the introduction of the Codes that racial and gender earnings gaps started narrowing. This suggests that the details of AA legislation matter: policies that lack clear and quantifiable goals that are poorly monitored, or that offer weak incentives to comply, are unlikely to be effective. The Latin American experience suggests that removing discriminatory policies is insufficient to reduce group differences (Gradín 2019). South Africa’s experience shows that implementing active antidiscrimination laws that do not have sufficient provisions for monitoring and implementation also has limited effect on changing informal practices that have persisted over centuries. However, the precondition for affirmative action to have an effect – however small – is that the laws are clear to implement and that they are monitored. Strengthened monitoring and implementation has, in this case, been achieved by clearer General Sector Codes that support the legal framework; this has decentralized implementation to specific economic sectors, moving legal requirements higher up the priorities of business decision-making. Further, monitoring was decentralized to private sector accreditation agencies who work more closely with business to meet requirements and to expand the capacity for monitoring the legislation. Although this chapter demonstrates that South African affirmative action policies were partially successful in reducing between-group earnings gaps, some caveats apply. Earnings gaps only narrowed at the top of earnings distribution and remained substantial even after the enactment of aggressive affirmative action legislation. Marginalized groups also continue to have much higher rates of unemployment. This suggests that even effective affirmative action policies on their own alone are unlikely to eradicate income inequality or poverty. Affirmative action policies need to be supported by other interventions, such as continued improvement in access to high-quality schools, on-thejob training programs, and active labor market policies aimed at marginalized groups.
Appendix Data and Methodology The empirical analysis uses pooled cross-sectional data to investigate the evolution of earnings gaps in the South African labor market. South Africa’s national statistics agency, Statistics South Africa, has collected nationally representative
35
The Effectiveness of Affirmative Action Policies in South Africa
815
household surveys with an extensive labor module since 1994. These surveys were conducted annually as the October Household Survey (OHS) between 1994 and 1999, semiannually as the Labour Force Survey (LFS) between 2000 and 2008, and quarterly as the Quarterly Labour Force Survey (QLFS) since 2008. We use the Postapartheid Labour Market Series (PALMS) version 3.3 (Kerr et al. 2019) for our analysis. There are several challenges to comparing the different cross-sectional surveys to obtain time trends. The early surveys experienced several changes in sampling frames and questionnaire design. By the 1997 OHS, many of these issues had been resolved, so we use this as the first survey for our analysis. Although the QLFS contained questions about labor earnings since its inception in 2008, no earnings data were officially released for the surveys between 2008 and 2010. Furthermore, whereas prior to 2008 the unimputed earnings data was made available, since 2010 earnings data included imputations undertaken by Statistics South Africa. As noted in Kerr and Wittenberg (2020: 2), these imputations are of “low quality,” especially after 2012Q2. We therefore ignore all the data for 2013 onward. Differences in earnings across demographic groups can occur for several reasons, including differences in educational attainment and life-cycle participation patterns. Since this chapter is primarily interested in how affirmative action legislation has affected discrimination that occurs after workers leave school and enter the labor market, we need to control for such differences before estimating earnings gaps. In our empirical analysis, we achieve this in two ways. First, we limit the sample to those with overlapping support for all demographic groups. Effectively this means dropping everyone with fewer than 10 years of schooling from the sample, since this outcome was very rare for the white population group, and increasingly also for young workers from the other population groups. We also reweight our sample to produce demographic groups with comparable schooling and age distributions across all years. We achieve this by following a method similar to that originally proposed by DiNardo et al. (1996). A logit regression is used to regress each of the eight binary demographic group indicators on a series of education dummy variables and age splines, both interacting with survey dummies. The inverse of the predicted value from this regression is then used to adjust the survey weights, and this adjusted weight is used when calculating the percentiles of the wage distribution in different years for different groups. This approach effectively reweights each of the groups in each of the years so that the education and age distributions more closely resemble that of the entire population for the full period. This implies that our estimated earnings gaps will not be driven by any between-group differences in schooling or age. Furthermore, our estimated time trends will be unaffected by time trends in the schooling levels or age distribution of workers. Confidence intervals that accommodate reweighting are bootstrapped with replicate weights using the procedures outlined by Kolenikov (2010).
816
R. Burger et al.
References Acemoglu D, Gelb S, Robinson J (2007) Black economic empowerment and economic performance in South Africa. Report of the Harvard Advisory Panel. National Treasury, Pretoria Allanson P, Atkins JP, Hinks T (2002) No end to the racial wage hierarchy in South Africa? Rev Dev Econ 6(3):442–459 Banerjee A, Galiani S, Levinsohn J, McLaren Z, Woolard I (2008) Why has unemployment risen in the New South Africa? Econ Transit 16(4):715–740 B-BBEE Commission (2020) National status and trends on broad based black economics empowerment, VO4.00. [Online]. Available: https://www.bbbeecommission.co.za/wp-content/ uploads/2020/07/National-Status-and-Trends-on-Broad-Based-Black-EconomicEmpowerment_.pdf Becker G (1993) Nobel lecture: the economic way of looking at behavior. J Polit Econ 101(3): 385–409 BEE Commission (2001) Black economic empowerment commission report. Skotaville Press, Johannesburg Bezuidenhoud A, Bischoff C, Buhlungu S, Lewins K (2008) Tracking progress on the implementation and impact of the Employment Equity Act since its inception. Research Commissioned by the Department of Labour, Cape Town, South Africa Bhorat H, Hodge J (1999) Decomposing shifts in labour demand in South Africa. South Afr J Econ 67(3):349–380 Bhorat H, Hirsch A, Kanbur R, Ncube M (eds) (2014) “The Oxford Companion to the Economics of South Africa,” Oxford University Press, Oxford Bowmans (2020) Guide 1-B-BBEE in South Africa: an overview. [Online]. Available: https://www. bowmanslaw.com/insights-landing/guides/ Brookes M, Hinks T (2004) The racial employment gap in South Africa. South Afr J Econ 72(3): 573–580 Burger RP, Jafta R (2006) Returns to race: racial discrimination in the South African labour market. Stellenbosch economic working paper 04/06. Stellenbosch University, Stellenbosch Burger RP, Jafta R (2010) Affirmative action in South Africa: an empirical assessment of the impact on labour market outcomes. CRISE Working paper No. 76. University of Oxford, Oxford Burger RP, von Fintel DP (2014) Rising unemployment in a growing economy: a business cycle, generational and life cycle perspective of post-transition South Africa’s labour market. Stud Econ Econ 38(1):35–64 Burger RP, Jafta R, von Fintel D (2016) Affirmative Action policies and the evolution of postApartheid South Africa’s racial wage gap. UNU-WIDER working paper 2016-66 Casale D, Posel D (2002) The continued feminization of the labour force in South Africa: an analysis of recent data and trends. South Afr J Econ 70(1):156–184 Cobb J (2013) Mandela and the politics of forgiveness [Online]. Available: https://www.newyorker. com/news/news-desk/mandela-and-the-politics-of-forgiveness Delius P, Trapido S (1982) Inboekselings and Oorlams: the creation and transformation of a servile class. J South Afr Stud 8(2):214–242 Department of Trade & Industry (2003) A strategy for broad-based black economic empowerment. Pretoria Department of Trade and Industry (2000) Trade and Industry. 2000. Black Economic Empowerment Commission: briefing by Cyril Ramaphosa C & Ndlovu, L [Online] Available: https:// pmg.org.za/committee-meeting/191/ DiNardo J, Fortin N, Lemieux T (1996) Labor market institutions and the distribution of wages, 1973–1992: a semiparametric approach. Econometrica 64:1001–1044 DTI (2015) The dti Black Industrialist Policy, towards full-scale industrialization and inclusive growth. [Online]. Available: http://www.thedtic.gov.za/wp-content/uploads/BI_Policy2015-1. pdf
35
The Effectiveness of Affirmative Action Policies in South Africa
817
DTIC (2020) Overview of the B-BBEE Commission in Transforming the Economy and Promoting BEE, October 30. [Online]. Available: http://www.thedtic.gov.za/wp-content/uploads/B-BBEE30-Oct-2020.pdf Erichsen G, Wakeford J (2001) Racial wage discrimination in SA before and after the first democratic election. DPRU working paper, 001/49. DPRU, University of Cape Town, Cape Town Fourie J, von Fintel DP (2010) The dynamics of inequality in a newly settled, pre-industrial society: the case of the cape colony. Cliometrica 4(3):229–267 Fourie J, von Fintel DP (2011) A history with evidence: income inequality in the Dutch Cape Colony. Econ Hist Dev Regions 26(1):16–48 Gradín C (2019) Occupational segregation by race in South Africa after apartheid. Rev Dev Econ 23:553–576 Hofmeyr A (2010) Social networks and ethnic niches: an econometric analysis of the manufacturing sector in South Africa. South Afr J Econ 78(1):107–130 Kassen S (2019) Recent Amendments to the B-BBEE Codes, ENS Africa, [Online]. Available: https://www.worldservicesgroup.com/publications.asp?action=article&artid=12234 Kerr A, Lam D, Wittenberg M (2019) The Post-Apartheid labour market series v3.3 (PALMS) Keswell M, Poswell L (2004) Returns to education in South Africa: a retrospective sensitivity analysis of the available evidence. South Afr J Econ 72:834–860 Klasen S, Minasyan (2020) Affirmative action and intersectionality at the top: evidence from South Africa. GLO discussion paper 467. Global Labor Organisation, Essen Knight J, McGrath MD (1977) An analysis of racial wage discrimination in South Africa. Oxf Bull Econ Stat 39(4):245–271 Kolenikov S (2010) Resampling variance estimation for complex survey data. Stata J 10(2): 165–199 KPMG (2012) BEE Survey 2012. Shifting the BEE landscape. [Online]. Available: kpmg.co.za KPMG (2013) BEE Survey 2013. BEE coming of Age. [Online]. Available https:// transformationpartners.co.za/documents/library/bee/research/KPMG_BEESurvey_2013.pdf Leibbrandt M, Woolard C, Woolard I (2009) Poverty and inequality dynamics in South Africa: postapartheid developments in the light of the long-run legacy. In: Aron J, Kahn B, Kingdon G (eds) South African economic policy under democracy. Oxford University Press, Oxford Mariotti M (2012) Labour markets during apartheid in South Africa. Econ Hist Rev 65:1100–1122 Moll P (2000) Discrimination is declining in South Africa but inequality is not. J Stud Econ Econom 24(3):91–108 Mparadzi A (2014) BEE - basis, evolution, evaluation – a critical appraisal of black economic empowerment in South Africa. University of Cape Town. [Online]. Available: https://open.uct. ac.za/handle/11427/4516?show¼full Msimang S (2000) Affirmative action in the New South Africa: the politics of representation, law and equity. Women in Action 1(2):1–6 Mwabu G, Schultz TP (2000) Wage premiums for education and location of South Africa workers, by gender and race. Econ Dev Cult Chang 48(2):307–334 National Skills Authority (2020) Roles and functions. [Online] Available: http://www. nationalskillsauthority.org.za/roles-and-functions/ Ponte S, Roberts S, van Sittert L (2007) Black economic empowerment. Business and the State in South Africa. Dev Chang 38(5):933–955 Reddy V, Wildschut A, Luescher T, Petersen I, Rust J (2018) Skills development legislation as a lever of change to reduce poverty, inequality and unemployment. (Commissioned by a High Level Parliamentary Panel (HLP) of the South African Parliament, March) Republic of South Africa (1996) Constitution of the Republic of South Africa, Act 9 (3). Government Printer, Pretoria Robinson JA (2014) Elites and institutional persistence. In: Amsden AH, DiCaprio A, and Robinson JA (eds) The Role of elites in economic development. Oxford: Oxford University Press, pp 29–49
818
R. Burger et al.
Rospabé S (2002) How did labour market racial discrimination evolve after the end of apartheid? An analysis of the evolution of employment, occupational and wage discrimination in South Africa between 1993 and 1999. South Afr J Econ 70(1):185–217 Ross R (1993) Emancipations and the economy of the Cape Colony. Slavery Abolit 14(1):131–148 Rumney R (2010) BBBEE code versus sector charters, Mail and Guardian, 23 December. [Online]. Available: https://mg.co.za/article/2010-12-23-bbbee-code-versus-sector-charters/ Schoer V, Rankin N, Roberts G (2014) Accessing the first job in a slack labour market: job matching in South Africa. J Int Dev 26:1–22 Sherer G (2000) Intergroup economic inequality in South Africa: the post-apartheid era. Am Econ Rev 90(2):317–321 Strata-g (2018) Sector codes and.or transformation charters. [Online] Available: https://www.stratag . c o . z a / w p - c o n t e n t / u p l o a d s / S E C TO R - C O D E S - a n d o r - T R A N S F O R M AT I O N CHARTERS-.pdf Van der Berg S (2007) Apartheid’s enduring legacy: inequalities in education. J Afr Econ 16(5): 849–880 Van der Berg S, Burger C, Burger R, de Vos M, du Rand G, Gustafsson M, Moses E, Shepherd D, Spaull N, Taylor S, van Broekhuizen H, von Fintel D (2011) Low quality education as a poverty trap. Stellenbosch economics working papers 25/11 Werkmans (2015) The amendment to the BBBEE and the codes explained. [Online] Available: https://www.werksmans.com/wp-content/uploads/2015/04/042870-WERKSMANS-bbbeebooklet-sp.pdf Workinfo (2017). http://www.workinfo.org/index.php/articles/item/1805-minister-oliphant-employ ment-equity-stats-remain-unchanged-9-may-2017
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need of a Rethink
36
Hwok-Aun Lee
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Foundations, Contexts, and Frameworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Constitutional Premises and Political Imperatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Policy Platforms and Public Discourse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mechanisms and Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Policy Achievements, Shortfalls, and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
820 821 822 824 829 831 836 837
Abstract
Malaysia maintains one of the world’s most extensive affirmative action regimes, buttressed by the transformative and iconic New Economic Policy (NEP). Constitutional provisions, political imperatives, and socioeconomic conditions gave rise to the establishment of preferential policies in four broad sectors – higher education, employment, enterprise, and ownership – favoring the political dominant but economically disadvantaged Bumiputera majority. This chapter elucidates the origins, programs, outcomes, and implications of affirmative action in Malaysia. A brief historical overview explains the language and context of the constitutional authorization of Bumiputera quotas and the modest implementation in the early post-independence years, followed by policy expansion, centralization, and intensification from 1971 under the NEP, which was forged in the aftermath of May 13, 1969, racial conflagration. The NEP judiciously conceptualized a two-pronged strategy of poverty eradication regardless of race and “social restructuring” through Bumiputera-targeted affirmative action as distinct but complementary elements of the ultimate goals of national integration, which H.-A. Lee (*) ISEAS-Yusof Ishak Institute, Singapore, Singapore e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_40
819
820
H.-A. Lee
entails redressing imbalances and ultimately rolling back overt preferential treatment. However, the NEP lacked a systematic articulation of policy objectives, instruments, and outcomes. Malaysia has registered immense progress in facilitating Bumiputera access, participation, and upward mobility in the four designated policy sectors. Recent discourses have popularized misguided notions of reform that conflate the NEP’s twin elements and omit attention to the decisive shortfall of affirmative action – its inefficacy in building capability and competitiveness among the Bumiputera beneficiaries, which are requisite for Malaysia to attain the ultimate NEP goals. Malaysia has substantially remedied destabilizing inequalities but, moving forward, must fundamentally rethink affirmative action. Keywords
Malaysia · New Economic Policy · Affirmative action · Inequality
Introduction In Malaysia’s storied history of development planning, the New Economic Policy (NEP) stands out for its transformative impact and iconic stature. Launched in 1971, the NEP was a vision statement, but also a pivotal template of change, an imprimatur for sweeping measures that rolled out in subsequent decades. The country’s conditions – having a politically dominant but economically disadvantaged majority group, known as the Bumiputeras – reinforced the policy’s political imperative and socioeconomic rationale and the extensive scale and scope of affirmative action. The magnitude of preferential treatment and Bumiputera-exclusive programs heightens both the potential and peril of these measures while also underscoring the need for effective implementation such that the Bumiputeras can graduate out of institutionalized and overt group preferences. Undeniably, modifying and rolling back the policy pose profound challenges. While successful in remedying imbalances and Bumiputera underrepresentation, through facilitating access and expanding opportunity, Malaysia has faltered on the attendant mission of developing capability and competitiveness. Beyond facilitating educational degrees and mobility into professional and management positions, the policy must also engender acquisition of knowledge and skills, resourcefulness, and self-reliance. Affirmative action has unquestionably expanded Bumiputera presence in the middle and upper socioeconomic strata, bolstering the community’s esteem and influence and enhancing inter-racial integration. The policy also persists as a source of tension, due to the ways it attenuates minority group opportunities. Malaysia’s experience shows the imperative of going beyond expanding access to equipping policy beneficiaries to graduate out of overt preferential treatment. It is also important to note that, while AA is mandated to contribute to national integration, this ultimate objective derives from settlements in other domains besides economic opportunity. Safeguarding minority groups’ preservation and expression of culture, religion, and language, especially through Chinese and Tamil vernacular
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
821
schools, have contributed to a sense of compromise and mutual accommodation. However, these other mainstays of Malaysia’s integration endeavors stretch beyond the scope of this chapter. It is timely to reflect on the NEP today, as its presence in Malaysia continues beyond 50 years. The NEP demonstrates the efficacy of a two-pronged strategy with poverty eradication and affirmative action as distinct but reinforcing elements which have narrowed income gaps, grown middle classes, and fostered upward mobility through group preferential programs in higher education, high-level occupations, enterprise, and ownership, undergirded by broad social programs and basic provisions in education, health, infrastructure, and connectivity. The enormous scope and substantial achievements of Malaysia’s affirmative action stand out internationally. However, Malaysia’s experience must be viewed with caveats and in context. Quotas and reservations for the majority Bumiputeras stem from constitutional provisions, on the basis of their “special position” – which grants exceptional force to such interventions, including the creation of racially exclusive programs that are non-replicable in most countries. In practice, Malaysia immensely facilitated Bumiputera access, mobility, and ownership, but lack of attention to developing capability and competitiveness of policy beneficiaries constrained the policy’s overall attainment and detracted from needed reforms in line with the NEP’s long-term objective of being a transitory intervention. Recent policy discourses, while stemming from noble principles, have also become muddled by conflating the NEP’s two prongs and setting up need-based, pro-poor assistance as a substitute for affirmative action, instead of viewing the former as complementary and reinforcing the latter. This chapter surveys and analyzes the foundations, mechanisms, outcomes, and implications of affirmative action in Malaysia. The next section discusses constitutional, historical, and political aspects that planted the policy from Malayan independence in 1957 and vastly expanded it from 1971 under the NEP. This is followed by a discussion of policy practice and rhetoric, particularly the NEP’s two prongs – poverty eradication and “social restructuring” – with a focus on the second which corresponds with affirmative action. The NEP judiciously distinguished both elements and clarified their complementarity but inadequately pursued capability development while giving inordinate attention to equity ownership and, in recent times, mainstreaming misguided policy alternatives. The chapter then outlines the mechanisms and specific programs that Malaysia has implemented and evaluates the policy outcomes. The conclusion unpacks some policy implications, for both the country’s continual implementation of affirmative action and other countries looking to draw lessons from the Malaysian experience.
Foundations, Contexts, and Frameworks Malaya gained independence in 1957. Malaysia was formed in 1963 through the merging of Malaya, subsequently referred to as Peninsular Malaysia, with the Borneo states of Sabah and Sarawak and Singapore. Singapore departed Malaysia
822
H.-A. Lee
in 1965. Demographically, Malaysia’s current citizen population of 31 million comprises 70% Bumiputeras (Malay for “sons of the soil”), an official category subsuming the Malays (56%) who predominantly reside in Peninsular Malaysia and non-Malay Bumiputeras, or other indigenous groups mainly in Sabah and Sarawak (14%). Among minority groups, Chinese make up 23%, Indians 7%, and “others” about 1%. This chapter adopts these racial categories but with circumspection, mindful of problematic conceptual frameworks and colonial origins (Hirschman 1986).
Constitutional Premises and Political Imperatives Malaysia introduced affirmative action while emerging from colonialism and its legacies of racial stratification and complex inequalities. The country intensified the policy amidst robust and sustained economic growth, urbanization, industrialization, and expanding higher education. The extensive scope of AA derives from the historical, demographic, and structural conditions, of a politically dominant and economically disadvantaged majority group, within an economy offering a broad range of opportunity available for distribution. Affirmative action was institutionalized at Malaya’s independence. The 1957 Malayan constitution incorporated provisions for affirmative action that were first articulated in the 1948 Federation of Malaya Agreement. Article 153 of the constitution authorizes the national king to safeguard Bumiputera “special position. . . in such manner as may be necessary,” through reserving a “reasonable proportion” of positions in public sector employment, scholarships, training, and licensing. Bumiputera reservations in college and university admissions were added in 1971 through a constitutional amendment (Malaysia 2010). These provisions arose out of socioeconomic contexts and political bargains. Social stratification in colonial Malaya corresponded with race, most evidently in the residential and occupational locations of the main constituent groups. The overall population comprised 49.8% Malays, 37.2% Chinese, and 11.3% Indians, but the urban population was starkly different: 22.6% Malays, 63.9% Chinese, and 10.7% Indians. Overwhelmingly, Malays lived in rural settings and worked as farmers or government officers especially in the police and security services, while Chinese settled in urban areas and got involved in mining, manufacturing, and commerce, and Indians were mainly plantation workers and civil servants (Andaya and Andaya 2001). Malays accounted for 95.8% of rice farmers, 83.2% of police and security services, 76.8% of the armed forces, and 52.4% of government services. The Chinese were the vast majority of the workforce in manufacturing (72.2%), mining (68.3%), and commerce (65.1%), while Indians were over-represented in government services (26.3%), rubber cultivation (24.5%), as well as commerce (16.8%). British companies ruled the apex sectors of the economy – tin mining and plantations – under colonial administration that engineered these structures of racial stratification and separation. Poverty was common among all communities, but most acutely among the Malays and Indians. Official sources estimated poverty incidence at
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
823
70.5% in the Malay population in 1957, compared to 35.7% among Indians and 27.4% among the Chinese (Khoo 2005). Economic insecurity arising from the above disparities and Malay disadvantage were heightened by extraordinary political circumstances. Malay royalty and the ruling class, having secured a commanding position in the independence movement and in negotiations with the British, galvanized Malay nationalism around an agenda of institutionalizing special provisions and establishing the monarchies and political elites as the community’s protectors (Amoroso 2014). Malaysia’s affirmative action joins a few other countries in favoring a majority group but stands alone for conferring “special position” on the policy beneficiaries. Some countries’ constitutions also authorize quotas and reservations for minorities, such as India, Pakistan, and Ghana, among former British colonies whose legal systems draw on common traditions. South Africa’s democratic constitution, a notable comparator to Malaysia for also explicitly permitting majority group affirmative action, premises such actions on the designated group being disadvantaged due to unfair discrimination. South Africa’s constitution was passed in 1996, amid unique domestic circumstances and a different global milieu. The marked contrast in the affirmative action laws and policies of Malaysia and South Africa underscores the impact of each country’s distinctive constitutional underpinnings (Lee 2016, 2021a). The constitution shaped affirmative action into two consequential ways. First, the articulation of the majority group’s “special position” as the basis for preferential treatment imputes a measure of permanence and immutability on the policy, since this status does not reference socioeconomic disadvantage or historical conditions that may change over time. In popular discourses, and even in academic literature, the term has also transmuted to Malay “special rights,” reinforcing its aura of indelibility and even painting a veneer of sanctity. Having omitted the principle of socioeconomic disadvantage as grounds for AA, Article 153 also scarcely provides reference for balancing majority preference with equitable opportunity for minorities.1 These issues have been negotiated politically and pragmatically, although the misleading invocation of “special rights” arguably pre-empts more substantive and constructive engagements toward modifying and enhancing affirmative action. Malaysia’s constitutional grounds for AA are truly exceptional, as noted by Sabbagh (2012) in his scholarly comparative work on the subject. The constitution drafting process was particularly engrossed in deliberations on Article 153, primarily surrounding whether reservations would be indefinite or time-bound or whether the
Article 153 stipulates that it is the national king’s responsibility to safeguard both the “special position” of the groups classified as Bumiputera and the “legitimate interests of other communities,” suggesting a mandate to balance contending claims. However, a closer reading finds that the specific provisions for the other communities only pertain to protection against arbitrary dismissal, confiscation, or non-renewal of employment, property, or contracts and licenses already in possession. Article 153’s protections for minority groups were limited to the independence transition; jobs, property, and contracts held before independence could not be taken away when the new nation was founded. Article 153 provides no constitutional oversight on minority group interests for any new, post-independence recruitment, property acquisition, or the award of contracts.
1
824
H.-A. Lee
provisions would be reviewed after an interval. Ultimately, limitations or schedules for review were not included (Fernando 2015). Conventional readings preclude the possibility of change over time. However, Article 153(2) also stipulates that the king shall exercise his functions “in such manner as may be necessary,” which can be interpreted to mean that implementing Bumiputera quotas is conditional upon the need for such measures. A second set of implications relate to the scope and mechanisms of affirmative action. In practice, affirmative action in Malaysia derives from Article 153 in that it predominantly operates in publicly owned entities – the civil service, public universities, and public procurement – and has largely omitted private sector operations and private education institutions. Interventions to compel workforce representation in the private sector lack the moral underpinnings present in countries with history of systemic discrimination or oppression, warranting some form of restitution in all segments of the labor market. Private entities in Malaysia were not former participants of systemic discrimination against the Malays, although the largely nonMalay-owned private sector has arguably privileged its established networks and supply chains. The colonial legacies of Malay exclusion and disadvantage and of racial stratification were real and serious, but the minority groups, especially the urbanized and economically advancing Chinese, were not politically dominant, which precluded the case for redistribution on the grounds of recompense. Operationally, Malaysia’s preponderant mode of AA is also traceable to Article 153. The constitution was drawing on precedent and reference to other countries but, intentionally or otherwise, solidified the practice of quotas and reservations, largely to the exclusion of other mechanisms of preferential selection.
Policy Platforms and Public Discourse Affirmative action rolled out in a gradual and piecemeal manner in independent Malaya’s early years. The state took a rather laissez-faire stance in general, with some pursuit of import substitution industrialization and rural development programs, but limited scope of affirmative action. Policies abided more literally to Article 153, in being confined to the public sector, the scholarships, and the specified areas of intervention. Some targeted initiatives unfolded in the 1960s, notably Bank Bumiputera which financed Bumiputera business and self-employment and MARA (Majlis Amanah Rakyat or Council of Trust for the People) which also provided financial services, in addition to education institutions and scholarships. However, the government faced mounting pressure to more vigorously promote Bumiputera economic participation, largely induced by discontent toward the slow socioeconomic progress, particularly in commerce and industry. The May 13, 1969, racial riots and bloodshed plunged the nation into crisis. Malaysia was reset, politically and economically. The ensuing years saw reassertion of Malay primacy in politics and policy and increased centralization of power in the executive branch (Ooi 2013). Intensive deliberations on economic policy settled on
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
825
more emphatic, concerted, and voluminous interventions to promote growth, reduce poverty, and remedy inter-racial inequality (Faaland et al. 1990). The New Economic Policy, launched in 1971, forcefully and comprehensively articulated a transformation vision and soberly outlined racial inequality statistics conspicuously absent in preceding planning documents (Malaysia 1971, Malaysia 1976). The empirical analysis zoomed in on three areas where the magnitude of disparity was especially stark: income, employment, and corporate wealth. In 1970, an estimated 49% were considered poor, but the incidence was much higher among Malay households (65%) than Indians (49%) and Chinese (26%). Malays accounted for 47.2% of professional and technical positions and 22.4% of administrative and managerial positions but were concentrated in the public sector, agriculture, and services (Malaysia 1976). The racial composition of occupation groups was sectorally disaggregated into primary (agriculture), secondary (manufacturing, mining, and construction), and tertiary (services). Malays accounted for 55.9% of professionals in the primary sectors and 48.7% in tertiary sectors but only 26.0% in the secondary sectors. Among managers, the Malay shares were 19.8% (primary), 34.5% (tertiary), and 14.7% (secondary) (Malaysia 1976). The NEP also heavily emphasized disparities in equity ownership. Share capital held by Malay individuals or trust agencies representing Malay interests accounted for a mere 2.4% of the total, alongside 34.3% held by other Malaysians and 60% under foreign ownership (Malaysia 1976). Education, of course, was a key component of the NEP but was outshone by the spotlight on employment and wealth. The disparities were substantial, even if not as prominently reported in the national planning documents. Malay students had seen a rise in their share of university enrolment from a very low bar of 20% of total enrolment in 1963/1964 to slightly above 40% in 1970, but the community’s participation in higher education was still lagging (Tan and Santhiram 2017). Notably, Malay students were grossly underrepresented in scientific and professional fields. At the University of Malaya, then the only comprehensive university in Malaysia, Malay graduates numbered 22 out of a total 493 in science, 4 out of 67 in medicine, 1 per 71 in engineering, and 15 per 49 in agriculture (Selvaratnam 1988). One of the most important contributions of the NEP was its judicious two-pronged agenda, of eradicating poverty regardless of race and accelerating social restructuring to reduce and eventually eliminate the identification of race with economic function (Malaysia 1971). Importantly, the NEP grasped the conceptual differences between interventions that ameliorate the conditions of poor households, principally by addressing basic needs and upholding equality, and measures that promote the representation, upward mobility, and participation of disadvantaged groups, which are premised on the political imperatives and socioeconomic desirability of fostering equitable representation. Persisting absence of Bumiputeras in higher education, in the upper echelons of employment, and in ownership and management of economic entities, emphatically corroborated by empirical evidence shown above, were deemed untenable, particularly in the aftermath of the May 13 conflagration.
826
H.-A. Lee
However, the NEP lacked a framework that clearly and coherently set out affirmative action policy objectives, mechanisms, and long-term implications. The NEP demonstrated some self-awareness that its prescriptions would broaden and intensify the system of preferential treatment to facilitate access and participation of Bumiputeras, to enable them to break out of their predominantly rural and agricultural economies. But the NEP did not systematically integrate the policy areas characterized by entry barriers and resultant problems of underrepresentation, which called for the application of Bumiputera preferential treatment. Such a perspective might have inclined the agenda to recognize higher education, highlevel occupations, enterprise development, and wealth ownership as inter-dependent policy sectors and to accordingly set objectives and targets centered on increasing participation and cultivating capability. Instead, two policy goals predominated: first, the workforce at occupational levels should reflect the national racial composition and, second, Bumiputera ownership of corporate equity should be increased to 30%. Appended to these overarching goals were the quest to develop a Bumiputera Commercial and Industrial Community (BCIC) and for education and skills training to broaden occupational participation. The NEP, presented as 9-page vision statement in 1971, also boldly proclaimed that “no particular group will experience any loss or feel any sense of deprivation” (Malaysia 1971: 1). While a noble principle and empathetic assurance, and perhaps a reasonable premise from the perspective that economic expansion would continuously avail more opportunities for all, this was also a wishful ideal and inflated promise with regard to the specific workings of affirmative action. A growing economy can continuously generate jobs and new wealth, but fundamental scarcity in various socioeconomic opportunities – public universities, public sector employment, and public procurement – would inescapably entail, to some extent, certain parties benefiting while others miss out. The NEP omitted grappling with the tensions between preferential treatment and equitable opportunity. The question of policy permanence versus transience subtly permeated the NEP. Its 1971–1990 bracket denoted a timeline, but in substance, the NEP did not articulate terminal plans. An ultimate objective of Bumiputera self-reliance – graduating out of the need for preferential treatment, which entailed rolling back affirmative action – was not explicitly declared, although the initial NEP of 1971 aspired for Bumiputeras to be “full partners” in the economy “within one generation,” implying an intention to make breakthroughs by 1990 (Malaysia 1971). Nevertheless, the two-decade time frame was not presented as a binding constraint, and no clear guidance was provided on the implications of achieving or missing numerical targets. A 1969 policy paper written by the NEP’s chief author Just Faaland had emphasized the need for “determined and consistent efforts over a long period well into the next century” (Faaland et al. 1990, emphasis added). Toward the mid-1970s, priorities also apparently shifted from broad capacity building to narrow ownership goals. When the NEP was fleshed out as a 40-page document in 1976, the “within one generation” quest was reduced to Bumiputeras holding 30% of total equity (Malaysia 1976).
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
827
Beyond 1990, the affirmative action regime remained solidly intact, albeit with modifications in line with political economic developments – particularly, the rapid propulsion of the BCIC agenda through privatization of formerly state-owned entities (Gomez and Jomo 1999). The overriding justification was Bumiputera equity ownership which by official accounts fell short of the 30% threshold. Indeed, discourses leading up to the succession of the NEP by the NDP were dominated by the question of whether the quota would be retained (Means 1990). Still, the quest for Bumiputera self-confidence and self-reliance has resurfaced at critical junctures. In 1991, the galvanizing Vision 2020 proclaimed the goal of a “fully competitive Bumiputera community. . . on par with the non-Bumiputera community” (Mohamad 1991). That the NEP could not complete its mission in 1990, and persists until today, continually spur debate. It is difficult to contend that racial disparities, as wide as they were, could be bridged within two decades. Indeed, no scholarly argument has been put forward claiming that the NEP had sufficiently fulfilled its objectives for the Bumiputera community and hence has outlived its purpose. The literature critiquing the policy’s perpetuation largely focuses on the political vested interests in continuing Bumiputera preferential treatment (Chin and Teh 2017; Gomez 2015). While the Malay political-business elites clearly favor continuity of privileged access, and many gained handsomely from wealth transfers, this explanation for the persistence of affirmative action critically omits the question of whether the Bumiputera population as a whole was sufficiently equipped to undertake change. It fails to account for the more consequential shortcoming of policy succession – that the NEP remained fixated on providing access and ownership, rather than enhancing capability and competitiveness. A more purposeful emphasis on the fact that the NEP’s second prong involved preferential treatment for the majority, and primarily involved facilitating access and opportunity, might have attuned the NEP toward the need for these interventions to cultivate capability, competitiveness, and confidence as vital supplementary objectives. Instead, over time, Bumiputeras owning 30% of total equity became a defining goal, shifting attention inordinately toward wealth acquisition. Inter-racial household income disparities were continuously tracked as benchmarks of the system’s efficacy in facilitating Bumiputera catch-up, but without clear causal links drawn between these empirical outcomes – which derive from a multitude of factors – and affirmative action. The lack of clarity surrounding policy objectives, instruments, and outcomes also detracted from a constructive differentiation of quantitative outcomes (such as enrolment in university or recruitment as professionals and managers) versus more qualitative outcomes (academic achievement of employability of graduates, mobility across private and public sectors) that are essential to affirmative action’s full realization. In the early stages, education access and quantitative policy goals accordingly took precedence, but education quality was not given added emphasis over time, as Bumiputeras continuously made quantitative progress. Affirmative action proceeded in the ensuing two decades after 1990, despite official handovers to the National Development Policy (NDP, 1991–2000) and the National Vision Policy (NVP, 2001–2010). The NVP never gained traction and
828
H.-A. Lee
dissipated in collective memory. The NDP featured in policy discourses and was often presented as a departure from the NEP, as a growth- rather than redistributioncentered program of action. Although investment regulations were liberalized from the late 1980s and Malaysia embarked on a massive privatization project, and private higher education burgeoned from the mid-1990s, all the major affirmative action instruments instituted under the NEP endured, and many were expanded, especially in post-secondary and higher education,2 while some new programs started, such as Bumiputera microfinance schemes Tekun and PUNB. Malaysia’s privatization did not proceed in a market fundamentalist manner of relinquishing the government’s stake, but amounted to corporatization, and in many cases public listing, of those entities by the state, which selected the new Malay corporate titans and retained effective control. Further reconfiguration took place with the renationalizing of many of the same entities – formerly corporatized – which had collapsed in the aftermath of the 1997–1998 Asian financial crisis (Gomez et al. 2017). These governmentlinked companies (GLCs) assumed the driving position in the Bumiputera enterprise development agenda. Pragmatism also prevailed rather than any ideological shift away from the NEP. For instance, the expansion of private higher education from the late 1990s relieved some of the socio-political pressure arising from non-Bumiputera communities, since it provided alternatives for those who perceived their chance of admission to public universities was too slim or fallback options for those who applied but did not get a place. The New Economic Model (NEM), introduced in 2010, arguably impacted on affirmative action discourses more than any other policy document since the NEP – but mainly in a regressive direction, by conflating the NEP’s twin elements’ prongs and muddling policy thought (NEAC 2010). The notion that Malaysia should, instead of targeting the Bumiputeras, help the poor – who need help the most – had percolated for years but became mainstreamed and massively popularized by the NEM. A raft of “reform” slogans is proposed to replace “race-based” affirmative action with “need-based,” “merit-based,” and “market-based” affirmative action. These anodyne notions, despite amounting to hollow slogans, resonated even in academic circles (Gomez 2015). Need-based affirmative action, as the NEM itself articulated, focused on basic needs provision and the livelihoods of low-income households. It did not constitute a systemic replacement for race-based affirmative action, despite the NEM’s pronouncements of reform. The NEM presented no coherent explanation for how the existing pro-Bumiputera affirmative action would be systemically replaced with pro-poor assistance or “merit-based” selection. It overwhelmingly spotlighted the problems of rent-seeking, patronage, and corruption, which are specifically pertinent to government contracting and enterprise development initiatives. However, the NEM
2
A number of new public universities and pre-university matriculation colleges were founded in the 1990s. Some Bumiputera-exclusive pre-university programs introduced a 10% non-Bumiputera quota in 2001–2002. However, university admissions – operating through a centralized system – clearly maintained Bumiputera preferences, increasingly through these pre-university programs.
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
829
superimposed that concern onto the whole system. The report seemed unaware that its proposals took the form of enhancing the need-based and merit-based selection among Bumiputeras – simply put, extending more assistance to low-income Bumiputeras instead of high-income Bumiputeras or favoring more capable Bumiputeras over less capable Bumiputeras – rather than removing the race-based, pro-Bumiputera regime altogether. Nonetheless, its ambiguity opened the window for misreading – that it was a plan to abolish Bumiputera privileges – which rattled segments of the political establishment and, in turn, unleashed a fierce backlash. The NEM withdrew in the face of political pressure but was fundamentally crippled by its incoherence and opacity. Notwithstanding the incoherencies, the past decade has seen some positive developments on the policy front. The emphasis on the bottom 40% (B40) households has inclined implementation toward distributing more benefits to low-income Bumiputera households where viable and to Bumiputera SMEs and dynamic Bumiputera-owned and Bumiputera-operated businesses. However, the requisite systematic rethink has not transpired (Lee 2022). Wave after wave of policy rebranding in recent decades has neglected the paramount challenge of Malaysia’s affirmative action: having immensely facilitated access and participation, the regime needs to effectively cultivate capability and competitiveness while systematically finding ways to transition away from the application of overt quotas and preferences.
Mechanisms and Instruments How has affirmative action operated in Malaysia? The full range and changes across time are too vast to be catalogued here. Nevertheless, to acknowledge the magnitude and variety of Malaysia’s affirmative action, and to appreciate the challenges of reforming the regime, it is pertinent to survey the major programs currently being implemented. Many of these have been in place for decades, while some, as noted in Table 1, have been introduced in the past decade or so. This section considers the policy regime sector by sector. In higher education, Bumiputera students enjoy special access to secondary-level residential colleges and pre-university matriculation programs, which were exclusively Bumiputera for some years until 90% quotas came into effect in 2001–2002. MARA-operated technical colleges and scholarships remain Bumiputera preserves, while public sector department scholarships have applied group quotas. University admissions likewise implemented quotas for many years, but since 2002, Bumiputera preference has operated through quotas at the pre-university stage, mainly in matriculation colleges. Yayasan Peneraju Pendidikan Bumiputera (YPPB), founded in 2012, offers financial aid for technical and professional training, targeting disadvantaged Bumiputera students, even more specifically for those growing up in difficult circumstances, such as having a single parent or disabled sibling. Malaysia’s efforts to promote Bumiputera upward occupational mobility have been less formalized. The public sector has not expressly declared a quota policy.
830
H.-A. Lee
Table 1 Major affirmative action programs in Malaysia, by policy sector Policy sector Higher education
High-level employment Enterprise development
Wealth and property ownership
Specific programs and key information • Residential colleges and pre-university matriculation colleges (90% Bumiputera quota) • Exclusively Bumiputera MARA technical institutions and university • Exclusively Bumiputera MARA education sponsorship • Racial quotas in public services department scholarships • Yayasan Peneraju Pendidikan Bumiputera (YPPB) scholarships for disadvantaged students for technical/professional training • De facto Bumiputera preference in the public sector and governmentlinked companies (GLCs) • Public procurement: 7 classes, from G1 smallest to G7 largest: • G1: reserved for 100% Bumiputera-owned companies; G2–G6: carve-outs and price preference for majority-Bumiputera companies •“Carve out and compete” (reserved contracts) in megaprojects • Teras selection of competitive and high-growth enterprises – to be given consideration in public procurement • GLC procurement, vendor development (GLC Transformation Program 2006–2015) • Loans and support for small and micro firms (MARA, Tekun Nasional), SMEs primarily in retail and distribution (PUNB); Bumiputera programs within SME Bank and SME Corp • INSKEN Entrepreneurship Institute, Bumiputera Facilitation Fund (reconfigured to Bumiputera Prosperity Fund), SUPERB grant for youth entrepreneurs • Private equity (Ekuinas) • Amanah Saham Bumiputera (unit trust) • Public listing equity requirements • Property purchase discounts
Source: Adapted from Lee (2021a)
Nonetheless, the NEP mandate for all occupations to reflect the population composition, buttressed by the constitutional authorization for reservations specifically in government employment, firmly laid the groundwork for de facto Bumiputera preference. Enterprise development, encapsulated in the BCIC and spanning a broad range of activities and agencies, is distinct from the other AA spheres in the ways it has traversed across different modes and varying mixes of state and market functions. The focal point of the BCIC agenda shifted, from a heavier reliance on state-owned enterprises and state-sponsored credit in the 1970s to an experiment with heavy industries – more directly under state ownership – in the early 1980s, followed by massive privatization that accelerated from the late 1980s and into the 1990s, as a gargantuan vehicle for Bumiputera capitalist development, particularly the promotion of lionized Malay “captains of industry.” As alluded to earlier, the privatization project collapsed in the aftermath of the 1997–1998 Asian financial crisis, prompting a massive renationalization exercise. The emergent entities, rebranded as GLCs, were commissioned to spearhead the BCIC agenda, directly through employing managers and indirectly through Bumiputera vendor and supplier development
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
831
programs. For micro and small businesses, parallel interventions have become embedded from the 1990s – notably the Tekun and PUNB (Perbadanan Usahawan Nasional Berhad or National Entrepreneurship Corporation Ltd) credit schemes – and SME financing for Bumiputera SMEs have also been in place from the early 2000s. Public procurement has since the 1970s played a major role in facilitating Bumiputera business opportunity. Bumiputera contractors enjoy exclusive access to the smallest class of contracts (out of seven tiers, scaled by company and project size) and, according to official policy, receive price handicaps in the larger contracts. The most recent measures, promulgated after 2010, have placed more emphasis on capability and performance, through modifying SME financing and procurement mechanisms. Saliently, the “carve out and compete” scheme has set aside megaproject contracts for Bumiputera players to be awarded through rigorous selection, accompanied by mentoring and capacity building. However, while many programs are in place, efforts at invigorating the system are more piecemeal than comprehensive. In the wealth ownership policy sector, Malaysia has predominantly focused on corporate equity. Among the expansive and enduring institutions are unit trust funds, most prominently Amanah Saham Bumiputera, which has attained very broad participation and leveraged on substantial funding and political clout to engage in some takeovers of British companies in the early 1980s. Bumiputera equity requirements, which reallocate existing shares by fiat, have flowed and ebbed. In the mid-1970s, new legislation compelled medium- and large-scale manufacturing establishments to relinquish 30% of equity to Bumiputera investors. This drastic move triggered fierce backlash; subsequent compromises exempted a larger segment of firms, and all export-oriented operations, from these equity rules. Over time, more of such regulations have been rolled back. Initial public offerings, formerly also subjected to the 30% Bumiputera quota, have since 2009 had to reserve 12.5% for Bumiputera investors. Private equity institution Ekuinas was established in conjunction with dismantling of Bumiputera equity requirements in various service industries, to serve the dual role of fostering Bumiputera ownership and entrepreneurship. Bumiputera households receive property purchase discounts for new housing developments, the terms of which vary by state.
Policy Achievements, Shortfalls, and Implications The outcomes of Malaysia’s affirmative action and New Economic Policy are as expansive as the programs in place. This section critically surveys the literature and then evaluates the most salient policy achievements and shortfalls. The analysis remains constantly mindful of policy implications, specifically the ways that AA in Malaysia, as the NEP passes its half century, can build on its progress and chart pathways toward its ultimate goal of being a transitory intervention. While the range of affirmative action programs is wide and varied, the empirical literature is somewhat narrow and repetitive. This is partly due to lack of data.
832
H.-A. Lee
Research on this subject seek statistics that are representative of the country and reported according to racial or ethnic categories and hence inevitably rely on official sources, such as national labor force surveys or household income surveys. These datasets are generally inaccessible in raw form; studies have largely settled for statistics disclosed in the 5-year Malaysia Plans or Department of Statistics’ summary reports. A limitation in the empirical literature, from the perspective of affirmative action, is the pattern in most studies of focusing on income inequality and extrapolating the findings onto affirmative action (Hashim 1998; Shari 2000; Jamaludin 2003; Jomo 2004; Chakravarty and Abdul-Hakim 2005; Leete 2007; Mat Zin 2008; Meerman 2008; Abdul Aziz 2012). A few studies have evaluated the immediate and direct outcomes of affirmative action (Khoo 2005; Lee 2005; Yusof 2012) or compiled data from sources besides the Malaysia Plans (Lee 2012, 2017). Unsurprisingly, the bulk of the literature, in observing the same data series, converges around a storyline resembling the official account. On the positive side, Malaysia has succeeded in narrowing inter-racial income gaps, increasing Bumiputera educational attainment, and boosting representation in professional and managerial positions. On the negative side, attention also concentrates on the same indicator: Malaysia’s failure to reach the 30% Bumiputera equity ownership target. Debate mainly revolves around the reasons for this failure, or the empirical record, which stems from alternative claims that the 30% threshold has been surpassed (CPPS 2006). Confining this analysis to the data reported in Malaysia’s planning documents reproduces the biases and limitations stemming from their inordinate emphases on equity ownership and extrapolation of inter-racial household income gaps – which derive from numerous factors, many of which are unrelated to Bumiputera-targeted programs – to make conclusions about AA. The literature has also inadequately accounted for differences between increased access and participation and capability and competitiveness, especially in education, employment, and enterprise. The NEP’s two prongs constitute a cogent starting point for empirical evaluation – with adherence to the caveats noted in the conceptual overview above that poverty reduction complements affirmative action and that household income is a broader policy objective dependent on many other factors beyond affirmative action. Malaysia’s sustained rural development, economic growth, industrialization, and educational expansion have alleviated poverty and laid vital groundwork for the country’s progress in affirmative action. The incidence of poverty steadily declined from 49.3% in 1970 to 8.7% in 1995 and 0.2% in 2019 (DOSM 2020a).3 Bumiputera household income sustained more rapid growth, thus narrowing disparities between groups. Chinese households on average received 130% times more income than 3 The long series of poverty estimates from 1970 refers to the absolute poverty headcount ratio – the share of households with per capita income below the poverty line – which hinges on the pre-determined level of poverty line income (PLI). Malaysia’s near-zero official poverty rate in recent years has been criticized as unrealistically low. A PLI revision exercise in 2019 raised the bar, which subsequently raised the 2019 poverty rate from 0.2% to 5.6% (DOSM 2020a). Regardless of the PLI level, though, Malaysia’s long-term record of poverty reduction will assuredly stand.
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
833
Bumiputera households in 1970–1976. This difference diminished to 75% in 1990–1995 and, most recently, 40% (2014–2019). The Indian-Bumiputera gap ratio similarly registered a downward trend, from 67% (1970–1976) to 38% (1990–1995) and 15% (2014–2019) (author’s calculations from DOSM 2020a). These outcomes are germane to the NEP and to inequality in general, but must be placed in proper context, to align the appraisal of affirmative action outcomes with the policies from which the outcomes derive. Inter-racial mean income ratios importantly reflect the general distribution of well-being across groups, but the relationship with affirmative action is not straightforward; accordingly, convergence of household income cannot be equated with efficacy of affirmative action, nor can it be taken as an immediate signal of readiness to relinquish preferential treatment. Downward trends in Bumiputera to non-Bumiputera income ratios mark progress in narrowing disparity in average welfare, but do not automatically prescribe AA policy implications. Indeed, if narrowing inter-racial income gaps demonstrate that beneficiaries are empowered and ready to phase out the preferential regime, then such reforms should already be underway. Furthermore, in view of the concentration of Chinese and Indian populations in urban areas, it is more credible to compute interracial income gaps among urban households. In the process, the disparity between Bumiputera and Chinese drops to 28%, while Indian average income is only 5% higher than Bumiputera (DOSM 2020a). These converging trends, to the point of near parity between urban Bumiputera and urban Indian incomes, have manifestly not fostered confidence and impetus to undertake reforms to Bumiputera preferential treatment. Political interest in maintaining privileged access, of course, constitutes a source of resistance to change. However, the socioeconomic conditions are equally if not more important. Arguably, the primary obstacle to change is the inadequate development of capability, competitiveness, and confidence in the Bumiputera community, such that the community is under-equipped to transition away from the preferential system. Affirmative action must be evaluated primarily with reference to its direct policy objectives and with a view to account for access and participation as well as capability and competitiveness. In the policy sector of higher education, Bumiputera attainment continuously advanced; the community’s representation in public university enrolment rose from 40% in 1970 to 67% in 1985 and 81–83% in 2005–2008 (Malaysia 1986; Mukherjee et al. 2017). The Bumiputera share of the workforce with higher education qualifications increased from 5.7% in 1990 to 34.3% in 2018 – at a faster rate than the other major groups. Simultaneously, the proportion of diploma or degree holders within the Chinese workforce increased from 5.6% to 29.9%. However, these quantitative gains have also been accompanied by concerns over the quality of education, with Bumiputera graduates on average lagging in academic achievement, employability, and workplace mobility (Lee 2021a). Malaysia’s higher education policies do facilitate inter-generational upward mobility, and this policy sector also provides greater scope for shifting away from racial quotas to preferential treatment based on socioeconomic disadvantage. In recent years, the government has increased quotas for B40 students in some Bumiputera residential secondary school programs. Studies have recorded
834
H.-A. Lee
substantial inter-generational upward mobility within Bumiputera households (Khalid 2018). Nevertheless, college and university admissions and financial support do not systematically incorporate socioeconomic disadvantage into selection processes. Efforts to allow non-Bumiputeras access to the numerous Bumiputeraexclusive institutions are exceedingly difficult, but the YPPB program, being founded to serve disadvantaged Bumiputeras, even more specifically those with proven family hardships, is a distinct candidate among programs that can shift away from being an exclusive Bumiputera domain toward opening to disadvantaged students of all backgrounds. In the high-level occupations targeted by affirmative action, Bumiputera representation has steadily risen over the course of the NEP, albeit with significant difference between professional and technical positions compared to management. This progress, based on employment statistics obtained in labor force surveys, was charted in Malaysia’s planning documents. From 1971 to 1990, the Bumiputera share of professionals and technicians rose from 47% to 61% and that of managerial positions grew from 24% to 30% (Malaysia 1985, Malaysia 1995). In 2019, Bumiputeras comprised 66% of professionals and 41% of managers (DOSM 2020b). As expected, government departments and government-linked companies, where affirmative action policies are in place, register disproportionately higher Bumiputera presence. The public sector is close to 90% Bumiputera. The “G20” largest GLCs’ overall workforce comprises 79% Bumiputera, 10% Chinese, 8% Indian, and 3% others; in executive positions, 73% are Bumiputera, 19% Chinese, 6% Indian, and 2% others (Lee 2017). Of course, Bumiputeras have also made inroads in the private sector, with their increasing acquisition of higher education qualifications – although quality differentials, and discriminatory treatment as revealed by field experimental research, have likely hindered their opportunity to advance (Lee and Khalid 2016). Fair employment legislation, to administer the ideal of equal opportunity specifically in Malaysia while also clarifying the scope of affirmative action in the public sector and diversity-promoting measures in the private sector, constitutes a potentially constructive initiative. In enterprise development, Bumiputera advancement has conspicuously fallen short, in line with the lesser presence in managerial positions. This policy sector poses some of the most acute challenges, due to lack of cumulative experience, uncertainty of success, and long gestation periods. Nonetheless, the deficits still stand out. Despite phases of different interventions, involving public and private enterprise and allocation of ownership, contracts, and credit, Malaysia has not made a breakthrough in terms of cultivating sizable, dynamic, and competitive Bumiputera enterprise. Privatization of erstwhile state-owned enterprises, the most voluminous program of the BCIC agenda, is also perhaps the most significant failure. While the AFC triggered the collapse of privatized entities and renationalization of corporations hitherto helmed by Malay captains of industry, the selection and monitoring processes were undermined by patronage, profiteering, and weak governance (Tan 2008). Renationalized entities were subsequently rebranded as government-linked companies and commissioned to spearhead Bumiputera enterprise. GLCs undertook a
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
835
10-year GLC Transformation Programme (2006–2015) and continually drive Bumiputera managerial and commercial development through direct and demonstrative roles of employment and executive leadership and indirectly through vendor development. The G20 largest GLCs have performed steadily on commercial and financial fronts, but Bumiputera vendor development faces continual challenges and shortfalls (PCG 2015). GLC workforce composition is more diverse than the public sector, as noted earlier, but CEO positions remain a Bumiputera preserve. GLCs also contribute to Bumiputera, specifically Malay, representation on the boards of Malaysia’s largest companies. Among the 837 directorships of the 100 top publicly listed companies at end 2019, 42% were Chinese, 41% Malay, and 5% Indian. A substantial proportion of Malay directors are retired bureaucrats, and a preponderance of seniors in their ranks suggests the presence of a Malay “old guard” holding non-executive board positions (Lee 2021a). Among micro, small, and medium enterprises (MSMEs), which account for 39% of Malaysia’s GDP and 48% of employment, Bumiputera participation continually lags. In 2005, Bumiputera-owned MSMEs consisted of 83.5% micro, 15.1% small, and 1.4% medium; non-Bumiputera MSMEs were 69.9% micro, 25.9% small, and 4.3% medium (Lee 2021b). Likewise, in public procurement, three quarters of Bumiputera contractors are in the smallest G1 class (out of seven), and the overwhelming majority remain there – largely due to reservation of 100% of G1 contracts for Bumiputeras and a policy that generally disincentivizes efforts to scale up. Procurement programs by government and GLCs, alongside numerous programs to cultivate SMEs and managerial capacity – through loans, grants, and private equity – have yielded mixed results that, on the whole, fall short of the needed breakthrough. It is paramount that affirmative action in this policy sector operates effectively and equitably, toward cultivating dynamic and actively owned Bumiputera enterprise, with attention to growth, innovation, and competitiveness. On the ownership front, the record is also checkered – and the data controversial. Unit trust giant Amanah Saham Bumiputera’s widespread participation, with 8.6 million individual accounts in 2014, has contributed to Bumiputera ownership, albeit with the vast majority holding a small number of units (Lee and Khalid 2020). By official estimates, the Bumiputera portion of total equity, hovering around 20% since 1990, has persistently fallen short of the targeted 30%. This is repeatedly referenced to justify the continuation of the entire regime of Bumiputera preferences. However, the presence of nominees, accounting for 8% of total holdings in 2015 (Ministry of Economic Affairs 2019), masks the identity of a considerable proportion of equity owners, while foreign ownership, which has increased its share of the total in the past decade and hence depressed the Bumiputera share, may paint misleading pictures of declining Bumiputera ownership. These discrepancies warrant a thorough revision of the measurement method. However, the more fundamental problem is that persistent fixation with Bumiputera equity holdings misaligns policy priorities and targets and detracts from the paramount need to cultivate enterprise. There is significant follow-through from the government’s tacit acknowledgment of the need to get back to some basics rather than repeat previous rather grandiose schemes, in that the current policy
836
H.-A. Lee
framework places more emphasis on Bumiputera-controlled SMEs and their expansion and technological advancement (Malaysia 2021). These objectives must decidedly take precedence over wealth acquisition, especially when narrowly construed as equity ownership.
Conclusion Malaysia remains a widely cited case of affirmative action, for its immense magnitude and long experience, coupled with sustained achievements under the NEP. Where should Malaysia go from here, and what can other countries learn from Malaysia’s experience? The appraisal above has hopefully provided systematic, critical, and nuanced perspectives on the country’s achievements and shortfalls. This chapter concludes with three broad lessons. First, country context profoundly matters. Malaysia’s socioeconomic disparities demanded more extensive intervention than is necessary in most countries, but its constitutional provisions for Bumiputera preference and political imperatives also created conditions for more aggressive and exclusivist forms of action. Malaysia’s narrowing of inter-racial inequalities and promotion of Bumiputera socioeconomic advancement, which may stand out positively in comparative study, have been attained through overt preferential treatment and a wide range of racially exclusive programs. Malaysia can benefit from a greater self-awareness of the potential and peril of its particular modes of affirmative action; other countries must view the operational aspects of Malaysia’s policy with circumspection. Second, it is imperative to apply a systematic framework that (1) clarifies affirmative action’s principal objectives of increasing a disadvantaged group’s representation in specific areas, the extent of which should correspond with each country’s historical and present context, and (2) recognizes the supplementary effects of economic growth, employment generation, socioeconomic development, and poverty reduction. Malaysia’s NEP laid out a judicious template that distinguished poverty eradication regardless of race from social restructuring that involved Bumiputera targeting. However, these distinct policy mainstays have become conflated, such that assistance for the poor is repeatedly invoked as a systemic alternative to affirmative action, instead of a complement and reinforcement. The scope and interaction of race-based preferential treatment and class- or need-based preferential treatment also warrant clarity and rigor. In some, but not all, affirmative action sectors, socioeconomic disadvantage can be increasingly applied to foster more equitable distribution within the beneficiary group or even to increasingly operate in place of race-based preferences. Such options are much more applicable to higher education and microfinance, while in certain policy sectors, notably employment and enterprise development, selecting Bumiputeras based on ability and potential, and limiting repeat reception of preferential treatment, presents more productive and constructive paths forward. Third, the design, conduct, and analysis of affirmative action must be mindful of specific structures and constraints of the major policy sectors – higher education,
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
837
high-level occupations, enterprise, and ownership – and of the further policy goal of broadly cultivating capability, competitiveness, and confidence. This lesson also derives more from what Malaysia has not done; the country did not start out with a systematic conception of the core problems, objectives, and instruments of AA and hence has significantly lacked direction and coherence. While various policies since the NEP have promoted opportunity and participation over five decades, the country has fallen short in the paramount goal of capability development. Affirmative action in Malaysia, driven by constitutional foundations, political imperatives, and socioeconomic conditions, passes the 50th anniversary since the policy momentously expanded under the New Economic Policy. The NEP has significantly remedied racial disparities, particularly by increasing Bumiputera access, participation, and mobility, in higher education, high-level occupations, enterprise, and ownership. Malaysia must make progress such that Bumiputeras can undertake appropriate change and rollback of quotas and overt preferences, because the community has become sufficiently empowered, and such interventions are no longer necessary. This undoubtedly lofty ambition also aligns with the constitutional provisions for affirmative action, which are conditional on the necessity for such extraordinary measures. The imperative of fostering equitable group representation, especially in public universities, government departments, and decision-making positions, will remain, but Malaysia can pursue such objectives through less divisive means besides racial quotas and overt preferences. Moving in these directions, however, requires a systematic rethink and a rekindling of the NEP’s original aspiration for the Bumiputeras to be full partners in the economic life of the nation.
References Abdul Aziz R (2012) New economic policy and the Malaysian multiethnic middle class. Asian Ethnicity 13(1):29–46 Amoroso DJ (2014) Traditionalism and the ascendancy of the Malay ruling class in Malaya. NUS Press and Strategic Information and Research Development Centre, Singapore Andaya BW, Andaya LY (2001) A history of Malaysia. University of Hawai’i Press, Honolulu Chakravarty SP, Abdul-Hakim R (2005) Ethnic nationalism and income distribution in Malaysia. Eur J Dev Res 17(2):270–288 Chin YW, Teh BCG (2017) Malaysia’s protracted affirmative action policy and the evolution of the Bumiputera commercial and industrial community. SOJOURN J Soc Issues Southeast Asia 32(2):336–373 CPPS (2006) Corporate Equity Distribution: Past Trends and Future Policy. Kuala Lumpur, Centre for Public Policy Studies DOSM (2020a) Household income and basic amenities survey report 2019. Putrajaya, Department of Statistics Malaysia DOSM (2020b) Labour force survey report 2019. Putrajaya, Department of Statistics Malaysia Faaland J, Parkinson JR, Saniman R (1990) Growth and ethnic inequality: Malaysia’s new economic policy. Hurst, London Fernando JM (2015) Special rights in the Malaysian constitution and the framers’ dilemma, 1956–57. J Imperial Commonw Hist 43(3):535–556
838
H.-A. Lee
Gomez ET (2015) The 11th Malaysia plan: covertly persisting with market-friendly affirmative action? Round Table 104(4):511–513 Gomez ET, Jomo KS (1999) Malaysia’s political economy: power, profits, patronage. Cambridge University Press, Cambridge Gomez ET, with Padmanabhan T, Kamaruddin N, Bhalla S, Fisal F. 2017. Minister of finance incorporated: ownership and control of corporate Malaysia. Singapore, Palgrave Macmillan Hashim SM (1998) Income inequality and poverty in Malaysia. Lanham, Rowman and Littlefield Hirschman C (1986) The making of race in colonial Malaya: political economy and racial ideology. Sociol Forum 1(2):330–359 Jamaludin F (2003) Malaysia’s new economic policy: has it been a success? In: Darity W, Deshpande A (eds) Boundaries of clan and color: transnational comparisons of inter-group disparity. Routledge, London and New York, pp 152–174 Jomo KS (2004) The new economic policy and interethnic relations in Malaysia. Identities, conflict and cohesion programme paper no. 7, UNRISD, Geneva Khalid MA (2018) Climbing the ladder: socioeconomic mobility in Malaysia. Asian Econ Papers 17(3):1–23 Khoo BT (2005) Ethnic structure, inequality and governance in the public sector: Malaysian experiences. Democracy, governance and human rights programme paper 20. Geneva, UNRISD Lee HG (2005) Affirmative action in Malaysia. Southeast Asian Affairs 2005:211–228 Lee H-A (2012) Affirmative action in Malaysia: education and employment outcomes since the 1990s. J Contemp Asia 42(2):230–254 Lee H-A (2016) Affirmative action regime formation in Malaysia and South Africa. J Asian Afr Stud 51(5):511–527 Lee H-A (2017) Malaysia’s Bumiputera preferential regime and transformation agenda: modified programmes, unchanged system. Trends in Southeast Asia 2017 No. 22. Singapore, ISEAS Lee H-A (2021a) Affirmative action in Malaysia and South Africa: preference for parity. Routledge, London and New York Lee H-A (2021b) The new economic policy beyond fifty: assessing its strengths and weaknesses to chart a cohesive Malaysian society. IDEAS policy paper 73, November 2021. Kuala Lumpur, IDEAS Lee H-A (2022) Malaysia’s new economic policy: fifty years of polarization and impasse. Southeast Asian Stud 11(2) page numbers will be available by end August 2022 Lee H-A, Khalid MA (2016) Discrimination of high degrees: race and graduate hiring in Malaysia. J Asia Pac Econ 21(2):53–76 Lee H-A, Khalid MA (2020) Is inequality really declining in Malaysia? J Contemp Asia 50(1): 14–35 Leete R (2007) From Kampung to twin towers: 50 years of economic and social development. Petaling Jaya, Oxford Fajar Malaysia (1971) The second Malaysia plan, 1971–75. Kuala Lumpur, Government of Malaysia Malaysia (1976) The third Malaysia plan, 1976–80. Kuala Lumpur, Government of Malaysia Malaysia (1986) The fifth Malaysia plan, 1986–90. Kuala Lumpur, Government of Malaysia Malaysia (2010) Federal Constitution. Kuala Lumpur, Commissioner of Law Revision, Malaysia Malaysia (2021) The twelfth Malaysia plan, 2021–25. Putrajaya, Government of Malaysia Mat Zin R (2008) Income inequality in Malaysia. Asian Econ Policy Rev 3(1):114–132 Means G (1990) Malaysia in 1989: forging a plan for the future. Southeast Asian Affairs 1990: 183–203 Meerman J (2008) The Malaysian success story, the public sector, and inter-ethnic inequality. In: Nelson JM, Meerman J, Embong AR (eds) Globalization and national autonomy: the experience of Malaysia. Singapore, ISEAS, pp 76–115 Ministry of Economic Affairs (2019) Shared prosperity vision 2030. Ministry of Economic Affairs, Putrajaya Mohamad M (1991) Malaysia: the way forward. Working paper presented to the Malaysian Business Council, February 28, 1991
36
Malaysia’s New Economic Policy and Affirmative Action: A Remedy in Need. . .
839
Mukherjee H, Singh JS, Fernandez-Chung RM, Marimuthu T (2017) Access and equity issues in Malaysian higher education. In: Malakolunthu S, Nagappan CR (eds) Policy discourses in Malaysian education: a nation in the making. Routledge, London, pp 45–70 NEAC (2010) New economic model – Part 1. Putrajaya, National Economic Advisory Council Ooi KB (2013) The new economic policy and the centralisation of power. In: Gomez ET, Saravanamuttu J (eds) The new economic policy in Malaysia: affirmative action, ethnic inequalities and social justice. Singapore and Petaling Jaya, NUS Press, ISEAS and SIRD, pp 317–334 PCG (2015) GLC transformation graduation report. Putrajaya, Putrajaya Committee on GLC High Performance (PCG) Sabbagh D (2012) Affirmative action. In: Rosenfeld M, Sajó A (eds) The Oxford handbook of comparative constitutional law. Oxford University Press, Oxford, pp 1124–1141 Selvaratnam V (1988) Ethnicity, inequality and higher education in Malaysia. Comp Educ Rev 32(2):173–196 Shari I (2000) Economic growth and income inequality in Malaysia, 1971–95. J Asia-Pacific Econ 5(1/2):112–124 Tan J (2008) Privatization in Malaysia: regulation, rent-seeking and policy failure. Routledge, London and New York Tan YS, Santhiram R (2017) Race-based policies and practices in Malaysia’s education system. In: Samuel M, Tee MY, Symaco L (eds) Education in Malaysia: developments and challenges. Springer, Singapore, pp 17–32 Yusof ZA (2012) Affirmative action in Malaysia: an overview of progress and limitations. In: Brown G, Stewart F, Langer A (eds) Affirmative action in plural societies: international experiences. Palgrave Macmillan, Basingstoke, pp 128–150
Stereotype Threat Experiences Across Social Groups
37
Insights for Understanding Affirmative Action Valerie Jones Taylor, C. Finn Siepser, Juan Jose´ Valladares, and Rita Knasel
Contents Discrimination and Affirmative Action: Definitions and Policy in the United States . . . . . . . . . Through the Lens of Stereotype Threat: Workplace Experiences of Discrimination and Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereotype Threat Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Foundational Stereotype Threat Theory Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consequences of Stereotype Threat Beyond Academic Performance . . . . . . . . . . . . . . . . . . . . . . Social Groups, Stereotype Threat, and Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereotype Threat and Affirmative Action Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Insights from Stereotype Threat Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Insight One: Misconceptions and Mischaracterizations of AA Cues Stereotype Threat . . . Insight Two: Creating Identity-Safe Environments in the Context of Affirmative Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gaining Critical Historical Knowledge About Discrimination and Social Inequities . . . . . . Accurate and Strategic Messaging of AAPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Renewing Commitments to AAPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Concluding Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
843 846 848 849 850 851 853 854 855 856 856 857 859 860 861 861
Abstract
This chapter argues that stereotype threat theory (Steele et al., 2002) can demystify the relationship between experiences of discrimination, identity (e.g., race/ ethnicity, gender), and perceptions of affirmative action (AA) and related policy (AAP). Following the failure of anti-discrimination laws in the US, the federal government instituted affirmative action policies (1964–1967), or “proactive approaches” designed to redress centuries of de jure and de facto discrimination toward historically marginalized groups. Since their inception, AAPs in education and employment have become increasingly misunderstood and controversial, V. J. Taylor (*) · C. F. Siepser · J. J. Valladares · R. Knasel Lehigh University, Bethlehem, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_6
841
842
V. J. Taylor et al.
while debates about their fairness, efficacy, and implementation abound. Stereotype threat theory can inform divergent understandings of AA and AAPs by delineating a framework that explains how and why concerns about being judged through the lens of negative group-based stereotypes can impact cognition and behavior for historically marginalized and privileged groups (i.e., assumed affirmative action beneficiaries and non-beneficiaries). After defining discrimination and affirmative action in the US historical context, this chapter illustrates the various stereotype threat-related concerns AAPs may raise for people from different social groups who have disparate experiences with discrimination. Next, findings from stereotype threat research and its relationship to affirmative action perceptions are reviewed to demonstrate that properly implemented and effectively communicated AAPs are critical in reducing the threat associated with them. Integrating insights from stereotype threat theory, the chapter concludes by advocating for (a) education on critical historical knowledge about discrimination and AAPs, (b) accurate and strategic AAP messaging, and (c) renewed commitments to instituting evidence-based AAPs to reduce systemic group-based discrimination. Keywords
Stereotype threat · Affirmative action · Affirmative action policy · Discrimination · Identity-safe environments
I champion sensibly designed racial affirmative action, not because I have benefited from it personally – though I have. I support it because, on balance, it is conducive to the public good. (Randall Kennedy)
While some people agree with esteemed law professor Randall Kennedy that affirmative action is “conducive to the public good,” others regard it as divisive, unfair, and counterproductive. This range of views is connected to a general lack of awareness about the relationship between discrimination and affirmative action. Furthermore, it is associated with misunderstandings and mischaracterizations of the intent, content, and implementation of affirmative action policy in academic and employment settings (Crosby et al. 2006). This chapter argues that stereotype threat theory – a framework that centers the concerns and experiences of members of social groups who are negatively stereotyped – provides a valuable lens to examine these processes and their impact on academic and employment experiences. Stereotype threat is the concern or worry that one will be perceived or judged through the lens of negative group-based stereotypes which, under specific conditions, can undermine outcomes in academic and other important life domains (Steele 1997). The broader theory outlines situational triggers, processes, and boundary conditions of the stereotype threat phenomenon and applications to various populations, contexts, and outcomes (Inzlicht and Schmader 2012; Murphy et al. 2022; Steele et al. 2002). This chapter argues that stereotype threat theory can
37
Stereotype Threat Experiences Across Social Groups
843
demystify the relationship between experiences of discrimination, identity (e.g., race/ethnicity, gender), and perceptions of affirmative action policy. Likewise, it offers theoretically driven insights for actions to enable identity-safety among diverse groups surrounding these issues. Specifically, this theory provides critical insights for constructing identity-safe environments – spaces of respect, trust, and belonging among different groups, which stand to mitigate the misunderstanding of and potential backlash to affirmative action policies. Identity-safe environments communicate that an individual’s social identities – particularly those historically and systemically devalued and mistreated – are valued and treated equitably (Hall et al. 2018). Insights into the experience of stereotype threat and the value of identity-safe environments thereby afford a more accurate and equitable understanding of affirmative action policy and its relationship to race- and gender-based discrimination in particular. This chapter centers social identities and provides an evidence-based lens through which people may understand discrimination and affirmative action. Specifically, it highlights how social group memberships and the negative stereotypes associated with them impact discrimination and affirmative action in academia and workplaces. This chapter begins with brief definitions of discrimination and affirmative action in the USA, highlighting how social identities influence perceptions and views. Then, an example is provided, detailing how these factors might function differently for a Black woman and a White man, who also have disparate experiences of stereotype threat. Next, findings from the stereotype threat literature are reviewed and integrated with research on minoritized groups’ experiences with affirmative action. From this foundation, the next section describes how stereotype threat theory provides a useful framework to explain divergent perceptions of affirmative action. The chapter ends by describing how the complex relationships between stereotype threat, discrimination, and affirmative action can inform the creation of identity-safe environments by considering differing perceptions and experiences of affirmative action among members of different social groups (Leslie et al. 2014; Van Laar et al. 2008). With these insights, this work suggests that stereotype threat theory can shape one’s understanding and support of affirmative action policies that stand to be “conducive to the public good.”
Discrimination and Affirmative Action: Definitions and Policy in the United States To ground this discussion of affirmative action and stereotype threat, this section provides definitions of discrimination from a social psychological lens and then broadly describes US affirmative action, and associated laws and policies (hereafter referred to as AA and AAPs). The term discrimination describes the differential treatment of individuals and groups, and unjustified negative behaviors directed toward people or groups, due to their social group membership. Overt discrimination is the blatant and often hostile mistreatment (e.g., behaviors, judgments, decision-making) of people or groups based on their social group membership
844
V. J. Taylor et al.
(Jones et al. 2016). Subtle discrimination, equally as psychologically damaging as overt forms (and arguably more damaging), includes ambiguous or negative treatment that can be enacted consciously or unconsciously toward people or groups and often conveys ambiguous intent (Jones et al. 2016). In both form and across time, discrimination has significantly impeded historically minoritized groups’ access to a variety of educational and employment opportunities; it has also contributed to persistent disparities in education, incarceration, health, and wealth (Alexander 2010; Rothstein 2015; Trawalter et al. 2020). AAPs have the potential to counter overt and subtle discrimination as well as centuries of systemically racist and sexist policies. Originally designed as a proactive approach to redressing discrimination, AAPs provide opportunities for historically minoritized groups impacted by discriminatory practices (e.g., Crosby et al. 2003; Crosby and Cordova 1996). The US federal government more widely instituted these policies (1964–1967) when anti-discrimination laws were deemed ineffective at correcting historical patterns of racial and gender discrimination (e.g., low-wage job stagnation; exclusion from academic institutions, jobs, and entire fields). Since their formal adoption, AAPs have become increasingly misunderstood and controversial (Fryer and Loury 2005; Plous 2003). The few aspects of AAPs that remain in place today are frequently misunderstood and mischaracterized, largely due to unclear conceptualizations about the laws, scope, and intent of AA (Fryer and Loury 2005; Plous 2003). For example, the common ill-founded assumption of a quota system often fuels debates about AA’s current relevance and exacerbates narratives about supposed unfairness (e.g., Unzueta et al. 2008). Current legal AAPs in the USA do not and cannot include quotas (see University of California Regents v. Bakke 1978). Additionally, people are often unaware that there is a spectrum of AA preferential selection procedures. Mischaracterizations of preferential selection procedures assume that people from underrepresented groups will be admitted or hired despite being deemed “unqualified” (e.g., not having the requisite skills) – a practice that is illegal (Bureau of National Affairs 1979). Finally, many legal battles have ensued over the complexity of using race in college admissions and many decisions have determined that colleges and universities cannot use race as a primary preferential factor in the admissions selection process, either by law or in practice (see Gratz v. Bollinger 2003; Schuette v. Coalition 2014). Moreover, since the earliest legal challenges to AAPs, conservative narratives have erroneously characterized AA as a “theft” by Black and Latinx students against White and also Asian students (Garces and Poon 2018). Given the nuances and general lack of education about AAPs, most Americans are unaware of the characteristics of these policies and their implementation (Crosby et al. 2003). Today, practices resulting from AAPs and laws vary widely and may be voluntary or required, though there remains little consensus on the core definition of AA (Crosby et al. 2006; Crosby and Cordova 1996). The original goals of AA legislation were to monitor and evaluate whether recruitment and hiring practices may be unintentionally discriminatory and subsequently eliminate those practices (Crosby et al. 2003; Crosby and Cordova 1996). For nearly sixty years, AAPs have primarily
37
Stereotype Threat Experiences Across Social Groups
845
focused on discrimination in government contract hiring, hiring policies at large corporations, and admissions policies at highly selective colleges and universities (Crosby and Cordova 1996). AAPs have also included funding for businesses run by individuals with marginalized identities and employee support programs (Crosby et al. 2006) – proactive policies designed to undermine corporate redlining, discriminatory loan practices, and disenfranchisement to name a few. Additionally, administrations may monitor the overall demographic makeup of their organization and make a “good faith” effort to advertise to and recruit and hire qualified individuals from groups that have experienced discrimination. They may also incorporate some specific and clearly delineated form/s of preferential selection procedures (Bureau of National Affairs 1979). In higher education, three US supreme court cases between 1978 and 2013 have upheld aspects of AAP (Fisher v. University of Texas at Austin 2013; Gratz v. Bollinger 2003; University of California Regents v. Bakke 1978). Collectively, these cases ruled that legal AAPs in higher education can only be introduced or preserved if they aim to increase or maintain “diversity,” though the future of such policies is uncertain (Students for Fair Admissions, Inc. v. University of North Carolina, et al. 2019; Students for Fair Admissions, Inc. v. President & Fellows of Harvard College 2020). Experiences with discrimination and views on AA vary greatly between individuals and social groups (e.g., Kaiser et al. 2021; Slaughter et al. 2002). In addition to coping with ongoing discrimination, historically minoritized groups have faced centuries of legally sanctioned and de facto exclusion across all aspects of life and resulting disparities. Such experiences are not shared by members of historically privileged groups (Alexander 2010; Rothstein 2015). These divergent experiences with discrimination can partially explain minoritized and privileged groups’ differing views on AA (e.g., Slaughter et al. 2002). Further, scholars have noted two foundational American myths – meritocracy and individualism – that obscure how inequities function and thus impact understanding of AA (Liu 2011). Meritocracy is the belief that advancement and success are solely dependent on individual ability, effort, and hard work. Individualism is the belief that individuals are self-reliant and independent. Opponents of AAP contend that such policies violate these principles: AAPs are misperceived to reward lack of effort and discount hard work and fail to hold people accountable for their failures and successes (e.g., Kemmelmeier 2003). Relatedly, many individual difference variables, including but not limited to prejudice, conservatism, and experiencing self or group threat, also correlate with these ideals and opposition toward AAPs (e.g., Renfro et al. 2006). Given this backdrop, minoritized and historically privileged groups’ divergent experiences with discrimination and negative stereotyping are argued to be critical factors that influence perceptions of and experiences with AA. AAPs and individuals associated with them are frequently negatively stereotyped (e.g., ineffective policies; incompetent people) and stigmatized (e.g., unfair policies; undeserving people) (e.g., Leslie et al. 2014). Likewise, when perceptions of racial progress increase, majority group members perceive a loss of individual or group status or resources (see also Wilkins et al. 2015). These perceived losses are reflected in widespread and inaccurate claims of “reverse discrimination” or “reverse racism” (Phillips and Lowery
846
V. J. Taylor et al.
2018; Pincus 2000). Although, individual prejudice can exist toward those with privileged identities but discrimination, and racial and gender discrimination in particular, is an empirically-documented historical and ongoing pattern of systemic and institutionally supported marginalization. Accordingly, such discrimination has not and does not exist in “reverse.” Since these realities of discrimination, stereotyping, and their misunderstanding impact everyday experiences, these ideas will now be explored in a hypothetical engineering workplace scenario with a Black woman and a White man. This example highlights how empirically-documented experiences and stereotype threat processes (Steele et al. 2002) can contribute to divergent perceptions of and experiences with AA and obscure the benefits.
Through the Lens of Stereotype Threat: Workplace Experiences of Discrimination and Affirmative Action The presence of unclear, misunderstood, or mischaracterized AAPs in academic and workplace contexts not only interacts with experiences of discrimination, but can ironically trigger stereotype threat for historically minoritized and privileged group members. Consider the experience of a Black woman in science, technology, engineering, and math (STEM) who faces interpersonal and institutional instances of discrimination, which coincide with or cause stereotype threat. She has worked diligently in school toward her goal of becoming an engineer. Nonetheless, negative racial and gender stereotypes and discrimination routinely permeated her K-12 classroom experiences, perpetuating stereotypes that undermine Black students’ value and competence (e.g., La Salle et al. 2020; Rothstein 2015; Steele 1997). During college and graduate school, she encountered only a few other Black women, leading her to feel pressure to positively represent Black women (Sekaquaptewa and Thompson 2002). As high-profile court cases and debates about AAPs filtered into the classroom, she became concerned that others may negatively stereotype her as unintelligent and an unqualified AA beneficiary (Leslie et al. 2014; Van Laar et al. 2008). Despite working diligently, professors and peers believed her to be less capable of grasping challenging math concepts and questioned her belonging and fit in the field (e.g., LaCosse et al. 2020). As a Black woman in STEM, such experiences stand to undermine her sense of belonging, persistence, performance, and career interests (e.g., Ong et al. 2011). Additional barriers exist for this Black woman as she transitions from graduate school to working at a large engineering firm. As Black women comprise only 0.62% of the engineering workforce (National Center for Science and Engineering Statistics, NCSES 2019), being one of the few at her firm can feel isolating (Ong et al. 2011). She is partnered with a White man on a high-profile project and when she asks about unexplained workplace specific procedures, the White man expresses concern about her seeming lack of knowledge. She wonders how to attribute his concern: Does he perceive her to be unintelligent or unqualified? Are these procedures things she should know or things to be learned on the job from others? Is she facing racial discrimination and/or shifting standards? Such attributional ambiguity
37
Stereotype Threat Experiences Across Social Groups
847
is taxing (Major et al. 2002) and women and members of racially minoritized groups are judged by shifting standards of competence (Biernat et al. 2009). When interactions with her partner become challenging, she worries that her behavior will confirm negative stereotypes about Black people and women as lacking strong math skills (Richeson and Shelton 2012). At work, her ideas and contributions are often attributed to majority group members, her work is scrutinized more heavily than her peers, and she is passed over for deserved promotions (Ong et al. 2011). Moreover, the company’s key performance indicators (KPIs) on diversity and inclusion are routinely sidelined for other company objectives, suggesting to her that such goals are not a priority. Occasionally, colleagues mention increased hiring of “diverse” employees and connecting it with how their White peers somehow “didn’t make the cut.” This subtle negative reference to the company’s AAPs, especially when such policies are unspecified or not made public, erroneously suggests that people of color were hired because of their race and not their qualifications. Likewise, she wonders whether her colleagues assume she is an AA recipient and therefore question her abilities (Leslie et al. 2014). These compounding experiences are common and well-documented across multiple literatures, highlighting the psychological and interpersonal challenges incurred beyond the daily pressures and stress of an employee’s work. Moments like these throughout this woman’s educational and work experiences trigger stereotype threat through negative situational cues – i.e., characteristics of a setting signaling that one’s social group may be negatively stereotyped. At times, her preoccupation and frustration with these experiences impact her job satisfaction and retention and mental and physical well-being (e.g., Aronson et al. 2013). Now, consider the experience of a White man at the same engineering firm. He also has worked diligently in school and at work. However, his perceptions of and experiences with discrimination and stereotype threat are quite different. Though his educational experiences were challenging, he found supportive professors, comfortably joined study groups with classmates, and never wondered if engineering was a good field for a White man (e.g., Li et al. 2017). Given the demographic make-up of the engineering field, he is likely to have several connections in this industry. His classmate’s father had connections at a company where he secured college summer internships. If one of the few students of color made a mistake during his internships, unfounded rumors surfaced about them being unqualified AA beneficiaries (cf. Biernat et al. 2009). When the other White interns joked about the supposed AA recipients, he was uncomfortable but remained silent to “keep the peace.” Although he is concerned about being seen as racist by others – particularly the interns of color – he avoids more substantive interactions with the interns of color (Taylor et al. 2022). These experiences, however, do not impact his performance or sense of belonging in the field of engineering. Transitioning to a large engineering firm as an employee, as a White man, he remains in the numeric majority at the firm where many people share his identities and past experiences (NCSES 2019). He has been partnered with the previouslymentioned Black woman on the high-profile project. He notices that she does not
848
V. J. Taylor et al.
take the same shortcuts and is unaware of particular strategies he has learned during impromptu dinners with senior leaders (e.g., Li et al. 2017). He wonders if she is capable, believing the negative stereotypes about her engineering ability (e.g., incompetence), rather than realizing she does not have the extra advantage of learning inside information that expedites the work. When interactions with her become challenging, he worries that his behavior will confirm negative stereotypes about White people as prejudiced, or worse, racist (Goff et al. 2008; Richeson and Shelton 2012). Though he and his partner are excelling on their project, he feels stifled by the diversity and inclusion KPI requirement on all projects and is resentful when his White college friend is not hired. He is concerned about “reverse discrimination” by being passed over for a promotion in favor of someone more “diverse” (Phillips and Lowery 2018). Consequently, given this common but misinformed understanding of AAPs, he wonders if his Black partner was hired because of optics or to fill a “quota.” He assumes anyone hired this way must not have the requisite skills or be as intelligent and is hesitant to work with her on future projects. Overall, in this engineering work context where his identities are privileged, these perceptions and experiences do not negatively impact his job performance or mental and physical well-being in the ways that it impacts the Black woman (e.g., Aronson et al. 2013; Hall et al. 2015). These examples highlight how negative stereotyping and the presence or absence of racial and gender discrimination can impact cognition, motivation, and behavior in strikingly different ways for people with different identities and experiences. Likewise, stereotyping and discrimination can impact experiences and perceptions of poorly understood, mischaracterized, and/or implemented AAPs. Against this backdrop, the Black woman was concerned about assumptions and negative connotations that she might be an AA beneficiary, which in turn causes adverse outcomes (Leslie et al. 2014; Van Laar et al. 2008). The White man was threatened by diversity and inclusion KPI requirements and was concerned that he and his peers would experience “reverse discrimination” (Pincus 2000). Thus, in this context, the presence of unclear, misunderstood, or mischaracterized AAPs are likely perceived as negative cues for members of historically minoritized and privileged groups. Stereotype threat theory provides a lens through which these processes might better understood.
Stereotype Threat Theory People understand everything in their social environment through the lenses of their social identities (Tajfel and Turner 1986). Moreover, individuals are aware of how others perceive their social identities, the stereotypes associated with them, and the potential impacts of these perceptions and stereotypes (e.g., Vorauer et al. 1998). Individuals also bring their social identities into different environments, and aspects of those environments may signal back to them the value and meaning of their identities in those contexts. Likewise, awareness of cultural stereotypes about one’s group in these spaces can set the stage for the interpretation of potentially ambiguous
37
Stereotype Threat Experiences Across Social Groups
849
environmental and social cues (i.e., situational cues; Steele and Aronson 1995). Individuals are likely to experience stereotype threat when they are aware of the negative cultural stereotypes and situational cues that signal the value and meaning of their identities in a specific setting (Steele 1997). Stereotype threat is the concern or worry that one will be treated, judged, or evaluated through the lens of negative group stereotypes; and further, that one’s behavior might confirm such negative stereotypes (Steele 1997; Steele et al. 2002). Nearly three decades of empirical research indicates that stereotype threat can impact a wide range of cognitions, motivations, and behaviors in academic and organizational settings (Casad and Bryant 2016; Inzlicht and Schmader 2012; Murphy et al. 2022; Schmader et al. 2008). It follows that stereotype threat can affect people’s experiences with and perceptions of AA (Brown et al. 2000; Leslie et al. 2014; Van Laar et al. 2008).
Foundational Stereotype Threat Theory Research Stereotype threat theory was among the first to offer a psychosocial explanation for group-based underperformance that does not cite supposedly innate biological or cultural group differences as the basis for such underperformance. Biological explanations point to genetic differences among racial and gender groups and cultural explanations point to divergent socialization and cultural factors (Herrnstein and Murray 1994; Jencks and Phillips 1998). In contrast, stereotype threat focuses on the negative stereotypes that emerge in particular domains. Racism, sexism, discrimination, and other systemic inequities generate and reinforce stereotypes and these stereotypes could themselves impinge performance (Steele 1997; Steele et al. 2002). Although other factors also might undermine performance, negative stereotypes were an important and vastly understudied factor at the time. Classic stereotype threat studies examined whether test diagnosticity and reminders of supposed group differences function as situational cues that could trigger stereotype threat underperformance effects (Steele et al. 2002). In early work, researchers described a test as either diagnostic or nondiagnostic of intellectual abilities. These descriptions functioned as situational cues signaling (or reducing) the possibility that Black students’ performance on the task could confirm the stereotype of intellectual inferiority. Black students at a highly selective institution were found to perform significantly worse on a verbal GRE test and reported more stereotype-relevant words when the test was labeled as diagnostic (threat condition) compared to non-diagnostic (no threat condition). Moreover, Black and White students’ performance did not differ in the non-diagnostic (no threat) condition (Steele and Aronson 1995). In similar early investigations of stereotype threat, researchers found that highly math-identified women performed significantly worse on a test described as having found gender differences in math performance in the past (threat condition) compared to no gender differences (no threat condition) (Spencer et al. 1999).
850
V. J. Taylor et al.
These studies highlighted central tenets of stereotype threat theory. Specifically, stereotype threat is theorized to arise in certain circumstances and with certain individual factors (Steele 1997). These include (1) when group members are aware of group-based negative stereotypes, (2) when group stereotypes are relevant to a social group’s outcomes in the setting, (3) when people value doing well in that domain (i.e., domain identification, and, particularly in the case of performance), (4) when a task or action is difficult. When these factors converge, negative situational cues triggering stereotype threat most strongly undermine performance (Steele et al. 2002). Hundreds of studies document stereotype threat underperformance effects (Inzlicht and Schmader 2012; Nguyen and Ryan 2008; Taylor et al. 2021a). However, some studies have failed to find support on various performance measures (e.g., Ganley et al. 2013). This makes sense, as stereotype threat performance effects may vary since they are: tied to the specific cultural stereotypes of a historical timeframe and place, often moderated by individual difference variables, and are dependent on the particular operationalization of the phenomenon itself (Murphy et al. 2022; Taylor et al. 2021a).
Consequences of Stereotype Threat Beyond Academic Performance Although important, underperformance is only one of the many downstream consequences of stereotype threat. First, it is critical to understand the social cognitive processes that most proximally highlight the experience of stereotype threat (Murphy et al. 2022). Since stereotype threat involves both general beliefs about how one will be perceived and treated (i.e., meta-perceptual concerns) and specific beliefs that others might apply group stereotypes to the self (i.e., meta-stereotypes; Vorauer et al. 1998), it is an inherently social cognitive process. It can initiate attentional, perceptual, and/or physiological vigilance processes. Such vigilance includes perceiving and attending to situational cues that suggest one will be negatively stereotyped and marked physiological responses to these cues (e.g., greater heart rate, greater skin conductance; Murphy et al. 2007). It also causes working memory and executive functioning impairments (e.g., for a review, see Schmader et al. 2008). Finally, stereotype threat can also increase rumination (i.e., the tendency to repetitively focus on negative feelings, thoughts, or situations) and, ironically, reduce effective attempts to suppress the very thoughts that drive such rumination (e.g., Logel et al. 2009). Social cognitive stereotype threat processes thereby impact important downstream outcomes. They include reduced career interests and aspirations (e.g., Cheryan et al. 2009), lower trust and comfort in a company (Purdie-Vaughns et al. 2008), decreased work satisfaction (e.g., von Hippel et al. 2013), increased workrelated burnout and intentions to quit (e.g., Hall et al. 2015), reduced performance (Nguyen and Ryan 2008; Taylor et al. 2021a), and overcompensation and avoidance behaviors in interracial interactions (e.g., Goff et al. 2008; Taylor et al. 2018). Importantly, these outcomes are likely to be most impactful among members of social groups who have historically faced de jure and de facto discrimination and
37
Stereotype Threat Experiences Across Social Groups
851
marginalization. Thus, stereotype threat can significantly impact life outcomes (e.g., opportunities, employment, health, and wealth; Aronson et al. 2013; Casad and Bryant 2016).
Social Groups, Stereotype Threat, and Affirmative Action What insights does stereotype threat theory provide for understanding divergent perceptions of and experiences with AA? To address this question, this section reviews research examining negative situational cues that elicit experiences of stereotype threat among minoritized and historically privileged groups. It highlights research suggesting that while minoritized group members are often concerned with being the targets of prejudice and discrimination, majority group members are more concerned about appearing prejudiced (Richeson and Shelton 2012). Various situational cues could signal to minoritized group members that they are likely to be the targets of prejudice and discrimination. For example, in one series of studies, researchers examined the effect of two situational cues – a company’s racial representation and its diversity policy – on Black professionals’ trust and comfort in the company (Purdie-Vaughns et al. 2008). The cues were manipulated through a brochure that conveyed either a high or a low numeric representation of employees of color and either a “colorblind ideology” (i.e., “we don’t see color”; focusing on similarities without recognizing differences) or a “value-diversity” ideology (i.e., “we value differences”; embrace one anothers’ diversity). The results showed that the combination of low racial representation and a colorblind ideology, compared to all other combinations of cues, increased a range of identity-related concerns among the Black professionals. Moreover, greater identity concerns (i.e., possibly facing discrimination, additional scrutiny, group devaluation) decreased Black professionals’ trust and comfort in the company (Purdie-Vaughns et al. 2008). This work was among the first to demonstrate the interactive effects of negative situational cues (i.e., numeric underrepresentation and colorblind company policies) on work-related outcomes in settings where group stereotypes are relevant. More recently, research highlights stereotype threat processes in interracial interactions among minoritized groups and provides insights into the impact of AAPs. In this research, people of color witnessed or imagined seeing a member of their racial group (i.e., ingroup) behave in a manner that does or does not confirm negative group stereotypes in interracial contexts (e.g., answering questions incorrectly, stealing vs. purchasing items). Following the ingroup members’ stereotypicallynegative behavior during a dyadic interracial interaction (with a White person) compared to an intraracial interaction (with a same-race person), participants expressed increased concerns about being viewed stereotypically (i.e., increased meta-stereotyping) and greater anxiety about the interaction. Heightened metastereotypes and anxiety, in turn, increased participants’ desire to disprove negative stereotypes but decreased their desire to engage with the White partner in the future (Taylor et al. 2018, 2021b). This work highlights how the behaviors of racial ingroup members can impact how one believes others see the self (i.e., meta-stereotypes) and
852
V. J. Taylor et al.
engages in interracial interactions. It also suggests that witnessing an ingroup member behave stereotypically is conceptually similar to being aware of negative references to AAPs. Both represent negative situational cues suggesting to individuals that they also face the possibility of being viewed stereotypically, triggering stereotype threat. To majority group members, such as White Americans, situational cues in interracial interactions may heighten their concern of appearing prejudice or racist (Goff et al. 2008; Richeson and Shelton 2012; Taylor et al. 2022). In one series of studies, White men had a conversation with two White or two Black men on either a generally race-neutral topic (i.e., love and relationships) or a race-salient topic (i.e., racial profiling). For the White participants, anticipating a conversation about racial profiling with Black partners brought the stereotype of Whites-as-racist to mind (Study 1) and increased worries about being stereotyped as racist (Study 4). This experience of stereotype threat then led them to sit farther away from their Black conversation partners, providing evidence of their physical discomfort and avoidance of the conversation (Goff et al. 2008). Relatedly, the results showed that Whites also express heightened concerns about being seen as racist when they imagine overhearing a White colleague tell an antiBlack joke during an interaction with a fellow Black (vs. White) colleague. Whites’ heightened concern about being seen as racist and anxiety, in turn, predicted greater motivation to disprove such stereotypes. However, they also expressed a greater desire to avoid future interactions with their Black colleague who witnessed the racist joke (Taylor et al. 2022). Likewise, these findings suggest that seeing a member of one’s group behave in a racist manner and awareness of negative references to AA are distinct negative situational cues for Whites in intergroup contexts. Both raise concerns that one could also be seen as racist. Moreover, Whites’ interactions with minoritized group members may suffer if unclear AAPs and negative references to them trigger an unsubstantiated concern that they will face “reverse discrimination” and fears about appearing prejudiced for harboring such thoughts. This situation can be particularly challenging for minoritized group members because they may face exclusion, suspicion, and resentment from White peers and colleagues, which can adversely impact their work-related outcomes. This work suggests that stereotype threat processes are likely involved when AAPs are salient features of the environment. It also indicates that when erroneous perceptions of AAPs abound, women and members of racially minoritized groups may face concerns and anxiety about the possibility of being seen as an unqualified recipient; this in turn adversely impacts their work and interactions with classmates and colleagues. Likewise, erroneous, “motivated,” or self-serving understandings and perceptions of AA (see Phillips and Lowery 2018) can trigger a different set of concerns among Whites that also adversely impact interactions with colleagues from minoritized groups. That is, for White Americans, unclear AAPs could function as negative situational cues, spurring unfounded concerns about “reverse discrimination” and worries about appearing racist or sexist, especially if such thoughts are made known to minoritized colleagues. In all, these concerns could lead to avoidance, exclusion, or rejection of these colleagues.
37
Stereotype Threat Experiences Across Social Groups
853
Stereotype Threat and Affirmative Action Research This chapter contends that experiences of stereotype threat, especially in the context of discrimination, can engender divergent perceptions of and experiences with AA for members of minoritized and privileged groups. Critics of AA have long argued that such policies will stigmatize minoritized groups and ultimately undermine their outcomes in academic or employment settings (e.g., Steele 1992). This section reviews several studies that have examined aspects of this assertion and test instances in which AAPs could serve as a negative situational cue suggesting to members of racially minoritized groups and women that they are unqualified and inferior (e.g., Leslie et al. 2014; Van Laar et al. 2008). This work also provides evidence that it is not AAPs themselves that cause adverse outcomes, but individuals’ misunderstanding and/or mischaracterization of such policies that lead to stereotype threat. There is mixed evidence that AAPs make negative group stereotypes salient and thereby cause downstream stereotype threat effects. For example, Van Laar et al. (2008) investigated whether Black/Latinx students who perceived themselves to be AA recipients experienced greater stereotype threat and lower academic performance. Results showed that perceiving oneself to be an AA recipient was associated with decreased academic performance when students also reported experiencing stereotype threat. Racially minoritized students who were suspicious of receiving preferential admission to college (conceptualized as experiencing a form of stereotype threat) showed a similar pattern: greater suspicion was related to lower selfreported college GPA (Brown et al. 2000, Study 2). Additionally, Fischer and Massey (2007) found mixed results when examining the relationship between being a possible AA recipient and college outcomes. They reported that being a potential AA recipient (operationalized by calculating a student’s SAT score relative to all students’ scores) led to slightly higher GPAs and greater retention. However, though modest, being a likely recipient who attended institutions where Black/Latinx students’ SAT scores fell below other students’ average SAT score was associated with lower GPA, college life satisfaction, and retention. Notably, covariates more strongly predicted the latter outcomes (i.e., parental education and academic preparedness), leading the authors to conclude that the net effect of AAPs serves to enhance minority students’ college experiences. Leslie et al. (2014) also found evidence that greater perceptions of being an AA recipient was associated with heightened concern that others hold negative stereotypes about them (i.e., meta-stereotyping). Further, meta-analytic data showed that the relationship between the presence of AAPs increased meta-stereotyping, lower self-competence, and negative affect, which in turn predicted lower self-evaluated and objective performance (Leslie et al. 2014). Relatedly, Brown et al. (2000, Study 1) provided experimental evidence of affirmative action’s role in producing stereotype threat performance effects. In this study, women were told they were assigned to a leadership role during a group task either randomly, due to their gender (i.e., preferential treatment), or according to their gender and qualifications (preferential treatment and qualifications). They then
854
V. J. Taylor et al.
completed a difficult GRE mathematical task. Results showed that women performed better on the task when told they were selected to the leadership role because of preferential treatment and their qualifications rather than because of their gender alone. However, other work found that women reported lower competence to perform an engineering communication task (a domain where women are negatively stereotyped) when they believed their selection was due to preferential treatment rather than merit-based (Heilman and Alcott 2001). These findings suggest that the negative effects of perceiving oneself as an AA recipient dissipate when there is a more accurate understanding of the processes involved in role selection (i.e., combination of preferential treatment and merit contributing to selection). Taken together, correlational and experimental findings suggest that perceived AA status is a negative situational cue that can, at times, undermine performance and self-confidence. Indeed, marginalized group members are well aware of the stereotypes associated with being perceived as a recipient of AA (Heilman and Alcott 2001; Leslie et al. 2014). However, previous research found that when students were less likely to perceive themselves to be AA recipients, they were significantly less likely to experience adverse academic outcomes (Brown et al. 2000; Van Laar et al. 2008). Thus, the meaning of an AA situational cue can vary and be a source of identity safety, specifically when this cue communicates that individuals’ social identity will be valued, welcomed, and supported in environments that include AAPs. If viewed without context or through a narrow lens, one might argue that AA is to blame for documented stereotype threat underperformance effects. This would be a misconception. Stereotype threat theory argues that situational cues and stereotypes themselves are the root problem. AAPs are important in creating rich, diverse environments and may be critical for providing spaces for students and employees to flourish (cf. Fisher and Massey 2007; Kalev et al. 2006). Removing these policies under the assumption that it might improve the experience of beneficiaries would thus be unproductive at best. Therefore, properly implemented and effectively communicated AAPs are posited to be critical in reducing the threat that may be associated with them.
Insights from Stereotype Threat Theory Stereotype threat theory can inform and explain understandings of discrimination and legal efforts to address past group harms in two major ways. First, mischaracterizations and misunderstandings of AAPs can function as cues that trigger stereotype threat for members of minoritized and historically privileged groups. Second, these considerations can inform the creation of identity-safe environments that attend to the concerns of both groups. Specifically, this section describes how stereotype threat theory pioneers a path for diverse groups through (a) education on critical historical knowledge, (b) accurate and strategic AAP messaging, and (c) renewed commitments to instituting evidence-based AAPs to reduce systemic group-based discrimination.
37
Stereotype Threat Experiences Across Social Groups
855
Insight One: Misconceptions and Mischaracterizations of AA Cues Stereotype Threat Misunderstood and/or mischaracterized AAPs can be negative situational cues that trigger stereotype threat. However, such cues often function differently for members of minoritized and privileged groups. At its base, misconceptions and mischaracterizations of AA lead to the unfounded conclusion that AA recipients are undeserving and unqualified (Leslie et al. 2014). However, the causal link likely operates in reverse as well – that is, undeserving and unqualified people are assumed to be the beneficiaries of “unfair” AAPs. Further, people often believe that quotas for traditionally underrepresented groups are used in AAP selection procedures, unless the specific process of selection is made clear (Heilman and Blader 2001). These are erroneous, pervasive, and damaging narratives for members of minoritized and historically privileged groups (e.g., Fryer and Loury 2005; Plous 2003). The presence of AAPs will likely raise divergent concerns for members of different groups. For instance, minoritized groups face concerns about being seen as unqualified AA recipients. Hence, they risk being stigmatized by the AA label, undermining their confidence and performance (Leslie et al. 2014; Van Laar et al. 2008). This stigma arises from prejudice and thus the discrimination is the root problem, yet the impact remains. Members of historically privileged groups face concerns that they may experience “reverse discrimination” due to beliefs about the presence of quotas or other practices deemed unmeritocratic (Pincus 2000; Unzueta et al. 2008), as well as concerns about appearing prejudiced for having these views. Accordingly, they could express resentment toward minoritized groups, exclude and socially reject them, and express reduced support for AAPs. These processes do not hinge on AAPs themselves but on the mischaracterization or misunderstanding of AAPs. How might these processes play out in the workplace? Consider again the experiences of the Black woman engineer described earlier. In this context, particularly virulent situational cues include unspecified and opaque AAPs in the workplace and negative comments about AA and assumed beneficiaries. Such cues may signal that she could be judged, treated, or evaluated in light of negative stereotypes about her race and gender at work. It is possible that the superficial understanding of AAPs and concerns about being seen as a beneficiary cause her to monitor her behavior to reduce the possibility of being viewed stereotypically. She may feel uncomfortable, lack trust and organizational commitment, experience increased burnout and intentions to quit, and ultimately decide to leave her position (e.g., Hall et al. 2015; Purdie-Vaughns et al. 2008). By comparison, identical situational cues may differentially impact the White male engineer’s experience of stereotype threat. The presence of an unspecified AAP and negative comments about AA and assumed beneficiaries could signal that he could be negatively judged, treated, or evaluated at work in terms of negative stereotypes about race and gender identity. Prevalent misunderstandings or mischaracterizations of AAPs may lead him to be concerned about “reverse discrimination,” despite his privileges. He may also be concerned about appearing racist if he
856
V. J. Taylor et al.
shares his thoughts, leading him to avoid interactions with colleagues of color, and leaving him without the knowledge that might lead to positive change. Given his privileged identities in this setting, these negative situational cues are counteracted by a myriad of positive cues (e.g., a significant number of employees that share his identities and his informal access to senior colleagues). These cues may buffer against or prevent him from experiencing the negative downstream outcomes of stereotype threat. As these examples illustrate, stereotype threat theory suggests that researchers and practitioners must consider social group membership to understand how diverse groups are likely to perceive and experience AAPs. In particular, it is crucial to assess their potentially inconsistent meaning and how they can serve as negative situational cues triggering stereotype threat for minoritized and majority group members alike. Understanding the divergent meanings of AA/AAPs and what they represent can therefore inform the creation of identity-safe environments for diverse groups.
Insight Two: Creating Identity-Safe Environments in the Context of Affirmative Action Social groups’ divergent perceptions of and experiences with AA call attention to an important component of stereotype threat theory. That is, to successfully alleviate stereotype threat and improve outcomes, negative or threatening situational cues must be successfully reframed or removed from the setting to create and maintain identity-safe environments. Identity-safe environments are spaces that provide positive situational cues of respect for and equitable treatment of individuals’ social identities (Hall et al. 2018). Below, several important considerations are suggested for effective messaging and learning about AAPs that will contribute to creating and maintaining identity-safe environments. The first vital step is that of creating spaces where people are afforded opportunities to learn critical historical knowledge that situates discrimination and AA. Next, stereotype threat theory suggests that AAPs must be accurately and strategically reframed (though not removed) to function as a positive cue that considers the divergent concerns of minoritized and privileged group members. Finally, stereotype threat research and theory supports a renewed commitment to instituting and maintaining AAPs, given the cascade of protective factors engendered by identity-safe environments.
Gaining Critical Historical Knowledge About Discrimination and Social Inequities First, being educated about the history and legacy of discrimination and group-based inequities (i.e., critical historical knowledge) is necessary for understanding the intent and function of AAPs. Specifically, the intent and function of AAPs must be
37
Stereotype Threat Experiences Across Social Groups
857
contextualized within the historical patterns, laws and policies, and discrimination that created and reinforced group-based inequities in the USA (Rothstein 2015). Teaching about the histories of groups that have faced discrimination is important for at least two reasons. First, many people are unaware of or misunderstand the extent to which historically minoritized groups have experienced and currently experience systemic racism, sexism, and discrimination (e.g., Kraus et al. 2017). Second, more accurate knowledge about the role various forms of oppression (including but not limited to racism, sexism, heterosexism, classism, ableism, and their combinations) have played in the lives of different social groups throughout history is associated with increased recognition of current systemic forms of discrimination (Nelson et al. 2013). For example, research suggests that lack of knowledge about the history of US racism predicts Whites’ denial of systemic racism (Bonam et al. 2019; Nelson et al. 2013). Fortunately, Bonam et al. (2019) demonstrated that a short educational intervention that provides critical historical knowledge about how racism can operate in institutions and structures (i.e., racially discriminatory public policy) increased Whites’ acknowledgment of systemic racism and its impact. Thus, critical historical knowledge can help individuals understand justifications for efforts to rectify past harms and present disparities. In other words, as one begins to understand the systemic nature of past discrimination, one can come to understand the value of a systemic approach (e.g., AAPs) to reduce present inequities. Such knowledge lays the foundation for understanding AAP content, intent, and implementation.
Accurate and Strategic Messaging of AAPs Messaging about AAPs must be clear and accurate to counteract misconceptions and mischaracterizations that can function as negative situational cues contributing to stereotype threat. For example, inaccurate messaging may suggest that AAPs include quotas and rewards unqualified individuals (Fryer and Loury 2005; Plous 2003). Also, unclear messaging about what AAPs entail allows for ambiguity surrounding these policies (Brown et al. 2000). Such messaging, or lack thereof, can contribute to negative perceptions of and experiences with AA among both minoritized and historically privileged groups. Accurate messaging of AAPs should include factual descriptions of policy content (to dispel myths or assumptions) and explanations of intent and implementation. Next, additional insights are offered for why accurate and strategic messaging of AAPs is necessary for effectively reframing the meaning of these situational cues. For minoritized groups, accurate and strategic messaging of AA is essential for dispelling negative stereotypes about their intelligence, inferiority, and deservedness. Research supports this contention. For example, women who thought that their partner held the view that they had been preferentially selected for a leadership position because of their gender and merit rather than their gender alone believed their partner held more positive views about their competence (Heilman and Alcott
858
V. J. Taylor et al.
2001). These findings suggest that AA messaging that includes clear information about women’s merit changed the meaning of this situational cue for recipients and reduced their concerns about being negatively stereotyped. These and related findings (e.g., Brown et al. 2000) are encouraging because such messaging will alleviate the ambiguity that one might not be deserving (i.e., attributional ambiguity) and associated adverse outcomes. Moreover, this messaging aligns with current AAPs that cannot and do not allow for the admittance or hiring of members of historically underrepresented groups who are “unqualified.” Thus, accurate AA messaging addresses critics’ concerns about the stigmatization related to being labeled an AA recipient (S. Steele 1992). Accurate and strategic messaging is also critical for increasing knowledge about the content, intent, and implementation of AAPs among historically privileged groups (e.g., White Americans, White men). AAPs and negative stereotypes about recipients’ intelligence, inferiority, and deservedness impact privileged group members’ attitudes toward AAPs and beneficiaries. Accurate messaging is vital, but given the threat that AA can arouse for Whites (e.g., Renfro et al. 2006), AAP messaging must also be strategic. Research by Lowery et al. (2012) on inequality framing offers insight into how accurate AAP messaging should be structured to reduce opposition, and by extension, reduce concerns of reverse discrimination and/or appearing prejudiced. Specifically, they found that when instances of racial inequality were framed as Whites having an advantage compared to racially minoritized groups having a disadvantage, Whites expressed more support for redistributive policies that correct for inequalities. Moreover, they found that the White advantage frame reduced Whites’ esteem for their racial group, thereby boosting support for redistributive policies. These and related findings (e.g., Lowery et al. 2009) suggest that accurately framing AAPs by highlighting the advantages Whites as a group have had throughout US history may be unsettling and create dissonance for Whites. It focuses attention on White’s unearned privileges, which contradict core beliefs about meritocracy and individualism. To resolve this dissonance, Whites become more supportive of redistributive policies for minoritized groups that, by extension, restore their group image. Importantly, this work accurately and strategically frames AAPs with a focus on White advantage rather than the traditional focus on the disadvantage of racially minoritized groups. Thus, AAP messaging that highlights White advantage, bolstered by critical historical knowledge about discrimination and social inequities, will mitigate concerns of reverse discrimination and appearing prejudiced. Taken together, such messaging is important for improving Americans’ understanding of AAPs more generally but is also vital for students and employees in academic and organizational contexts. Moreover, accurate and strategic AAP messaging supports the creation and maintenance of identity-safe environments for members of diverse social groups though their purpose and implementation may be different for minoritized and historically privileged groups. If done well, such messaging highlights the efficacy of what Dr. Randall Kennedy calls “sensibly designed” AAPs. It accurately reframes the meaning of these policies to dispel
37
Stereotype Threat Experiences Across Social Groups
859
negative stereotypes, mitigate identity-related concerns, and supports the advancement of underrepresented and minoritized groups. Given this strategy, there is (at least) one important caveat to acknowledge. For some, exposure to such AAP messaging will not change their attitudes about AA. That is, their opposition is intertwined with core values of meritocracy and individualism that suggest AAPs are unmeritocratic and unfair. These views may also be associated with individuals’ prejudice and resistance to acknowledging the deeply imbued and often invisible systemic racism, sexism, etc. in the USA (e.g., Rothstein 2015). Still, others may view any AA approach as unfair despite over four centuries of discriminatory practices and policies and less than 60 years of AAPs to redress them. Although it is difficult to precisely estimate the proportion of people holding these views, is likely a relatively small portion of the US population. Research on general support for equal opportunity initiatives (Crosby et al. 2003) and recent polls on support for AA in particular (Saad 2021) corroborate this contention. Thus, a portion of AA critics may simply disagree about the best way to rectify discrimination. For most people, accurate and strategic AAP messaging may be a critical and necessary strategy to address opposition. Coupled with critical historical knowledge of discrimination and social inequities, such AAP messaging stands to create identity-safe environments for both minoritized and historically privileged groups.
Renewing Commitments to AAPs Lastly, insights from stereotype threat theory, and research on identity-safe environments, in particular, suggest that policymakers and practitioners should continue or renew their commitment to instituting and maintaining AAPs. Importantly, research finds that AA works to increase the representation of historically excluded and marginalized groups in education and employment and improve group outcomes (Fisher and Massey 2007; Kalev et al. 2006). From a stereotype threat perspective, increasing the representation of women, racially minoritized groups, and other historically excluded groups is vital for creating diverse identity-safe environments. Increased numeric representation of minoritized groups in academic and work environments is an identity-safe cue that increases feelings of belonging, trust, and comfort (Murphy et al. 2007; Purdie-Vaughns et al. 2008; Sekaquaptewa and Thompson 2002). Belonging, trust, and comfort are important factors contributing to academic achievement and organizational commitment for stereotyped individuals (Casad and Bryant 2016). Moreover, increased numeric representation can support the development of ingroup role models, which also improves stereotyped individuals’ outcomes in these settings (e.g., Cortland and Kinias 2019). Likewise, such increased numeric representation inevitably increases an institution’s diversity, making interactions among individuals from different social groups (i.e., intergroup contact) more likely. Intergroup contact has been shown to reduce intergroup anxiety and increase empathy, which in turn decreases prejudice, particularly among majority group members (Pettigrew and Tropp 2008). Such intergroup interactions, and
860
V. J. Taylor et al.
cross-race friendships, in particular, expose majority group members to different perspectives (Davies et al. 2011). Intergroup interactions also increase majority group members’ psychological investment in the welfare of outgroup members, which heightens their consideration of social inequities (Tropp and Barlow 2018) and can inform and improve their attitudes about AAPs and beneficiaries. Taken together, adopting accurately and strategically messaged AAPs can benefit institutions, members of minoritized and historically privileged groups, and help to create and maintain identity-safe environments. With the adoption of such AAPs, beneficiaries will face significantly less negative stereotyping and stigmatization. Moreover, minoritized and historically privileged groups will have more opportunities to interact with those unlike themselves, form friendships and close relationships, and expand their perspectives and attitudes on AA and other issues. Thus, committing to AAPs provides a cascade of protective factors that can engender identity-safe environments for all.
Concluding Thoughts Widespread misunderstanding and mischaracterization of discrimination and AA are critical factors that can thwart support and adoption of AAPs in the United States. Stereotype threat theory and research provide insights into how one’s social identities can impact divergent experiences and perceptions of discrimination and AA. Furthermore, stereotype threat theory informs strategies that consider the concerns of AA beneficiaries and non-beneficiaries to engender a more accurate understanding of AAPs that may support their adoption. Currently AA/AAPs can function as negative situational cues that trigger different concerns for diverse groups. For members of racially minoritized groups and women, poorly communicated AAPs may trigger concerns about being negatively stereotyped, in turn impacting their academic or work-related outcomes. By comparison, these same policies may raise unsubstantiated concerns for historically privileged groups that they will face “reverse discrimination” or appear prejudiced for opposing AAPs, in turn impacting their interactions with perceived AA beneficiaries. To address these divergent concerns and adverse outcomes, three strategies to support implementing “sensibly designed” AAPs for diverse groups are proposed. Specifically, seemingly opposed viewpoints on AA and the divergent concerns they arouse can be resolved by creating and maintaining identity-safe environments. Such identity-safe environments (a) educate individuals about discrimination and social inequities, (b) provide accurate and strategic AAP messaging, and (c) renew their commitment to “sensibly designed” AAPs that work to increase minoritized groups’ representation, improve intergroup contact, and redress systemic inequities. In summary, stereotype threat theory provides critical insights for understanding discrimination and AA for groups traditionally considered AA beneficiaries (i.e., women, Black/African Americans, Latinx Americans) and members of less discussed groups, including but not limited to members of 2SLGBTQIA+ communities, veterans, pregnant women, and people with disabilities. Further, given that
37
Stereotype Threat Experiences Across Social Groups
861
negative stereotypes can more severely impact people at the intersection of various identities, it is equally important to consider their perspectives on and concerns about AAPs. Thus, stereotype threat theory provides insight into why attuning to diverse identity threat concerns is necessary to create a more accurate and meritocratic understanding of and experiences with AAPs to achieve equitable academic and work environments. In all, AAPs will be good for individuals and “conducive to the public good.”
Cross-References ▶ Evidence of Covariation Between Regional Implicit Bias and Socially Significant Outcomes in Healthcare, Education, and Law Enforcement Acknowledgments This chapter was supported by a National Science Foundation (NSF) Career grant awarded to V.J.T. (HRD-2047105).
References Alexander M (2010) The new Jim Crow mass incarceration in the age of colorblindness. The New Press, New York Aronson J, Burgess D, Phelan SM, Juarez L (2013) Unhealthy interactions: the role of stereotype threat in health disparities. Am J Public Health 103(1):50–56. https://doi.org/10.2105/AJPH. 2012.300828 Biernat M, Collins EC, Katzarska-Miller I, Thompson ER (2009) Race-based shifting standards and racial discrimination. Personal Soc Psychol Bull 35(1):16–28 Bonam CM, Nair Das V, Coleman BR, Salter P (2019) Ignoring history, denying racism: mounting evidence for the Marley hypothesis and epistemologies of ignorance. Soc Psychol Personal Sci 10(2):257–265. https://doi.org/10.1177/1948550617751583 Brown RP, Charnsangavej T, Keough KA, Newman ML, Rentfrow PJ (2000) Putting the “affirm” into affirmative action: preferential selection and academic performance. J Pers Soc Psychol 79(5):736–747. https://doi.org/10.1037/0022-3514.79.5.736 Bureau of National Affairs (1979) Uniform guidelines on employee selection procedures Casad BJ, Bryant WJ (2016) Addressing stereotype threat is critical to diversity and inclusion in organizational psychology. Front Psychol 7. https://doi.org/10.3389/fpsyg.2016.00008 Cheryan S, Plaut VC, Davies PG, Steele CM (2009) Ambient belonging: how stereotypical cues impact gender participation in computer science. J Pers Soc Psychol 97(6):1045–1060. https:// doi.org/10.1037/a0016239 Cortland CI, Kinias Z (2019) Stereotype threat and women’s work satisfaction: the importance of role models. Arch Sci Psychol 7(1):81–89. https://doi.org/10.1037/arc0000056 Crosby FJ, Cordova DI (1996) Words worth of wisdom: toward an understanding of affirmative action. J Soc Issues 52(4):33–49. https://doi.org/10.1111/j.1540-4560.1996.tb01847.x Crosby FJ, Iyer A, Clayton S, Downing RA (2003) Affirmative action: psychological data and the policy debates. Am Psychol 58(2):93–115. https://doi.org/10.1037/0003-066X.58.2.93 Crosby FJ, Iyer A, Sincharoen S (2006) Understanding affirmative action. Annu Rev Psychol 57: 585–611. https://doi.org/10.1146/annurev.psych.57.102904.190029 Davies K, Tropp LR, Aron A, Pettigrew TF, Wright SC (2011) Cross-group friendships and intergroup attitudes: a meta-analytic review. Personal Soc Psychol Rev 15(4):332–351. https://doi.org/10.1177/1088868311411103
862
V. J. Taylor et al.
Fisher MJ, Massey DS (2007) The effects of affirmative action in higher education. Soc Sci Res 36(2):531–549. https://doi.org/10.1016/j.ssresearch.2006.04.004 Fisher v. University of Texas at Austin, et al., 570 US 297 (F.5d 2013) Fryer RG, Loury GC (2005) Affirmative action and its mythology. J Econ Perspect 19(3):147–162. https://doi.org/10.1257/089533005774357888 Ganley CM, Mingle LA, Ryan AM, Ryan K, Vasilyeva M, Perry M (2013) An examination of stereotype threat effects on girls’ mathematics performance. Dev Psychol 49(10):1886–1897. https://doi.org/10.1037/a0031412 Garces LM, Poon O (2018) Asian Americans and race-conscious admissions: understanding the conservative opposition’s strategy of misinformation, intimidation & racial division. The Civil Rights Project. https://www.civilrightsproject.ucla.edu/ Goff PA, Steele CM, Davies PG (2008) The space between us: stereotype threat and distance in interracial contexts. J Pers Soc Psychol 94(1):91–107. https://doi.org/10.1037/0022-3514.94. 1.91 Gratz v. Bollinger, 539 US 244 (Federal District Court 2003) Hall WM, Schmader T, Croft E (2015) Engineering exchanges: daily social identity threat predicts burnout among female engineers. Soc Psychol Personal Sci 6(5):528–534. https://doi.org/10. 1177/1948550615572637 Hall WM, Inness M, Schmader T, Aday A, Croft E (2018) Climate control: the relationship between social identity threat and cues to an identity-safe culture. J Pers Soc Psychol 115(3):446–467 Heilman ME, Alcott VB (2001) What I think you think of me: women’s reactions to being viewed as beneficiaries of preferential selection. J Appl Psychol 86(4):574–582. https://doi.org/10. 1037/0021-9010.86.4.574 Heilman ME, Blader SL (2001) Assuming preferential selection when the admissions policy is unknown: the effects of gender rarity. J Appl Psychol 86(2):188–193. https://doi.org/10.1037/ 0021-9010.86.2.188 Herrnstein RJ, Murray CA (1994) The bell curve: intelligence and class structure in American life. Free Press Inzlicht M, Schmader T (eds) (2012) Stereotype threat: theory, process, and application. Oxford University Press Jencks C, Phillips M (1998) The Black-White test score gap. Brookings Institution Press Jones KP, Peddie CI, Gilrane VL, King EB, Gray AL (2016) Not so subtle: a meta-analytic investigation of the correlates of subtle and overt discrimination. J Manag 42(6):1588–1613. https://doi.org/10.1177/0149206313506466 Kaiser CR, Dover TL, Small P, Xia G, Laura M, Brady LM, Major B (2021) Diversity initiatives and White Americans’ perceptions of racial victimhood. Person Soc Psychol Bull. https://doi. org/10.1177/01461672211030391 Kalev A, Dobbin F, Kelly K (2006) Best practices or best guesses: assessing the efficacy of corporate affirmative action and diversity policies. Am Sociol Rev 71(4):589–617. https://doiorg.ezproxy.lib.lehigh.edu/10.1177/000312240607100404 Kemmelmeier M (2003) Individualism and attitudes toward affirmative action: evidence from priming experiments. Basic Appl Soc Psychol 25(2):111–119. https://doi.org/10.1207/ S15324834BASP2502_2 Kraus MW, Rucker JM, Richeson JA (2017) Americans misperceive racial economic equality. Proc Natl Acad Sci 114(39):10324–10331. https://doi.org/10.1073/pnas.1707719114 La Salle TP, Wang C, Wu C, Neves JR (2020) Racial mismatch among minoritized students and white teachers: implications and recommendations for moving forward. J Educ Psychol Consult 30(3):314–343. https://doi.org/10.1080/10474412.2019.1673759 LaCosse J, Murphy MC, Garcia JA, Zirkel S (2020) The role of STEM professors’ mindset beliefs on students’ anticipated psychological experiences and course interest. J Edu Psychol. https:// doi.org/10.1037/edu0000620
37
Stereotype Threat Experiences Across Social Groups
863
Leslie LM, Mayer DM, Kravitz DA (2014) The stigma of affirmative action: a stereotyping-based theory and meta-analytic test of the consequences for performance. Acad Manag J 57(4): 964–989. https://doi.org/10.5465/amj.2011.0940 Li S, Rincon R, Williams J (2017) Climate control? Gender and racial bias in engineering. 2017 ASEE Annual Conference & Exposition Proceedings, Columbus Liu A (2011) Unraveling the myth of meritocracy within the context of US higher education. High Educ 62(4):383–397. https://doi.org/10.1007/s10734-010-9394-7 Logel C, Iserman E, Davies P, Quinn D (2009) The perils of double consciousness: the role of thought suppression in stereotype threat. J Exp Soc Psychol 45(2):299–312. https://doi.org/10. 1016/j.jesp.2008.07.016 Lowery BS, Chow RM, Randall-Crosby J (2009) Taking from those that have more and giving to those that have less: how inequity frames affect corrections for inequity. J Exp Soc Psychol 45(2):375–378. https://doi.org/10.1016/j.jesp.2008.09.010 Lowery B, Chow R, Knowles E, Unzueta M (2012) Paying for positive group esteem: how inequity frames affect whites’ responses to redistributive policies. J Pers Soc Psychol 102(2):323–336. https://doi.org/10.1037/a0024598 Major B, Quinton WJ, McCoy SK (2002) Antecedents and consequences of attributions to discrimination: theoretical and empirical advances. Adv Exp Soc Psychol 34:251–330 Murphy MC, Steele CM, Gross JJ (2007) Signaling threat: how situational cues affect women in math, science, and engineering settings. Psychol Sci 18(10):879–885. https://doi.org/10.1111/j. 1467-9280.2007.01995.x Murphy MC, Taylor VJ, Steele CM (2022) Stereotype threat: a situated theory of social cognition. In: A situated theory of social cognition. Oxford University Press National Center for Science and Engineering Statistics (2019) Women, minorities, and persons with disabilities in science and engineering. National Science Foundation. www.nsf.gov/statistics/ wmpd/ Nelson JC, Adams G, Salter PS (2013) The Marley hypothesis: denial of racism reflects ignorance of history. Psychol Sci 24(2):213–218. https://doi.org/10.1177/0956797612451466 Nguyen H-HD, Ryan AM (2008) Does stereotype threat affect test performance of minorities and women? A meta-analysis of experimental evidence. J Appl Psychol 93(6):1314–1334. https:// doi.org/10.1037/a0012702 Ong M, Wright C, Espinosa L, Orfield G (2011) Inside the double bind: a synthesis of empirical research on undergraduate and graduate women of color in science, technology, engineering, and mathematics. Harv Educ Rev 81(2):172–209. https://doi.org/10.17763/haer.81.2. t022245n7x4752v2 Pettigrew TF, Tropp LR (2008) How does intergroup contact reduce prejudice? Meta-analytic tests of three mediators. Eur J Soc Psychol 38(6):922–934. https://doi.org/10.1002/ejsp.504 Phillips LT, Lowery BS (2018) Herd invisibility: the psychology of racial privilege. Curr Dir Psychol Sci 27(3):156–162. https://doi.org/10.1177/0963721417753600 Pincus FL (2000) Reverse discrimination vs white privilege: an empirical study of alleged victims of affirmative action. Race Soc 3(1):1–22. https://doi.org/10.1016/S1090-9524(01)00018-3 Plous S (2003) Ten myths about affirmative action. In: Understanding prejudice and discrimination. McGraw-Hill, pp 206–212 Purdie-Vaughns V, Steele CM, Davies PG, Ditlmann R, Crosby JR (2008) Social identity contingencies: how diversity cues signal threat or safety for African Americans in mainstream institutions. J Pers Soc Psychol 94(4):615–630. https://doi.org/10.1037/0022-3514.94.4.615 Renfro CL, Duran A, Stephan WG, Clason DL (2006) The role of threat in attitudes toward affirmative action and its beneficiaries. J Appl Soc Psychol 36(1):41–74. https://doi.org/10. 1111/j.0021-9029.2006.00003.x Richeson JA, Shelton JN (2012) Stereotype threat in interracial interactions. In: Stereotype threat: theory, process, and application. Oxford University Press, pp 231–245 Rothstein R (2015) The racial achievement gap, segregated schools, and segregated neighborhoods: a constitutional insult. Race Soc Probl 7(1):21–30. https://doi.org/10.1007/s12552-014-9134-1
864
V. J. Taylor et al.
Saad L (2021, July 30) Americans’ confidence in racial fairness waning. Gallup Organization, Washington, DC. https://news.gallup.com/poll/352832/americans-confidence-racial-fairnesswaning.aspx Schmader T, Johns M, Forbes C (2008) An integrated process model of stereotype threat effects on performance. Psychol Rev 115(2):336–356. https://doi.org/10.1037/0033-295X.115.2.336 Schuette v. Coalition, 572 US 291 (United States District Court for the Eastern District of Michigan 2014) Sekaquaptewa D, Thompson M (2002) The differential effects of solo status on members of highand low-status groups. Personal Soc Psychol Bull 28(5):694–707. https://doi.org/10.1177/ 0146167202288013 Slaughter JE, Sinar EF, Bachiochi PD (2002) Black applicants’ reactions to affirmative action plans: effects of plan content and previous experience with discrimination. J Appl Psychol 87(2): 333–344. https://doi.org/10.1037/0021-9010.87.2.333 Spencer SJ, Steele CM, Quinn DM (1999) Stereotype threat and women’s math performance. J Exp Soc Psychol 35(1):4–28. https://doi.org/10.1006/jesp.1998.1373 Steele S (1992, August) The new segregation. Imprimis 21(8) Steele CM (1997) A threat in the air: how stereotypes shape intellectual identity and performance. Am Psychol 52(6):613–629. https://doi.org/10.1037/0003-066X.52.6.613 Steele CM, Aronson J (1995) Stereotype threat and the intellectual test performance of African Americans. J Pers Soc Psychol 69(5):797–811. https://doi.org/10.1037/0022-3514.69.5.797 Steele CM, Spencer SJ, Aronson J (2002) Contending with group image: the psychology of stereotype and social identity threat. In: Advances in experimental social psychology, vol 34. Elsevier, pp 379–440. https://doi.org/10.1016/S0065-2601(02)80009-0 Students for Fair Admissions, Inc. v. University of North Carolina, et al. 954 F.3d 190 (4th Cir. 2019) Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll. 980 F.3d 157 (1st Cir. 2020) Tajfel H, Turner JC (1986) The social identity theory of intergroup behavior. In: Psychology of intergroup relations. Hall Publishers, pp 7–24. https://doi.org/10.4324/9780203505984-16 Taylor VJ, Garcia RL, Shelton JN, Yantis C (2018) “A threat on the ground”: the consequences of witnessing stereotype-confirming ingroup members in interracial interactions. Cult Diver Ethnic Minor Psychol 24(3):319–333. https://doi.org/10.1037/cdp0000190 Taylor VJ, Forscher PS, Walton GM (2021a) Stereotype threat meta-analysis: the strength of the stereotype threat effect across multiple populations and domains. Unpublished manuscript Taylor VJ, Yantis C, Bonam C, Hart A (2021b) What to do? Predicting coping strategies following ingroup members’ stereotypical behaviors in interracial interactions. Person Soc Psychol Bull 47(7):1084–1100. https://doi.org/10.1177/0146167220960269 Taylor VJ, Yantis C, Valladares JV (2022) Whites’ reactions to ingroup members’ stereotypeconfirming behavior: consequences for interracial contact. Unpublished manuscript Trawalter S, Bart-Plange D-J, Hoffman KM (2020) A socioecological psychology of racism: making structures and history more visible. Current Opin Psychol 32:47–51. https://doi.org/ 10.1016/j.copsyc.2019.06.029 Tropp LR, Barlow FK (2018) Making advantaged racial groups care about inequality: intergroup contact as a route to psychological investment. Curr Direct Psychol Sci 27(3):194–199. https:// doi.org/10.1177/0963721417743282 University of California Regents v. Bakke, 438 US 265 (Supreme Court of California 1978) Unzueta M, Lowery B, Knowles E (2008) How believing in affirmative action quotas protects White men’s self-esteem. Organ Behav Hum Dec Proc 105(1):1–13. https://doi.org/10.1016/j. obhdp.2007.05.001
37
Stereotype Threat Experiences Across Social Groups
865
Van Laar C, Levin S, Sinclair S (2008) Social identity and personal identity stereotype threat: the case of affirmative action. Basic Appl Soc Psychol 30(4):295–310. https://doi.org/10.1080/ 01973530802502200 von Hippel C, Kalokerinos EK, Henry JD (2013) Stereotype threat among older employees: relationship with job attitudes and turnover intentions. Psychol Aging 28(1):17–27. https:// doi.org/10.1037/a0029825 Vorauer JD, Main KJ, O’Connell GB (1998) How do individuals expect to be viewed by members of lower status groups? Content and implications of meta-stereotypes. J Person Soc Psychol 75(4):917–937. https://doi.org/10.1037/0022-3514.75.4.917 Wilkins CL, Wellman JD, Babbitt LG, Toosi NR, Schad KD (2015) You can win but I can’t lose: bias against high-status groups increases their zero-sum beliefs about discrimination. J Exp Soc Psychol 57:1–14. https://doi.org/10.1016/j.jesp.2014.10.008
Part VII Miscellaneous
Feminism as Racist Backlash: How Racism Drove the Development of Nineteenthand Twentieth-Century Feminist Theory
38
Tommy J. Curry
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Against Savage (Black) Manhood: Elizabeth Cady Stanton, Complementarianism, and the Uniting of the White Race . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Evolutionary Origin of Patriarchy is Feminine – Gilman’s Claiming of White Supremacy as the Gift of White Womanhood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Racial Proxemic: [W]hite Womanhood, Feminism, and the Dangers of Black Men to White Civilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feminist Theory’s Adoption of Subculture of Violence Criminology as a Response to Desegregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
870 872 879 885 889 892 893
Abstract
American feminism’s anti-Black racism is often presented as a failure of white feminists to integrate Black women into their movement. This historiographic approach presumes that feminism was a progressive movement that merely suffered from blind spots in its approach to women’s rights due to the biases of some white women. Unlike previous research which has pointed out the individual racism of suffragettes and mid-twentieth-century feminists, this chapter argues for an understanding of the theories created and endorsed by feminists from 1860 to 1980 as the consequence of feminism’s dependence on racist theories predicated on research from ethnologists and evolutionary theorists in the nineteenth and criminologists in the twentieth centuries. By looking at the primary racial target of feminist thought and activism over the centuries, the Black male, I argue scholars can more accurately trace the theories feminists used to derail Black American’s struggle for civil rights. T. J. Curry (*) Philosophy, University of Edinburgh, Edinburgh, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_48
869
870
T. J. Curry
Keywords
Anti-Black racism · Black males · Suffragism · Ethnology · Subculture of violence · Feminism’s racism
Introduction Quite often, the more popular an idea becomes, the harder it is to question the rise of the idea. Introducing skepticism, or questioning the propagandist aspects of what has become a cherished idea(l) – a belief thought to be obvious or intuitively correct – is often seen a heresy. Feminism has become such belief. Over the last several decades, the history of white women’s rights movements and its leaders has become increasing described as not only influenced by racism but intimately committed to the preservation and imperial projection of white supremacy (Sneider 2008). For example, the historian Louise Newman (1999) has argued that “racism was not just an unfortunate sideshow in the performances of feminist theory. Rather it was center stage: an integral, constitutive element in feminism’s overall understanding of citizenship, democracy, political self-possession, and equality” (p. 183). Rather than being a generative critique of the racist legacy suffragists, segregationists, and the competing political ideologies of twentieth-century America, much of the interpretive fervor given to the racism of American feminism has focused on the historical failure of integrating Black women into the feminist movement and the contemporary absence of intersectionality among white women. The explanation of feminism’s failures has been articulated as failures of inclusion – the exclusion of Black women. Little to no research has evaluated the cost of feminism racist caricatures and political appeal to white supremacy to Black men (Curry 2021b). The quest for white women’s rights is not merely a history of feminism’s resistance against patriarchy and oppression. In many cases, feminism did not challenge patriarchy at all (Curry 2017; Newman 1999). In actuality, women’s rights are often a cataloguing of the theories developed by white women under the mantle of feminism which sought to justify the capability of white women to rule over the darker races alongside white men. For over a century, feminism has embraced the sciences of its time be it evolutionary ethnology in the nineteenth century or criminology in the twentieth century to construct racialized others as deviant and uncivilized populations able to be transformed through white women’s benevolence (Bhattacharyya 2013; Zakaria 2021). The racism of white women should not be misunderstood to be mere personal predilections or the shared view of the women of their time. The racism of white feminists and suffragists was forged by deliberate action, theoretical innovation, and white women’s investment in womanhood throughout history. White women’s participation in the slave trade and the owning, selling, and abuse of enslaved Black men and women (Jones-Rogers 2019) created a basis upon which womanhood was claimed to be above the savagery of the Negro. Even the right to abortion was used by white feminist as threat against white men. Suffragettes leveraged their wombs for political power and threatening white men
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
871
with racial suicide if they did not comply. In the latter half of the nineteenth century, it was thought only savage women who did not value motherhood would abort their babies, and white women making such a threat were thought to devolve the race (Beisel and Kay 2004). This chapter will argue that it is essential to understand the racialist discourse and political theories of nineteenth- and twentieth-century feminism as a backlash against the enfranchisement of Black men and the acquisition of political and economic power by Black Americans. Understanding feminism as white women’s backlash to Black social and political mobility is proposed as an intervention into current historiographic accounts of women’s rights and liberal political theories which assert a narrative of an ever-expanding democratic freedom for US citizens. While there has been a swell of substantial works looking at the political ideology, imperial justifications, and racial reactions of white women to Black enfranchisement, there has been no engagement with how white feminism has positioned Black males as the border of racial advancement. The theories white feminists introduced throughout the nineteenth and twentieth centuries to limit Black male economic advancement and social mobility energized the coalescence of white power against Black civil rights for over a century. In using the term racial backlash to describe feminism, I mean to convey a deliberate and dedicated program of action by a dominant group, in this case white women, who along with various white supremacist entities sought to eradicate the perceived gains of suffrage and citizenship for Blacks through the criminalization of Black men. Like well-established conceptualizations of racial backlash as “a forceful swing against a perceived unwelcome change to the status quo. . . [indicated by] a strong adverse reaction against various racial remedies adopted by national governments for the effects of centuries long racial discrimination” (Ansell 2013, p. 17), this chapter conceptualizes the deployment of ethnology and criminology by white feminist leaders and thinkers as a response to the perceived gains Black Americans achieved from the mid-1800s to the 1980s. This chapter begins with a history of suffrage and the theory of gender complementarianism underlying much of Elizabeth Cady Stanton’s analysis of male power during the 1860s. Following the science of ethnology, Stanton and other suffragettes believed women’s rights helped to evolve the white race toward more effective rule over other lesser racial groups. In the second section entitled “The Evolutionary Origin of Patriarchy is Feminine – Gilman’s Claiming of white Supremacy as the Gift of white Womanhood,” the theme of racial evolution is read through the work of Charlotte Perkins Gilman. Gilman is often underappreciated as an ethnological theorist. Her theory of racial development situates the maternal as the foundation of white patriarchy and white imperial power. Her book, Women and Economics, provides invaluable resources for understanding how white women’s liberation was not solely a political movement but an evolutionary necessity for the white race to direct the course of human civilization. Gilman believed the imperial ventures of the white race could only be sustained by understanding that the place of white women was not the home but as a colonizer alongside white men. The third section, “The Racial Proxemic: [w]hite Womanhood, Feminism, and the Dangers of
872
T. J. Curry
Black Men to white Civilization,” concerns itself with the women’s rights activists of the 1920s to the 1960s. Following the declines of America’s imperial ventures, white women’s rights activists used the language of white women’s vulnerability and their newly won enfranchisement to support racial segregation policies and establish Ku Klux Klan organizations, especially throughout the South. The last section “Feminist Theory’s Adoption of Subculture of Violence Criminology as a Response to Desegregation” focuses on the legislation and theories white feminists endorsed at the expense of Black Americans who believed their protests and spilled blood would bring about equality in the 1960s and 1970s. Ultimately this chapter aims to provide insight into white feminist theories about race and Black men that drove their reactionary politics and insidious claims about Black male deviance from the 1860s to the 1980s.
Against Savage (Black) Manhood: Elizabeth Cady Stanton, Complementarianism, and the Uniting of the White Race Elizabeth Cady Stanton was a racist. She believed whole heartedly that white women were more evolved than Blacks and more developed and civilized than Black men. In the winter months of 1865, Stanton was increasingly frustrated by the Republicans’ and abolitionists’ advocacy that favored giving Black men the right to vote above white women. As Corinne T. Field (2014) writes in Struggles of Equal Adulthood, “Stanton did all she could to convince Republican leaders that adulthood rather than manhood should determine political rights. To press her argument, she made a fateful choice: she decided to attack the political capacity of adult men by invoking scientific and popular theories of racial difference” (p. 133). The highlighting of Stanton’s beliefs should not distract the reader from the larger point concerning the specific ethnological theories, the racism, and the particularly misandric and dangerous views concerning the existence of racialized men as citizens with political rights. One must not be distracted by the historical content of this work. This chapter is not about the individual racism of Elizabeth Cady Stanton, Susan B. Anthony, or Phoebe Couzins, but rather this research concerns itself with how the individual and collective sentiments the nineteenth-century suffragettes expressed reflected the specific ideological paradigms and political goals of white feminists at the turn of the century. Louise Newman (2007) has been unwavering in her position that “feminism, assimilation, and imperialism were historical siblings, the offspring of a marriage between democratic ideals and social evolutionary beliefs: equal citizenship was possible only for those who conformed to the racialized and gendered precepts of (white) civilization” (p. 173). These biases were reflected in the leadership, member, and the overarching goals of the feminism in the late nineteenth century and throughout the twentieth century that are rarely situated as deliberate dedications of feminism as an idea. As the historian Jen McDaneld (2013) explains:
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
873
Works rarely explore or explicate white suffragist racism— they only confirm it. As with our acquaintance with Stanton herself, this racism is now notorious even as it remains obscure. Scholars persistently produce this racism as the reason for the so- called split within the suffrage movement over whether to support or oppose the Fifteenth Amendment and the enfranchisement of black men before women, but in doing so they make that racism into an explanation for an event, a way to understand the fissures and failures of the movement. (p. 244)
Current research into suffragism makes racism seem necessary and natural among white suffragettes, a defensive and reactionary response to the expansion of rights for Black men, rather than a political or ideological program of conscious choosing. Ironically, while suffragette writings and speeches often specifically name and target Black men as the predominate threats to women’s rights and white civilization, few if any writings on the racism of the suffrage movement addresses the misandric vitriol of the women’s rights movement toward Black males generally (Curry 2021b). The previous research by the historian Michelle Mitchell, specifically her article entitled “Lower Orders, Racial Hierarchies, and Rights Rhetoric: Evolutionary Echoes in Elizabeth Cady Stanton’s Thought During the Late 1860s,” is exceptional in showing the dangerous racialist thinking of white suffragettes such as Elizabeth Cady Stanton. Mitchell does not see Stanton’s ideas a blameworthy of future discrimination against Black men and women. Mitchell (2007) argues the conclusion of the aforementioned article that “Elizabeth Cady Stanton can hardly be blamed for the active, frequently ruthless discrimination faced by black women and men-along with Asian Americans, Latina/os, Native Americans, and immigrants who are now occasionally referred to as ‘white ethnics’—that had a devastating impact on citizenship rights and continued well into the twentieth century” (p. 146). I am however of a different opinion. The violence of feminism – its pursuit of imperialism, its violence and criminalization of Black men and other racialized males, its demand for segregationist logics to protect womanhood – has not yet been popularly pursued by Black theorists in such a way that it has become part of dominant historiography, despite Black males being the group specifically targeted by white suffragettes during the debates over the 15th Amendment and by women’s rights groups who feared engaging a public filled of free Black men, who were thought to be rapists. These ideas not only mirrored the ethnological thinking of the day but sought to advance such racial theories as public opinion and social consciousness on matters not only tied to race, but racial manhood. If scholars accept, as they often do, that nineteenth-century feminism was a transformative of the social consciousness of its day, then we must accept that its racism was not solely isolated to the minds of its architects and leaders or a contingent political stratagem as maintained by Kraditor (1965). When Elizabeth Cady Stanton decided to abandon the plight of Blacks and side with Democrats, under the leadership of proslavery advocate George Francis Train, she made a deliberate and calculated decision to demonize Black men as threats to womanhood and exclude Black women from the ballot through educated suffrage. By 1867, Elizabeth Cady Stanton and Susan B. Anthony had begun advocating for educated suffrage, a position that would be adopted as part of the National Women’s
874
T. J. Curry
Suffrage Association’s actual platform in 1903 (Terborg-Penn 2002, p. 71). Educated suffrage was a damaging discourse aimed at convincing a racial American public that Black women were too uneducated and ignorant to desire the ballot. As historian Rosalyn Terborg Penn explains: the educated suffrage argument was elitist, because advocates believed that a majority of foreign-born, poor, and Afro-American women would be excluded from the franchise. Consistent with the discriminatory nature of this woman suffrage view, the majority of white, middle-class suffragists abandoned the mid-nineteenth century cry for universal suffrage. As a result of the racism inherent in this position, the policies of the national woman suffrage leaders during the twentieth century either justified the enfranchisement of white women only, or took a states-rights position which, in essence, would have excluded black women in much of the South. (Terborg-Penn 1977, p. 16–17)
Black women were excluded from the suffrage movement, while Black men were being constructed as political threats that must be controlled by death. The actions of Stanton were remarkably consistent with the theories of racial inferiority held by George Francis Train. Train did not believe the Negro was civilized enough for freedom. In a speech entitled “Slavery and Emancipation,” Train clearly articulated his views on the innate inferiority and immorality of the African race. “The African has no social ties, no sacred rights, no family pleasures, and is a cannibal,” said Train. Slavery brought Africans to church, to God. Africa had no Christian ministers or schoolteachers, so there is no way to expect improvement and development. The ignorance of the African had become genetic. Train was adamant that “Religiously and morally, all the heads under which I have classified the arguments are subordinate to this—the barbarian meets civilized man and improves as far as he can. Education may develop, but cannot originate mind. Color is not the only thing that marks him. You must first put inside his thick skull nine cubic inches more of brain!” (Train 1862, p. 22). “So far as America is concerned—slavery, if not a curse to the white man, is, compared to his native habits, a blessing to the slave. America is no more a hell to the negro, than Africa is a heaven,” writes Train (1860, p. 22). In “Manhood Suffrage,” Stanton does object to what she perceives as the tendencies of male government – tendencies that have come to be in our present day synonymous to and with patriarchy. Stanton (1868b) believed that “the male element is a destructive force; stern, selfish, aggrandizing; loving war, violence, conquest, acquisition; breeding discord, disorder, disease and death.” Slavery and war emerged because of the excess of the masculine element. Stanton was interested in the balance white womanhood had to white manhood: Mid violence and disturbance in the natural world, we see a constant effort to maintain an equilibrium of forces. Nature, like a loving mother, is ever trying to keep land and sea, mountain and valley, each in their place, to hush the angry wave? and winds, balance the extremes of heat and cold of rain and drought, that harmony and beauty may reign supreme. There is ever a striking analogy in the world of matter and mind, and the present disorganization of our social state warns us that in the dethronement of woman we have lot loose the elements of violence and ruin, that she, only, has power to curb. (Stanton 1868b)
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
875
“If the civilization of the age calls for an extension of suffrage, a government of the most virtuous, educated men and women would better represent the whole humanitarian idea, and more perfectly protect the interests of all, than could representation of either sex alone,” says Stanton (1868b). Eerily absent from Stanton’s urging for suffrage is a call for equality. Stanton (1868b) is reminding her readers that a “man’s government” is much worse and more unjust than a white man’s government, “because in proportion as you multiply the tyrants, you make the conditions of the subjects more hopeless and degraded.” Stanton is not against white men’s government, or what we would now understand as patriarchy, but she is against being ruled by men of the darker races, men with a degraded and inferior manhood. She argues this throughout her reflections on male suffrage. “If woman finds it hard to bear the oppressive laws of a few Saxon Fathers, of the best orders of manhood, what may she not be called to endure when all the lower orders, natives and foreigners, Dutch, Irish, Chinese, and African, legislate for her daughters” (Stanton 1868b). So when Stanton (1868b) asks readers to “Think of Patrick and Sambo and Hans and Yung Tung who do not know the difference between a Monarchy or Republic, who never read the Declaration of Independence or Webster’s spelling book making laws for Lydia Maria Childs, Lucretia Mott, or Fanny Kemble,” she is asking them to imagine superior and educated white women being ruled by savage men, who possess more severe brutish natures. Empowering savage males politically was against the laws of science. Stanton (1868b) argued that “All late writers on the science of government recognize in woman the great humanizing element of the new era we are now entering, in which moral power is to govern brute force.” Again appealing to the complementarian nature of white educated womanhood, Stanton urges white men to allow them to join in governance to subdue these harsher expressions of manhood. “The need of this hour is not territory, gold mines, railroads, or specie payments but a new evangel of womanhood, to exalt purity, virtue, morality, true religion, to lift man into the higher realms of thought and action,” writes Stanton (1868a). The “lower orders” of men threaten white womanhood. They are not fit so to speak to rule over a more civilized white racial group or educated class. Stanton is reminding her readers, and her Democratic allies, that the enfranchisement of Black men is merely the first step to male governance by inferiors – savage men. Racialized men are sexist and misogynist toward women. According to Stanton, “The lowest classes of men are invariably the most hostile to the elevation of woman as they have known her only in ignorance and degradation and ever regarded her in light of a slave” (1868). Implying that the conditions of women under Black and non-white, read uncivilized, cultures are worse than the actual slavery of whites is not unlike previous comments made by Stanton trying to convince women of the ever looming dangers associated with the Black male vote. It was not so long ago in 1865 that Stanton asked “Have not Black male citizens been heard to say they doubted the wisdom of extending the right of suffrage to women” (1865). Because the racial backwardness of Black men is brutish, Stanton infers that the political empowering of Black men would be another form of slavery for Black women who were still actually enslaved. “If the two millions of Southern Black women are not secured in
876
T. J. Curry
their rights of person, property, wages, and children, their emancipation is but another form of slavery. In fact, it is better to be the slave of an educated white man, than of a degraded, ignorant Black one” (1865). Similar to her previous misandric renderings of Black men, Stanton again restates the idea previously presented in her letters on the pages of the Revolution as an equivocation between the institution of American chattel slavery and the marriage of a Black woman to a Black man. In her testimony over the 15th Amendment, Phoebe Couzins, the first female US Marshal and an early member of the National Women’s Suffrage Association, reiterates Stanton’s views of Black men. She says: I have had opportunities of seeing and knowing the condition of both sexes, and will bear my testimony, that the black women are, and always have been, in a far worse condition than the men. As a class, they are better, and more intelligent than the men, yet they have been subjected to greater brutalities, while compelled to perform exactly the same labor as men toiling by their side in the fields. . . The black men, as a class, are very tyrannical in their families; they have learned the lesson of brute force but too well, and as the, marriage law allows the husband entire control over his wife’s earnings and her children, she is in worse bondage than before; because in many cases the task of providing for helpless children and an idle, lazy, husband, is imposed on the patient wife and mother; and, with this sudden elevation to citizenship, which the mass of stupid, ignorant negroes look upon as entitling them to great honor, I regard the future state of the negro woman, without the ballot in- her hand, as deplorable. And what is said of the ignorant black man can as truthfully be said of the ignorant white man; they all regard woman as an inferior being. (Couzins 1881, p. 387)
There are several instructive moments in Couzin’s statement concerning Black men. For instance, while she explains that Black women are better off as a class insofar as Black women are thought to be more intelligent and just as hard working, the position of Black women as a group is endangered because of the dangers and violence threatened by Black men. Black men are tyrannical and violent in their families according to Couzins, because they were brutalized in slavery somehow that violence is how they attempt to rule within their homes. Again reiterating the relationship between chattel slavery and marriage, Couzins suggests that like slave masters, Black men rob Black women of their earnings. Black women are then forced to take care of helpless children and lazy husbands. Like much of the intersectional discourse of our present day, Black women are framed as being burdened by racism and sexism; however, an important distinction is being raised. Because there was no idea of women as oppressed as simply because of their sex and patriarchy, womanhood was framed as that of mother often within the sphere of the home. In this case, the Black woman is oppressed because of her race and the abuse she receives at the hands of Black men. There is great similarity between the aforementioned statements and Sojourner Truth’s often celebrated statement made at the American Equal Rights Association inaugural meeting in 1867, which said: There is a great stir about colored men getting their rights, but not a word about the colored women; and if colored men get their rights, and not colored women theirs, you see the colored men will be masters over the women, and it will be just as bad as it was before. So I
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
877
am for keeping the thing going while things are stirring; because if we wait till it is still, it will take a great while to got it going again. White women are a great deal smarter, and know more than colored women, while colored women do not know scarcely anything. They go out washing, which is about as high as a colored woman gets, and their men go about idle. (Truth 1867)
While Paula Giddings (1985) has suggested that Truth was a victim of the brutish conditions of slavery and took on both the misrepresentations of Black men and the supposed inferiority of Black women to white women, what if her statement says something more fundamental about the relationship of feminism to Black men? Could Truth not in fact articulating the script of the movement in fact see herself very much a part of? When Elise Johnson McDougald (1925) writes that “In this matter of sex equality, Negro women have contributed few outstanding militants, a notable instance being the historic Sojourner Truth. On the whole the Negro woman’s feminist efforts are directed chiefly toward the realization of the equality of the races, the sex struggle assuming the subordinate place” (pp. 380–381), she is describing the fundamental break between Truth’s prioritization of the sex struggle and the larger program of racial advancement Black men and women had in not siding with white suffragists who sought to demonize Black men and by the effect the whole Black race. Convincing the world of Black men’s antagonism to women’s rights was part of feminism’s racist program. As Rosalyn Terborg Penn (1977) explains, “Despite the evidence that many blacks opposed anti-woman suffrage arguments, there is a popular view which says that blacks have historically opposed woman suffrage” (p. 256). Part of the investment in reacting to Black male emancipation was to show that Black men were less forward thinking than white men and ultimately spelled doom for the republic. In “Women and Black Men,” Stanton reiterated that in the vote for Black men, “manhood suffrage not only rouses woman’s prejudice against the negro, but on the other hand his contempt and hostility towards her as an equal” (Stanton 1869). Restating Lucretia Mott’s reports of two Black men, Mr. Downing and Mr. Purvis, Stanton implored her readers to see the misogyny of Black men toward women. “Mr. Downing, a colored man. . .said that in his opinion Nature intended that the male should dominate over the female everywhere. . .[and] young Dr. Purvis remarked that woman was the Black man’s worst enemy” (Stanton 1869). Because “this is the feeling among the majority of all colored men,” white women were asked to think about and fear the “the enfranchisement of Africans, Chinese, and all the ignorant foreigners the moment they touch our shores” (Stanton 1869). Terborg Penn (1977) explained that “Negative reports about black male voters became consistent with the rhetoric of white suffrage leaders after the post-Civil War years, especially among southern suffragists who expediently attacked the black male electorate for failing to support the woman’s movement” (p. 260). Efforts to depict Black men as threats to the unifying force of white supremacy politically and the intimate lives of women, both Black and white, were popular throughout the South despite the growing prevalence of educated suffrage and its call to exclude Black women from the ballot (Jones 2020).
878
T. J. Curry
Louise Newman (1999) has previously written that “white women’s expressions of resentment over the enfranchisement of black men and these women’s subsequent decision to keep the movement clear of ‘race’ questions were part of a larger postReconstruction retreat from support of racial justice” (p. 5); however, this view does not take into full view the call for violence made to enforce Black male disenfranchisement. At the turn of the nineteenth century, white feminists demanded white men to protect the racial order of the South through violence against Black men. Rebecca Latimer Felton, a Georgian feminist and suffragist, argued that giving Black males the right to vote gave Black men a false sense of manhood that would permit them to rape white women. In her 1898 speech to the Georgia Agricultural Society, she explained to the audience of several 100 members that: when you take the negro into your embraces on election day to control his vote and use liquor to befuddle his understanding and make him believe he is a man and your brother. . .lynchings prevail, because the cause will grow and increase with every election when there is not enough religion in the pulpit to organize a crusade against this sin, nor justice in the court-house to promptly punish the crime, nor manhood enough in the nation to put a sheltering arm about innocence and virtue. If it requires lynching to protect woman’s dearest possession from ravening, drunken human beasts, then I say lynch a thousand negroes a week, if it is necessary. (Felton 1898, p. 625)
This newly enfranchised Black man threatened the future of the American republic. His freedom thwarted the reign of the white race and endangered the safety of the white woman. Black men were believed to be beasts, and rape according to Felton was “the brutal lust of these half-civilized gorillas” (Litwack 1998, p. 213). The brutish nature of Black men not only was used to justify the lynching of Black men in the United States but also justified the imperial interventions into the darker world to save primitive women from the violence savage men. The influence of social Darwinism among white feminists and ethnologists presumed a racist evolutionary schema. The position of a civilization was often assessed by the status of its women, so it is not surprising that “one justification for Western colonialism was formulated in terms of protecting primitive women from various forms of social, economic, and sexual mistreatment. For over a century, Westerners had presumed that primitive women were overworked, sexually abused, or otherwise badly treated by men of their cultures” (Newman 1999, p. 160). While women from the darker races were not thought to be evolved women who could appreciate the need for homes and the evolutionary import of patriarchal order, it was the belief of suffragettes that primitive women need help from the civilized races to fend off the attacks of their savage male counterparts (Newman 1999; Bederman 1995). The presumed savagery of non-white males legitimized white feminists’ imperial and colonial projects as a progressive endeavor to defend women’s rights and made their presence within burgeoning colonies essential (Jacobs 2009). “Primitive races, lacking the biological capacity to develop racially advanced traits like manliness of character, would require many generations to slowly acquire manliness and pass these civilized capacities on to their offspring” (Bederman 1995, p. 29). Consequently, there was no hope for patriarchal manhood among savage races which brought with it values of
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
879
chivalry, sexual restraint, and paternalism to protect the sanctity of womanhood. The salaciousness of primitive man was insatiable and required intervention to introduce civility between men and women outside of Europe.
The Evolutionary Origin of Patriarchy is Feminine – Gilman’s Claiming of White Supremacy as the Gift of White Womanhood The nineteenth-century feminist Charlotte Gilman (1860–1935) is most widely recognized for her short story “The Yellow Wall-Paper” published in The New England Magazine in 1892. Gilman is often celebrated as a feminist novelist and activist noted for her reflections on the repressive nature of the home and the grave psychological and existential harm caused by women’s traditional roles in society. In Women and Economics: A Study of the Economic Relation Between Men and Women as a Factor in Social Evolution (1898), Gilman seeks to examine an objective truth of racial and sexual evolution in civilized peoples. Specifically, she aims to understand the role that women play as the authors of white supremacy and the creators of men. As she explains, Women and Economics purpose is to “reach in especial the thinking women of today, and urge upon them a new sense, not only of their social responsibility as individuals, but of their measureless racial importance as makers of men” (Gilman 1898, vii). The almost overbearing tenor of Gilman’s white supremacy may be somewhat off-putting to contemporary twenty-first-century readers; however, it must be emphasized that Women and Economics was praised by women and feminists in her own time as a monumental success. The historian Mary A. Hill (1980) explains that “It was Charlotte Gilman’s single most important work, the beginning of her rapid climb toward notoriety as one of the major feminist theorists in turn of the century America. Jane Addams of Hull House thought Women and Economics a ‘masterpiece.’ Her coworker Florence Kelley believed it was ‘the first real substantial contribution made by a woman to the science of economics” (p. 295). Gilman’s vision for white women in Women and Economics earned her recognition as the “brains of the women’s movement” in the eyes of her peers (Davis 2005, p. 1). Rosita Schwimmer (1933) wrote a short editorial review in the New York Times proclaiming Women and Economics to be “. . .considered by feminists of the whole world as the outstanding book on feminism, [and]. . . has been accepted all over the world as a standard work on the movement.” The popularization of evolutionary theory among suffragettes energized new theorizations of the role that femininity and white womanhood played in the global supremacy of the white race. Instead of suggesting that white femininity was the natural complement to the male element or nature’s attempt to create harmony within civilized races, as previous suffragettes such as Elizabeth Cady Stanton or Susan B. Anthony did, Gilman argued that only white womanhood could evolve white men toward patriarchy and develop within in a psychology of white supremacy that would allow him to care for all white people around the world as one of his own. Gilman wrote Women and Economics to fundamentally alter the minds of the women of her day and “urge upon them a new sense, not only of their social responsibility as
880
T. J. Curry
individuals but of their measureless racial importance as makers of men” (1898, p. vii). The development of the human species is both constrained and enabled by the laws of nature. Gilman believed despite the best efforts of individuals against their environment, “it remains true that the human creature is affected by his environment, as is every other living thing” (1898, p. 1). The human species mirrors the animal kingdom and its hierarchies of sexual reproduction and survival of the fittest with one exception – the dependency of the female on the male for sustenance. Gilman believed that female dependency was the exceptional characteristic of the human species. It is important to note however that for Gilman, the humans capable of creating the infrastructures (e.g., the home) necessary to create female dependency were solely comprised of the white races. As Gilman explained, “the only animal species in which the female depends on the male for food, the only animal species in which the sex-relation is also an economic relation. With us an entire sex lives in a relation of economic dependence upon the other sex, and the economic relation is combined with the sex-relation” (p. 5). The sexual economic relationship between white men and white women was not an exploitative one; it was a stage of racial evolution among white races that placed them above the savage races. Economic production is a masculine endeavor that benefits white women because it enriches the race (p. 8). The woman benefits the race through motherhood. Among the primitive races, “motherhood bears no relation to their economic status” (p. 16). The nuclear family of white civilization is built upon the presumed moral rightness of monogamy among civilized peoples. “All the way up, from the promiscuous horde of savages, with their miscellaneous matings, to the lifelong devotion of romantic love, social life has been evolving a type of sex-union best suited to develop and improve the individual and the race” (p. 26). The sexual and economic relationship between white men and white women was an expression of their racial capacity as civilized rather than savage. Like early suffragettes, Gilman supported the ethnological sciences of the time. For several decades, there is the idea that patriarchy or the elevation of the masculine over the feminine was proof of a racial group’s evolutionary stage. Racial superiority was indicated by the difference between the sexes. The inequality of white men and white women, or more accurately the different roles white men and women assume to build white supremacy, was the necessary for civilization. Unlike non-white (savage) races, who were thought to be nomadic and promiscuous, the white race evolved the capacity to build homes and value motherhood. Gilman (1898) explained: Developed and increased by use, the distinction of sex increased in the evolution of species. As the distinction increased, the attraction increased, until we have in all the higher races two markedly different sexes, strongly drawn together by the attraction of sex, and fulfilling their use in the reproduction of species. These are the natural features of sex-distinction and sex-union, and they are found in the human. (p. 29)
Gilman argues that the need for women’s liberation is not based in the oppression of women within the home but that the home has become a relic that makes the white
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
881
woman oversexed and excessively feminine. The maternal instinct grew and developed in the home since it was the order of propagation ordained by nature. For the race to evolve, the white woman was made to depend on the white man economically. The white woman, unlike her Black counterpart who could not blush (Van Evrie 1863, p. 89) or seduce primitive men driven solely by sexual instincts, used her femininity and the pleasure of sex to create affection from white men. This affection and love earned by the white woman created a family unit where motherhood and fatherhood have a purpose for the propagation of the race. The savage man and woman are driven to have sex without consciousness of the larger racial goals reproduction serves toward the enhancement of civilization. The home became a mode of racial propagation for centuries that not only increased the numerical superiority of white population but evolved the white male toward the maternal instinct of love. The white race’s evolutionary advantage allowed patriarchal sex roles to dominate their social and economic life. The demands of physical labor and industry required the male to provide for the home and care for the woman and child, but which was not his sole endeavor. The white man had various roles and racial duties to care for throughout society, while the white woman was limited to the regeneration of the white race. By accepting her maternal role in the home, the white woman became a weaker race worker. As Gilman explains, “the woman, specially modified to sex and denied racial activity, pours her whole life into her love” (p. 48) and neglects the now present duty to rule the white world alongside white men. The racial superiority of the white man required him to be an economic thinker and individualistic competitor within the world. The world needed to be remade for the white race, and his industrious spirit required his rugged manliness to conquer savage lands. The capacity and instinct to care for the (white) race throughout history “has been almost entirely a female function, sometime absolutely so,” writes Gilman (p. 131). This made white men inadequate white supremacists, since white men could compete to see who among them might be the fittest, but do not understand the need for cooperation among all whites to protect the legacy and civilizational triumph of the white race globally. For white men to learn to care for all whites as he would for the mother or child of his family, white women had to give white men an object of care. The home became an institution throughout which white men could develop the capacity to love his race which is against his competitive masculine instincts. “Maternal energy is the force through which have come into the world both love and industry” (p. 126), writes Gilman. The white man evolves the capacity to love the white race through his responsibility to care for his wife and children, so the “subjection of woman has involved to an enormous degree the maternalizing of man” (p. 126). The relationship that bound the white man to the white woman and child in the home changed his individualistic and competitive nature by developing within him the capacity for maternal energy. This allowed him to be both man and mother – both patriarchal leader and racial caretaker – “new functions [previously] impossible to male energy alone” (p. 126). The white male, acting through his natural instincts, steadily encroached upon the freedom of the female until she was reduced to the state of
882
T. J. Curry
economic dependence, thereby assumed the position of provider for this creature no longer able to provide for herself. He was not only compelled to serve her needs but to fulfill in his own person the thwarted uses of maternity. This bond created a white man that “was made part mother; and so both man and woman were enabled to become human. It was an essential step in our racial progress, a means to an end” (p. 128). Women’s subjugation to (white) men was not thought to be oppressive throughout Gilman’s 1898 treatise. She believed that it was the white male who is sacrificing himself for the betterment of the white race. “It should not be considered as an extreme maternal sacrifice, but as a novel and thorough system of paternal sacrifice, the male of genus homo coerced by sex-necessity into the expression of maternal energy” (p. 128). The transubstantiation of the white male from man to man-mother allowed white men to express the maternal element necessary to care for the white race as he does himself. The expression of maternal energy completed his evolution to a complete patriarch that contributed to the propagation of the white race not only through the biological act of sex but through the spiritual and psychological attributes of white racial domination. In this sense, the white male is elevated from his lowly biological contribution to racial reproduction to an equal standing with white women. Gilman explains that: sexual equality has been slowly evolved, not only by increasing the importance of the male element in reproduction, but by developing race-qualities in the male, so long merely a reproductive agent. The last step of this process has been the elevation of the male of genus homo to full racial equality with the female, and this has involved her temporary subjection. Both her physical and psychical tendencies have been transplanted into the organism of the male. He has been made the working mother of the world. (p. 132).
Women’s subjugation was not oppression, but a transitory phase that enabled the white race to dominate global economies and foreign lands and continue their quest for white racial superiority. Gilman believed the patriarch is perfected by the labor of woman within the home, her care for others. To dominate other savage races, white supremacy needed cooperation and collaboration between white groups. This feminine element of racial superiority was cultivated by the sacrifice of white women within the home. To colonize the world, “it was essential that the expansive forces of masculine energy be combined with the preservative and constructive forces of feminine energy. The expansive and variable male energy, struggling under its new necessity for constructive labor, has caused that labor to vary and progress more than it would have done in feminine hands alone” (p. 132). The white woman could not conquer the world. Colonization was a male element that the author of nature decided she could not possess. Her contribution to the great arc of civilization was to evolve the white patriarch beyond himself and his immediate nuclear family. The maternal element required white races to realize the great truth of history was that races are families. For the white race to rule the world, “Women can well afford their period of subjection for the sake of a conquered world, a civilized man” (p. 134). This aspect
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
883
of Gilman’s evolutionary thinking has largely been ignored by feminist authors who focus their analysis on Gilman’s wish for women to escape the home. Gilman wanted to escape the home because white women had evolved beyond it. In her view, white women were warrior for the race and should rule over the darker races alongside white men in the world made by patriarchy. While the subjugation of women within the home was no longer desirable given the stage of racial evolution achieved by the white race, Gilman was clear that the greatest victims of the white race’s march toward civilization were the men of the darker races. She reminds white women that women’s liberation should be understood as their success in cultivating the capacity for global white supremacy, not as an escape from their oppression as women: When the centuries of slavery and dishonor, of torture and death, of biting injustice and slow, suffocating repression, seem long to women, let them remember the geologic ages, the millions and millions of years when puny, pygmy, parasitic males struggled for existence, and were used or not, as it happened, like a half-tried patent medicine. (p. 134)
The march toward white supremacy required sacrifice on behalf of white men and women so that they can enjoy economic and political dominance. Gilman conceptualized white races as a global family unit where white racial groups around the world lived in a world created by the white patriarch. The white patriarch was the creation of the white woman’s maternal instincts, so the white woman entrusts herself to men because it was her womb that perfected him to be the white patriarch. The evolutionary claims outlined by Gilman defined the emergent feminist ideology of the twentieth century. The racial superiority of the white race rightly dictated the order and borders of civil society throughout Gilman’s writings. In a lesser known essay, “A Suggestion on the Negro Problem,” Gilman puts forth the idea that Black social integration should be based on the ability of Black people to reach a proximal relationship to white culture. The Negro cannot simply achieve the accomplishment of the white races, since the “transfusion of blood is a simple matter compared with the transfusion of civilization”; however, some Negroes have shown that with contact with whites, they could slowly improve themselves (Gilman 1908, p. 78). Unfortunately, according to Gilman, while the elevation of some Negroes did prove the ability of lesser races to assimilate, this was an incredibly difficult and arduously slow process of racial evolution. The Negro was an American problem, because his “unavoidable presence of a large body of aliens, of a race widely dissimilar and in many respects inferior, whose present status is to us a social injury” (Gilman 1908, p. 78). The backwardness of the Black race produced a racial disposition for deviance and crime in Gilman’s view, so arresting this devolution of America’s civil society caused by Black emancipation is of special concern. To remedy the Negro problem, Gilman calls for an honorific re-enslavement of the Black race – a program she calls enlistment. Gilman proposes to: Let each sovereign state carefully organize in every county and township an enlisted body of all negroes below a certain grade of citizenship. Those above it -the decent, self-supporting,
884
T. J. Curry
progressive negroes-form no problem and call for nothing but congratulation. But the whole body of negroes who do not progress, who are not self-supporting, who are degenerating into an increasing percentage of social burdens or actual criminals, should be taken hold of by the state. (1908, p. 80–81)
This conscription of lesser Blacks was meant to slow the degeneration of the Black race toward savagery. Gilman’s proposal was to confine Black people who showed lesser potential for social evolution to menial labor. This compulsory program was thought to educate lower Blacks into becoming productive members of the Black race and more tolerable citizens within a civil society. “By the same methods in which a state or county arbitrarily provides for its poor, its defectives, or for the education of its children; so it could now bestir itself to provide for this large class of comparatively backward citizens,” writes Gilman (1908, p. 84). Gilman believed the enlistment of Black people into menial jobs would hold a special attraction to the Negro race, if it was offered to them as an honor. “The whole system should involve fullest understanding of the special characteristics of the negro; should be full of light and color; of rhythm and music; of careful organization and honorable recognition” (p. 83). The colors and music were thought to appeal to the cherub nature of the unevolved Negro so much so that Gilman insisted “If the arrangement were made very clear and visibly attractive, and volunteers were called for, with some special honor and recognition for them, it is quite possible that numbers would enlist of their own accord” (p. 84). Women’s rights at the dawn of the twentieth century were predominately framed by the scientific presumptions of evolutionary theory. Even the right to the vote was discussed by suffragists as a macroevolutionary process that drove white society to a higher plane of civilization. Carrie Chapman Catt, the president of the NAWSA following Elizabeth Cady Stanton, argued for white women’s suffrage by appealing to the arguments that white women had evolved far beyond the lower races and stood upon the same plane as white men (Amidon 2007). Catt, like Gilman, saw women’s liberation as a driver of white civilization. In an editorial entitled “Mrs. Catt on Feminism,” Catt argued that “feminism is a universal movement due to natural and world-old feeling” (Catt 1917, p. 12). Catt was adamant that feminism was a “worldwide revolt against all artificial barriers which laws and customs interpose between women and human freedom” that valued (white) women as coworkers, “the comrades of the men and of her family” (Catt 1917, p. 12). Feminism was not merely a political cause, but “an evolution like the enlightenment and democracy” (Catt 1917, p. 12). Like many of the feminists of her day, Catt believed the suffrage right strengthened white supremacy and empowered the white race to overcome any resistance by Black voters, even with the support of newly enfranchised Black women. In her contribution to Women Suffrage by Federal Amendment (1917), Carrie Catt addresses the worry that suffrage rights would impede white supremacy. Using census data from 18 states, she explains that “white supremacy will be strengthened, not weakened, by woman suffrage” (p. 76). Catt did not only claim that the demographic majority of voting age (over the age of 21) white women outnumbered
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
885
Black women but insisted the ongoing discrimination against Black men throughout the South and assured discrimination that would limit Black women’s vote would protect white supremacy for the foreseeable future. “In South Carolina, voters must read, own and pay taxes on $300 worth of property. In Mississippi, voters must read the Constitution. The other four states of the ‘black belt’—Georgia, Florida, Alabama and Louisiana—impose an educational test. Women voters would be compelled to submit to the same qualifications. In the other nine states white women exceed the total negro population. Woman suffrage in the South would so vastly increase the white vote that it would guarantee white supremacy if it otherwise stood in danger of overthrow,” writes Catt (1917, p. 76–77). Catt was adamant that the Black vote was not to be feared and in the words of Chief Justice Walter E. Clark, “the votes of 260,000 white women can be relied on to stand solid against any measure or any man who proposes to question Anglo-Saxon supremacy” (Catt 1917, p. 79). In fact, she insisted that “If the South really wants White Supremacy, it will urge the enfranchisement of women” (Catt 1917, p. 77).
The Racial Proxemic: [W]hite Womanhood, Feminism, and the Dangers of Black Men to White Civilization The empowering of inferior/savage/raced males intensifies the violent element of manhood and the efforts of feminism sought to advance (white) women’s liberation within an ever-expanding medium of anti-Black and misandric phobics, which reduced Black men not only to caricatures and stereotypes but impeding and dangerous entities that were inevitably going to spell doom for the republic. The American South was seen as the pinnacle of white Anglo-Saxon civilization. Like many of the colonies the English set up around the world, perhaps even as locally as Jamaica, Americans sought to continue the legacy of colonization and rule as part of their Anglo-Saxon heritage (Curry 2018). The idea of the colonial heritage was often mixed with the question and position of women in society, since they were ultimately the transmitters of racial legacy. Suffragettes asserted that it was the white woman who held the key to the people and the fate of the white race in the United States. Throughout the American South, white women valorized their Anglo-Saxon heritage and birthright to racial superiority. Belle Kearney was a Mississippian feminist and officer of the National American Women’s Suffrage Association (NAWSA). In an address to the NAWSA, Kearney said: The same blood was in the veins of the men and women who founded those new commonwealths as flowed in the veins of Washington, Jefferson and Madison; the same as that which coursed through the veins of the heroes of Lexington, Concord and Blinker Hill; the same dominant blood that has coursed through the earth, and will rule through the ages. The people of the South have remained true to their royal inheritance. Today the Anglo-Saxon triumphs in them more completely than in the inhabitants of any portion of the United States—the Anglosaxon blood, the Anglo-Saxon ideals, continue the precious treasure of 2000 years of effort and aspiration. (Kearney 1903, p. 106)
886
T. J. Curry
Women’s rights, specifically white women’s right to the ballot, was not about equality but the ability of the white race to exercise unbridled power and authority over the Black race. Kearney (1903) explained that: The enfranchisement of women would insure immediate and durable white supremacy, honestly attained; for, upon unquestionable authority, it is stated that in every Southern State but one. There are more educated women than all the illiterate voters, white and black, native and foreign, combined. As you probably know, of all the women in the South who can read and write, ten out of every eleven are white. When it comes to the proportion of property between the races, that of the white outweighs that of the black immeasurably. The South is slow to grasp the great fact that the enfranchisement of women would settle the race question in politics. (Kearney 1903, p. 107)
Kearney explained that Black men’s freedom threatened the spiritual unity of America as a white republic. The foundations of white supremacy depended on the preservation of Anglo-Saxon heritage and the regeneration of white civilization against the recently emancipated Black male. “The white people of the North and South are children of the same heroic souls who laid the foundations of civil and religious liberty in this new world, and built thereon this great Republic,” writes Kearney (1903, p. 107). The division of North and South is meaningless given the newly emerging threat of freed Black men. This call for national racial unity was prefigured by the need to ensure that America was a “mighty country safe for the habitation of Anglo-Saxon” (Kearney 1903, p. 107). “Thank God the black man was freed! I wish for him all possible happiness and all possible progress, but not in encroachments upon the holy of holies of the Anglo-Saxon race,” says Kearney (1903, p. 107). The sexual threat Black men were said to pose to white women understood as a need for white racial unity to protect white civilization against the pestilence of the Black freedman (Stein 2012). Popular periodicals such as Good Housekeeping sometimes included editorials of professional white women commenting on the ever-present threat of the Negro man. One such article entitled “The White Woman and the Negro Man” by Ellen B. Ligon insisted that the burden of Black enfranchisement is shouldered by the white woman. The threat of developing an “alien black race, but one century removed from barbarism,” was that it could jeopardize the white races’ “sacred responsibility—the preservation of an untainted white civilization developed to it utmost possibility” (Ligon 1903, p. 426). The Negro was misled in believing they could be the equals of whites. In fact, writes Ligon, “the greatest wrong ever done civilization was when thousands of beings but two degrees removed from naked barbarism, were declared by law, and taught that they were the social equals of a race which represents the refinement, the chivalry, the bravery of thousands of years of civilization” (p. 427). Black male enfranchisement and the quest for Black social equality were framed as a threat to white women’s future liberties. The freed Black man, incapable of controlling his insatiable lust for white women, uses rape as a weapon to exact vengeance against the white race. Ligon urged the white world to hear her pleas: “As Let all the world listen while the south calls on you to hear: The white woman is the coveted desire of the negro
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
887
man. The despoiling of the white woman is his chosen vengeance. The white woman must be saved!” (p. 428). Throughout the Southern states, white women articulated women’s suffrage as realizable only when the Black male was controlled and managed through violence. The struggle for women’s rights was asserted to be a great racial cause that allowed the white race to maintain control over the alien Black population in America through lynching and segregation. The white women, her physical body and political aspirations outside the home, came to define the borders of white geopolitics. The white women, corporeally and conceptually, came to define the proxemics of white violence. When Black men, who were no longer bound by shackles, physically or politically encroached too close to the white woman, violence and repression was triggered to reestablish racial order. This proxemical dynamic became even more evident through white women’s activism against Black civil rights in the 1920s forward. The focus on white women’s bodies, homes, and family set up a geopolitical border, conceptual and cultural boundaries, that became armed and aggressively protected by white men and women alike. As the historian Crystal Feimster (2009) argues: . . .white women’s participation as actors and audience at lynchings had opened up a wider space for them in public and commercial life. If they could lynch they could do anything. They had faced their worst fears. They had survived the Civil War and the violent politics of Reconstruction and Populism–coming out on the other side strong and confident in their ability to engage the racial and sexual politics of the New South. From Demanding legal protection to participating in mob violence, the new southern woman exercised all the privileges of white supremacy to demand women’s rights. (p. 155)
The use of violence against Black men became a defining feature of white womanhood at the turn of the twentieth century and energized several new political organizations to meet the now evolved status of white womanhood outside of the home (Hale 2010). The early twentieth century not only saw a federal amendment granting (white) women the right to vote as an expansion of the voting rights they enjoyed in roughly half of the states prior to the passage of the 19th Amendment (McConnaughy 2013; McCammon and Campbell 2001), but the disenfranchisement of Black voters throughout states through literacy tests and lynch mob violence. The disenfranchisement of Blacks, particularly Black men, was a platform for newly enfranchised white women who saw Jim Crow segregation as a vital policy that ensured their safety and the triumph of white civilization. To protect the spatial order of the South that confined Black males to the bottom of civil society (Dollard 1937), white women began to turn away from imperial feminism toward domestic ethnonationalism. Ironically, this was also the same time that saw white women attempt to relinquish their historical argument that they were intended by God and nature to be rulers of the darker races alongside white men and move toward a political and academic narrative asserting that they were now oppressed by white men in ways that mirrored the racist oppression of several racial minority groups like Blacks and Jews (Curry 2022).
888
T. J. Curry
In the early 1900s, white women not only began participating in male Ku Klux Klan organizations but sought to establish their own leagues to protect and enforce women’s rights. As the historian Kathleen Blee (1991) explained, “women who interpreted the struggle for women’s votes through the prism of racial, ethnic, and class privilege thus experienced an apparently easy transition from women’s suffrage to the plethora of white supremacist, nativist, and racist political movements of the early twentieth century” (p. 58). Rather than simply being passive observers of the mid-twentieth-century white male vigilantism and violence, white women orchestrated and participated in terroristic acts designed to discourage Black political participation and spread anti-immigrant sentiment. While acts of racial terrorism are classically associated with widely accepted ideas of white manhood and ethnonationalism, the mobilization of women into the 1920s’ Klan linked the racist, nationalistic zeal, which also motivated men to join the KKK, to a specific gendered notion of the preservation of family life and women’s rights (p. 67). The WKKK held strong allegiances to ideas of racial and religious purity while rejecting “messages of white female vulnerability” (Blee 2009, p. 41). Instead of vulnerability, the WKKK emphasized their “support for women’s rights and a challenge to white men’s political and economic domination” through white nationalism and an assertion of white women’s superiority to non-white races (Blee 2009, p. 41). Through the home and womanhood, the WKKK emphasized power of white women had to direct and engineer racial progress. This posture departed from the white women’s burden ideology of the early twentieth century by radicalizing the duties white women had to safeguard white racial superiority. In the postenfranchisement era, the race question turned to propagating anti-Black and pro-segregationist stances to guard the white race against the encroachment of alien (non-white) peoples. In short, the 19th Amendment empowered white women to take on more active roles to protect their racial legacy and support extralegal violence to safeguard the homes and families of the white race. The ongoing demographic pressures the white race faced from the death of white men from theaters of war, increased immigration, and the burgeoning movement for Black civil rights motivated white women to join their white men in the attempt to restore white supremacist values and politics to the fore of the national imagination. From the 1920s to the 1970s, white women supported segregationist policies, pro-Jim Crow politicians, and even petitions against UN opinions and education materials supporting desegregation (McRae 2018). While WKKK organizations were one extreme manifestation of white women’s activism around women’s rights, it was not the limit of white women’s everyday support of American segregation policies known as Jim Crow. The expansion of white women’s political landscape after the 1920s saw a multitude of various political and community-based activism that allowed white women as mothers, teachers, and journalists to continue the fight against Black civil rights. White women claimed schools, neighborhoods, and their homes as being under the providence of white womanhood. The racial proxemics established by white womanhood saw the subjugation of Black men and the separation of their white daughters from Black men as being linked to their ability to protect their families and ultimately the
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
889
white race. To do so, space and distance from the men of the Black race was seen as an integral part of white women’s exercise of liberty throughout the United States, but especially in the Southern states.
Feminist Theory’s Adoption of Subculture of Violence Criminology as a Response to Desegregation The desegregation of public institutions following Brown v. Board of Education did not arrest the proliferation of negative stereotypes or the endorsement of erroneous pseudoscientific explanations of Black male savagery by white feminist scholars or activists. Feminist organizations often acted to impede singular concessions for Black civil rights without specific provisions aimed at increasing the political opportunities for white women. The strategy of the National Women’s Party, a wealthy exclusively white feminist organization dedicated to the political and economic interests of white women (Rhode 2009, p. 35–37), wanted to see sex added to Title VII legislation despite the dangers it posed to the passing of the legislation. The National Women’s Party decided to approach Congressman Howard C. Smith. Smith was an 81-year-old conservative Democrat from Virginia who had a history of being against Black civil rights initiatives and equal pay. According to Caroline Bird’s Born Female: The High Cost of Keeping Women Down (1969), Smith’s efforts “extended the promise of equal opportunity from the seven million Negro men and women workers that Title VII was drafted to protect, to 21 million white women who did not get as high compensation for their work as the Negro members of the majority sex” (p. 3). The effect of adding sex to the antidiscrimination laws was that it ensured white women captured the majority of the jobs and employment initiatives meant to address the racist distribution of economic wealth and labor participation that historically disadvantaged Black people. The economic vulnerabilities of Black men and women were well understood by policymakers and feminist organizations at the time. In fact, many “liberals and women’s organizations in 1964 opposed adding sex to the Civil Rights Bill primarily because they did not want to endanger protection for Negroes, but also because absolute equality between sexes before the law might endanger rights and immunities favoring women” (Bird 1969, p. 7). Smith intended to sink Title VII because it was intended to protect Black civil rights and ensure Black labor participation well past the twentieth century (Bird 1997; Gold 1981). While the bill did manage to pass, the addition of sex so radically altered the scope of the legislation that it became impossible to maintain that remedying Black racial injustice was or ever could be the priority of the legislation ever again. The addition of the sex category was meant to prioritize the rights of white women over those of Blacks. Congresswoman Martha Griffith argued on the floor of Congress that without sex being added to Title VII legislation, “white women would be last at the hiring gate” (Bird 1969, p. 6). This argument was persuasive to white legislators who saw the benefit and racial advantage of safeguarding white women over Black workers be they men or women. The legal historian Michael Gold (1981) additionally observed that:
890
T. J. Curry
The other reason Congress added sex to Title VII was a fear that a black woman would enjoy greater protection than a white woman. In a legal sense, this fear was unjustified. If an employer preferred a black woman because of her race, a white woman would have had a prima facie claim for racial discrimination under the bill as it stood before sex was added to it; Congress had already decided that whites as well as blacks would be protected by Title VII. In a practical sense, however, the fear for the white woman was realistic enough. As the debate shows, the Representatives sensed that an employer, choosing between a black and a white, might lean towards the black because of a belief that the black would be more likely to sue and win than the white. (p. 468)
The addition of sex to Title VII legislation was rooted in anti-Black sentiment and the political vulnerability of Black Americans which placed their economic prosperity in the hands of white lawmakers. The effect of adding sex to Title VII was that it allowed the discrimination of white women to be placed alongside that of historically disadvantaged racial groups that have endured enslavement and genocide as equals. This small legislative change radically changed not only the employment practices but the political initiatives of hiring associated with affirmative action that allowed white women to benefit more than the Black men and women who marched and died for racial equality (Hu-DeHart 1997; Goodwin 2013; Massie 2016). By the late 1960s, the political abandonment of Black economic vitalization through civil rights initiatives had resonated throughout white academic institutions. Criminologists, sociologists, and feminists decided that the inability of poor Blacks, particularly poor Black males, to culturally integrate and economically compete with whites doomed them to subcultures of violence and mimeticism. In 1967, Wolfgang and Ferracutti developed a theoretical framework insisting American society was comprised of a dominant culture and various subcultures. Subcultures were marked by separate cultural patterns and worldviews that emphasized the “potent theme of violence current in the cluster of values that make up the lifestyle, the socialization process, the interpersonal relationships of individuals living in similar conditions” (Wolfgang and Ferracuti 1967, p. 140). The Black subculture was marked not only by poverty and exceptionally high rates of homicide. However, in 1971, Menachem Amir, a student of Wolfgang and Ferracutti, would develop a theory of Black subcultural violence patterns including rape insisting that: The Negro subculture is characterized by the revolving of life around some basic focal concerns which include a search for thrills through aggressive actions and sexual exploits. . . Young boys are imbued with negative, or at least ambivalent, feelings toward masculine functions. Sexual and aggressive behavior becomes the main vehicle for asserting their worthiness. They, therefore, idealize personal violence and prowess which substitute for social and economic advantages. (Amir 1971, pp. 327–328)
These theories of a Black male subculture driven by lack and violence held a peculiar appeal for white feminists who began theorizing that rape and sexual violence was a central feature of the masculinity created in poverty and racial and ethnic group disadvantage. Amir’s study made intra-racial rape a theoretical concern for white feminists going forward, and feminist analyses of Black masculinity appear
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
891
progressive despite advocating the premises of racist criminology and ghetto studies depicting Black men as culturally predisposed to crime, rape, and aggression. The inability of Black men to achieve economic or political independence, or patriarchy, prompted white feminists to insist that Black males were childlike and imitative of white men who were/are their masters. Shulamith Firestone (1970) argued in The Dialectics of Sex that “the relationship between the Black man and the white man duplicates the relationship of the male child to the father. We have seen how at a certain point in order to assert his ego, the child must transfer his identification from the female (powerless) to the male (powerful)” (p. 111). In the attempt to imitate white men, Black men distorted white culture and embrace pathology as a substitute for power. With the publication of Susan Brownmiller’s Against Our Will: Men, Women and Rape (1975), feminist theory had successfully linked subcultural theories of violence with the analytic concept of gender. Brownmiller had great admiration for the explanatory power of subcultural theories of sexual violence. According to Brownmiller (1975), “the single most important contribution of Amir’s Philadelphia study was to place the rapist squarely within the subculture of violence. The rapist, it was revealed, had no separate identifiable pathology aside from the individual quirks and personality disturbances that might characterize any single offender who commits any sort of crime” (p. 181). This was especially true of poor young Black males who were culturally predisposed to such violence (Brownmiller 1975, p. 180). By the 1980s, pathological theories of Black men and boys became normalized in white feminist analyses of rape and gender. In The Second Assault (1981), Joyce Williams and Karen Holmes develop a theory of racial-sexual stratification which focuses on the intra-racial Black male rapist. Williams and Holmes believed: “Rape, or the threat of rape, is an important tool of social control in a complex system of racial-sexual stratification. Fear of rape keeps not only the female in her place, but fear of the accusation of raping a white woman keeps minority males in their place as well” (Williams and Holmes 1981, p. 26). Because there was such a high cost involved with interracial rape, Williams and Holmes asserted that intra-racial patterns of rape were used to by Black men to assert their masculinity. Black men were not real men and had no actual power, so “in raping minority women, minority males frequently are doing no more than imitating the white male” (p. 27). Like the subcultural violence criminologists before them, Williams and Holmes asserted that Black masculinity is compensatory and defined by its lack of real patriarchal manhood. While Williams and Holmes admit that “there is no empirical evidence. . .nor is there any empirical validation for either the myth of Black male sexuality or that of sex as compensatory behavior” (Williams and Holmes 1981, p. 35), their theory of compensatory Black masculinity and racial-sexual stratification would come to define key aspects of intersectionality and Black feminist thought from the 1980s forward (Curry 2021a). Compensatory masculinity, or the conceptualization of Black masculinity as lacking manhood and seeking to restore this lack through violence, dominated early Black feminist and intersectional writings about Black men throughout late twentieth and early twenty-first century. Explanations for domestic violence, intra-
892
T. J. Curry
racial rape and sexual assault, and conflicts among intimate partners were often explained as a product of Black male’s compensatory pathologies. In Mapping the Margins, Kimberle Crenshaw (1991) cites Williams’ and Holmes’ The Second Assault (1981) quite positively as evidence that rape is a compensatory means of social control among Black men (Curry 2021a). In We Real Cool, for example, bell hooks (2004) warns that describes intimate relationships with Black men as a perilous endeavor. “Since so many black males, especially young black males, feel that they are living on borrowed time, just waiting to be locked down (imprisoned) or taken out (murdered), they may as well embrace their fate—kill and be killed” (p. 57). Like the subcultural criminologists before her (Amir 1971; Curtis 1975, 1976), Hooks asserts, with no ethnographic or statistical data to support this assertion, that challenges to the masculinity of Black men causes them to respond with “anger and sexual predation to maintain their dominator stance” (2004, p. 57). The depiction of Black males as pathological, violent, and sexually predatory due to their savagery and lack of real (civil) manhood has been a core tenet of American feminism from the mid-nineteenth century to the present. By making the freedom of Black males synonymous with the doom of Western civilization and women’s rights, feminism had a role in supporting and engineering racist social policies and criminal justice initiatives in the name of women’s rights.
Conclusion The quest for women’s rights in the United States originating in the nineteenthcentury writings of suffragists to the writings of Black and white feminists in the mid-twentieth century relied on racist theories of white evolution and Black savagery to explain the threats that a freed Black male population posed not only to white women but also to womankind. Despite decades of scholarship illuminating the relationship that feminism, and the political struggle for women’s rights, has had to anti-Black racism, colonial conquest, and classism, present-day scholars are met with opprobrium and sanction when attempting to discuss the limits and harms of American feminism. Academic discussions of white feminism’s racism are permitted if feminism itself is not questioned. When feminism’s racism is framed as a question of inclusion, or white women’s failure to include the Black and other non-white women’s experiences, the history of feminism as a movement for the freedom and liberation of women across the world remains intact. From this perspective, the historical movement and political aims of feminism were not incorrect or malicious, merely non-inclusive and incomplete. However, when feminism is interpreted as a historical dynamic that set itself against Black civil rights and deliberately campaigned for the elimination and lethal targeting of Black men for over a century, such conversations are often outright dismissed within the academy. The history of feminism’s racism reveals an insidious program of social control and political deflection that inhibited Black civil and human rights. Despite this history, academic theory proceeds as if feminism is largely compatible with the goals of racial equality and democratic progress in the United States. The fear of Black
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
893
male militancy and the encroachment of racial equality saw white feminist theory embrace an insidious form of racist criminology that viewed the poverty of Black people and the sexual insatiability of Black men to be incompatible with the safety and economic progress of white women in the United States. Further research is needed to assess the impact that American feminism’s anti-Black male rhetoric and theories have had on the economic, political, and cultural development of Black Americans.
References Amidon KS (2007) Carrie chapman catt and the evolutionary politics of sex and race, 1885–1940. J Hist Ideas 68(2):305–328 Amir M (1971) Patterns in forcible rape. University of Chicago Press, Chicago Ansell A (2013) Race and ethnicity: the key concepts. Routledge, New York Bederman G (1995) Manliness and civilization: a cultural history of gender and race in the United States, 1880–1917. University of Chicago Press, Chicago Beisel N, Kay T (2004) Abortion, race, and gender in nineteenth century America. Am Sociol Rev 69(4):498–518 Bhattacharyya G (2013) Dangerous Brown men: exploiting sex, violence and feminism in the ‘War on Terror’. Zed Books, London Bird C (1969) Born female: the high cost of keeping women down. Pocket Books, New York Bird RC (1997) More than a congressional joke: a fresh look at the legislative history of sex discrimination of the 1964 civil rights amendment. William & Mary J Women Law 3:137–162 Blee KM (1991) Women in the 1920s’ Ku Klux Klan movement. Fem Stud 17(1):57–77 Blee KM (2009) Women of the Klan: racism and gender in the 1920s. University of California Press, Berkeley Brownmiller S (1975) Against our will: men, women, and rape. Fawcett Columbine, New York Catt CC (1917) Objections to the federal amendment. In: Catt CC (ed) Woman suffrage by federal constitutional amendment. National Woman Suffrage Publishing, New York, pp 69–92 Couzins P (1881) Phoebe Couzins law student. In: Stanton EC, Anthony SB, Gage MJ (eds) History of suffrage Vol II, vol 387. Charles Mann, Rochester Crenshaw K (1991) Mapping the margins: intersectionality, identity politics, and violence against women of color. Stanford Law Rev 43(6):1241–1299 Curry TJ (2017) The man-not: race, class, genre and the dilemmas of black manhood. Temple University Press, Philadelphia Curry TJ (2018a) Another white man’s burden: Josiah Royce’s quest for a philosophy of white racial empire. SUNY Press, Albany Curry TJ (2018b) Killing Boogeymen: Phallicism and the Misandric mischaracterizations of black males in theory. Res Philosophica 95(2):235–272 Curry TJ (2021a) Decolonizing the intersection: black male studies as a critique of intersectionality’s indebtedness to subculture of violence theory. In: Critical psychology Praxis: psychosocial non-alignment to modernity/coloniality. Routledge, New York, pp 132–154 Curry TJ (2021b) He wasn’t man enough: black male studies and the ethnological targeting of black men in nineteenth-century suffragist thought. In: Davidson J (ed) African-American studies. Edinburgh University Press, Edinburgh, pp 209–224 Curry TJ (2022) Reconstituting the object: black male studies and the problem of studying black men and boys within patriarchal gender theory. In: Tate SA, Rodríguez EG (eds) The Palgrave handbook of critical race theory and gender. Springer International Publishing, New York, pp 525–544 Curtis LA (1975) Violence, race, and culture. Lexington Books, Lexington, MA
894
T. J. Curry
Curtis LA (1976) Rape, race, and culture: some speculations in search of theory. In: Walker MJ, Brodsky SL (eds) Sexual assault: the victim and the rapist. Lexington Books, Lexington, pp 117–134 Davis CJ (2005) Love and economics: Gilman on ‘The Woman Question’. ATQ 19(4):63–86 Dollard J (1937) Caste and class in a southern town. Doubleday, New York Evrie JV (1863) Negroes and negro slavery: the first an inferior race: the latter its normal condition. Van Evrie, Horton, & Co, New York Feimster C (2009) Southern horrors: women and the politics of rape and lynching. Harvard University Press, Cambridge Felton RL (1898) Mrs. Felton vs Manly. Literary Dig 17(22):625 Field CT (2014) Struggle of equal adulthood: gender, race, age, and the fight for citizenship in antebellum America. UNC Press, Chapel Hill Firestone S (1970) The dialectic of sex: the case for feminist revolution. Bantam Books, New York Giddings P (1985) When and where I enter: the impact of black women on race and sex in America. William Morrow Co., New York Gilman C (1898) Women and economics: a study of the economic relation between men and women as a factor in social evolution. Small, Maynard, and Company, Boston Gilman C (1900) Women and economics: a study of the economic relation between men and women as a factor in social evolution. Small, Maynard, and Company, Boston Gilman C (1908) A suggestion on the negro problem. Am J Sociol 14(1):78–85 Gold ME (1981) A Talke of two amendments: the reasons congress added sex to title VII and their implication for the issue of comparable worth. Duquesne Law Rev 19(3):453–477 Goodwin M (2013) The death of affirmative action? Wis Law Rev 2013(3):715–726 Hale GE (2010) Making whiteness: the culture of segregation in the south, 1890–1940. Vintage Books, New York Hill MA (1980) Charlotte Perkins Gilman: the making of a radical feminist, 1860–1896. Temple University Press, Philadelphia Hooks B (2004) We real cool: black men and masculinity. Routledge, New York Hu-DeHart E (1997) Affirmative action – some concluding thoughts. Univ Colorado Law Rev 68(4):1209–1216 Jacobs MD (2009) White mother to a dark race: settler colonialism, maternalism, and the removal of indigenous children in the American West and Australia, 1880–1940. University of Nebraska Press, Lincoln Jones MS (2020) Vanguard: how black women broke barriers, won the vote, and insisted on equality for all. Basic Books, New York Jones-Rogers S (2019) They were her property: white women as slave owners in the American south. Yale University Press, New Haven Kearney B (1903) The south and Women’s rights. The Women’s Journal, April 4, p 106 Kraditor A (1965) The ideas of women’s suffrage movement 1890–1920. Columbia University Press, New York Ligon EB (1903) The white woman and the negro man. Good Housekeeping 37:426–429 Litwack LF (1998) Trouble in mind: black southerners in the age of Jim crow. Knopf, New York Massie VM (2016) White women benefit most from affirmative action – and are among its fiercest opponents. Vox.com, June 23. https://www.vox.com/2016/5/25/11682950/fisher-supremecourt-white-women-affirmative-action McCammon HJ, Campbell KE (2001) Winning the vote in the west: the political successes of the women’s suffrage movements, 1866–1919. Gend Soc 15(1):55–82 McConnaughy CM (2013) The woman suffrage movement in America: a reassessment. Cambridge University Press, New York McDaneld J (2013) White suffragist dis/entitlement: the revolution and rhetoric of racism. Legacy: J Am Women Writers 30(2):243–264 McDougald EJ (1925) The task of negro womanhood. In: Locke A (ed) The New Negro: voices of the Harlem renaissance. Albert and Charles Boni, Inc., New York, pp 369–384
38
Feminism as Racist Backlash: How Racism Drove the Development. . .
895
McRae EG (2018) Mothers of massive resistance: white women and the politics of white supremacy. Oxford University Press, New York Mitchell M (2007) Lower orders, racial hierarchies, and rights rhetoric: evolutionary echoes in Elizabeth Cady Stanton’s thought during the late 1860s. In: DuBois EC, Smith RC (eds) Elizabeth Cady Stanton, feminist as thinker: a reader in documents and essays. NYU Press, New York, pp 128–154 Newman L (1999) White women’s rights: the racial origins of feminism in the United States. Oxford University Press, Oxford Newman L (2007) Women’s rights, race, and Imperilism in U.S. HIstory. In: Campbell J (ed) Race, nation, and empire in American history. UNC Press, Chapel Hill, pp 157–179 Oyewumi O (1997) The invention of woman: making an African sense of Western gender discourses. University of Minnesota Press, Minneapolis Penn RT (1977) Afro-Americans in the struggle for Women’s suffrage. Howard University, Washington, DC Penn RT (2002) The politics of the anti-woman agenda: African Americans respond to conservativism. In: Tate GT, Randolph LA (eds) Dimensions of black conservativism in the United States. Palgrave, New York, pp 64–89 Rhode DL (2009) Justice and discrimination: sex discrimination and the law. Harvard University Press, Cambridge Schwimmer R (1933) Letter to the editor. The New York Times, September 3 Sneider A (2008) Suffragists in an Imperial age U.S. expansion and the woman question, 1870–1929. Oxford University Press, London Stanton EC (1865/2001) This is the negro’s hour. In: Stanton EC, Anthony SB, Gordon AD (eds) The selected papers of Elizabeth Cady Stanton and Susan B. Anthony: in the school of antislavery 1840–1866. Rutgers University Press, New Brunswick, pp 564–566 Stanton EC (1868a) The destructive male. Women’s Suffrage Convention, Washington, DC, January 1 Stanton EC (1868b) Manhood Suffrage. The Revolution, December 24 Stanton EC (1869) Women and Black Men. The Revolution, February 4 Stein M (2012) Nature is the author of such restrictions: science, ethnological medicine, and Jim Crow. In: Cole S, Ring NJ (eds) The Folly of Jim Crow: rethinking the segregated south. University of Texas at Arlington Press, Arlington, pp 124–149 Train GF (1860) The facts, or at whose door does the sin(?) lie? R.M. DeWitt, New York Train GF (1862) Slavery and emancipation. T.B. Peterson & Brothers, Philadelphia Truth S (1867) Address to the first annual meeting of the American Equal Rights Association. New York City, May 10 Williams J, Holmes KA (1981) The second assault: rape and public attitudes. Praeger, Santa Barbara Wolfgang M, Feracutti F (1967) The subculture of violence: towards an integrated theory in criminology. Tavistock Publications, London Zakaria R (2021) Against white feminism: notes on disruption. W.W. Norton & Company, New York
Inequality and Inefficiency
39
Pranab Bardhan
Abstract
In this chapter, we consider several reasons why inequality can have serious inefficiency consequences even from the narrow point of view of economists, belying their traditional dogma of equality-efficiency trade-off. Inequality is thus not just ethically distasteful; it can be economically harmful, even ignoring problems of absolute poverty. Keywords
Equality-efficiency trade-off · Fairness · Inequality of opportunity · Cooperation
In many developing countries in the last three decades, up until the devastation of the pandemic, poverty by most criteria had by and large declined, but inequality had mounted (or remained high). The pandemic, of course, exacerbated inequality in various ways. The question we want to address in this chapter is why we should worry about rising inequality even if mass poverty were declining. In other words, even when the conditions of the poor improve, but suppose those for the rich improve much more, should we be concerned about rising inequality, and, if so, why? To many moral philosophers steeped in the theory of justice, inequality in society may simply be ethically distasteful. Even to those who are consequentialists, say, the utilitarians, many of whom may be economists, controlling for the average income, increasing income inequality reduces social welfare simply because the same dollar has lower marginal utility for the rich than for the poor. Many non-utilitarian philosophers find some kinds of inequality simply unconscionable. P. Bardhan (*) University of California, Berkeley, CA, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5_46
897
898
P. Bardhan
There is, however, an important discussion among philosophers (and some economists) about the kinds of inequality which are morally permissible. The latter, for example, may pertain to cases where between two individuals facing similar life chances, one may end up richer than the other simply because the former is more ambitious or hardworking than the latter. This brings to the fore a distinction between inequality of opportunity and that of outcome. As philosophers, public commentators, and the general public increasingly find the issue of personal responsibility in one’s choice or life decision quite socially salient, one can make a clear distinction between opportunity egalitarianism – seeking to offset only those inequalities that are due to circumstances beyond an individual’s control (like the characteristics of a family or neighborhood a child is born in or its biological characteristics) – and outcome egalitarianism that, often on grounds of social norms, seeks to offset even those differences in outcome that are due to an individual’s own choice (say, in blowing away one’s opportunity by indulging in drugs or alcohol) or initiative (or lack of it). Many economists take a much narrower approach on inequality, even when they are sensitive to issues of absolute poverty. They find relative inequality acceptable if it does not harm economic performance or efficiency; in fact, they are prepared to tolerate even a large dose of inequality if it improves the aggregate economic performance – the presumption is that the larger surplus keeps open the possibility of redistributing some to the poor (a possibility that is infrequently realized in actual politics), thus making everyone better-off. The main purpose of this chapter is to indicate why we may have sufficient reasons to worry about inequality even from this narrow economist’s efficiency point of view. Most undergraduate economics textbooks to this day discuss an equality-efficiency trade-off. This is mainly about the disincentive effects of attempts to redistribute income from the rich to the poor. Progressive taxes to fund such redistribution may discourage work effort, investment, and risk-taking, and the consequent shrinking of the pie may leave the poor worse-off compared to the case of a larger pie but with the same share as before. Okun (1975) compared such redistributive transfers with carrying water in “leaky buckets.” In general, it is argued allowing inequality is a way of encouraging entrepreneurs and other fortune-seekers, whose enterprise, new ideas, and innovations enrich them but also improve the conditions of the whole economy. The wealthy are more willing and able to take risks, and hence redistribution may reduce productive risk-taking if it transfers wealth from less to more risk-averse agents (unless there is a public risk insurance policy to cover the latter). It is also the case that the rich save more than the poor, so inequality can generate more savings and investible funds, which can expand the productive base of society. Such expansion or growth may eventually trickle down to the poor (though economists on the opposite side usually argue that such trickle-down is often not enough). Both sides, however, can agree that high inequality may weaken the povertyreducing powers of the same growth rate.
39
Inequality and Inefficiency
899
But the last few decades of advance in economic theory and empirical findings have raised questions about the general applicability of the equality-efficiency tradeoff. In the rest of this paper, I shall enumerate and examine some of the issues that are of relevance here. (a) When there is information asymmetry between the two sides in a given transaction, the trade-off may not hold. Creditors do not have enough information about the viability of a project brought to them by a potential borrower. You may have a project that you know is very much worthwhile from both private and social points of view, but the creditor may not be aware or convinced of it, and you do not have sufficient collateral to persuade the creditor. A rich man with an inferior project may get the loan, not you, because of the former’s larger assets and hence collateral value. Thus, inequality here promotes the less efficient outcome. Similarly, your low savings or collateral may not permit you to finance or borrow for investment in higher education for which you may otherwise have the talent and proficiency, whereas the less talented children of your rich neighbor go through college and university, while you drop out. This is a loss to society as well as yourself. (b) There is, of course, a great deal of socially unproductive risk-taking by the rich (with “collateral damage” for the poor), say, in financial or real-estate speculation. Even if we ignore this, it is important to keep in mind that not all dynamic innovations and productive risk-taking are by private fortune-seekers. In the USA, much of the basic or foundational research and great innovations of recent times (like the Internet, GPS, Digital Search Engine, supercomputers, Human Genome Project, magnetic resonance imaging, smartphone technology, hydraulic fracturing for shale gas, and a whole host of others) have been facilitated by or been the outcome of public investment funded to a large extent by taxpayer money. Scandinavian countries with a high-tax redistributive economy have not been lagging in innovations. In the Bloomberg ranking of countries by the Global Innovation Index, the USA is in the ninth position, and three Scandinavian countries (Denmark, Sweden, and Finland) have a rank above that of the USA. Even in the private sector, assuring temporary monopoly and thus great fortune for the innovator through the patent system has not been the only or the best way of encouraging innovations. Patents on a new technology often make things costly or obstructive for future innovators and thus may hamper further advances in technology. A prize for solving a well-defined technology problem may be better. Also, open-source programs are often more conducive to new developments of technology. (c) If inequality is generated by market power of big firms in product and labor markets, then there is a direct loss of efficiency (in output and employment) if that market power enables those firms in their attempt to maximize profits to restrict output (and labor hiring) below the amounts for a competitive firm. (d) Historically, the case where inequality and inefficiency have stubbornly persisted together relates to land, which is usually very unequally distributed.
900
P. Bardhan
In traditional (and in some non-traditional) agriculture, the empirical evidence suggests that economies of scale in farm production are insignificant (except in some plantation crops) and that the small farm is often the most efficient unit of production. Yet the violent and tortuous history of land reform in many countries suggests that numerous road blocks on the way to a more efficient reallocation of land rights are put up by the powerful landed interests for many generations. Why don’t the large landlords instead voluntarily lease out or sell their land to small farmers and grab in the bargaining process much of the surplus arising from this efficient reallocation? There clearly has been some leasing out of land, but problems of monitoring the tenant’s work and application of inputs, insecurity of tenure (discouraging long-term land improvements by the tenant), and the landlord’s fear that the tenant will acquire occupancy rights on the land have limited the scope for such efficiency gains and the extent of tenancy. The land sales market is often rather thin (and in many developing countries, the sales sometimes go the opposite way – from distressed small farmers to landlords and money-lenders). With low household savings and collaterals, the potentially more efficient small farmer is often incapable of affording the going market price of land. Landlords on the other hand often resist land reforms particularly because the leveling effects reduce their social and political power and their ability to control and dominate even non-land transactions in the village. Large land holdings may give their owner special social status or political power in a lumpy way – so that the status or political effect from owning 100 hectares is larger than the combined status or political effect accruing to 50 new buyers owning 2 hectares each. Thus, the social or political rent of land ownership for the large landowner may not be compensated by the offer price of numerous small buyers. Under the circumstances, the former will not sell, and inefficient (from the productivity point of view) land concentration persists. (e) Then there is the demand-side impact of inequality. There is a story of a Ford company executive in conversation with a union leader, pointing to the arrival of a bunch of robots in the factory and asking, “can you collect union dues from them?”, to which the union leader replied, “can you get them to buy Ford cars?”. Particularly, in times of depressed aggregate demand and idle capacity, inequality may hurt macroeconomic performance by making it difficult to stimulate enough mass consumer spending. Recent research suggests – for example, by Auclert and Rognile (2020) – that a long-term rise in inequality can push the economy into a deep recession. There is also some literature which shows that inequality encourages excessive risktaking in the financial sector and along with household indebtedness may have been responsible for the financial crisis that originated in the USA in 2007–2008 – see, for example, Rajan (2010), and for a more formal analysis, see Kumhof et al. (2015). (f) Under inequality, not merely the aggregate consumer demand may be deficient, but also the pattern of consumer spending may also get distorted. Certain types of
39
Inequality and Inefficiency
901
consumer spending on status goods or what Hirsch (1976) called “positional goods” (houses, cars, clothings, and other easily visible conspicuous consumption items) can, in the context of inequality and community norms for emulation and the resultant “expenditure cascade,” lead to a race to the bottom among neighbors and reference groups, as Frank (2007) has abundantly illustrated: clearly an inefficient outcome. (g) The link between inequality and crime has often been pointed out both in scholarly and popular outlets. For an empirical confirmation of the association between inequality of visible or conspicuous expenditure and violent crime, see Hicks and Hicks (2014). For a review of the time-series evidence from several countries, which clearly shows a positive relation between income inequality and property crimes and violent crimes like robbery, homicide, and murder, see Rufrancos et al. (2013). (h) Similarly, the link between inequality and social and political conflicts (and hence economic disorder and instability) is also often suggested. Here, the evidence is more mixed, as the literature review by Ray and Esteban (2017) shows. There are several reasons for this. Firstly, one has to be clear about the nature of conflict, which can range from industrial strikes and agitations all the way to violent civil wars. Secondly, for most of these conflicts, grievance arising out of economic disparity is not enough, and one needs resources, organization, agency, initiative, and leadership to pave the way for action that takes the form of conflict and, more importantly, to sustain it. Thirdly, when group conflict arises out of intergroup disparity, the measure of disparity that may be more relevant than the usual inequality is what Esteban and Ray (1994) measure as polarization, which takes into account the depth of cleavage or distance between the groups as well as their size. Fourthly, conflict may originate in the tension that may grow more from the change in the relative income status of two even similar groups rather than the overall level of inequality. Then, there is what is called the Tocqueville paradox – in his words: “The hatred that men bear to privilege increases in proportion as privileges become fewer and less considerable, so that democratic passions would seem to burn most fiercely just when they have least fuel.” (i) Just as the evidence on the relation between inequality and conflict is mixed, that between inequality and the obverse of conflict, i.e., cooperation (in team production efforts, in resolving disputes, and in the management of the local commons) is also not straightforward. Bardhan and Dayton-Johnson (2007) review the empirical literature on the relation between inequality and cooperation in management of water resources for irrigation in developing countries. There are sometimes important initial setup costs in an irrigation management regime, which the rich and powerful people in the village often provide or take the leadership in mobilizing and sustaining. On the other hand, they cite quite a bit of evidence that relative equality helps in the formation and maintenance of water user associations, in the following of water allocation rules, and in the broad-based resolution of water disputes.
902
P. Bardhan
Baland and Platteau (1997) cite similar evidence from Africa in the community management of forests, fisheries, and grazing lands. The empirical evidence is, however, often deficient in providing sufficiently refined data to discern among varied theoretical hypotheses about norms, bargaining power, and perceptions of fairness under situations of inequality. Experimental evidence suggests that subjects whose fallback positions are very different are less likely to come to agreements than are more equally situated subjects – see, for example, Lawler and Yoon (1996). Also under inequality, bargaining failures may occur because inequality heightens informational asymmetries among the bargaining partners or because very unequal offers based on disparities in initial wealth or bargaining power are likely to be perceived as unfair and rejected, as in experimental play of the ultimatum game. For a more theoretical discussion on the issue of initial wealth inequality and cooperation, see Bardhan et al. (2000) and the chapters by Baland and Platteau and by Bardhan, Ghatak, and Karaivanov in Baland et al. (2007). (j) It is quite common to observe in the labor market (which is qualitatively different from markets in, say, vegetables, as it more directly involves social institutions and norms) that perceived inequality may disrupt norms of individual dignity and autonomy of the worker and thus affect labor discipline, loyalty, turnover, and ultimately productivity. Looking particularly to the future patterns of work, as the nature of human work is likely to involve more personalized or customized (including caregiving) services and more of production and dissemination of knowledge, the role of intrinsic motivation (i.e., when you do something largely for your own satisfaction/esteem not just for external rewards) and of shared norms inherent in these lines of work will be increasingly important – perceived inequality palpably affects these motivations and norms. (k) It has been widely observed and commented upon that economic inequality enables the rich and the corporate sector to pour resources in the political influence machine to get the system to work in their favor, particularly through lobbying (not just in improving access, but in the USA, the lobbyists now actually develop and draft the legislation in some cases) and election finance. This often results in laws and regulations in favor of wealth concentration and perpetuation of plutocratic power and away from efficient outcomes, apart from undermining democracy. For a recent incisive general account of how the business lobbies work in the USA and their adverse effects, see Drutman (2015). Corporate power in lobbying, bribing, election-funding, and media-shaping is mounting in many developing countries as well. In India, the recently introduced electoral bond system has made an already highly corrupt system of election finance even more murky. (l) Another political mechanism through which inequality can affect efficiency in the delivery of public services is what is called “secession of the rich”: rising inequality is usually associated with the rich opting out of public services and turning to private providers (private schools, private clinics and nursing homes, gated communities for safety, etc.); this “exit” results in a lowering of the general quality of public services as they lose influential political support (“voice”).
39
Inequality and Inefficiency
903
(m) Social inequalities also have adverse economic efficiency effects. In countries of acute gender inequality, women’s education, health, and work participation suffer, and this has negative consequences not merely for the women themselves but also for the children that these women bring up. Thus, society pays the price of gender inequality across generations. Similarly, if there are serious inequalities across neighborhoods and localities, a child born in a backward area will have inferior schools, roads, and other facilities and less exposure to good networks, peer groups, and role models in the neighborhood and other forms of social capital. This has obvious effects on future economic performance of the child. We have thus considered several reasons why inequality can have serious inefficiency consequences even from the narrow point of view of economists, belying their traditional dogma of equality-efficiency trade-off. Inequality is thus not just ethically distasteful; it can be economically harmful, even ignoring problems of absolute poverty.
References Auclert A, Rognlie M (2020) Inequality and aggregate demand. unpublished Baland J-M, Platteau J-P (1997) Wealth inequality and efficiency in the commons, part i: the unregulated case. Oxf Econ Pap 49(4):451–482 Baland J-M, Bardhan P, Bowles S (2007) Inequality, cooperation, and environmental sustainability. Princeton University Press, Princeton Bardhan P, Dayton-Johnson J (2007) Inequality and the governance of water resources in Mexico and South India. In: Baland J-M, Bardhan P, Bowles S (eds) Inequality, cooperation, and environmental sustainability. Princeton University Press, Princeton Bardhan P, Bowles S, Gintis H (2000) Wealth inequality, wealth constraints and economic performance. In: Atkinson AB, Bourguignon F (eds) Handbook of income distribution, vol I, Amsterdam Drutman L (2015) The business of America is lobbying. Oxford University Press, New York Esteban J, Ray D (1994) On the measurement of polarization. Econometrica 62(4):819–851 Frank RH (2007) Falling behind: how the rising inequality harms the middle class. University of California Press, Berkeley Hicks DL, Hicks JH (2014) Jealous of the joneses: conspicuous consumption. Inequality and Crime Oxford Economic Papers 66(4):1090–1120 Hirsch F (1976) Social limits to growth. Harvard University Press, Cambridge, MA Kumhof M, Ranciere R, Winant P (2015) Inequality, leverage, and crises. Am Econ Rev 105(3): 1217–1245 Lawler EJ, Yoon J (1996) Commitment in exchange relations: test of a theory of relational cohesion. Am Sociol Rev 61(1):89–108 Okun AM (1975) Equality and efficiency: the big tradeoff. Brookings Institute, Washington, D.C. Rajan R (2010) Fault lines: how hidden fractures still threaten the world economy. Princeton University Press, Princeton Ray D, Esteban J (2017) Conflict and development. Annual Review of Economics 9:263–293 Rufrancos HG et al (2013) Income inequality and crime: a review and explanation of the time–series evidence, sociology and criminology. Open Access 1(1):1–9
Index
A Abolition of slavery, 357, 360, 361 Abortion, 50 Abrahmani, 429 Abrahmani-gender comple, 428–432 Absolute position, 52 Absolute poverty, 354 Achievement distribution, 123, 126 Active corrective policies, 64 Activism, 95 Act on the Promotion of the Elimination of Buraku Discrimination (APEBD), 513 Affirmative action, 11, 13, 52, 54–57, 63–65, 514, 720–725, 730–735, 751, 755–757, 762, 774, 820–824, 826–837, 842, 843 academic preparedness, 123 active corrective policy, 64 adaptations, 125 arguments, 55 barriers types, 123 beneficiaries, 125 cascade effect, 125 case studies, 51 caste-based, 125 class-based inequality, 55 cooperation, 727 criticisms, 55 discriminatory mindsets, 125 discriminatory system, 124 dishonest behavior, 727 diversity, 54 education and public sector employment, 531–533 effectiveness, 126 efficiency loss, 723 equalizing opportunities, 122 evaluation of beneficiaries, 726 goal-based policies, 123 in government contracting, 624–625
higher education, 125 in India, 785–786 macro level, 55 meritocracy, 122 mis-targeting, 124 participation, 722 polarized (and politicized) debates, 123 policy design, 122 as policy intervention, 785 politics, 533–534 preferential treatment programs, 51 racial animus, 55 redistribution interventions, 122 resistance, 55 segregation, 59 social obstacles, 123 structural disadvantages, 122 subaltern population, 65 underprivileged students, 124 underserved groups, 124 unintended consequence, 124 Affirmative action, in South Africa amendments, 809–811 critiques, 805–806 description, 803 empirical analysis, 812–814 multidimensional measures, 806–808 redress and access, 804–805 Sectoral Charters and Codes, 808–809 Affirmative action policies (AAPs), 495, 842–848, 851–856, 860, 861 accurate and strategic messaging of, 857–859 renewing commitments, 859–860 Africa National Congress (ANC), 492 African women, 91 Afro-Brazilian activists, 360 Afro-Brazilian population, 359 Aggregate decomposition, 136
© Springer Nature Singapore Pte Ltd. 2023 A. Deshpande (ed.), Handbook on Economics of Discrimination and Affirmative Action, https://doi.org/10.1007/978-981-19-4166-5
905
906 Aggressive and frequent stops, 574 Agraharam, 676 Agricultural innovation, 658 Alexander, M. Jacqui, 95 Alumni associations, 110 Amalgamation, 59 Amaltas, 313 Amanah Saham Bumiputera, 830, 831, 835 Ambedkar, B.R., 64, 426–427 American justice system, 587 American Revolution, 57 America’s feminism, 13 Analogous discrimination, 424 Animus, 18 Anonymity, 316 Anti-colonialisms, 97 Anti-colonial politics, 95 Anti-discrimination law, 459, 844 Anti-discrimination policies, 8 Anti-fat/pro-thin attitudes, 598, 606 Anti-imperial politics, 95 Anti-Muslim discrimination, 446 Anti-poverty programs, 55 Anti-racist politics, 95 Apartheid-based discrimination, 482 Apartheid policies, 484, 494 Approach-avoidance conditioning, 212 Article 153, 822–824 Asian Americans, 594, 606 Asian financial crisis, 828, 830 Asian financial crisis of 1997, 76 Assimilation, 445 Atrocity, 433 Attention discrimination, 177 Atwater, L., 51, 54, 60 Audit studies, 156, 158, 319, 670, 676, 677, 683 labor market, 158 limitations of, 160 Pager method, 159 Australia, 316 Avarna, 518 Average treatment effect (ATE), 140 B Bail decisions, 585 Ban the box, 286 Basu’s analysis, 4 Becker, G., 53 Becker’s classic theory, 3 Becker’s theory bias, 25, 28, 32, 37 competition, 23 conceptual frameworks, 19–22
Index customer discrimination, 24, 25 empirical assessments, 22–24 ethnic gaps, 23 inaccurate beliefs, 34 prejudice, 26–28, 30, 31, 36–40 residual approaches, 25, 26 social cognition, 30, 31 socio-historical roots, 31–33 stereotype content model, 29, 30 “Behavioral” decision maker, 237 Beijing, China, 248 Being out, 304 Between-type inequality, 121 Bhanwari Devi, 433 Bhuria Commission, 753 Biased behaviors, 595 “Bias of Crowds” model, 608 Big data, 318, 320 Bilge, Sirma, 425 Bisexual, 301, 302, 309, 310, 313 Black American, 600, 603–608 Black Britain, 92 Black Economic Empowerment, 485 Black feminism, 90, 91 Black Hand, 577 Black Lives Matter movement, 115, 581 Black males, 873, 878, 887, 890, 892 Black mothers, 604 Black-owned businesses in USA, 636 Black women, 410 Black workers, 484, 491 Blinder-Oaxaca decomposition method, 5 Blue Diamond Society, 311 Borderlands, 86, 91, 93, 94 “Border” theory, 93 Bosnia-Herzegovina, 252 Brahmins, 676, 781, 792 Brazilian demographics, 367 Brazilian Institute of Geography and Statistics, 355 Brazilian National Survey by Household Sampling, 355 Brazilianness, 359 Brazil’s National Poverty Line (NPL), 373 Brexit movement, 467 Broad-Based Black Economic Empowerment Act (B-BBEE), 806–808 Budaun, 435 Bumiputera Commercial and Industrial Community (BCIC), 826, 827, 830, 834 Bumiputeras, 820–822, 825–827, 829, 834–837 Buraku Liberation League (BLL), 501
Index Burakumin, 500 discrimination, 501 identity, 500 postwar Buraku liberation movement, 507–513 prewar outcaste emancipation, 501–507 1918 Rice Riots, 505 Bureau, 62 Bureau of Refugees, 61 C Call-back rates, 451, 452 Call log data, 679 Call-tracking, 680 Canada, 8 economic disparities, 472–476 employment, 467 exclusionary discrimination, 476 immigration system, 464 internationally trained immigrants, 476 migration, 466 occupational integration, 476 points system, 464 provincial/territorial government, 476 racialized population, 465 region of birth, 467 social and economic opportunities, 476 social disparities, 469–472 social services, 465 socioeconomic disparities, 468, 469 urban centers, 464 Canadian labor market, 467, 469, 470 Capitalism, 87, 88, 91, 96, 98 Caste and religious discrimination, 671 Caste-based discrimination, 637 Caste Development Index (CDI), 675 Caste disparities, 530 regional patterns, 535–537 Caste quotas contemporary reality, 791–792 criticism of, 792–794 demands for reservation, 790–791 for economically weaker sections, 789–790 in legislature, 788 in public sector Educational Institutions, 787–788 in public sector occupation, 786 Caste system, 247 Catholic Croats, 252 “Ceteris paribus” condition, 107 Chamar community, 326 Chaturvarna, 780
907 Civil Rights Act of 1964, 63, 64 Civil services, 460 Clearance rate, 575–577, 580 Cognitive empathy, 328–335, 339, 342–345 Collective silence, 578 Collins, Patricia Hill, 425 Colonial epistemologies, 97 Colonialism, 95, 97, 464 Colonial Malaya, 822, 824 Colour Bar Act, 800 Communas, 341 Comparative Feminism, 96 Compensation principle, 120, 121 Computer-administered test, 294 Conceptual frameworks, 19, 20 Conflict resolution, 579 Connections, 95 Conscious beliefs, 570 Consciousness-raising strategies, 210 Conservative modernization, 357, 394 Consortium of Gallup Europe, 315 Constitutional premises, 822 Consumer discrimination, 624 Contact hypothesis, 206 Contraband, 293 Convergence, 524 Cooperation, 249–256, 575, 577, 578, 901, 902 Corrective policies, 123 Correspondence studies, 156, 160 academia, 171 in labor market, 161–169 limitations of, 173–178 in rental markets, 169–171 retail, 171 Counterfactual wages, 136, 140, 142 Counter-geographies, 93 COVID-19 pandemic, 617 Craigslist, 680 Credit-scoring algorithms, 641 Creditworthiness, 59 Crenshaw, Kimberle Williams, 425 Crimes, 568, 575, 578, 579, 581, 585 Criminal homicides, 575, 576, 587 Criminal justice system, 286 Critical historical knowledge, 856–859 Critical Race Studies, 90, 91 Critical Refugee Studies, 100 Critical security studies, 100 Cross-country analysis, 651 Crowding-out hypothesis, 731, 735, 751 Cultural anthropology, 324, 326 Cultural environment, 295 Cultural health capital, 654
908 Cultural mental models, 238, 239, 256 Cultural norms, 409 Cultural Studies, 92 Current Population Survey (CPS), 22 Customer discrimination, 22, 24 Customer preferences, 620–622 D Dalits, 247 Death rates, 604 Decomposition conditional quantiles, 146–147 formal identification, 141–142 issues with, 137–139 Oaxaca-Blinder method, 135–137 recentered influence functions, 147–148 residual imputations, 143–144 reweighting methods, 144–146 treatment effect interpretation, 139–141 Delhi Metro, 678 Delivery services unit, 107 Demand and supply of health, 653 Democracy, 763, 764, 772 Democratic era, 485, 495 Democratic futures, 96 Demographic representation, 598, 606–608 Department of Housing and Urban Development (HUD), 670, 673 Department of Research and Statistics (DARES), 450 Depression, 358 1877 Desert Land Act, 57 Destruction game, 253 Developmental aid policies, 702 Development outcomes of LGBTI people, 301, 304, 310, 311, 313, 319 Development research, 87 Diaspora, 92–93, 99 Diasporic Hegemonies, 99 Dictator game, 262, 265, 266, 269, 274–277 Dignity, 112 Disability, 492–493 Disaggregate data, 413 Discrimination, 2, 52, 53, 63–65, 134, 139, 142, 263, 264, 269, 272, 273, 276, 277, 280, 411, 413, 414, 467, 469, 500, 502, 504, 506, 510, 570–573, 575, 584–587, 595, 596, 598, 603, 604, 607–609, 616, 618, 620–622, 624, 628, 824 audit studies, 158–160 benefits of diversity, 197–200 colorism, 53
Index correspondence studies (see Correspondence studies) and corruption, 199–200 dimensions of, 9–11 endogenous responses to bias, 194–195 expectancy effects and self-fulfilling prophecies, 192–195 game, 109, 110 Gary Becker’s models of, 53 Goldberg paradigm experiments, 182–183 identity and preferences, 190–191 implicit association test, 178–182 index, 493–494 intergroup contact, 206–209 and intersectionality, 7–8 labor market, 3 law of small numbers, 200 leaders and role models, 200–205 against LGBT candidates, 168 list randomization, 183–184 markets and, 107–110 in politics and inequality across groups, 196 Pygmalion and Golem effects, 192–194 social and regional dimensions of, 8–9 socio-cognitive de-biasing strategy, 209–218 statistical, 3 statistical model, 153 stereotype threat and underperformance, 187–190 taste-based, 154, 155 technological de-biasing, 218–222 willingness to pay, 184–186 workhorse models of, 153 Discrimination and affirmative action, 842, 843 definitions and policy, 843–846 workplace experiences of, 846–848 Discrimination in credit evidence for, 635–641 implications, 641–642 law, rules and institutions, 638–641 mechanisms and methods, 642–644 Discriminatory patterns of lending, 639 Dissimilarity index (DI), 365 Distributions, 143–147 Diversity, 416, 465 Doll test, 291 Domain identification, 850 Domestic responsibilities, 72 Domestic violence, 333 Dōwa Special Measures Law (SML), 508 Downward occupational mobility, 477 Dynamic efficiency, 78 Dynamic general equilibrium model, 659
Index E Eagly’s hypothesis, 182 Echantillon Démographique Permanent (EDP), 447 Econometric analysis, 447 Econometric decomposition, 449 Econometric techniques, 670 Economic(s), 106, 108, 112, 324, 325, 327–335, 343–345 assimilation, 455 departments, 244 of discrimination, 468 liberalization policies, 773 migration model, 470 repercussions, 458 Economic disparities accreditation/licensing procedures, 472 average hourly wages, 472 education level, 472 labor market, 474 online job postings, 474 over-education, 473 prejudice, 474 race and gender inequities, 475 racist perception, 474, 475 unemployment rates, 472, 475 Economic Freedom Front (EFF), 492 Economic stratification ascriptive characteristics, 287 intergroup inequality, 287 prejudice, 288 See also Stereotypes Education, 596, 598, 602, 606, 609 achievement, 118, 469 attainment, 477 bridging programmes, 454 mismatch, 456 system, 495 Efficiency loss, 723 Electoral quotas, 731, 734, 748, 752 Elementary occupations, 76 Emotional bias, 241–242 Empirical economics, 328, 329, 340 Empirical findings, 113, 114 Employment, 731–733, 739, 741, 756 discrimination, 474, 475, 477 equity, 477, 488–491, 803–811 trends, 487, 488 Employment Equity Act, 485 Endogamy, 426 Enterprise development, 807, 809, 810 Entitlement effect, 248 Entrepreneurship, 616, 619, 620, 622, 628
909 Equal Credit Opportunity Act of 1974, 636 Equality, 771, 773 Equality-efficiency trade-off, 898, 899 Equilibrium fiction, 248 Equitable distribution of incomes/payoffs, 112 Ethnic gaps, 23 Ethnic homophily, 469 Ethnicity, 408, 620, 621, 624 “Ethnic salience” test, 254 Ethnic studies, 90, 91, 94 Ethnography, 325, 326, 328, 329, 332, 334, 340–343 Ethnology, 871 Ethno-racial income inequalities, 370 Ethno-racial inequalities, 372, 378 European American, 606 European Union Agency for Fundamental Rights (FRA), 305, 314 Exclusion of blacks, 483, 484 Exogenous preferences, 237 Exo-group, 451 Experimental designs affirmative action, 721 baseline inequality, 721 salient identity, 722 Experimental field studies, 595 Experiments, 262, 264, 271, 273, 279, 319 Explicit attitudes and beliefs, 596 Explicit beliefs, 570 Exposure index (EI), 365 “The Extended Case-Study Method”, 328 Extensive margin, 733–734 F Facebook, 318 Fair Labor Standards Act (FLSA), 60 Fairness, 902 Family Health Programme, 396 Federal Housing Administration (FHA), 59 Federal job guarantee, 64 Female mortality, 650 Female share, 487, 494 Feminisms, 86, 91, 95, 96 Feminism’s racism, 892 Feminist, 86–88, 90, 91, 93–100 movements, 86, 87, 90, 95, 98, 100 networks, 97 NGO’s, 98 praxis, 96 research, 87, 88, 96, 97, 100 theory, 889–892 Feminization, 471
910 Ferguson Police Department, 574 Field experiments in academia, 171 attention discrimination, 177 Field-oriented economists, 335 Financial diaries, 335, 336 Firearms, 571, 575, 579 Five-year Malaysia plans, 832 Football games, 241 Formation Qualification Professionnelle (FQP), 448 France employment opportunities, 460 ethnic minorities, 454 geographic regions, 447 Islamic attacks, 444 labor migration, 448 Maghreb populations, 447 practical professional life, 456 working population, 445 See also Religious discrimination Fraternities, 110 Freedmen and Abandoned Lands, 61 Freedmen’s Bureau, 61, 62 1865 Freedmen’s Bureau bill, 61 French Labor Code, 460 French Microfinance Institution (MFI), 639 French socio-economic system, 444 French spirit, 458 Frictional environments, 453 FUNDEF programme, 396 G Gantt, H., 50 Gay, 99 men, 300, 301, 313, 319 sex, 99 Gender, 490, 492, 495, 496 expression, 303 identity, 302–304, 307, 308, 310–314, 316, 317, 319 inequality, 70–73, 76, 289, 458, 471 pay gap, 75, 82 segregation, 470 Gender-ascriptive social relation, 70 Gender-based discrimination, 11, 843 Gender discrimination, 70, 650 economic costs in health, 659 within households, 650–654 in labor markets, 70, 77, 83 measurement issues, 655–656 policy insights, 660–661
Index Gender quotas, in politics conceptual problems, 771–774 as fast track to equal representation, 766–771 types of, 766 Gender-Science IAT, 243 GI Bill, 59, 60, 63 Gilman’s white supremacy, 879–885 Gini Index, 368 Global feminism, 98 Global Financial Crisis of 2009, 76 Global housing discrimination, 672–674 Globalization, 86, 92, 94, 98 Global sisterhood, 94, 96 Global social change, 86 Global South, 98 Goldberg paradigm experiments, 182–183 Golem effect, 193 Google U.S. consumer survey, 309 Government-linked companies (GLCs), 828, 830, 834, 835 Graduate Record Examination (GRE) standardized test, 189 Gram Panchayat, 534 Gram sabhas, 343 Grewal, Inderpal, 95 Grewal’s analysis, 4 Gross national income (GNI) per capita, 73 Group-based income inequalities, 370 Group-based inequalities, 355, 360 Group discrimination, 112, 113, 115 Group identification, 51, 52 Group justification theories, 290 Group status, 51 Group Weighted Coefficient of Variation (GWCOV), 370 Group Weighted Gini Index (GWGini), 370 Group Weighted Theil Index (GWTheil), 370 H Hate crimes, 236 Hathras, 434 Healthcare, 596, 598, 604–606, 609 Health insurance programmes, 655 Health outcomes, 650, 655, 657, 658, 660 Health utilization, 653, 660 Helms, J., 50 Heteronormative family, 91 Hetero-patriarchies, 95 Heterosexual applicants, 319 Hijras, 313 Hispanic immigrants, 673
Index Hispanic women, 414 Hit-rate test, 293, 571–573, 585 Home-based work, 72 Homelands, 483, 484 Homeownership, 455, 642 Home Owners’ Loan Corporation, 59 Homestead Acts Homestead Act of 1862, 57, 58 homesteading as practice, 57 Southern Homestead Act of 1866, 57, 62 Homesteaders, 62, 63 Homesteading, 57, 58, 61, 62, 64 Homicide(s), 575, 579–582, 587 clearance rate, 575–577, 579, 580 Homonationalism, 99 Homosexuality, 302 Horizontal inequality, 112 Horizontal’ model of inequality, 355, 356, 359, 372, 387, 388, 402 Hostile environments, 424, 432–435 Household registration system, 248 Housing discrimination in India applicant and application characteristics, 682 caste and religion, 675 CDI, 675 data and analysis, 679 features of the properties, 681 historical experience, 674 Muslims, 675 muslims applicants, 685 names and contact strategy, 678 residential segregation, 676 responders and responses by applicant type, 683 SC/OBC applicants, 687, 689 SC/ST, 675 tracing callers, 680 UC applicants, 687, 689 Housing Discrimination Study, 673 Housing market, 449, 454–455, 669–674, 692 Housing Market Practices Survey (HMPS), 670, 673 Howard, O. O., 62 Hukou, 248 Human capital, 468, 469, 476, 524–527 Human resource workers, 245 I IAT D scores, 602 Identities, 110, 263, 265, 269 Identity-safe environments, 843, 854, 856, 858–860
911 Ignorability, 142 Illiterate agro-export outpost, 357 ILO Maternity Protection Convention, 80 Immigrant, 86 Immigration, 445, 446 Immigration Act of 1976, 464 Immigration policy compounding factors, 467 point system, 464 structure, 464 Imperial cultures, 95 Imperial feminism, 91 Imperialism, 86, 91, 95, 98, 99 “Impersonal” or “automated” audit studies, 671 Implicit association test (IAT), 6, 10, 178–182, 243, 253, 273, 294 Implicit attitudes and beliefs, 596, 597 Implicit bias, 238, 291, 294, 596–598, 602–608 and demographic representation, 606 and individual discriminatory behavior, 597 and macro-level societal variables, 607 and systemic behaviors, 608 and systemic discriminatory behavior, 597, 598 systemic outcomes, 606 of teachers against females in science, 243, 244 Implicit discrimination, 237 Implicit stereotypes, 602, 603, 605–607 Inaccurate beliefs, 33 Inaccurate statistical discrimination, 618 Inclusive society, 416 Income inequality, 368 Income inequality between households, 76 India, 9, 247, 313, 784 affirmative action in, 785–786 caste quotas (see Caste quotas) housing discrimination, 671, 674–678, 683, 693 local government and development, 739 NREGS, 739–741 rural roads and public goods, 741–742 Indian Census, 733 Indian entrepreneurship, 619 Indian Muslims in Dravidian land, 557–559 education, 549–550 in Government and public sector jobs, 554 income of, 545–549 jobs, 550–552 in Northern and Eastern India, 560 politics ad policies of affirmative action, 554–561
912 Indian Muslims (cont.) socio-economic situation and educational marginalization, 545 statistics, 562–563 timidity and bias at the center, 555–557 youth in higher education, 552–553 Indian National Sample Survey, 343 Indian SIM cards, 679 India’s National Sample Survey, 675 Indigenous groups, 822 Individual discriminatory behavior, 597 Industrialization, 358 Inequality(ies), 87–89, 100, 496, 825, 832, 833 and community normsI, 901 and cooperation, 901 and crime, 901 demand-side impact of, 900 by earnings and work status, 383 economic, 902 and inefficiency, 899 market power of big firms, 899 public services, 902 regimes, 477 relative, 898 social, 903 and social and political conflicts, 901 Inequality of opportunity (IOp), 5, 898 acceptable measure, 121 accountability and responsibility, 119 affirmative action, 119, 122–126 aspirations, 118 compensation principle, 121 conceptualization, 120 distortionary behavioral responses, 127 educational outcomes, 118 estimation, 122 ex-ante and ex-post approach, 121 exclusionary practices, 126 external barriers, 119 framework, 118 higher education, 127 intuition, 120 lifespan, 120 microdata, 122 non-discriminatory practices, 119 personal efforts, 126 principles, 121 social hierarchies, 126 socioeconomic status, 118 theoretical and empirical challenges, 118 utilitarian reward, 121 within-group inequality, 121 Infant health outcomes, 596, 604 Information channel, 456
Index Innovations, 899 Institutional discrimination, 635 Integrated classrooms, 252–253 Integration, 445 Intensive margin, 734 Interactive work, 114 Inter-ethnic cooperation, 251 Intergenerational inherited deprivation, 64 Intergenerational transmission, 296 Intergroup contact theory, 206–209 Intergroup inequality, 52, 295 International (Extreme) Poverty Line (IPL), 373 International Labor Organization, 459 International Lesbian Gay Bisexual Trans and Intersex Association (ILGA), 315 Internet-based housing portals, 673 Interracial couples, 411 Inter-racial marriage, 366 Intersecting inequalities in Brazil access to, and control over land, 385, 386 ‘conservative modernization’ to ‘liberal neo-developmentalism, 394 educational outcomes, 388, 390, 391 ethno-racial group poverty, 375 from horizontal to vertical inequality, 359 illiterate agro-export outpost, 357, 358 incidence of poverty by gender, 374 income inequality, 368, 370 inequalities by earnings and work status, 383 labour market segregation by occupation and work status, 378 in late 20th century, 362, 363, 366 poverty by income groups, 374 poverty by region, 376 poverty trends, 366 rural urban locations poverty, 376 social movements, 398–400 from vertical to intersecting inequalities, 360, 361 wage inequalities, 377, 378 Intersectional approach, 414, 416, 418 Intersectionality, 775 Intersex, 303 Islamophobia, 444, 457, 458 J Japan, 9 Jatis, 518 Jharkhand, 737 Ji’ichirō Matsumoto, 509 Jinkenren, 513 Jiyū Dōwakai, 510
Index Job access, 447, 450–452 Job audit methodology, 287 Job grading systems, 71 Job segregation, 289 Job training, 60 Johnson, A., 61, 62 Johnson, L. B., 54 Judges, 241–242 K Kahneman, Daniel, 237 Kaplan, Caren, 95 Katrina, H., 53 Kennedy, J. F., 54 Kerner Commission, 576 Key performance indicators (KPIs), 847, 848 Khairlanji, 435 Kshatriyas, 781 Kyoto, 508 L Lab-in-the-field experiments, defined, 237 Labor force participation, 73, 74 Labor force shares, 487 Labor Force Survey (LFS), 448 Labor legislation, 71 Labor market, 107, 448, 452–454, 487, 495, 496 acquisition, 469 discrimination, 670, 800–803 outcomes, 813 policies, 814 segregation by occupation and work status, 378 standards, 82 Labor market discrimination, 3 dynamic efficiency, 78 employment rates, 74 female labour force participation, 73, 74 gender-ascriptive social relation, 70 gender discrimination and inefficiency, 77, 78 gender earnings gaps, 75 gendered occupational and sectoral segregation, 75 gender inequality, 76 income inequality between households, 76 job grading systems, 71 labor market standards, 82 macro-efficiency, 79 micro-efficiency, 78 social efficiency, 79
913 sphere of reproduction, 71 static efficiency, 78 transformatory strategies, 79 Landlords, 669, 671–673, 677–681, 683–685, 688–693 Land ownership, 483, 486 Land sales market, 900 Language communities, 197 Law of Manu, 784 Left-wing Workers Party, 395 Legislated candidate quotas, 767 Legislated reserved seats, 767 Lesbian, gay, bisexual, transgender, and intersex (LGBTI) people Australia, 316 being out, 304 big data, 318 coming out, 304 data from other countries, 310 data from the United States, 308 Europe, 315, 316 experimental approaches to measure discrimination and exclusion, 319 gender identity, 303 India, 313 intersex, 303 Nepal, 311 Nepal’s 2011 census, 311 population size estimation, 308 sexual orientation, 302 surveys, 317 Lesbian, gay, bisexual, transgender and intersex (LGBTI) individuals, 7 Lethal force, Black Americans, 605 Lethal force, 580 LGBTQ group, 115 LGBTQ movements, 98 Liberal neo-developmentalism, 394 List randomization, 183–184 Local-global” spatial constructs, 95 Local improvements, 505 Logistic regressions, 448 Logit models, 449 Logit regression, 293, 815 Lower prenatal health care and nutrition, 604 Lucy’s conversion, 427 M Machine Learning (ML), 337–340 Macro-efficiency, 79 Mahatma Gandhi National Rural Employment Guarantee Act, 331 Majlis Amanah Rakyat (MARA), 824, 830
914 Malabar region, 558 Malaysia, 495 Malaysia, New Economic Policy (NEP), 820–826, 828–837 Male chauvinism, 458 Mandatory gender-related functions, 361 Marginal workers, 675 Markets and discrimination, 107–110 Market-based affirmative action, 828 Marxist framework, 360 Maternal energy, 881 May 13th 1969 racial riots, 824 Median wage, 488 Medicaid spending, 604 Memory confusion protocol, 250 Men who have sex with men (MSM), 302, 313 Merit-based admissions, 123 Merit-based affirmative action, 828 Meritocracy, 55, 122 Meta-analysis, 627 Micro and small enterprises (MSEs), 625 Microcredit, 638 Micro-efficiency, 78 Micro-entrepreneurship training program, 627 Migrant labor, 484 Migration, 99 Minimal group paradigm (MGP), 264–270 Minority groups, 820, 822, 824 Mismatch hypothesis, 125 Misra report, 557 Missing women, 651 Mis-targeting, 124 Mobility, 93–94 Mohanty, Chandra Talpade, 95 Monte-Carlo simulation approach, 638 Mukkuvars, 427 Mulatto escape hatch, 359 Multiculturalism, 465 Multi-equations econometric modelling, 447 Multiple identities, 110 Murder, 568, 575, 576, 578, 580, 584 Muslim, 236, 671, 672, 674–678, 681, 684, 685, 691, 692 Bosniaks, 252 community, 459 economic sectors, 446 extremism, 458 face discrimination, 451 in France, 8 immigrants, 455 and Jewish women, 451 male applicant, 450
Index origin, 447 population, 446 slum-dwellers, 251 sounding names, 458 unemployment, 453 “Mystery shopping” studies, 670, 673 N Narrative data, 335–337 Narrative economics, 329 Nash equilibria, 109–111 National Capital Region (NCR), 677 National Committee for Buraku Liberation, 507 National comparison, 96 National Development Policy (NDP), 827 Nationalisms, 86, 89, 92, 94–96, 98 National Labor Relations Act (NLRA), 60 National Qualification Framework (NQF), 804 National Rural Employment Guarantee Scheme (NREGS), 12, 82, 731, 732, 739–741, 753 impacts on, 746–748 National Socio-Economic Characterization Survey (CASEN), 310 National Vision Policy, 827 Native American, 90, 91 Natural identities, 270–274 Naturalization Act of 1790, 57 Natural Language Processing (NLP), 337–340 Naz Foundation vs. NCT Delhi, 436 Need-based affirmative action, 828 Neo-colonialism, 88 Nepal, 311 Nepal and India surveys, 317 New Deal, 59–61, 64 FLSA, 60 GI Bill, 59, 60, 63 Social Security Act, 60 New Economic Policy (NEP), 625, 820, 821, 825–828, 830–834 NITI Aayog, 561 Non-targeted minorities, 734 Non-untouchable’ groups, 676 Normative implications, 114 O Oaxaca-Blinder decomposition method, 653 Oaxaca-Blinder method, 135–137 Occupation, 528 Occupational segregation, 477, 487 Omitted group problem, 138
Index Online data collection tool, 316 Online housing markets, 692 Online housing search platforms, 677 Online rental application form, 679 2016 online survey in 9 EU countries, 310 Online surveys, 317 Open-source programs, 899 Opportunity, disparity in, 603 Opportunity sets, 126 Optimal groups, 110–112 Organizational environment, 293 Organizational hierarchy, 470 Orientalism, 89 Osaka, 503 Other Backward Classes (OBC), 672, 678, 679, 682–684, 688, 689 Outcastes, 501–507 Over-education, 473 P Paid Domestic Work, 82 Palma Index, 368, 371 Panchayats Extension to Scheduled Areas Act of 1996 (PESA), 737, 753 Participation, 330, 723 Participatory-Tracking, 343 Participatory tracking, 344 Part-time work, 72 Party quotas, 767 Past studies and surveys, 317 Patriarchy, 88, 91, 94 Payment systems, 71 Pay structures, 71, 73–75, 80, 81 Peer-to-peer lending, 638 People with disabilities, 487 Perbadanan Usahawan Nasional Berhad (PUNB), 831 Performance hypothesis, 731, 735, 751 Performance-related payment systems, 71 Personal services category, 364 Peru, 246 PETI programmes, 396 Phule, Jotiba, 430 Phule, Savitribai, 425 Pink-washing, 99 Pleven Act, 459 Point system, 464 Police data on shooter bias, 237 Police discrimination, 583 Policing, 596, 598, 603, 606, 609 disparities in, 605–606 Policy achievements, 831
915 Policy dilemma, 112–113 Policy implications, 751–752 Policy platforms, 824 Political affirmative action, 730, 733, 755–757 Political imperatives, 825, 836, 837 Politically connected elite, 496 Political reservation, 533 Political resistance, 60 Population size, LGBTI people, 308 Portuguese colonization, 357 Positional objectivity, 655 Positive discrimination (PD) policies, 700 beneficiary communities eligible for preferences, 708–709 configuration of, 709–715 critics of, 700 extent of support for under-prepared beneficiaries, 714–715 identifiability of beneficiaries, 713 magnitude of preference, 711–712 quotas vs. preferences, 710–711 sensitivity of selection process, 712 sphere of applicability of, 705–708 Post-apartheid period, 486, 495, 496 Postcolonial contributions, 88–92 Post-stop outcomes, 292 Poverty, 354–356, 360, 362, 373–377, 396, 529–530, 822, 825, 832, 836 eradication, 821, 836 Pradhan Mantri Gram Sadak Yojana (PMGSY), 732, 741–743, 745, 751, 753, 754 impacts on, 748 Praxis, 97 Preemption, 578–580 Preemptive motive for killing, 579 Preferential procurement, 807, 810 Preferential selection policies, 702 Prejudice, 23, 53, 288, 572 Prejudiced discrimination, 572 Pre natal sex detection technologies, 658 Priming, 294 Probit model, 145 Process, 329 Profiling, 571 Project Implicit data, 601 Pro-White/anti-Black attitudes, 597, 599, 600, 602, 604, 605 Psycho-emotional threat, 54 Publication bias, 165 Public discourse, 824 Public funds, 511 Public goods, 732, 733, 739, 741–743, 751, 756 impacts on, 749
916 Public policies, 460 Public procurement, 831 Pygmalion effect, 192–194 Q Quantitative data, 324, 330, 333, 340–342, 345 Quantitative preferential-boost system, 711 Quasi-audit studies, 670 R Race, 360, 409–413, 416, 488, 490, 492, 495, 496, 620, 621, 623, 626 ethnicity, 86, 90 Race-based statistical discrimination, 643 Racial and ethnic discrimination, 361 Racial bias, 107, 114 Racial categories, 409 Racial democracy, 359–361, 364, 365 Racial discrimination, 286, 287, 292, 293, 296, 623, 636 Racial disparities, 574, 575, 585, 586 Racial empires, 96 Racial employment discrimination, 801 Racial heterogeneity, 90 Racial hierarchy, 90 Racial inequalities, 362 Racialization, 471 Racialized immigrants, 467, 469–477 Racially dominant group, 288 Racial profiling, 292 in police stops, 574 Racial segregation, 59 Racial stereotypes, 6 Radio Rwanda, 255 Radio station, 254, 255 Ramabai, 429 Randomized field studies, 217 Rational actors, 237 Recentered influence function (RIF) regressions, 147–148 Reflexive science, 328, 329, 344 Reflexivity, 328, 338–340, 344, 345 Regression analysis, 286 Regression-based decompositions, 619 Regression discontinuity (RD), 732 Relative inequality, 898 Relative position, 52, 54, 65 Relative status, 56, 59 Religious discrimination domains, 444 economic consequences, 460 immigration, 445, 446 increase/decrease of, 458
Index methodologies, 446–449 physical and social violences, 444 public policies, 458 social issue, 444 society’s opinion and drivers, 457, 458 socio-economic dimensions, 444 theme wise analysis, 449–457 Religious penalty, 452 Remote’ audits, 671 Renewals/revitalizations, 59 Rental market, 449 studies, 169–171 Representation, 90, 763 Republicanism, 459 “Research for Action” method, 331 Reservations, 765, 770, 771, 774, 786, 789, 790, 794 Residential segregation, 366, 669, 676 Residual approaches, 25 Resistance, 88, 93, 96 Resource-transfer programmes, 701 Reverse discrimination, 124, 845, 855 Reverse racism, 845 Reward principle, 120, 121, 124 Reweighting methods, 144–146 1918 Rice Riots, 505 Right of natural expansion, 58 Rig Veda, 780 Rukhmabai, 429 Rural Social Insurance, 396 Rwanda, 254, 255 S Sahariya, 326 Salah, Mohamed, 236 Sales department, 107 Sarvagod, Mukta, 431 Scheduled Areas, in India, 736–737, 743 analysis of balance with census data, 745–746 control variables, 743 data construction, 742–743 electoral mechanism, 752–754 geographic regression discontinuity, 744–745 impact of, 746–749 Indian identity categories, 738–739 local elections in, 753 PESA, 737 quota overlap, 754 quotas and political conflict, 737–738 targeted minority electoral influence, 753–754
Index Scheduled Castes (SC), 672, 678, 679, 682–684, 688, 689, 691, 692, 736, 738 Scheduled Castes/Tribes (SC/ST), 675 Scheduled Tribes (ST), 731, 736, 737 Schematic discrimination, 238, 239, 245, 249, 256 School discipline, 603 Science, technology, engineering, and math (STEM), 846 Scientific racism, 295 Seasonal Agricultural Workers’ Program (SAWP), 471 Secularism, 459 Sedentarist metaphysics, 94 Self-classification, 365 Self-defense, 579 Self-employment and entrepreneurship affirmative action, in government contracting, 624–625 credit access, 620–622 customer preferences, 622–624 demographic characteristics, 618–620 reducing unconscious bias, 626–627 training programs and role models, 627–628 Self-esteem, 290 Self-identify, 65, 302, 308–310, 318 Self reported health, 656 Semi-structured interviews, 448 Sensitive selection processes, 713 Servicemen’s Readjustment Act of 1944, 59 Sexuality, 99 Sexual orientation, 302–305, 307, 308, 310–314, 316, 317, 319 Shinde, Tarabai, 425 Shooter bias, 240, 241 tests, 294 Shudras, 781 Singh, Bhrigupathi, 326 Skills Development Act, 804 Slave codes, 246 Slavery, 90, 91 Small business entrepreneurship, 641 Snowball surveys, 312 Social capital, 619 Social citizenship, 471 Social cognitive process, 850 Social cognitive stereotype, 850 Social constructs, 408–409 Social Darwinism, 878 Social disparities accreditation bodies, 470 attributes, 470 disempowerment, 471 employment, 471
917 ethnic diversity, 470 ethnic minority groups, 470 exclusionary discrimination, 469 fabricated resumes, 470 gender-based analysis, 471 health profession, 471 immigration, 470 occupations, 471 organizational structure, 470 Social dissent, 506 Social efficiency, 79 Social group membership, 843 Social identity, 154, 191 Social inequalities, 903 Social Insurance Fund, 396 Social media, 318 Social mobility, 360 Social movements (Sawyer & Gampa, 2017), pathogens, 607 Social networks, 460 Social psychology, 52, 213 audit studies, 287 causal mechanisms, 286 domains, 286 economic opportunities, 286 economic stratification and stereotypes, 287–291 employer decision-making, 287 field experiments, 286 racial bias, 286 stereotypes, discrimination and policing, 292–295 wage regressions, 286 Social reproduction of caste system, 521–524 Social restructuring, 821, 825, 836 Social role theory, 289, 291 Social Security Act, 60 Socio-cognitive de-biasing strategy, 209–218 Socioeconomic disparities, 468, 469 Socioeconomic High-resolution Rural-Urban Geographic Dataset for India (SHRUG), 742 Socioeconomic inequalities, 118 Socioeconomic status (SES), 246 Sociology, 52, 108 of mobility, 94 SOGIE, 301, 307, 314 Solidarity, 95 hypothesis, 731, 732, 734, 751 Solipsistic work, 114 Son-preferring fertility-stopping rules, 652 South Africa, 8 disability, 492–493 discrimination index, 493–494
918 South Africa (cont.) employment, 487 Employment Equity, 488–491 executive and legislative representation, 492 higher educational institutions, 491 post-1994 reforms, 485–486 racial discrimination in, 483–485 wages by gender and population group, 489 South Africa’s Gini coefficient, 801 Southern strategy, 51, 54, 60 Spatial inequalities, 669, 676 Special Buraku, 503 Special position, 823 Sphere of production, 71 Srinivasan, TN, 331, 332, 334 Stable working-class black families, 327 Stanton, E.C., 872–879 Static efficiency, 78 Statistical discrimination, 3, 237, 255, 256, 287, 288, 291, 618, 622, 623, 628 in health, 654 model, 153 STEM tertiary education, 602 STEM workforce, 602 Stereotype(s), 30, 237–239, 241–244, 248, 251, 253, 255, 256, 569–570, 585 group position, 290–291 norms, 289, 290 See also Economic stratification Stereotype content model (SCM), 29 Stereotype threat, 52, 247, 849 accurate and strategic messaging of AAPs, 857–859 and affirmative action research, 853–854 consequences, 850–851 critical historical knowledge, discrimination and social inequities, 856–857 foundational stereotype threat theory research, 849–850 misconceptions and mischaracterizations of AA cues, 855–856 renewing commitments to AAPs, 859–860 social groups and affirmative action, 851–852 workplace experiences of discrimination and affirmative action, 846–848 Strategic complementarity, 108, 110, 113, 114 Strategic discrimination, 571 Stratification economics, 3, 51–54, 56, 63, 65 affinity, 56–61 application, 52 dominant group, 56 element, 53
Index framework, 51 group’s relative performance in hierarchy, 52 lens of, 53 mortality rates, 53 principles, 52 rational and functional, 53 tenets, 51 wage gaps, 53 white negativism, 61–63 Stratification theory, 287 Stree Purusha Tulana, 425 Street gender, 413 Street race, 412 Structural economics, 329 Structural inequalities, 123 Students for Fair Admissions v. Harvard, 54 Student Teams Achievement Divisions, 207 Subculture of violence, 889–892 Subtle discrimination, 844 Suffragism, 873 Suiheisha, 504 Survey of Health, Ageing and Retirement in Europe (SHARE), 656 Surveys, 302, 304, 305, 308, 311–313, 315–318, 320 Systemic discrimination, 596, 598, 599, 603, 607 System justification theory, 290 T Targeted minorities, 734 Taste-based discrimination, 18, 19, 21, 23–25, 237, 238, 244, 255, 256, 618, 624 Taste for discrimination, 286, 293 theory, 470 Taxi-driver syndrome, 472 Taylor, F. W., 55 Technological de-biasing, 218–222 Technology, 113 Tekun, 828, 831 Temporary Foreign Worker Program (TFWP), 471 Terry stops, 571 Theme wise analysis housing market, 454–455 job access, 450–452 labor market outcomes, 452–454 marriage and language skills, 455 return migration, 456, 457 “The National Poverty Study”, 340 Theory of justice inequality, 897
Index Theory of menace, 451 Third gender, 310, 311 Third world woman, 88 Three-fold decomposition, 137 1873 Timber Culture Act, 57 Tokyo, 502 Top management, 488 Traffic policing, 292, 293 Traffic stops, 601, 605 Transformatory employment policy, 79 Transgender, 303, 304, 308, 310, 311, 315 Translocational positionality, 428 Transnational activist networks, 98 Transnational advocacy network, 97 Transnational feminism, 4, 94–100 Transnational feminist practices, 94 Transnational networks, 96 Transnational opportunity structure, 98 Transnational sexuality, 98 Travancore region, 558 Trojan Horse, 65 Truecaller, 680, 681 Trump, D., 51, 54 Trust game, 262, 263, 270, 271, 274, 278–280 Tuberculosis, 653 Tversky, Amos, 237 U Ultimatum game, 262, 281, 282 Unconscious stereotypes, 155 Unemployment, 447, 453 rates, 487 Unified Health System, 396 United States, 308 Universal Declaration of Human Rights, 459 Unobserved productivity differentials, 286 Unpaid work, 72, 74, 83 Untouchability, 9, 783 Untouchables, 247 Upper-caste Hindus, 672, 674, 676–678, 684, 688, 689 Upset losses, 241 USA, 10 U.S. citizens, 416 V Vaishyas, 781 Valmiki, Omprakash, 432 Varna system, 518 Vedic Hindu culture, 782
919 Vegetarian tenants, 693 Veil of darkness test, 292 Vemula, Radhika, 424 Vertical inequality, 112, 360, 364, 366, 368, 372, 386, 388 Vertical to intersecting inequalities, 360 Veterans Administration, 59 Violence criminology, 889–892 Vishakha vs. State of Rajasthan, 433 Voluntary party quotas, 492, 494 Vote, 568 W Wage gap, 447 Wage inequality, 377, 802 Wage structure effect, 136 Wealth, 530 ownership policy sector, 831 Wealth-building mechanism, 62 Web-based audit, 671 Web-based call-tracking resources, 680 Wellbeing, 418 White ethnics, 873 “White” feminism, 91 Whiteness, 409 White women, 410 Willingness to pay, 184–186 Women and Economics, 879 Women entrepreneurs, 637 Women in Development, 87 Women in Informal Employment: Globalizing and Organizing (WIEGO), 82 Women of color, 90, 91, 411 Women’s participation in politics, 764, 767, 769 Women’s studies, 87 Working age population, 487 World Conference against Racism, 435 Y Yayasan Peneraju Pendidikan Bumiputera (YPPB), 829 Yoshikazu Kawamoto, 512 YPPB programme, 834 Yūwa, 505 Z Zenkairen, 509 Zennosuke Asada, 509 Zones of priority (ZEP), 459