New Directions in Public Opinion [3 ed.] 2019027919, 9781138483552, 9781138483569, 9781351054621, 9781351054614, 9781351054591, 9781351054607

Citation preview

Praise for Prior Editions “Berinsky has brought together a world-class group of scholars working on cutting-edge puzzles in the causes and consequences of public opinion. The book is an excellent resource for students, containing foundational insights about measurement and research design, electoral dynamics, and campaign media strategy. The largest contribution of this volume, however, may be the treasure trove of ideas for future research on the impact of group attitudes and identities in structuring our politics.” — Nicholas A. Valentino, University of Michigan “Berinsky’s masterful structuring of the book will allow it to be used as either a main text or a supplementary reader in a wide variety of courses. With just the right mix of analytical depth and topical breadth, this text will provide a solid grounding in six decades of public opinion research while also engaging and exciting students with its timely essays on pioneering new developments in the field.” — Matthew Jacobsmeier, West Virginia University “In the latest edition of Berinsky’s New Directions in Public Opinion, leading scholars of public opinion have written engaging chapters that survey recent research. Berinsky’s volume continues to be the best overview of the topic, perfect for students who want to learn about new developments.” — Gabriel Lenz, University of California-Berkeley

New Directions in Public Opinion

The 2016 elections called into question the accuracy of public opinion polling while tapping into new streams of public opinion more widely. The third edition of this well-​established text addresses these questions and adds new perspectives to its authoritative line-​up. The hallmark of this book is making cutting-edge research accessible and understandable to students and general readers. Here we see a variety of disciplinary approaches to public opinion reflected including psychology, economics, sociology, and biology in addition to political science. An emphasis on race, gender, and new media puts the elections of 2016 into context and prepares students to look ahead to 2020 and beyond. New to the third edition: • • •

Includes 2016 election results and their implications for public opinion polling going forward. Three new chapters have been added on racializing politics, worldview politics, and the modern information environment. New authors include Shanto Iyengar, Michael Tesler, Vladimir E. Medenica, Erin Cikanek, Danna Young, Jennifer Jerit, and Jake Haselswerdt.

Adam J. Berinsky is the Mitsui Professor of Political Science at Massachusetts Institute of Technology and Director of the MIT Political Experiments Research Lab. He is the author of In Time of War:  Understanding American Public Opinion from World War II to Iraq and Silent Voices: Public Opinion and Political Participation in America. He has received numerous awards for his research.

New Directions in American Politics

The Routledge series New Directions in American Politics is composed of contributed volumes covering key areas of study in the field of American politics and government. Each title provides a state-​of-​the-​art overview of current trends in its respective subfield, with an eye toward cutting-edge research accessible to advanced undergraduate and beginning graduate students. While the volumes touch on the main topics of relevant study, they are not meant to cover the “nuts and bolts” of the subject. Rather, they engage readers in the most recent scholarship, real-​world controversies, and theoretical debates with the aim of getting students excited about the same issues that animate scholars. Titles in the Series: New Directions in American Political Parties Edited by Jeffrey M. Stonecash New Directions in the American Presidency, Second Edition Edited by Lori Cox Han New Directions in Campaigns and Elections Edited by Stephen K. Medvic New Directions in Congressional Politics Edited by Jamie L. Carson New Directions in Public Opinion, Second Edition Edited by Adam J. Berinsky New Directions in Judicial Politics Edited by Kevin T. McGuire New Directions in American Politics Edited by Raymond J. La Raja New Directions in Interest Group Politics Edited by Matt Grossmann

New Directions in Public Opinion

Third Edition

Edited by Adam J. Berinsky

Third edition published 2020 by Routledge 52 Vanderbilt Avenue, New York, NY 10017 and by Routledge 2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business © 2020 Taylor & Francis The right of the Adam J. Berinsky to be identified as the author of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. First edition published by Routledge 2011 Second edition published by Routledge 2016 Library of Congress Cataloging-​in-​Publication Data Names: Berinsky, Adam J., 1970– editor. Title: New directions in public opinion / edited by Adam J. Berinsky. Description: Third edition. | New York, NY : Taylor & Francis, 2020. | Series: New directions in American politics | Includes bibliographical references and index. | Summary: “The 2016 elections called into question the accuracy of public opinion polling while tapping into new streams of public opinion more widely. The third edition of this well-established text addresses these questions and adds new perspectives to its authoritative line-up. The hallmark of this book is making cutting edge research accessible and understandable to students and general readers. Here we see a variety of disciplinary approaches to public opinion reflected including psychology, economics, sociology, and biology in addition to political science. An emphasis on race, gender, and new media puts the elections of 2016 into context and prepares students to look ahead to 2020 and beyond. New to the Third Edition: Includes 2016 election results and their implications for public opinion polling going forward. Three new chapters have been added on racializing politics, worldview politics, and the modern information environment. New authors include Shanto Iyengar, Michael Tesler, Vladimir E. Medencica, Erin Cikanek, Danna Young, Jennifer Jerit and Jake Haselswerdt”– Provided by publisher. Identifiers: LCCN 2019027919 | ISBN 9781138483552 (hardback) | ISBN 9781138483569 (paperback) | ISBN 9781351054621 (ebook) | ISBN 9781351054614 (adobe pdf) | ISBN 9781351054591 (mobi) | ISBN 9781351054607 (epub) Subjects: LCSH: Public opinion–United States. | United States–Politics and government–Public opinion. Classification: LCC HN90.P8 N49 2020 | DDC 303.3/80973–dc23 LC record available at https://lccn.loc.gov/2019027919 ISBN: 978-​1-​138-​48355-​2  (hbk) ISBN: 978-​1-​138-​48356-​9  (pbk) ISBN: 978-​1-​351-​05462-​1  (ebk) Typeset in Minion Pro by Newgen Publishing UK

Contents

List of Illustrations  List of Contributors  Acknowledgments  Introduction and Overview  Adam J. Berinsky

x xii xx 1

PART I

The Meaning and Measurement of Public Opinion 

19

1

The Practice of Survey Research: Changes and Challenges  D. Sunshine Hillygus

21

2

Citizen Competence and Democratic Governance  Martin  Gilens

41

PART II

Foundations of Political Preferences 

73

3

Ideology and Public Opinion  Christopher M. Federic o

75

4

Affective Polarization or Hostility Across the Party Divide: An Overview  Shanto Iyengar

99

viii Contents

5 Racial Attitudes and American Politics  Michael Tesler

118

6 Race, Ethnicity, and Public Opinion  Erica Cz aja and Vl adimir E. Medenica

137

7 Categorical Politics in Action: Gender and the 2016 Presidential Election  D onald Kinder, Molly E. Reynolds, AND NANCY BURNS

159

8 Worldview Politics  Marc Hetherington

177

9 The Emotional Foundations of Democratic Citizenship  Ted Brader and Erin Cikanek

202

10 No Longer “Beyond our Scope”: The Biological and Non-​Conscious Underpinnings of Public Opinion  Frank J. Gonz alez, John R. Hibbing, and Kevin B. Smith

236

PART III

The Public and Society 

259

11 The “Daily Them”: Hybridity, Political Polarization and Presidential Leadership in a Digital Media Age  Mat thew A. Baum and Dannagal G. Young

261

12 How People Learn About Politics: Navigating the Information Environment  Jennifer Jerit

282

13 Campaigns and Elections  John Sides and Jake Haselswerdt

297

14 Ambivalence in American Public Opinion About Immigration  315 Deb orah J. Schildkrau t

Contents ix

15 Public Opinion and Public Policy  Andrea L ouise Campbell and Eliz abeth  Rigby

338

Conclusion: Assessing Continuity and Change  David O. Sears

363

Index 

386

List of Illustrations

Figures 1.1 The Tital Survey Error Perspective  22 1.2 Steps in Survey Process  25 2.1 The Preference/​Policy Link for Respondents at the 10th, 50th, and 90th Income Percentiles  60 2.2 The Preference/​Policy Link when Preferences at the 10th or 50th Income Percentiles Diverge from the 90th Income Percentile  61 3.1 What Attracts People to Different Ideological Positions?  79 3.2 The Relationship between Authoritarianism and Ideology is Stronger among those High in Political Information  90 4.1 ANES Party Feeling Thermometers (1976–​2016)  101 4.2 Out-​Party Animus in Feeling Thermometer Scores  101 5.1 Distribution of White Racial Attitudes  120 5.2 Whites’ Support for Presidential Candidates by Racial Attitudes  124 5.3 Democratic Identification among Non-​College Educated Whites by Racial Attitudes, 1986–​2012  128 5.4 Views of Race-​Related Issues by Party, 2000–​2018  129 7.1 Democratic Presidential Vote by Gender 1980–​2016  163 7.2 Turnout in Presidential Elections by Gender 1964–​2016  165 7.3 Women’s Vote for Clinton by Gender Identification  167 7.4 Vote for Clinton by Modern Sexism and by Women’s Role  170 8.1 Worldview and Partisanship, 1992 versus 2016  186 8.2 Worldview and Racial Attitudes  190 8.3 Worldview and Gender Attitudes  191 8.4 Worldview and Immigration  191 8.5 Party Feeling Thermometers, Carter Through Obama  197 9.1A Emotional Reactions to Democratic Presidential Nominees, 1980–​2016  206 9.1B Emotional Reactions to Republican Presidential Nominees, 1980–​2016  206 9.2 News Stories Elicit Distinct Patterns of Emotional Reactions  208

Illustrations xi

9.3 Fear Cues in Political Advertising Trigger Greater Attention and Information-seeking for Relevant Information  9.4 Fear Cues in Political Advertising Weaken the Impact of Predispositions on Voting Decisions, while Enthusiasm Cues Strengthen the Impact of Predispositions  10.1 Upper Panel—​Traditional View of Public Opinion  Lower Panel—​Biologically Informed View of Public Opinion  11.1 Partisan Trend in “Regular” Viewing of Network TV News  11.2 Trend in Partisan Viewing of CNN, Fox, and MSNBC  12.1 Knowledge of Clinton Proposals on Health Care from 1997 State of the Union Address  13.1 Economic Growth and Presidential Election Outcomes, 1952–​2016  14.1 Foreign-born as Percentage of U.S. Population, 1850–​2016  14.2 Immigrants in the United States by Region of Birth, 1960–​2016  14.3 In Your View, Should Immigration Be Kept at Its Present Level, Increased or Decreased?  14.4 Public Opinion on a Path to Citizenship for Unauthorized Immigrants, 2007–​2017  14.5 Public Opinion on a Path to Citizenship by Selected Demographic Groups, 2016  14.6 Public Opinion on Building a Wall on the US–​Mexico Border by Selected Demographic Groups, 2016 

211 218 241 265 269 286 301 317 317 319 320 324 325

Tables 9.1 Some Appraisals and Action Tendencies of Common Emotions  9.2 The Impact of Emotions on Political Participation in the 2008 Election  9.3 The Impact of Emotions on Political Participation in the 2012 Election  9.4 The Impact of Emotions on Political Participation in the 2016 Election  12.1 Knowledge of Current Events in 2017 Pew Research Center Poll, by Education Level 

204 221 222 223 284

Contributors

Matthew A. Baum is the Marvin Kalb Professor of Global Communications and Professor of Public Policy at Harvard University’s John F.  Kennedy School of Government and Department of Government. His research focuses on domestic political influences on international conflict and cooperation in general and American foreign policy in particular, as well as on the role of the media and public opinion in contemporary American politics. His research has appeared in over a dozen leading scholarly journals, such as Science, the American Political Science Review, American Journal of Political Science, and the Journal of Politics. His books include Soft News Goes to War: Public Opinion and American Foreign Policy in the New Media Age (2003, Princeton University Press), War Stories: The Causes and Consequences of Public Views of War (2009, Princeton University Press, co-​authored with Tim Groeling), and War and Democratic Constraint: How the Public Influences Foreign Policy (2015, Princeton University Press, co-​authored with Phil Potter). He has also contributed op-​ed articles to a variety of newspapers, magazines, and blog sites in the United States and abroad. Before coming to Harvard, he was an associate professor of political science and communication studies at UCLA. Adam J.  Berinsky is the Mitsui Professor of Political Science at MIT. He received his Ph.D.  from the University of Michigan. He is a specialist in the fields of political behavior and public opinion with over 20  years of experience in survey design and analysis. He is the author of In Time of War:  Understanding American Public Opinion from World War II to Iraq (2009, University of Chicago Press) and Silent Voices:  Public Opinion and Political Participation in America (2004, Princeton University Press) and has published articles in many academic journals. He has won several scholarly awards, is the recipient of multiple grants from the National Science Foundation, was a fellow at the Center for Advanced Study in the Behavioral Sciences, and has been a John Simon Guggenheim Memorial Foundation Fellow. He is also the founding director of the MIT Political Experiments Research Lab.

Contributors xiii

Ted Brader is Professor of Political Science at the University of Michigan and Research Professor in the Center for Political Studies at the Institute for Social Research. He is Principal Investigator of the American National Election Studies and Associate Principal Investigator for Time-​ sharing Experiments for the Social Sciences. His research focuses on the role of emotions in politics, political partisanship, media effects on public opinion, experimental and survey methods, and other topics in political psychology. He is the author of Campaigning for Hearts and Minds (2006, University of Chicago Press) and numerous scientific articles. He received his Ph.D. from the Department of Government at Harvard University in 1999. Nancy Burns is the Warren E. Miller Collegiate Professor of Political Science, Chair of the Department of Political Science, and Research Professor at the Center for Political Studies at the University of Michigan. Her work focuses on gender, race, public opinion, and political action. She is the co-​author with Kay Lehman Schlozman and Sidney Verba of The Private Roots of Public Action:  Gender, Equality, and Political Participation (2001, Harvard University Press) and the author of The Formation of American Local Governments: Private Values in Public Institutions (1994, Oxford University Press). Her work has appeared in the American Political Science Review, the American Journal of Political Science, the Journal of Politics, Political Analysis, Politics & Gender, Studies in American Political Development, and the Annual Review of Political Science, among other outlets. Burns served as Principal Investigator of the National Election Studies from 1999 to 2005. She is a Fellow of the American Academy of Arts and Sciences. Andrea Louise Campbell is Arthur and Ruth Sloan Professor of Political Science at the Massachusetts Institute of Technology. She is the author of Trapped in America’s Safety Net: One Family’s Struggle (2014, University of Chicago Press), How Policies Make Citizens:  Senior Citizen Activism and the American Welfare State (2003, Princeton University Press) and, with Kimberly J. Morgan, The Delegated Welfare State: Medicare, Markets, and the Governance of Social Provision (2011, Oxford University Press). Her research has appeared in the American Political Science Review, Political Behavior, Comparative Political Studies, Politics & Society, Studies in American Political Development, and Health Affairs, among others, and has been funded by the National Science Foundation, Robert Wood Johnson Foundation, and Russell Sage Foundation. She was elected to the American Academy of Arts and Sciences in 2016 and the National Academy of Social Insurance in 2007 and served on the National Academy of Sciences Commission on the Fiscal Future of the United States. She holds a Ph.D. in Political Science from UC Berkeley. Erin Cikanek is a Ph.D. student in political science at the University of Michigan. Her main areas of interest are political psychology, public opinion, and

xiv Contributors

methods. Her current work focuses on better measuring emotional signals in news media and the impact of these signals on the American electorate. Erica Czaja is the Director of the Community Research Institute in the Dorothy A. Johnson Center for Philanthropy at Grand Valley State University. She earned her Ph.D.  in Politics and Social Policy from Princeton University (2013), and was a Robert Wood Johnson Foundation Scholar in Health Policy Research at the University of California, Berkeley. Her research has been supported by the National Science Foundation, the Mamdouha S. Bobst Center for Peace and Justice, and the Center for the Study of Democratic Politics at Princeton University. Her doctoral dissertation, Revolutionary Emotion: Empathy and Equality in the United States, examines the effects of empathy across difference on Americans’ social policy preferences. She is the recipient of the American Political Science Association’s Distinguished Junior Scholars Award in political psychology, and she has taught courses at Hope College and Princeton University on race politics, political psychology, gender politics, and campaigns and elections. She also serves on the Board of Directors for the Lakeshore Ethnic Diversity Alliance and on the Institutional Review Board at Calvin College. Christopher M. Federico is Professor of Political Science and Psychology at the University of Minnesota, and he is affiliated with the University’s Center for the Study of Political Psychology. His research focuses on the nature of belief systems in mass publics; the psychological foundations of ideology, partisanship, and issue preferences; political expertise and political engagement; and intergroup attitudes. He is the recipient of several awards, including the 2007 Erik Erikson Award for Early Career Research Achievements and the 2018 David O. Sears Book Award (with Christopher Johnson and Howard Lavine) from the International Society of Political Psychology (ISPP). He is currently a Vice President of ISPP, and a former President of the American Political Science Association’s Organized Section on Political Psychology from 2008 to 2009. Martin Gilens is Professor of Public Policy at UCLA. His research examines representation, public opinion, and mass media, especially in relation to inequality and public policy. He is the author of Affluence & Influence:  Economic Inequality and Political Power in America (2012, Princeton University Press) and Why Americans Hate Welfare:  Race, Media and the Politics of Antipoverty Policy (1999, University of Chicago Press), and co-​author (with Benjamin I.  Page) of Democracy in America? What Has Gone Wrong and What We Can Do about It (2017, University of Chicago Press). He has published widely on political inequality, mass media, race, gender, and welfare politics. He earned a Ph.D.  in sociology at the University of California, Berkeley, and has held fellowships at the Institute for Advanced Study in Princeton, the Center for Advanced Study

Contributors xv

in the Behavioral Sciences at Stanford, and the Russell Sage Foundation. He is a member of the American Academy of Arts and Sciences, and taught in the political science departments at Yale, UCLA, and Princeton universities before joining the UCLA Luskin School of Social Welfare in 2018. Frank J. Gonzalez is an Assistant Professor in the School of Government and Public Policy at the University of Arizona. He received his Ph.D. from the Department of Political Science at the University of Nebraska-​ Lincoln. His research uses social psychology, biology, and neuroscience to understand group-​related social and political phenomena such as racial prejudice, national identity, transnational policy attitudes, and views regarding class and inequality. His work has been funded by the National Science Foundation, and in 2018 he received the Best Dissertation Award from the International Society of Political Psychology. His published work can be found in peer-​reviewed journals in political science as well as psychology. Jake Haselswerdt is an Assistant Professor in the Department of Political Science and the Truman School of Public Affairs at the University of Missouri. His research focuses on the politics of public policy in the United States, particularly health and social welfare policy. He earned his Ph.D. from George Washington University, and is an alumnus of the Robert Wood Johnson Scholars in Health Policy Research Program and the American Political Science Association Congressional Fellowship Program. Marc Hetherington is the Raymond Dawson Bicentennial Professor of Political Science at the University of North Carolina-​Chapel Hill. He is the author of several books, the most recent of which is Prius or Pickup: How the Answers to Four Simple Questions Explains America’s Great Divide (co-​authored with Jonathan Weiler; 2018, Houghton Mifflin). He has published articles in a wide range of scholarly journals, mostly about American public opinion. Over the years, he’s also won a number of undergraduate teaching awards. John R.  Hibbing is the Foundation Regents Professor of Political Science and Psychology at the University of Nebraska-​Lincoln, where he studies the interaction of biology and politics. He has been named a Guggenheim Fellow, a NATO Fellow in Science, a Senior Fulbright Fellow, and a Fellow in the American Association for the Advancement of Science (AAAS). His articles have been published in journals such as Science, Behavioral and Brain Sciences, Current Biology, and the American Political Science Review and he is the co-​author of numerous books including Stealth Democracy (2002, Cambridge University Press) and Predisposed: Liberals, Conservatives, and the Biology of Political Differences (2014, Routledge), and his work has been featured in popular outlets such as The New York Times, The Washington Post, Star Talk, NPR, Fox News, the BBC, the Hidden Brain, and The Daily Show.

xvi Contributors

D. Sunshine Hillygus is Professor of Political Science and Public Policy and Director of the Initiative on Survey Methodology at Duke University. She is co-​author of The Hard Count: The Political and Social Challenges of Census Mobilization (2006, Russell Sage Foundation), The Persuadable Voter: Wedge Issues in Presidential Campaigns (2008, Princeton University Press), and Making Young Voters:  Converting Civic Attitudes into Civic Action (forthcoming, Cambridge University Press). She serves as Associate Principal Investigator of the American National Election Study, associate editor of Political Analysis, and Chair of the Advisory Committee for Public Opinion Quarterly. She earned her Ph.D. in political science from Stanford University in 2003. Shanto Iyengar is Professor of Political Science at Stanford University and Co-​ Principal Investigator of the American National Election Studies. His areas of interest include mass communication, polarization, and experimental design. His most recent book is Media Politics:  A Citizen’s Guide (2019, Norton). Jennifer Jerit is Professor of Political Science at Stony Brook University. She has interests in public opinion and political communication, with a focus on the features of news coverage that influence whether people learn about politics. Several of her current projects examine best practices for the measurement of public opinion through survey and experimental methods. She received the Erik Erikson Early Career Award for Excellence and Creativity in the field of Political Psychology and her work has been funded by the National Science Foundation and the Economic and Social Research Council. Her research has appeared in the American Political Science Review, American Journal of Political Science, Journal of Politics, Public Opinion Quarterly, and other scholarly journals. She received her Ph.D. from the University of Illinois, Urbana-​Champaign. Donald Kinder is the Philip E. Converse Distinguished University Professor at the University of Michigan. He has held fellowships from the Guggenheim Foundation and the Center for Advanced Study in the Behavioral Sciences and is a Fellow of the American Academy of Arts and Sciences and the National Academy of Sciences. His books include News that Matters (with Shanto Iyengar; 1987, University of Chicago Press), Divided by Color (with Lynn Sanders; 1996, University of Chicago Press), Us Against Them (with Cindy Kam, 2009, University of Chicago Press), The End of Race? (with Allison Dale-​Riddle; 2011, Yale University Press), and Neither Liberal nor Conservative (with Nathan Kalmoe; 2017, University of Chicago Press). Vladimir E.  Medenica received his Ph.D.  in Politics and Social Policy from Princeton University and is currently an Assistant Professor of Political Science at the University of Delaware. Prior to joining the University of Delaware, he served as a Postdoctoral Scholar at the University of Chicago

Contributors xvii

working on the GenForward Survey, a pioneering research endeavor examining how race and ethnicity structure the attitudes and experiences of young adults in the United States. His work has appeared in both peer-​reviewed journals as well as mainstream media outlets including The Washington Post, Vox, and The Hill. Molly E. Reynolds is a senior fellow in Governance Studies at Brookings. She studies Congress, with an emphasis on how congressional rules and procedure affect domestic policy outcomes. She is the author of the book, Exceptions to the Rule: The Politics of Filibuster Limitations in the U.S. Senate (2017, Brookings Institution Press), which explores creation, use, and consequences of the budget reconciliation process and other procedures that prevent filibusters in the U.S. Senate. She received her Ph.D. in political science and public policy from the University of Michigan and her A.B. in government from Smith College. Elizabeth Rigby is an Associate Professor of Public Policy and Political Science in the Trachtenberg School of Public Policy and Public Administration at George Washington University where she teaches courses on the role of politics in the policymaking process. Her research examines the interplay of politics, policy, and social inequality, particularly as it impacts vulnerable children and families. She holds a Ph.D.  in Politics and Education from Columbia University. In addition, she received post-​doctoral training in population health at the University of Wisconsin-​Madison as a Robert Wood Johnson Health and Society Scholar. Deborah J. Schildkraut is a Professor of Political Science at Tufts University. She received her Ph.D.  from Princeton University and her B.A.  from Tufts University. She is the author of Americanism in the Twenty-​First Century:  Public Opinion in the Age of Immigration (2011, Cambridge University Press), Press “One” for English: Language Policy, Public Opinion, and American Identity (2005, Princeton University Press), and The Challenge of Democracy:  American Government in Global Politics (with Ken Janda, Jeff Berry, Jerry Goldman, and Paul Manna; 2018, Cengage Learning). Her research examines the implications of the changing ethnic composition of the United States on public opinion in a variety of domains. David O. Sears received his A.B. in History at Stanford University (1957) and his Ph.D. in Psychology at Yale University (1962). He began as an Assistant Professor in Psychology at UCLA (1961). He remains there as Distinguished Research Professor and Professor Emeritus of Psychology and Political Science, along the way serving as Dean of Social Sciences (1983–​1993) there and as President of the International Society of Political Psychology. His books include Public Opinion (1964, Prentice Hall), The Politics of Violence:  The New Urban Blacks and the Watts Riot (1973, University Press of America), Tax Revolt:  Something for Nothing in California (1982,

xviii Contributors

Harvard University Press), Political Cognition (1986, Erlbaum), Racialized Politics: The Debate about Racism in America (2000, University of Chicago Press), The Diversity Challenge: Social Identity and Intergroup Relations on the College Campus (2008, Russell Sage Foundation), Obama’s Race:  The 2008 Election and the Dream of a Post-​Racial America (2010, University of Chicago Press), American Identity and the Politics of Multiculturalism (2014, Cambridge University Press), 12 editions of Social Psychology (1970–​ 2006), and the Oxford Handbook of Political Psychology (2003, 2013, Oxford University Press). John Sides is Professor of Political Science at Vanderbilt University. He studies political behavior in American and comparative politics. He is an author of Identity Crisis:  The 2016 Presidential Campaign and The Battle for the Meaning of America (2018, Princeton University Press), The Gamble: Choice and Chance in the 2012 Election (2013, Princeton University Press), and Campaigns and Election: Rules, Reality, Strategy, Choice (2013, W. W. Norton). He helped found and serves as Editor-​in-​Chief of The Monkey Cage, a site about political science and politics at the Washington Post. Kevin B. Smith is the Leland J. and Dorothy H. Chair of Political Science, and Chair of the Department of Political Science at the University of Nebraska-​ Lincoln. His research is focused on the biology and psychology of political attitudes and behaviors. His research has been published in Science, Behavioral and Brain Sciences, Physiology and Behavior, the American Political Science Review, the American Journal of Political Science, and the Journal of Politics. Michael Tesler is Associate Professor of Political Science at UC Irvine. He is author of Post-​Racial or Most Racial? Race and Politics in the Obama Era (2016, University of Chicago Press), co-​author with David O.  Sears of Obama’s Race:  The 2008 Election and the Dream of a Post-​Racial America (2010, University of Chicago Press), and co-​author with John Sides and Lynn Vavreck of Identity Crisis:  The 2016 Presidential Campaign and the Battle for the Meaning of America (2018, Princeton University Press). Dannagal G. Young received her Ph.D. from the University of Pennsylvania’s Annenberg School for Communication in 2007 and is an Associate Professor University of Delaware where she studies media and political psychology. Her book Irony and Outrage:  The Polarized Landscape of Rage, Fear, and Laughter in the U.S. (2019, Oxford University Press) examines the genres of satire and outrage as the logical extensions of the respective psychological profiles of liberals and conservatives. She has published over 40 academic articles and book chapters, and her research has been covered by the popular press, including The Atlantic, The New  York Times, Variety, and National Public Radio. She is co-​editor of two books, A Crisis of Civility:  Political Discourse and its Discontents (2018, Routledge) and Breaking Boundaries

Contributors xix

in Political Entertainment Studies (2013, USC Annenberg Press). She is a Research Fellow with the University of Delaware’s Center for Political Communication and was awarded the University of Delaware’s Excellence in Teaching Award in 2014. She is a Distinguished Fellow of the University of Pennsylvania’s Annenberg Public Policy Center and a Researcher with the National Institute for Civil Discourse.

newgenprepdf

Acknowledgments

An edited volume is only as good as the chapters it contains. Every one of the authors who contributed to this volume produced an excellent chapter. The final product is a testament to their hard and careful work. Above all, I am grateful to them. I am also grateful to Michael Kerns who approached me to write the first edition of this manuscript and guided it seamlessly through the editorial process. I also thank Jennifer Knerr for guiding this most recent edition through the editorial process. Seth Dickinson and Timothy McKinley both provided outstanding research assistance for the first edition. Daniel Guenther provided outstanding help for the second edition. Kathryn Treder and Anna Weissman did the same for this most recent edition. Gabe Lenz, Michele Margolis, and Michael Sances provided helpful comments on drafts of the introduction. This book is dedicated to my former advisors Don Kinder and Nancy Burns, who sparked my interest in public opinion while I was a graduate student at Michigan (and provided a timely chapter for this volume to boot).

 

Introduction and Overview Adam J. Berinsky

Why do some citizens approve of the health care reforms passed under the 2010 Affordable Care Act, while others are vehemently opposed? Why can’t Democrats and Republicans seem to agree on anything? Should we trust opinion polls? Are some polls better than others? How much does the average American really care about politics anyway? These are the types of important questions that are central to the study of public opinion. The field of public opinion is one of the most diverse in political science. Over the last 70 years, scholars have drawn upon the disciplines of psychology, economics, sociology, and even biology to answer these questions and more. As a field we have learned a great deal about how ordinary people come to understand the complicated business of politics. This diversity makes the study of public opinion an especially interesting area of political science, but diversity comes with a cost. Much path-​breaking research in the field of public opinion is published in professional journals, whose target audience is professors and advanced graduate students. These papers take up fairly narrow questions one at a time and often require advanced statistical knowledge to understand. As a result, the study of public opinion can seem confusing and incoherent to novices and specialists alike. This book is the third edition of the New Directions in Public Opinion. As with the first and second editions, it is intended to expand the audience for cutting-​ edge public opinion research by providing an accessible and coherent overview of the current state of the field. In this volume, I have brought together leading scholars of public opinion to provide an understandable introduction to new and exciting research. Each of the authors of the chapters in this volume both provide an overview of their field of specialization as well as a lively review of their own research. In order to stimulate interest among readers and students, these examples focus on contemporary and ongoing political controversies. My hope is that this book will serve as a comprehensive introduction to the field, while at the same time piquing the interest of students to further explore the frontiers on their own. Each chapter is intended to be self-​contained, but together, this volume serves as a gateway to the study of public opinion.

2  Adam J. Berinsky

The Measure and Meaning of Public Opinion Before we can begin to study public opinion, we need to have some sense of what we mean by “public opinion.” As you might expect, this is easier said than done. As the eminent political scientist V. O. Key aptly noted 50 years ago, “To speak with precision of public opinion is a task not unlike coming to grips with the Holy Ghost.”1 Indeed, the term has long been used to mean many different things. The ancient Greeks assessed opinion through public rhetoric and oratory, embracing a notion of public opinion that befit their restrictive notions of citizenship.2 With the emergence of expanded male suffrage throughout the nineteenth century, a notion of public opinion as an aggregation of the preferences of individuals began to take hold. But this view did not go unchallenged. In the 1940s, sociologist Herbert Blumer argued that public opinion was a purely collective phenomenon. Public opinion emerged through the communication and clash of group interests and, as a result, could not be gauged through individual survey responses. It was, instead, “the product of a society in operation.”3 It is no surprise then, that a review of the academic literature on public opinion in the 1960s uncovered scores of different definitions of this central concept.4 While it may be difficult to come to a consensus on a single definition of public opinion, Key ultimately arrived at a working view of public opinion—​ one that is a useful starting point for this book: “Those opinions held by private persons which governments find it prudent to heed.”5 Key’s definition is an expansive one. Public opinion is a property of individuals, but acquires its power in the public sphere. Moreover, there is a place both for the strongly formed, crystallized opinions of citizens that we might think of as reasoned public opinion and the lightly held beliefs and transient preferences that are decried by politicians and journalists as fickle judgments, but which sometimes guide government. Once we have defined public opinion, we must figure out how best to measure it. Almost certainly, the most familiar technique for the readers of this volume is opinion polls or surveys. Over the course of the twentieth century, polls emerged as an important tool to measure the public will. Today, polls pervade the political scene. Indeed, writing for the 50th anniversary of the leading journal for public opinion research, Public Opinion Quarterly, noted political scientist Philip Converse (whose crucial contributions to the field are discussed below) argued that opinion polls were public opinion. This view is at least implicitly shared by nearly all scholars of public opinion. Indeed, in this volume, when scholars speak of public opinion, they almost always mean the results of polls. Given that polls are the dominant way to measure opinion, practitioners should be careful about how they conduct surveys—​and citizens should be careful about how they interpret them. In Chapter  1, D.  Sunshine Hillygus describes the key challenges that face those who wish to conduct and use surveys

Introduction and Overview  3

as a measure of public opinion. Above all, Hillygus reminds us that we need to pay close attention to the quality of the data gathered by particular surveys. Data quality is especially important in the twenty-​first century, because we are undergoing great changes in the way we measure opinion through surveys. In the early days of surveys, pollsters went door-​to-​door to interview survey respondents. Beginning in the 1970s, the industry started conducting polls by telephone, a cheaper and more convenient form of data collection that facilitated an explosion in the number of polls in the United States. In recent years, however, this model of data collection has been threatened on several fronts. For one, the rise in the use of cell phones puts the representativeness of phone polls at risk. Over half of all Americans have dropped their land lines in favor of a cell phone and these numbers are growing, especially among younger Americans. As of December 2017, more than three-​quarters of Americans aged 25 to 29 lived in households that had a cell phone but no traditional land line telephone.6 Polling exclusively through land lines will therefore miss a large portion of the citizens that comprise the mass public. In addition, the Internet holds promise as a method of data collection but, as Hillygus notes, it is difficult to translate the time-​tested mechanics of collecting opinions to this new medium. For instance, it is almost impossible to define the universe of Internet users, making it a challenge to draw representative samples. With the future of opinion polling uncertain, scholars and practitioners must be even more careful about how they collect and report measures of public opinion. Consider, for example, the reporting of poll results by the media. Traditionally, media outlets such as The New York Times and The Washington Post (in conjunction with CBS and ABC, respectively) maintained polling units that would conduct polls across the course of a campaign. In the last few election cycles, websites that aggregate all available polling information—​such as Real Clear Politics and Nate Silver’s FiveThirtyEight—​have become extremely popular. These sites provide polling consumers with a vast array of information. But this information is not all of equal quality. Some surveys, Hillygus reminds us, are more worthy of trust than others. In the end, as Hillygus argues, no survey is perfect, but by making the decisions that go into collecting survey data more transparent, it is possible for policymakers, journalists, and ordinary citizens to decide how much faith to put into any single measure of the public will.

The Question of Democratic Competence Even if we draw a high-​quality sample and conduct our poll to minimize survey errors, sometimes it is not clear what it is that we can measure with opinion polls. If, as Walter Lippmann once wrote, politics is a “swarming confusion of problems,” can ordinary people make sense of it?7 Perhaps, when confronted with a complex world, ordinary citizens simply don’t think about politics in a coherent manner—​or even at all.

4  Adam J. Berinsky

Lippmann did not have much faith in the common man. And he was not alone in his beliefs: for most of American history politicians and commentators have been skeptical that the mass public is—​or ever could be—​engaged enough with the political world to make reasonable and temperate political decision. There was a reason, after all, that the Founders established a representative democracy rather than let citizens directly govern themselves. However, it was not until the early twentieth century that the capability of the citizenry became a pressing concern. In the early republic, the political sphere was restricted in numerous ways to keep “undesirables” away from the political process. But gradually more and more citizens were brought into the political sphere. With the expansion of the franchise to women in 1920, all Americans (at least nominally) had the right to vote. The capacity of the mass public, then, became a critical subject of interest to politicians and scholars alike. A. Lawrence Lowell, for example, argued that individual differences in political knowledge must be taken into account when determining the proper role of the public in the process of government—​some people, according to Lowell, should count more than others.8 To return to Key’s conception of public opinion, Lowell’s worry was that without some mechanism to filter opinion, public opinion would reflect only lightly held views and ill-​though-​out transient demands. The early days of opinion polling confirmed the worst suspicions of the skeptical. Survey researchers in the 1930s and 1940s painted a rather bleak picture of the capabilities of the mass public. Summing up the findings of one of the first large-​scale academic studies of voter decision-​making in the 1948 election, Bernard Berelson, Paul Lazarsfeld, and William McPhee wrote that their surveys “reveal that certain requirements commonly assumed for the successful operation of democracy are not met by the behavior of ‘average’ citizen.”9 For instance, in 1947, one-​quarter of the public could not even name the Vice President and 40  percent did not know who controlled the Senate, essentially failing a multiple choice question with two options. The passage of time has done little to alleviate these concerns. Even with a remarkable increase in the educational attainment of the average American and the rapid development of the mass media since in the 1940s, American citizens remain largely ignorant of the goings-​on in the political world. Moreover, survey researchers have consistently found that even minor differences in the wording of questions can dramatically change the shape of opinions expressed in polls. Americans support “assistance to the poor” but reject spending on “welfare.” It is a legitimate question, then, whether American citizens hold meaningful policy preferences worthy of serving as the basis of democratic governance. Martin Gilens addresses this very question in Chapter  2. As he concedes, much evidence suggests that citizens lack the knowledge or motivation to form sensible policy preferences. Few Americans, for example, understand the content of complex legislation like the health care reforms that passed under President Obama. But, as Gilens notes, individuals do not navigate the political

Introduction and Overview  5

world on their own. For instance, citizens can draw on cues from prominent politicians to help them form meaningful attitudes. And while any individual might fall short of the democratic ideal, groups of citizens can be drawn together into a meaningful aggregate. As James Stimson has argued, once we think of public opinion as the property of whole electorates, rather than individuals, it appears orderly and functions as a sensitive barometer to events in the political world.10 Aggregation does not, however, solve all the problems of public opinion. The aggregation of disparate preferences may lead to a rational and reasonable opinion, but in practice not all individuals have an equal voice. For instance, Gilens finds that the policies enacted by government reflect aggregate public preferences, but this policy is strongly biased toward the preferences of the most economically advantaged Americans. Still, Gilens does not blame the public; the inequality in democratic responsiveness, he concludes, results not from a failure of the broader public to form meaningful preferences, but from the failure of the political decision makers to take those preferences into account.

The Foundations of Political Preferences That said, the opinions of individual members of society remain a critical object of study. Shifting the focus to the aggregate level, as Gilens argues we should, might address some of the concerns raised by critics of direct democracy, such as Berelson and his colleagues. But it remains a fact that even if the American public, taken as a whole, can reason effectively, individual citizens are often distracted by more pressing concerns. Setting aside the noble impulse to accept the wisdom of the collective, to fully understand the relationship between the mass public and politicians, we must not lose our focus on individual citizens. After all, as Christopher Achen aptly noted, if individuals do not possess even meaningful attitudes, let  alone well-​defined policy preferences, then “democratic theory loses its starting point.”11 A great deal of public opinion research therefore tries to understand exactly how people do reason about politics. Answering this question is the task of the second section of this book.

Ideology and Political Reasoning For the first few decades of the academic study of public opinion, the search for the principles that shaped public opinion did not stray far from ideology. From the 1950s through the 1970s, scholars were occupied by the search for a single overarching belief system that could guide political opinions. The seminal work in this tradition was Philip Converse’s 1964 chapter, “The Nature of Belief Systems in Mass Publics.” Converse defined a belief system as “a configuration of ideas and attitudes in which the elements are bound together by some form of constraint or functional interdependence.”12 Political ideology, as commonly conceived, is a belief system with a broad range of political objects

6  Adam J. Berinsky

serving as referents for the ideas and attitudes in the system. Ideology therefore provides a relatively abstract and far-​reaching structure for a large variety of political attitudes and preferences. Converse argued that very few citizens thought of politics in an ideological manner. Analyzing survey data from the 1950s, Converse found that the vast majority of citizens did not use ideological terms, such as “liberal” or “conservative,” when talking about politics. Moreover, their opinions were largely unconnected across different issues and seemed to vary in random ways across time. Converse’s work is perhaps the most provocative piece of public opinion scholarship of the last 50 years. It is provocative not only because it focused researchers on questions central to the functioning of democracy, but also because it generated a tremendous amount of reaction—​most of it trying to rehabilitate the picture of the general public. For instance, in the 1970s, Norman Nie, Sidney Verba, and John Petrocik published The Changing American Voter. These authors found that “constraint”—​the correspondence between attitudes on related issues examined by Converse—​suddenly increased in 1964. Nie, Verba, and Petrocik attributed this change to a shift in the larger political context—​from the placid Eisenhower era of the 1950s to the ideologically charged events of the 1960s. These findings, however, turned out to be largely illusory. Both Converse and Nie, Verba, and Petrocik examined data collected by the American National Election Study (ANES). While the ANES is careful to preserve the continuity of its survey questions, at times the ANES does change its wordings. One important such change in question wording was … in 1964! It turns out that the ANES shifted from asking questions in a form where respondents were given a statement that asked if they agree or disagree with it, to a format where they were asked to choose between a pair of competing alternatives. In a series of clever experiments, John Sullivan, James Pierson, George Marcus, and Stanley Feldman demonstrated that most of the change found by Nie, Verba, and Petrocik could be attributed to these changes in question wording.13 By the end of the 1970s, the field was in many ways back where it started in the 1950s—​with the dismal findings of the early days of survey research at the forefront. Was there, then, nothing left to say about public opinion? Perhaps the answer was to ask different kinds of questions. In the early 1980s, Donald Kinder made a call to reframe the study of public opinion from the search for a single overarching principle to a finer grained investigation of what, exactly, does structure the political thought of ordinary Americans. This call to expand the study of public opinion beyond a debate over the power of ideology has been taken up to good effect. Indeed the chapters in this book on the role of race, religion, emotion, personality, and other topics provide testament to the fruitfulness of this research agenda. However, what may have gotten somewhat lost in the renaissance of the last 30 years is that ideology is still an important topic, as demonstrated by Christopher Federico’s contribution to this volume. In Chapter 3, Federico presents evidence that ideology has

Introduction and Overview  7

deep roots in citizens’ social circumstances, their psychological characteristics, and perhaps even in biology. Federico also notes that there are conditions under which ideology is a powerful determinant of citizens’ opinions on the matters of the day. Admittedly, not all citizens are able to reason in an ideological manner—​they must possess both political information and a strong desire to appraise things as “good” or “bad” in order to think ideologically and express ideologically consistent opinions. Among these citizens, though, ideology is a powerful force.

Moving Beyond Ideology That said, Kinder’s call to move beyond ideology has greatly expanded the scope of the field of public opinion. For those citizens who do not reason in an ideological manner, there are many other possible bases of political reasoning. Perhaps the most obvious determinant of public opinion is partisanship—​the degree to which individual citizens identify with one of the major political parties in America. Republicans and Democrats, after all, differ greatly on the major issues of the day. For instance, a poll taken right after the 2010 midterm election by CBS News demonstrated that Americans were evenly split on the question of whether Congress should try to repeal the health care bill passed in March of 2010. Forty-​five percent of respondents supported repeal, 45 percent opposed repeal, and another 10 percent were undecided. This seemingly balanced judgment, however, concealed a strong divide between the parties; 76 percent of Republicans favored the repeal, compared with 19 percent of Democrats. Such divisions continue through today. These stark partisan differences over policy extend to evaluations of major political figures. In that same poll, 78 percent of Democrats approved of the way Barack Obama was handling his job as president, but only 10  percent of Republicans expressed support.14 This partisan gap may have widened even further during the Trump administration. A  poll conducted by the Pew Research Center in June 2018 found that 84 percent of Republicans approved of the way Donald Trump was handling his job as president, compared to only 7 percent of Democrats.15 Given the magnitude of these differences it might surprise today’s readers to learn that for much of the history of the field, the power of partisanship was not a given. In fact, if this book were written 30 or 35 years ago, we would be bemoaning the death of parties. While the heyday of the literature on the decline of parties occurred in the 1970s and 1980s—​with works such as David Broder’s 1972 book, The Party’s Over and William Crotty’s 1984 book, American Parties in Decline—​Martin Wattenberg’s seminal work, The Decline of American Political Parties was updated several times into the late 1990s. Ironically, at the precise moment that political scientists were writing about the decline of parties and attachments to parties as a central determinant of public opinion, parties were beginning a steep ascent. Examining the votes that occur on the floor of the House and the Senate, Keith Poole and Howard Rosenthal

8  Adam J. Berinsky

have found that beginning in the mid-​1970s Democrats and Republicans began voting in more unified and distinct ways. Over time, the emergence of these distinct ideological positions among politicians began to trickle down to the level of ordinary citizens. In Chapter 4, Shanto Iyengar chronicles the change in mass conception of party identification over the last 30  years. Traditionally, political scientists have debated whether partisanship is better thought of as an overall evaluation of individuals’ ongoing interactions with politics, or as a psychological attachment to a political group. Some scholars argued that party identification is akin to a summary judgment about the current state of the political world. If times are good when a Democrat is in power, as they were in the early 1960s, individuals are more likely to think of themselves as Democrats. If times are bad—​think, for instance, 2010—​they are less likely to identify as Democrats. The more dominant view, however, is to think of party identification as an attachment to a political group. Over a lifetime, party identification is generally stable for most people. Identification with a particular party therefore provides even casually engaged citizens with a useful shortcut to arrive at an understanding of a complicated political world. What has changed recently, however, is the intensity of that attachment. At its heart, partisanship is about identifying with the “Democrat” group or the “Republican” group. Over time, the enmity between these groups has grown considerably, mirroring the widening gap between politicians observed by Poole and Rosenthal. Iyengar details these changes and explores the causes and consequences of this “affective polarization” in the U.S.

The Importance of Groups Attachments to political parties are not the only group connections that matter in American politics. As noted above, Converse’s landmark work on belief systems is primarily remembered for its conclusions regarding the limits of ideological thinking among members of the mass public. But Converse did not merely document the shortcomings of the citizenry; he also considered the ways that individuals could come to reasoned political decisions, even in the absence of an overarching guiding ideology. Chief among these was social groups. Converse claimed that visible groups in a society provide structure to individual political judgments, mentioning race, religion, and nationality as clear referents on the political scene in the 1950s. Converse placed a great deal of weight on the power of groups because they were relatively simple concepts, requiring a lower threshold of sophistication than needed to employ abstract concepts, such as ideology. As Converse argued, to make use of group-​based reasoning, citizens need only “be endowed with some cognitions of the group and with some interstitial ‘linking’ information indicating why a given party or policy is relevant to the group.”16 Converse concluded that reference group

Introduction and Overview  9

cues could serve as the foundation of “ideology by proxy,” creating meaningful patterns in the attitudes and behaviors of ordinary citizens.17 In America, for better or for worse, race has long served as such a guide. The political implications of the relationships between Blacks and Whites have a long history in this country, from pre-​Civil War debates about the role of slavery, through today. In Chapter 5, Michael Tesler explores this landscape. He begins by describing the ways in which social scientists have measured racial attitudes over the last 75 years. The open endorsement of white supremacy is clearly a thing of the past. But the demise of old fashion racism, such as the belief in the biological inferiority of Blacks and a desire for social distances between Blacks and Whites does not mean that racial animus has faded completely from the political scene. “Racial resentment”—​a moral feeling that Blacks violate such traditional American values as individualism and self-​reliance, work ethic, obedience, and discipline—​characterizes this new form of symbolic racism. Tesler then goes on to show how powerful a force racial resentment has become in recent years. The reason is simple. The election of Barack Obama in 2008 and 2012, and the subsequent election of Donald Trump in 2016, heightened the salience of peoples’ understandings of which groups are thought to be helped or hurt by the president’s policies. In the wake of Obama’s election as the first African American president in 2008, commentators from across the political spectrum speculated that the U.S. would enter an era in which race would diminish as a force in American politics. But Obama’s rise to the executive office did not create a “post-​racial” era. Instead, it expanded the reach of race, ushering in a new “most racial” political era. Such dynamics were clear from the start of Obama’s presidency. His first term of office induced the “spillover of racialization” from evaluations about Obama to the policies and people that he has become associated with. That is, racial attitudes became stronger determinants of public opinion when Barack Obama took visible positions on issues. Tesler demonstrates that attitudes toward health care reform, votes in congressional elections, and evaluations of major political figures were shaped by attitudes about race. If anything, the ascendency of Donald Trump has heightened the racialization of politics. As Tesler notes, Trump repeatedly pushed the boundaries of political rhetoric, making explicit appeals to racial resentment, anti-​ immigrant sentiments, Islamophobia, and white grievances. Where Obama polarized public opinion by racial attitudes primarily because of who he was, Trump polarized opinion by what he said. While Black/​White relations have always been important in American politics, racial politics is not simply about the relationship between Blacks and Whites. As America becomes a more diverse society, other racial groups have risen in prominence. For instance, over the last 30 years, the percentage of the population that is Hispanic has increased greatly. Moreover, the “multi-​racial” population—​individuals who identify with more than one racial group—​is among the fastest-​growing group in America. Thus, as Erica Czaja and Vladimir

10  Adam J. Berinsky

Medenica remind us, racial politics in the United States is in flux. In Chapter 6, they argue that the contours of the relationships between racial group identity, racial group consciousness, and public opinion—​especially for Latinos and Asian Americans—​are particularly challenging for scholars. In their chapter, they try to make sense of these diverse measurements by focusing on individual-​ level measurements of psychological attachment to groups, namely group identity and consciousness, which are critical intervening variables between racial group classification and the formation of political preferences. Czaja and Medenica are primarily concerned with the consequences of attachment to particular groups for public opinion. This in-​ group identification is indeed an important component of political attitudes. But also important is an individual’s relationships with other groups in society. In Chapter 7, Donald Kinder, Molly Reynolds, and Nancy Burns explicitly take up both these concerns in the context of gender. Specifically, they consider how the groupings of “man” and “woman” matter for politics. As a mental category, gender provides an important guide to how we think about ourselves and others. Like race, gender is a source of persistent and serious inequality in America. However, in practice, race and gender generally have different effects on opinion. Examining a wide range of policy attitudes, Burns, Kinder, and Reynolds find that the differences between men and women in the domain of gender policy are present, but often quite small. That said, there are times when gender differences can help shape political views. The 2016 presidential election was one of those times. Through these examples, Burns, Kinder, and Reynolds remind us that the power of group identity is in many ways contextually dependent. Different aspects of group identity may come into play in different ways depending on the particular social and political circumstances that create, shape, and activate that identity.

The New Psychological Foundations of Opinion In addition to political and group attachments, there are a host of other factors that help to shape public opinion. An important new direction in recent years is the incorporation of insights from psychology and biology about the roots of public opinion. Marc Hetherington considers a wholly new kind of divide in political thinking: one shaped by the fundamental worldview that Americans possess. In describing this worldview, Hetherington reaches beyond the kinds of belief systems that the scholars of the 1960s and 1970s went in search of. As he writes, this worldview is “someone’s deeply ingrained beliefs about the nature of the world and the priorities a good society should have.”18 And at the core of such a worldview is a notion how dangerous individuals perceive the world to be. People who see the world as a scary and frightening place see potential threats everywhere and think that public policies must be designed to protect individuals. Those who do not adopt such a view recognize value in allowing

Introduction and Overview  11

individuals to freely explore the world. The former—​individuals who hold a “fixed” worldview—​see danger around every corner and are most interested in survival. The latter—​those with a “fluid” worldview—​see a world of safety and worry about missed opportunities for connection and free expression. The key insight from Hetherington’s worldview theory is that different individuals may react in very different ways to the exact same situation, depending on whether they see the world as a dangerous place or not. This reaction can profoundly shape not only their lifestyle choices, but their political views as well. What makes this process especially powerful is that, as Hetherington shows in his chapter, the correspondence between people’s worldview and their partisan attachment has converged greatly over the last three decades. In the early 1990s, there was no association between whether people held a fixed or fluid view and whether they were Democrats or Republicans. But by 2016, among those with a fixed worldview, Republicans outnumber Democrats by a ratio of almost three to one. Among those with a fluid worldview, Democrats outnumbered Republicans by an even greater proportion. Clearly, individual psychology plays a huge role in determining where people stand in the modern political landscape. Hetherington details the origins and implications of such a divide in Chapter 8. Another line of research concerns the role of emotion in shaping political preferences. At the forefront of this fast-​growing subfield is Ted Brader, who provides an overview of the emotional foundations of democratic citizenship in Chapter 9. While emotions are a complex set of reactions to external circumstances, individuals experience emotions via their gut reactions to particular stimuli. Thus, as Brader notes, emotions are best understood as the processes that generate feelings, and serve as a motivation for action. For instance, a feeling of fear causes individuals to become more alert and focused on external threats. People therefore shift from the status quo to actively reconsider their options. While emotions are not explicitly political, such feelings often have important political implications. Of the plethora of emotional reactions, scholars have found that three stand out: fear, anger, and enthusiasm. These emotions affect public opinion by altering whether and how citizens pay attention to political events, learn about political developments, think through their decisions, and act on their opinions. For example, in his work on campaign advertising reviewed in Chapter 9, Brader found that campaign ads that elicit fear—​through the use of ominous music and unsettling images—​cause voters to reconsider their standing decisions and make decisions based on their assessment of a candidate’s issue positions and leadership qualities. In more recent elections, Brader and Cikanek show that enthusiasm, anger, and fear continue to shape how ordinary citizens make the decision to participate in the political sphere. In Chapter 10, Frank Gonzalez, Kevin Smith, and John Hibbing turn to the related question of the role of biology in politics. The authors, who are some of the leading figures in the new field of biopolitics, begin by exploring the

12  Adam J. Berinsky

importance of non-​conscious information processing and then describe the important work that they and others are doing at the intersections of biology and politics. Some of this research studies how genetics can shape (though not determine) political attitudes through the heritability of personality traits and cognitive styles. Other research explores brain activity during the process of political reasoning through functional Magnetic Resonance Imaging (fMRI). This work is still in its early stages, and studies of biology and politics have generated a great deal of controversy—​as the authors note in their chapter. However, this is certainly a promising and interesting area that will remain on the forefront of the field of public opinion in the years to come.

The Public and Society Understanding the structure of public opinion is an important step in assessing its political significance. But it is also important to actually detail the role it plays in the political world. The last section of this book explicitly examines how public opinion is shaped by political events and the rhetoric of politicians and how, in turn, public opinion shapes and guides the conduct of politics. In Chapter  11, Matthew Baum and Danna Young explore how changes in the media environment have revolutionized the ways in which politicians communicate with the mass public when building support for their policies and programs. With the rise of the Internet, social media, and the fragmentation of the television audience among a growing number of cable channels, the media landscape has transformed; the line between news and entertainment has blurred and is increasingly permeable. Using the rise and subsequent fall of the Roseanne sitcom reboot over a few months in early 2018 as an example, Baum and Young demonstrate how partisan politics both informed and shaped peoples’ reactions to the show. In essence, political news has spilled over to multiple channels in the modern media environment. In addition to the legacy news media, which was the focal point for political news through the 1980s, Americans can learn about politics from partisan media (such as Fox News and talk radio shows), online digital media (including both social media and online outlets from legacy media) and a wide variety of hybrid political entertainment media (including entertainment programs like Roseanne and satirical comedy shows on the left). As the media become more specialized—​and as voters follow their interests and flock to different sources that share their pre-​existing political views—​it becomes harder to find any common political ground. While a common civic space for public affairs has not entirely disappeared, the potential for reaching across the partisan aisle through the media has certainly been shrinking in recent years. In Chapter 12, Jennifer Jerit examines the consumer side of the media/​mass public relationship, returning to some of the same themes taken up by Gilens in Chapter 2. How, she asks, do people navigate the modern information landscape to learn (or not) about the political world? She investigates the factors that

Introduction and Overview  13

shape whether or not individuals are able to become informed about politics, considering three pathways to knowledge. First, people differ in their ability to learn about politics. Second, the political context affords different people different opportunities to engage with the political world. Finally, there is wide variation in citizens’ motivation to learn about politics. This three-​path framework is especially important in the modern age because, increasingly, pundits and scholars have become concerned not only about the lack of political information, but the prevalence of political misinformation. Another important topic of research is the dynamics of democratic elections. Though the study of campaigns and elections is itself a thriving subfield in political science, it has been deeply informed by insights from the field of public opinion. In fact, many of the fundamental insights about the nature of public opinion discussed in this book came out of data collected on voter reactions to political campaigns in the 1940s and 1950s. Consider the Columbia studies led by Berelson, Lazarsfeld, and McPhee, discussed above. These scholars went into their study expecting that voters during a campaign would treat candidates as something akin to products in a supermarket—​citizens would shop around, consider (and reconsider) their options, and finally arrive at a choice on Election Day. In fact, the Columbia researchers found something very different. Most people did not change their preferences. The campaign, it seemed, had little effect on voters’ decisions. Until the 2016 election, most political scientists would have argued that what was true in 1948 is true today. Beginning in the 1980s, a cottage industry developed in the prediction of presidential contests. Months before an election, political scientists could predict election results. For instance, at a meeting of political scientists held over Labor Day weekend in 2012—​even before the start of the Democratic National convention—​a panel of political scientists each presented a forecast of the final vote tally. Almost all of them predicted the outcome within a few percentage points. The average of their predictions was a narrow Obama win of 50.2 percent two-​party vote share for the Democrats, just 1.5 percent below the national percentage of the vote Barack Obama received on Election Day. But why is this? There are a number of reasons why reasonable people could think campaigns should matter. Journalists focus incessantly on the ups and downs of campaigns and analyze each day’s events in the search for critical turning points. Furthermore, polls are variable, changing from day to day and week to week. Taken together, the political science models of stability, and the journalistic accounts of wildly shifting indicators led to an important question: How is it possible that elections are so predictable even amidst the apparent volatility of the campaign? At least, that is, until the election of Donald Trump over Hillary Clinton in 2016. Trump’s election surprised not just many political scientists, but even his own campaign team. The question on most people’s minds in the wake of that election is “do we know anything about elections?”

14  Adam J. Berinsky

In Chapter  13, John Sides and Jake Haselswerdt argue that we do. While Trump’s election defied most predictions, that does not mean that elections are inherently unpredictable. Voters, Sides remind us, almost never arrive at a campaign as a blank slate, devoid of ideas about politics or the candidates. In most elections, voters can draw on their long-​standing political identities to guide their choices, even without any detailed information about the candidates. These identities include race, ethnicity, socioeconomic status, religion, and—​ above all—​partisanship. The predictability of elections stems from how these fundamental factors affect how citizens vote and by association, who wins the election. Put another way, most campaign outcomes are predictable because preexisting ties to particular choices are so strong. Even in 2016, this was true. Remember, though Clinton lost the Electoral College vote by a margin of 304 to 227, she won the popular vote by two percentage points—​a narrow victory that was very much in line with the political science models based on the economic and political conditions of the country which predicted a close election. Still, as the 2016 contest reminds us, this is not to say that campaigns serve no function in a democratic society. True, most voters can draw on social and partisan identities to make decisions about candidates, but there will still be some citizens who are uncertain about, or unfamiliar with, the candidates. For these citizens, campaigns can serve a critical function. Moreover, in a primary or nonpartisan election, voters cannot draw on familiar heuristics and decision rules, such as straight-​party voting. Here too, campaigns can matter greatly. In the second half of their chapter, Sides and Haselswerdt therefore detail how campaigns can reinforce some existing decisions and change others’ minds. In Chapter  14, Deborah Schildkraut explores one of the most salient political controversies in recent years—​public opinion concerning immigration policy. Taking a broad historical perspective, Schildkraut makes the case that American public opinion toward immigration is characterized by great ambivalence. On the one hand, while many citizens support creating a path to citizenship for undocumented immigrants, they also favor allowing law enforcement officers check on the immigration status of aliens. In the aggregate, this has led to a situation in which there is no clear view of immigrants—​half of Americans say that immigrants strengthen the country because of their hard work and talents, while 4 out of 10 Americans say that immigrants are a burden “because they take our jobs, housing, and healthcare.” In this mix of competing views, many of the themes addressed elsewhere in this volume find expression, most notably the power of elite rhetoric to guide the public and the role of groups in shaping mass opinion. Finally, in Chapter 15, Andrea Campbell and Elizabeth Rigby examine the relationship between public opinion and public policy, including processes of policy feedback. Political scientists have spent decades examining whether public opinion shapes the policies of government. But, as Campbell and Rigby note, an exclusive focus on the public opinion side of the equation paints an incomplete picture. On the one hand, the public’s preferences influence

Introduction and Overview  15

policy outcomes, although this varies across time and across issues. Moreover, privileged and politically active groups are more likely to see their preferences fulfilled than others. But once policies are created, they acquire a life of their own. Thus, public policies themselves influence public opinion, so that existing policy shapes the political landscape and the possibilities for future policy. Opinion and policy move together, each influencing the other. Campbell and Rigby use a case study of the Affordable Care Act (ACA)—​ the 2010 federal health reform bill nicknamed “ObamaCare”—​to illustrate these effects. The ACA was one of the largest social policy reforms in generations. The political debate surrounding the passage of the bill illustrates many of the themes explored by Campbell and Rigby. On the one hand, public opinion can shape policy; differences in public support across the states influenced the pace of the implementation of the ACA. States where support was high, such as New York, were more likely to move forward to implement state-​based health insurance exchanges, as specified in the federal legislation. States where support was relatively low, like Utah, were less likely to move forward. On the other hand, the case of the ACA also demonstrates that Americans’ policy positions are shaped by the political discourse they encounter. For instance, through the early years of the ACA, survey questions that referred to the ACA as “Obamacare” showed significant less support for the reforms enacted by the ACA than did questions that did not. The future of the ACA—​and popular reaction to healthcare—​will continue to be shaped by these forces.

Final Thoughts Though I have tried to provide a comprehensive overview of the field of public opinion, this volume has only scratched the surface of the excellent research that is out there. Of necessity, there is a great deal of material that I could not cover. For instance, besides political and group attachments and personality and emotion, there are a host of other factors that help to shape public opinion. There is a large and rich literature on the power of core political values, such as subscription to principles of political equality or individualism.19 Other scholars have explored the relationship between self-​interest and opinion.20 For instance, some political scientists have examined whether those individuals with the greatest stake in a policy—​parents who might be directly affected by policies designed to ensure school integration, for instance—​are more likely to take a clear stance on questions relating to that policy. Here, the evidence is very mixed; somewhat surprisingly, the relationship between self-​interest and opinion is often quite weak. Recently, however, scholars have begun exploring other ways in which direct economic interest could matter for political choice.21 Beyond research on the structure of opinion, there is also a great deal of interesting work being done on the role of public opinion in society. For example, a number of researchers have explored the nature of social influence

16  Adam J. Berinsky

in opinion formation and dissemination, some of which harkens back to Blumer’s work in the 1940s.22 And even this work merely scratches the surface of the field. But this book is intended to be just the starting point for most students. Hopefully, the chapters will pique your interest and give you the tools and motivation you need to go off and explore the field on your own.

Notes 1 Key, V. O. (1963) Public Opinion and American Democracy. New  York:  Alfred Knopf, p. 8. 2 Herbst, Susan (1993) Numbered Voices: How Opinion Polling has Shaped American Politics. Chicago, IL: University of Chicago Press. 3 Blumer, Herbert (1948) “Public Opinion and Public Opinion Polling.” American Sociological Review 13: 542–​9. 4 Childs, H. (1964) Public Opinion. Princeton, NJ: D. Van Nostrand. 5 Key (1963), p. 14. 6 National Health Interview Study. “Wireless Substitution: Early Release of Estimates From the National Health Interview Survey.” June 2018, www.cdc.gov/​nchs/​data/​ nhis/​earlyrelease/​wireless201806.pdf. 7 Lippmann, Walter. (1925) The Phantom Public. New York: Macmillan & Co., p. 24. 8 Lowell, Abbot Lawrence (1922) Public Opinion in War and Peace. Cambridge, MA: Harvard University Press. 9 Berelson, Bernard R., Paul F.  Lazarsfeld, and William N.  McPhee (1954) Voting. Chicago, IL: University of Chicago Press, p. 207. 10 Stimson, James A. (2004) Tides of Consent: How Public Opinion Shapes American Politics. Cambridge: Cambridge University Press. 11 Achen, Christopher H. (1975) “Mass Political Attitudes and the Survey Response.” The American Political Science Review 69(4): 1227. 12 Converse, Philip (1964) “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, ed. D. Apter (New York: Free Press), p. 207. 13 Sullivan, John L., James Pierson, and George E.  Marcus (1979) “An Alternative Conceptualization of Political Tolerance: Illusory Increases, 1950s–​1970s.” American Political Science Review 73: 233–​49. 14 CBS News. (2011). “Looking Ahead to the 112th Congress.” Retrieved from www. cbsnews.com/​stories/​2010/​11/​11/​politics/​main7045964.shtml. 15 www.pewresearch.org/​fact-​tank/​2018/​08/​01/​trumps-​approval-​ratings-​so-​far-​are-​ unusually-​stable-​and-​deeply-​partisan/​. 16 Converse (1964), pp. 236–​7. 17 Converse, Phillip (1975) Public Opinion and Voting Behavior. In Handbook of Political Science, ed. F.  W. Greenstein and N.  W. Polsby, Vol. 4 (Reading, MA: Addison-​Wesley), pp. 75–​169. 18 Hetherington, Marc, and Jonathan Weiler (2018) Prius Or Pickup?: How the Answers to Four Simple Questions Explain America’s Great Divide (Boston, MA: Houghton Mifflin), pp. x–​xii. 19 For an excellent review, see Feldman, Stanley (2003) “Values, Ideology, and the Structure of Political Attitudes.” In Oxford Handbook of Political Psychology, ed.

Introduction and Overview  17 David O.  Sears, Leonie Huddy, and Robert Jervis (New  York:  Oxford University Press), pp. 477–​508. 20 See, for example, Sears, David O., Richard Lau, Tom Tyler, and Harris Allen, Jr. (1980) “Self-Interest vs. Symbolic Politics in Policy Attitudes and Presidential Voting.” The American Political Science Review 74(3):  670–​84. Also, Citrin, Jack, Beth Reingold, and Donald Green (1990) “American Identity and the Politics of Ethnic Change.” The Journal of Politics 52: 1124–​54. 21 Gerber, Alan S. and Gregory A. Huber (2010) “Partisanship, Political Control and Economic Assessments.” American Journal of Political Science 54: 153–​73. 22 See, for example, R.  Huckfeldt and J.  Sprague (1995) Citizens, Politics, and Social Communication:  Information and Influence in an Election Campaign (New York: Cambridge University Press); Baybeck, Brady and Scott McClurg (2005) “What do they know and how do they know it? An Examination of Citizen Awareness of Context.” American Politics Research 33(4): 492–​250; Mutz, Diana (2006) Hearing the Other Side: Deliberative versus Participatory Democracy (New York: Cambridge University Press); Nickerson, David (2008) “Is Voting Contagious? Evidence from Two Field Experiments.” American Political Science Review 102(01): 49–​57.

Part I

The Meaning and Measurement of Public Opinion

Chapter 1

The Practice of Survey Research Changes and Challenges D. Sunshine Hillygus

The 2016 U.S. presidential election understandably rattled public confidence in the polling industry. Pre-​election polls and forecasts consistently predicted that Democrat Hillary Clinton would win a resounding victory over Republican Donald Trump, but he surprised the world by winning 56.8 percent of the electoral college vote to capture the presidency. Media headlines asked “Can We Ever Trust the Polls Again?” and pundits suggested that the polling industry was “teetering on the edge of disaster.”1 There is little doubt that polling or survey research has faced increasing methodological challenges in recent years. Surveys rely on the cooperation of people to check boxes and answer questions, yet people today are harder to reach, and when contacted they are less likely to answer questions. At the same time, there has been a proliferation in the amount of polling—​from horserace numbers in the newspaper headlines to opt-​in “polls” predicting sports outcomes on ESPN.com or judging celebrity outfits in Instyle magazine. With so many polls, it is no wonder that it can be difficult to figure out if polls are trustworthy. In this chapter, I will outline some of the key methodological challenges in conducting, using, and evaluating surveys as a measure of public opinion. There are three “take-​home” messages:  First, I  will explain why all surveys are not created equal. Some surveys should be trusted more than others, and, unfortunately, it is not sufficient to make assumptions about survey quality based on polling topic (say, politics rather than entertainment), sample size, or sponsorship. The total survey error perspective provides a framework for evaluating how various aspects of the survey method can influence the validity and reliability of the resulting survey statistics. Second, I hope this chapter makes clear that no survey is perfect. While there is significant variation in survey quality, not even our “gold standard” surveys like the American National Election Study should be immune from scrutiny. Finally, I will appeal for journalists and scholars at all levels to provide enough information about their survey methods for readers to assess the knowledge claims being made.

22  D. Sunshine Hillygus

The Data Stork Myth Despite increasing concerns about survey quality, surveys remain the cornerstone of research on economic, political, and social phenomena across academic, commercial, nonprofit, and government sectors. When properly designed, surveys are a powerful tool for collecting information about the attitudes, characteristics, and behaviors of individuals, households, and organizations. Too often, however, scholars and journalists tend to treat survey data as if they have simply been delivered by a data stork, failing to question where they came from, how they were produced, and by what methodology. Yet a survey involves numerous steps and decisions, and with each one, error can be introduced into the resulting survey statistics. A significant part of the difficulty in establishing survey quality standards is not that our scientific understanding of survey methodology is flawed or inadequate, but rather that scientific research in survey methodology has not permeated the broader community of survey consumers. In the survey methodology literature, scholars have adopted a total survey error perspective that recognizes the need to consider a variety of different types of error in evaluating survey quality.2 Figure  1.1, reproduced from Herb Weisberg’s textbook The Total Survey Error Approach, summarizes these

Sampling error

Measurement error

Nonresponse error

Due to interviewer Due to respondent

At unit level At item level

Coverage error

Post survey error Types of survey error

Figure 1.1 The Tital Survey Error Perspective

The Practice of Survey Research  23

various sources of error in the survey process.3 In this chapter, I  discuss some of these—​sampling error, coverage error, nonresponse error, and measurement error—​highlighting specific challenges and controversies. I first provide an overview of the survey process and introduce some key terminology. A high-​quality survey is one that tries to minimize all sources of error within the inevitable time and budgetary constraints of the project. The goal is to produce survey results that are valid and reliable, terms that have specific meaning and usage in scientific research. Validity refers to the accuracy of the results, while reliability refers to consistency or stability in the results if it were to be repeated in identical conditions. Archery offers a common analogy to clarify the difference between the concepts of validity and reliability. Research that is reliable but not valid is like an archer who always hits about the same place but not near the bullseye. Research that is valid but not reliable is like the archer who hits various places centered around the bullseye, but not very accurately. Again, the goal is to be both reliable and valid, like an archer who hits consistently close to the bullseye.

Overview of the Survey Process When we think of surveys, we often have in mind the resulting survey statistics. For example, a recent news story reported that 61 percent of Americans support legalizing marijuana. This survey statistic about public opinion is the product of a very specific survey process that involves a series of consequential methodological decisions and assumptions. In small print at the end of the article, we find some of that methodological information: “Conducted April 11–​April 15, 2017 and based on 1,011 telephone interviews. Sample: National adult. The margin of sampling error is plus or minus 3 percentage points.” In this section, I will outline the basic process involved in reaching conclusions about public opinion on the basis of a smaller sample of respondents. The first step in the survey process is deciding on the target population; that is, the group to whom the survey is intended to generalize. This survey obviously did not ask all Americans their opinion on this issue; rather, they surveyed 1,011 individuals they believed were representative of the broader American public. Their target population was the entire adult US population.4 Many polls, especially pre-​election polls, use instead a target population of only adults who are registered to vote. Other surveys are interested in even more specialized populations; for example, a recent survey on alcohol and drug use at Duke University was meant to represent only those undergraduates currently enrolled at the university. After determining the target population, the next step in the survey process is specifying a sample frame—​lists or procedures that identify all elements of the target population. The sample frame may be a list of telephone numbers, maps of areas in which households can be found, or a procedure (like random

24  D. Sunshine Hillygus

digit dialing) that could identify the target population. At their simplest, sampling frames just list the phone numbers, addresses, or emails of individuals in the target population, such as the list of student email addresses for the Duke University student survey. In the case of this poll about marijuana attitudes, random digit dialing (RDD) was used. In random digit dialing, a computer generates a random set of seven-​digit numbers. Compared to using a telephone book or other list of telephone numbers, an RDD sample frame has the advantage of including unlisted numbers. Often, the list will not perfectly capture the entire target population. For example, telephone surveys that only call landline telephone numbers will miss individuals who only have a cell phone. This creates coverage error—​the error that arises when the sampling approach does not include all of the target population. That is, when there is a failure to give some persons in the target population a chance of selection into the sample. If those included in the sample frame differ from those who are not, that coverage error can create coverage bias, affecting the accuracy of the resulting survey statistics. Once a sample frame has been identified, individual cases are randomly selected to be in the survey. Because the survey is administered to a sample, rather than all, of the target population, it is subject to random sampling error. This is the “margin of error” mentioned in the methodological disclosure of the poll. Of course, these selected cases are just the people asked to be in the survey—​many of them will be difficult to reach, will refuse to participate, or will drop out during the survey. Nonresponse error occurs when the individuals invited to take the survey are not interviewed. And the respondents are the subsample of the selected cases who actually complete the survey and on which the analysis is conducted.5 Critically, if those who respond are different from those who do not (either because of coverage error or nonresponse error) the resulting survey statistics can be biased.6 Figure 1.2 illustrates the key steps in the survey sampling process using the poll about marijuana attitudes as an example. As shown in the figure, each step in the survey sampling process can introduce uncertainty and bias in the resulting survey statistics. These errors can threaten the ability to generalize from the sample to the target population. Traditionally, survey users have focused on sampling error as the metric for evaluating survey quality. As mentioned, sampling error represents the uncertainty or imprecision in estimates based on random chance that occurs simply because we observe data from a sample of individuals in the population rather than every individual in the population. Sampling error is often reported as margin of error. In the case of this poll, we should interpret the results as showing that public support for marijuana legalization is 61 percent +/​–​3 percentage points. This tells us how precise we are in our estimate of public opinion on this issue—​the larger the margin of error, the less confidence we have in our estimate. The literal interpretation of the margin of error is somewhat long-​winded and awkward: if the survey were repeated many times, 95 percent

The Practice of Survey Research  25

Target population: Americans Coverage error Sample frame: U.S. Adults with telephone numbers Random sampling error Sampled cases: 10,000 telephone numbers (estimated) Nonresponse error Respondents: 1,010 individuals who completed survey

Figure 1.2 Steps in Survey Process

of the samples of this size would be expected to produce a margin of error that captures the true percentage of Americans supporting the legalization of marijuana.7 Critically, the size of sampling error depends only on the size of the sample collected—​the larger the sample, the less uncertainty in the estimate. Sampling error does not tell us about whether our estimates are biased or inaccurate. Instead, it is only a measure of reliability. Thus, despite the traditional focus on sampling error in the reporting of survey statistics, it is actually the least important aspect of survey error; for a survey of a given size, sampling error simply “is what it is,” whereas other sources of error—​coverage error, nonresponse error, measurement error—​can be minimized through various design decisions.8 The total survey error perspective highlights the need to take into account not only sampling error but also the nonsampling errors, like coverage error and nonresponse error. The total survey error perspective recognizes that the substantive conclusions drawn from surveys also depend on the measurement process, in which scholars have to make decisions about how to operationalize and measure their theoretical constructs and then have to make decisions about how to code and adjust the resulting data. Nonsampling errors are different from sampling error in that it can affect not only the reliability of the results, but also the accuracy. These errors can introduce bias in the survey results if they are systematic rather than random. In the remainder of this chapter, I will use the total survey perspective to outline some of the key contemporary threats to survey quality.

26  D. Sunshine Hillygus

Probability vs. Nonprobability Sampling Surveys are typically conducted in order to make generalizations about a target population from data collected from a smaller subset—​the sample. The ability to generalize from the sample to the population rests on the use of probability sampling. Probability samples are ones that use random selection. As pollsters like to joke, “If you don’t believe in random sampling, the next time you have a blood test tell the doctor to take it all.” Random selection of respondents means that errors—​both those observed and unobserved—​cancel out over the long run. In order to have a random selection method, it’s necessary for each member of the target population to have a chance of being selected into the sample. With a random probability sample, the results will be close (within the “margin of error”) to what we would have found had we interviewed the entire population. In contrast, nonprobability samples select respondents from the target population in some nonrandom manner, so that some members of the population have no chance of selection. Nonprobability samples instead recruit respondents through advertisements, pop-​ up solicitations, and other approaches. For example, many media organizations invite visitors to their websites to answer “straw polls.” Sometimes nonprobability samples rely on quota sampling, which identifies a set of groups (e.g., men, women, 18–​25-​year-​olds, 26–​40-​year-​olds, etc.) and specifies a fixed number of people to be recruited for each group. Interviewing then proceeds until the quota is reached for each group. With quota sampling, nonprobability samples can be designed so that they match the population proportions on age, gender, and socioeconomic status or any other known population characteristics. The Literary Digest polling fiasco of 1936 is the classic example of how nonprobability samples can lead to biased conclusions. The popular magazine had correctly predicted the winner in the previous five presidential elections, but in 1936 incorrectly predicted that Alf Landon would beat FDR in that year’s election by 57 to 43 percent (FDR won with 60.8 percent of the popular vote). The Digest had mailed over 10 million survey questionnaires to their subscribers and to names drawn from lists of automobile and telephone owners. More than 2.3 million people responded, but it turns out that, in 1936, those who owned automobiles, telephones, or had the disposable income to subscribe to a magazine were not a random cross-​section of the voting public. More recently, a Scientific American online poll illustrated the perils of straw polls. The popular science magazine’s website invited their readers to answer questions about climate change. The poll attracted the attention of climate skeptic bloggers who directed their own readers to participate in the poll, so that the poll results showed that 80  percent of respondents denied climate change and 84 percent answered that “The Intergovernmental Panel on Climate Change is … A  corrupt organization, prone to groupthink, with a political agenda.” Although it’s not unusual for online polls to be hijacked by activists,

The Practice of Survey Research  27

these skewed polling results have since been reported in an editorial in The Wall Street Journal and included in Congressional testimony with no mention of the unscientific methodology.9 Probability sampling allows us to calculate sampling error so we can estimate how much our sample might differ from the target population (the margin of error). In nonprobability sampling, in contrast, the degree to which the sample differs from the population remains unknown and unknowable. Even if the sample looks demographically similar to the target population (as with quota sampling), we have no way to evaluate if the sample is representative of unobserved characteristics. One of the key contemporary debates in public opinion research regards the quality of nonprobability-​based online panel surveys. New technologies have both made probability sampling more difficult and made nonprobability sampling—​especially with online panels—​fast, easy, and inexpensive compared to traditional probability samples. Online nonprobability surveys now make up a significant portion of academic, business, and government survey work.10 The main concern with Internet-​based surveys is not just that they will miss those without Internet access—​Internet use in the United States population has risen steadily since the turn of the century, increasing from 52  percent in 2000 to 88 percent in 2016 in the U.S. The key hurdle is that, in most cases, it is difficult to define an appropriate sample frame from which to draw a random sample that is a reasonable approximation of the target population.11 In other words, there is typically no list of email addresses from which a random sample can be drawn. While not a problem in cases where a population list exists and is reachable online (e.g., email addresses of students at a university), for general population surveys, the nature of the Internet means that “frames of internet users in a form suitable for sampling do not—​and likely will not—​exist.”12 Despite their pervasiveness, there remains considerable debate and confusion about online survey methods, fueled by the wide variation in sampling design and implementation approaches. Online surveys use a variety of different sampling designs, recruitment strategies, and implementation procedures, with implications for data quality. In cases where the population list is known and reachable online (e.g., email addresses of students at a university or business CEOs), web surveys are appropriate—​even preferable.13 It is also possible to draw a probability-​based sample using a traditional technique (such as RDD or address-​based sampling), and then provide Internet access to those without it. For example, the 2016 American National Election Study mailed a random sample of addresses a letter (and a large financial incentive) inviting them to answer the questionnaire online.14 But the majority of web-​based surveys rely on nonprobability online panels. In such cases the respondents are (nonrandomly) recruited through a variety of techniques: website advertisements, targeted emails, and the like.15 Individuals are then signed up in an online panel in which they are regularly invited to answer surveys in exchange for financial incentives or other awards. Even if a

28  D. Sunshine Hillygus

given sample is randomly selected from this online panel, the pool of potential respondents are all people who initially “opted in” to the respondent pool so it is not representative of the general population.16 Surveys from such panels often rely on quota sampling from the panel. A second source of confusion is that nonprobability samples are often claimed to be “representative” because the sample looks like the target population on a set of observed characteristics, often through adjustments (e.g., weighting and/​or matching) of the opt-​in sample to census benchmarks.17 There are, however, only a limited number of benchmarks on which the sample can be compared, so these samples still require the untestable assumption that unmatched characteristics are ignorable.18 And research has shown, for instance, that those who volunteer to participate in surveys are often more informed, knowledgeable, and opinionated about the survey topic even if they look demographically similar to the general population.19 A taskforce of the American Association for Public Opinion Research (AAPOR), the leading professional organization of public opinion and survey research professionals in the U.S., tackled the issue of online panels and forcefully concludes that “There currently is no generally accepted theoretical basis from which to claim that survey results using samples from nonprobability online panels are projectable to the general population … Claims of ‘representativeness’ should be avoided.” Pollsters Gary Langer and Jon Cohen offer a similar, if more colorful, conclusion: anyone following the polls is probably finding it increasingly difficult to separate20 signal from noise … In reality, there are good polls and bad, reliable methods and unreliable ones. To meet reasonable news standards, a poll should be based on a representative, random sample of respondents; “probability sampling” is a fundamental requirement of inferential statistics, the foundation on which survey research is built. Surrender to “convenience” or self-​selected samples of the sort that so many people click on the internet, and you’re quickly afloat in a sea of voodoo data … Probability sampling has its own challenges, of course. Many telephone surveys are conducted using techniques that range from the minimally acceptable to the dreadful. When it’s all just numbers, these, too, get tossed into the mix, like turpentine in the salad dressing.21 There is considerable variation in the quality of nonprobability samples, with some methods being better or worse depending on the particular outcome of interest. A  recent report by Pew Research Center examined the variation in the data quality and demographic composition of online panel surveys.22 The authors fielded the same survey on several different online survey vendors, comparing survey estimates across a number of demographic and political benchmarks, and they find wide variability in the accuracy of estimates across the different samples.

The Practice of Survey Research  29

In sum, nonprobability samples—​ no matter their size—​ require often unrealistic, untestable, and unstated assumptions to make inferential claims. There is often no way to know how respondents and nonrespondents might differ across an infinite number of characteristics related to the outcome of interest. Procedures such as quota sampling, matching, or weighting ensure that a convenience sample looks like the target population on a set of observed characteristics but inherently assumes that unobserved characteristics do not influence the phenomenon being studied. A  Pew Research Center study compared different approaches to adjusting nonprobability samples and finds that they often are ineffective at correcting the bias in the samples.23 To be sure, this does not mean that nonprobability samples should never be conducted. There are also many research questions for which a probability sample will not be a priority. For example, scholars conducting survey experiments are often more concerned with internal validity (a clear causal effect) than external validity (generalizability). Likewise, focused exploratory research might use a nonprobability sample to generate hypotheses or pilot various measurements. There may also be times when the researcher simply wants to demonstrate that a particular trait occurs in a population. These are all cases in which the researcher does not intend to draw inferences to the broader population, so a nonprobability sample can be a cost-​effective method for the research goals. The key point is that analysts are transparent about the methodology being used. AAPOR, for example, recommends the following wording when documenting surveys with non-​probability samples: Respondents for this survey were selected from among those who have [volunteered to participate/​registered to participate in (company name) online surveys and polls]. The data (have been/​have not been) weighted to reflect the demographic composition of (target population). Because the sample is based on those who initially self-​selected for participation [in the panel] rather than a probability sample, no estimates of sampling error can be calculated. All sample surveys and polls may be subject to multiple sources of error, including, but not limited to sampling error, coverage error, and measurement error. Given the challenges and costs of traditional probability surveys, nonprobability online surveys will continue to play an important role in measuring public opinion. At the same time, it makes it more important to understand variation in data quality. I next consider specific sources of nonsampling error that can affect the quality of survey estimates.

Nonresponse Error Nonresponse errors refer to errors introduced by the practical reality that surveys almost never collect data from all sampled cases. People are often

30  D. Sunshine Hillygus

difficult to reach or they refuse to participate. In fact, most of us have probably contributed to nonresponse in a survey if we have ever hung up the phone when we realized it was a pollster on the other end of the line interrupting our dinner. There has been considerable focus on nonresponse error in recent decades and rightfully so. In recent decades, response rates have declined precipitously across government, academic, and media surveys. Given the barrage of telemarketing calls, spam, and junk mail, people are increasingly hesitant to participate in surveys. And technologies like voicemail and caller id make it easier than ever to avoid intrusions from strangers. Nonresponse error will create nonresponse bias if those who respond are different from those who do not. The most common marker for nonresponse error has traditionally been the survey response rate. In its most basic form, the response rate is calculated as the number of people you actually surveyed divided by the number of people you tried to survey. Even high-​budget “gold standard” academic surveys, such as the General Social Survey and the American National Election Study (ANES), have seen sharp declines in response rates. For example, the ANES response rate declined from 74 percent in 1992 to less than 60 percent in 2008 to 50 percent in 2016.24 Moreover, much of the decline is attributable to increasing rates of refusal. The ANES refusal rate increased from less than 15 percent in 1972 to over 24 percent in 2004. Similarly, the GSS refusal rate jumped from 16.9 percent in 1975 to 25 percent in 2010.25 Response rates for media polls have been especially hard hit by declining cooperation. Although the response rate was not reported for the poll in our example, most media telephone polls these days have response rates that rarely exceed 10 percent. The question is whether these lower response rates actually lessen data quality. Certainly, low response rates of telephone polls are often used as justification for using nonprobability samples. Some argue that the bias introduced by those who “opt out” from survey requests (nonresponse) is no different from the bias introduced by people choosing to “opt in” to online nonprobability panels. An increasing body of research has evaluated the link between response rate and nonresponse bias, and, perhaps surprisingly, has concluded that a low response rate by itself does not indicate the results are inaccurate.26 Nonresponse bias depends not just on the rate of nonresponse but the extent to which those who answer are different from those who did not.27 So, a low response rate indicates a risk of lower accuracy, but does not guarantee it. The reassuring news on response rates does not mean we can ignore nonresponse error. To the contrary, it remains a significant concern—​we have just been using an incomplete metric for evaluating its impact. In thinking about nonresponse error, it’s first worth clarifying that nonresponse can be classified in two different categories:  unit and item nonresponse. Unit nonresponse is where an individual fails to take part in a survey. This is the basis of response rate calculations. Another type of nonresponse, item nonresponse, occurs when the

The Practice of Survey Research  31

individual answering the questionnaire skips a question, giving us incomplete data on an individual respondent. Questions on income, for instance, are often susceptible to item nonresponse. Once again, the key concern is with potential differences between nonrespondents and respondents. For instance, in his book Silent Voices, Adam Berinsky shows that item nonresponse in racially sensitive survey questions can reflect prejudicial sentiments.28 For both unit and item nonresponse, the most important step in reducing the potential for nonresponse bias is to create an appropriately designed survey in the first place. Many of the fundamental design decisions, including mode, interviewer characteristics, length of survey, question wording and response options, can directly affect the extent of nonresponse error. For example, self-​administered surveys (mail and Internet) have higher levels of item nonresponse than interviewer-​administered surveys, but answers in self-​ administered surveys tend to be more accurate because of reduced pressures to give a socially desirable answer. Respondents are more likely to skip questions that are long, burdensome, confusing, vague, or that do not provide the preferred response, so it becomes especially important that the questionnaire itself follows best practice principles for the particular mode being used. Again, while response rates are perhaps not the key marker of nonresponse bias, it is nonetheless important for those conducting surveys to try to minimize nonresponse error and those consuming surveys to consider the nature and extent of nonresponse bias in any reported data.

Coverage Error One of the growing issues of concern about survey quality comes from coverage error. Coverage error is the failure to give some persons in a target population a chance of being selected into the sample, such as when those without Internet access have no chance of ending up in an Internet survey. The extent of bias resulting from coverage error depends both on the rate of noncoverage and the difference between those covered by the survey and those not. Consider, for instance, surveys that rely on a list of registered voters as a sample frame—​they will inherently miss individuals who are not yet registered (including all those individuals who use same-​day registration in states where it is available). These individuals are likely to be younger, and more Democratic—​ a fact that led polls to underestimate support for Barack Obama in 2008. These has been a lot of attention to the rise in the number of cell-​only households as a source of coverage error in telephone surveys. Cell phone-​ only respondents tend to be younger, lower income, and more likely to be racial or ethnic minorities, so omitting them can create biased estimates of public opinion. For a long time, survey vendors were hesitant to call cell phones to conduct interviews. Although there are some quality issues

32  D. Sunshine Hillygus

(e.g., blurry geographic associations, shorter questionnaires possible, lower response rates), cost is the primary reason for the exclusion of cell phone-​ only households from telephone surveys. Specifically, the 1991 Telephone Consumer Protection Act (TCPA) prohibits the use of automated dialers for all unsolicited calls to cell phones, including surveys. Pollsters typically use automated dialers—​a device that automatically calls telephone numbers until reaching a live respondent—​because it is much faster (and thus less expensive) than having interviewers manually dial the numbers. Today, cell phone-​only households are the norm—​more than 50% of all households and roughly 75% of those under the age of 30 are cell phone-​only. Given the clear implications of cell phone-​only households on the generalizability of survey estimates, many top media organizations, including ABC News, CBS News, and The New York Times—​have started routinely including cell phone respondents in their samples, despite the increased cost. Nonetheless, there remains considerable variability in the relative numbers of cell phone versus landline rates across firms. For example, some firms use just 20  percent cell phone-​only respondents, whereas Pew recently announced they would include 75 percent cell phone-​only respondents.

Measurement Error Thus far, we have focused on sources of error that shape the ability to generalize from a sample of respondents to a population of interest. But the quality of a survey depends not only on the ability to generalize, but also on the ability to accurately measure the theoretical concepts of interest. Ideally, the survey questions result in measures that are both valid—​they fully and accurately measure the concept that is supposed to be measured, and reliable—​they measure the concept in a reproducible manner. Measurement error occurs when recorded responses to a survey fail to reflect the true characteristics of the respondents, and it can influence both the accuracy and reliability of our results. There are many different sources of measurement error: the questionnaire, the data collection method, the interviewer, and the respondent. Questionnaire factors like question wording, question order, length of questions and questionnaire, number of response categories, and the presence of a “don’t know” or middle response option can all influence measurement error. Even very small differences in question wording can generate very different findings. For example, asking about attitudes toward “assistance for the poor” generates much higher levels of support than a question asking about attitudes toward “welfare.”29 In another example, party identification questions that are otherwise identical besides the beginning phrase, either “In politics today” or “Generally speaking,” result in entirely different conclusions regarding the stability of partisanship.30

The Practice of Survey Research  33

Measurement error can also be affected by the mode of survey administration (e.g., telephone, in-​person, mail). A survey that uses an interviewer in the administration, for example, can introduce measurement error from that interaction. Numerous studies have found that Whites express more liberal racial attitudes to black interviewers than to white interviewers.31 Finally, respondents themselves introduce error based on their comprehension or interpretation of the question in addition to any editing of the responses they might make because of fears of disclosure, concerns about privacy, or a desire to give a response that would be viewed favorably by others. People are especially reluctant to provide honest answers on sensitive topics, like sexual history, drug use, or racial attitudes. Voter turnout is another sensitive question—​people tend to over-​report voting because they want to appear to be good citizens. Thus, the ANES does not simply ask “Did you vote in the last election? (yes or no?).” Rather, they attempt to reassure the respondent that it really is okay to admit to not voting: In talking to people about elections, we often find that a lot of people were not able to vote because they weren’t registered, they were sick, or they just didn’t have time. Which of the following statements best describes you: 1) 2) 3) 4)

I did not vote (in the election this November) I thought about voting this time –​but didn’t I usually vote, but didn’t this time I am sure I voted

For those conducting their own surveys, it is worth remembering that substantive expertise on a topic is not the only skill needed to conduct a survey. There is a rich body of research on the nature and extent of measurement error in surveys, and emerging best practices for reducing that error.32 The single best way to improve measurement is to do extensive pretesting of the survey instrument.33 For those introducing a new measure, it is especially important to explicitly evaluate the operationalization of that measure for validity and reliability. In this regard, political science as a field could take guidance from fields like psychology or education, where it is standard practice to take measurement seriously. For those using secondary survey data, there is often a tendency to take for granted that the survey questions adequately measure the concepts of interest. However, many questions in major infrastructure surveys were written before the development of rigorous question-​wording practices. Moreover, because over time inferences depend on having identical question wording, recurring surveys like the American National Election Study face a tension between the need for continuity in question wording and the need for innovation to keep up with developing knowledge in the field of survey methodology. Ultimately, we often must “work with what we got,” but any analysis that uses survey research should pay careful attention to the potential for measurement error.

34  D. Sunshine Hillygus

The Future of Polls? Given the challenges outlined here, it is natural to ask if there might be alternatives to polls as a measure of public opinion. Should we instead look at markers like protest numbers, rallies, town hall meetings, or Twitter posts as a measure of public opinion? Fundamentally, these methods are not better than polls, unfortunately. The opinions of those who make the effort to show up at a political rally or a protest have attitudes that are quite different from the general American public. For example, the white supremacists who participated in the 2017 “Unite the Right” rally in Charlottesville, VA—​chanting racist and anti-​Semitic slogans and carrying Nazi symbols—​do not represent the views of the Trump voters or Republican identifiers broadly. And while there have been recent attempts to measure public opinion from Twitter, tweets are likewise unrepresentative of public opinion. It is well documented that Twitter users are not only much younger than the general population, but also that only a small percentage of Twitter users actually do most of the tweeting. Those who choose to tweet about politics are likely more politically engaged and extreme than the general population. Ultimately, despite the significant challenges to conducting and interpreting polls, they remain a powerful tool for understanding elections and American democracy.34 Indeed, I  would argue that some of the biggest problems with polls is in how they are misused and misinterpreted. Journalists focus too much on horserace polling during political campaigns. Surveys are better suited for measuring public opinion than for predicting election outcomes. Recall that the scientific basis for polls rests on drawing a random sample from a population of interest. In the case of elections, we don’t know the population of interest. That is, we don’t know who will show up to vote. So, pollsters have to guess. Pollsters make a guess about the population by using likely vote models or (incorrectly) assuming that anyone who is registered will vote. Pollsters often fail to share details about their likely voter model. And they often get it wrong. It turns out that most Americans say they plan to vote when asked months before the election. Many fewer will actually do so. People also change their minds about whether to vote or who they plan to vote for. The polls are designed to get respondents to pick a side: they are asked how they would vote if the election were held today; those who say undecided are pressed to make a choice. All of this means that horserace numbers should be viewed with caution and skepticism. Polling averages can capture trends in candidate support during the campaign, but polls should be viewed as just one piece of the election prediction puzzle. Polls are better suited for giving meaning to the election than for predicting the outcome. Don’t expect too much out of a single poll 90 days out from election day, but I also don’t think we should judge the polling industry based only on the pre-​election polls.

The Practice of Survey Research  35

Conclusion It would be easy to blame the media for blurring the line between quality and junk polls. After all, many mainstream news organizations sponsor both open-​ access “straw polls” on their websites as well as traditional, scientific surveys—​ and fail to distinguish the methodological differences between the two. ABC News polling director Gary Langer chides the news media for indulging in “the lazy luxury of being both data hungry and math phobic.” Journalists value the credibility and authority that survey numbers add to a story, but they often fail to scrutinize those numbers for methodological rigor. The media, however, are not the only ones to blame. In academia, we have also seen increasing variability in survey quality. Surveys that would fail to meet the minimum quality standards of the top news organizations are currently being published in social science journals. Even with academic surveys, the increased variability in survey design makes it more difficult, but also more important, to assess survey quality. Ultimately, the ability to assess survey quality—​across all sources of survey error—​rests on having sufficient information about the survey methodology. Although most academic journals and media organizations do not have formal disclosure requirements in place, there are increasing pressures on survey users to improve methodological transparency. In the last few years, there have been at least two well-​publicized incidents in which survey firms appear to have made-​up or manipulated survey results. The liberal blog, DailyKos, discovered that weekly polling results they had paid for and featured from the organization Research 2000 (R2K) were “largely bunk.”35 Likewise, blogger Nate Silver of fivethirtyeight.com concluded that pollster Strategic Vision, LLC was “disreputable and fraudulent.”36 AAPOR publicly reprimanded Strategic Vision for failure to disclose basic methodological information about the studies. Not long after, they announced a transparency initiative aimed at encouraging and making it as easy as possible for survey firms to be transparent about their research methods. Basic standards for minimal disclosure include reports of the following information about a survey:37 1 . Who sponsored the survey, and who conducted it. 2. The exact wording of questions asked, including the text of any preceding instruction or explanation to the interviewer or respondents that might reasonably be expected to affect the response. 3. A definition of the population under study, and a description of the sampling frame used to identify this population. 4. A description of the sample design, giving a clear indication of the method by which the respondents were selected by the researcher, or whether the respondents were entirely self-​selected.

36  D. Sunshine Hillygus

5. Sample sizes and, where appropriate, eligibility criteria, screening procedures, and response rates computed according to AAPOR Standard Definitions. At a minimum, a summary of disposition of sample cases should be provided so that response rates could be computed. 6. A discussion of the precision of the findings, including estimates of sampling error, and a description of any weighting or estimating procedures used. 7. Which results are based on parts of the sample, rather than on the total sample, and the size of such parts. 8. Method, location, and dates of data collection. Full methodological disclosure should make clear that every survey is flawed in some way. There is no perfect survey design in part because there are inevitable trade-​offs involved in balancing the various sources of survey error. In reducing one source of survey error a researcher could inadvertently increase another source of error. For example, best practices for measurement error would have multiple questions about each concept of interest, but doing so lengthens the survey and thus might increase the number of people who skip questions or drop out of the survey because of the time burden. Because no survey is perfect, every analysis of survey data should explicitly discuss how the results might or might not be affected by various survey errors. Greater levels of transparency will give readers the ability to evaluate whether the knowledge claims being made are warranted given the methodology used. Increased transparency might also offer incentives to researchers to employ higher quality methods because it should make clear that not all survey methods are equal. Currently there seem to be two standards for surveys: gold and tin. The budgets of some of the most important federal and academic “gold standard” surveys are increasing dramatically in an effort to maintain the same levels of quality by traditional metrics; yet even these budgets are often not sufficient to maintain traditional metrics. The proliferation and variability in survey methods increases the importance of high-​quality government-​ funded benchmark surveys, like the American National Election Study and the Current Population Survey, because they provide the standard against which nonprobability samples can be evaluated and adjusted. At the same time, an extraordinary amount of research is currently conducted on modest budgets, yet falls dramatically short on many standards. A clearer understanding of the sources of survey errors and a full disclosure of survey methodology will help survey practitioners and consumers better understand and evaluate the potential trade-​offs involved in using new or emerging technologies. Most importantly, it will make clear that there is no one answer to the question, “Can we ever trust the polls again?”

Notes 1 https://​newrepublic.com/​article/​139158/​2016-​can-​ever-​trust-​polls-​again.

The Practice of Survey Research  37 2 Herb Weisberg, The Total Survey Error Approach:  A Guide to the New Science of Survey Research (Chicago, IL: The University of Chicago Press, 2005); Robert Groves et al., Survey Methodology (New York: John Wiley and Sons, 2004). 3 Weisberg, The Total Survey Error Approach. 4 We could even be more specific in defining our target population, in this case, as adults reachable at home April 11–​15, 2017. 5 Not mentioned are a number of other important steps, including choosing the precision level necessary, choosing the response mode, drafting the questionnaire, pretesting the instrument, data processing, and analysis. These steps, too, can introduce error in the resulting statistics. 6 The distinction between error and bias is worth highlighting. Error can be either systematic or random; systematic error creates bias, decreasing the validity of survey estimates. Random error creates uncertainty, decreasing the reliability of the estimates. 7 The game of horseshoes offers an analogy for thinking about the correct interpretation of margin of error. Consider a blindfolded horseshoe champion who makes ringers 95 percent of the time. After a single pitch of the shoe, she doesn’t know whether or not she made a ringer, but she is 95  percent confident that a ringer was made. 8 It is perhaps also worth noting that estimates of sampling error (margin of error; standard errors) almost always are calculated assuming the survey was collected using simple random sampling. Yet, most major data collections use a more complex probability sampling design such as clustered or stratified sampling. Although procedures exist in statistical packages like R and STATA for correcting the standard errors to account for complex designs, it is rarely ever done in political science. As such, published political science research often underestimates standard errors. 9 http://​online.wsj.com/​article_​email/​SB100014240527487033054045756104021 16987146-​lMyQjAxMTAwMDEwNjExNDYyWj.html; http://​democrats.science. house.gov/​ Media/​ f ile/​ C ommdocs/​ h earings/​ 2 010/ ​ E nergy/ ​ 1 7nov/ ​ M ichaels_​ Testimony.pdf. 10 Nora Cate Schaeffer and Jennifer Dykema (2011) “Questions for Surveys: Current Trends and Future Directions.” Public Opinion Quarterly 75.5 (2011): 909–​961. 11 There are also heightened concerns about data quality, particularly for Internet panels. For a detailed overview, see Mario Callegaro, Reg P. Baker, Jelke Bethlehem, Anja S. Göritz, Jon A. Krosnick, and Paul J. Lavrakas (Eds.), Online Panel Research: A Data Quality Perspective (New York: John Wiley & Sons, 2014). For instance, it can be difficult for researchers to verify that the person taking the survey is the desired respondent. Some have found such respondents provide “don’t know” responses at a higher rate, are less likely to provide differentiated responses across items, and are more likely to avoid responding to individual items altogether. 12 Mick Couper and Peter Miller, “Web Survey Methods.” Public Opinion Quarterly 72 (2008). 13 Indeed, the web, as a mode, has a number of unique advantages. For instance, web-​ based surveys are convenient for both interviewers and subjects—​respondents can decide when to answer rather than having dinner interrupted by a phone survey. Researchers have shown that the web-​based mode is quite resistant to social desirability biases. See Frauke Kreuter, Stanley Presser, and Roger Tourangeau, “Social

38  D. Sunshine Hillygus Desirability Bias in CATI, IVR, and Web Surveys.” Public Opinion Quarterly 72 (2008); Don Dillman, “Why Choice of Survey Mode Makes a Difference.” Public Health Reports 121 (2006). 14 The 2016 ANES is a terrific example of how to conduct a probability-​based online survey, but it is also worth noting that the cost of the survey remains significantly higher than nonprobability online surveys. ANES staff estimated that cost-​per-​case was roughly $400 per completed interview, which is significantly less expensive than the $3,000 per completed interview for the probability-​based face-​to-​face interview, but is much higher than the $5–​$20 cost-​per-​case of online nonprobability samples. 15 Researchers have evaluated the success rates from various recruitment strategies. See R. Michael Alvarez, Robert Sherman, and Carla VanBeselaere, “Subject Acquisition for Web-​Based Surveys.” Political Analysis 11 (2003). They found, for instance, that their banner ad was displayed over 17  million times, resulting in 53,285 clicks directing respondents to the panel Web site, and ultimately 3,431 panel members. 16 Another concern for online panels is that they create a pool of professionalized survey-​ takers. See, for instance, D.  Sunshine Hillygus, Natalie Jackson, and M.  Young. “Professional Respondents in Non-​probability Online Panels.” Online Panel Research: A Data Quality Perspective (2014): 219–​237. People also drop out of panels over time (i.e., attrition), leading to a panel of respondents who are more interested in taking online surveys. 17 The matching procedure might work as follows. First, the survey is administered to a sample of opt-​in respondents. Next, a random sample of individuals from existing consumer and voter registration files is drawn, but not administered in the survey. Finally, a matching procedure is used to find the opt-​in respondent (who answered the survey) who most closely matches the randomly selected individual (who did not answer the survey). Survey weighting is a post-​survey procedure that adjusts the sample to look more representative on some observed characteristics. For example, if the sample of respondents is 60 percent female, 40 percent male, but the target population is evenly split between the two, then we might weigh each man in the sample a bit more and each woman a bit less. For a more detailed description of the matching procedure used by one survey firm, see Stephen Ansolabehere and Douglas Rivers. “Cooperative Survey Research.” Annual Review of Political Science 16 (2013): 307–​29. 18 A large number of studies—​at least 19—​have examined survey results with the same questionnaire administered to probability samples and online to nonprobability samples. See, for instance Yeager et al., “Comparing the Accuracy of RDD Telephone Surveys and Internet Surveys Conducted with Probability and Non-​Probability Samples.” Working Paper (Knowledge Networks, 2009), www.knowledgenetworks. com/​insights/​docs/​Mode-​04\_​2.pdf. All but one found significant differences in the results that could not be substantially reduced by weighting. Unfortunately, most of these studies cannot adequately distinguish differences due to sampling design effects and differences due to mode effects. 19 Mick Couper, “Web Surveys: A Review of Issues and Approaches.” Public Opinion Quarterly 64 (2000); Jill Dever, Ann Rafferty, and Richard Valliant, “Internet Surveys:  Can Statistical Adjustments Eliminate Coverage Bias?” Survey Research Methods 2 (2008); Linchiat Chang and Jon Krosnick, “National Surveys via RDD

The Practice of Survey Research  39 Telephone Interviewing versus the Internet: Comparing Sample Representativeness and Response Quality.” Public Opinion Quarterly 73 (2009); Neil Malhotra and Jon Krosnick, “The Effect of Survey Mode and Sampling on Inferences about Political Attitudes and Behavior: Comparing the 2000 and 2004 ANES to Internet Surveys with Nonprobability Samples.” Political Analysis 15 (2007). 20 http://​assets.pewresearch.org/​wp-​content/​uploads/​sites/​12/​2018/​01/​26170902/​ Weighting-​Online-​Opt-​In-​Samples.pdf. 21 Gary Langer and Jon Cohen, “5 Tips for Decoding Those Election Polls.” The Washington Post, December 30, 2007 Sunday B03. 22 Courtney Kennedy, Andrew Mercer, Scott Keeter, Nick Hatley, Kyley McGeeney, and Alejandra Gimenez, Evaluating Online Non-​probability Surveys. (Washington, DC: Pew Research Center, 2016). Available online www.pewresearch.org/​2016/​05/​ 02/​evaluating-​online-​nonprobability-​surveys/​. 23 http://​assets.pewresearch.org/​wp-​content/​uploads/​sites/​12/​2018/​01/​26170902/​ Weighting-​Online-​Opt-​In-​Samples.pdf. 24 These response rates were computed according to AAPOR’s “minimum response rate” (RR1). It is referred to as the “minimum” because it assumes that all households in which the eligibility of residents was not determined, at least one eligible adult lived there. Numbers come from www.electionstudies.org/​overview/​dataqual.htm. and individual study year codebooks. 25 In contrast to the ANES, the GSS estimates are calculated based on AAPOR RR5, sometimes called the “maximum response rate,” because it assumes that there are no eligible cases among the cases of unknown eligibility. Calculations come from James Allan Davis, Tom W. Smith, and Peter V. Marsden, General Social Surveys, 1972–​2008: Cumulative Codebook. Chicago, IL: National Opinion Research Center. 26 Emilia Peytcheva and Robert Groves, “Using Variation in Response Rates of Demographic Subgroups as Evidence of Nonresponse Bias in Survey Estimates.” Journal of Official Statistics 25 (2009). 27 Scott Keeter et al., “Gauging the Impact of Growing Nonresponse on Estimates from a National RDD Telephone Survey.” Public Opinion Quarterly 70 (2006); Penny Visser et al., “Mail Surveys for Election Forecasting? An Evaluation of the Columbus Dispatch Poll.” Public Opinion Quarterly 60 (1996). 28 Adam Berinsky, Silent Voices: Opinion Polls and Political Representation in America (Princeton, NJ: Princeton University Press, 2004). 29 Tom Smith, “That Which we Call Welfare by any other Name Would Smell Sweeter:  An Analysis of the Impact of Question Wording on Response Patterns.” Public Opinion Quarterly 51 (1987). 30 Paul Abramson, and Charles Ostrom, Jr., “Macropartisanship:  An Empirical Reassessment.” American Political Science Review (1991). Paul Abramson et  al., “Question Form and Context Effects in the Measurement of Partisanship: Experimental Tests of the Artifact Hypothesis.” American Political Science Review 88 (1994). 31 See, for example, Darren Davis, “Nonrandom Measurement Error and Race of Interviewer Effects Among African-​ Americans.” Public Opinion Quarterly 61 (1997). 32 Paul Biemer et al., Measurement Errors in Surveys (New York: John Wiley and Sons, Inc., 1991); Colm O’Muircheartaigh, Measurement Error in Surveys:  A Historical Perspective (1997).

40  D. Sunshine Hillygus 33 Mick Couper et  al., Methods for Testing and Evaluating Survey Questionnaires (New York: Wiley, 2004). 34 www.nytimes.com/ ​ r oomfordebate/ ​ 2 015/ ​ 1 1/ ​ 3 0/ ​ d oes- ​ p olling- ​ u ndermine-​ democracy/​polls-​can-​give-​people-​a-​stronger-​voice. 35 For more complete discussion of the controversy and evidence, see www.dailykos. com/​storyonly/​2010/​6/​29/​880185/​-​More-​on-​Research-​2000. 36 www.fivethirtyeight.com/​search/​label/​strategic%20vision. 37 www.aapor.org/​ A M/​ Template.cfm?Section=Standards_​ a ndamp_​ E thics& Template=/​CM/​ContentDisplay.cfm&ContentID=2397.

Chapter 2

Citizen Competence and Democratic Governance Martin Gilens

The eminent political theorist Robert Dahl asserted that “… a key characteristic of democracy is the continuing responsiveness of the government to the preferences of its citizens, considered as political equals.”1 This formulation implies first that citizens must have meaningful preferences for democratic government to be possible, and second that in order to gauge the democratic quality of any given government we must be able both to discern what its citizens’ preferences are, and to assess how strongly and how equally government policy responds to those preferences. The American public, it is often claimed, simply lacks the interest, motivation, or ability to form meaningful preferences on most political issues. Political issues in contemporary societies are numerous, complex, and often remote from citizens’ everyday lives. And the preferences citizens do express often seem to shift erratically over time, or change in response to seemingly minor differences in how a question is posed, or reflect a blind willingness to adopt whatever position their own party’s leaders endorse. In this chapter I  ask first whether American citizens hold meaningful policy preferences and whether such preferences, if they exist, are accurately reflected in surveys of political attitudes. I argue that despite the limited political knowledge and engagement typically displayed by the American public, public preferences—​at least in the aggregate—​are “enlightened enough” to serve as a reasonable basis for guiding government decision-makers on a wide range of issues. Given this positive evaluation, I then ask how the preferences of the public are related to the policy decisions of our national government, and how equally influence over government policy extends to more and less well-​off Americans. Analyzing hundreds of survey questions about public policies over a 40-​year period, I find that only affluent Americans have any discernable influence over what their government does. Policymakers essentially ignore the preferences of poor and middle-​class Americans who only see the policies they favor adopted when their preferences happen to coincide with the powerful and well-​to-​do.

42  Martin Gilens

Competing Views of Citizen Competence In his seminal 1964 paper “The Nature of Belief Systems in Mass Publics,” Philip Converse2 painted a bleak picture of the American public as largely lacking coherent political preferences. Converse observed that survey respondents were apt to express different preferences when presented with the identical question on different occasions, that preferences on one policy issue were at best weakly associated with preferences on seemingly related issues, and that broad organizing principles like liberalism or conservatism were poorly understood by most Americans. Confronted with this evidence, Converse concluded that the preferences respondents report on surveys consist largely of “non-​ attitudes” and that “large portions of [the] electorate do not have meaningful beliefs, even on issues that have formed the basis for intense political controversy among elites for substantial periods of time.”3 Many contemporary scholars have come to similar conclusions. The most trenchant recent analysis of Americans’ shortcomings as democratic citizens is by political scientists Christopher Achen and Larry Bartels.4 Achen and Bartels describe the “folk theory of democracy” as the notion that citizens form preferences about what policies they want their government to adopt and elect leaders who enact those policies. When officeholders fail to adopt popular policies, or produce undesirable outcomes like a poor economy or an unsuccessful war, they are voted out of office. By Achen and Bartels’ account, this idealized model does not reflect the reality of contemporary America (or of any large-​scale democratic society of any era, for that matter). Achen and Bartels argue that citizens are uninformed about policy issues, that the information they do hold is biased to conform with their preexisting partisan leanings, that the policy preferences citizens express are inconsistent over time and easily altered by the way a policy is described, that citizens tend to blindly adopt the policy positions espoused by their own party’s leaders, and that voters do a very poor job of holding incumbents accountable for their performance in office (often focusing exclusively on the most recent trends or events and frequently blaming incumbents for developments they could not possibly control). Other scholars take a more sanguine view of the quality of citizens’ policy preferences. These scholars point to three aspects of mass political attitudes to explain how a public with minimal political information can nevertheless form meaningful issue preferences. First, citizens with modest levels of information might turn to more knowledgeable others for “cues” about the desirability of alternative policies or politicians. Second, individual citizens are not equally interested in the full range of political issues in play at any given time but tend to “specialize” in a subset of issues about which they are more knowledgeable and have more stable and well thought out preferences. The division of citizens into these “issue publics” means that the ability of any individual citizen to meaningfully participate in shaping government policy should be judged relative

Citizen Competence and Democratic Governance  43

to the set of issues that that individual cares about; all citizens need not hold equally well-​developed preferences on all issues for the public to fulfill its role in democratic governance. Finally, the fickle element of individual citizens’ policy preferences will, to some degree at least, tend to cancel out when preferences are aggregated across the public as a whole (or across distinctive subgroups of the public). Aggregate opinion, by this reckoning, will typically be more stable, with a higher “signal to noise ratio” than the individual opinions that make it up. Scholars also disagree about citizens’ ability to evaluate incumbent officeholders’ performance. “Pessimists” like Achen and Bartels, argue that voters often rely on irrelevant factors or attend only to recent events. “Optimists” counter that voters are far from perfect but nonetheless use their available information in sensible ways to shape their vote choices. From this perspective, shortcomings in the information environment rather than voters’ inadequacies are often to blame when retrospective evaluations of incumbents’ performance go awry.

Citizens’ Information Deficits In the most comprehensive assessment of Americans’ political knowledge (based on hundreds of survey measures of political information) Michael Delli Carpini and Scott Keeter conclude that “… large numbers of American citizens are woefully underinformed,” and “overall levels of knowledge are modest at best.”5 In recent surveys, for example, only six in ten Americans know that it is the Supreme Court (not Congress or the President) that determines if a law is constitutional, only four in ten know that a U.S. Senate term lasts six years, and only one in four Americans can name one of their two U.S. Senators (indeed, only 46 percent of American adults are aware that each state has two U.S. Senators).6 Americans’ knowledge of politically relevant social conditions is also spotty. Most Americans, for example, believe that crime has been getting worse over the past decades (when crime rates have actually plummeted),7 and most think that foreign aid accounts for over 20 percent of the federal budget (when the true figure is under two percent).8 Of course, exactly what kind of knowledge citizens need in order to fulfill their role in democracy is unclear.9 “Civics tests” may reveal the public’s lack of information about basic aspects of American government, but knowing how long a Senator’s term lasts, or how the constitutionality of a law is determined, may be irrelevant to most of the tasks citizens are asked to perform in a democracy. Similarly, keeping tabs on how one’s Senators have voted, or even what their names are, may not be necessary if the relevant information is available at the time the next election comes around. Scholars recognize that no citizen could possibly keep track of all the issues that are (or should be) on the political agenda, all the actions their elected officials take, or all the politically relevant conditions and events that impact

44  Martin Gilens

their lives. A relatively few “news junkies” may habitually acquire a lot of knowledge about a wide range of issues, but many citizens attend only to the few issues they care about most and others may pay attention to public affairs only during election seasons, or not at all. But even the most knowledgeable and engaged citizens depend on intermediaries to identify the most salient and pressing information. The media (of all stripes) play a primary role, as do political leaders and advocacy groups. Studies that show that most people lack solid information on most issues in fact tell us little about the ability of citizens to form meaningful preferences on the issues they care about when those issues become politically salient. As democratic “optimists” argue, citizens’ lack of information can be overcome by taking cues from trusted others, by “specializing” in a small set of issues that they care about most, and by combining the views of thousands of individuals.

Cue-​t aking as a Basis for Political Preferences In a landmark study in the 1950s, Berelson, Lazarsfeld, and McPhee explained how citizens take cues about specific issues from better-​informed “opinion leaders” who they believe hold compatible interests and outlooks. Social networks allow for a division of labor in which the more informed provide policy insights and endorsements to their less-​informed friends and acquaintances. “The political genius of the citizenry,” they wrote, “may reside less in how well they can judge public policy than in how well they can judge the people who advise them how to judge public policy.”10 Taking cues from more knowledgeable elites or acquaintances is a sensible strategy for citizens who lack the ability or inclination to gather the information needed to formulate a preference on a given policy issue. Anthony Downs, writing shortly after Berelson, Lazarsfeld, and McPhee, notes that the average citizen “cannot be expert in all the fields of policy that are relevant to his decision. Therefore, he will seek assistance from men who are experts in those fields, have the same political goals he does, and have good judgment.”11 A substantial literature has developed over the past decades which identifies the wide range of cue-​givers that citizens can rely on in forming political judgments.12 Most cue-​taking models posit that citizens adopt the policy positions expressed by “like-​minded” elites (judged on the basis of partisan or ideological compatibility, or the more specific affinities associated, for example, with a citizen’s religious, union, or professional organization) and either ignore those of the “non-​like-​minded” or adopt the opposite position from the one that they espouse.13 Cue-​givers of this sort can be either social leaders whose views are transmitted through the media, or individual acquaintances who are perceived as comparatively well informed on the issue at hand.

Citizen Competence and Democratic Governance  45

The strategy of turning to those with greater knowledge when faced with a challenging decision is hardly limited to political novices. Even citizens who follow politics closely will inevitably lack sufficient information (or technical expertise) to form opinions “from scratch” on many issues. In a modern nation, there are simply too many detailed and technical issues for even the most motivated members of the public to possibly keep abreast of. Indeed, even elected representatives who have abundant informational resources and who “follow politics” for a living turn to experts in specific issue areas for advice and take cues from other representatives in their own party who “specialize” in particular issue domains.14 Not all cue-​taking is equally desirable from the perspective of democratic governance. If citizens are unable to identify cue-​givers who share their interests and outlooks (perhaps because those cue-​givers are dishonest about their true motivations), or if cue-​takers are too willing to adopt any position endorsed by the leaders of their party (or union or religious congregation), even when it conflicts with the cue-​takers’ interests and values, then citizens can end up holding positions that serve them poorly. Americans who identify as Democrats or Republicans frequently adopt their chosen candidate or party’s position as new issues come onto the agenda. As Gabriel Lenz has shown, this tendency to “follow the leader” is much more common than the reverse tendency to choose a party or candidate on the basis of a voter’s existing stance on an issue.15 But voters do sometimes abandon their party or withhold their vote from a same-​party candidate on the basis of a strongly held issue preference. As Achen and Bartels demonstrate, when the Democratic and Republican parties began to distinguish themselves on abortion policy in the 1970s, a large number of voters shifted their abortion views to match their party identification while another large group of voters shifted their party identification to match their views on abortion (and yet other voters simply maintained their preferences on both abortion policy and partisan identification, despite the new tension between the two).16 That voters sometimes hold issue preferences that conflict with the positions of the party they identify with suggests the limits to “follow the leader” type behavior. Sunshine Hillygus and Todd Shields offer an in-​depth look at these kinds of “cross-​pressured” citizens. They find that most Americans—​whether strong or weak partisans, well or poorly educated, most or least informed about politics—​disagree with their party on at least one prominent issue.17 Taking cues from better-​informed “opinion leaders” offers no guarantee that cue-​takers will arrive at the same positions they would have if they had the time, inclination, and expertise to evaluate the issue on their own. I’ll return below to the question of how well cue-​taking (and other mechanisms) work in allowing citizens to form meaningful policy preferences. For now, it is enough to note that citizens appear to rely heavily on cues from others to form their preferences but that they are willing and able to resist those cues when the cues conflict with their strongly held preferences or beliefs.

46  Martin Gilens

Issue Publics Among the many enduring contributions of Converse’s seminal paper was the concept of issue publics—​the obvious but often overlooked fact that different people care about different political issues. In order to participate in democratic governance, citizens must be able to form meaningful preferences on the policy issues that government addresses. But that does not mean that every citizen must have a preference on every issue. Given the broad range of backgrounds, interests, and situations that citizens in a large and diverse society face, it would be surprising if there were not substantial variations in the specific political issues that different citizens care about and attend to. Converse based his negative assessment of the mass public’s political preferences in part on the substantially stronger “constraint” (i.e., the stronger associations among preferences on related issues) among the political elites he surveyed. (Converse’s sample of political elites consisted of candidates for the U.S. Congress, arguably an unrealistically sophisticated comparison group.) Nevertheless, when Converse restricted his analysis of the public’s policy positions in a given issue domain like foreign aid or racial policy to those respondents who he judged to be members of a given issue public,18 he found that the inter-​correlations among ordinary Americans resembled those among his political elites: “… removal from analysis of individuals who, through indifference or ignorance, lie outside the issue publics in question serves to close much of the gap in constraint levels between mass and elite publics.”19 Subsequent analyses confirm Converse’s insight regarding issue publics.20 Jon Krosnick, for example, sorted survey respondents into issue publics on the basis of the level of importance they attached to a dozen different political issues.21 Krosnick reported that the greater the importance a respondent attached to a given policy issue, the more likely they were to mention that issue as a reason for liking or disliking the presidential candidates, the less likely they were to change their issue preference in response to persuasive communications, and more stable their reported issue preference was over time. Another technique for identifying issue publics is to rely on demographic group membership on the assumption that members of particular groups are, at least on average, more interested in some issues than others. Vincent Hutchings, for example, identifies union members and people living in union households as more likely to have an interest in labor issues while abortion policy is likely to be of greater interest to women and religious conservatives.22 Consistent with these expectations, he finds that members of these groups are more attentive to Senate and gubernatorial campaigns when “their” issues were raised and more likely to base their Senate votes on their Senator’s record on the particular issues associated with their group. Research on issue publics suggests that assessments of the quality of public preferences that look only at the average level of knowledge, preference stability, or other measures across the public as a whole may strongly understate

Citizen Competence and Democratic Governance  47

the degree to which a typical citizen holds meaningful policy preferences. True, the typical citizen may attend to only a few of the many issues facing the country at any point in time. But if citizens have sensible, stable, and reasonably informed preferences on the subset of issues that they care most about, and if they use those issues disproportionately as a basis for choosing among parties and candidates, then the public can fulfill its assigned role in democratic governance, even if most citizens lack meaningful opinions on most issues.

The “Magic of Aggregation” and the Quality of Public Preferences Cue-​taking suggests that even citizens with minimal information may be able to form meaningful preferences by relying on others who share their general outlooks or political orientations, and the division of the citizenry into issue publics suggests that meaningful participation in democratic governance does not require all citizens to hold meaningful preferences on all issues. A  third factor relevant to the assessment of the public’s role in democracy is that the aggregate preferences of the public as a whole have different characteristics than the individual preferences that make them up. The eighteenth-​century French philosopher and mathematician Condorcet explained in his famous “jury theorem” that if each individual in a group has even a modest tendency to be correct, the group as an aggregate can have a very high probability of reaching the correct decision (and the larger the group, the higher the probability that the collective judgment will be correct). This insight has been applied to the political attitudes expressed on surveys to suggest that the “errors” in respondents’ reports of their own preferences will, at least under some circumstances, tend to cancel out, resulting in a measure of aggregate opinion that is more stable and more reliable than the individual opinions that make it up.23 But how can respondents be “wrong” about their own preferences? Two different kinds of “errors” in survey-​based measures of policy preferences can be distinguished. First, even if respondents had perfectly fixed and certain views on a particular policy option, the reports of those views as captured on surveys would contain some degree of error. The ambiguities of question wording, the difficulty in matching a specific sentiment to the available response options, and mistakes in reading or hearing the survey question or recording the respondent’s answer will all introduce some degree of “measurement error” (see Chapter 1 of this volume). Second, most respondents are not likely to have perfectly fixed and certain views on most political issues. As John Zaller’s influential studies of public opinion show, citizens typically hold a variety of considerations relevant to a given policy issue and use those considerations to construct a position on a policy question when asked by a survey interviewer.24 For example, if asked whether they favor cutting government spending on foreign aid, a survey

48  Martin Gilens

respondent might consider his or her views about taxes and government spending, about humanitarian needs in developing countries, about waste and corruption in those countries, about competing domestic needs, and so on. This process of canvassing considerations and constructing positions is an imperfect one, however. Given the time and motivational constraints typical of a survey interview, only a subset of all possible considerations bearing on a particular question are likely to be brought to mind. Moreover, this subset of considerations may be biased toward those that are at the “top of the head” as a result of earlier questions in the survey, stories that have been in the news, recent experiences the respondent may have had, specific aspects of the question wording, or any number of other reasons. From this perspective, most citizens cannot be said to “have attitudes” corresponding to a particular survey question on a political issue, in the sense that those attitudes existed in a crystallized form before the question was asked.25 But individual citizens can be said to have “response tendencies” or “long-​term preferences” which represents their (hypothetical) average opinion if it were to be ascertained repeatedly over time.26 This Platonic “true attitude” is nothing more than the imperfectly revealed average of these hypothetical repeated preference constructions (in the same way that a “true circle” is a hypothetical shape that can only be approximated by any actual circle in the real world). It is impractical, of course, to measure citizens’ “long term preferences” by repeatedly surveying the same individuals. But aggregating survey responses across many individuals will produce much the same result (without the problem of dealing with new information or changed circumstances which might alter the set of relevant considerations). To the extent that the biases in formulating a preference from a given set of considerations are randomly distributed across individuals, they will balance out, just as the errors in individuals’ judgments in a jury context cancel each other out. If randomly distributed idiosyncratic factors lead some citizens to under-​report support for a given policy, they will lead others to over-​report such support. With a large enough sample of citizens, these errors will cancel out resulting in aggregate preferences that closely match the average of the individuals’ long-​term preferences. Of course, not all factors that lead citizens to wrongly report their issue preferences will be random and therefore offsetting: a concern I’ll return to below. The most thorough examination of aggregate opinion toward public policy is Benjamin Page and Robert Shapiro’s influential book The Rational Public.27 Page and Shapiro do not view aggregation as a cure for all of the shortcomings of public opinion. But they argue that collective preferences display a degree of stability and cogency that far exceeds the typical individual level preferences that make them up. While we grant the rational ignorance of most individuals, and the possibility that their policy preferences are shallow and unstable, we maintain

Citizen Competence and Democratic Governance  49

that public opinion as a collective phenomenon is nonetheless stable (though not immovable), meaningful, and indeed rational … it is able to make distinctions; it is organized in coherent patterns; it is reasonable, based on the best available information; and it is adaptive to new information or changed circumstances.28 Moreover, they maintain, “surveys accurately measure this stable, meaningful, and reasonable collective public opinion.”29 The collective rationality of public opinion stems, Page and Shapiro argue, from the aggregation of individual opinions which cancel out both random measurement errors in surveys and temporary fluctuations in individuals’ opinions. The aggregate preferences that result from this process tend to be quite stable, but also exhibit sensible responsiveness to changing conditions. For example, public support for unemployment assistance increases as unemployment rates rise, public support for defense spending increases when the threat of war goes up, public support for tax cuts declines when tax rates are lowered, and so on. Two principal objections have been raised about the “miracle of aggregation.” The first, which Page and Shapiro discuss at some length, is that “errors” in individuals’ policy preferences will not always be randomly distributed. One source of non-​random “error” in preference formation is misinformation that leads most or all members of the public to shift their policy preferences in the same direction. For example, John F. Kennedy and others claimed during the 1950s that the U.S. was facing a nuclear “missile gap” with the Soviet Union. In retrospect it is clear that not only was there no missile gap (the U.S. had and maintained a considerable advantage in nuclear missiles) but that good evidence was available at the time, demonstrating the absence of such a gap. This sort of misinformation will inevitably “pervert” the preferences that the public would otherwise hold on related policy issues (in this case, defense spending and foreign policy). Shared misinformation need not result from purposeful attempts to mislead the public. Sizable misperceptions of changes in the crime rate, spending levels on foreign aid, the racial composition of the poor, and the typical length of time beneficiaries receive welfare have all been widespread among the American public at various points in time.30 The extent of collectively held misinformation among the public is difficult to assess, in part because the truth about many politically relevant facts may not become known until later (if ever). After canvassing some of the sources and content of misinformation held by Americans, Page and Shapiro conclude: we cannot hope to offer a precise or definitive account of the extent (or, for that matter, the nature) of information biases in the United States. But if we are on track concerning important instances of opinion manipulation and general patterns of biased and misleading information, these pose troubling implications for the workings of democracy.31

50  Martin Gilens

Just how troubled we should be about biased or misleading information is difficult to judge. To the extent that misinformation is universal (or nearly universal) among elites and the public at large, it is hard to see how any form of government could make optimal decisions. The consequences of misinformation that are unique to democracy, on the other hand, are those in which large numbers of citizens fall prey to avoidable misperceptions or biases. For example, if the preferences of the majority of citizens were influenced by misinformation that the best-​informed citizens knew to be untrue, then a democratic government that reflected the public’s collective preference might do a poor job of serving the public’s true interests.

How Well does Cue-​t aking and Aggregation Work? To what extent do cue-​taking, preference aggregation, and issue publics ameliorate concerns about low levels of political information and the low quality of public preferences on political issues? No actual public in a large society is likely to meet the classical expectations of the well-​informed citizen. But does the existing public display enough “wisdom” in its political preferences to recommend a system of governance that strongly reflects the preferences of the public? We know that cue-​taking can be an effective strategy for forming policy preferences on complex issues. In one study, for example, respondents who were poorly informed about the details of five competing insurance-​reform initiatives on a California ballot, but who knew where the insurance industry stood on each initiative, were able to closely emulate the voting behavior of their better-​informed peers.32 But just because cues can serve as effective shortcuts doesn’t mean the necessary cues are always available or that citizens will make use of them when they are. One way to assess the quality of public preferences that emerge from the processes described above is to compare the actual preferences expressed on surveys to some hypothetical standard of “well-​informed preferences” that citizens would hold if they had the ability, time, and inclination to gather the relevant information on a given set of policy issues. The most straightforward way to assess how far actual preferences diverge from hypothetical well-​informed preferences is to inform a representative group of citizens about some set of policy issues and see how their preferences shift as a result. James Fishkin and Robert Luskin have done just this in a series of “deliberative polls.”33 For example, the 1996 National Issues Convention brought 466 participants, selected at random from the U.S.  population, to Austin, Texas for four days, during which time they read briefing materials on various economic, foreign policy, and family issues, discussed those issues in small groups, and participated in question-​and-​answer sessions with experts. When initially contacted, and once again at the end of their stay in Austin, participants answered identical questions concerning their policy preferences

Citizen Competence and Democratic Governance  51

in these three issue areas. To provide a comparison group, members of the initial sample who elected not to come to Austin completed the same surveys. The participants in the National Issues Convention did shift their preferences somewhat on many of the 48 political attitude questions they were asked. But the average change in aggregate preferences was not large and barely exceeded the aggregate change of preferences expressed by the control group which was not provided with the information or opportunity to deliberate. On a 100-​ point scale, the average net (i.e., aggregate) difference in pre–post preferences across these 48 issue questions was about five points for the deliberation group and about three points for the control group.34 The four days of focused study and deliberation, it appears, resulted in a two percentage point greater aggregate change in policy preferences than would otherwise be expected by simply resurveying the same respondents with no intervening activity. The results of the National Issues Convention study suggest that on the topics addressed, participants’ pre-​ existing aggregate preferences closely resembled the “well-​informed preferences” they expressed after four days of education and deliberation. But these conclusions hinge on the specific information provided to the deliberating respondents. If the information provided was not new to the participants, or was not different enough from what they already knew, or was not relevant enough to the policy judgments they were asked to make, then the possibility remains that different information might have shifted aggregate preferences to a greater degree. Nevertheless, since the organizers’ goal was to provide just the sort of educational experience that critics of the quality of public opinion view as lacking, these results do lend some credibility to the notion that cue-​taking and aggregation result in collective judgments that differ little from what a well informed and engaged citizenry would express. A very different way to compare actual to hypothetical “well-​informed” preferences is to use statistical tools to simulate a well-​informed citizenry. This approach takes advantage of the fact that, as Philip Converse observed, the mean level of political knowledge among the electorate is very low, but the variation in knowledge is very high.35 By modeling the vote choices or policy preferences of the most well-​informed segment of the electorate, one can impute preferences for citizens who share a given set of characteristics but have lower levels of political information. Larry Bartels, for example, compared the presidential votes of the most well-​ informed respondents with those of less-​informed respondents of the same age, education, income, race, sex, occupational status, region, religion, union membership, urban residence, homeowner status, and labor force participation.36 Bartels found an average individual deviation of about 10  percentage points between actual and “well-​ informed” votes for the six presidential elections between 1972 and 1992. Many of these deviations were off-​setting, however—​some poorly informed citizens reported casting a Republican vote when they would have been predicted to vote Democratic if well informed,

52  Martin Gilens

but other poorly informed citizens “mistakenly” voted Democratic when they would have been predicted to vote Republican. The more relevant aggregate deviation between actual and well-​informed presidential votes was only three percentage points.37 In an even more directly relevant study that used a similar methodology, Scott Althaus compared respondents’ expressed preferences on 235 political opinion questions with imputed preferences calculated by assigning to each respondent the predicted preference of someone with the maximum level of political knowledge but otherwise identical in terms of education, income, age, partisan identification, race, sex, marital status, religion, region, labor force participation, occupational category, union membership, and homeownership.38 Althaus found that in the aggregate, imputed “fully informed preferences” differed from expressed preferences by an average of about 6.5  percentage points. Not a trivial amount, but hardly enough to dismiss existing preferences as an unsuitable guide to government decision-​making. Two lessons can be drawn from the research on “enlightened preferences.” First, while heuristics or informational shortcuts might, in theory, be extremely effective at allowing citizens to reach the same preferences they would if they were more fully informed, in practice a gap remains between actual and hypothetical “well-​informed” preferences, whether those preferences are statistically imputed or arrived at after exposure to new information and deliberation. Second, the size of the aggregate gap is rather modest. The two most directly relevant analyses that focus on policy preferences find gaps of 2 and 6.5 percentage points, with a 3 percentage point gap in presidential voting. Differences of this size might be enough to swing a close election or to shift aggregate preferences from slightly favorable toward some policy option to slightly opposed. But the policy proposals I analyze at the end of this chapter range widely from strong opposition to strong support (about two-​thirds of the proposed policy changes in my dataset were favored by under 40 or over 60 percent of the respondents). Thus, the relatively small differences in favorability that might be expected from a better-​informed, more “enlightened” citizenry, would be unlikely to lead to substantially different conclusions.

Question Wording and Framing Effects Even casual consumers of survey data are aware that subtle differences in how a question is worded can sometimes produce large differences in responses. Advocacy groups sometimes take advantage of this phenomenon by asking “loaded” or “biased” questions which are designed to portray public sentiment as highly favorable toward the group’s preferred policies. But many observers are skeptical that even careful and well-​crafted surveys can avoid this problem. One popular book aimed at explaining surveys and their use in American politics claims:

Citizen Competence and Democratic Governance  53

Even when the sponsor has no obvious ax to grind, question wording choices greatly influence the results obtained. In many instances, highly reputable polling organizations have arrived at divergent conclusions simply because they employed different (although well-​ constructed) questions on a particular topic.39 But just how ubiquitous and how consequential are such question wording effects? This is a difficult question to answer because there is no clear way to define the range of plausible question wordings on a given topic or the set of topics that should be considered. Some of the most frequently cited examples of question wording effects do raise doubts about the ability of survey measures to accurately capture the public’s policy preferences. For example, Tom Smith reports that 64 percent of Americans thought the government was spending too little on “assistance to the poor” but only 22 percent thought too little was being spent on “welfare.”40 Howard Schuman and Stanley Presser found that in the 1970s two in five Americans felt that the United States should “not allow” public speeches against democracy, but only half that number felt that United States should “forbid” public speeches against democracy.41 Finally, George Quattrone and Amos Tversky found that 64  percent of their respondents preferred a program that would increase inflation somewhat while reducing unemployment from 10  percent to 5  percent, but only 46  percent made the same choice when the program was described as increasing employment from 90 percent to 95 percent.42 Each of these examples reveals substantial effects from apparently minor changes in the words used to describe a policy choice and each has been replicated numerous times, so we cannot dismiss them as statistical flukes. Yet their implications for how we understand citizens’ policy preferences (or their lack of preferences) and our ability to gauge those preferences is far from clear. For example, the greater appeal of “assisting the poor” over “welfare” has often been interpreted as indicating the sensitivity of the public to particular positively or negatively loaded terms. If the preferences expressed toward the same policy can be shifted so dramatically by calling it one thing rather than another, can we even say the public has a “real” and discernable preference toward that policy? Yet this example can be viewed another way entirely. There are many different government programs that assist the poor by providing medical care, housing subsidies, legal aid, child care, job training, and so on. For some respondents, all of these programs might be included under the rubric of “welfare” but for many Americans, welfare is understood as cash assistance to the able-​bodied working-​age unemployed poor. The public tends to be strongly supportive of these other anti-​poverty programs, so the lesser appeal of “welfare” in comparison to “assisting the poor” can be understood not as a superficial response to an emotionally laden term, but a sophisticated differentiation between different sorts of government anti-​poverty programs.43

54  Martin Gilens

The broader lesson from this alternative perspective on the “welfare” question wording experiment is that much of what passes for question wording effects are in actuality differences in responses resulting from differences in the policy that respondents are asked to respond to. The same survey that showed more support for “assisting the poor” than for “welfare” also found greater support for “halting the rising crime rate” than for “law enforcement,” and greater support for “dealing with drug addiction” than for “drug rehabilitation” (General Social Survey). But these alternative question wordings are not simply different formulations of identical policies; they are references to different aspects of their respective issues. In the second example above, which contrasts “forbid” and “not allow,” the alternative wordings do appear to have identical meanings. The substantial differences in responses to these two formulations are a bit of a mystery, especially since the alternative question wordings sometimes produce dramatically different responses (like the case of “speeches against democracy” described above), sometimes modest differences (e.g., in a parallel experiment focused on “speeches in favor of communism”), and sometimes no differences at all (e.g., in questions about “showing x-​rated movies” or “cigarette advertisements on television”).44 Sometimes respondents seem to react differently to “forbid” and “not allow” but at other times these alternative wordings seem to make no difference. The third example above revealed different evaluations if a policy choice was presented in terms of its effect on the percentage of the work force that would be employed or on the percent of the work force that would be unemployed.45 These sorts of mathematically equivalent alternative descriptions are what James Druckman calls “equivalency frames.”46 They trouble some scholars of democracy because they seem to preclude the possibility of determining whether the policy in question is in fact supported or opposed by a majority of the public. But such equivalency framing effects are not a failing of democracy, but of human reasoning. In a study comparing ordinary citizens with members of the national legislature in three different countries, Lior Sheffer and his colleagues found that political elites were just as prone to such inconsistencies as the mass publics that they represented.47 While framing effects have led many scholars to doubt whether the public can plausibly be said to have preferences on the underlying policies, others point out that such framing effects in survey experiments take place under highly artificial conditions. In the real world, alternative ways of characterizing a policy choice are typically encountered not in isolation (as in typical survey experiments) but simultaneously as part of the political debate. The availability of competing frames, and the give and take of political debate have been shown to undermine framing effects, reducing or eliminating differences in responses.48 Question wording and framing effects potentially challenge the notion that the public holds meaningful preferences and that we can use survey interviews to

Citizen Competence and Democratic Governance  55

discern what those preferences are. Yet the real-world impact of these problems may be small, as two recent examples suggest. In the first example, opponents of the inheritance tax were said to have boosted their cause by relabeling it the “death tax.”49 But the best evidence suggests that the label made little difference. In a survey experiment using two alternative wordings administered to randomly selected halves of the sample, 69 percent of respondents favored doing away with the “estate tax” while 73 percent favored doing away with the “death tax.”50 In a parallel example, observers have claimed that the label “climate change” generates greater concern among the public than “global warming.”51 But the only randomized survey experiment to pit these two formulations against each other found little difference:  57  percent of Americans believed that “global warming” would become a “very serious” or “extremely serious” problem if nothing was done, compared with 60 percent who felt that way about “climate change” and 58 percent about “global climate change.”52 In sum, we cannot dismiss concerns about question wording and framing effects entirely. The evidence is strong that how a policy is described can have an impact on the level of support or opposition expressed toward that policy. These effects, however, do not imply that the public has no “real” attitudes toward these policies, or that we cannot know (at least approximately) what those attitudes are. As one expert who has himself conducted numerous studies of framing effects concludes, “… framing effects appear to be neither robust nor particularly pervasive. Elite competition and heterogeneous discussion limit and often eliminate framing effects.”53

Retrospective Evaluations Assessing incumbent officeholders’ performance (or making “retrospective evaluations”) is sometimes thought to be less demanding of citizens than forming policy preferences on issues. In his influential work on retrospective voting, Morris Fiorina argues that “In order to ascertain whether the incumbents have performed poorly or well, citizens need only calculate the changes in their own welfare.”54 But observers like Achen and Bartels argue that voters fall short of even this less demanding standard. First, citizens sometimes appear to let irrelevant events influence their support for incumbent officeholders. Achen and Bartels argue, for example, that shark attacks off the New Jersey coast in 1916 (and the resulting loss of tourism dollars) was associated with a ten percentage point drop in support for incumbent President Woodrow Wilson in those communities where the attacks occurred.55 In a broader analysis of calamities equally beyond the control of incumbent politicians, Achen and Bartels conclude that “voters throughout the 20th century punished incumbent presidents at the polls for droughts and floods ….”56 While these examples of voters’ responses to events outside of incumbents’ control have garnered considerable attention, it is not clear how frequent or

56  Martin Gilens

consistent they are. As Achen and Bartels note, events must be made politically relevant to voters, typically by challengers who claim not that incumbents caused the calamity but that they did not prepare adequately or respond to the disaster as they should have. Since many calamities are not associated with reduced support for incumbents, it is difficult to distinguish a “rational” reaction of voters to politicians’ poor response to a disaster from an “irrational” reaction to the disaster itself.57 Voters may at times have sensible reasons to hold incumbents accountable for the aftermath of natural disasters, but the same cannot be said for the outcome of sporting events. If voters punish incumbents when the local sports team loses, that would seem a sure indication that voting behavior is irrationally shaped by irrelevant events. One study found exactly this pattern with regard to local college football teams—​incumbents benefited electorally when the hometown team won in the weeks before an election and suffered electorally when they lost.58 But a subsequent study, using a different approach, found no evidence of electoral effects for either college or professional football outcomes.59 While scholars debate the evidence on voters’ responses to natural disasters or sporting events, there is a large literature documenting the electoral impact of economic conditions. Presidential elections, in particular, are strongly affected by national economic conditions in the year or so prior to the election. While few argue that economic conditions are an inappropriate criterion for voters to consider (in contrast to sports outcomes or shark attacks), scholars disagree over the rationality and effectiveness of economic-​based voting in rewarding or sanctioning incumbent officeholders. Achen and Bartels again present the most thorough and compelling evidence that voters tend to rely on inappropriate economic considerations and are ineffective and even counter-​productive in their efforts to hold incumbents accountable for economic performance. First, they point out, perceptions of the economy are subject to strong partisan biases; supporters of the incumbent party tend to see the economy significantly more positively than supporters of the out-​party. Second, they argue that voters do a poor job of distinguishing between economic trends that can reasonably be attributed to incumbent officeholders and those that are clearly beyond their control. Finally, they note that voters’ myopic focus on recent economic performance, and neglect of performance earlier in an incumbent’s term, lead to perverse incentives for officeholders and result in an essentially random process based on incumbents’ luck rather than any enduring characteristic of their economic stewardship. The first of these concerns—​partisan biases in economic perceptions—​is likely to have only a modest impact on the public’s success in holding incumbents accountable for economic outcomes. First, these biases are strongest among the most committed partisans and such voters are highly unlikely to be pulled away from their standing partisan loyalty on the basis of economic conditions.60 Second, partisan biases in economic perceptions are weakest during unusually

Citizen Competence and Democratic Governance  57

good or bad economic times and most pronounced when the economy is performing about average.61 This means that such biases will least interfere with electoral accountability exactly when economic conditions are most conducive to either retaining or removing the incumbent. Finally, some of the partisan bias in economic evaluations expressed on surveys does not in fact reflect true differences in the beliefs of Democratic and Republican respondents, but rather a kind of partisan “cheerleading.” When survey respondents are incentivized to provide their true perceptions of the economy (for example, by being paid for correct answers to questions such as whether unemployment has gone up or down), differences between the perceptions reported by Democrats and Republicans are dramatically reduced.62 A second critique of citizens’ retrospective economic voting behavior raised by scholars concerns voters’ ability to distinguish between economic outcomes that can be plausibly attributed to the incumbent officeholder and those that are beyond his or her control. For example, if rising inflation in the U.S. results from an increase in the global price of oil due to unrest in the Middle East, it may be unreasonable to blame the sitting U.S. president. But such assessments are frequently difficult to make. In this example, perhaps a different president could have tempered conflicts in the Middle East or could have boosted domestic oil production in order to reduce price increases. Studies of subnational or cross-​national electoral outcomes can help to shed light on this aspect of voters’ behavior. One study, for example, found that governors of oil-​producing states are punished electorally when declines in the world price of oil (over which they clearly have no control) hurt their state’s economy.63 Another study found that the global economic downturn starting in 2008 led voters in many countries to turn against their incumbent governments, even if their own country had fared better than average under the difficult international circumstances.64 On the other hand, studies show that voters do often distinguish between different causes of economic prosperity or adversity, holding incumbents more responsible when the relevant economic factors are more plausibly under their leaders’ control. For example, when a state’s economy improves or worsens along with the national economy, voters do not hold their governors responsible. But when state economic conditions diverge from those of the nation as a whole, governors tend to get the credit or blame.65 Similarly, the electoral consequences of domestic economic conditions are stronger in countries that are less dependent on foreign trade (and where national governments therefore have more influence over their own domestic economies).66 In short, voters appear to be neither fully competent nor wholly incompetent in assigning credit and blame for economic conditions. As in many other domains of decision-​making (like personal finance or consumer behavior), citizens make imperfect use of available information in forming political evaluations.

58  Martin Gilens

A final concern raised about the electoral consequences of economic conditions involves the time frame that voters take into account. As decades of studies have shown, presidential elections are influenced primarily by economic conditions in the election year while earlier years in a president’s term have little or no impact. If election-year conditions were a good approximation for economic performance during an entire term in office, or a good predictor of future economic performance for incumbents who were reelected, this might not be problematic. But as Achen and Bartels demonstrate, neither of these holds true.67 Consequently, voters’ focus on recent economic trends rewards incumbents who are lucky to preside over good election-​year conditions even if the overall economic performance during their term was poor. This creates perverse incentives for elected officials to “goose” the economy during election years, even if such manipulations reduce economic prosperity in the long run. Achen and Bartels conclude that voters’ myopic focus on recent economic conditions wholly undermines economic accountability such that voters are essentially flipping a coin to decide whether the incumbent party deserves to remain in office or not. In a clever series of studies, Andrew Healy and Gabriel Lenz explore the reasons voters focus on recent economic conditions.68 They conclude that voters would prefer to base their evaluations on the incumbent’s performance across their entire term in office. But that information is either not as readily available, or requires an effortful and sometimes complex integration of information about economic conditions at different periods during an incumbent’s term. Healy and Lenz show that when information about average or cumulative economic performance is provided to voters (in hypothetical elections) they weigh each year in office roughly equally.69 The authors conclude that “… relatively simple changes in the information context may enable voters to hold their leaders accountable more effectively.”70 In sum, citizens are clearly imperfect at evaluating the performance of incumbent officeholders. Events outside incumbents’ control sometimes shape voters’ choices, although the extent to which this distorts democratic accountability remains a subject of scholarly debate. Consistent evidence shows that economic evaluations are subject to a strong “recency bias” which undermines the effectiveness of elections in holding incumbents accountable for economic performance. While more research on the causes and potential cures for such shortcomings is clearly called for, experimental studies suggest that the kinds of information available to citizens can have an important impact on the extent to which retrospective evaluations work for or against citizens’ interests.

Public Preferences and Government Policymaking If the policy preferences expressed by the mass public are meaningful (at least in the aggregate) and reflective of Americans’ genuine attitudes toward alternative government actions, then one criterion for assessing the degree of democratic

Citizen Competence and Democratic Governance  59

legitimacy is the strength of the association between public preferences and policy outcomes. Of course we would not expect or desire a perfect match between majority preferences and government policy. First, there are issues of minority rights to be considered. Second, the public is not capable of guiding policy on all questions that come before the government. Issues like alternative high-​definition TV standards, or which government regulatory agency should be responsible for agricultural futures trading, are simply too obscure for most citizens to have meaningful preferences on. Finally, one subset of the public might care intensely about a particular issue about which others are indifferent. If I  care deeply about foreign policy and am indifferent to education, and you care strongly about education and not foreign policy, a government that responds to my preferences on foreign policy and yours on education would make us both happier than one that took each of our views equally into consideration in both issue domains. These considerations notwithstanding, it remains true that a government that responds only weakly or not at all to the preferences of the governed, or that systematically responds to some citizens but ignores others, has but a weak claim to being considered a democracy. The association of government policy with public preferences measured by surveys is only one basis for judging government responsiveness, but it is a useful starting point for assessing the nature and degree of representation. To estimate the association between public preferences and government policy I make use of the dataset mentioned briefly above. These data consist of 2,245 survey questions asked between 1964 and 2006.71 Each question asked whether respondents favored or opposed some specific change in federal government policy. In my dataset, I  collected the responses to these questions separately for respondents at different income levels in order to compare the strength of the preference/​policy link for more and less well-​off Americans. As we would expect, the more support a given policy has among the public, the more likely it is that that policy will be adopted, and this pattern holds true for respondents at all income levels. Figure 2.1 shows this relationship separately for respondents at the 10th, 50th, and 90th income percentiles.72 The far left side of the figure shows that policies with strong public opposition (with fewer than 20 percent favoring the proposed change) have a low probability of being adopted, with the probability of adoption increasing as support increases. However, the far right side of the figure shows that even policies with strong public support (at least 80 percent of the public favoring the proposed change) have a less than even chance of being adopted. This pattern suggests that the political system is responsive to public preferences, but with a strong status quo bias. Given that our federal government was designed by its framers to inhibit as much as facilitate lawmaking (with its separation of powers, multiple veto points within congress, supermajority requirements in the Senate, and so on), this status quo bias should not be surprising.

60  Martin Gilens

Percent of proposed changes adopted

Figure 2.1 also shows that the probability of a policy being adopted is somewhat more strongly related to the preferences of the affluent than those of the middle class or the poor: the solid line, representing respondents at the 90th income percentile, is somewhat steeper than the lines for the 50th and 10th income percentiles. But the differences among income groups are modest, and at every level of income, favored policies are substantially more likely to be adopted than unfavored policies. To better gauge the true influence over policymaking of Americans at different income levels, we need to take into account the fact that poor and well-​off Americans agree on many policy questions. If affluent Americans are better able to influence policy outcomes than the less well off, the association of policy outcomes with the preferences of the poor or the middle class shown in Figure 2.1 may simply reflect those proposed changes on which Americans at all income levels agree. Figure 2.2 shows the same relationships shown in Figure 2.1, but restricted to proposed policy changes for which low-​and high-​income Americans’ preferences, or middle-​and high-​income Americans’ preferences, diverge by at least 10 percentage points. Here we see a very different picture: the preference/​policy link for the affluent remains strong, but when the preferences of less well-​off Americans diverge from those at the top of the income distribution, the preferences of the less well off appear to have virtually no relationship with policy outcomes. In other publications I explore these relationships in greater detail. I find that the basic pattern shown in Figure 2.2 is similar with regard to foreign policy, economic policy, social welfare, and moral or religious issues and cannot be explained by the higher levels of voting among 60% 10th percentile 50%

50th percentile

40%

90th percentile

30%

Linear (10th percentile)

20%

Linear (50th percentile)

10%

Linear (90th percentile)

0% 60% 80% 100% 0% 20% 40% Percent favoring proposed policy change

Figure 2.1 The Preference/​Policy Link for Respondents at the 10th, 50th, and 90th Income Percentiles

Percent of proposed changes adopted

Citizen Competence and Democratic Governance  61

60% 50% 40% 10th percentile 30%

90th percentile

20%

Linear (10th percentile)

10%

Linear (90th percentile)

0%

Percent of proposed changes adopted

60% 80% 100% 0% 20% 40% Percent favoring proposed policy change

60% 50% 40%

50th percentile

30%

90th percentile

20%

Linear (50th percentile)

10%

Linear (90th percentile)

0% 0% 20% 40% 60% 80% 100% Percent favoring proposed policy change

Figure 2.2 The Preference/​Policy Link when Preferences at the 10th or 50th Income Percentiles Diverge from the 90th Income Percentile

the affluent, or a lack of strong preferences among the middle class or the poor.73 A number of different factors may contribute to the influence over policy exerted by affluent Americans and the lack of influence among the less well off, but they all relate directly or indirectly to the importance of money in the political system. The well-off contribute to parties, candidates, and interest organizations at far higher rates than the middle class or the poor. In addition,

62  Martin Gilens

they tend to share the policy preferences of an even smaller and more powerful group of truly rich Americans who help determine who runs for and wins public office (and therefore what sorts of policies they subsequently pursue). Finally, government policymakers are themselves far better off economically than the average American, and their own policy preferences are more likely to reflect those of their economic peers than those of less well-​off citizens.

Conclusions The American public’s knowledge of political issues and understanding of the policy choices government faces is clearly limited. And studies of framing and question wording show the potential difficulties of measuring public preferences. Yet on balance, the evidence indicates that framing and question wording effects in the real world are infrequent, and that aggregate preferences as measured by surveys reflect much the same set of attitudes that a more fully informed and engaged public would express. This evidence, I suggest, is sufficient to conclude that survey measures of public opinion are sensible bases on which to judge the extent to which government policy reflects the preferences of the governed. Based on the findings described briefly above, American democracy is found wanting. The problem lies not in the failure of the public to form meaningful policy preferences, but in the failure of policymakers to respond to the public. Affluent Americans appear to have substantial influence over policy outcomes, but when less well-​off Americans’ preferences diverge from those of the affluent, government policymakers appear to take into account only the desires of the most economically advantaged. In every society the well off have more influence over government than the less economically advantaged. But the degree of inequality in how government responds to its citizens is a fundamental gauge of how truly democratic a society is. In this regard, we have a long way to go before we can claim to be a democracy characterized by “… the continuing responsiveness of the government to the preferences of its citizens, considered as political equals.”74

Notes 1 Robert Alan Dahl, Polyarchy:  Participation and Opposition (New Haven, CT:  Yale University Press, 1973), p.1. 2 “The Nature of Belief Systems in Mass Publics,” in Ideology and Discontent, ed. David E. Apter (New York: Free Press, 1964). 3 Ibid., p.245. 4 Christopher H. Achen and Larry M. Bartels, Democracy for Realists: Why Elections Do Not Produce Responsive Government (Princeton, NJ:  Princeton University Press, 2017). 5 Michael X. Delli Carpini and Scott Keeter, What Americans Know About Politics and Why It Matters (New Haven, CT: Yale University Press, 1996), p.270.

Citizen Competence and Democratic Governance  63 6 Length of U.S. Senate term:  2016 American National Election Study; Name own U.S. Senators/​how many U.S. Senators does each state have: Edward M. Kennedy Institute for the U.S. Senate, survey of 1,025 adults conducted by landline and cell phone by February 24–​28, 2016, www.emkinstitute.org/​resources/​national​civics-​survey-​results-​2016. 7 www.pewresearch.org/​ f act-​ t ank/​ 2 016/​ 1 1/​ 1 6/​ v oters-​ p erceptions-​ o f-​ c rime-​ continue-​to-​conflict-​with-​reality/​ accessed July 20, 2018. 8 https://​kaiserfamilyfoundation.files.wordpress.com/​2016/​01/​8825-​t-​americans-​ views-​on-​the-​us-​role-​in-​global-​health.pdf accessed July 20, 2018. 9 Arthur Lupia, Uninformed:  Why People Know So Little About Politics and What We Can Do About It (New York: Oxford University Press, 2016). Doris A. Graber, “Why Voters Fail Information Tests:  Can the Hurdles Be Overcome?” Political Communication 11 (1994). 10 Bernard R. Berelson, Paul F. Lazarsfeld, and William N. McPhee, Voting: A Study of Opinion Formation in a Presidential Campaign (Chicago, IL: University of Chicago Press, 1954), p.109. 11 Anthony Downs, An Economic Theory of Democracy (New  York:  Harper Collins, 1957), p.233. 12 Samuel L. Popkin, The Reasoning Voter: Communication and Persuasion in Presidential Campaigns (Chicago, IL: University of Chicago Press, 1991); Paul M. Sniderman, Richard A.  Brody, and Philip E.  Tetlock, Reasoning and Choice:  Explorations in Political Psychology (Cambridge: Cambridge University Press, 1991); Arthur Lupia, “Shortcuts Versus Encyclopedias:  Information and Voting-​Behavior in California Insurance Reform Elections,” American Political Science Review 88, no.  1 (1994); Arthur Lupia and Mathew D. McCubbins, The Democratic Dilemma: Can Citizens Learn What They Need to Know? (Cambridge: Cambridge University Press, 1998); Martin Gilens and Naomi Murakawa, “Elite Cues and Political Decision-​Making,” in Research in Micropolitics, ed. Michael X.  Delli Carpini, Leonie Huddy, and Roberty Y. Shapiro (Oxford: Elsevier, 2002); Richard R. Lau and David P. Redlawsk, “Voting Correctly,” American Political Science Review 91, no. 3 (1997); How Voters Decide:  Information Processing During Election Campaigns, Cambridge Studies in Public Opinion and Political Psychology (Cambridge; New  York:  Cambridge University Press, 2006); John R.  Zaller, The Nature and Origins of Mass Opinion (Cambridge: Cambridge University Press, 1992); Edward G. Carmines and James H.  Kuklinski, “Incentives, Opportunities, and the Logic of Public Opinion in American Political Representation,” in Information and Democratic Processes, ed. John A.  Ferejohn and James H.  Kuklinski (Urbana, IL:  University of Illinois Press, 1990). 13 Arthur Lupia, “Who Can Persuade? A Formal Theory, a Survey and Implications for Democracy,” Paper (1995). 14 For example, John W. Kingdon, Congressmen’s Voting Decisions, 3rd ed. (Ann Arbor, MI: University of Michigan Press, 1989). 15 Gabriel S. Lenz, Follow the Leader?: How Voters Respond to Politicians’ Policies and Performance (Chicago, IL: University of Chicago Press, 2012). 16 Achen and Bartels show that “cross-​pressured” men were more likely to shift their abortion views while women (who they speculate cared more about abortion policy) were more likely to shift their partisanship. Christopher H.  Achen and Larry M. Bartels, Democracy for Realists: Why Elections Do Not Produce Responsive

64  Martin Gilens Government: With a New Afterword by the Authors, Princeton Studies in Political Behavior (Princeton, NJ: Princeton University Press, 2017), p.264. 17 D. Sunshine Hillygus and Todd G.  Shields, The Persuadable Voter:  Wedge Issues in Presidential Campaigns (Princeton, NJ:  Princeton University Press, 2008). See also John Bullock, 2011 APSR, “Elite Influence on Public Opinion in an Informed Electorate.” 18 Converse assessed issue public membership by dint of the consistency of responses to the same questions over time. 19 Converse, “The Nature of Belief Systems in Mass Publics,” p.246. 20 Shanto Iyengar et al., “Selective Exposure to Campaign Communication: The Role of Anticipated Agreement and Issue Public Membership,” Journal of Politics 70, no. 1 (2008); Amy R. Gershkoff, “How Issue Interest Can Rescue the American Public,” (Doctoral dissertation, Department of Politics, Princeton University, 2006); Jon A. Krosnick, “Government Policy and Citizen Passion: A Study of Issue Publics in Contemporary America,” Political Behavior 12, no. 1 (1990). 21 “Government Policy and Citizen Passion: A Study of Issue Publics in Contemporary America.” 22 Public Opinion and Democratic Accountability:  How Citizens Learn About Politics (Princeton, NJ: Princeton University Press, 2003). 23 Nicholas R.  Miller, “Information, Electorates, and Democracy:  Some Extensions and Interpretations of the Condorcet Jury Theorem,” in Information Pooling and Group Decision Making, ed. Bernard Grofman and Guillermo Owen (Greenwich, CT: JAI, 1986); Benjamin I. Page and Robert Y. Shapiro, The Rational Public: Fifty Years of Trends in Americans’ Policy Preferences (Chicago, IL: University of Chicago Press, 1992); Philip E. Converse, “Popular Representation and the Distribution of Information,” in Information and Democratic Processes, ed. John A.  Ferejohn and James H. Kuklinksi (1990); Christopher H. Achen, “Mass Political Attitudes and the Survey Response,” American Political Science Review 69, no. 4 (1975). 24 Zaller, The Nature and Origins of Mass Opinion; John Zaller and Stanley Feldman, “A Simple Theory of the Survey Response: Answering Questions Versus Revealing Preferences,” American Journal of Political Science 36, no. 3 (1992). 25 Larry M. Bartels, “Democracy with Attitudes” (New York, May 9 1998). 26 Page and Shapiro, The Rational Public:  Fifty Years of Trends in Americans’ Policy Preferences. 27 Ibid. 28 Ibid., p.14. 29 Ibid., p.14. 30 Martin Gilens, “Political Ignorance and Collective Policy Preferences,” American Political Science Review 95, no. 2 (2001); James H. Kuklinski et al., “Misinformation and the Currency of Democratic Citizenship,” Journal of Politics 62, no. 3 (2000); Martin Gilens, Why Americans Hate Welfare:  Race, Media, and the Politics of Antipoverty Policy (Chicago, IL: University of Chicago Press, 1999). 31 The Rational Public: Fifty Years of Trends in Americans’ Policy Preferences, p.381. 32 Lupia, “Shortcuts Versus Encyclopedias:  Information and Voting Behavior in California Insurance Reform Elections.” Achen and Bartels (2017, p.80) argue that Lupia’s “better informed voters” were in fact misguided in supporting an initiative that was “superficially appealing but shortsighted.” But this critique is itself open

Citizen Competence and Democratic Governance  65 to question. See www.consumerwatchdog.org/​prop-​103-​california-​insurance-​ reform (accessed July 31, 2018). 33 Robert C.  Luskin and James S.  Fishkin, “Deliberative Polling, Public Opinion, and Democracy:  The Case of the National Issues Convention” (paper presented at the American Political Science Association, Boston, 1998); James S. Fishkin and Robert C. Luskin, “Bringing Deliberation to the Democratic Dialogue,” in A Poll with a Human Face, ed. Max McCombs and Amy Reynolds (1999); “Experimenting with a Democratic Ideal: Deliberative Polling and Public Opinion,” Acta Politica 40, no. 3 (2005). 34 See Gilens and Murakawa, “Elite Cues and Political Decision-​making” for an analysis of these results. 35 Philip E. Converse, “Assessing the Capacity of Mass Electorates,” Annual Review of Political Science 3 (2000). 36 Larry M. Bartels, “Uninformed Votes: Information Effects in Presidential Elections,” American Journal of Political Science 40, no. 1 (1996). 37 Ibid., table 3. 38 Scott L. Althaus, Collective Preferences in Democratic Politics: Opinion Surveys and the Will of the People (New York: Cambridge University Press, 2003). 39 Herbert B. Asher, Polling and the Public: What Every Citizen Should Know, 7th ed. (Washington, DC: CQ Press, 2007), p.56. 40 “That Which We Call Welfare by Any Other Name Would Smell Sweeter:  An Analysis of the Impact of Question Wording on Response Patterns,” Public Opinion Quarterly 51, no. 1 (1987). 41 Howard Schuman and Stanley Presser, Questions and Answers in Attitude Surveys (San Diego: Academic Press, 1981), p.277. 42 “Contrasting Rational and Psychological Analyses of Political Choice,” American Political Science Review 82, no. 3 (1988). 43 Gilens, Why Americans Hate Welfare: Race, Media, and the Politics of Antipoverty Policy. 44 Schuman and Presser, Questions and Answers in Attitude Surveys, pp.281–​3. 45 Quattrone and Tversky, “Contrasting Rational and Psychological Analyses of Political Choice,” p.728. This example is explained by recognizing that people tend to evaluate differences in magnitude (like the employment or unemployment rates) at least partly in terms of ratios. The difference between 10 percent and 5 percent unemployment appears large because the former is twice as big as the latter. In contrast, the difference between 90 percent employment and 95 percent employment appears small because their ratio is close to one. 46 James N. Druckman, “The Implications of Framing Effects for Citizen Competence,” Political Behavior 23, no. 3 (2001). 47 Lior Sheffer et al., “Nonrepresentative Representatives: An Experimental Study of the Decision Making of Elected Politicians,” American Political Science Review 112, no. 2 (2018). 48 James N. Druckman, “Political Preference Formation: Competition, Deliberation, and the (Ir)Relevance of Framing Effects,” American Political Science Review 98, no. 4 (2004); Paul M. Sniderman and Sean M. Theriault, “The Structure of Political Argument and the Logic of Issue Framing” (1999). 49 Michael J. Graetz and Ian Shapiro, Death by a Thousand Cuts: The Fight over Taxing Inherited Wealth, 1st ed. (Princeton, NJ: Princeton University Press, 2005).

66  Martin Gilens 50 Larry M.  Bartels, Unequal Democracy:  The Political Economy of the New Gilded Age (Princeton, NJ: Princeton University Press, 2008). Whether Americans would still object to the estate tax if they knew how few estates were actually subject to the tax is uncertain. Bartels, Unequal Democracy, and Joel Slemrod “The Role of Misconceptions in Support for Regressive Tax Reform,” National Tax Journal 59, no. 1 (2006) offer evidence that misinformation about the breadth of the estate tax’s impact contributes only modestly to public opposition. In contrast, Ilyan Kuziemko, Michael I.  Norton, Emmanuel Saez, and Stefanie Stantcheva, “How Elastic are Preferences for Redistribution? Evidence from Randomized Survey Experiments,” NBER Working Paper no. 18865 (2013), find a substantial impact. 51 Katy Butler, “George Lakoff Says Environmentalists Need to Watch Their Language,” Sierra Magazine 89, July/​August (2004). 52 Ana Villar and Jon A.  Krosnick, “Global Warming vs. Climate Change, Taxes vs. Prices: Does Word Choice Matter?” Climate Change 105, no. 1 (2010). 53 Druckman, “Political Preference Formation:  Competition, Deliberation, and the (Ir)Relevance of Framing Effects,” p.683. 54 Morris P. Fiorina, Retrospective Voting in American National Elections (New Haven, CT: Yale University Press, 1981), p.5. 55 Achen and Bartels, Democracy for Realists: Why Elections Do Not Produce Responsive Government: With a New Afterword by the Authors, p.127. 56 Ibid. p.143. 57 A debate has emerged over the robustness of the data and statistical analyses supporting these findings on the political impacts of shark attacks and natural disasters. Studies supporting such impacts include Achen and Bartels (2017) and Achen and Bartels (2018). Studies arguing either that these events are unrelated to voting behavior or that any such association is based on voters’ reasonable responses to the adequacy of government responses include Healy and Malhotra (2009), Healy and Malhotra (2010), and Fowler and Hall (2018). 58 Andrew J.  Healy, Neil Malhotra, and Cecilia Hyunjung Mo, “Irrelevant Events Affect Voters’ Evaluations of Government Performance,” Proceedings of the National Academy of Sciences 107, no. 29 (2010). 59 Fowler and Montagnes (2015). See also Healy, Malhotra, and Mo (2015) in response. 60 Larry M. Bartels, “Beyond the Running Tally: Partisan Bias in Political Perceptions,” Political Behavior (2002). 61 Andrew Healy and Neil Malhotra, “Retrospective Voting Reconsidered,” in Annual Review of Political Science (2013). 62 Markus Prior, Gaurav Sood, and Kabir Khanna, “You Cannot Be Serious: The Impact of Accuracy Incentives on Partisan Bias in Reports of Economic Perceptions,” Quarterly Journal of Political Science 10, no. 4 (2015). John G. Bullock et al., “Partisan Bias in Factual Beliefs About Politics,” ibid. 63 Justin Wolfers, “Are Voters Rational? Evidence from Gubernatorial Elections,” (Stanford: GSB, 2002). 64 Larry M.  Bartels, “Ideology and Retrospection in Electoral Responses to the Great Recession,” in Mass Politics in Tough Times: Opinions, Votes, and Protest in the Great Recession, ed. Nancy Bermeo and Larry M.  Bartels (New  York: Oxford, 2014). 65 Wolfers, “Are Voters Rational? Evidence from Gubernatorial Elections.”

Citizen Competence and Democratic Governance  67 66 Raymond M. Duch and Randolph T. Stevenson, The Economic Vote: How Political and Economic Institutions Condition Election Results, Political Economy of Institutions and Decisions (New York: Cambridge University Press, 2008). 67 Achen and Bartels (2017) ch. 6. 68 Healy and Lenz (2013). 69 In a similar set of experiments, Huber, Hill, and Lenz (2012) found that participants who were told at the outset of the experiment that they would be making a decision to “reelect” or “replace” the incumbent did not put extra emphasis on economic payouts during the most recent time periods, unlike those who were not made aware in advance of their opportunity to replace the incumbent. The authors suggest that many voters may not focus on a distant election and by the time they are paying attention to their electoral choice, their recollections of the previous years’ economic conditions may be vague or lacking in salience. In a separate experiment, they found that a very large random payout unrelated to the incumbent’s performance resulted in higher “reelection rates” consistent with a “good mood” rather than a rational reward/​punishment account of economic voting. 70 Healy and Lenz (2013), p.45. 71 For reasons explained in Martin Gilens, “Paying the Piper:  Economic Inequality and Democratic Responsiveness in the United States,” (Unpublished book manuscript, Princeton University, n.d.), only survey questions asked during 1964–​68, 1981–​2002, and 2005–​06 were included. 72 Figures  2.1 and 2.2 show the mean proportion adopted for policies favored by less than 20  percent, 20–​30, 30–​40, 40–​50, 50–​60, 60–​70, 70–​80, and 80  percent of respondents or more (for respondents at the 10th, 50th, and 90th income percentiles). See “Inequality and Democratic Responsiveness,” Public Opinion Quarterly 69, no. 5 (2005) for an explanation of how preferences at different income percentiles are calculated. 73 Ibid.; “Preference Gaps and Inequality in Representation,” PS: Political Science and Politics 42, no. 2 (2009); “Policy Outcomes and Representational Inequality,” in Who Gets Represented?, ed. Peter K. Enns and Christopher Wlezien (New York: Russell Sage Foundation, 2011); “Paying the Piper: Economic Inequality and Democratic Responsiveness in the United States.” 74 Dahl, Polyarchy: Participation and Opposition, p.1.

Bibliography Achen, Christopher H. “Mass Political Attitudes and the Survey Response.” [In English]. American Political Science Review 69, no. 4 (1975): 1218–​31. Achen, Christopher H., and Larry M. Bartels. Democracy for Realists: Why Elections Do Not Produce Responsive Government: With a New Afterword by the Authors. Princeton Studies in Political Behavior. Princeton, NJ: Princeton University Press, 2017. Achen, Christopher H., and Larry M. Bartels. “Statistics As If Politics Mattered: A Reply to Fowler and Hall.” The Journal of Politics  80, no. 4 (2018): 1438–​53. Althaus, Scott L. Collective Preferences in Democratic Politics: Opinion Surveys and the Will of the People. New York: Cambridge University Press, 2003. Asher, Herbert B. Polling and the Public:  What Every Citizen Should Know. 7th ed. Washington, DC: CQ Press, 2007.

68  Martin Gilens Bartels, Larry M. “Uninformed Votes:  Information Effects in Presidential Elections.” American Journal of Political Science 40, no. 1 (February 1996): 194–​230. ———. “Democracy with Attitudes.” New York, May 9, 1998. ———. “Beyond the Running Tally:  Partisan Bias in Political Perceptions.” Political Behavior (2002). ———. Unequal Democracy:  The Political Economy of the New Gilded Age. Princeton, NJ: Princeton University Press, 2008. ———. “Ideology and Retrospection in Electoral Responses to the Great Recession.” In Mass Politics in Tough Times: Opinions, Votes, and Protest in the Great Recession, edited by Nancy Bermeo and Larry M. Bartels, 185–​223. New York: Oxford, 2014. Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. Voting:  A Study of Opinion Formation in a Presidential Campaign. Chicago, IL:  University of Chicago Press, 1954. Bullock, John G. “Elite Influence on Public Opinion in an Informed Electorate.” American Political Science Review no. 105 (2011): 496–​515. Bullock, John G., Alan S. Gerber, Seth J. Hill, and Gregory A. Huber. “Partisan Bias in Factual Beliefs About Politics.” Quarterly Journal of Political Science 10, no. 4 (2015): 519–​78. Butler, Katy. “George Lakoff Says Environmentalists Need to Watch Their Language.” Sierra Magazine 89, July/​August (2004): 54–​6. Carmines, Edward G., and James H. Kuklinski. “Incentives, Opportunities, and the Logic of Public Opinion in American Political Representation.” In Information and Democratic Processes, edited by John A. Ferejohn and James H. Kuklinski, 240–​68. Urbana, IL: University of Illinois Press, 1990. Converse, Philip E. The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, edited by David E. Apter. New York: Free Press, 1964. ———. “Popular Representation and the Distribution of Information.” In Information and Democratic Processes, edited by John A. Ferejohn and James H. Kuklinksi, 1990. ———. “Assessing the Capacity of Mass Electorates.” Annual Review of Political Science 3 (2000): 331–​53. Dahl, Robert A. Polyarchy: Participation and Opposition. New Haven, CT: Yale University Press, 1971. Delli Carpini, Michael X., and Scott Keeter. What Americans Know About Politics and Why It Matters. New Haven, CT: Yale University Press, 1996. Downs, Anthony. An Economic Theory of Democracy. New York: Harper Collins, 1957. Druckman, James N. “The Implications of Framing Effects for Citizen Competence.” [In English]. Political Behavior 23, no. 3 (Sep 2001): 225–​56. ———. “Political Preference Formation: Competition, Deliberation, and the (Ir)Relevance of Framing Effects.” [In English]. American Political Science Review 98, no. 4 (Nov 2004): 671–​86. Duch, Raymond M., and Randolph T. Stevenson. The Economic Vote: How Political and Economic Institutions Condition Election Results. Political Economy of Institutions and Decisions. New York: Cambridge University Press, 2008. Fiorina, Morris P. Retrospective Voting in American National Elections. New Haven, CT: Yale University Press, 1981. Fishkin, James S., and Robert C. Luskin. “Bringing Deliberation to the Democratic Dialogue.” In A Poll with a Human Face, edited by Max McCombs and Amy Reynolds, 1999.

Citizen Competence and Democratic Governance  69 ———. “Experimenting with a Democratic Ideal:  Deliberative Polling and Public Opinion.” [In English]. Acta Politica 40, no. 3 (Sep 2005): 284–​98. Fowler, Anthony, and Andrew B. Hall. “Do Shark Attacks Influence Presidential Elections? Reassessing a Prominent Finding on Voter Competence.” The Journal of Politics 80, no. 4 (2018): 1423–​37. Fowler, Anthony, and B. Pablo Montagnes. “College Football, Elections, and False-​ Positive Results in Observational Research.” Proceedings of the National Academy of Sciences 112, no. 45 (2015): 13800–​4. Gershkoff, Amy R. “How Issue Interest Can Rescue the American Public.” Doctoral dissertation, Department of Politics, Princeton University, 2006. Gilens, Martin. “Paying the Piper: Economic Inequality and Democratic Responsiveness in the United States.” Unpublished book manuscript, Princeton University, n.d. ———. Why Americans Hate Welfare: Race, Media, and the Politics of Antipoverty Policy. Chicago, IL: University of Chicago Press, 1999. ———. “Political Ignorance and Collective Policy Preferences.” American Political Science Review 95, no. 2 (2001): 379–​96. ———​ . “Inequality and Democratic Responsiveness.” [In English]. Public Opinion Quarterly 69, no. 5 (2005): 778–​96. ———​. “Preference Gaps and Inequality in Representation.” [In English]. PS: Political Science and Politics 42, no. 2 (April 2009): 335–​41. ———​. “Policy Outcomes and Representational Inequality.” In Who Gets Represented?, edited by Peter K. Enns and Christopher Wlezien. New  York:  Russell Sage Foundation, 2011. Gilens, Martin, and Naomi Murakawa. “Elite Cues and Political Decision-​Making.” In Research in Micropolitics, edited by Michael X. Delli Carpini, Leonie Huddy and Robert Y. Shapiro, 15–​50. Oxford: Elsevier, 2002. Graber, Doris A. “Why Voters Fail Information Tests: Can the Hurdles Be Overcome?” Political Communication 11 (1994): 331–​46. Graetz, Michael J., and Ian Shapiro. Death by a Thousand Cuts: The Fight over Taxing Inherited Wealth. 1st ed. Princeton, NJ: Princeton University Press, 2005. Healy, Andrew, and Neil Malhotra. “Myopic Voters and Natural Disaster Policy.” American Political Science Review 103, no. 3 (2009): 387–​406. Healy, Andrew, and Neil Malhotra. “Retrospective Voting Reconsidered.” In Annual Review of Political Science, 2013. Healy, Adam J., Philip A. Ash, Oliver Lenz, and Kylie A. Vincent. “Attenuated Total Reflectance Infrared Spectroelectrochemistry at a Carbon Particle Electrode; Unmediated Redox Control of a [NiFe]-​Hydrogenase Solution.” Physical Chemistry Chemical Physics 15, no. 19 (2013): 7055–​9. Healy, Andrew J., Neil Malhotra, and Cecilia Hyunjung Mo. “Irrelevant Events Affect Voters’ Evaluations of Government Performance.” Proceedings of the National Academy of Sciences 107, no. 29 (2010): 12804–​9. Healy, A. J., Neil Malhotra, and Cecilia Hyunjung Mo. “Reply to Fowler and Montagnes (2015): Discussion of auxiliary tests.” Available at dx.doi.org/​10.7910/​DVN/​FKEZH6. Accessed October 26 (2015): 2015. Hillygus, D. Sunshine, and Todd G. Shields. The Persuadable Voter:  Wedge Issues in Presidential Campaigns. Princeton, NJ: Princeton University Press, 2008. Huber, Gregory A., Seth J. Hill, and Gabriel S. Lenz. “Sources of bias in retrospective decision making:  Experimental evidence on voters’ limitations in controlling incumbents.” American Political Science Review 106, no. 4 (2012): 720–​41.

70  Martin Gilens Hutchings, Vincent L. Public Opinion and Democratic Accountability:  How Citizens Learn About Politics. Princeton, NJ: Princeton University Press, 2003. Iyengar, Shanto, Kyu S. Hahn, Jon A. Krosnick, and John Walker. “Selective Exposure to Campaign Communication: The Role of Anticipated Agreement and Issue Public Membership.” Journal of Politics 70, no. 1 (Jan 2008): 186–​200. Kingdon, John W. Congressmen’s Voting Decisions. 3rd ed. Ann Arbor, MI: University of Michigan Press, 1989. Krosnick, Jon A. “Government Policy and Citizen Passion: A Study of Issue Publics in Contemporary America.” Political Behavior 12, no. 1 (1990): 59–​92. Kuklinski, James H., Paul J. Quirk, Jennifer Jerit, David Schwieder, and Robert F. Rich. “Misinformation and the Currency of Democratic Citizenship.” [In English]. Journal of Politics 62, no. 3 (August 2000): 790–​816. Ilyan Kuziemko, Michael I. Norton, Emmanuel Saez, and Stefanie Stantcheva, “How Elastic Are Preferences for Redistribution? Evidence from Randomized Survey Experiments.” NBER Working Paper no. 18865 (2013). Lau, Richard R., and David P. Redlawsk. “Voting Correctly.” American Political Science Review 91, no. 3 (1997): 585–​98. ———​. How Voters Decide:  Information Processing During Election Campaigns. Cambridge Studies in Public Opinion and Political Psychology. Cambridge; New York: Cambridge University Press, 2006. Lenz, Gabriel S. Follow the Leader? How Voters Respond to Politicians’ Policies and Performance. Chicago, IL: University of Chicago Press, 2012. Lupia, Arthur. “Shortcuts Versus Encyclopedias:  Information and Voting Behavior in California Insurance Reform Elections.” American Political Science Review 88, no. 1 (March 1994): 63–​76. ———​ . “Who Can Persuade? A  Formal Theory, a Survey and Implications for Democracy.” 1995. ———​. Uninformed:  Why People Know So Little About Politics and What We Can Do About It. New York: Oxford University Press, 2016. Lupia, Arthur, and Mathew D. McCubbins. The Democratic Dilemma:  Can Citizens Learn What They Need to Know? Cambridge: Cambridge University Press, 1998. Luskin, Robert C., and James S. Fishkin. “Deliberative Polling, Public Opinion, and Democracy:  The Case of the National Issues Convention.” Paper presented at the American Political Science Association, Boston, 1998. Miller, Nicholas R. “Information, Electorates, and Democracy:  Some Extensions and Interpretations of the Condorcet Jury Theorem.” In Information Pooling and Group Decision Making, edited by Bernard Grofman and Guillermo Owen. Greenwich, CT: JAI, 1986. Page, Benjamin I., and Robert Y. Shapiro. The Rational Public: Fifty Years of Trends in Americans’ Policy Preferences. Chicago, IL: University of Chicago Press, 1992. Popkin, Samuel L. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago, IL: University of Chicago Press, 1991. Prior, Markus, Gaurav Sood, and Kabir Khanna. “You Cannot Be Serious: The Impact of Accuracy Incentives on Partisan Bias in Reports of Economic Perceptions.” Quarterly Journal of Political Science 10, no. 4 (2015): 489–​518. Quattrone, George A., and Amos Tversky. “Contrasting Rational and Psychological Analyses of Political Choice.” [In English]. American Political Science Review 82, no. 3 (September 1988): 719–​36.

Citizen Competence and Democratic Governance  71 Schuman, Howard, and Stanley Presser. Questions and Answers in Attitude Surveys. San Diego, CA: Academic Press, 1981. Sheffer, Lior, Peter John Loewen, Stuart Soroka, Stefaan Walgrave, and Tamir Sheafer. “Nonrepresentative Representatives: An Experimental Study of the Decision Making of Elected Politicians.” American Political Science Review 112, no. 2 (2018): 302–​21. Joel Slemrod “The Role of Misconceptions in Support for Regressive Tax Reform.” National Tax Journal 59, no. 1 (2006). Smith, Tom W. “That Which We Call Welfare by Any Other Name Would Smell Sweeter:  An Analysis of the Impact of Question Wording on Response Patterns.” Public Opinion Quarterly 51, no. 1 (1987): 75–​83. Sniderman, Paul M., and Sean M. Theriault. “The Structure of Political Argument and the Logic of Issue Framing.” Unpublished manuscript, Stanford University, 1999. Sniderman, Paul M., Richard A. Brody, and Philip E. Tetlock. Reasoning and Choice:  Explorations in Political Psychology. Cambridge:  Cambridge University Press, 1991. Villar, Ana, and Jon A. Krosnick. “Global Warming Vs. Climate Change, Taxes Vs. Prices: Does Word Choice Matter?” Climate Change 105, no. 1 (August 19 2010): 1–​12. Wolfers, Justin. “Are Voters Rational? Evidence from Gubernatorial Elections.” Stanford, CA: GSB, 2002. Zaller, John R. The Nature and Origins of Mass Opinion. Cambridge:  Cambridge University Press, 1992. Zaller, John, and Stanley Feldman. “A Simple Theory of the Survey Response: Answering Questions Versus Revealing Preferences.” American Journal of Political Science 36, no. 3 (1992): 579–​616.

Part II

Foundations of Political Preferences

Chapter 3

Ideology and Public Opinion Christopher M. Federico

Few concepts in the study of public opinion have attracted as much attention as that of ideology. While social scientists have always shown a keen interest in the nuts and bolts of belief systems, the resurgence of polarization in the ­contemporary era has brought the topic back to the forefront of scholarship and lay discussion alike. In this chapter, I review past and present work on the nature of ideology and its consequences for public opinion. I begin by addressing the definitional question of what ideology is. Next, I provide an overview of several decades’ worth of research on what attracts individuals to different ideological positions and when individuals think and make judgments about issues and candidates in ways that reflect ideology. As we shall see, these two aspects of ideology do not always go hand-​in-​hand. While most citizens willingly identify themselves with various ideological labels (e.g., “liberal” or “conservative”), only those who have absorbed a great deal of information from political leaders think about politics in terms of these ideological categories and express opinions that consistently lean to the left or the right. Finally, I present an integrated perspective describing how ideological affinities and belief systems are shaped by the interplay between top-​down processes (e.g., elite discourse and cue-​giving) and bottom-​up processes (e.g., variation in personal characteristics).

What is Ideology? While the concept of ideology has a familiar ring to it, students of public opinion—​including political scientists, psychologists, and sociologists—​have struggled to converge on a common definition of what it is.1 As a result, those who dive into the topic may find themselves adrift on a sea of competing interpretations. Nevertheless, in public opinion research, the most important working definitions of ideology agree on a number of crucial claims.2 First, ideologies are belief systems or networks of inter-​related ideas. That is, they consist of opinions, values, and beliefs about society that can be grouped together under some common social theme, for example, moving society in the direction of greater justice and equality. In the language of psychology, this claim suggests that ideologies are schemas—​i.e., organized clusters of ideas about

76  Christopher M. Federico

social and political life that have been stored away in long-​term memory.3 Second, ideologies are shared by, and reflect the life situations of, groups of individuals. In this respect, ideologies are usually not idiosyncratic; they are held in common by a group of people living in a particular social and historical context, and they express the opinions, values, and concerns those people have developed as a result of the challenges and opportunities present in that context. Third, ideologies are both descriptive and prescriptive in nature. That is, they both provide an interpretation of society as it currently exists and set out aspirations for how society should ideally be organized and how political goals should be attained. Thus, ideologies can be thought of as shared belief systems that reflect a group’s understanding of the social world and its vision of what that world should ideally look like. However, this definition tells us very little about how ideological belief systems are organized. On this point, there has also been a good bit of debate among scholars about the nature of ideology. On one hand, a standard assumption is that ideological positions can be ordered according to the familiar “left–right” spectrum. This perspective suggests that ideological phenomena can be boiled down to a single dimension anchored by preferences for equality and openness to social change on the left end of the spectrum and by preferences for hierarchy and preservation of the status quo on the right end.4 In this usage, the terms “left” and “right” date back to seating arrangements in the French National Assembly during the revolutionary era, which placed the more conservative factions on the right side of the hall and the more radical factions on the left.5 This understanding of the structure of ideology continues to inform current scholarship, and it receives support from a number of sources. To begin with, use of a single left–right spectrum is clearly the norm among those most involved in political action and political decision-​making, that is, “political elites” in government, party and activist organizations, the media, and academia.6 Similarly, the most well-​informed and politically active members of the general public—​who are more likely to attend to signals from the aforementioned elites about “what goes with what”—​also rely strongly on the basic left–right dimension in their thinking and political judgments. Indeed, this tendency has grown stronger in recent decades as elites from the Democratic and Republican parties have ideologically diverged.7 Finally, a great deal of evidence suggests that individuals who place themselves at different positions on the left–​right spectrum tend to adopt correspondingly different opinions about issues connected with the core distinctions of equality versus inequality and openness to change versus the status quo, though mainly when they are politically well informed and attentive to political elites.8 That is, self-​described liberals are more likely to take liberal positions on specific issues, whereas self-​described conservatives are more likely to take conservative stances on the same issues. Of course, ideological self-​placement and the general liberalism or conservatism of one’s issue preferences rarely align in a simple one-​to-​one fashion.

Ideology and Public Opinion  77

In this regard, researchers have been careful to distinguish between symbolic (or philosophical) ideology and operational ideology.9 Symbolic ideology refers to whether one generally identifies oneself as a “liberal,” “conservative,” or some moderate position in between, while operational ideology refers to one’s average left–right position across issues, especially those relevant to government spending. Importantly, evidence suggests that these two types of ideology do not coincide for many citizens. An example of this is Free and Cantril’s classic finding that many Americans living in the middle of the twentieth century were simultaneously “philosophical conservatives” and “operational liberals,” opposing “big government” in the abstract but offering strong support for the individual programs that made up “big government.”10 More recently, Ellis and Stimson have noted that over two-​thirds of those who identify as symbolic conservatives are operational liberals on the issues.11 Thus, even in situations where citizens do appear to rely on a single left–right spectrum, their beliefs may differ at the levels of general identification and actual judgments about issues. We return to this point about the lack of correspondence between left–right self-​placement and issue opinions below. While the notion of a single ideological dimension is simple and consistent with a great deal of evidence, other scholars have argued that citizens’ ideological views may have more than one dimension. Multidimensional approaches of this sort come in a number of forms, but the most common variant argues that ideology can be characterized in terms of two different content dimensions.12 Broadly speaking, the first dimension reflects one’s preference for equality versus inequality, while the second dimension reflects one’s preference for openness versus social order. “Dual process” models of this sort suggest that the two dimensions are governed by distinct but related sets of psychological processes, and that the dimensions may operate somewhat independently.13 Thus, while the notion of a single left–right spectrum implies that a “right-​wing” orientation on one dimension (e.g., a preference for inequality) should be accompanied by a right-​wing orientation on the other (e.g., support for traditional values), the dual-​process approach suggests that views on the two dimensions need not be perfectly aligned with one another. Indeed, bi-​dimensional structures like this recur repeatedly in analyses of various kinds of social and political beliefs. For example, researchers interested in human values—​abstract beliefs about desirable social goals or modes of conduct—​ suggest that the former can be arrayed according to two such dimensions.14 The first dimension deals with “self-​transcendence versus self-​ enhancement,” and it reflects concerns about power and rank. This dimension is anchored by values like benevolence and universal concern for others at one end and values like power and achievement at the other. The second deals with “openness versus conservation,” and it reflects concern about security and order. This dimension is anchored by values like self-​direction at one end and tradition and conformity at the other. Similar dual dimensions have been identified in classic studies of political extremism, which suggest that ideologies

78  Christopher M. Federico

can be distinguished in terms of their level of support for equality versus inequality, on one hand, and their level of intolerance for alternative points of view on the other.15 The dual-​process approach also finds an echo in public-​ opinion research suggesting that there are distinct dimensions corresponding to attitudes about “economic” issues (e.g., the extent to which government should regulate business, how much should the government spend on social welfare, etc.) and “social” issues (e.g., should the government protect gays and lesbians from discrimination, should immigration be increased or decreased).16 A similar dichotomy manifests itself at the level of differences between political parties across nations, with research suggesting that parties compete with one another for votes along two dimensions corresponding to concern for equality and preferences regarding tradition versus change.17 So, which of these perspectives is correct? Is ideology best thought in terms of a single left–right dimension, or multiple dimensions corresponding to preferences regarding equality and social openness? While the bulk of the data suggests that two dimensions of ideology are in fact present, it is also clear that these dimensions are not completely independent of another and that they may be very highly aligned among the politically engaged.18 In this vein, a number of theorists have suggested that having multiple dimensions of political evaluation makes many political decisions more difficult.19 Many common political choices—​such as who to vote for or which party to affiliate with—​are dichotomous in nature:  there is a left-​wing option and a right-​wing option. If one’s preferences on the two dimensions are not aligned, political choices become more fraught with conflicts and trade-​offs. This implies that individuals who are most involved in political decision-​making—​including not just political professionals, but also the most informed and politically involved segments of the mass public—​would benefit most from a belief system in which the two dimensions of ideology overlap with one another to form a single left–right axis. Consistent with this argument, attitudes associated with the equality and openness dimensions are more likely to be aligned with one another among the highly informed and involved and among elected officials.20 Research also suggests that intensified political competition may lead multiple ideological dimensions to collapse more cleanly into a single left–right dimension. For example, cross-​ national comparisons reveal that the equality and openness dimensions are more likely to be positively correlated in countries with established systems of party competition with distinct left-​wing and right-​wing options.21 Moreover, once a second dimension of ideological competition arises in a political system, it tends to become aligned with the existing dimension, such that parties and politicians that support equality also tend to support openness and freedom.22 Thus, while there may be multiple dimensions of ideology, they are seldom fully independent of one another—​and the need to effectively organize political competition and decision-​making may lead to considerable overlap between the dimensions among those most engaged in politics.

Ideology and Public Opinion  79

What Attracts People to Different Ideological Positions? Having discussed the issue of what ideology is, I turn to the question of what attracts people to various ideological positions. Working from a variety of theoretical perspectives, political scientists, psychologists, and sociologists have brought a great deal of data to bear on this question. From this welter of research, a number of answers have emerged. While some of these answers have a decidedly “social” feel and center on features of an individual’s social environment, such as his or her social position and group memberships, others focus more on intra-​individual characteristics and processes, such as psychological needs, personality traits, moral predispositions, and even genetics. At a glance, Figure 3.1 summarizes some of the most important factors that attract people to particular ideological positions. Below, I review these factors in more detail. Identifications and relationships. As one might expect, social relationships—​ both with other individuals and with groups—​are a major influence on citizens’ ideological sympathies. In general, we tend to adopt the views of those we like, identify with, or otherwise have some social connection to. For example, activating a desire for closeness to a significant other leads individuals to align their own social and political opinions more closely with the perceived views of the

Social Factors Relationships with parents and family

Does ideological position strengthen social ties? Does ideological position serve the in-group interests?

Identifications with reference groups Group interests with respect to power and resources

Self-placement on the general left–right dimension Individual Factors Needs: certainty and security Goals: power and order Personality: openness and conscientiousness Genes

Does ideological position meet needs and goals? Is ideological position consistent with personality? Is ideological position consistent with intermediate phenotypes coded by genes?

Figure 3.1 What Attracts People to Different Ideological Positions?

80  Christopher M. Federico

other. In turn, this increased similarity in opinions produces a “shared reality” that helps solidify the relationship in question.23 Research on political socialization—​the process by which individuals learn about the political world and acquire their opinions and views from others—​ also illustrates the role of relational influences on the acquisition of ideological positions.24 In this vein, decades of research suggest that ideological positions are likely to be transmitted from parents to children, especially if both parents have similar beliefs and discuss politics frequently and if bonds within the family are close.25 Similarly, reference groups—​groups that people use as a standard for evaluating themselves and their behavior—​also have a strong influence on ideological affinity.26 That is, we tend to adopt the views of groups we relate to in some way. This was famously demonstrated by Theodore Newcomb in his study of female undergraduates at Bennington College.27 Newcomb’s undergraduates were largely conservative in their views when they first arrived at Bennington, having come from fairly well-​off Eastern families. However, the Bennington environment itself—​consisting of faculty and older students—​was largely liberal. Accordingly, Newcomb’s undergraduate subjects moved further to the left the longer they were at Bennington, as they came to identify more with the campus social environment; the only exceptions were undergraduates who remained unusually close to their families while at college. Naturally, these reference-​group effects are not limited to one’s college environment. They are also quite evident with respect to a number of more politically relevant group identities, such as one’s religion, region, or occupational category.28 These relational influences on ideology also appear to be stronger at some times than at others. In general, parents, peers, and reference groups have their strongest effects on ideological affinity during late adolescence and early adulthood, before a person’s identify is fully fleshed out.29 Moreover, the resulting identifications tend to persist as long as a person’s pattern of relationships with other individuals and groups does not change markedly. For example, this pattern was notably evident in Newcomb’s Bennington study: while Newcomb observed profound changes in his subjects’ attitudes over the course of college, follow-​up interviews later in their adult lives revealed little additional change. Group interests. Another relational influence on ideological affinity follows from identification with groups, namely, the pursuit of those groups’ collective political interests. In general, a long line of theorizing—​derived mainly from economics—​suggests that individuals should make political choices that reflect self-​interest. However, a wide variety of research suggests that personal self-​ interest has only minor effects on what people believe ideologically; that is, calculations about one’s own economic interests or the interests of one’s family have little influence on political attitudes.30 Nevertheless, the perceived collective interests of the social groups one identifies with do tend to influence ideology, especially when people are highly aware of their identity as a member of a particular group. Generally speaking, members of groups with low social status, less power, and/​or fewer resources

Ideology and Public Opinion  81

tend to be more egalitarian and left-​wing in their political outlook, while members of more privileged groups gravitate toward the right.31 However, this effect is not a simple “reflective” one: not all individuals who identify with a group adopt group-​interested ideological positions. In fact, for some members of less well-​off groups, the group-​interest motive may be countered by other motives, such as the need to avoid the uncertainty or insecurity that might come from challenging the political status quo.32 Thus, the effect of group interest—​while often present—​is rarely total. Psychological needs. Moving away from the level of broader social influences on ideological affinity, a great deal of research suggests that preferences for the left versus the right may be heavily shaped by various psychological needs. In particular, attraction to different positions on the left–right spectrum is not random but systematically rooted in one’s underlying level of comfort with uncertainty and threat. All other things being equal, this body of work finds that strong needs for certainty and security correlate with greater conservatism—​as support for the status quo allows individuals to stick with what is known, familiar, and safe—​whereas weaker needs for certainty and security are associated with greater liberalism. This argument was first and perhaps most famously made by the authors of The Authoritarian Personality, who suggested that attraction to far-​right ideological positions (as opposed to liberal or left-​leaning positions) was driven by an “authoritarian personality” type consisting of nine interrelated tendencies, such as rigid moral conventionalism, aggressiveness, submission to idealized leaders, and a preoccupation with power and toughness.33 In turn, this intolerant type was explained using some of Sigmund Freud’s ideas about the management of anxiety. Specifically, Adorno and his colleagues argued that harsh childrearing led authoritarians to “repress” hostility toward their parents and other traditional authorities and “project” it outward onto the scapegoats and outsiders often targeted by right-​ wing political ideologies (e.g., minorities, those who desire social change, etc.). This need-​based perspective was echoed by other early theorists as well.34 While these models—​and the Adorno et al. model in particular—​were later the subject of numerous theoretical and methodological criticisms,35 the notion that ideological affinity may be rooted in some underlying feature of psychological functioning has persisted. In this vein, psychologist Bob Altemeyer has updated the authoritarianism construct, re-​labeling it right-​wing authoritarianism (RWA) and characterizing it more simply as a learned constellation of three attitudes:  conventionalism, “authoritarian submission” to traditional social authorities, and “authoritarian aggression” toward disliked outgroups.36 Others have reduced the concept even further, boiling it down to a general desire for social conformity.37 More recently, John Jost and his colleagues—​ echoing the scholars mentioned above—​have reiterated and provided much new evidence for the view that support for the right is associated with need for certainty and security, whereas support for the left is associated with greater tolerance for uncertainty and potential social danger.38

82  Christopher M. Federico

Other psychological approaches have suggested that different sets of needs may account for preferences in different domains of political belief. As noted above, many researchers have suggested that ideology may consist of two dimensions—​one corresponding to preferences for equality versus inequality and the other corresponding to preferences for openness versus order. In turn, these two dimensions may each be related to a distinct set of psychological needs. For example, Duckitt and Sibley argue that the equality dimension reflects orientations toward social hierarchy and depends on the value an individual places on superiority, achievement, and power, whereas the openness dimension reflects orientations toward traditional morality and social conformity and depends on the value an individual places on order and security in social life.39 Moreover, each dimension may be associated with a distinct worldview—​either the extent to which one believes that the world is a competitive, violent place, in the case of the equality dimension, or the extent to which one believes that it is a dangerous place, in the case of the openness dimension. People who see the world as a dangerous place tend to prefer order, conformity, and security, which attracts them to conservative positions on social issues related to religion, gender, and social convention. In comparison, people who see the world as highly competitive place a premium on social hierarchy and social dominance, which attracts them to those aspects of conservatism which favor inequality, particularly in the economic realm. Thus, the needs for certainty and security highlighted by the “classic” models of ideological affinity reviewed earlier may actually be pertinent to only some of the political concerns that distinguish the left from the right—​namely, those dealing with openness and freedom versus tradition and order. Personality characteristics. Another set of psychological approaches to ideology have focused on how individual differences in personality might account for variation in ideological sympathies. By “personality,” these approaches refer to sets of characteristics possessed by a person that uniquely shape his or her thoughts, feelings, and behaviors across various situations.40 Speculation about links between personality traits and politics has long been a preoccupation among psychologists.41 In the present era, the dominant framework for examining the relationship between personality and ideology in recent years has been the “Big Five” model of personality traits.42 Using analyses of personality adjectives found in everyday language, this model boils variation in personality down to five key dimensions. These include extraversion, one’s level of sociability and assertiveness; agreeableness, one’s level of altruism and concern for others; conscientiousness, one’s level of concern for social duty, responsibility, and impulse control; emotional stability, one’s level of even-​temperedness or freedom from negative emotion; and openness to experience, one’s level of interest in novelty, complexity, and originality. Numerous studies by psychologists and political scientists alike have examined relationships between these five dimensions and support for the left

Ideology and Public Opinion  83

versus the right. This impressive body of work is reviewed in detail elsewhere,43 so I will touch only on its key findings. In this regard, the Big Five dimensions that have the most consistent relationships with ideological affinity are openness to experience and conscientiousness. While openness is typically associated with greater support for the left, conscientiousness is usually associated with greater support for the right.44 The relationships between each of these dimensions and ideology are not negligible in size; indeed, they are similar in magnitude to the relationships between ideology and key demographic variables like education and income. In contrast, the relationships between the other three Big Five dimensions and ideology are far less consistent across studies, and they also tend to vary across issue domains (i.e., economic issues versus social issues) and social groups (e.g., Blacks versus Whites).45 Morality. Noting the highly moralistic tone of many contemporary ideological disputes, other perspectives argue that left–right differences may be rooted in divergent moral sensibilities. One well-​known proponent of this view is the social psychologist Jonathan Haidt, who suggests that moral judgment rests on five intuitive moral foundations: avoidance of harm to others, concern for fairness, group loyalty, respect for authority, and purity from contamination.46 The first two foundations are considered individualizing: they function to protect individual rights. The other foundations—​related to loyalty, authority, and purity—​are considered binding: their function is to protect the integrity and cohesion of social groups. Importantly, Haidt and his colleagues have shown that liberals and conservatives differ in their relative emphasis on these two sets of concerns: conservatives tend to emphasize all five foundations equally, whereas liberals emphasize the two individualizing foundations and downplay the importance of the three binding ones.47 Other studies have suggested the two sets of foundations may relate to different dimensions of ideology, with the individualizing foundations predicting concern for equality and the binding foundations predicting preferences for conformity over openness.48 This pattern is echoed by research on ideology and human values, some of which was discussed previously. Specifically, studies on this topic indicate that those who identify with the left place a greater emphasis on self-​transcendence values like benevolence and universal concern for humanity, while those who identify with the right emphasize conservation values focused on security, order, and tradition.49 Thus, evidence from a variety of quarters suggests that ideology may at least in part reflect deeper individual differences in moral understanding. The role of genetics. Finally, researchers in a number of disciplines have begun to explore the possibility that ideological sympathies may at least in part be genetically shaped and transmitted.50 This view is a sharp departure from most public-​ opinion research on ideology, which has traditionally assumed that ideological affinity is socially learned from parents, peers, and important social groups. This line of work generally relies on what is known as the “classic twin design” in order to estimate what proportions of the variation among individuals in political

84  Christopher M. Federico

opinion is due to genes, common environmental influences (i.e., those shared by members of a family), and unique environmental influences (i.e., those not shared by family members).51 The method does this by comparing the attitudes of identical twins, who effectively share 100 percent of their genetic makeup, and fraternal twins, who share roughly 50 percent of one another’s genetic heritage. This fixed difference between identical and fraternal twins in genetic relatedness, along with the assumption that a given pair of twins—​whether identical or fraternal—​is subject to similar environmental influences, allows the researcher to tease apart the relative impact of genes and environment. Studies using this and other related techniques (such as molecular genetic methods) have found relatively large effects of genes on ideological opinions. Research on this topic is reviewed more comprehensively by Gonzalez, Hibbing, and Smith in Chapter 10 of the present volume, who make a compelling argument that the influence of genetics on political attitudes and behavior is more extensive than traditionally believed and that preferences can be transmitted from parents to children even in the absence of direct socialization.52 As such, I only discuss the most important findings here. Specifically, some 40 percent to 50 percent of the variability in left–right political opinions among individuals appears to be attributable to genetic differences as opposed to differences in social environment.53 Interestingly, however, genetics does not appear to contribute to differences in partisanship, although it may have an influence on the strength of partisan identification.54 Thus, while social influences on ideological affinity are undoubtedly important, at least some portion of what people believe politically may be linked to their genetic makeup. Not surprisingly, this conclusion has been controversial. Among other things, the validity of the assumption that identical and fraternal twins are subject to the same environmental influence has come in for criticism, and others have noted that heritability of political attitudes may differ across populations living in different environments. Nevertheless, statistical corrections for potential differences in environment across types of twins fail to rule out environmental effects, suggesting that biological influences must be considered by any full treatment of the bases of ideology. Moreover, other common deviations from assumptions of twin studies—​such as the widespread tendency for individuals to choose mates with similar political attitudes—​may cut in the other direction, leading to an underestimate of genetic influences.55 One open question in this literature has to do with factors that mediate the connection between genes and ideology. In this vein, some research suggests that genes influence intermediate phenotypes or observable traits related to social behavior (e.g., orientations toward threat or social order), which then affect specific attitudes in the domain of politics (e.g., social conservatism). However, other perspectives suggest “pleiotropy” rather than mediation (i.e., a process in which both ideology and supposedly intermediate traits are independently shaped by the same genetic influences).56

Ideology and Public Opinion  85

When Do People “Use” Ideology? Thus far, I have discussed ideology primarily in terms of attraction to particular ideological identities or positions along the left–right spectrum. However, as noted at the outset, ideologies are also belief systems. Besides some crowning posture or identity—​ like liberalism or conservatism—​ they consist of an interlocking web of opinions, values, and perceptions of social reality that go along with that posture. For example, an identification with the political left implies a wide range of issue opinions across a variety of domains—​support for welfare spending, support for gay rights, support for diplomacy over the use of force in international relations, and so on. Moreover, it implies support for general value postures (e.g., preferences for equality and self-​direction) as well as certain beliefs about the nature of the social world (e.g., inequality stems from structural factors like discrimination as opposed to an individual lack of ability or effort). Indeed, as we have seen, the general ideological label a person adopts typically has consequences for their opinions about specific issues:  those who identify themselves as conservatives tend to adopt conservative issue positions, while those who identify themselves as liberals tend to adopt liberal issue positions. But where do these broader belief packages come from? Public opinion researchers have long noted that the sheer force of logic is not sufficient to explain why certain issue positions get linked together as part of an ideological whole; as just one example, there is no apparent reason why opposition to legal abortion should go together with support for lower taxes as part of the contemporary “conservative” belief package. As such, most scholars have come to regard ideologies as products of convention—​or more specifically, the culture of the groups that share the ideology.57 Nevertheless, most perspectives on ideology suggest that the social activities which give rise to ideological content are disproportionately the province of narrow elites within the groups that share different ideologies—​usually powerful and unrepresentative ones. This emphasis is perhaps most evident in the classical sociological tradition and in Marxist approaches to social science, both of which have argued—​ albeit with different evaluative implications, depending on the writer—​that the discursive content of ideologies should disproportionately represent the interests of powerful groups and justify states of affairs the latter benefit from. Other approaches have placed a similar emphasis on the construction of belief packages by small, highly involved segments of the population. However, in these models, the focus is less on the role of dominant groups whose interests color the content of ideologies and more on the role of the “political elites” discussed earlier—​the politicians, activists, and media figures whose activity develops the constellations of positions, values, and interpretations of reality that make up different ideological positions.58 From this perspective, groups of political elites in competing political parties adopt belief packages reflecting particular interests and values—​ belief packages that are often articulated

86  Christopher M. Federico

ahead of time by intellectuals and transmitted to party elites by activists and other “intense policy demanders.”59 These opposed packages of opinions and positions then serve to psychologically “anchor” the ends of the left–right spectrum in a particular context; they make up the ideological “menu” from which members of the mass public typically make their political choices (e.g., votes). The flip side of this notion of “elite opinion leadership” is that the beliefs held by citizens at the mass level are typically not constructed by the citizens themselves. Rather, the content associated with different ideological positions is acquired by members of the mass public when they “take cues” about what to believe from political figures that share their basic partisan and ideological identifications.60 Thus, for most people, the packages of issue opinions, values, and views associated with a particular ideological posture are learned from those more highly involved in politics. This raises an important question:  to what extent does the content associated with particular ideological positions fully diffuse to the broader public? By and large, decades of research suggest that most citizens do not learn the full set of “correct” opinions and views associated with the ideological identity they claim, even when they do claim a left–right position.61 On one hand, as Donald Kinder and Nathan Kalmoe recently showed using data from surveys collected between 1972 and 2012, 72.5 percent of respondents were willing to place themselves on the left–right spectrum and label themselves as liberals, conservatives, or something in between.62 On the other hand, most citizens—​even if they do adopt an ideological position—​fail to pick up the broader systems of opinions and views associated with particular ideological positions in elite discussion.63 Perhaps the most important demonstration of this incomplete learning of ideology was provided by Philip Converse in a famous essay on “The Nature of Belief Systems in Mass Publics.”64 In this essay, Converse reviewed findings taken from large surveys of American political elites and everyday citizens conducted during the 1950s. To begin with, Converse demonstrated that most members of the general public showed what political scientists refer to as a low “level of conceptualization”—​that is, only a minority typically characterized political parties and candidates in terms of ideological categories like liberalism and conservatism. Similarly, most people were not able to explain the philosophical differences between conservatism and liberalism, and they are not able to accurately indicate which issue positions “go along” with each of these two ideological categories. Moreover, most citizens showed relatively low levels of ideological constraint, that is, they do not take consistently liberal or consistently conservative positions across different issues. Finally, the issue opinions of most citizens showed little stability over time—​that is, they tend to fluctuate randomly over time, which is not what we would expect if opinions were more deeply anchored in an overarching ideological posture like liberalism or conservatism. Importantly, in each of these cases, Converse found that his samples of political leaders showed far more ideological sophistication: on average, they revealed a higher level of conceptualization, they understood the meaning of

Ideology and Public Opinion  87

ideological labels better, and their issue opinions showed greater constraint and stability over time. To Converse, this suggested that the issue opinions of a large portion of the general public were effectively “non-​attitudes.” That is, in most cases, survey respondents were neither interested in, nor informed about, the issues they were queried on, and they definitely did not use a common left–right standard when making judgments about them. Accordingly, they offered off-​the-​cuff, “doorstep” opinions that showed little structure, ideological or otherwise. Although there has been some debate about whether the public’s apparent ideological innocence is really an artifact of imperfect measurement65 and about the extent to which the average citizen makes greater use of ideology than in the past,66 much of the research in this area has followed Converse in concluding that most citizens’ political preferences are not heavily structured by ideology.67 Other lines of work have demonstrated a similar lack of ideological structure in citizens’ opinions. For instance, to refer back to an example from earlier, recall that several researchers have noted a lack of concordance between symbolic and operational ideology—​ that is, there are many citizens who label themselves as “conservatives” while taking issue positions that lean toward the left.68 Naturally, the conclusions reached by Converse and others have somewhat negative implications for democracy, as they suggest that much of the citizenry is too disengaged from politics to form “real” opinions about crucial issues and organize their opinions in an ideological fashion. This has led many scholars to ask what factors allow citizens to think in ideological terms and adopt opinions that are consistent with the ideological positions they claim. At the mass level, the main factor governing the acquisition of ideological content is exposure to flows of information from leading political figures, which is highest among those with a strong interest in politics, the highly educated, those who see themselves as politically competent, and members of relatively privileged social groups.69 Individuals who receive a good deal of political information over time eventually build up elaborate political knowledge structures in long-​term memory, leading to the development of political expertise.70 As a result, highly informed citizens are more likely to have learned what goes with what politically, that is, the specific issue positions and views about the world that go along with being a liberal, a conservative, or something in between. In turn, this knowledge results in patterns of thinking and opinion which more closely resemble those of political leaders. So, for example, well-​informed citizens are more likely to make active use of concepts like liberalism and conservatism in explaining differences between parties and candidates,71 and they show higher levels of ideological constraint in their opinions toward different issues.72 They are also more likely to possess operational issue positions that match their symbolic ideological identifications.73 On the whole, the sheer volume of data suggesting that the well informed are more likely to think and make judgments in ways that reflect a mastery

88  Christopher M. Federico

of ideological content has had a profound effect on how public opinion researchers understand the phenomenon of ideology. Indeed, even a brief look at the literature on the topic makes it clear that the consensus political-​science view is that citizens will think in an ideological way and adopt an ideologically consistent set of opinions to the extent that they have successfully received political information from elites and stored it away in long-​term memory in the form of an organized knowledge structure. Of course, researchers have not concluded that information is the only factor governing the use of ideology. For example, recent work has qualified the information story somewhat, finding that political information is more likely to be associated with the use of ideology among those motivated to evaluate various objects as “good” or “bad.”74 As a final note, it should be emphasized that the poor mastery of ideological content that follows from a lack of information does not make citizens completely incapable of being political. In this respect, other orientations that are not strictly ideological but which have some political relevance—​like values and social group memberships—​may serve many of the same functions as the more complex ideological belief systems discussed above:  they provide cues about what positions to take on various political issues, justify one state of affairs over another, and so on.75 Among others, these substitutes for ideology include core political values such as egalitarianism, moral traditionalism, and self-​reliance, which even information-​poor citizens may be able to use in making political judgments.76 Similarly, much evidence suggests that reference groups—​like one’s social class, racial or ethnic group, or religious affiliation—​ can be used as cues about what positions to take or how to vote as individuals at all levels of information.77 Finally, standard surveys may not be able to detect discursive frameworks that differ from the ones offered by political leaders. Since these techniques are best used to detect “ideology” in the conventional left–right sense, they may miss idiosyncratic belief systems that are as elaborate and internally consistent as those offered by elites and which serve all of the important functions of ideology (e.g., organizing different opinions together under common themes, justifying political action, explaining the world, etc.). Indeed, in-​depth interviews of citizens with low-​to-​average levels of political information have shown that normal conversation can reveal coherent “ideological” understandings of political reality that nevertheless depart from the left–right framework used by political elites.78

Integrating Bottom-​U p and Top-​D own Perspectives on Ideology In reviewing the sizable, interdisciplinary body of research on ideology, this chapter has highlighted two basic perspectives on the topic: a “bottom-​up” perspective focusing on how personal and social characteristics attract different citizens to different ideological postures, and a “top-​ down” perspective

Ideology and Public Opinion  89

focusing on how information derived from attention to elite signals allows (some) citizens to construct ideologically consistent belief systems. Research reflecting these perspectives has often proceeded on separate tracks—​and often in different disciplinary silos, with psychologists emphasizing bottom-​up processes and political scientists emphasizing top-​down ones. In an effort to surmount this divide, my colleagues and I have recently offered an integrated perspective on ideological processes. Specifically, we argue that only those citizens who receive sufficient information from elite signals will sort into ideological identities that “match” their personal and social characteristics and then adopt ideologically consistent viewpoints.79 Put another way, the information that citizens acquire from attention to political elites governs not only their ability to adopt “constrained” belief systems that are consistent with their overall ideological identity, but also their likelihood of adopting an ideological identity that reflects characteristics such as psychological needs, personality, and moral orientations. In recent studies, we have presented several converging lines of evidence in favor of the idea that top-​down processes related to attention to elite discourse may also play a role in the bottom-​up shaping of ideological affinity. For example, a growing body of evidence suggests that psychological needs and personality traits are more likely to predict ideological affinity among those who are high in political information and interest. In this vein, our analyses consistently indicate that the relationship between psychological variables reflecting needs for certainty and security are more strongly related to conservative ideological identification among those who are more informed about, and interested in, political matters.80 As an illustration of this from one of our studies, Figure 3.2 shows this effect with respect to the relationship between authoritarianism and identification as conservative in the 2004 American National Election Study. Paralleling this result, we have also shown similar patterns with respect to partisanship, which has become increasingly intertwined with ideology in the United States. For instance, authoritarianism is more strongly related to partisanship in general—​and to attraction to the socially conservative Republican Party in particular—​among those who are politically well informed.81 In sum, we find that individuals differentially sort into varying ideological and partisan identities as a function of attention to elite signals: those who are interested and well informed are more likely to sort into political identities that “match” their underlying psychological characteristics. Importantly, this may also have downstream consequences for attitudes toward specific political issues. Citizens who pay more attention to elite signals, having sorted into ideological and partisan identities that fit their needs, should then take cues from like-​minded elites about what issue positions go with those identities. As a result, the relationship between needs for certainty and security and conservative issue positions is stronger among the politically well informed.82 The moderating effect of attention to elite discourse on the relationship between personal characteristics and issue positions takes an especially

90  Christopher M. Federico

1

Ideological conservatism

.8

.6

.4

.2

0 Low authoritarianism

High authoritarianism Low information

High information

Figure 3.2 The Relationship between Authoritarianism and Ideology is Stronger among those High in Political Information. Scores on the ideology variable have been recoded to run from 0 (most liberal) to 1 (most conservative).

interesting form in the realm of economic preferences. One interesting quirk in the literature on the relationship between psychological variables and political attitudes is that the connection breaks down in the case of attitudes toward redistribution and government intervention in the economy:  needs for certainty and security strongly predict social-​issue conservatism, but not conservatism in economic preferences (i.e., favoring minimal government). Our work suggests that this null result may be due not to a complete absence of any effect of psychological variables in the economic domain, but to heterogeneity in the relationship between needs for security and certainty and economic preferences between those low and high in attention to elite discourse. Specifically, we argue that strong needs for security and certainty should predict right-​leaning, market-​oriented positions among the interested and informed, but left-​leaning, interventionist views among those low in interest and information.83 This is because individuals high and low in interest and information differ in how they translate their relative desires for certainty and security into economic preferences. Among the interested and well informed, economic preferences serve an expressive function. As noted previously, citizens with strong needs for certainty and security are more likely to sort into conservative ideological and partisan identities. In turn, they take cues about what economic positions to adopt from elites who share those identities, leading them to adopt

Ideology and Public Opinion  91

conservative, small-​government economic positions that reflect their political affiliations. In contrast, among those lower in interest and information, economic preferences serve an instrumental function. Since these individuals pay less attention to elite signals, they are less likely to have sorted in ideological and partisan identities that match their psychological characteristics and less likely to receive cues from political elites about what economic views they should adopt. Nevertheless, needs for certainty and security have an unmediated impact on these individuals’ economic views—​one that is not channeled through political identities. For those low in interest and information, needs for certainty and security directly influence the extent to which they prefer that government “insure” them against uncontrollable risks via redistribution (e.g., unemployment insurance, universal health care) and regulation (e.g., workplace safety rules). Rather than adopting right-​leaning economic positions to express a conservative identity, low-​engagement citizens with strong needs for certainty and security adopt left-​leaning economic positions in a direct effort to avoid uncertainty and risk in the economic sphere. Recent evidence provides relatively strong support for this “reversal” or “menu-​dependence” hypothesis. For example, Christopher Johnston, Howard Lavine, and I found evidence for it in the United States in 21 tests using ten large nationally representative datasets. Variables reflecting strong needs for security and certainty (such as authoritarianism) generally predicted opposition to redistribution and regulation among the interested and informed, but support for redistribution and regulation among those relatively low in political interest and information.84 Echoing this, analyses by Ariel Malka and his colleagues found similar results in a multinational sample from over 40 different countries.85 Thus, attention to elite political discourse—​and the information it provides—​may structure not only the relationship between ideological affiliation and broader belief systems, but also the relationship between ideological affiliation and its pre-​political determinants—​and it may do so in fairly complex ways in particular issue domains.

Conclusion As Philip Converse noted a half-​century ago, ideologies have often “served as primary exhibits for the doctrine that what is important to study cannot be measured and that what can be measured is not important to study.”86 Since then, public-​opinion research has made considerable progress not only in overcoming the difficulties inherent in studying ideology, but also in reinforcing the importance of ideology as an explanatory concept. In this chapter, I have attempted to provide the reader with an overview of this body of work. Specifically, I  have tried to address the questions of what ideology is, which social and psychological factors attract citizens to one ideological posture rather than another, and when citizens “use” ideology to make political judgments and adopt political postures that reflect their basic psychological dispositions.

92  Christopher M. Federico

As we have seen, ideologies are clearly relevant to mass opinion in certain respects. Ideologies of the left and right play a key role in the organization of opinion-​holding and debate among political leaders and other elites. Moreover, under the influence of various identifications and interests, psychological needs and traits, and perhaps even their genetic inheritance, members of the general public are sufficiently attracted to one ideological posture or another to place themselves on the left–right spectrum as liberals, conservatives, or something in between. However, at the mass level, far fewer citizens actually adopt the full range of views and opinions that “go along” with these postures. In an effort to explain this state of relative ideological innocence, researchers have focused in particular on the role of information received from political leaders, repeatedly observing that well-​informed citizens are more likely to think about politics in ideological terms and express opinions that are ideologically consistent with one another. Extending this perspective, my own work suggests that “top-​ down” processes associated with attention to elites may also condition the “bottom-​up” effects of individual psychological characteristics on ideological affinity. Thus, while ideology plays a key role in the organization of political discussion in the political system as a whole, its influence in the political lives of individual citizens depends heavily on how much they attend to elite cues and signals and how much information they acquire as a result.

Notes 1 Philip Converse, “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, ed. D.  Apter (New  York:  Free Press, 1964); John T.  Jost, Christopher M.  Federico, and Jaime L.  Napier, “Political Ideology:  Its Structure, Functions, and Elective Affinities,” Annual Review of Psychology 60 (2009):  307–​337; Donald R.  Kinder and Nathan P.  Kalmoe, Neither Liberal nor Conservative:  Ideological Innocence in the American Public (Chicago, IL: University of Chicago Press, 2018). 2 Arthur D. Denzau and Douglass C. North, “Shared Mental Models: Ideologies and Institutions,” in Elements of Reason: Cognition, Choice, and the Bounds of Rationality, ed. Arthur Lupia, Matthew McCubbins, and Samuel L. Popkin (New York: Cambridge University Press, 2000). 3 Christopher M. Federico and Monica Schneider, “Political Expertise and the Use of Ideology:  Moderating Effects of Evaluative Motivation,” Public Opinion Quarterly 71 (2007): 221–​252; Susan T. Fiske, Richard R. Lau, and Richard A. Smith, “On the Varieties and Utilities of Political Expertise,” Social Cognition 8 (1990): 31–​48; Charles M. Judd and Jon A. Krosnick, “The Structural Bases of Consistency Among Political Attitudes:  Effects of Expertise and Attitude Importance,” in Attitude Structure and Function, ed. Anthony Pratkanis, Steven Breckler, and Anthony Greenwald (Hillsdale, NJ: Erlbaum, 1989). 4 Kenneth Benoit and Michael Laver, Party Policy in Modern Democracies (New York: Routledge, 2006); R. S. Erikson and K. L. Tedin, American Public Opinion (8th ed.) (New  York:  Longman, 2011); Jost et  al., “Political Ideology”; Seymour M. Lipset, Political Man (New York: Doubleday, 1960). 5 N. Bobbio, Left and Right (Cambridge: Polity Press, 1996).

Ideology and Public Opinion  93 6 Angus Campbell et al., The American Voter (New York: Wiley, 1960); Converse, “The Nature of Belief Systems”; Nolan McCarty, Keith T. Poole, and Howard Rosenthal, Polarized America: The Dance of Ideology and Unequal Riches (Cambridge, MA: MIT Press, 2006); Hans Noel, Political Ideologies and Political Parties in America (New York: Cambridge University Press, 2013). 7 Kinder and Kalmoe, Neither Liberal nor Conservative; John Zaller, The Nature and Origins of Mass Opinion (New  York:  Cambridge University Press, 1992). On polarization, see Alan I. Abramowitz, The Disappearing Center:  Engaged Citizens, Polarization, and American Democracy (New Haven, CT: Yale University Press, 2010); John G.  Bullock, “Elite Influence on Public Opinion in an Informed Electorate.” American Political Science Review 105 (2011): 496–​515; Matthew Levendusky, The Partisan Sort (Chicago, IL:  The University of Chicago Press, 2009); and Lilliana Mason, Uncivil Agreement (Chicago, IL: University of Chicago Press, 2018). 8 Erikson and Tedin, American Public Opinion. 9 Lloyd A. Free and Hadley Cantril, The Political Beliefs of Americans (New Brunswick, NJ:  Rutgers University Press, 1967); Christopher Ellis and James A.  Stimson, Ideology in America (New  York:  Cambridge University Press, 2012); Matthew Grossmann and David A. Hopkins, Asymmetric Politics: Ideological Republicans and Group Interest Democrats (New York: Oxford University Press, 2016). 10 Free and Cantril, The Political Beliefs of Americans. 11 Ellis and Stimson, Ideology in America. Note that the opposite pattern—​a pairing of symbolic liberalism and operational conservatism—​is comparatively rare (only 2  percent of the population). As such, Ellis and Stimson argue that the relatively high proportion of symbolic conservatives with operationally liberal beliefs may arise from political discourse that associates the “liberal” label with negative cultural connotations without calling for reductions in the most popular (and most expensive) government programs, such as Social Security, Medicare, and education spending (see also Grossmann and Hopkins, Asymmetric Politics). 12 Edward G. Carmines and Nicholas J. D’Amico, “The New Look in Political Ideology Research,” Annual Review of Political Science 25 (2015): 205–​216; Stanley Feldman, “Political Ideology,” in The Oxford Handbook of Political Psychology, ed. David O. Sears, Leonie Huddy, and Jack Levy (New York: Oxford University Press, 2013). 13 John Duckitt and Christopher G.  Sibley, “A Dual Process Model of Ideological Attitudes and System Justification,” in Social and Psychological Bases of Ideology and System Justification, ed. John T.  Jost, Aaron C.  Kay, and Hulda Thorisdottir (New York: Oxford University Press, 2009). 14 Shalom H. Schwartz, “Universals in the Content and Structure of Values: Theoretical Advances and Empirical Tests in 20 Countries,” Advances in Experimental Social Psychology 25 (1992):  1–​ 65; Milton Rokeach, The Nature of Human Values (New York: Free Press, 1973). 15 Hans J. Eysenck, The Psychology of Politics (New York: Praeger, 1954); Rokeach, The Nature of Human Values. 16 See Carmines and d’Amico, “The New Look”; Geoffrey Evans, A.  Heath, and M.  Lalljee, “Measuring Left–Right and Libertarian–Conservative Attitudes in the British Electorate,” British Journal of Sociology 47 (1996):  93–​112; Ariel Malka, Yphtach Lelkes, and Christopher J. Soto, “Are Cultural and Economic Conservatism Positively Correlated? A Large-​Scale Cross-​National Test,” British Journal of Political Science (2017): 1–​25.

94  Christopher M. Federico 17 Liesbet Hooghe, Ryan Bakker, Anna Brigevich, Catherine de Vries, Erica Edwards, Gary Marks, Jan Rovny, and Marco Steenbergen, “Reliability and Validity of Measuring Party Positions:  The Chapel Hill Expert Surveys of 2002 and 2006,” European Journal of Political Research 49 (2010):  687–​ 703; Hanspeter Kriesi, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier, and Timotheos Frey, West European Politics in the Age of Globalization (Cambridge:  Cambridge University Press, 2008); Seymour M. Lipset and S. Rokkan, Party Systems and Voter Alignment (New York: Free Press, 1967); but see Wouter Van der Brug, and Joost van Spanje, “Immigration, Europe, and the ‘New’ Cultural Dimension,” European Journal of Political Research 48 (2009): 309–​334. 18 Jost et  al., “Political Ideology”; Robert N.  Lupton, William M.  Myers, and Judd R.  Thornton, “Political Sophistication and the Dimensionality of Elite and Mass Attitudes, 1980–​2004,” Journal of Politics 77 (2015):  368–​380; Malka et  al., “Are Cultural and Economic Conservatism.” 19 Anthony Downs, An Economic Theory of Democracy (New York: Harper and Row, 1967); Paul M.  Sniderman and John Bullock, “A Consistency Theory of Public Opinion and Political Choice: The Hypothesis of Menu Dependence,” in Studies in Public Opinion:  Attitudes, Nonattitudes, Measurement Error, and Change, ed. Willem E.  Saris and Paul M.  Sniderman (Princeton, NJ:  Princeton University Press, 2004). 20 On the informed and involved, see Abramowitz, The Disappearing Center, and Judd and Krosnick, “The Structural Bases of Consistency”; on elected officials, see Converse, “The Nature of Belief Systems”; Lupton et al., “Political Sophistication”; McCarty et al., Polarized America. 21 For example, Duckitt and Sibley, “A Dual Process Model”; Gary Marks, Liesbet Hooghe, Moira Nelson, and Erica Edwards, “Party Competition and European Integration in the East and West: Different Structure, Same Causality,” Comparative Political Studies 39 (2006): 155–​175. 22 Liesbet Hooghe, Gary Marks, and Carole J. Wilson, “Does Left/​Right Structure Party Positions on European Integration?” Comparative Political Studies 35 (2002): 965–​ 989; Geoffrey C.  Layman, Thomas M.  Carsey, and Juliana M.  Horowitz, “Party Polarization in American Politics:  Characteristics, Causes, and Consequences,” Annual Review of Political Science 9 (2006):  83–​110; James A.  Stimson, Tides of Consent:  How Public Opinion Shapes American Politics (New  York:  Cambridge University Press, 2004). 23 John T.  Jost, Alison Ledgerwood, and Curtis D.  Hardin, “Shared Reality, System Justification, and the Relational Basis of Ideological Beliefs,” Social and Personality Psychology Compass 2 (2008): 171–​186. 24 For a review, see David O. Sears and Christia Brown, “Childhood and Adult Political Development,” in The Oxford Handbook of Political Psychology, ed. Leonie Huddy, David O. Sears, and Jack S. Levy (New York: Oxford University Press, 2013). 25 M. Kent Jennings and Richard G. Niemi, Generations and Politics: A Panel Study of Adults and Their Parents (Princeton, NJ: Princeton University Press, 1981). 26 Duane F. Alwin, R. L. Cohen, and Theodore Newcomb, Political Attitudes Over the Life Span (Madison, WI: University of Wisconsin Press, 1991). 27 Theodore M.  Newcomb, Personality and Social Change:  Attitude Formation in a Student Community (New York: John Wiley and Sons, 1943).

Ideology and Public Opinion  95 28 B. R. Berelson, Paul F. Lazarsfeld, and W. N. McPhee, Voting: A Study of Opinion Formation in a Presidential Campaign (Chicago, IL:  University of Chicago Press, 1954); Campbell et al., The American Voter. 29 Alwin et al., Political Attitudes Over the Lifespan; Sears and Brown, “Childhood and Adult Political Development.” 30 David O. Sears and Cary L. Funk, “The Role of Self-​Interest in Social and Political Attitudes,” Advances in Experimental Social Psychology 24 (1991): 1–​91. 31 James R. Kluegel and Eliot R. Smith, Beliefs about Inequality: Americans’ Views of What Is and What Ought to Be (New  York:  Aldine de Gruyter, 1986); Sears and Funk, “The Role of Self-​Interest”; Mark J. Brandt, “Do the Disadvantaged Legitimize the Social System? A Large-​Scale Test of the Status–​Legitimacy Hypothesis,” Journal of Personality and Social Psychology 104 (2013):  765–​785; Jason Weeden, and Robert Kurzban, The Hidden Agenda of the Political Mind: How Self-​Interest Shapes Our Opinions and Why We Won’t Admit It. (Princeton, NJ:  Princeton University Press, 2014). 32 Jost et al., “Political Ideology.” 33 Theodor W. Adorno, Else Frenkel-​Brunswik, D. J. Levison, and R. N. Sanford, The Authoritarian Personality (New York: Harper and Row, 1950). 34 For reviews, see John T. Jost, Jack Glaser, Arie W. Kruglanski, and Frank J. Sulloway, “Political Conservatism as Motivated Social Cognition,” Psychological Bulletin 129 (2003): 339–​375; Jost et al., “Political Ideology”; see also Christopher M. Federico, and Ariel Malka, “The Contingent, Contextual Nature of the Relationship Between Needs for Security and Certainty and Political Preferences: Evidence and Implications,” Advances in Political Psychology 39 (2018): 3–​48. 35 For a discussion of these criticisms, see Robert A.  Altemeyer, The Authoritarian Specter (Cambridge, MA: Harvard University Press, 1996), 45–​47; Karen Stenner, The Authoritarian Dynamic (Cambridge: Cambridge University Press, 2005). 36 Altemeyer, The Authoritarian Specter. 37 Stanley Feldman, “Enforcing Social Conformity:  A Theory of Authoritarianism,” Political Psychology 24 (2003):  41–​74. Stenner, The Authoritarian Dynamic; Marc Hetherington and J.  D. Weiler, Authoritarianism and Polarization in American Politics (Cambridge: Cambridge University Press, 2009). 38 Jost et al., “Political Conservatism as Motivated Social Cognition.” Note that recent work has qualified this generalization somewhat, as needs for security and certainty are sometimes associated with “left-​wing” preferences in some domains; see Federico and Malka, “The Contingent, Contextual Nature.” I discuss this in more detail below. 39 Duckitt and Sibley, “A Dual Process Model.” 40 Alan S. Gerber et al., “The Big Five Personality Traits in the Political Arena,” Annual Review of Political Science 14 (2011):  265–​287; Jeffery J.  Mondak and Matthew V. Hibbing, “Personality and Public Opinion,” in New Directions in Public Opinion (2nd ed.), ed. Adam J. Berinsky (New York: Routledge, 2016). 41 Jost et al., “Political Conservatism as Motivated Social Cognition.” 42 Alan S. Gerber et al., “The Big Five Personality Traits.” 43 Alan S.  Gerber et  al., “The Big Five Personality Traits”; Dana R.  Carney, John T.  Jost, Sam D.  Gosling, and Jeff Potter, “The Secret Lives of Liberals and Conservatives: Personality Profiles, Interaction Styles, and the Things They Leave

96  Christopher M. Federico Behind,” Political Psychology 29 (2008): 807–​840; Mondak and Hibbing, “Personality and Public Opinion.” 44 Alan S. Gerber et al., “Personality and Political Attitudes: Relationships Across Issue Domains and Political Contexts.” American Political Science Review 104 (2010): 111–​ 133; Mondak and Hibbing, “Personality and Public Opinion.” 45 Gerber et al., “Personality and Political Attitudes.” 46 Jesse Graham, Jonathan Haidt, and Brian A.  Nosek, “Liberals and Conservatives Rely on Different Sets of Moral Foundations,” Journal of Personality and Social Psychology 96 (2009): 1029–​1046. 47 For a review of this work, see Jonathan Haidt, The Righteous Mind:  Why Good People Are Divided by Politics and Religion (New York: Pantheon, 2012). 48 Christopher M.  Federico, Christopher R.  Weber, Damla Ergun, and Corrie Hunt, “Mapping the Connections Between Politics and Morality:  The Multiple Sociopolitical Orientations Involved in Moral Intuition,” Political Psychology 34 (2013): 589–​610; Matthew Kugler, John T. Jost, and Sharareh Noorbaloochi, “Another Look at Moral Foundations Theory: Do Authoritarianism and Social Dominance Orientation Explain Liberal-​Conservative Differences in ‘Moral’ Intuitions?” Social Justice Research 27 (2014): 413–​431. 49 Schwartz, “Universals in the Content and Structure of Values”; Paul Goren, On Voter Competence (New York: Oxford University Press, 2012). 50 For additional review, see Frank Gonzales, John R. Hibbing, and Kevin B. Smith, “No Longer ‘Beyond our Scope’: The Biological and Non-​Conscious Underpinnings of Public Opinion,” Chapter  10 of this volume; Aleksander Ksiazkiewicz and Amanda Friesen, “Genes and Politics,” in Handbook of Biology and Politics, ed. Steven A. Peterson and Albert Somit (Cheltenham, UK: Elgar Publishing, 2017). 51 John R. Alford, Carolyn L. Funk, and John R. Hibbing, “Are Political Orientations Genetically Transmitted?” American Political Science Review 99 (2005):  153–​ 167; John R.  Hibbing, Kevin B.  Smith, and John R.  Alford, Predisposed:  Liberals, Conservatives, and the Biology of Political Differences (New York: Routledge, 2014); Gonzales, Hibbing, and Smith, Chapter 10 of this volume. 52 See Gonzales, Hibbing, and Smith, Chapter 10 of this volume. 53 Ibid. 54 Peter K. Hatemi et al., “Is There a ‘Party’ in Your Genes?” Political Research Quarterly 62 (2009): 584–​600. 55 For an example of such criticisms, see Evan Charney, “Genes and Ideologies,” Perspectives on Politics 6 (2008):  299–​ 319; for responses, see John R.  Alford, Carolyn L.  Funk, and John R.  Hibbing, “Beyond Liberals and Conservatives to Political Genotypes and Phenotypes,” Perspectives on Politics 6 (2008): 321–​328; and Ksiazkiewicz and Friesen, “Genes and Politics.” 56 On mediation models, see Kevin Smith et  al., “Biology, Ideology, and Epistemology: How Do We Know Political Attitudes Are Inherited and Why Should We Care?” American Journal of Political Science 56 (2012): 17–​33; on pleiotropy, see Peter K. Hatemi and Brad Verhulst, “Political Attitudes Develop Independently of Personality Traits,” PLOS ONE 10 (2015): e0118106. For a review, see Ksiazkiewicz and Friesen, “Genes and Politics.” 57 Converse, “The Nature of Belief Systems.” 58 Converse, “The Nature of Belief Systems”; Malka et al., “Are Cultural and Economic Conservatism Positively Correlated?”; Kinder and Kalmoe, Neither Liberal Nor Conservative; Zaller, The Nature and Origins of Mass Opinion.

Ideology and Public Opinion  97 59 Sniderman and Bullock, “A Consistency Theory of Public Opinion.” On the process by which ideological packages are transmitted by intellectuals and activists, see Noel, Political Ideologies and Political Parties. 60 Zaller, The Nature and Origins of Mass Opinion. 61 Converse, “The Nature of Belief Systems”; Kinder and Kalmoe, Neither Liberal Nor Conservative. 62 Kinder and Kalmoe, Neither Liberal Nor Conservative; see also Erikson and Tedin, American Public Opinion; Stimson, Tides of Consent. 63 Zaller, The Nature and Origins of Mass Opinion. 64 Converse, “The Nature of Belief Systems.” 65 Christopher H.  Achen, “Mass Political Attitudes and the Survey Response,” American Political Science Review 69 (1975): 1218–​1223. 66 For example, Norman H.  Nie, Sidney Verba, and John Petrocik, The Changing American Voter (Cambridge, MA: Harvard University Press, 1976). 67 For an up-​ to-​ date treatment, see Kinder and Kalmoe, Neither Liberal Nor Conservative; see also Michael X. Delli Carpini and Scott Keeter, What Americans Know About Politics and Why it Matters (New Haven, CT:  Yale University Press, 1996); Zaller, The Nature and Origins of Mass Opinion. 68 Ellis and Stimson, Ideology in America; Stimson, Tides of Consent. 69 Delli Carpini and Keeter, What Americans Know About Politics; Martin Gilens, “Citizen Competence and Democratic Governance,” Chapter  2 of this volume; Kinder and Kalmoe, Neither Liberal Nor Conservative. 70 Fiske et al., “On the Varieties and Utilities of Political Expertise.” 71 Converse, “The Nature of Belief Systems.” 72 See Delli Carpini and Keeter, What Americans Know About Politics. 73 Stimson, Tides of Consent. 74 Federico and Schneider, “Political Expertise”; Christopher M. Federico, “Expertise, Evaluative Motivation, and the Structure of Citizens’ Ideological Commitments,” Political Psychology 28 (2007): 535–​562. 75 Jost et al., “Political Ideology.” 76 Goren, On Voter Competence. Here, “values” should be distinguished from ideologies in that the latter are usually more abstract and encompassing. In this respect, ideologies are usually thought of as tying together multiple values into a larger posture. 77 Berelson et al., Voting; Campbell et al., The American Voter. 78 Robert E. Lane, Political Ideology (New York: Free Press, 1962). There is one caveat worth mentioning in the context of this argument: it is not clear that all individuals who fail to show an understanding of the discursive content associated with the left–right distinction are in fact using their “own” ideologies. If this were the case, we would observe substantial attitude stability even in the absence of left–right understanding. However, such stability is rarely observed, and it tends to be higher among those who are politically well informed; on this point, see Converse, “The Nature of Belief Systems,” and Zaller, The Nature and Origins of Mass Opinion. 79 Federico and Malka, “The Contingent, Contextual Nature”; Christopher M.  Federico, Christopher D.  Johnston, and Howard G.  Lavine, “Context, Engagement, and the (Multiple) Functions of Negativity Bias,” Behavioral and Brain Sciences, 37 (2014):  311–​312; Christopher D.  Johnston, Howard G.  Lavine, and Christopher M. Federico, Open Versus Closed: Personality, Identity, and the Politics of Redistribution (Cambridge: Cambridge University Press, 2017).

98  Christopher M. Federico 80 Christopher M. Federico, Emily L. Fisher, and Grace Deason, “Political Expertise and the Link Between the Authoritarian Predisposition and Conservatism,” Public Opinion Quarterly 75 (2011):  686–​708; Christopher M.  Federico and Paul Goren, “Motivated Social Cognition and Ideology: Is Attention to Elite Discourse a Prerequisite for Epistemically Motivated Political Affinities?” in Social and Psychological Bases of Ideology and System Justification, ed. John T.  Jost, Aaron C.  Kay, and Hulda Thorisdottir (New  York:  Oxford University Press, 2009); Christopher M. Federico, Corrie V. Hunt, and Damla Ergun, “Political Expertise, Social Worldviews, and Ideology: Translating ‘Competitive Jungles’ and ‘Dangerous Worlds’ Into Ideological Reality,” Social Justice Research 22 (2009):  259–​ 279; Johnston et al., Open Versus Closed. 81 Christopher M.  Federico and Michal Reifen-​Tagar, “Zeroing in on the Right: Education and the Partisan Expression of Authoritarianism,” Political Behavior 36 (2014): 581–​603; Johnston et al., Open Versus Closed. 82 For a review, see Federico and Malka, “The Contingent, Contextual Nature.” 83 Federico and Malka, “The Contingent, Contextual Nature”; Johnston et  al., Open Versus Closed. 84 Johnston et al., Open Versus Closed. 85 Ariel Malka, Christopher J.  Soto, Michael Inzlicht, and Yphtach Lelkes, “Do Needs for Security and Certainty Predict Cultural and Economic Conservatism? A Cross-​National Analysis,” Journal of Personality and Social Psychology 106 (2014), 1031–​1051. 86 Converse, “The Nature of Belief Systems,” p. 206.

Chapter 4

Affective Polarization or Hostility Across the Party Divide An Overview Shanto Iyengar

The symptoms of party polarization are by now all too familiar: intransigent policy positions, unwillingness to compromise, and intensified political conflict. In the United States, research into the extent of party polarization has focused mainly on the question of ideological divergence. Have the two major political parties moved far apart on the major issues of the day? By this standard, there is clear evidence that party elites have polarized over the past five decades.1 However, does this trend apply to partisan voters? Here the evidence is far from clear. Some scholars present data showing that the median citizen has remained centrist on most issues despite the gravitation of party elites to the ideological extremes.2 Others contest this description of the masses, citing a gradual decline in the number of ideological moderates and a near doubling of the average distance between the ideological self-​placement of non-​activist Democrats and Republicans between 1972 and 2004.3 Ideological disagreement is but one way of defining partisan polarization. An alternative definition considers mass polarization as the extent to which partisans view each other as a disliked out-​group. In the U.S., partisanship is about identifying with the “Democrat” group or the “Republican” group.4 When people identify with a political party, they instinctively categorize the world into an in-​group (their own party), and an out-​group (the opposing party).5 A vast literature in social psychology demonstrates that any such in-​group versus out-​ group distinction, even one based on the most trivial of shared characteristics, triggers both positive feelings for the in-​group, and negative evaluations of the out-​group.6 The more salient the group to the sense of personal identity, the stronger these intergroup divisions.7 For Americans, partisanship is a particularly salient and powerful identity for several reasons. First, it develops at a young age, and rarely changes over the life cycle, notwithstanding significant shifts in personal circumstances.8 Second, political campaigns—​the formal occasions for expressing citizens’ partisan identity—​recur frequently, and last for many months (or even years). The frequency and duration of campaigns means that individuals constantly receive partisan cues from elites. It is not surprising, therefore, that ordinary Americans view themselves in terms of their partisan affiliation. As described

100  Shanto Iyengar

below, in recent decades this sense of group identity has elicited strong feelings of hostility toward political opponents, a phenomenon political scientists refer to as affective polarization.9

Affective Polarization: The Evidence There are now multiple studies documenting the extent to which partisans treat each other as a disliked out-​group. The evidence of out-​party animus derives primarily from survey data, but also includes behavioral indicators of discrimination and implicit or sub-​conscious markers of partisan prejudice.

Self-​R eported Partisan  Affect Survey data on respondents’ feelings toward the parties and their followers are the most widely used measure of affective polarization. The one item with the greatest longevity—​dating back to the 1970s—​is the “feeling thermometer” question. Introduced into the American National Election Studies, the measure has since been widely adopted by other survey organizations.10 The question asks respondents to rate the two parties, or “Democrats” and “Republicans”, on a scale ranging from (0)  indicating coldness to (100) indicating warmth. Since the measure targets attitude objects representing both parties, it is possible to track the difference between in-​group and out-​group affect over the past 40 years.11 As widely reported in the scholarly literature and popular press, the trends in the feeling-thermometer scores reveal substantially increased affective polarization over time. As shown in Figure 4.1, which plots the in-​party and out-​party thermometer scores in the ANES time series, the gap between the in-​ and out-​party thermometer scores steadily increased from around 23 degrees in 1978 to 41 in 2016.12 As the figure makes clear, virtually all the increase in affective polarization has occurred because of increased animus toward the opposing party. Warm feelings for one’s own party have remained stable across the entire period. Stronger hostility for the out-​party is a recent, but rapidly escalating trend that began at the turn of the century. Figure 4.2 shows that while the percentage of partisans who rated the out-​party between 1 and 49 on the thermometer has increased steadily since the 1980s, the share of partisans expressing intense negativity for the out-​party (ratings of 0) remained quite small until 2000. Post-​ 2000, the size of this group has surged dramatically—​from 8 percent in 2000 to 21 percent in 2016. Thus, the first two decades of the twenty-​first century represent an acute era of polarization, in which what was only mild dislike for political opponents now appears to be a deeper form of animus.13 A parallel pattern reappears when we track respondents’ feelings toward the presidential candidates. Until about 2000, partisans reported only ambivalent feelings toward the opposing party’s nominee (feeling thermometer scores of

80

+ + + +

Mean therm rating

+ +

+

+

+

+ +

+

+

+

+

+

60

×

×

× ×

40

×

× ×

× ×

×

×

× ×

× ×

20 1980

1990

2000

Democrat-In party

+ Republican-In party

2010

Democrat-Out party

× Republican-Out party

Figure 4.1 ANES Party Feeling Thermometers (1976–​2016)

Percent

0.4

0.2

0.0 1970

1980

1990

0 < Out party therm < 50

2000

2010

Out party therm = 0

Figure 4.2 Out-​Party Animus in Feeling Thermometer Scores

×

102  Shanto Iyengar

around 40).14 However, beginning in 2004, feelings toward the out-​party candidate turn colder, with average thermometer scores dropping to around 15 in 2016. As in the case of the party thermometers, partisans’ feelings toward their own party nominee are unchanged; the strengthened polarization occurs because of increased hostility toward the opposing party nominee. The feeling thermometer data show clearly that the party divide elicits affective polarization. It is important, however, to place the findings on partisan affect in some context. How does partisanship compare with other salient cleavages as a source of group polarization? Fortunately, the feeling thermometers have been applied to multiple groups making it possible to compare in-​group versus out-​group evaluations based on party with evaluations based on race, religion, region, and other relevant groupings. The comparisons reveal that party is easily the most affectively laden group divide in the U.S. Social out-​groups including Muslims, Hindus, atheists, Latinos, African-​Americans, gays, and poor people all elicit much warmer thermometer scores than the out-​party.15 This contrast between the party divide and socio-​cultural divides should alert us to a major limitation of self-​reported indicators of group affect. Survey responses are highly reactive and susceptible to intentional exaggeration/​suppression based on normative pressures. In the case of race, religion, gender, and other social divides, the expression of animus toward out-​groups is tempered by strong social norms.16 Most individuals are prone to comply with applicable norms when asked sensitive questions. In the case of the party divide, however, there are no corresponding pressures to moderate disapproval of political opponents. If anything, the rhetoric and actions of political leaders convey to their followers that hostility directed at the opposition is not only acceptable, but also appropriate.

Implicit Measures The normative pressures facing survey respondents make it difficult to establish a fair comparison of social with political divides as a basis for out-​group animus. Fortunately, psychologists have developed an array of implicit or sub-​ conscious measures of group prejudice. These implicit measures provide a more valid comparison of the bases for prejudice because they are much harder to manipulate than explicit self-​reports and less susceptible to impression management or political correctness.17 Iyengar and Westwood18 developed a Party Implicit Association Test (IAT) (based on the brief version of the Race IAT) to document unconscious partisan bias.19 Their results showed ingrained implicit bias with approximately 70  percent of Democrats and Republicans showing a bias in favor of their party. Interestingly, implicit bias proved less extensive than explicit bias as measured through survey questions; 91 percent of Republicans and 75 percent of Democrats in the same study explicitly evaluated their party more favorably. This is an important reversal from the case of race or religion where

Affective Polarization or Hostility  103

social norms restrain the expression of conscious hostility toward out-​groups resulting in higher levels of implicit over explicit prejudice. To place the results from their party IAT in context, Iyengar and Westwood also administered the race IAT. Surprisingly, relative to implicit racial bias, implicit partisan bias proved more widespread. The difference in the D-​score—​ the operational indicator of implicit bias across the party divide—​was 0.50, while the corresponding difference in implicit racial bias across the racial divide was only 0.18.20 Thus, prejudice toward the out-​party exceeded comparable bias directed at the racial out-​group by more than 150 percent!

Indicators of Social Distance An even more unobtrusive measure of partisan affect is social distance: the extent to which individuals feel comfortable interacting with out-​ group members in a variety of different settings. In 2018, the media highlighted several high-​profile instances of social “shunning” directed at political opponents. Opponents of Supreme Court nominee Brett Kavanaugh heckled Senator Ted Cruz and his wife at a posh Washington restaurant. Harvard Professor Alan Dershowitz, a prominent supporter of President Trump, complained to The New York Times—​“My liberal friends have stopped inviting me for dinner.” More generally, the argument is that if partisans take their political affiliation seriously, they should be averse to entering into close interpersonal relations with their opponents. The most vivid evidence of increased social distance across the party divide concerns inter-​party marriage. In the early 1960s, the percentage of partisans expressing concern over the prospect of their son or daughter marrying someone from the opposition party was in the single digits, but some 45 years later, it had risen to more than 25 percent.21 Among Republicans, one-​ half expressed dismay at the prospect of their offspring marrying a Democrat. Today, the party divisions and resulting out-​party animus are sufficiently strong to motivate partisans to associate with like-​minded others. Of course, survey data on social distance is limited to questions of hypothetical social interactions across the party divide. More compelling evidence of increased social distance based on party affiliation comes from online dating sites and other available sources of “big data,” including national voter files indicating that the party cue does in fact influence the decision to enter into interpersonal relations. The level of spousal partisan agreement has increased significantly over the period marked by heightened polarization. In a longitudinal analysis spanning 1965 to 2015, Iyengar, Konitzer, and Tedin find that spousal agreement on partisanship moved from 73 to 82  percent, while disagreement fell from 13 to 6 percent.22 Since the 1965 sample of spouses had been married for decades, they had many opportunities to persuade their partner, thus inflating the observed level of agreement. When the researchers limited the focus to younger couples, they found a more impressive shift in spousal agreement; among recently married couples in 1973, spousal agreement

104  Shanto Iyengar

registered at 54.3 percent. For the comparable group of recently married couples in the 2014 national voter file, spousal partisan agreement reached 73.9 percent. This is an increase of 36 percent in partisan agreement among couples who have had little opportunity to persuade each other. Online dating sites are a rich source of data on the politics underlying interpersonal attraction. Huber and Malhotra leverage data from a major dating website where they gained access to both the daters’ personal profiles as well as their messaging behavior.23 They found that partisan agreement increases the likelihood of two people exchanging messages by 10 percent. To put that difference in perspective, the comparable difference for couples matched on socioeconomic status (using education as the indicator) was 11 percent. Thus, partisanship appears to be just as relevant as social standing in the process of selecting a romantic partner. The authors replicated this result in the context of a survey experiment where they demonstrated that after exposure to a dating profile that includes the target individual’s political ideology, ideological agreement significantly enhanced the participant’s interest in dating the target individual. The fact that individuals date and marry co-​partisans does not necessarily mean that politics was the basis for their choice. Agreement on partisanship may be a byproduct of spousal selection on some other attribute correlated with partisan identity, such as economic status. While some researchers argue that partisan agreement among couples is in fact “induced” or accidental,24 others provide evidence in favor of an active selection model in which the political affiliation of the prospective partner is the point of attraction. Huber and Malhotra, for instance, show that ideology and partisanship both predict reciprocal online messaging on dating sites even after controlling for alternative bases of spousal attraction.25 Iyengar et al. present similar results, showing that spousal agreement in the current era is more attributable to selection based on politics than alternative mechanisms including induced selection, the homogeneity of marriage markets, and agreement due to one spouse gradually persuading the other.26 Dating and marriage both entail long-​term and more intimate relationships. Does politics also impede the initiation of more casual friendships? Surveys by the Pew Research Center suggest that it does. About 64 percent of Democrats and 55 percent of Republicans say they have “just a few” or “no” close friends who are from the other political party.27 Similarly, Huber and Malhotra report that discordant partisanship decreases people’s likelihood to being friends with someone even when they are not seeking a romantic relationship.28 Thus, partisanship appears to act as a litmus test even at the level of casual social encounters.

Behavioral Evidence of Partisan Bias Survey measures of partisan affect are subject to several limitations, since people can answer questions in ways that do not reveal their true feelings. In

Affective Polarization or Hostility  105

response, scholars have turned to behavioral manifestations of partisan animus in both lab and naturalistic settings. Iyengar and Westwood29 and Carlin and Love30 have used economic games as a platform for documenting the extent to which partisans are willing to endow or withhold financial rewards from players who either share or do not share their partisan affiliation. In the trust game, the researcher gives Player 1 an initial endowment ($10) and instructs her that she is free to give all, some, or none to Player 2 (said to be a member of a designated group). Player 1 is further informed that any amount she donates to Player 2 will be tripled by the researcher, and that Player 2 is free (although under no obligation to do so) to transfer an amount back to Player 1. The dictator game is an abbreviated version of the trust game in which there is no opportunity for Player 2 to return funds to Player 1 and where the researcher does not add to the funds transferred. Since there is no opportunity for Player 1 to observe the strategy of Player 2, variation in the amount Player 1 allocates to the different categories represented by Player 2 in the dictator game is attributable only to group dislike and prejudice. As Fershtman and Gneezy put it, “any transfer distribution differences in the dictator game must be due to a taste for discrimination.”31 The trust and dictator games provide a consequential test of out-​group bias, for they assess the extent to which participants are willing to transfer money they would otherwise receive themselves to co-​partisans while simultaneously withholding money from opposing partisans. For both the trust game and the dictator game, partisan bias emerges as the difference between the amount allocated to co-​partisans and opposing partisans. The results reported by Iyengar and Westwood show the expected pattern:  co-​partisans consistently receive a bonus while opposing partisans are subject to a financial penalty.32 As in the case of implicit bias, the effects of party affiliation on donations exceeded the effects of ethnicity. In fact, the effects of racial similarity proved negligible and not significant—​co-​ethnics were treated more generously (by eight cents) in the dictator game, but incurred a loss (seven cents) in the trust game. As in the case of the survey data, social norms appear to suppress racial discrimination in the trust and dictator games. Iyengar and Westwood shed further light on the extent of affective polarization by comparing the effects of partisan and racial cues on non-​political judgments.33 In one study, they asked participants to select one of two candidates for a college scholarship. The candidates (both high school students) had similar academic credentials, but differed in their ethnicity (White or African American) or partisanship (Democrat or Republican). The results indicated little racial bias; Whites, in fact, preferred the African American applicant (55.8 percent). In contrast, partisan favoritism was widespread; 79.2 percent of Democrats picked the Democratic applicant and 80  percent of Republicans picked the Republican applicant. These results held even when the out-​partisan candidate had a significantly higher GPA (4.0 v. 3.5); in fact, the probability of a partisan selecting the more qualified out-​party candidate was never above 30 percent.

106  Shanto Iyengar

The scholarship study showed that partisan cues spilled over into non-​ political arenas. As noted above, economic decision-​making provides an especially interesting test of affective spillover because the expression of in-​party favoritism or out-​party prejudice often entails monetary costs. Michelitch, for instance, found that Ghanaian taxi drivers accepted lower prices from co-​ partisans and demanded higher prices from counter-​partisans.34 Specifically, taxi drivers demanded 16  percent higher fares from passengers whose ethnicity and party affiliation differed from their own. In the case of passengers who differed only in their ethnicity, the fare penalty was smaller (6 percent). McConnell et  al. conducted a field experiment in the U.S.  in which they provided people the opportunity to buy a heavily discounted gift card.35 They assigned some buyers to conditions in which they learned that the seller was either a co-​partisan or counter-​partisan. They found no evidence of out-​group animus; the purchasing rate remained stable across same party and opposite party sellers. However, interacting with a co-​partisan seller nearly doubled the purchasing price of the gift card. The effects were even larger among strong partisans. Panagopoulos et al., on the other hand, found evidence of out-​group animus:  15 to 20  percent of participants in their study were less willing to accept a gift card from a company that gives PAC (Political Action Committee) donations to the opposing party.36 In addition to product markets, partisanship can distort labor markets. Using an audit design, Gift and Gift mailed out résumés signaling job applicants’ partisan affiliation in a heavily Democratic area and a heavily Republican area.37 They found that in the Democratic county, Democratic résumés were 2.4 percentage points more likely to receive a callback than Republican résumés; the corresponding partisan preference for Republican résumés in the Republican county was 5.6  percentage points. Whereas Gift and Gift examine employer preferences, McConnell et al. examine the other side of the labor market and study how partisanship affects employee behavior.38 The researchers hired workers to complete an online editing task and subtly signaled the partisan identification of the employer. Unlike Gift and Gift, they mainly find evidence of in-​group favoritism as opposed to out-​group prejudice. The only significant differences occurred between the co-​partisan condition and the control group. People exhibited a willingness to accept lower compensation (by 6.5 percent) from a partisan congruent employer. At the same time, they performed lower-​ quality work and exhibited less effort. Although the mechanism for this performance deficit is unclear, one possibility is that they perceive the employer to be of higher quality and therefore less likely to make copy-​editing mistakes. In summary, evidence from self-​reported feelings toward the parties, sub-​ conscious partisan prejudice, increased social distance based on political affiliation, and multiple instances of behavioral discrimination against opposing partisans all converge on the finding of intensified party polarization in the U.S. We turn next to consider the factors that may have contributed to this phenomenon.

Affective Polarization or Hostility  107

Possible Explanations The period over which mass polarization has intensified (from 1980 to today) coincides with several major changes in American society and politics, including changes in the media environment, increased social homophily, and partisan sorting or greater differentiation between Democrats and Republicans. In addition to independently inducing hostility toward opponents, each of these factors reinforces the others, further contributing to the rise of affective polarization. First, in the last 50  years, the percentage of “sorted” partisans, that is, partisans who identify with the party most closely reflecting their ideology, has steadily increased.39 When most Democrats (Republicans) are also liberals (conservatives), they are less likely to encounter conflicting political ideas and identities,40 and are more likely see non-​identifiers as socially distant. Sorting likely leads people to perceive both opposing partisans and co-​partisans as more extreme than they really are, with misperceptions being more acute for opposing partisans.41 As partisan and ideological identities have come into alignment, other salient social identities, including race and religion, also converged with partisanship. Democrats are increasingly the party of women, non-​Whites, professionals, and residents of urban areas, while Republican voters are disproportionately older white men, evangelical Christians, and residents of rural areas. This decline of crosscutting identities is at the root of affective polarization according to Mason.42 She has shown that those with consistent partisan and ideological identities become more hostile toward the out-​party without necessarily changing their ideological positions, and those that have aligned religious, racial, and partisan identities react more emotionally to information that threatens their partisan identities or issue stances. In essence, sorting has made it much easier for partisans to make generalized inferences about the opposing side, even if those inferences are inaccurate. However, in one of the few efforts to estimate the impact of ideological sorting on the partisan affect, Lelkes finds that changes over time in affective polarization have occurred at the same rate among voters who are either ideologically sorted or unsorted.43 A second potential cause of strengthened polarization is social homophily. We have described studies documenting strengthened processes of socialization by which families come to agree on their partisan loyalties. Family agreement creates an interpersonal echo chamber that facilitates polarization. When family members identify with the same party, they also express more extreme positions on the issues and harbor hostile views toward their opponents. In the case of a 2015 national survey of married couples, respondents evaluated the presidential candidates Hillary Clinton and Donald Trump (using the ANES 100 point feeling thermometer). Among spouses who agreed on their party identification, the average difference between the in-​and out-​party candidate

108  Shanto Iyengar

thermometer score was 59 points (70 vs. 11 degrees). Among the few pairs consisting of spouses with divergent loyalties (Democrat–​Republican pairings), this margin of difference fell by more than 30 degrees. Partisan agreement within the family strengthens polarization.44 Given the importance of family socialization to the development of partisan attitudes, the rate at which any given society undergoes polarization will be conditional on the extent to which partisans grow up in homogeneous environments. Recent simulations by Klofstad, McDermott, and Hatemi suggest that spousal agreement induces ideological polarization within the family fairly rapidly, with most of the increased polarization occurring as early as the fifth generation.45 We would similarly expect generations to move increasingly apart on their feelings toward the opposing party to the extent family members share these sentiments. Finally, a third potential contributor to affective polarization is technology. The revolution in information technology has empowered consumers to encounter news on their own terms. The availability of 24-​hour cable news channels provided partisans with their first real opportunity to obtain news from like-​minded sources (Fox News first for Republicans, and MSNBC later for Democrats). The development of the Internet provided a much wider range of media choices, which greatly facilitated partisans’ ability to obtain political information and commentary consistent with their leanings. In a break with the dominant paradigm of non-​partisan journalism, a growing number of outlets, motivated in part by the commercial success of the Fox News network, offered reporting in varying guises of partisan commentary. Many of these online partisan outlets depict the opposing party in harsh terms,46 and focus disproportionately on out-​party scandals (real or imagined). The political blogosphere, with hundreds of players providing news and analysis—​often vitriolic—​developed rapidly as a partisan platform, with very little cross-​party exposure.47 The creation of vast online social networks permitted extensive recirculation of news reports, even to those not particularly motivated to seek out news. Several scholars have thus singled out the technologically enhanced media environment and partisans’ ability to encounter “friendly” information providers as an especially influential agent of polarization.48 While there are good reasons to believe that the new media environment has contributed to the growth in partisan animus, by facilitating access to partisan news and commentary, it is possible that enhanced consumer choice also sets in motion processes that weaken polarization. As media platforms have multiplied, consumers gain access not only to more news providers, but also to entertainment providers. The availability of entertainment programming on demand enables some to drop out of the political arena entirely.49 Thus, the net impact of the increased empowerment of consumers is unclear. In fact, despite the myriad changes in the media environment, the evidence to date demonstrating that news consumption exacerbates polarization is less than unequivocal. While experimental studies of online browsing behavior

Affective Polarization or Hostility  109

confirm the tendency of partisans to self-​select into distinct audiences,50 more generalizable real-​world studies find few traces of audience segregation. In their pioneering analysis of Americans’ web-​browsing behavior (conducted in 2008), Gentzkow and Shapiro found that online audiences were only slightly more segregated than audiences for network or cable news.51 They concluded, “Internet news consumers with homogeneous news diets are rare. These findings may mitigate concerns … that the Internet will increase ideological polarization and threaten democracy” (p. 1831). However, more recent work—​ also based on large-​scale tracking of online-browsing behavior—​suggests that the segregation of news audiences is increasing. A  2013 study showed that although most people relied on ideologically diverse online sources such as web aggregators, audience segregation tended to increase among individuals who used search engines to locate news stories and among social media users who encountered links in their news feed.52 Both these pathways to news exposure feature personalized algorithms, making it more likely that individuals encounter information consistent with their political loyalties. In the case of Facebook, now a major source of news, most individuals find themselves in politically homogeneous networks, increasing the likelihood of exposure to polarizing messages.53 To the extent partisans do gravitate to like-​minded news providers, has the diffusion of high-​speed Internet facilitated this behavior? Here, too, the evidence is mixed. In those parts of the country where broadband is more available, traffic to partisan news sites is greater.54 Moreover, Lelkes et al. go on to show that broadband diffusion has strengthened partisan affect. Moving from a county with the fewest number of broadband providers to a county with the highest number increased affective polarization by roughly 0.07 (an effect roughly half as large as the effect of partisans’ political interest). On the other hand, Boxell, Gentzkow, and Shapiro demonstrate that affective polarization has increased the most among those least likely to use social media and the Internet.55 Given these inconsistent results, it is too early to conclude that Internet usage (and the availability of a wider array of information) plays a causal role in the growth of affective polarization.

Conclusion The phenomenon of affective polarization—​the tendency of Democrats and Republicans to treat each other as a stigmatized out-​group—​has far-​reaching consequences for the behavior of politicians. For one thing, it creates incentives for politicians to use inflammatory rhetoric and demonize their opponents. The most frequent and enthusiastic chant at Republican rallies in 2016 was “lock her up.” Illegal immigrants, in Trump’s words, were “rapists and drug dealers.” Earlier, during the debate over the passage of the Affordable Care Act, some Republicans likened the mandatory insurance requirement in the law to the forced deportation of Jews by the Nazis. In response, the liberal commentator

110  Shanto Iyengar

Keith Olbermann declared that Republicans’ opposition to the law was tantamount to racism. Symptomatic of the pressures facing politicians to demonstrate their party colors, a study found that taunting of the opposition party is the most frequent theme in congressional press releases.56 At the level of electoral politics, heightened polarization has made it almost impossible for partisans to abandon their party’s candidates, no matter their limitations. The release of the Access Hollywood tape—​in which Trump made crude references to his willingness and ability to grope women—​would surely have ended the candidacy of any presidential candidate in any election cycle from the 1980s or 1990s. Yet the impact on Donald Trump’s poll numbers was miniscule. And in Alabama, in the 2017 Senate election, evidence of Republican candidate Roy Moore’s inappropriate relations with under-​age women hardly caused concern among Republican voters, a mere seven percent of whom defected. Partisans have become so committed to their party that scholars have had to update the standard finding of public opinion research—​voter ignorance of current events. Today, partisans are not only uninformed, but also misinformed, and deliberately misinformed.57 Partisan voters have become reliable team players whose loyalty provides politicians considerable leeway to guide and lead public opinion. As a result, when candidates make claims that are false, there is the very real possibility of voter manipulation. Well before he became a presidential candidate, Donald Trump was the principal sponsor of the conspiracy-​oriented “birther” theory concerning former president Barack Obama’s place of birth and citizenship. Since taking office, Trump has continued to show little respect for facts and evidence. He claimed that extensive voter fraud was responsible for his loss in the popular vote and that charges of possible collusion between his campaign and the Russian government amount to a “hoax.” What’s more, he frequently attacks the credibility of the American press by referring to stories critical of his leadership as “fake news.” Trump’s rhetoric has persuaded Republicans, many of whom believe in Trump’s false claims. All told, intensified affective polarization portends serious repercussions, especially during times of political turmoil. There are multiple parallels between Watergate and the current era, yet polarization has fundamentally altered the political dynamics of scandal. Investigative news reports that brought to light the cover-​up in the Nixon White House became widely accepted as credible evidence of official wrongdoing. The media spotlight resulted in a significant erosion of President Nixon’s approval among both Democrats and Republicans. In contrast, the multiple investigations swirling around the Trump administration have, to date, done little to undermine his standing among Republicans. Partisans’ willingness to ignore information that challenges their sense of political identity is disturbing and undermines the ability of the press to act as the “fourth branch of government.” President Trump famously claimed that he could stand in the middle of Fifth Avenue and shoot somebody at no cost to his electoral support. We can only hope that he is mistaken.

Affective Polarization or Hostility  111

Notes 1 Richard Fleisher and Jon R. Bond, 2001. “Evidence of Increasing Polarization among Ordinary Citizens,” in American Political Parties: Decline or Resurgence? edited by Jeffrey E. Cohen, Richard Fleisher, and Paul Kantor, 55–​77. Washington, DC: CQ Press; Marc Hetherington, 2002. “Resurgent Mass Partisanship:  The Role of Elite Polarization.” American Political Science Review 95: 619–​31; Nolan McCarty, Keith Poole, and Howard Rosenthal, 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, MA: MIT Press. 2 Morris Fiorina, Samuel Abrams, and Jeremy Pope, 2008. “Polarization in the American Public: Misconceptions and Misreadings.” Journal of Politics 70: 556–​60. 3 Alan I. Abramowitz and Kyle L. Saunders, 2008. “Is Polarization a Myth?” Journal of Politics 70: 542–​55. 4 Donald Green, Bradley Palmquist, and Eric Schickler, 2002. Partisan Hearts and Minds:  Political Parties and the Social Identities of Voters. New Haven, CT:  Yale University Press; Leonie Huddy, Lilliana Mason, and Lene Aaroe, 2015. “Expressive Partisanship:  Campaign Involvement, Political Emotion, and Partisan Identity.” American Political Science Review 109: 1–​17. 5 See Henri Tajfel and John C.  Turner, 1979. “An Integrative Theory of Intergroup Conflict,” in The Social Psychology of Intergroup Relations, W. G. Austin and Stephen Worchel (Eds), 33–​47. Monterey, CA: Brooks/​Cole. 6 See, for instance, M.  G. Billig and Henri Tajfel, 1973. “Social Categorization and Similarity in Intergroup Behavior.” European Journal of Social Psychology 3: 27–​52. 7 Samuel Gaertner, John Dovidio, Phyllis Anastasio, Betty Bachman, and Mary Rust, 1993. “The Common Ingroup Identity Model: Recategorization and the Reduction of Intergroup Bias.” European Review of Social Psychology 4: 1–​26. 8 David O.  Sears, 1975. “Political Socialization,” in Handbook of Political Science, Fred I.  Greenstein and Nelson W.  Polsby (Eds.), Vol. 2, 93–​153. Reading, MA: Addison-​Wesley. 9 For a review of this literature, see Shanto Iyengar, Yphtach Lelkes, Matthew Levendusky, Neil Malhotra, and Sean Westwood, 2018. “The Origin and Consequences of Affective Polarization in the United States.” Annual Review of Political Science, forthcoming. 10 Herbert F.  Weisberg and Jerrold G.  Rusk, 1970. “Dimensions of Candidate Evaluation.” American Political Science Review 64: 1167–​85. 11 Following conventional practice, scholars of affective polarization measure party identification using the standard ANES seven-​point question ranging from strongly Republican to strongly Democratic. Most scholars classify independent “leaners” as partisans and exclude pure independents from consideration (this group represents less than 15 percent of the electorate in the 2016 ANES). Democratic and Republican evaluations of their own side constitute the measure of in-​group affect while partisans’ evaluations of their opponents provide the measure of out-​group affect. 12 See Shanto Iyengar, Gaurav Sood, and Yphtach Lelkes, 2012. “Affect, not Ideology: A Social Identity Perspective on Polarization.” Public Opinion Quarterly 76(3): 405–​ 31; Shanto Iyengar and Masha T. Krupenkin, 2018. “The Strengthening of Partisan Affect.” Political Psychology 39: 201–​18. 13 The ANES survey is administered through in-​ person interviews and online questionnaires. The level of expressed hostility toward the opposing party tends to

112  Shanto Iyengar be greater in the anonymous setting of the online survey. When faced with an interviewer in their home, respondents assign less extreme thermometer scores to their opponents (see Shanto Iyengar and Masha T. Krupenkin, 2018. “The Strengthening of Partisan Affect.” Political Psychology 39: 201–​18). 14 The fact that the thermometer scores for the individual candidates are more positive than the scores given the parties may represent a “person positivity” bias (David O. Sears, 1983. “The Person-​Positivity Bias.” Journal of Personality and Social Psychology 44:  233–​50). Psychologists have demonstrated that attitudes toward specific individuals are typically more positive than attitudes toward the group represented by those individuals. 15 See Shanto, Iyengar and Masha T. Krupenkin, 2018. “The Strengthening of Partisan Affect.” Political Psychology 39: 201–​18. 16 Eleanor E. Maccoby and Nathan Maccoby, 1954. “The Interview: A Tool of Social Science,” in Gardner Lindzey (Ed.) The Handbook of Social Psychology, 449–​87. Cambridge, MA: Addison-​Wesley. 17 Guy A.  Boysen, David L.  Vogel, and Stephanie Madon, 2006. “A Public Versus Private Administration of the Implicit Association Test.” European Journal of Social Psychology 36: 845–​56. 18 Shanto Iyengar and Sean J.  Westwood, 2015. “Fear and Loathing Across Party Lines.” American Journal of Political Science 59: 690–​707. 19 The PIAT measures the speed with which partisans associate positive and negative terms with images associated with the in-​and out-​party. When the target stimulus is the donkey, the Democratic mascot, the expectation is that Democratic respondents will more quickly categorize the mascot as “good” since they have come to associate “good” with Democrats. Conversely, Republican respondents should take more time to associate the Democratic mascot with “good.” Iyengar and Westwood constructed the partisan IAT using the standard set of good stimuli (Wonderful, Best, Superb, Excellent), and the standard set of bad stimuli (Terrible, Awful, Worst, Horrible) used in other IATs. 20 For an application of implicit measures to the study of partisanship, see also Alexander Theodoridis, 2017. “Me, Myself, and (I), (D), or (R)? Partisanship and Political Cognition through the Lens of Implicit Identity.” Journal of Politics 79: 1253–​67. 21 Shanto Iyengar, Gaurav Sood, and Yphtach Lelkes, 2012. “Affect, Not Ideology: A Social Identity Perspective on Polarization.” Public Opinion Quarterly 76: 405–​31. 22 Shanto Iyengar, Tobias Konitzer, and Kent L. Tedin, 2018. “The Home as a Political Fortress: Family Agreement in an Era of Polarization.” Journal of Politics, forthcoming. 23 Gregory A.  Huber and Neil Malhotra, 2017. “Political Homophily in Social Relationships: Evidence from Online Dating Behavior.” Journal of Politics 79: 269–​83. 24 See, for instance, Casey Klofstad, Rose McDermott, and Pete K. Hatemi, 2013. “The Dating Preference of Liberals and Conservatives.” Political Behavior 35: 519–​38. 25 Gregory A.  Huber and Neil Malhotra, 2017. “Political Homophily in Social Relationships: Evidence from Online Dating Behavior.” Journal of Politics 79: 269–​83. 26 Shanto Iyengar, Tobias Konitzer, and Kent Tedin, 2018. “The Home as a Political Fortress: Family Agreement in an Era of Polarization,” The Journal of Politics 80, no. 4: 1326–1338. 27 Pew Research Center, 2017. “The Partisan Divide on Political Values Grows Even Wider.” Technical Report. Available online at: http://​pewrsr.ch/​2z0qBnt.

Affective Polarization or Hostility  113 28 Gregory A.  Huber and Neil Malhotra, 2017. “Political Homophily in Social Relationships:  Evidence from Online Dating Behavior.” Journal of Politics 79: 269–​83. 29 Shanto Iyengar and Sean J. Westwood, 2015. “Fear and Loathing Across Party Lines.” American Journal of Political Science 59: 690–​707. 30 Robert Forsythe, Joel L. Horowitz, Nathan E. Savin, and Martin Sefton, 1994. “Fairness in Simple Bargaining Experiments.” Games and Economic Behavior 6: 47–​69. 31 Chaim Fershtman and Uri Gneezy, 2001. “Discrimination in a Segmented Society: An Experimental Approach.” Quarterly Journal of Economics 116: 351–​77. (354). 32 Shanto Iyengar and Sean J. Westwood, 2015. “Fear and Loathing Across Party Lines.” American Journal of Political Science 59: 690–​707. 33 Shanto Iyengar and Sean J. Westwood, 2015. “Fear and Loathing Across Party Lines.” American Journal of Political Science 59: 690–​707. 34 Kristin Michelitch, 2015. “Does Electoral Competition Exacerbate Interethnic or Inter-​Partisan Economic Discrimination? Evidence from a Market Price Bargaining Experiment in Ghana.” American Political Science Review 109: 43–​61. 35 Christopher McConnell, Neil Malhotra, Yotam Margalit, and Matthew S. Levendusky, 2018. “The Economic Consequences of Partisanship in a Polarized Era.” American Journal of Political Science 62: 5–​18. 36 Costas Panagopoulos, Donald P. Green, Jonathan Krasno, Michael Schwam-​Baird, Eric Moore, and Kyle Endres, 2016. “Risky Business:  Does Corporate Political Giving Affect Consumer Behavior?” Unpublished Manuscript, Fordham University. 37 Karen Gift and Thomas Gift, 2015. “Does Politics Influence Hiring? Evidence from a Randomized Experiment.” Political Behavior 37: 653–​75. 38 Christopher McConnell, Neil Malhotra, Yotam Margalit, and Matthew S. Levendusky, 2018. “The Economic Consequences of Partisanship in a Polarized Era.” American Journal of Political Science 62: 5–​18. 39 Matthew S. Levendusky, 2009. The Partisan Sort: How Liberals Became Democrats and Conservatives Became Republicans. Chicago, IL: University of Chicago Press. 40 Sonia Roccas and Marilynn B. Brewer, 2002. “Social Identity Complexity.” Personality and Social Psychology Review 6: 88–​106. 41 Matthew S.  Levendusky and Neil Malhotra, 2016. “Misperceptions of Partisan Polarization in the American Public.” Public Opinion Quarterly 80: 378–​91. 42 Lilliana Mason, 2015. “I Disrespectfully Agree: The Differential Effects of Partisan Sorting on Social and Issue Polarization.” American Journal of Political Science 59:  128–​45; Lilliana Mason, 2018. Uncivil Agreement:  How Politics Became Our Identity. Chicago, IL: University of Chicago Press. 43 Yphtach Lelkes, 2018. “Affective Polarization and Ideological Sorting: A Reciprocal, albeit Weak, Relationship.” The Forum, forthcoming. 44 Shanto Iyengar, Tobias Konitzer, and Kent L.  Tedin, 2018. “The Home as a Political Fortress: Family Agreement in an Era of Polarization.” Journal of Politics, forthcoming. 45 Casey Klofstad, Rose McDermott, and Pete K. Hatemi, 2013. “The Dating Preference of Liberals and Conservatives.” Political Behavior 35: 519–​38. (pp. 530–​1). 46 Jeffrey Berry and Sarah Sobieraj, 2014. The Outrage Industry:  Political Opinion Media and the New Incivility. Oxford: Oxford University Press. 47 Lada A. Adamic and Natalie Glance, 2005. “The Political Blogosphere and the 2004 U.S. Election: Divided They Blog.” Paper presented at the Annual Workshop on the

114  Shanto Iyengar Weblogging Ecosystem, WWW2005, Chiba, Japan, May 10–​14; E.  D. Lawrence, John Sides and Henry Farrell, 2010. “Self-​ Segregation or Deliberation? Blog Readership, Participation, and Polarization in American Politics.” Perspectives on Politics 8: 141–​58. 48 See, for instance, Cass R. Sunstein, 2017. #Republic: Divided Democracy in the Age of Social Media. Princeton, NJ: Princeton University Press. 49 Markus Prior, 2007. Post-​ Broadcast Democracy. How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. New  York:  Cambridge University Press. 50 See Shanto Iyengar and Kyu Hahn, 2009. “Red Media, Blue Media:  Evidence of Ideological Selectivity in Media Use.” Journal of Communication 59:  19–​39; Natalie J. Stroud, 2011. Niche News: The Politics of News Choice. New York: Oxford University Press. 51 Matthew Gentzkow and Jesse M.  Shapiro, 2011. “Ideological Segregation Online and Offline.” Quarterly Journal of Economics 126: 1799–​839. 52 Seth Flaxman, Sharad Goel, and Justin M.  Rao, 2016. “Filter Bubbles, Echo Chambers and Online News Consumption.” Public Opinion Quarterly 80: 298–​320. 53 Eytan Bakshy, Solomon Messing, and Lada A.  Adamic, 2015. “Exposure to Ideologically Diverse News and Opinion on Facebook.” Science 348: 1130–​2. 54 Yphtach Lelkes, Gaurav Sood, and Shanto Iyengar, 2017. “The Hostile Audience: The Effect of Access to Broadband Internet on Partisan Affect.” American Journal of Political Science 61: 5–​20. 55 Levi Boxell, Matthew Gentzkow, and Jesse M. Shapiro, 2017. “Greater Internet Use is Not Associated with Faster Growth in Political Polarization among US Demographic Groups.” Proceedings of the National Academy of Science, 114: 10612–​17. 56 Justin Grimmer and Gary King, 2011. “General Purpose Computer-​ assisted Clustering and Conceptualization.” Proceedings of the National Academy of Science 108: 2643–​50. 57 Adam Berinsky, 2017. “Rumors and Health Care Reform:  Experiments in Political Misinformation.” British Journal of Political Science 47:  241–​62; Donald J.  Flynn, Brendan Nyhan, and Jason Reifler, 2017. “The Nature and Origins of Misperceptions:  Understanding False and Unsupported Beliefs about Politics.” Political Psychology 38: 127–​50.

Bibliography Abramowitz, Alan I., and Kyle L. Saunders. 2008. Is Polarization a Myth? Journal of Politics 70: 542–​55. Adamic, Lada A., & Glance, Natalie. 2005. The Political Blogosphere and the 2004 U.S. Election: Divided They Blog. Paper presented at the Annual Workshop on the Weblogging Ecosystem, WWW2005, Chiba, Japan, May 10–​14. Bakshy, Eytan, Messing, Solomon, and Lada A. Adamic. 2015. Exposure to ideologically diverse news and opinion on Facebook. Science 348: 1130–​2. Berinsky, Adam. 2017. Rumors and Health Care Reform:  Experiments in Political Misinformation. British Journal of Political Science 47: 241–​62. Berry, Jeffrey, and Sarah Sobieraj. 2014. The Outrage Industry: Political Opinion Media and the New Incivility. Oxford: Oxford University Press.

Affective Polarization or Hostility  115 Billig, M. G., and Henri Tajfel. 1973. Social Categorization and Similarity in Intergroup Behavior. European Journal of Social Psychology 3: 27–​52. Boxell, Levi, Gentzkow, Matthew, and Jesse M. Shapiro. 2017. Greater Internet use is not associated with Faster Growth in Political Polarization among US Demographic Groups. Proceedings of the National Academy of Science, 114: 10612–​17. Boysen, Guy A., Vogel, David. L., and Stephanie Madon. 2006. A Public versus Private Administration of the Implicit Association Test. European Journal of Social Psychology 36: 845–​56. Fershtman, Chaim, and Uri Gneezy. 2001. Discrimination in a Segmented Society: An Experimental Approach. Quarterly Journal of Economics 116: 351–​77. Fiorina, Morris, Abrams, Samuel, and Jeremy Pope. 2008. Polarization in the American Public: Misconceptions and Misreadings. Journal of Politics 70: 556–​60. Flaxman, Seth, Goel, Sharad, and Justin M. Rao. 2016. Filter Bubbles, Echo Chambers and Online News Consumption. Public Opinion Quarterly 80: 298–​320. Fleisher, Richard, and Jon R. Bond. 2001. Evidence of Increasing Polarization among Ordinary Citizens, in American Political Parties:  Decline or Resurgence?, edited by Jeffrey E. Cohen, Richard Fleisher, and Paul Kantor, 55–​ 77. Washington, DC: CQ Press. Flynn, Donald. J., Nyhan, Brendan, and Jason Reifler. 2017. The Nature and Origins of Misperceptions:  Understanding False and Unsupported Beliefs about Politics. Political Psychology 38: 127–​50. Forsythe, Robert, Joel L. Horowitz, Nathan E. Savin, and Martin Sefton. 1994. “Fairness in Simple Bargaining Experiments.” Games and Economic Behavior 6: 47–​69. Gaertner, Samuel, Dovidio, John, Anastasio, Phyllis, Bachman, Betty, and Mary Rust. 1993. The Common Ingroup Identity Model: Recategorization and the Reduction of Intergroup Bias. European Review of Social Psychology 4: 1–​26. Gentzkow, Matthew, and Jesse M. Shapiro. 2011. Ideological Segregation Online and Offline. Quarterly Journal of Economics 126: 1799–​839. Gift, Karen, and Thomas Gift. 2015. Does Politics Influence Hiring? Evidence from a Randomized Experiment. Political Behavior 37: 653–​75. Green, Donald, Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts and Minds:  Political Parties and the Social Identities of Voters. New Haven, CT:  Yale University Press. Grimmer, Justin, and Gary King. 2011. General Purpose Computer-​assisted Clustering and Conceptualization. Proceedings of the National Academy of Science 108: 2643–​50. Hetherington, Marc. 2002. Resurgent Mass Partisanship: The Role of Elite Polarization. American Political Science Review 95: 619–​31. Huber, Gregory A. and Neil Malhotra. 2017. Political Homophily in Social Relationships: Evidence from Online Dating Behavior. Journal of Politics 79: 269–​83. Huddy, Leonie, Mason, Lilliana, and Lene Aaroe. 2015. Expressive Partisanship: Campaign Involvement, Political Emotion, and Partisan Identity. American Political Science Review 109: 1–​17. Iyengar, Shanto and Kyu Hahn. 2009. “Red Media, Blue Media: Evidence of Ideological Selectivity in Media Use.” Journal of Communication 59: 19–​39. Iyengar, Shanto, and Masha T. Krupenkin. 2018. The Strengthening of Partisan Affect. Political Psychology 39: 201–​18. Iyengar, Shanto and Sean J. Westwood. 2015. Fear and Loathing Across Party Lines. American Journal of Political Science 59: 690–​707.

116  Shanto Iyengar Iyengar, Shanto, Konitzer, Tobias, and Kent L. Tedin. 2018. The Home as a Political Fortress: Family Agreement in an Era of Polarization. Journal of Politics, Forthcoming. Iyengar, Shanto, Yphtach Lelkes, Matthew Levendusky, Neil Malhotra and Sean Westwood. 2018. The Origin and Consequences of Affective Polarization in the United States. Annual Review of Political Science, Forthcoming. Iyengar, Shanto, Gaurav Sood and Yphtach Lelkes. 2012. Affect, Not Ideology: A Social Identity Perspective on Polarization. Public Opinion Quarterly 76(3): 405–​31. Klofstad, Casey, Rose McDermott, and Pete K. Hatemi. 2013. The Dating Preference of Liberals and Conservatives. Political Behavior 35: 519–​38. Lawrence, E. D., Sides, John, & Henry Farrell. 2010. Self-​Segregation or Deliberation? Blog Readership, Participation, and Polarization in American Politics. Perspectives on Politics 8: 141–​58. Lelkes, Yphtach. 2018. Affective Polarization and Ideological Sorting:  A Reciprocal, albeit Weak, Relationship. The Forum, Forthcoming. Lelkes, Yphtach, Sood, Gaurav, and Shanto Iyengar. 2017. The Hostile Audience:  The Effect of Access to Broadband Internet on Partisan Affect. American Journal of Political Science 61: 5–​20. Levendusky, Matthew S. 2009. The Partisan Sort: How Liberals Became Democrats and Conservatives Became Republicans. Chicago, IL: University of Chicago Press. Levendusky, Matthew S., and Neil Malhotra. 2016. Misperceptions of Partisan Polarization in the American Public. Public Opinion Quarterly 80: 378–​91. Maccoby, Eleanor E., and Nathan Maccoby. 1954. The Interview:  A Tool of Social Science, in Gardner Lindzey (Ed.) The Handbook of Social Psychology. 449–​87, Cambridge, MA: Addison-​Wesley. Mason, Lilliana. 2015. I Disrespectfully Agree: The Differential Effects of Partisan Sorting on Social and Issue Polarization. American Journal of Political Science 59: 128–​45. Mason, Lilliana. 2018. Uncivil Agreement: How Politics Became Our Identity. Chicago, IL: University of Chicago Press. McCarty, Nolan, Keith Poole, and Howard Rosenthal. 2006. Polarized America:  The Dance of Ideology and Unequal Riches. Cambridge, MA: MIT Press. McConnell, Christopher, Malhotra, Neil, Margalit, Yotam, and Matthew S. Levendusky. 2018. The Economic Consequences of Partisanship in a Polarized Era. American Journal of Political Science 62: 5–​18. Michelitch, Kristin. 2015. Does Electoral Competition Exacerbate Interethnic or Inter-​partisan Economic Discrimination? Evidence from a Market Price Bargaining Experiment in Ghana. American Political Science Review 109: 43–​61. Panagopoulos, Costas, Donald P. Green, Jonathan Krasno, Michael Schwam-​Baird, Eric Moore, and Kyle Endres. 2016. “Risky Business:  Does Corporate Political Giving Affect Consumer Behavior?” Unpublished Manuscript, Fordham University. Pew Research Center. 2017. “The Partisan Divide on Political Values Grows Even Wider.” Technical report Available Online at: http://​pewrsr.ch/​2z0qBnt. Prior, Markus. 2007. Post-​Broadcast Democracy. How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. New York: Cambridge University Press. Roccas, Sonia, and Marilynn B. Brewer. 2002. “Social Identity Complexity.” Personality and Social Psychology Review 6: 88–​106. Sears, David O. 1975. Political Socialization, in Fred I. Greenstein and Nelson W. Polsby (Eds.), Handbook of Political Science, Vol. 2. 93–​153, Reading, MA: Addison-​Wesley.

Affective Polarization or Hostility  117 Sears, David O. 1983. The Person-​Positivity Bias. Journal of Personality and Social Psychology 44: 233–​50. Stroud, Natalie J. 2011. Niche News:  The Politics of News Choice. New  York:  Oxford University Press. Sunstein, Cass R. 2017. #Republic:  Divided Democracy in the Age of Social Media. Princeton, NJ: Princeton University Press. Tajfel, Henri, and John C. Turner. 1979. An Integrative Theory of Intergroup Conflict, in The Social Psychology of Intergroup Relations, W. G. Austin and Stephen Worchel (Eds), 33–​47. Monterey, CA: Brooks/​Cole. Theodoridis, Alexander. 2017. “Me, Myself, and (I), (D), or (R)? Partisanship and Political Cognition through the Lens of Implicit Identity.” Journal of Politics 79: 1253–​67. Weisberg, Herbert F., and Jerrold G. Rusk. 1970. Dimensions of Candidate Evaluation. American Political Science Review 64: 1167–​85.

Chapter 5

Racial Attitudes and American Politics Michael Tesler

Race has been one of the most important issues in American politics since the nation’s founding. Yet, the effects of racial attitudes on public opinion have waxed and waned over the course of American political history. As is the case for the effects of all core predispositions and values, the impact of racial attitudes on public opinion depends fundamentally on political information.1 When political communications depict policies, candidates, and parties as benefiting one racial or ethnic group over another, racial attitudes inevitably become more important in Americans’ political preferences. The elections of Barack Obama and Donald Trump, for example, effectively simplified the politics of race by providing average citizens with more information about where the two parties’ policies and candidates stand on racial issues.2 The upshot, this chapter shows, is a growing alignment between racial attitudes and public opinion that has polarized the electorate and helped make American politics increasingly vitriolic.

Measuring Racial Attitudes Before documenting this growing alignment, though, it’s important to understand how social scientists measure racial attitudes. This task is easier said than done, as the conceptualization and measurement of racial prejudice is one of the most contentious issues in public opinion research.3 Still, there are some important points of consensus within this large body of academic literature. All agree that large majorities of white Americans openly subscribed to the ideology of white supremacy at the midpoint of the twentieth century. That ideology, often called old-​fashioned racism, had three main elements: (1) desire for social distance between the races, (2)  beliefs in the biological inferiority of Blacks, and (3) support for public policies insuring racial segregation and formalized discrimination.4 A number of factors, including World War II, the new scientific consensus of biological equality between the races, Supreme Court cases and federal legislation dismantling Jim Crow segregation, the civil rights movement, the Cold War, and rising education levels all helped transform white Americans’ racial attitudes during the second half of the twentieth

Racial Attitudes and American Politics  119

century.5 Old-​fashioned racism dramatically declined during that time period, accompanied by a growing acceptance of racial equality in principle.6 Yet despite this decline in racism, most Whites remained strongly opposed to government action aimed at remedying the profound and enduring racial inequality left from centuries of formal discrimination in such areas as education, income, employment, health, housing, and wealth. As a result, new social-science theories arose in the post-​civil rights era to explain the decline in old-​fashioned racism on the one hand and continued opposition to government action to produce racial equality on the other. These “new racism” theories, which are variously described as symbolic racism, modern racism, and this chapter’s preferred term, racial resentment, suggested that a new form of racial animus best explains the influence of anti-​black sentiment in contemporary American politics.7 Unlike old-​fashioned racism, racial resentment does not embrace notions of black biological and social inferiority. Instead, it is characterized by “a moral feeling that Blacks violate such traditional American values as individualism and self-​reliance, work ethic, obedience, and discipline.”8 That is, racially resentful Whites cite deficiencies in black culture, rather than innate inferiority, to explain the ongoing racial inequality in America. These beliefs are most frequently measured with questions from Donald Kinder and Lynn Sanders’ racial resentment scale. The questions are presented as assertions, asking survey respondents how strongly they agree or disagree with the following statements: It’s really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites. Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class. Irish, Italian, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors. Over the past few years, blacks have gotten less than they deserve. Kinder and Sanders designed the scale to distinguish between racially sympathetic Americans who see racial inequality as the result of structural causes like discrimination, and racially resentful Americans who point the finger at African Americans’ cultural deficiencies for their lower socioeconomic status.9 In fact, recent research convincingly argues that the scale should be recast as “Structural versus Individual Attributions for Black Americans’ Economic and Social Status.”10 According to these four items, at least, America is a racially conservative country. The first panel of Figure 5.1 shows that the average white racial resentment score was around 60 in 2016 on a scale where a score of 0 represents the most racially sympathetic response and 100 the most racially resentful. Informatively, this 2016 racial resentment score was significantly lower than it

120  Michael Tesler

had ever been in American National Election Study (ANES) surveys conducted from 1986 to 2012—​a result consistent with several other surveys showing that the country became more liberal on matters of race and ethnicity as a backlash against Donald Trump’s racially charged 2016 presidential campaign.11 Racial resentment is the measure most commonly used by social scientists to explain the role of racial attitudes in public opinion.12 Yet some political scientists have argued that racial resentment confounds anti-​black animus with ordinary political conservatism, so that its effects may only reflect a relatively unprejudiced aversion to liberal big government.13 These scholars, instead, prefer to measure prejudice with more blatantly racist measures like negative racial stereotypes. The racial stereotype scale typically asks respondents to rate how hardworking, intelligent, and/​or violent racial and ethnic groups are on seven-​point scales, ranging from 1 (lazy/​unintelligent/​violent) to 7 (hardworking/​intelligent/​peaceful). Measured this way, racial stereotyping has been described as “subtle but consistent.”14 A majority of Whites consistently rate their own group as more hardworking and peaceful than African Americans. Yet, almost no Whites say that their racial group is categorically better than Blacks. For example, the second panel of Figure 5.1 shows that on a 0–​100 stereotype scale, where a score of 50 means rating Blacks and Whites equally and a score of 100 represents the most prejudiced response possible, the average white American’s stereotype score is in the high 50s. That average stereotype score has been relatively constant in ANES surveys conducted from 1992 through 2016.15 Racial resentment is significantly correlated with overt measures of prejudice like negative stereotypes. But stereotypes are less-​rooted political ideology and have weaker impacts on Americans’ political preferences than racial resentment. At the same time, though, the more blatant nature of racial stereotypes is also a major weakness. Stereotypes are especially susceptible to misreporting due to social desirability pressures to rate all racial groups equally.16 Moreover, 40%

40%

30%

30%

20%

20%

10%

10%

0% Sympathetic

Resentful

0% Pro-black

Racial resentment scale

Figure 5.1 Distribution of White Racial Attitudes

Pro-white Stereotype scale

Racial Attitudes and American Politics  121

the low-​end of the racial resentment scale does a much better job of tapping into racial sympathy than the low-​end of the stereotype scale. Figure 5.1 shows that nearly one-​third of whites fall on the sympathetic side of the racial resentment scale, compared to almost none who rate Blacks more favorably than their own racial group on the stereotype scale. The stereotype scale’s inability to tap into racial sympathy inevitably underestimates the political effects of racial attitudes.17 Like past academic disputes, the degree to which racial attitudes influence Americans’ political preferences depends on how we measure attitudes toward African Americans. More importantly, though, several different measurements—​ whether racial resentment, anti-​ black stereotypes, overtly unfavorable ratings of African Americans, or even old-​fashioned racist opposition to interracial dating and marriage—​indicate that racial attitudes have become increasingly aligned with Americans’ political opinions during Barack Obama and Donald Trump’s presidencies.

Racial Attitudes and Voting for President For over half of a century now, the two major political parties’ presidential nominees have differed in their policies and rhetoric about race. The enactment of civil rights legislation by the Democratic Kennedy and Johnson administrations, which was opposed by the Republican’s 1964 presidential candidate, Barry Goldwater, helped solidify the partisan realignment on race that had been taking place since the Great Depression.18 Those partisan divisions were solidified by the two major parties’ subsequent presidential platforms. Most notably, the Republican Party explicitly denounced racial quotas and race-​ based preferences in every one of their party platforms from the 1970s through Barack Obama’s elections, while every Democratic platform openly endorsed affirmative action during that same time period.19 Along with those policy differences, there was a growing divide between Democratic and Republican presidential candidates’ racial rhetoric in the latter decades of the twentieth century. Implicit racial appeals became a staple of Republican presidential campaigns when Richard Nixon deployed the “southern strategy” in 1968 to try to win over disaffected white Southern Democrats with subtle appeals to anti-​black stereotypes and racial resentments.20 It’s not surprising, then, that racial attitudes have long been important ingredients in Americans’ vote choices for president. Racially resentful Whites were over 30  percentage points more likely to support the Republican candidate in presidential elections from 1988 to 2004 than their racially sympathetic counterparts, even after statistically controlling for the fact that liberal Democrats tend to be more racially tolerant than conservative Republicans.21 Pre-​ Obama political-science research also showed that this existing link between racial attitudes and Americans’ preferences for president could be further strengthened through racial priming—​informational cues in political

122  Michael Tesler

messages as subtle as race-​coded words (i.e., “inner city” or “welfare”), black imagery, and especially some combination of the two.22 Racial attitudes then became an even stronger predictor of vote choice in Barack Obama’s two elections. Several studies showed that racial resentment, anti-​black stereotypes, ethnocentric favoritism of whites over other racial groups, and old-​fashioned racist opposition to intimate interracial relationships all had significantly stronger effects on voter preferences in 2008 and 2012 than their influence on pre-​Obama presidential contests.23 These same racial attitudes were also significantly stronger predictors of Americans’ vote preferences than they would have been had John McCain faced a white Democrat like Hillary Clinton or John Edwards instead of Barack Obama in the 2008 election.24 Even living in areas with a high volume of racist Google searches was more strongly linked to Republican vote choice in Obama’s two elections than it was in either the 2004 presidential election or in 2008 trial heats that matched John McCain against Hillary Clinton rather than Barack Obama.25 It is important to note that the outsized effects of racial attitudes on voter preferences in Obama’s 2008 and 2012 presidential election victories cut both ways. That is, Barack Obama performed particularly poorly among racial resentful Whites, but garnered more votes from African Americans and white racial liberals than a similarly situated white Democratic candidate like Hillary Clinton would have. We described this phenomenon as the two sides of racialization because prior research primarily focused on racially resentful opposition to black candidates and racial policies.26 The large effects of racial attitudes in Obama’s election, therefore, did not so much hurt him electorally as they polarized voter preferences based on their feelings about African Americans. Those unusually large effects of both positive and negative racial attitudes on Americans’ 2008 and 2012 vote preferences are rather straightforward: Barack Obama’s omnipresent position as a historic racial figure, and his embodiment of race as the first African-​American president, made racial attitudes an easily accessible consideration in public opinion about his candidacy and presidency. As Donald Kinder and Allison Dale-​Riddle aptly asserted, “Whatever Obama said about society and government and about problems and policies, at the end of the day, every time American voters caught a glimpse of him, he was black.”27 That strong and easy link between race and the president made it difficult for Obama’s race-​neutral rhetoric and policies to neutralize the effects of racial attitudes during his presidency.28 But the impact of racial attitudes on presidential vote choice is not simply a matter of the candidates’ racial backgrounds. Perhaps even more important is which groups are thought to be helped or hurt by the candidates’ policies. In other words, do voters think presidential candidates are for us or for them?29 Despite Obama’s embodiment of race, he spoke about race less than recent Democratic presidents and when he did, he often preached a message of black personal responsibility that should have resonated with racial conservatives.30 He was also criticized by black leaders and intellectuals for refusing to push

Racial Attitudes and American Politics  123

for race-​specific policies. There was little about Obama’s governing behavior to make Whites think he wasn’t for them. Indeed, Obama polarized public opinion by racial attitudes primarily because of who he was, not because of what he said or did. Hillary Clinton, by contrast, made racial equality one of her 2016 presidential campaign’s defining messages, speaking early and often about the pernicious consequences of “systematic racism” and the “implicit biases we all have.” “As a result,” Clinton wrote in her 2017 memoire, “some white voters may have decided I wasn’t on their side.”31 Meanwhile, it was abundantly clear to white voters whose side Donald Trump was on. Trump repeatedly went where prior Republican presidential candidates were unwilling to go: making explicit appeals to racial resentment, anti-​ immigrant sentiments, Islamophobia, and white grievances. Hillary Clinton even said the following about Trump in an August 2016 campaign speech, “There’s always been a paranoid fringe in our politics, steeped in racial resentment. But it’s never had the nominee of a major party stoking it, encouraging it, and giving it a national megaphone. Until now.”32 The American public was paying attention to Trump’s rhetoric. More than half of Americans consistently said that the term “racist” described Donald Trump in YouGov/​Economist surveys conducted from August to November of the election year. Likewise, 61 percent of likely voters in a September 2016 Quinnipiac University Poll thought that “the way Donald Trump talks appeals to bigotry.” And a late October 2016 Pew Poll showed that most Americans thought Donald Trump had more respect for Whites than Blacks (55 percent), but virtually no one thought he respected Blacks more than Whites.33 Most importantly, the American public saw a much wider gulf between Clinton and Trump’s positions on issues like immigration and federal aid to African Americans than they had perceived between prior Democratic and Republican presidential candidates.34 That set the stage for attitudes about race and ethnicity to matter more in 2016 than they had in modern times. Across several different racial attitude measures in a number of different surveys, views about race and ethnicity were more strongly related to vote choice in 2016 than they were in Obama’s elections. The top of Figure 5.2 shows the results from one of those surveys—​ RAND’s Presidential Election Panel Survey, which interviewed the exact same individuals in both 2012 and 2016. The left-​hand side of the display shows that racial resentment was more closely linked to white voters’ support for Donald Trump in 2016 than their support for Mitt Romney in 2012. After controlling for partisanship and ideology, the most racially resentful Whites were significantly more likely to support the Republican presidential candidate in 2016 than they were in 2012. Those results are consistent with several other studies showing that racial resentment, anti-​black stereotypes, anti-​Muslim sentiments, anti-​immigrant attitudes, and thinking Whites are discriminated against in society were all more closely linked to white Americans’ vote choices for president in 2016 than they were in 2008 or 2012.35

124  Michael Tesler

The right-​hand panel of Figure 5.2 suggests that Donald Trump’s campaign was largely responsible for the enhanced relationship between racial attitudes and voting for president in 2016. Much the way that racial attitudes mattered more in hypothetical matchups conducted during the 2008 Democratic primaries that pitted John McCain against Barack Obama, racial resentment was a stronger predictor of opposition to Hillary Clinton when she was matched up against Donald Trump than it was when she faced other Republican presidential candidates (e.g., Ted Cruz and Marco Rubio) in Spring 2016 surveys. In addition to those results for racial resentment in the right-​hand display of Figure  5.2, the exact same pattern was found using other racial attitude measures. In fact, ethnocentrism, anti-​ black sentiments, and perceptions of discrimination against Whites were all stronger predictors of opposition to Hillary Clinton when she was matched up against Donald Trump than they were when she was pitted against other Republican challengers in 2016 surveys.36 It appears that Donald Trump’s racialized campaign helped make racial attitudes matter more in 2016 than they would have if another Republican had been the nominee. All told, then, the impact of racial attitudes on voting for president has evolved over time based on the positions and rhetoric of the two parties and their candidates for the presidency. Racial attitudes became a significant predictor of Americans’ votes after presidential candidates from the two parties began taking an increasingly dissimilar position on racial issues during the 1960s. That relatively stable relationship increased in 2008 when Barack Obama’s presidency made it even easier for Americans to link their racial attitudes to their vote choices. The 2016 campaign then further extended that strong polarization of the electorate by making it increasingly clear to Americans—​even Americans who don’t pay much attention to politics—​that the two parties and their presidential candidates had very different positions on matters of race and ethnicity.

100%

Whites’ support for the Republican presidential candidate in 2012 and 2016 100%

White’s support for Republican presidential candidates against Hillary Clinton in 2016

2016 80%

80% 60%

2012

60%

40%

40%

20%

20%

0%

0%

Sympathetic

Racial attitudes

Resentful

Clinton vs. Cruz & Rubio

Clinton vs. Trump

Sympathetic

Racial attitudes

Figure 5.2 Whites’ Support for Presidential Candidates by Racial Attitudes

Resentful

Racial Attitudes and American Politics  125

Racial Attitudes and Public Policy Preferences The impact of racial attitudes on Americans’ public policy preferences also depends on which racial groups Americans think those policies benefit. Or, as the political scientists Jon Hurwitz and Mark Peffley surmised, When messages are framed in such a way to reinforce the relationship between a particular policy and a particular group, it becomes far more likely that individuals will evaluate the policy on the basis of their evaluations of the group.37 For some issues, there is a clear and direct link between public policies and the racial and ethnic groups who benefit from them. So, it’s not surprising that racial attitudes are important determinants of public opinion about policies like affirmative action, immigration, and government assistance to racial minorities.38 But racial attitudes also influence public opinion about several ostensibly nonracial policies. This process, in which certain nonracial policies become racialized, often occurs after mass communications heighten the association between racial groups and public policies. For example, the emergence of media coverage linking welfare benefits with “undeserving blacks” in the 1960s and 1970s helped white Americans bring their racial antagonisms to bear on opposition to this policy.39 Similarly, media coverage connecting Social Security to both white recipients and to such symbolically white attributes as hard work and just rewards helped make this nonracial policy about race.40 And political scientists have argued that “the race-​coded rhetoric public officials use to talk about crime and media coverage that exaggerates black violence,” has caused racial attitudes to influence white Americans’ support for the death penalty.41 Social science experiments provide even stronger causal evidence that media coverage connecting racial groups to certain issues can make racial attitudes a more important ingredient of white Americans’ policy preferences. In fact, racial attitudes have been significantly activated in white Americans’ policy positions by experiments that randomly assigned respondents to receive racialized messages (e.g., race coded language and/​or racial imagery) about such issues as crime, gun control, welfare, government spending, early education programs, Social Security, the minimum wage, and even the Iraq War.42 Barack Obama’s association with specific policies had a similar “racializing” effect on public opinion during his presidency. Racial attitudes became stronger determinants of public opinion when Barack Obama took visible positions on issues—​a phenomenon we labeled the spillover of racialization. That is, the president’s position as the center of the political universe strongly connected him to several different political evaluations (e.g., mass assessments of public figures, public policy preferences, vote choices for congressional candidates, party identification, etc.), many of which became more racialized

126  Michael Tesler

because of their newfound association with Barack Obama. Indeed, the spillover of racialization extended into several different policies during Obama’s presidency, including taxes on the rich, the 2009 Recovery and Reinvestment Act (“the Stimulus”), and same-sex marriage.43 But health care was easily the issue most affected by this phenomenon. Barack Obama’s strong association with the policy that ultimately took his name (Obamacare) made health care an issue particularly ripe for the spillover racialization. Multiple studies showed that racial resentment and anti-​ black stereotypes were much stronger predictors of health-care attitudes after Obama became the face of the presidency than they were before the 2008 election.44 Moreover, racial attitudes were significantly stronger determinants of health-care attitudes in experiments that framed health-care proposals as Obama’s policies than experiments that framed the exact same policies as Bill Clinton’s 1993 proposals.45 To be sure, racial attitudes were by no means the only, or even the most important, determinants of Americans’ healthcare opinions during Obama’s presidency. But the effects of racial attitudes on public policies tended to increase in importance relative to other factors, such as ideology and support for limited government, after Barack Obama was connected to an issue. The spillover of racialization into health care had important consequences, too. Before Obamacare took effect in 2013, racial attitudes were entirely unrelated to Americans’ decision to purchase health insurance. But after 2013, the most racially sympathetic Americans were significantly more likely than the most racially resentful to have health insurance (after controlling for partisanship, ideology, and several factors associated with having health-insurance coverage).46 It appears, then, that the spillover of racialization extended even into especially important decisions like purchasing health insurance.

Racial Attitudes and Party Identification As important as those health-care results are, though, the most consequential effects of racial attitudes are found in their relationship with party identification. Party identification influences just about everything in contemporary American society. Partisanship is not only the most important determinant of our vote choices and policy preferences, but it shapes countless other beliefs and behaviors. Party identification has even been linked to who we find attractive and who we decide to marry, how we perceive objective conditions like the unemployment rate and federal budget deficit, which neighborhoods we want to live in, and the type of TV shows and cars we like.47 As such, the racialization of party identification is by itself the racialization of American politics and society. Like voting for president, the relationship between partisanship and racial attitudes has its origins in the growing divisions between the two parties over civil rights issues in the 1960s and 1970s. Americans’ partisan identities became

Racial Attitudes and American Politics  127

increasingly sorted by racial attitudes in the post-​civil rights era, as older citizens who came of age before the two parties sharply diverged on racial issues were gradually replaced by younger Americans whose partisan attachments were formed amidst the new racial schism in party politics.48 That growing relationship between racial attitudes and party identification was particularly consequential in the American South, where negative racial attitudes helped turn the once solidly Democratic region into a Republican stronghold by the end of the twentieth century.49 Yet the link between racial attitudes and party identification was far from complete before Obama’s presidency. Moral issues had largely displaced racial issues from partisan discourse during the presidencies of Bill Clinton and George W. Bush. As a result, the growing relationship between racial resentment and Republican partisanship leveled off by the late 1990s after controlling for traditional views about morality and family values.50 Racial resentment did not predict changes in white partisanship from 2000 to 2006. Moreover, blatant racial attitudes, such as opposition to interracial dating and explicitly unfavorable views of African Americans, were entirely unrelated to partisanship for decades leading up to Obama’s election; and the influence of racial resentment on party identification was largely confined to politically informed individuals who were aware of the two parties’ differences on matters of race and ethnicity.51 This all changed with the election of Barack Obama. Racial attitudes became a much stronger predictor of party identification after Obama was elected president.52 In fact, racial resentment was consistently the strongest predictor of changes in partisanship during Obama’s presidency. Moreover, overt racial attitudes that were unrelated to partisanship throughout the pre-​ Obama era, such as opposition to interracial marriage and negative views of African Americans, now predicted Republican Party identification for the first time in decades during Obama’s presidency.53 This increased association between racial attitudes and partisanship during Obama’s presidency was particularly strong for Whites who did not attend college. College-​educated Whites and Whites who live in highly educated areas of the country have long been much more racially tolerant than other white people.54 Yet, while poorly educated Whites as a group are more racially intolerant, prejudice has typically played a weaker part in their political preferences than it has for better-​educated Americans.55 At the root of the paradox is that Whites with more education followed politics more closely and thus knew about long-​standing divisions between the Democratic and Republican parties on civil rights and other policies related to race. Whites with less education, who tended to follow politics less closely, hadn’t fully learned this. After Obama’s first election, Whites who did not attend college became much more aware of the fact that the Democratic Party is more supportive of racially progressive policies than the Republicans.56 In other words, the election of an African American Democratic president simplified the politics of race

128  Michael Tesler

80%

60%

Democratic party identification (Non-college Whites in ANES cumulative file)

Lower racial resentment

40%

20%

80%

60%

Democratic party identification (Non-college Whites in Pew values cumulative file)

Discrimination against Blacks is NOT rare

40%

High racial resentment

0%

20%

Discrimination against Blacks is rare

0% 1984 1988 1992 1996 2000 2004 2008 2012

1984 1988 1992 1996 2000 2004 2008 2012

Figure 5.3 Democratic Identification among Non-​College Educated Whites by Racial Attitudes, 1986–​2012

and helped shrink the diploma divide in awareness of the two parties’ differing positions on race. That information had important effects on their partisanship. After decades of being unrelated to non-​college Whites’ partisanship, racial attitudes began to influence their partisanship during Obama’s presidency. The left-​hand side of Figure 5.3 shows that non-​college Whites who scored high in racial resentment were about 15 percentage points less Democratic in the 2012 ANES than they had been in the pre-​Obama era. The second display shows a similarly sharp downturn in Democratic identification among poorly educated Whites in Pew Surveys who thought that discrimination against African Americans is rare in the Pew data—​a belief closely associated with racial resentment. Meanwhile, there was no such drop-​off among working-​class Whites with more progressive racial attitudes. Simply put, racially resentful Whites without a college degree were most likely to flee the Democratic Party during Barack Obama’s presidency. That exodus meant that there was a large group of disaffected Democratic voters for Donald Trump’s campaign appeals to pick off in 2016. Indeed, the largest swings from Obama to Trump in 2016 generally occurred in areas of the country with lots of white voters who did not attend college and occurred especially among racially resentful Whites without a college degree. Donald Trump obviously did not create these patterns, which were firmly established during the years leading up to his presidential campaign, but he effectively exploited and accelerated them.57 His explicit appeals to racial resentment and white identity made it even clearer where the two parties stood on racial issues and expanded the Obama-​era polarization of mass partisanship by racial attitudes.58

Racial Attitudes and American Politics  129

The Growing Racialization of American Politics and Society It’s hard to overstate the importance of these trends. For, as mentioned earlier, the growing racialization of partisanship by itself represents the growing racialization of American politics and society. Indeed, Democratic and Republican voters do not simply disagree about what the government should do on racially charged issues like immigration and affirmative action, they now inhabit increasingly separate realities about race in America. Democrats and Republicans viewed race-​related events, such as firing prominent individuals who made racist comments and police violence against unarmed African Americans, much differently during Obama’s presidency than they did before.59 Figure 5.4 expands on those results, showing just how dissimilar Democrats’ and Republicans’ views of race-​related controversies have become in recent years. The upper-​left panel of the display plots the percentage of Democrats 100%

The Confederate flag symbolizes racism

The n-word is offensive

100%

80%

80%

60%

Democrats

60% Democrats

40%

40%

20%

Republicans

20% Republicans

0%

100%

2000

2006

2012

2018

Not enought black history taught in schools

100%

Democrats

2006

2012

2018

Too much black history taught in schools

60% Republicans

40%

40% 20%

2000

80%

80% 60%

0%

Republicans

20% Democrats

0%

0% 2000

2006

2012

2018

2000

Figure 5.4 Views of Race-​Related Issues by Party, 2000–​2018

2006

2012

2018

130  Michael Tesler

and Republicans who think that the Confederate flag symbolizes racism rather than southern pride. In 2000, Democrats were only 24 percentage points more likely than Republicans to say that the flag symbolized racism (40 percent to 16 percent respectively). That relatively modest difference between Democrats and Republicans, however, expanded into a 44-​percentage point divide when the exact same question was asked in July 2015 and then to a 52-​point gulf the last time this question was asked in September 2016. These partisan divisions over confederate symbolism grew even larger after President Trump defended white nationalists protesting the removal of Robert E. Lee’s statue from a park in Charlottesville Virginia. In an August 2017 YouGov/​Economist Poll, 81 percent of Trump voters opposed removing the statue, compared to just 21 percent of Clinton supporters. The upper-right panel similarly shows a growing divide in how Democrats and Republicans view the n-​word. In a 2006 CNN/​ORC poll, 55  percent of Democrats and 49 percent of Republicans said that the n-​word was offensive. By 2018, though, that narrow gap had widened. More Democrats (75 percent) found the n-​word offensive. Meanwhile, fewer Republicans (43 percent) did. Additional polling from 2018 showed that few Trump voters think that using the n-​word makes Whites racist (18 percent), just 42 percent find the slur offensive, and only a quarter of Trump voters wouldn’t vote for a candidate who said it. By contrast, more than three-​quarters of people who voted for Hillary Clinton thought the n-​word is offensive, racist, and should disqualify political candidates who have said it.60 The bottom panels of Figure  5.4 show that Democrats and Republicans have even grown more divided on the value of teaching black history in public schools. The bottom-​left display shows that in 2000, Democrats were only about 20 percentage points more likely than Republicans to say there’s not enough black history taught in public schools. That partisan divide expanded to 57 points in 2018. Likewise, the bottom right-​hand display shows that Democrats and Republicans have also grown increasingly divided over whether schools are teaching too much black history. In 2000, Democrats and Republicans didn’t differ on this, and Democrats haven’t changed in the past two decades. Republicans, however, are now 30 points more likely to say schools should teach less black history. In fact, one-​third of President Trump supporters think that “American children should solely be taught about Western civilization and European/​U.S. history.”61

Conclusion The growing partisan divisions over race documented in this chapter are particularly consequential. They threaten to make political conflict less about what the government should do and more about which racial and ethnic groups’ identities are respected in American society. That, in turn, should make politics more vitriolic and divisive, since political-science research

Racial Attitudes and American Politics  131

convincingly shows that racial attitudes are emotionally charged and evoke anger in ways that nonracial ideological attitudes about the size of the government are not.62 In fact, a recent study by political scientists Nicholas Valentino and Kirill Zhirkov found that Americans’ growing dislike of the opposite party—​“negative partisanship”—​is related precisely to the increasing overlap between race and party in the minds of ordinary Americans.63 That overlap between race and party in the minds of ordinary Americans appears to be growing stronger during Donald Trump’s presidency. The upshot is a growing racialization of public opinion—​one that polarizes the electorate and makes American politics especially vitriolic.

Notes 1 John Zaller, for example, famously described every opinion as “a marriage of information and predisposition: information to form a mental picture of the given issue, and predisposition to motivate some conclusion about it.” Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge University Press. Online, p. 6. 2 Tesler, Michael. 2016. Post-​Racial or Most Racial? Race and Politics in the Obama Era. Chicago, IL: University of Chicago Press; Sides, John, Michael Tesler and Lynn Vavreck. 2018. Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America. Princeton, NJ: Princeton University Press. 3 For an overview, see: Sears, David O., Jim Sidanius, and Lawrence Bobo, eds. 2000. Racialized Politics: The Debate About Racism in America. Chicago, IL: University of Chicago Press. 4 This ideology of white supremacy has also been called Jim Crow racism, red-​neck racism, and biological racism. See:  Bobo, Lawrence, and James R.  Kluegel, 1997. Status, ideology, and dimensions of Whites’ racial beliefs and attitudes: Progress and stagnation. In Racial Attitudes in the 1990s: Continuity and Change, edited by Steven A. Tuch and Jack K. Martin, 93–​120. Westport, CT: Praeger; McConahay, John B., and Joseph C. Hough, Jr. 1976. “Symbolic racism.” Journal of Social Issues 322: 23–​45; McConahay, John B., Betty B. Hardee, and Valerie Batts. 1981. “Has racism declined in America? It depends upon who is asking and what is asked.” Journal of Conflict Resolution 254: 563–​579; Kinder, Donald R. 2013. Prejudice and politics. In Oxford Handbook of Political Psychology, 2nd Edition, edited by Leonie Huddy, David O. Sears and Jack S. Levy. Oxford: Oxford University Press. 5 Lee, Taeku. 2002. Mobilizing Public Opinion: Black Insurgency and Racial Attitudes in the Civil Rights Era. Chicago, IL: University of Chicago Press; Dudziak, Mary L. 2001. Cold War Civil Rights:  Race and the Image of American Democracy. Princeton, NJ: Princeton University Press; Schuman, Howard, Charlotte Steeh, Lawrence Bobo, and Maria Krysan. 1997. Racial Attitudes in America:  Trends and Interpretations. Cambridge, MA: Harvard University Press; Page, Benjamin I., and Robert Y. Shapiro. 1992. The Rational Public:  Fifty Years of Trends in Americans’ Policy Preferences. Chicago, IL:  University of Chicago Press; Zaller, The Nature and Origins of Mass Opinion, 1992. 6 Schuman et al. 1997. Racial Attitudes in America; Mendelberg, Tali. 2001. The Race Card. Princeton, NJ: Princeton University Press.

132  Michael Tesler 7 Kinder, Donald R., and David O.  Sears. 1981. “Prejudice and politics:  Symbolic racism versus racial threats to the good life.” Journal of Personality and Social Psychology 40(3):  414; McConahay, John B.  1986. Modern racism, ambivalence, and the modern racism scale. In Prejudice, Discrimination, and Racism, edited by John F.  Dovidio and Samuel L.  Gaertner, 91–​126. New  York:  Academic Press; Kinder, Donald R., and Lynn M. Sanders. 1996. Divided by Color: Racial Politics and Democratic Ideals. Chicago, IL: University of Chicago Press. 8 Kinder and Sears. “Prejudice and politics,” p. 416. 9 Kinder and Sanders. Divided by Color. 10 Kam, Cindy D.  and Camille D.  Burge. 2018. “Uncovering reactions to the racial resentment scale across the racial divide.” The Journal of Politics 80(1): 314–​320. 11 Sides et al. 2018. Identity Crisis. 12 Hutchings, Vincent L. and Nicholas A. Valentino. 2004. “The centrality of race in American politics.” Annual Review of Political Science 7: 383–​408. 13 Sniderman, Paul M. and Philip Tetlock. 1986. “Symbolic racism: Problems of motive attribution in political debate.” Journal of Social Issues 42(2):  129–​50; Sniderman, Paul M., Gretchen C. Crosby, and William G. Howell. 2000. “The politics of race.” In Racialized Politics: The Debate About Racism in America, edited by D. O. Sears, J.  Sidanius and L.  Bobo. Chicago, IL:  University of Chicago Press, pp.  236–​79; Hurwitz, Jon, and Mark Peffley. 1998. Perception and Prejudice: Race and Politics in the United States. London: Yale University Press. 14 Kinder, Donald. “Prejudice and politics,” p. 822. 15 Tesler. Post-​Racial or Most Racial? 16 Huddy, Leonie, and Stanley Feldman. 2009. “On assessing the political effects of racial prejudice.” Annual Review of Political Science 12:  423–​47; Iyengar, Shanto, Kyu Hahn, Christopher Dial, and Mahzarin R. Banaji. 2009. “Explicit and implicit attitudes: Black-​white and Obama-​McCain comparisons.” Prepared for Presentation at the Annual Meeting of the International Society of Political Psychology, Dublin, July  15–​18. 17 In fact, we show elsewhere that the different political effects of racial resentment and negative stereotypes are primarily due to the latter’s inability to adequately identify racially sympathetic Whites. Tesler, Michael and David O.  Sears. 2015. “How Standard Stereotype Measures Underestimate the Political Impact of Racial Attitudes.” Paper presented at American Political Science Association’s annual meeting. 18 Carmines, Edward G.  and James A.  Stimson. 1989. Issue Evolution:  Race and the Transformation of American Politics. Princeton, NJ:  Princeton University Press; Schickler, Eric. 2016. Racial Realignment:  The Transformation of American Liberalism, 1932–​1965. Princeton, NJ: Princeton University Press. 19 King, Desmond S. and Roger M. Smith. 2011. Still a House Divided: Race and Politics in Obama’s America. Princeton, NJ: Princeton University Press. 20 Mendelberg, Tali. 2001. The Race Card; Lopez, Ian Haney. 2014. Dog Whistle Politics:  How Coded Racial Appeals have Reinvented Racism and Wrecked the Middle Class. Oxford:  Oxford University Press; O’Reilly, Kenneth. 1995. Nixon’s Piano:  Presidents and Racial Politics from Washington to Clinton. New  York: Free Press. 21 Tesler. 2016. Post-​Racial or Most Racial?; Kinder, Donald R. and Allison Dale-​Riddle. 2012. The End of Race? New Haven, CT: Yale University Press. Tesler, Michael and

Racial Attitudes and American Politics  133 David O.  Sears. 2010. Obama’s Race:  The 2008 Election and the Dream of a Post-​ Racial America: Chicago, IL: University of Chicago Press. 22 Kinder and Sanders, Divided by Color; Mendelberg. The Race Card; Valentino, Nicholas A. 1999. “Crime news and the priming of racial attitudes during evaluations of the President.” Public Opinion Quarterly 63:  293–​320. Valentino, Nicholas A., Vincent L.  Hutchings, and Ismail K.  White. 2002. “Cues that matter:  How political ads prime racial attitudes during campaigns.” American Political Science Review, 96: 75–​90. 23 Tesler and Sears. Obama’s Race; Tesler. Post-​Racial or Most Racial?; Kinder and Dale-​ Riddle. The End of Race?; Kam, Cindy and Donald R. Kinder. 2012. “Ethnocentrism as a short-​term influence in the 2008 election,” American Journal of Political Science; Piston, Spencer. 2010. “How explicit racial prejudice hurt Obama in the 2008 election.” Political Behavior 32: 431–​451. Tesler, Michael. 2013. “The return of old-​ fashioned racism to white Americans’ partisan preferences in the early Obama era.” The Journal of Politics 75(1):110–​123. 24 Tesler and Sears. Obama’s Race; Kam and Kinder. “Ethnocentrism”; Tesler. “The return of old-​fashioned racism”; Jackman, Simon, and Lynn Vavreck. 2012. “How does Obama match-​up? Counterfactuals and the role of Obama’s race in 2008.” Unpublished manuscript; Kinder, Donald R., and Timothy J. Ryan. 2017. “Prejudice and politics re-​examined: The political significance of implicit racial bias.” Political Science Research and Methods 5(2): 241–​259. 25 Stephens-​Davidowitz, Seth. “The cost of racial animus on a black candidate: Evidence using Google search data.” Journal of Public Economics 118 (2014): 26–​40. 26 Tesler and Sears. Obama’s Race; Tesler. Post-​Racial or Most Racial? 27 Kinder and Dale-​Riddle. The End of Race?, p. 25. 28 Tesler. Post-​Racial or Most Racial?; Kinder and Dale-​Riddle. The End of Race?; Valentino, Nicholas A., Fabian G. Neuner, and L. Matthew Vandenbroek. “The changing norms of racial political rhetoric and the end of racial priming.” The Journal of Politics 80, no. 3 (2018). 29 Kinder. Prejudice and politics. 30 Coe, Kevin, and Michael Reitzes. 2010. “Obama on the stump:  Features and determinants of a rhetorical approach.” Presidential Studies Quarterly 40(3): 391–​413; Coe, Kevin, and Anthony Schmidt. 2012. “America in black and white: Locating race in the modern presidency, 1933–​2011.” Journal of Communication 62, no. 4: 609–​ 627; Gillion, Daniel Q.  2016. Governing with Words:  The Political Dialogue on Race, Public Policy, and Inequality in America. Cambridge: Cambridge University Press; Harris, Fredrick. 2012. The Price of the Ticket:  Barack Obama and the Rise and Decline of Black Politics. Oxford: Oxford University Press; McIlwain, Charlton, and Stephen M. Caliendo. 2011. Race Appeal: How Candidates Invoke Race in US Political Campaigns. Philadelphia, PA: Temple University Press. 31 Clinton, Hillary. 2017. What Happened. New York: Simon and Schuster, p. 415. 32 For a full transcript of this campaign speech, see: www.politico.com/​story/​2016/​08/​ transcript-​hillary-​clinton-​alt-​right-​reno-​227419. 33 For more on these data, see: Tesler, Michael. “Jemele Hill’s the mainstream: Most Americans think Donald Trump’s a racist.” Huffington Post/​Pollster. September 15, 2017, www.huffingtonpost.com/​entry/​jemele-​hills-​the-​mainstream-​most-​americans​think_​us_​59bad6ace4b0390a1564dbf6. 34 Sides et al. Identity Crisis.

134  Michael Tesler 35 Sides et  al. Identity Crisis; Mutz, Diana C.  2018. “Status threat, not economic hardship, explains the 2016 presidential vote.” Proceedings of the National Academy of Sciences; Collingwood, Loren, Tyller Reny and Ali Valenzuela. 2018. “Flipping for Trump.” Public Opinion Quarterly; Hopkins, Daniel. 2018. “Prejudice, priming and presidential voting:  Panel evidence from the 2016 election.” Working Paper; Schaffner, Brian F., Matthew Macwilliams and Tatishe Nteta. 2018. “Understanding white polarization in the 2016 vote for president: The sobering role of racism and sexism.” Political Science Quarterly 133(1): 9–​35. 36 Tesler, Michael. 2016. “In a Clinton–Trump match up, racial prejudice makes a striking difference.” Washington Post/​Monkey Cage; Sides, John, Michael Tesler and Lynn Vavreck. 2017. “Donald Trump and the rise of white identity politics.” Paper presented at annual meeting of the American Political Science Association. 37 Hurwitz, Jon and Mark Peffley. 2005. “Playing the race card in the post–​Willie Horton era:  The impact of racialized code words on support for punitive crime policy.” Public Opinion Quarterly 69: 99–​112, p. 105. 38 Kinder and Sanders, Divided by Color. Kinder, Donald R.  and Cindy D.  Kam. 2009. Us against Them:  Ethnocentric Foundations of American Opinion. Chicago, IL: The University of Chicago Press; Virtanen, Simo and Leonie Huddy. 1998. “Old-​ fashioned racism and new forms of racial prejudice.” Journal of Politics 60(2): 311–​32. 39 Gilens, Martin. 1999. Why Americans Hate Welfare:  Race, Media, and the Politics of Antipoverty Policy. Chicago, IL:  University of Chicago Press; Kellstedt, Paul M.  2003. The Mass Media and the Dynamics of American Racial Attitudes. Cambridge:  Cambridge University Press; Winter, Nicholas J.G. 2008. Dangerous Frames:  How Ideas about Race and Gender Shape Public Opinion. Chicago, IL: University of Chicago Press. 40 Winter, Nicholas J.G. 2006. “Beyond welfare: Framing and the racialization of white opinion on social security.” American Journal of Political Science 50(2):  400–​420; Winter. Dangerous Frames. 41 Soss, Joe, Laura Langbein, and Alan R. Metelko. 2003. “Why do white Americans support the death penalty?” Journal of Politics 65(2): 397–​421. 42 Hurwitz and Peffley. “Playing the race card”; Kinder and Kam, Us Against Them; Mendelberg, The Race Card; Winter. Dangerous Frames; Peffley, Mark and Jon Hurwitz. 2010. Justice in America:  The Separate Realities of Blacks and Whites. New York: Cambridge University Press; Federico, Christopher M. 2004. “When do welfare attitudes become racialized? The paradoxical effects of education.” American Journal of Political Science 48.2: 374–​391. Huber, Gregory. A. and John Lapinski. 2006. “The ‘race card’ revisited:  Assessing racial priming in policy contests.” American Journal of Political Science 50: 421–​440; Filindra, Alexandra, and Noah J. Kaplan. 2016. “Racial resentment and whites’ gun policy preferences in contemporary America.” Political behavior 38,(2): 255–​275; Strother, Logan and Daniel Bennett. 2018. “Which Americans support the second amendment? The answer depends on whether blacks or whites have the guns.” Washington Post/​Monkey Cage; White, Ismail K. 2007. “When race matters and when it doesn’t: Racial group differences in response to racial cues.” American Political Science Review 101: 339–​354. 43 Tesler and Sears. Obama’s Race; Tesler. “In a Clinton–Trump match-​up.” 44 Tesler. “In a Clinton–Trump match-​up.”; Tesler, Michael 2012. “The spillover of racialization into health care: How President Obama polarized public opinion by racial attitudes and race.” American Journal of Political Science, 56(3); Henderson,

Racial Attitudes and American Politics  135 Michael and D.  Sunshine Hillygus. 2011. “The dynamics of health care opinion, 2008–​2010:  Partisanship, self-​interest and racial resentment.” Journal of Health Politics, Policy, and Law, 36(6): 945–​960; Kinder, Donald, and Jennifer Chudy. 2016. “After Obama.” The Forum 14(1) pp. 3–​15. 45 Knowles, Eric D., Brian Lowery and Rebecca L. Schaumberg. 2010. “Racial prejudice predicts opposition to Obama and his health care reform plan.” Journal of Experimental Social Psychology, 46: 420−423; Tesler. “The spillover”; Tesler, Michael. 2015. “The conditions ripe for racial spillover effects.” Political Psychology, 36(S1), 101–​117. 46 Reny, Tyler and Michael Tesler. 2018. “Racial attitudes and post-​Obamacare health insurance coverage.” Presented at the American Political Science Association’s Annual Meeting. 47 Nicholson, Stephen P., Chelsea M. Coe, Jason Emory, and Anna V. Song. 2016. “The politics of beauty: The effects of partisan bias on physical attractiveness.” Political Behavior 38(4):  883–​898; Alford, John R., Peter K.  Hatemi, John R.  Hibbing, Nicholas G. Martin, and Lindon J. Eaves. “The politics of mate choice.” The Journal of Politics 73: 362–​379; Bishop, Bill. 2009. The Big Sort: Why the Clustering of Like-​ minded America is Tearing us Apart; Bartels, Larry M. 2002. “Beyond the running tally:  Partisan bias in political perceptions.” Political Behavior 24(2):  117–​150; Hetherington, Marc, and Jonathan Weiler. 2018. Prius or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA: Houghton Mifflin. 48 Carmines and Stimson, Issue Evolution. 49 Valentino, Nicholas A., and David O.  Sears. 2005. “Old times there are not forgotten:  Race and partisan realignment in the contemporary South.” American Journal of Political Science 49: 672–​688; Kuziemko, Ilyana, and Ebonya Washington. 2018. “Why did the Democrats lose the south? Bringing new data to an old debate.” American Economic Review 108, (10):  2830–​ 2867; Acharya, Avidit, Matthew Blackwell, and Maya Sen. Deep Roots:  How Slavery Still Shapes Southern Politics. Princeton, NJ: Princeton University Press. 50 As noted elsewhere, the growing overtime “correlations between racial conservatism and Republican Party identification reached a plateau by the late 1990s in the Pew and GSS surveys. Moreover, the increased correlation between racial resentment and PID from 2000 to 2004 in the ANES was spurious, as this relationship actually weakened from 2000 to 2004 after [moral traditionalism] was taken into account.” Tesler. “In a Clinton–Trump match-​up.” p.  150. See also, Stimson, James A.  2004 Tides of Consent:  How Public Opinion Shapes American Politics. Cambridge: Cambridge University Press. 51 Tesler. “In a Clinton–Trump match-​up”; Sides et al. “Donald Trump and the rise of white identity politics”; Layman, Geoffrey C., and Thomas M. Carsey. 2002. “Party polarization and ‘conflict extension’ in the American electorate.” American Journal of Political Science 46(4): 786–​802. 52 Tesler. “In a Clinton–Trump match-​up”; Belcher, Cornell. 2016. A Black Man in the White House: Barack Obama and the Triggering of America’s Racial-​aversion Crisis. Water Street Press. 53 Tesler. “In a Clinton–Trump match-​up”; Tesler. “The return of old-​fashioned racism.” 54 Sniderman, Paul M.  and Thomas Piazza. 1993. The Scar of Race. Cambridge, MA:  Harvard University Press. Oliver, J.  Eric, and Tali Mendelberg. 2000.

136  Michael Tesler “Reconsidering the environmental determinants of white racial attitudes.” American Journal of Political Science 44(3): 574–​589. 55 Federico. “When do welfare attitudes”; Huber and Lapinski. “The ‘race card’ revisited”; Sears, David O., Colette van Laar, Mary Carillo, and Richard Kosterman. 1997. “Is it really racism? The origins of white Americans’ opposition to race-​ targeted policies.” Public Opinion Quarterly 61(1): 16–​53. 56 Tesler. “In a Clinton–Trump match-​up.” 57 Indeed, by the fall of 2016 only about one-​fifth of racially resentful Whites who did not attend college identified with the Democratic Party. 58 Sides et  al. “Donald Trump and the rise of white identity politics”; Mutz, Diana C. 2018. “Status threat, not economic hardship, explains the 2016 presidential vote.” Proceedings of the National Academy of Sciences. 59 See: Tesler. “In a Clinton–Trump match-​up,” F ­ igure 9.1. 60 For more on these results, see: Tesler, Michael. 2018. “Democrats and Republicans used to agree about the N-​word. Now they don’t.” Washington Post/​Monkey Cage, August 30. 61 For more on these results, see: Tesler, Michael. 2018. “Democrats and Republicans are increasingly divided on the value of teaching black history.” Washington Post/​ Monkey Cage, February 28. 62 Banks, Antoine J.  2014. Anger and Racial Politics:  The Emotional Foundation of Racial Attitudes in America. Cambridge: Cambridge University Press. 63 Valentino, Nicholas and Kirill Zhirkov. 2018. “Blue is black and red is white? Affective polarization and the racialized schemas of US party coalitions.” Working Paper, University of Michigan.

Chapter 6

Race, Ethnicity, and Public Opinion Erica Czaja and Vladimir E. Medenica

Political-science research on race and ethnicity is in the midst of an important shift. Traditionally, scholarship on race and ethnic politics has focused almost exclusively on minority populations.1 The study of race as a “puzzle” in need of solving has often been treated as unique to people of color, namely African Americans, Latinxs, and Asian Americans, and has considered white Americans as the baseline or reference group. In the weeks, months, and even year following the election of Donald Trump in 2016, however, analysts pored over election data, dissecting the event to better understand what drove voters, particularly white voters in Rust Belt states like Michigan and Wisconsin, to elect a candidate that few considered a serious contender for the presidency. The distinguishing factor, it turns out, was race.2 The unifying factors across Trump voters, more so than individual factors like age, education, class, and gender, were identifying as white and attitudes toward racial issues, including immigration. While work on understanding how race and ethnicity shape the political attitudes and behavior of marginalized groups continues to be important and necessary, less attention has been given to the role of race and ethnicity in the political lives and experiences of white Americans.3 Today’s political and social climate has prompted scholars to expand traditional understandings of race and ethnicity. Demographic change in the form of immigration from Latin America and Asia as well as the rise, election, and governance of Barack Obama, the United States’ first African American president, have increased the salience of race in political life and disrupted the racial status quo. Once defined racially by black and white, today the U.S. population is characterized by a multiplicity of racial and ethnic group divisions. Hispanics are now the largest minority population in the U.S., followed by African Americans and then Asian Americans and Native Americans.4 The “multi-​ racial” population—​a category formed by counting more than one racial group and allowed by the census since 2000—​is among the fastest-​growing groups.5 The vast majority of the newest Americans are no longer from Europe as they once were in the nineteenth century. Instead, today’s immigrants come primarily from Latin America and Asia, with the U.S. Asian population growing faster than any other immigrant group. While black migrants from Africa and the

138  Erica Czaja and Vladimir E. Medenica

Caribbean constitute a much smaller share of new immigrants, their presence creates important diversity within the racial category of black.6 Studying how these groups incorporate into American politics is of profound importance. Nevertheless, as the 2016 election demonstrates, such changes may also impact the attitudes and behavior of white Americans, which have been historically overlooked by scholars of race and ethnicity and instead have been viewed as being without race or race-​neutral. In this chapter, we take the increased racial and ethnic diversity of the United States as a starting point, and analyze the significance of race, ethnicity, and the group bases of political preferences. We begin with a discussion of categories of race and ethnicity in the U.S.  and argue that these categories are based not in “objective” biological differences, but rather, have been “socially constructed” or created by the institutions and practices of U.S.  government and society, which have assigned different meanings and values to various racial categories.7 Next we focus on individual-​level measurements of psychological attachment to groups—​group identity and consciousness—​as critical intervening variables between racial group classification and the formation of political preferences. The contours of the relationships between racial group identity, racial group consciousness, and individuals’ opinions, particularly for Latinxs, Asian Americans, and Whites, are especially challenging for public opinion researchers because of the heterogeneity within these populations and their politics. Finally, we discuss additional factors that may differentially influence the political opinions of individuals, depending in part on their racial group classifications and attachments, including party identification and mobilization, interpersonal contact and the racial, economic, and political context, and perceptions of, and experiences with, discrimination.

Categorizing Race and Ethnicity The practice of official racial classification in the U.S.  dates to the nation’s founding. The institution of slavery made information on racial categorization vital to the apportionment of legislative seats in the federal government. The now-​infamous “Three-​Fifths” compromise found in Article I  Section 2 of the U.S. Constitution specifies that both taxes and the number of elected representatives be calculated by adding the number of free persons and three-​ fifths of all other persons, “excluding Indians not taxed.” The free population was white while the enslaved population was black, hence the enumeration by slave status was also an enumeration by race. In every decennial census since the first in 1790, race has been recorded for each person counted. Political scientist Melissa Nobles demonstrates how government agencies such as the U.S. Bureau of the Census constructed categories of race in order to meet the social and political goals of the time.8 It would take almost 100 years and a bloody civil war for the United States to abolish slavery, but by then, the idea of race as a meaningful social distinction was embedded

Race, Ethnicity, and Public Opinion  139

in the fabric of the polity. Skin color only has more meaning in politics than, for example, eye color or other arbitrary physical differences between people, because political actors have given racial categories particular meanings in order to serve their purposes. This is what scholars mean when they say race is “socially constructed.” The practices of categorizing people based on race and recording race have continued unabated through the present day, and definitions of racial categories have continued to evolve. Moreover, for the vast majority of the nation’s history, racial categorization has gone hand in hand with preferential treatment for those recognized as white—​from citizenship and property rights to eligibility to vote. Political scientists have documented clear patterns of the role of the American state in the maintenance and definition of racial categories, unequal treatment by race, and the accompanying white privilege.9 These scholars argue that racial discrimination is deeply embedded in American political institutions and culture. Even when discrimination on the basis of racial categories was prohibited by law, as in the Fourteenth Amendment, state and local governments as well as private individuals found creative ways to use ostensibly race-​neutral practices and rules to exclude racial minorities from public life, beginning in the 1860s and continuing even today.10 Some scholars draw an important distinction between systemic structures of discrimination, or institutionalized racism, such as election rules that prevented African Americans from voting, and individuals’ feelings of racial antipathy, arguing both that the latter do not necessarily lead to the former and that institutionalized racism is what matters most for political outcomes.11 The long-​standing patterns of racial categorization and white privilege in the United States have persisted at the same time that the categories themselves have undergone change. Individuals at any point in time may be designated as part of a racial group, not because they are objectively Latinx or black but instead because of a combination of social and political constructions that work together to ascribe a specific category of race to the person. Especially relevant is the move among “white ethnics” during the period of mass immigration in the late nineteenth and early twentieth centuries to be classified by the government as white.12 Some groups such as the Irish, Italians, and Jews (once considered non-​white) were successful. Others, including Asian Americans, were not able to get the courts to recognize them as white and thus eligible for the full privileges of U.S. citizenship.13 Federal law prohibited Asian immigrants from naturalization until 1952, when more than 70  years of explicit Asian exclusion from the United States was ultimately abolished.14 From the 1860s, local, state, as well as the national government of the United States enacted laws targeting Asian Americans that barred property ownership, leveed additional race-​based taxes, and forcibly interned Americans of Japanese descent during World War II.15 Claire Kim explains these dynamics in her theoretical description of Asian Americans in U.S. society as “triangulated” between Blacks and Whites. According to Kim, Asian Americans have been (and continue to be)

140  Erica Czaja and Vladimir E. Medenica

valorized by Whites as superior to African Americans on some cultural and racial dimensions but at the same time have been deemed forever foreign and unfit for assimilation and civic membership with Whites. This triangulation racializes Asian Americans and African Americans in different ways and enables the majority racial group, white Americans, to maintain its dominant position over both minority groups simultaneously.16 Complicating matters further is the introduction by the federal government of a fourth major category, “Hispanic or Latino” ethnicity. While developed decades earlier, the requirement of reporting Hispanic/​ Latinx ethnicity along with other racial categories was implemented by the federal Office of Management and Budget in the 1970s. The complexity of racial categories suggests that researchers of racial and ethnic groups must utilize the terms white, black, Latinx, and Asian American carefully and with an awareness of the role that cultural norms and politics play in shaping individuals’ ideas of race.

Key Concepts, Measurement, and Methodology Key Concepts in the Study of Race and Groups Researchers are concerned with three key concepts: racial group membership, or what we have referred to above as racial categorization, racial group identity, and racial group consciousness. According to Paula McClain and her colleagues and to a long tradition of research, simple membership does not tell us how strongly a person identifies with a group or whether she views politics as relevant to the group.17 “Group identification refers to an individual’s awareness of belonging to a certain group and having a psychological attachment to that group based on a perception of shared beliefs, feelings, interests, and ideas with other group members;” whereas [g]‌roup consciousness is in-​group identification politicized by a set of ideological beliefs about one’s group’s social standing, as well as a view that collective action is the best means by which the group can improve its status and realize its interests.18 The more strongly that society and politics define group members by their racial category, and the more isolated and discriminated against that people are because of their assigned group membership, the more likely they are to identify with their assigned racial group. These conditions increase the likelihood that group members will view their unequal treatment as a result of politics as well as the likelihood that they will thus organize for political change. Paradoxically then, the very conditions that stifle individuals can facilitate political mobilization of the group. The interactions between oppressive institutions and the politics of non-​ white minority individuals are well illustrated by another set of key concepts

Race, Ethnicity, and Public Opinion  141

developed by Michael Dawson in the study of African American politics: linked fate, the black counterpublic, and the black utility heuristic.19 Linked fate is the idea among African Americans that individual well-​being is inextricably linked with the fate of the race as a whole; essentially, they believe that their success depends on the success of the group, so what is good for the race is good for the individual. According to Dawson, African Americans’ unique history of racial subjugation and forced segregation has led to the transmission of notions of linked fate across generations, so that still today African Americans continue to receive messages that reinforce their sense of shared racial group interests through the black counterpublic—​mainly black media, predominantly black organizations, and the black church. Information shared in these segregated spaces, Dawson argues, enables and encourages African Americans to evaluate politics using a rational, mental shortcut that he calls the “black utility heuristic.” That is, African Americans form their political opinions about political parties, candidates, and public policies by using their perceptions of what is best for the entire racial group instead of what they think is best for them individually. The sense of linked fate is so strong that it overcomes the force of class interests for the black middle class and the lure of cultural conservatism, which resonates with many African Americans. According to Dawson, linked fate explains why African Americans vote nearly unanimously for the Democratic Party in presidential and many lower-​level electoral contests. Researchers interested in Asian American, Latinx, and even white Americans are beginning to use both sets of concepts, but, we argue, should do so with care because of the different historical and contemporary experiences of racial groups. Today, for example, Asian Americans and Latinxs are typically much closer to the immigration experience that helps shape political incorporation. Michael Jones-​Correa’s study of first-​generation Latinx immigrants in Queens, New York suggests that there are important psychological and material costs in renouncing homeland citizenship that prevent some immigrants from becoming citizens.20 He argues that Latinxs practice a “politics of in-​between,” being torn between two nations, neither fully politically engaged in their new homes nor in their homelands. However, beyond such individual factors, he as well as others also identified a lack of institutional mechanisms to aid in the incorporation of immigrants, including exclusive local party machines.21 Latinx organizations, such as churches, may be evolving to play an increasingly political role that could strengthen Latinxs’ identification with all Latinxs rather than merely their national origin group (e.g., Mexicans) and enhance their sense of linked fate and group consciousness. Scholars have recently identified relatively high levels of linked fate among white Americans, though less is known about how linked fate operates with respect to individual political attitudes and behavior.22 As a consequence of current social and political trends, there are reasons to believe that linked fate is growing both more common and, as the election of 2016 suggests, more politically consequential for Whites.23 At the same time, the political history

142  Erica Czaja and Vladimir E. Medenica

and experience of white Americans is so fundamentally different from that of African Americans in the United States that what linked fate means for Whites remains a very open question. Nevertheless, many Whites, regardless of characteristics like age and partisanship, racially identify as white, believe that Whites are discriminated against, see their fate as tied to the fate of Whites overall, and share common political preferences and behavior.24 Examining the effects of linked fate on the political outcomes of white Americans in the post-​ Trump era, especially as social and cultural diversification trends continue to accelerate toward a minority–​majority future, is an area ripe for study.

Measurement of Key Concepts There are a number of important challenges in the measurement of racial group membership, identity, and consciousness, and we highlight two of the primary challenges here. The first challenge arises because, as detailed above, group identity and consciousness are flexible for individuals, depending on context. In addition, historical, political, and social forces shape what it means to be a member of particular racial and ethnic groups. In-​person and telephone survey interviews are the most common ways to measure these concepts in the study of public opinion, but different individuals understand questions about race and ethnicity differently.25 Second, the survey questions that attempt to measure group-​based identities vary widely in their wording across surveys, making comparisons between groups and at different time periods difficult. Furthermore, the context in which the survey is administered, such as whether the interview is conducted by a same-​race interviewer or during an election campaign, can influence the racial identity and level of group consciousness that respondents report. In addition, there is a range of national origin groups that make up the pan-​ethnic categories of Latinx and Asian American, and therefore, whether respondents identify with their country of origin or with a broad pan-​ethnic category depends upon the options given to respondents in surveys. While Mexican-​ Americans make up the largest share of the Latinx population in the U.S., the category of Hispanic or Latino also includes Cubans, Caribbeans, Puerto Ricans, and people from other Latin American countries. Similarly, there are as many national origin and ethnic groups within the pan-​ethnic racial category of Asian American, with the six largest groups being Chinese, Asian Indian, Filipino, Vietnamese, Korean, and Japanese. Finally, while Blacks demonstrate the highest degree of racial group consciousness, the internal diversity of this group is also in flux, with U.S.-​born African Americans included in the same racial category as new arrivals from the African continent as well as large numbers of Afro-​Caribbeans. Thus, differences in the ways in which individuals understand the same questions, differences in the ways that survey questions are worded, and the contexts in which these questions are administered complicate the

Race, Ethnicity, and Public Opinion  143

measurement and comparison of group membership, identity, and consciousness across racial groups.

Methodological Challenges in Survey Research There are also methodological challenges in collecting data on race in the United States. First, identifying and recruiting responses from marginalized groups in the United States can be difficult, particularly those that are heavily comprised of immigrants. Geographic concentration and dispersion and the prevalence of speaking a language other than English among Latinx and Asian American populations today requires innovative methods of survey research designed specifically for these respondents to effectively reach individuals for interviews. Second, asking people about their racial attitudes can be a fraught exercise. Gathering honest and accurate responses requires researchers to broach sensitive topics, like race, with care and strategy in order to minimize social desirability bias or the tendency of respondents to answer questions in a way that they believe will be viewed favorably but does not necessarily reflect their true attitudes. Indeed, some analysts have posited that pollsters were inaccurate in their 2016 election forecasts largely due to Trump supporters being less likely to be contacted for surveys and more likely to exhibit social desirability bias in their responses. Asian Americans and Latinxs, and immigrant groups more generally, have increasingly complex patterns of geographic mobility. Once heavily concentrated in the Southwestern United States and large urban metropolitan areas such as Los Angeles and New York City, Latinxs are moving in increasing numbers to the South, the mid-​Atlantic, and the plains states.26 Asian Americans, while once heavily concentrated in a handful of states, are beginning to disperse as well, with sizable populations in states such as Virginia, Florida, and Nevada. Sampling these populations for survey interviews is challenging, but making sure that respondents are not drawn only from high-​density locations is critical for obtaining survey samples that are representative of the population. Similarly, because nearly three-​quarters of Asian American adults27 and approximately one in three Latinx adults are foreign-​born,28 writing surveys in languages other than English and hiring interviewers who can speak in the respondents’ native languages greatly increase the likelihood of acquiring both good samples and good data. While many immigrants speak English, it is a second language for many, and answering survey questions in their native languages is preferable.29 Finally, given the high degree of internal heterogeneity within each of these groups, the size of the sample must be large enough to include sufficient numbers of respondents from specific national origin groups. For example, Mexican-​Americans and Cuban-​Americans not only have different migration histories to the United States, but they are also distinctive in their political beliefs.30 National origin groups within the pan-​ethnic rubric of Asian Americans demonstrate similar differences.31

144  Erica Czaja and Vladimir E. Medenica

Racial Group Identity and Racial Group Consciousness There is no simple way to characterize the multiplicity of identities of Americans classified as racial minorities today. Moreover, the political influence of group identity and group consciousness may differ across racial categories and individuals. In this section, we examine the individual and contextual antecedents that impact racial group identity and consciousness as well as the ways in which racial identity and consciousness affect political attitudes.

Explaining Group Identity and Consciousness Several recent studies demonstrate the contextual nature of both group identity and group consciousness and the ways in which they operate differently for different groups. First, with respect to group identity, the labels “Asian American” and “Latino” are the least frequent identifiers adopted by group members themselves: only 19 percent of Asians and 24 percent of Latinxs say that they identify with these pan-​ethnic labels, while 62 percent of Asians32 and 51 percent of Latinxs33 self-​identify with their country of origin. These results might suggest that the country of origin is more central to the group identities of Asians and Latinxs living in the United States than the pan-​ ethnic identifiers of Asian American and Latinx. However, in an earlier study, Pei-​te Lien and colleagues found that when respondents who did not immediately self-​identify as Asian American were asked the follow-​up question, “Have you ever thought of yourself as an Asian American?” approximately 50 percent of respondents provided an affirmative response, illustrating that racial identification is a complex choice for group members, not a fixed, objective membership classification.34 These results highlight the multiple identity options for Asian Americans, and minority groups more generally, as well as the possibility of adopting different identities at different times. Secondly, with respect to racial group consciousness, the evidence suggests that environmental cues can play a role in whether one’s racial group identity becomes politicized. Jane Junn and Natalie Masuoka conducted a survey experiment intended to uncover the potential effects of descriptive representation—​ that is, representation by an elected official who shares a particular demographic characteristic, in this case race—​on African American and Asian American racial group consciousness.35 In the experiment, half of the participants in each racial group were randomly assigned to a treatment condition in which they were exposed to photographs and brief biographies of U.S. presidential cabinet members who shared their race while the remaining participants in each racial group were not. Junn and Masuoka hypothesized that African-​Americans’ typically high levels of group consciousness would be unlikely to increase much further as a result of cuing descriptive representation in the treatment condition. However, they expected Asian-​ Americans’ group consciousness, though lower than

Race, Ethnicity, and Public Opinion  145

that of African-​Americans overall, to be more malleable in response to contextual cues that remind them of “the political consequences of being Asian American,” such as exposure to same-​race political actors. They found Asian Americans who received the descriptive representation treatment scored significantly higher on measures of racial group consciousness than the control group of Asian Americans. Asian American respondents who were exposed to the treatment were more likely than control subjects to agree that their individual fates are linked to those of Asian Americans as a group and to say that being Asian/​Asian American is at least “somewhat important” to their political identity and beliefs.36 The treatment condition resulted in similar but weaker effects among African Americans, confirming Junn and Masuoka’s expectations that African Americans would be difficult to move any further since this group is already highly race-​conscious. These results support their contention that racial groups have different levels of racial group consciousness and, as a result, are influenced by the political environment to varying degrees. In his study of mayoral elections in five major U.S.  cities, Matt Barreto provides evidence that a similar latent group consciousness may operate among Latinxs.37 He compared consecutive mayoral elections in Houston, Los Angeles, San Francisco, Denver, and New York—​one in which a competitive Latinx candidate was on the ballot and one in which a Latinx candidate was not—​in order to test whether Latinx candidates would be more likely than non-​ Latinx candidates to mobilize Latinx voters. He finds that “[p]‌recincts with larger proportions of Latino registrants were more likely to evidence high rates of turnout when a Latino candidate was running for office.”38 Ethnic and racial identity may be a critical factor enabling racial minorities to overcome their relative disadvantage in resources such as education, employment, and interest in politics, which have proven crucial for participating in politics.39 Descriptive representation may activate and politicize these identities and help to level the political playing field. Beyond candidate co-​ethnicity, numerous other features of contemporary campaigns heighten Latinx voters’ awareness of their ethnic identity “in a way that directly connects Latino identity with politics.”40 Personalized mobilization of Spanish-​surname voters, targeted ads stressing the immigrant experience, Spanish-​ language campaign materials, and candidate endorsements by well-​known Latinxs may all serve to mobilize and engage Latinx voters.41 In addition, Barreto and Pedraza argue that a steady stream of immigration from Latin America anchors Latinx identities in the immigrant experience and garners popular attention for Latinxs, including negative attention in the form of discriminatory public discourse and policies.42 All of this serves to further politicize Latinx identity and elevate Latinx group consciousness, an effect we noted earlier with regard to African Americans’ experiences of racial discrimination. Finally, the socioeconomic context in which racial and ethnic minorities live matters. For example, Gay found that the lower the quality of one’s

146  Erica Czaja and Vladimir E. Medenica

neighborhood in terms of the maintenance and value of homes, cleanliness and safety of streets, and accessibility of public and private services like reliable trash removal and grocery stores, the higher was African Americans’ sense of racial group consciousness and linked fate.43

Diversity vs. Solidarity in Group Identity and Consciousness As noted earlier in this chapter, the study of the political impact of racial group consciousness began with Michael Dawson’s seminal work on racial linked fate within the African American population.44 Dawson’s work has been used to explain the apparent homogeneity in political opinions within the black community across other lines of difference, such as class, and to explain African Americans’ near universal support for the Democratic Party since the mid-​1960s. However, Cathy Cohen argues that the notion of linked fate itself is limited and that a more accurate characterization of the political positioning of most black Americans is that of a qualified linked fate, whereby not every black person in crisis is seen as equally essential to the survival of the community, as an equally representative proxy of our own individual interests, and thus as equally worthy of political support by other African Americans.45 Cohen demonstrates the consequences of this qualified linked fate through her in-​depth study of the African-​American political response to the HIV/​AIDS crisis in the 1980s and early 1990s. She focuses on the actions of black media, organizations, and leaders in New York City, and finds that, despite eventually acknowledging that AIDS severely affects many in the black community and attempting to provide services for afflicted individuals, these black elites ultimately failed to transform most African Americans’ thinking about the disease. African Americans do not view AIDS as a “black issue,” or an issue of primary importance to the black community, which Cohen calls “consensus issues.” Nor are those living with AIDS in the black community “embraced and ‘owned’ as essential members of the group.”46 Generalizing beyond the HIV/​AIDS case, Cohen contends that black politics has historically been focused on consensus issues, but increasingly, cross-​cutting issues relating to the particular concerns of vulnerable or stigmatized subpopulations within the black community—​usually along the lines of class, gender, and sexuality—​are competing for a place on the black political agenda.47 Cohen’s study challenges us to think more carefully about how racial minority groups address internal differences and inequality within the group, highlighting the complexities of politicized group consciousness and its dependency both on context for activation or development and on the subpopulation and issue area to which it is applied.

Race, Ethnicity, and Public Opinion  147

Building on Dawson’s historical account of the heterogeneity of black ideological traditions,48 Melissa Harris-​Lacewell examines the adult socialization processes that occur in the contemporary black counterpublic—​including social spaces like barbershops, churches, and media outlets. She demonstrates that ordinary African American citizens make sense of the world and form “identifiable patterns of public opinion that can be understood as ideologies” through processes of “everyday talk.”49 In the segregated spaces of the black counterpublic, African Americans can feel free to candidly talk to each other “beyond the gaze of racial others,” particularly Whites, and this conversation serves to socially (re)construct a variety of unique black worldviews.50 Harris-​ Lacewell identifies four black political ideologies that continue to operate today: Black Conservatism, Liberal Integrationism, Black Feminism, and Black Nationalism. While there are similarities between these ideologies and the traditional liberal-​conservative spectrum used in survey research (developed to understand white ideology), the relevant difference between the two overall frameworks is in whether there is a deliberate recognition of race as politically salient. Whereas the white ideological spectrum is, on its face, race-​neutral, Harris-​Lacewell argues that all of the black political ideologies are built upon a kind of black race consciousness or notion that being black matters politically, which she calls “black common sense.”51 Exactly how one believes that being black matters is proscribed by one’s ideology. Work on other racial groups also emphasizes the important types of diversity within each group. Abrajano, for example, argues that Latinxs who speak English are more politically knowledgeable and orient more toward the substance of issues in political campaigns, while Latinxs who speak only Spanish are more oriented toward easily digestible cues to their ethnic identity, such as Spanish-​language campaign materials and co-​ethnicity of political candidates.52 More generally, some scholars raise questions about the downside of group solidarity and political unanimity. Blacks have been called a “captured” group with the Democratic Party because they are the most loyal Democratic voters and their votes can be taken for granted. Thus, African Americans lack the influence that comes with the credible threat of switching their votes to the other party.53 Latinxs vote Democratic but in less consistent and uniform numbers, and this may give them leverage to get more of what they want from politics.54 In addition, when group membership becomes a simplistic cue, it can sometimes produce support for co-​ethnic leaders or for parties at odds with what voters would choose if they were fully informed and voting in their own best interests.55

What Influences Public Opinion? Among the multiple facets of public opinion and factors influencing political attitudes, we focus on:  1) party identification and mobilization, 2)  interpersonal contact and the racial and economic context, and 3) perceptions of and

148  Erica Czaja and Vladimir E. Medenica

experiences with discrimination. It is crucial to consider how and why the same antecedents might work in distinctive ways for different groups.

Partisan Identification Scholars have consistently identified partisanship as the most enduring, stable and powerful of all political predispositions.56 For white Americans, party identification is produced through an early emotional attachment to one party or the other, often learned through childhood socialization in the home or other institutions.57 The available evidence indicates that (overwhelmingly Democratic) partisanship is acquired through similar processes of institutional socialization for African Americans, though for this group, partisanship appears to be more instrumental and group-​interested than affective.58 It is unclear how immigrant-​based racial groups acquire partisanship when often their early and even adult political socialization does not occur in the United States and, as demonstrated by the work of Rogers59 and Jones-​Correa,60 they encounter numerous barriers to institutional incorporation once in the U.S. Wong61 argues that the longer an immigrant resides in the U.S., the greater political exposure they will have, the more likely they are to become a citizen, and the more likely they are to learn English proficiently; thus, the more likely they will be to identify with one of the political parties. Party mobilization (or lack of it) also seems to be a pivotal factor in whether and how immigrant groups are incorporated into the American polity. Being ignored or excluded by local political parties discourages naturalization,62 which in turn depresses the acquisition of partisanship, while becoming a citizen and being brought into the fold by the political parties encourages immigrants to adopt a partisan identification, likely that of whichever party is most welcoming.63

Race Relations Two primary hypotheses have been advanced to explain the impact of cross-​ racial exposure: the threat hypothesis and the contact hypothesis. According to the threat hypothesis, greater exposure between members of different races will increase negative attitudes and worsen race relations. But the contact hypothesis argues the opposite—​greater contact reduces prejudices and improves race relations. Classical formulations of the threat hypothesis predict that dominant groups will perceive increasing threats to their political and economic privileges as the population of subordinate group members in the immediate environment increases. As threats to resources increase, so too do dominant group hostilities toward subordinate groups.64 The contact hypothesis, on the other hand, predicts improved attitudes and cooperative relations through interpersonal contact as long as certain ideal conditions for the interactions are met, including equal status among individuals and shared goals.65

Race, Ethnicity, and Public Opinion  149

Researchers have attempted to adjudicate between the threat and contact hypotheses through careful study of the environmental context, as well as the relative size of groups at both the neighborhood and metropolitan levels. Oliver and Mendelberg find that the size of the African American population is unrelated to white racial attitudes at the neighborhood level, though it is moderately related to Whites’ anti-​black stereotypes at the metropolitan level. However, the strongest contextual effects come, not from racial composition, but from neighborhood educational composition, which they argue is a measure of white economic vulnerability. Whites living in economically vulnerable contexts are not only more prejudiced against African Americans but are also more anti-​Semitic and authoritarian than less economically vulnerable Whites. They attribute this generalized out-​group hostility to the psychological stresses of living in economically vulnerable environments and suggest that in the specific racial context of the U.S., such generalized out-​group hostility is most often directed at African Americans. Considering the rapidly changing racial topography of the U.S., future research should explore the impact of these psychological stresses on attitudes toward other racial minorities as well. Contrary to Oliver and Mendelberg’s findings, Claudine Gay finds that the overall economic conditions of a neighborhood do not influence African Americans’ expressions of anti-​Latinx prejudice.66 Instead, it is the relative economic positions of the two racial groups that matter. That is, African Americans who shared neighborhoods with economically advantaged Latinxs exhibited more prejudice against Latinxs, were less supportive of “special preferences in hiring and promotion” for Latinxs than they were for themselves, and agreed more with the statement “more good jobs for Latinos means fewer good jobs for Blacks.”67 Both racial prejudice and unsupportive policy attitudes intensified somewhat as the size of the Latinx population increased but only in contexts of Latinx economic advantage. When African Americans were better off than, or economically equal to, their Latinx neighbors, neighborhood economic conditions had no impact on Blacks’ attitudes toward Latinxs. These results lend partial support to both the threat and contact hypotheses, demonstrating that exposure produces threat and worsens race relations under conditions of economic inequality and also suggest that interpersonal contact may indeed only be effective at improving race relations under conditions of equality, as the contact hypothesis predicts. Using interview data from all four of the primary racial groups while analytically distinguishing between smaller neighborhood and larger metropolitan contexts, Oliver and Wong find that—​among Whites, African Americans, and Latinxs—​the more integrated the neighborhood (and thus the more opportunities for contact) the less hostility residents expressed toward racial out-​ groups.68 These effects were most apparent in metropolitan areas in which there were large populations of racial out-​groups. For example, Oliver and Wong compared the attitudes toward Latinxs of African American and white residents of Atlanta, where the population of Latinxs is relatively small, and

150  Erica Czaja and Vladimir E. Medenica

Los Angeles, where the population of Latinxs is relatively large. They found that African Americans and Whites living in less integrated, racially homogeneous neighborhoods in Los Angeles displayed much higher rates of anti-​ Latinx sentiment than their counterparts in Atlanta, and the researchers attribute this difference to the larger Latinx population in the L.A. metropolis relative to Atlanta.69 Asian Americans in Oliver and Wong’s study who were interviewed in English followed a similar pattern. However, Chinese and Korean respondents who were interviewed in their native languages reported greater prejudice when living in more integrated neighborhoods. The authors speculate that these findings may be related to the lower level of incorporation that non-​ English-​speaking Asian Americans experience, or possibly to the violence in Los Angeles against Asian-​American small businesses that occurred shortly before the survey was administered. Finally, Daniel Hopkins extends research on the threat hypothesis to long-​ time residents’ attitudes toward recent immigrants and highlights another important factor beyond the size of local and metropolitan populations:  the role of the national media in drawing attention to and politicizing issues as threatening political problems. Hopkins advances and finds support for what he calls the “politicized places hypothesis,” which predicts that residents of communities in which the number of immigrants has grown will perceive immigrants as threatening and adopt anti-​immigrant attitudes primarily when there are threatening, anti-​immigration cues in the national media environment.70 According to Hopkins’ research, residents are not threatened simply by increasing numbers of immigrants in their local communities if increasing immigration has not been politicized as a problem by political elites at the national level.71 Research on the emotion of anxiety by Shana Gadarian and Bethany Albertson supports Hopkins’ hypothesis and demonstrates how the effects of threat politicized by the national media might multiply. Gadarian and Albertson find that once people are made to feel anxious about immigration, they will seek out threatening immigration news with greater frequency than people who are not anxious about immigration. Biased information-​seeking among people who encounter threatening cues about immigration in the media likely further reinforces the negative effects on immigration attitudes identified by Hopkins.72 The results of all of these studies highlight the important role that is played by context—​at the neighborhood, city, and national levels—​in influencing the effects of integration and contact on relations between groups. Not only do the relative sizes and geographic distributions of racial and ethnic groups in neighborhoods and cities matter, but economic conditions, the relative economic positions of groups, and national politics played out in the media all impact how people respond to the increased presence of out-​group others in the neighborhoods and cities in which they reside.

Race, Ethnicity, and Public Opinion  151

Discrimination Dennis Chong and Dukhong Kim’s “theory of opportunities” echoes our theme, that “[t]‌he assimilation of a minority group into American society depends not only on the actions of group members but also on the reception accorded that group by the majority population.”73 Specifically, Chong and Kim ask why members with higher economic status sometimes continue to have strong racial group consciousness. They find that the effects of class will depend upon racial group members’ perceptions of opportunities for social mobility—​beliefs about their chances of moving up in the world.74 At the group level—​that is, looking at between-​group differences among African Americans, Asian Americans, and Latinxs—​Chong and Kim find that economic status has the smallest effect on African Americans’ levels of group consciousness. They find that support for policies that benefit the group is least affected by improved economic fortunes for African Americans, relative to other racial groups, because of frequent experiences with discrimination and perceptions that Blacks have fewer opportunities relative to Whites. In contrast, improved economic status for Asian Americans and Latinxs is often accompanied by fewer experiences with discrimination and a more positive outlook on U.S. society, making increased economic status for these groups a significant predictor of diminished support for racial group interests.75 Chong and Kim find the same dynamic at work at the individual level. In other words, when they focus on the between-​person differences within each of the three racial minority groups, they find that economic status has no effect on support for group interests for minority individuals who frequently experience discrimination and perceive unequal opportunities. On the other hand, among minority individuals who have little experience with discrimination and believe that U.S. society does offer equal opportunities for all, high economic status reduces support for group interests.76 Importantly, Chong and Kim’s research contradicts earlier scholarship on black public opinion. Sigelman and Welch found that African Americans’ perceptions of group discrimination influenced their views about the sources of racial disparities, and both of these perceptions and explanations influenced the policy solutions that African Americans preferred to remedy racial inequality.77 Furthermore, they found that African Americans perceived much higher levels of discrimination against Blacks as a group than they reported experiencing personally, and as such, personal experiences with discrimination had little effect on their attitudes. But why do personal experiences with discrimination impact African Americans’ opinions in Chong and Kim’s 2006 study but not in that of Sigelman and Welch in 1991? In answering this question, it is critical to look at the ways in which the different pairs of researchers measured personal discrimination. Chong and Kim used a combination of seven wide-​ranging questions to measure respondents’ levels of perceived discrimination, including questions

152  Erica Czaja and Vladimir E. Medenica

that ask whether respondents have experienced discrimination in the past ten years or have ever been “physically threatened or attacked” or “unfairly stopped by police.”78 Respondents in Chong and Kim’s 2006 study were also asked about the frequency with which they have been given “less respect” and “poorer service” (while shopping or dining) than others, as well as about how often people insult or call them names or seem fearful of them because of their race. In contrast, Sigelman and Welch used four questions about basic “quality of life” issues, which they acknowledged were “fairly crude,” including whether respondents had ever been discriminated against in getting “quality education” and “decent” housing, jobs, and wages.79 Sigelman and Welch astutely note that their measurements “ignore possible discrimination in the daily routines of life,” like shopping, eating at restaurants, and interacting with others in the community.80 As Chong and Kim’s measures highlight, Sigelman and Welch’s research also fails to capture discrimination at the hands of state actors like the police. The factors that influence others’ perceptions of discrimination against out-​ group members are also important to understand because of the consequences these perceptions have for public opinion about policies intended to benefit racial minority groups. Whites’ belief that Blacks are discriminated against is positively correlated with white support for a range of policies that serve to ameliorate racial inequality, like affirmative action, as well as less race-​ conscious policies.81 Believing that African Americans, Latinxs, and Asian Americans continue to be discriminated against goes hand in hand with support for policies intended to benefit all racial minorities, including job training, educational assistance, and preferential hiring and promotion programs, among white, African American, Latinx, and Asian American respondents.82

Concluding Remarks We began our review of research in political science on race, ethnicity, and the group bases of public opinion by describing the complexity and the socially constructed nature of racial categories in the United States. Despite the inherent difficulties in measuring these concepts, race and ethnicity remain among the most important divisions in political attitudes among Americans. To better understand the group bases of public opinion, researchers have attempted to define, measure, and examine the three key concepts of racial group membership (what we have referred to as racial categorization), racial group identity, and racial group consciousness. Most scholarship has focused on one of the four primary racial groups: Whites, African Americans, Latinxs, and Asian Americans. Michael Dawson developed the concept of linked fate from the experiences of African Americans in U.S. politics. This idea has been influential in scholarship in racial and ethnic politics; however, the extent to

Race, Ethnicity, and Public Opinion  153

which the concept is applicable to other populations facing different political circumstances, including Whites, Asian Americans and Latinxs, requires further examination. Differences in the ways in which individuals understand the same questions on surveys, the distinctive contexts in which surveys are administered, and difficulties in gathering honest responses to sensitive survey topics complicate the measurement and comparison of group membership, identity, and consciousness across groups. We conclude that the contours of the relationships between racial group identity, racial group consciousness, and public opinion are not well understood because of the dynamic nature of racial group populations and the still-​ early stage of systematic research. Particularly for members of pan-​ethnic racial groups, identification is a complex choice. For all racial and ethnic groups, membership and identity are fluid and primarily based on the forces of politics and the circumstances of society. They are not fixed or objective. Within all groups, there are important tensions between unity and difference, favored status and marginalization. Finally, we reviewed three widely studied predictors of public opinion, including party identification, race relations, and perceptions of discrimination. In terms of race relations, the mixture of resident groups in neighborhoods and cities along with the local economic and national political contexts in which they occur are key to understanding public opinion related to racial and ethnic out-​groups. Regarding discrimination, it is crucial to understand how the discriminatory treatment directed at a group member, and her interaction with society, vary systematically as a function of her group membership. The different historical and current circumstances of groups explain the varied outlooks their members adopt on individual opportunity.

Notes 1 Michael Dawson and Cathy J. Cohen, “Problems in the Study of the Politics of Race,” in Political Science:  The State of the Discipline III, eds. Ira Katznelson and Helen V. Milner (Washington, DC: American Political Science Association, 2003). 2 Vladimir E. Medenica, “Race and Millennials in the 2016 Election,” Journal of Race, Ethnicity, and Politics 3(1), (2017): 55–​76; Diana Mutz, “Status Threat, not Economic Hardship, Explains the 2016 Presidential Vote,” Proceedings of National Academies of Sciences (2018): 1–​10; Brian F. Schaffner, Matthew Macwilliams, and Tatishe Nteta, “Understanding White Polarization in the 2016 Vote for President: The Sobering Role of Racism and Sexism,” Political Science Quarterly 133 (2018): 9–​34. 3 Nicholas Weller and Jane Junn, “Racial Identity and Voting: Conceptualizing White Identity in Spatial Terms,” Perspectives on Politics 16(2), (2018): 436–​448; Cara Wong and Grace E. Cho, “Two-​Headed Coins or Kandinskys: White Racial Identification,” Political Psychology 26(5), (2005): 699–​720. 4 “United States Census Bureau State & County QuickFacts,” last modified July 8, 2014, http://​quickfacts.census.gov/​qfd/​states/​00000.html.

154  Erica Czaja and Vladimir E. Medenica 5 Natalie Masuoka, “Political Attitudes and Ideologies of Multiracial Americans: The Implications of Mixed Race in the United States,” Political Research Quarterly 61(2), (2008): 253–​267; Kim M. Williams, Mark One or More: Civil Rights in Multiracial America (Ann Arbor, MI: The University of Michigan Press, 2006). 6 Reuel Rogers, Afro-​Caribbean Immigrants and the Politics of Incorporation: Ethnicity, Exception or Exit (Cambridge, MA: Cambridge University Press, 2006); Yvette Alex-​ Assensoh, “African Immigrants and African-​Americans: An Analysis of Voluntary African Immigration and the Evolution of Black Ethnic Politics in America,” African and Asian Studies 8 (2009): 89–​124. 7 Michael Omi and Howard Winant, Racial Formation in the United States: From the 1960s to the 1990s (New York: Routledge, 1994); Anthony W. Marx, Making Race and Nation: A Comparison of the U.S., South Africa, and Brazil (New York: Cambridge University Press, 1998); David Roediger, The Wages of Whiteness:  Race and the Making of the American Working Class (New York: Verso Books, 1999). 8 Melissa Nobles, Shades of Citizenship:  Race and the Census in Modern Politics (Stanford, CA: Stanford University Press, 2000). 9 Ira Katznelson, When Affirmative Action Was White:  An Untold History of Racial Inequality in Twentieth-​Century America (New  York:  W. W.  Norton & Company, 2005); Rogers Smith, Civic Ideals:  Conflicting Visions of Citizenship in U.S. History (New Haven, CT: Yale University Press, 1997); Daniel J. Tichenor, Dividing Lines: The Politics of Immigration Control in America (Princeton, NJ:  Princeton University Press, 2002). 10 Rachael V. Cobb, D. James Greiner, and Kevin M. Quinn, “Can Voter ID Laws Be Administered in a Race-​Neutral Manner? Evidence from the City of Boston in 2008,” Quarterly Journal of Political Science 7, (2012): 1–​33. 11 Paul Frymer, Black and Blue:  African Americans, the Labor Movement, and the Decline of the Democratic Party (Princeton, NJ: Princeton University Press, 2008). 12 Victoria Hattam, In the Shadow of Race: Jews, Latinos, and Immigrant Politics in the United States (Chicago, IL: The University of Chicago Press, 2007); Noel Ignatiev, How the Irish Became White (New York: Routledge, 1995); Matthew Frye Jacobson, Whiteness of a Different Color:  European Immigrants and the Alchemy of Race (Cambridge, MA: Harvard University Press, 1999); Ariela Gross, What Blood Won’t Tell:  A History of Race on Trial in America (Cambridge, MA:  Harvard University Press, 2008). 13 Ian Haney Lopez, White by Law:  The Legal Construction of Race, revised and expanded edition (New York: New York University Press, 2006). 14 Roger Daniels, Guarding the Golden Door:  American Immigration Policy and Immigrants since 1882 (New York: Hill and Wang, 2004). 15 Mae Ngai, Impossible Subjects:  Illegal Aliens and the Making of Modern America (Princeton, NJ: Princeton University Press, 2004). 16 Claire Kim, “The Racial Triangulation of Asian Americans,” Politics & Society 27(1), (1999): 105–​138. 17 Paula McClain, Jessica D. Johnson Carew, Eugene Walton, Jr., and Candis S. Watts, “Group Membership, Group Identity, and Group Consciousness: Measures of Racial Identity in American Politics?” Annual Review of Political Science 12 (2009): 471–​485. 18 McClain et  al. “Group Membership, Group Identity, and Group Consciousness: Measures of Racial Identity in American Politics?” Annual Review of Political Science 12 (2009): 474, 476, emphasis in original.

Race, Ethnicity, and Public Opinion  155 19 Michael Dawson, Behind the Mule:  Race and Class in African American Politics (Princeton, NJ: Princeton University Press, 1994). 20 Michael Jones-​Correa, Between Two Nations: The Political Predicament of Latinos in New York City (Ithaca, NY: Cornell University Press, 1998). 21 Jones-​Correa, Between Two Nations, 1998; Rogers, Afro-​Caribbean Immigrants and the Politics of Incorporation, 2006. 22 Deborah J. Schildkraut, “White Attitudes About Descriptive Representation in the US:  The Roles of Identity, Discrimination, and Linked Fate,” Politics Groups, and Identities 5 (2017): 84–​106. 23 Ashley Jardina “Demise of Dominance:  Group Threat and the New Relevance of White Identity for American Politics,” Ph.D. dissertation (Department of Political Science, University of Michigan, 2014). 24 Vladimir E.  Medenica, “Race and Millennials in the 2016 Election,” Journal of Race, Ethnicity, and Politics 3(1), (2018):  55–​76; Deborah J.  Schildkraut, “White Attitudes”; Deborah J.  Schildkraut and Satia A.  Marotta, “Assessing the Political Distinctiveness of White Millennials:  How Race and Generation Shape Racial and Political Attitudes in a Changing America,” RSF: The Russell Sage Foundation Journal of the Social Sciences, 4(5), (2018): 158–​187. 25 Taeku Lee, “Race, Immigration, and the Identity-​to-​Politics Link,” Annual Review of Political Science 11 (2008):  457–​478; Dennis Chong and Reuel Rogers, “Racial Solidarity and Political Participation,” Political Behavior 27(4), (2005): 347–​374. 26 Paula D.  McClain, Niambi M.  Carter, Victoria M.  DeFrancesco Soto, Monique L.  Lyle, Jeffrey D.  Grynaviski, Shayla C.  Nunnally, Thomas J.  Scotto, J.  Alan Kendrick, Gerald F. Lackey, and Kendra Davenport Cotton, “Racial Distancing in a Southern City: Latino Immigrants’ Views of Black Americans,” Journal of Politics 68 (2006): 571–​584; Rogelio Saenz, “Latinos and the Changing Face of America,” in The American People:  Census 2000, eds. Reynolds Farley and John Haaga (New York: Russell Sage Foundation, 2005). 27 Pew Research Center, “The Rise of Asian Americans, Updated Edition,” Pew Research Social and Demographic Trends (2013), accessed November 6, 2014, www. pewsocialtrends.org/​2012/​06/​19/​the-​rise-​of-​asian-​americans/​#fn-​12979-​11. 28 Anna Brown and Eileen Patten, “Statistical Portrait of Hispanics in the United States, 2012,” Pew Research Hispanic Trends Project (2014), accessed November 6, 2014, www.pewhispanic.org/​2014/​04/​29/​statistical-​portrait-​of-​hispanics-​in-​the-​ united-​states-​2012/​. 29 Matt Barreto and Francisco Pedraza, “The Renewal and Persistence of Group Identification in American Politics,” Electoral Studies 28 (2009): 595–​605. 30 Barreto and Pedraza, “The Renewal and Persistence of Group Identification in American Politics,” 595–​605; Luis Ricardo Fraga, John A. Garcia, Rodney E. Hero, Michael Jones-​Correa, Valerie Martinez-​Ebers, and Gary M. Segura, Latino Lives in America: Making It Home (Philadelphia, PA: Temple University Press, 2010). 31 Pei-​te M. Lien, Margaret Conway, and Janelle Wong, “The Contours and Sources of Ethnic Identity Choices Among Asian Americans,” Social Science Quarterly 84(2), (2003): 461–​481; S. Karthick Ramakrishnan, Janelle Wong, Taeku Lee, and Jane Junn, “Race-​Based Considerations and the Obama Vote,” Du Bois Review 6(1), (2009): 219–​238. 32 Pew Research Center, “The Rise of Asian Americans.”

156  Erica Czaja and Vladimir E. Medenica 33 Paul Taylor, Mark Hugo Lopez, Jessica Martinez, and Gabriel Velasco, “When Labels Don’t Fit:  Hispanics and Their Views of Identity,” Pew Research Hispanic Trends Project (2012), accessed November 6, 2014, www.pewhispanic.org/​2012/​04/​04/​ when-​labels-​dont-​fit-​hispanics-​and-​their-​views-​of-​identity/​. 34 Lien et  al., “The Contours and Sources of Ethnic Identity Choices Among Asian Americans,” 465. 35 Jane Junn and Natalie Masuoka, “Asian American Identity: Shared Racial Status and Political Context,” Perspectives on Politics 6(4), (2008): 729–​740. 36 Junn and Masuoka, “Asian American Identity:  Shared Racial Status and Political Context,” 737. 37 Matt Barreto, “Si Se Puede! Latino Candidates and the Mobilization of Latino Voters,” American Political Science Review 101(3), (2007): 425–​441. 38 Barreto, “Si Se Puede! Latino Candidates and the Mobilization of Latino Voters,” 438. 39 Katherine Tate, From Protest to Politics: The New Black Voters in American Elections (New York: Russell Sage Foundation, 1993). 40 Barreto and Pedraza, “The Renewal and Persistence of Group Identification in American Politics,” 599. 41 Marisa A. Abrajano and R. Michael Alvarez, New Faces New Voices: The Hispanic Electorate in America (Princeton, NJ: Princeton University Press, 2010). 42 Barreto and Pedraza, “The Renewal and Persistence of Group Identification in American Politics.” 43 Claudine Gay, “Putting Race in Context:  Identifying the Environmental Determinants of Black Racial Attitudes,” American Political Science Review 98(4), (2004): 547–​562. 44 Dawson, Behind the Mule. 45 Cathy Cohen, The Boundaries of Blackness: AIDS and the Breakdown of Black Politics (Chicago, IL: University of Chicago Press, 1999): x–​xi, emphasis added. 46 Cohen, The Boundaries of Blackness, 118. 47 Cohen, The Boundaries of Blackness, 8. 48 Michael Dawson, Black Visions:  The Roots of Contemporary African-​ American Political Ideologies (Chicago, IL: The University of Chicago Press, 2001). 49 Melissa Harris-​Lacewell, Barbershops, Bibles, and BET:  Everyday Talk and Black Political Thought (Princeton, NJ: Princeton University Press, 2004): xxiii. 50 Harris-​Lacewell, Barbershops, Bibles, and BET, 9. 51 Harris-​Lacewell, Barbershops, Bibles, and BET, 23. 52 Abrajano and Alvarez, New Faces New Voices. 53 Paul Frymer, Uneasy Alliances: Race and Party Competition in America (Princeton, NJ: Princeton University Press, 1999). 54 Adrian D.  Pantoja, Ricardo Ramirez, and Gary M.  Segura, “Citizens by Choice, Voters by Necessity:  Patterns of Political Mobilization by Naturalized Latinos,” Political Research Quarterly 54 (2001): 729–​750. 55 James H. Kuklinski and Norman L. Hurley, “On Hearing and Interpreting Political Messages: A Cautionary Tale of Citizen Cue-​Taking,” The Journal of Politics 56(3), (1994): 729–​751; Tasha Philpot, Race, Republicans, and the Return of the Party of Lincoln (Ann Arbor, MI: The University of Michigan Press, 2007); Jane Mansbridge and Katherine Tate, “Race Trumps Gender: The Thomas Nomination in the Black Community,” PS: Political Science and Politics 25 (1992): 488–​492.

Race, Ethnicity, and Public Opinion  157 56 Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, The American Voter (Chicago, IL: The University of Chicago Press, 1960). 57 Campbell, Converse, Miller, and Stokes, The American Voter. 58 Dawson, Behind the Mule. 59 Rogers, Afro-​Caribbean Immigrants and the Politics of Incorporation. 60 Jones-​Correa, Between Two Nations. 61 Janelle S. Wong, “The Effects of Age and Political Exposure on the Development of Party Identification among Asian American and Latino Immigrants in the United States,” Political Behavior 22(4), (2000): 341–​371. 62 Rogers, Afro-​Caribbean Immigrants and the Politics of Incorporation; Jones-​Correa, Between Two Nations. 63 Wong, “The Effects of Age and Political Exposure on the Development of Party Identification among Asian American and Latino Immigrants in the United States,” 341–​371; Pei-​te Lien, Christian Collet, Janelle Wong, and S. Karthick Ramakrishnan, “Asian Pacific-​American Public Opinion and Political Participation,” PS:  Political Science & Politics 34(3), (2001): 625–​630. 64 V. O. Key, Southern Politics in State and Nation. (New York: Knopf, 1984 [1949]); Hubert M.  Blalock, Toward a Theory of Minority-​Group Relations (New  York: Wiley, 1967). 65 Gordon Allport, The Nature of Prejudice (Cambridge: Addison-​Wesley Publishing Company, 1954). 66 Claudine Gay, “Seeing Difference:  The Effects of Economic Disparity on Black Attitudes toward Latinos,” American Journal of Political Science 50(4), (2006): 982–​997. 67 Gay, “Seeing Difference,” 990. 68 Eric J.  Oliver and Janelle Wong, “Intergroup Prejudice in Multiethnic Settings,” American Journal of Political Science 47(4), (2003): 567–​582. 69 Oliver and Wong, “Intergroup Prejudice in Multiethnic Settings.” 70 Daniel J.  Hopkins, “Politicized Places:  Explaining Where and When Immigrants Provoke Local Opposition,” American Political Science Review 104(1), (2010): 40–​60. 71 Hopkins, “Politicized Places.” 72 Shana Kushner Gadarian and Bethany Albertson, “Anxiety, Immigration, and the Search for Information,” Political Psychology 35(2), (2014): 133–​164. 73 Dennis Chong and Dukhong Kim, “The Experiences and Effects of Economic Status among Racial and Ethnic Minorities,” American Political Science Review 100(3), (2006): 336–​337. 74 Chong and Kim, “The Experiences and Effects of Economic Status among Racial and Ethnic Minorities.” 75 Chong and Kim, “The Experiences and Effects of Economic Status among Racial and Ethnic Minorities.” 76 Chong and Kim, “The Experiences and Effects of Economic Status among Racial and Ethnic Minorities.” 77 Lee Sigelman and Susan Welch, Black Americans Views of Racial Inequality:  The Dream Deferred (Cambridge: Cambridge University Press, 1991). 78 Chong and Kim, “The Experiences and Effects of Economic Status among Racial and Ethnic Minorities,” 350. 79 Sigelman and Welch, Black Americans Views of Racial Inequality, 55, 59.

158  Erica Czaja and Vladimir E. Medenica 80 Sigelman and Welch, Black Americans Views of Racial Inequality, 59. 81 Donald Kinder and Lynn Sanders, Divide By Color: Racial Politics and Democratic Ideals (Chicago, IL:  The University of Chicago Press, 1996); David O.  Sears, Jim Sidanius, and Lawrence Bobo (eds.), Racialized Politics: The Debate About Racism in America (Chicago, IL: The University of Chicago Press, 2000). 82 Linda Lopez and Adrian D.  Pantoja. 2004. “Beyond Black and White:  General Support for Race-​Conscious Policies among African Americans, Latinos, Asian Americans and Whites,” Political Research Quarterly 57(4), (2004): 633–​642.

Chapter 7

Categorical Politics in Action Gender and the 2016 Presidential Election Donald Kinder, Molly E. Reynolds, and Nancy Burns

On July 26, 2016, the Democratic Party did what no major American political organization had ever done before: namely, nominate a woman to run for president. In her acceptance speech at the Democratic National Convention in Philadelphia, Hillary Clinton took notice of the historical significance of the occasion: Tonight we’ve reached a milestone in our nation’s march to a more perfect union … Standing here as my mother’s daughter, and my daughter’s mother, I’m so happy this day has come. Happy for grandmothers and little girls and everyone in between. Happy for boys and men, too—​ because when any barrier falls in America, for anyone, it clears the way for everyone. When there are no ceilings, the sky’s the limit. So let’s keep going, until every one of the 161 million women and girls across America has the opportunity she deserves.1 Clinton made history—​but she lost the election. Moreover, she lost the election to Donald Trump. During the campaign, Trump’s repeated displays of contempt for particular women—​for Clinton but for other women as well—​ made gender still more salient than it would have been had Clinton run against a more conventional opponent. To those of us with an interest in the relationship between gender and politics, the 2016 American presidential election presents a fascinating and irresistible case for analysis. Our chapter starts off by clarifying the meaning of gender itself (a perhaps surprisingly slippery concept). We distinguish gender from sex, on the idea that males and females differ not only in their physical bodies (sex) but also in their social position (gender). Next, we note the close relationship between gender and inequality. In every known society including our own, men enjoy substantial advantages over women in wealth, power, and status. Persistent, pervasive inequality—​over generations and across domains—​makes gender relevant to politics. In the next part of the chapter, we spell out why we would expect gender to play an especially important role in the 2016 American presidential election, drawing on the principle of framing. Finally—​and this is

160  Donald Kinder, Molly E. Reynolds, and Nancy Burns

really the heart of the chapter—​we test whether gender in fact was especially important in 2016, examining four aspects of politics in particular:  gender differences in the vote; gender differences in turnout; gender as a source of group solidarity; and gender as an object of attitude.

The Meaning of Gender Sex—​the distinction between male and female—​is a biological concept. Absent medical intervention, sex is fixed at conception. It is a matter of anatomy and hormones and, going deeper, genetic code. Females are endowed with the biological capacity to give birth, breast-​feed, and menstruate. Males are not. These biological facts of life are real—​but along with Erving Goffman, we suggest that biological differences between the sexes are modest when set against the far-​reaching differences between men and women visible in modern societies. Work, play, school, home, and politics:  all display evidence of the “awesomely ramified significance” of sex.2 This point is summed up, famously and epigrammatically, by Simone de Beauvoir in The Second Sex. “One is not born,” Beauvoir wrote, “but rather becomes, woman.”3 In the 1970s, feminist scholars in the social sciences and humanities began to use the term “gender” in place of “sex.” Roughly speaking, sex was reserved for differences between men and women understood to be biologically determined, while gender was applied to differences between men and women understood to be socially constructed.4 The distinction between sex and gender gets at something important, but it is neither entirely clean nor completely satisfying. All human behavior is a product of the interaction between genetic predispositions and environmental factors. The idea that gender is only a social construction, and that genetic predispositions can be ignored entirely, goes too far. But we need a term separate from sex, and for that gender will do. Gender, as contrasted with sex, allows us to recognize the part played by beliefs, practices, and institutions in the creation and maintenance of inequalities between men and women. Gender, we say, is social construction and biological predisposition. Gender is what society makes of sex.

Gender and Inequality According to Sherry Ortner, within every known society, women are subordinated to men. The particular form of female disadvantage varies enormously around the world and over time, but it is present to some degree everywhere. “The search for a genuinely egalitarian, let alone matriarchal, culture,” Ortner concludes, “has proved fruitless” (p. 24). Inequality between men and women appears to be a universal condition.5 In the United States, economic inequality in gender relations has been generated and maintained principally by the separation of men and women into distinctive occupational structures. For much of American history, paid

Categorical Politics in Action  161

labor was the exclusive province of men. Women’s economic activities were “enclosed” within households.6 As America industrialized, increasing numbers of women entered the paid labor force, but as they did so they were steered away from positions of influence and authority. Stenographer, waitress, typist, secretary, nurse, and filing clerk became women’s jobs; supervisors, managers, partners, and professionals were reserved for men. By the middle of the twentieth century, most firms had explicit policies designating certain positions for women and others for men. The prohibition of women from positions that promised greater earnings and mobility induced a piling up of women in certain jobs, putting downward pressure on their wages. When women and men did work the same jobs, women were paid less. In recent decades, gender’s role in the structure, organization, and operation of the labor market has diminished. With the rise of the service economy, American women poured into the paid labor force. They now constitute nearly one half of the labor force, a breathtaking change from just a generation or two ago. The gap in earnings between men and women is narrowing, though it remains substantial. At the beginning of the twentieth century, women engaged in full-​time year-​round work earned less than one half the pay of men; by the end of the century, the fraction had increased to about three-​quarters. Educational and employment opportunities for women are opening up, but progress on this front has been limited primarily to those best-​equipped to take advantage of the new opportunities: the highly educated and upper-​middle class.7 Finally, household labor, including care of children, constitutes roughly one-​third of all work done in the United States, and women do most of it. This inequality is most lop-​sided within traditional households, but it holds even for married women who work full-​time outside the home. In the United States today, as in other parts of the world, “women continue to perform the vast bulk of household labor, the essential work of reproducing labor power through feeding, cleaning, child care, health care, personal service, and emotional support.”8

Gender and the Framing of Electoral Choice In a series of influential essays written in the aftermath of World War I, Walter Lippmann, perhaps the most prominent public intellectual of the day, argued that the typical citizen—​parochial in interest, modest in intellect, and most of all preoccupied with private affairs—​lacks the wherewithal to grasp political matters in any deep way. People are busy; politics is hard.9 Lippmann, we now know, was right. While family, work, health, and friendships are central preoccupations, the events of political life remain, for most of us most of the time, peripheral curiosities. In modern societies like the United States, “politics is a sideshow in the great circus of life.”10 On top of this, we know from cognitive science that “human thinking powers are very modest when compared with the complexities of the environments in

162  Donald Kinder, Molly E. Reynolds, and Nancy Burns

which human beings live.”11 People form judgments and make decisions based on limited information in general, and this may be especially the case when it comes to politics. Voters form judgments and make decisions in light of a particular way of looking at a problem—​in light of a particular frame. Frames operate by altering the relative salience of different aspects of the decision, highlighting some features of the situation and masking others. Framing is powerful because people generally passively accept the frame they are given—​ in politics, as in other domains of life.12 The framing principle has direct implications for the place of gender in political decision-​making. For most of us, gender is an important aspect of our identity. But we belong to many social groups at once. We are black or white or brown; Catholic, Protestant, Jewish or atheist; bankers or carpenters; urbanites or suburbanites; Southerners or Yankees; and so on. This means we have available, in principle at least, many different ways of seeing ourselves. In parallel fashion, we possess attitudes toward many social groups—​toward women, Catholics, immigrants, bankers and more—​any number of which could be relevant to politics. Here’s the key point: which aspects of identity and attitude become important to political choice depends, in part, on how the choice is framed. In particular, the activation of identity and attitude depends on the clarity of cues linking candidates to groups. Such cues can take various forms. Candidates might propose particular policies that visibly favor some groups at the expense of others. They might emphasize or ignore problems that are of special concern to a particular group. They might spend time in the public eye with iconic representatives of one group or another. Perhaps the most effective signaling device of all is to embody membership, as Hillary Clinton embodied gender in 2016. Of course, Clinton’s opponent, Donald Trump also did his part to focus the electorate’s attention on gender. During the campaign, Mr. Trump claimed to “cherish” women, but he also said, without prompting, that Rosie O’Donnell was a “fat pig”; that Hillary Clinton couldn’t satisfy her husband’s sexual appetites; and that Carly Fiorina could never be elected president on account of her face. During a GOP debate, Fox News’ Megyn Kelly reminded Trump that he had called women he didn’t like “dogs,” “slobs,” and “disgusting animals.” Trump brushed aside the question, claiming he didn’t have the “time for total political correctness.” Shortly after the debate, Trump called Kelly a “bimbo” and implied that her unfriendly questions were caused by the fact that she was menstruating. And late in the campaign, the nation heard Trump gloating over what he felt free to do to women he found attractive: grab them by the pussy. Gender is always present in politics, and in presidential elections in particular, even when both candidates are men. In most cases, however, it is submerged, passed over, taken for granted. One party nominates a man; the other party does the same. Both candidates talk about many things during the campaign, but for the most part without referring to men or to women as politically relevant groups. Not in 2016. In 2016, gender moved to center stage: in

Categorical Politics in Action  163

part because one candidate “happened” to be a woman; and in part because her opponent seemed unable to quit dishing out insults to one woman or another or to women in general. In the remainder of the chapter we investigate how, under these extraordinary and unprecedented circumstances, gender affected the 2016 election.

Gender and the Vote in 2016

Percentage probability democratic vote

The gender gap familiar to us today—​women voting more Democratic than men—​was not always so. Women were less likely to vote Democratic than men in the immediate post-​World War II elections: slightly so in the Truman election, and then more substantially in the Eisenhower and Kennedy elections. If only women, and not men, had been permitted to vote in 1960, Richard Nixon would have been elected president. By the time Ronald Reagan entered the White House, the gender difference had reversed. Since then, women have voted more Democratic than men in presidential elections. (In the press, the gender gap is routinely portrayed as a problem for Republicans, forgetting that as women were voting disproportionately Democratic, men were voting disproportionately Republican. If Republicans have a woman problem, the Democrats have a man problem). The question before us here is whether the gender gap increased in 2016 over normal levels, given who Clinton is and what Trump said.13 To answer this question, we first need to determine the gender gap we would have expected to see in 2016, had that election been carried out under ordinary conditions with respect to gender (that is, with two male candidates). We estimate this value by making use of the massive exit polls carried out in presidential years by a consortium of news organizations. Figure  7.1 displays the 100 90 80 70 60 50 40 30 20 10 0 1980 1984 1988 1992 1996 2000 2004 2008 2012 2016 Men

Women

Figure 7.1 Democratic Presidential Vote by Gender 1980–​2016

164  Donald Kinder, Molly E. Reynolds, and Nancy Burns

percentage voting for the Democratic presidential candidate between 1980 and 2016, separately for men and women. The vertical distance between the two lines is the gender gap in voting. As the figure shows, over this period, women have regularly voted more Democratic than men. From 1980, when the modern gender gap became a fixture in American presidential elections, through 2012, this difference has averaged about 7 percentage points.14 In 2016, Hillary Clinton ran stronger among women than men, as Democrats tend to do. But in 2016, the gap was more pronounced than normal. According to the exit polls, and as shown in Figure 7.1, Clinton received 54 percent of the women’s vote, but just 41 percent of the men’s vote. Clinton won decisively among women and lost in a landslide among men. A survey of “verified” voters from the Pew Research Center reported an almost identical result: 54 percent of the women supported Clinton, compared to 43 percent of the men. Thanks, we suspect, to Clinton’s gender and Trump’s language, the gender gap widened in 2016.15

Gender and Turnout in 2016 Elections are determined not only by the choice voters make once they arrive at the polls but also by who shows up and who stays home. As we just learned, women were more likely to vote Democratic than men in 2016, and more than usual. The question we take up here is whether the 2016 election was out of the ordinary also with respect to turnout. Can we detect an increase in the number of women getting themselves to the polls in 2016? The existing literature on gender and turnout is mixed. Some studies report that the presence of a woman on the ballot increases turnout among women voters; others find no such effect. Let’s see what we find in 2016.16 Figure  7.2 presents turnout rates in presidential elections from 1964 to 2016, separately for men and women, based on the Census Bureau’s Current Population Surveys. As shown there, at the beginning of the series, men were more likely to turnout than women. The difference is modest and diminishes gradually, reversing in 1980. Starting with the first Reagan election, now it is women who outvote men. Does the gender gap in turnout jump in 2016? No, it does not. The difference in turnout favoring women is 4.7 percentage points in the first Obama election; 4.1 percentage points in the second Obama election; and 4.3 percentage points in the 2016 election. There is no evidence here, in the raw figures, that Clinton’s gender and Trump’s rhetoric had any effect on women turning out to vote. This may seem surprising to you. It surprised us. We took another look at gender and turnout in 2016, from a different angle, in an independent test, to make sure we had the story right. To do so we analyzed American National Election Study (ANES) data from every presidential contest between 1948 (the first election studied by ANES) and 2016. The ANES series provides two advantages to us here. First of all, we are able to control statistically for other factors that might influence turnout in addition to gender: notably, education,

Categorical Politics in Action  165

100 90

Percentage turnout

80 70 60 50 40 30 20 10 0 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 2012 2016 Men

Women

Figure 7.2 Turnout in Presidential Elections by Gender 1964–​2016

family income, age, and race. From previous research, we know that turnout is highest among well-​educated, affluent, older, white Americans.17 Our model tells us whether gender makes a difference to turnout, holding constant the effects due to these other relevant factors.18 A second advantage is that ANES election surveys ask respondents not only whether they voted but also whether they participated in the campaign in a variety of other ways:  by attending a rally, wearing a button or displaying a yard sign, trying to persuade someone to vote their way, giving time to a campaign, and donating money to a candidate or party. According to our analysis, women were generally less likely to vote in presidential elections than men over the first 30 years of the ANES series, from the Truman election in 1948 to the Ford election in 1976. From then on, men and women were generally equally likely to vote, once differences in education, income, age, and race are considered. This holds—​no effect of gender—​for the better part of 40 years, including 2016. Moreover, what we find for turnout we also find for other forms of political participation. Taken all together, we turn up no evidence that the Clinton–​Trump contest drew women deeper into politics.

Gender as an Object of Identification in 2016 For politics and no doubt for other things as well, it is one thing to be a woman. It is another thing to belong to the group psychologically, or as we will say, to identify with women.

166  Donald Kinder, Molly E. Reynolds, and Nancy Burns

Group identification is rooted in two kinds of attachments. Common fate refers to the extent to which individuals believe their life chances and outcomes are intertwined with the opportunities and experiences of their group, that what happens to their group will happen to them. Those highly identified with their group on grounds of common fate will come to a political choice with their group’s interests prominently in mind. Group identification may also be grounded in emotional interdependence, occurring when individuals feel close to their group, and experience pride when other group members do well and anger when they are treated unfairly. Emotional interdependence reflects the expressive side of politics. To the degree that Americans derive their sense of self from their membership in, and identification with, social groups, political choices become acts of affirmation and solidarity.19 People vary in the degree to which they identify with a group. For some group members, attachment is effectively zero; for others, identification with a social group constitutes a central aspect of identity; and there exist all shades in between. Strength of identification is a sign of a person’s priorities. The stronger the identification, the more powerful the political consequences of group membership will be. That, at least, is the theory. Hillary Clinton has been a political figure of national prominence for more than a quarter century: as First Lady in her husband’s White House; as Senator from New York; in her failed run for the Democratic presidential nomination in 2008; as Secretary of State; and most recently, of course, as her party’s candidate for president. Previous research, carried out before 2016, finds women who identify closely with their gender to be more likely to evaluate Hillary Clinton favorably than those who do not—​especially at times when Clinton is prominently featured in the national press.20 Running for president—​not to mention running as the first woman to lead a major party—​is about as good a way to get attention as there is. Accordingly, we expect gender group identification to have an elevated effect on the women’s vote in 2016. To find out, we estimate the effect of gender group identification on the presidential vote in 2016 and compare it to the effect of gender group identification in preceding presidential elections, looking for a bump up in the effect in 2016. We analyze a quartet of national studies: the 2004 National Election Study; the Cooperative Campaign Analysis Project (CCAP), a national panel study of registered voters, conducted online in six waves, commencing in December of 2007 and concluding in November of 2008 immediately after the presidential election; the 2012 Cooperative Congressional Election Study (CCES) carried out online, with interviews completed before and after the 2012 presidential election; and the 2016 CCES carried out online, with interviews completed before and after the 2016 presidential election. Each of the four studies included a standard measure of group identification. Women were asked how much they thought their own fate was tied up with the fate of women in general; how often they felt pride over the accomplishments of women; and how often they felt angry about the way society treated women.21

Categorical Politics in Action  167

We estimated the effect of gender group identification in a variety of different ways. We began with a simple model, one that assumed that voting is a consequence of just two factors—​party identification and gender group identification—​and then moved on to more complicated (and more realistic) models. However we ran the analysis, we obtained the same result. Gender group identification appeared to have nothing to do with the women’s vote in 2004, 2008, or 2012. But in 2016, the story changes. In 2016, women who felt closely attached to women in general were much more likely to support Clinton than were those who lacked such feelings. The magnitude of this effect is shown in Figure 7.3. Along the vertical axis is the predicted probability of a vote for Clinton. Along the horizontal axis is the standard measure of gender group identification, ranging from low on the left to high on the right. The predicted probability of a Clinton vote holds all the other variables in the analysis at their means (in this case, partisanship, belief in equal opportunity, belief in limited government, education, religion, region, and race). Put differently, the figure gives the predicted probability of a Clinton vote as a function of variation in gender group identification, for a woman who is otherwise average on partisanship, belief in equal opportunity, and all the rest.22 As Figure 7.3 shows, support for Clinton increases rapidly with gender group identification. Women who were average on control factors and who identified closely with women (a score of .80 on the gender group identification scale) were some 25 percentage points more likely to vote Democratic in 2016 than were otherwise comparable women who did not much think of themselves

Percentage probability democratic vote

1

.8

.6

.4

.2

0 0

.1

.2

.3

.4 .5 .6 Gender identification

.7

Figure 7.3 Women’s Vote for Clinton by Gender Identification

.8

.9

1

168  Donald Kinder, Molly E. Reynolds, and Nancy Burns

in terms of gender (a score of .20 on the gender group identification scale).23 Hillary Clinton’s historic run for the presidency succeeded in one respect for sure: in encouraging women voters to consider their sense of gender solidarity as they went to the polls.

Gender as an Object of Attitude in 2016 What gender can give, gender can take away. Gender solidarity generated votes for Clinton. But the attitudes Americans held toward the place of women in society pushed in the opposite direction. By “attitude” we refer to a psychological propensity to act in favor of, or in opposition to, some object. Attitudes are consequential. As Gordon Allport once put it, attitudes affect “perception, judgment, memory, learning, and thought” and “determine for each individual what he will see and hear, what he will think, and what he will do.”24 Attitudes toward social groups are especially important in politics. Scores of studies have shown that political preferences—​opinions on policy, support for candidates—​are group-​centric: that is, shaped in powerful ways by the attitudes citizens harbor toward the social groups they see as the principal beneficiaries or victims of a policy or candidate. For example, support for tightening welfare benefits derives, in part, from hostility toward the poor; resistance to immigration reflects, in part, suspicion that the new immigrants are somehow un-​American; voting against John Kennedy in the 1960 presidential election was an expression, in part, of animosity toward Catholics; and so on. Here we consider the attitudes Americans harbor toward women. Did Hillary Clinton and Donald Trump conspire (so to speak) to enhance the importance of such attitudes in the 2016 vote?25 For quite some time now, public discussion over gender has been preoccupied with the question of women’s place:  whether women’s primary, or even exclusive, responsibilities should lie in the private sphere of home and family. Politicians, activists, intellectuals, and religious leaders have all had something to say on this subject. The traditional view holds that women belong, properly and naturally, to the private sphere of home and family; that their fragile and delicate natures must be protected from the rough and tumble world of work; that their mission, ordained by biology if not by God, is to comfort their husbands and nurture their children. This perspective on gender relations came under intense and public challenge by the modern women’s movement. For the first time, Americans in visible numbers began to question the notion that men and women were essentially different, and that society must be organized in harmony with this fundamental fact of nature. Perhaps, some began to argue, women were equal in talent and ambition to men and could make significant contributions outside the home. To capture differences in attitudes toward women’s place, we have drawn upon research done by Janet Spence and her Attitude toward Women Scale (AWS),

Categorical Politics in Action  169

developed in the early 1970s.26 AWS is a measure of beliefs about women’s roles, rights, and responsibilities. The scale puts front and center the traditional view of the family, in which the husband works in the world and the wife maintains the home. Some of the questions are about where men and women belong (e.g., “It is much better for everyone involved if the man is the achiever outside the home and the woman takes care of the home and family”). Some, going further, propose that the traditional division of labor between men and women is a direct reflection of differences in their underlying essence (“Men are naturally better-​suited to the world of work than women are”). We are also interested in a second kind of attitude directed at women, more contemporary in origin, coming to prominence in response to the women’s movement. In 1963, the President’s Commission on the Status of Women issued its report documenting serious inequalities at work and before the law. The following year, passage of the Civil Rights Act prohibited discrimination on the grounds of race, religion, and—​as an afterthought—​sex. In 1966, the National Organization for Women came into being, providing women with organizational representation in Washington and some assurance that laws intended to protect their interests would be enforced. Demonstrations and protests erupted in the early 1970s and continued on through much of the decade. Women’s rights began to be discussed in press conferences, congressional hearings, and in national party platforms and conventions. In 1973, the Supreme Court issued its famous ruling (Roe v.  Wade) that state efforts to regulate abortion were unconstitutional. Things were changing and changing rapidly.27 Not everyone was applauding. The women’s movement generated animosity as well as support. Many Americans, men and women alike, were upset over efforts to upend traditional arrangements. Some were puzzled over the claim of systematic discrimination. Some were angry over the movement of women into public life. Some believed that women were demanding not equality but special favors. Taken together, these complaints constitute what we will call modern sexism:  modern, because we understand it to be a reaction to the modern women’s movement; and sexism, because we understand it to justify keeping women in their place.28 We measure modern sexism with questions that gauge resentment to changes in gender relations: that women are pushing too hard for equality (“When women demand equality these days, they are actually seeking special favors”); that women blame discrimination for their own shortcomings (rejecting the proposition that “Women often miss out on good jobs because of discrimination”); and that women are stirring things up for no good reason (“Women who complain about harassment cause more problems than they solve”). Measures of women’s place and modern sexism have been part of the ANES presidential election series since 2004. That’s the good news. It means we can compare the electoral impact of beliefs about women’s proper place and reactions to the modern women’s movement in the gender-​charged election

170  Donald Kinder, Molly E. Reynolds, and Nancy Burns

1 .8 .6 .4 .2 0 0

.1 .2 .3 .4 .5 .6 .7 .8 .9 Modem sexism

1

Percentage probability democratic vote

Percentage probability democratic vote

of 2016 to the effect due to the same attitudes in recent presidential elections where gender was a much less prominent part of the campaign. The not-​so-​ good news is this: the particular questions used to assess women’s place and modern sexism vary from one study to the next, complicating comparison. We would have preferred the questions to be exactly the same across all four ANES surveys. Proceeding cautiously, we estimated the effect of women’s place and modern sexism in a variety of different ways. Our models always included women’s place and modern sexism, of course, but they varied depending on which other factors were included: party identification, feelings of racial resentment (among Whites), views on the proper size and scope of government, beliefs regarding the importance of equal opportunity, religion, education, and region. Across all variations, results were much the same. First of all, both women’s place and modern sexism affected the vote in 2016, for men and women alike, and in roughly equal measure. Those who subscribed to the traditional understanding of relations between men and women and who were upset over the changes pushed by the modern women’s movement were more likely to vote against Clinton and for Trump. And second, women’s place and modern sexism, so important in 2016, were irrelevant in presidential elections preceding 2016: in 2012, in 2008, and in 2004. These effects of women’s place and modern sexism are on display in Figure  7.4.29 The left-​hand panel of the figure shows the predicted vote for Hillary Clinton (along the vertical axis) as a function of modern sexism (along the horizontal axis), controlling on other factors. The right-​hand panel shows the predicted vote for Clinton (the vertical axis) as a function of beliefs about women’s proper place (along the horizontal axis), again controlling on other factors. In both panels, support for Clinton declines sharply from left to right, as attitudes on women’s place in American society shift from a more contemporary view to a traditional one. 1 .8 .6 .4 .2 0 0

.1 .2 .3 .4 .5 .6 .7 .8 .9 Women’s role

Figure 7.4 Vote for Clinton by Modern Sexism and by Women’s Role

1

Categorical Politics in Action  171

Implications and Conclusion Gender is atmospheric. It is both profoundly important and yet, much of the time, invisible. In a way we don’t always notice, gender influences how we think about ourselves and how we think about others. It shapes our aspirations, opportunities, and achievements. It organizes our social lives. It defines the tasks we “naturally” take on and the tasks we “naturally” leave for others. In the 2016 presidential election, gender became visible. Indeed, it is hard to imagine a more provocative moment. In 2016, the American electorate was offered a real choice: Hillary Clinton or Donald Trump. Clinton was the first woman to run for the highest office in the land as the candidate for a major party. She campaigned as a feminist (“I’m with her”), promised to keep abortion legal, and generally welcomed the support of women. Clinton violated the traditional notion that women belong at home and became the figurehead for uppity women everywhere. Meanwhile, her opponent, Donald Trump, criticized Clinton for playing the “woman card,” suggested she had neither the strength nor the stamina to stand up to Russia, promised to appoint justices to the Supreme Court who would overturn Roe v. Wade, and said despicable things about particular women and about women in general. If we cannot find gender at work in the 2016 presidential election, where will we find it? Of course, we do find gender at work in the Clinton–​Trump contest. In 2016, Clinton and Trump generated the largest gender gap in presidential voting for 50 years—​and probably the largest gender gap ever.30 In 2016, feelings of group solidarity played an important role in the vote, but not in presidential elections leading up to the Clinton–​Trump contest. In 2016, Americans who subscribed to the traditional understanding of relations between men and women and who were upset over the changes pushed by the modern women’s movement voted against Clinton, but we saw no evidence of this in presidential elections prior to 2016.31 In just one respect did our analysis fail to find a gender effect in 2016. Try as we might, we could turn up no evidence that the 2016 campaign succeeded in mobilizing women into politics. Women were no more likely to vote in 2016, nor were they more likely to take part in the election in any number of other ways: attend a rally, wear a button or display a yard sign, try to persuade others, work for the campaign, donate money to a candidate or a party. What comes after 2016 may well be another matter. Women ran for office in record numbers in 2018. There were suggestions that they did so in reaction to Clinton’s defeat.32 Taken all round, our results carry a double-​edged lesson. We show, on the one hand, that under extraordinary circumstances, gender can play an important role in politics. On the other hand, we show that under more-or-less normal circumstances—​no woman on the ballot, no flagrantly sexist rhetoric from her opponent—​the political effects of gender diminish. Outside the gender-​ charged 2016 election, gender effects are modest or negligible.

172  Donald Kinder, Molly E. Reynolds, and Nancy Burns

Gender generally seems to play a small role in how men and women think about politics.33 Politics is group-​centric, but social categories other than gender are almost always more important. Party, religion, race, language, ethnicity, region: these are the categories that do most of the work in structuring political conflict. Consider race in particular. It is easy enough to compare the part played by gender in 2016 to the part played by race in 2008. History was made in 2008 when Barack Obama became the first African American to be elected president of the United States. He became president in a nation torn apart over whether slavery would be permitted in the new territories; in a nation whose legislative assemblies and high courts, for the better part of 200 years, provided legal justification for discrimination and segregation by race; in a nation where profound racial inequalities—​of income, wealth, education, health, longevity, and everyday justice—​persist into the 21st century. In such a nation, a black man was elected president. We can ask of race in 2008 exactly the same questions we asked of gender in 2016. In 2008, according to the American National Election Study, Obama won 98.8 percent of the black vote but just 44.3 percent of the white vote, a difference of nearly 55 percentage points. The 2008 election generated an enormous—​an unusually enormous—​racial divide. In 2008, black voters marched to the polls in record numbers, thereby contributing significantly to Obama’s victory. In 2008, black Americans identified more with their race than women identified with their gender in 2016, and those who did so supported Obama enthusiastically. And finally, in 2008, just as the Obama campaign succeeded in enhancing the political importance of racial solidarity among black voters, so, too, it succeeded (if that is the right word) in enhancing the political importance of racial resentment among white voters.34 Gender is, or can be, important in politics—​but race seems to be more important. Why this is so is a fascinating question, one we are struggling to answer in our ongoing research. We leave you with a pair of ideas we are currently investigating, two ways of understanding why it is that race intrudes more dramatically upon American politics than gender. The first has to do with inequality. Both gender and race are exemplary cases of durable, long-​lasting inequality—​but inequality organized around race is the more profound and pervasive. Slavery is gone and the Jim Crow regime of racial oppression that followed emancipation has been dismantled, but in American society today, race and disadvantage remain closely connected. The infant mortality rate is more than twice as high among Blacks than among Whites. Black children who survive their first year can look forward to poorer health, more illness, and a substantially shorter life, on average, than their white counterparts. Racial differences in unemployment and income are substantial. The racial divide in wealth is enormous. If political conflict and organization arise from inequality, it is easy to see why race has become such a central part of American politics.35

Categorical Politics in Action  173

A second speculation has to do with social organization. Gender and race are “made” by society, but they are made in very different ways. The social organization of gender emphasizes intimacy. Women spend much of their lives in intimate relationships with men: with fathers, husbands, brothers, and sons.36 In sharp contrast, a persistent feature of race relations in the United States is separation. Despite federal fair housing legislation passed in 1968, the United States remains today, in many respects, a segregated society. And if neighborhoods continue, by and large, to reflect the color line, then so do other important American institutions: schools, churches, work, and marriage.37 Intimacy, we suggest, impairs the development of gender as a basis of political action, while separation facilitates organizing around race.

Notes 1 Clinton, Hillary. July 28, 2016. Accessed Online: www.cnn.com/​2016/​07/​28/​hillary-​ clinton-​speech-​prepared-​remarks-​transcript/​index/​html. 2 Erving Goffman, “The Arrangement between the Sexes,” Theory and Society 4(3) (1977): 301–​330. 3 Simone de Beauvoir, The Second Sex (New York: Vintage Press, 1949/​2011, p. 283). Famously and epigrammatically—​and erroneously, we think. Beauvoir’s mistake is to erase biology entirely. We would say, instead, “One is born, and then becomes, woman.” And we would say the same about men: “One is born, and then becomes, man.” Beauvoir has a point, of course:  relatively slight differences in biology are parlayed into large differences in social, economic, and political status. But she goes too far. 4 Charlene L. Muehlenhard and Zoe D. Peterson. “Distinguishing between Sex and Gender: History, Current Conceptualizations, and Implications,” Sex Roles 64(11) (2011): 791–​803. 5 Sherry Ortner, Making Gender (Boston, MA: Beacon Press, 1972). 6 Nancy F.  Cott, The Bonds of Womanhood (New Haven, CT:  Yale University Press, 1977). 7 For evidence on these points, see Claudia Goldin, Understanding the Gender Gap. An Economic History of American Women (New  York:  Oxford University Press, 1990); Claudia Goldin, “A Grand Gender Convergence:  Its Last Chapter,” American Economic Review 104(4) (2014):  1–​30; Barbara F.  Reskin and Patricia A.  Roos, Job Queues, Gender Queues:  Explaining Women’s Inroads into Male Occupations (Philadelphia, PA: Temple University Press, 1990); Francine D. Blau, Mary C.  Brinton, and David B.  Grusky, The Declining Significance of Gender? (New  York:  Russell Sage Foundation, 2006); and Francine D.  Blau and Lawrence M.  Katz, “The Gender Wage Gap:  Extent, Trends, and Explanations,” Journal of Economic Literature 55(3) (2017): 789–​865. 8 Douglas S.  Massey, Categorically Unequal. The American Stratification System (New York: Russell Sage Foundation, 2007). 9 Walter Lippmann, Public Opinion (New York: Macmillan, 1922) and The Phantom Public (New York: Harcourt, Brace, 1925). 10 Robert A. Dahl, Who Governs? (New Haven, CT: Yale University Press, 1961).

174  Donald Kinder, Molly E. Reynolds, and Nancy Burns 11 Herbert A.  Simon, Models of Thought (New Haven, CT:  Yale University Press, 1979, p. 3). 12 On the idea of frame, see Amos Tversky and Daniel Kahneman, “The Framing of Decisions and the Psychology of Choice,” Science 211 (1981): 453–​458. For an overview of the cognitive science of judgment and choice, see Daniel Kahneman, “A Perspective on Judgment and Choice:  Mapping Bounded Rationality,” American Psychologist 58 (2003): 697–​720. 13 On the gender gap, see (among many others) Janet M. Box-​Steffensmeier, Suzanna De Boef, and Tse-​Min Lin, “The Dynamics of the Partisan Gender Gap,” American Political Science Review 98(3) (2004):  515–​528; Vincent L.  Hutchings, Nicholas Valentino, Tasha Philpot, and Ismail White, “The Compassion Strategy: Race and the Gender Gap in Campaign 2000,” Public Opinion Quarterly 68(4) (2004): 512–​ 541; Karen M. Kaufmann and John R. Petrocik, “The Changing Politics of American Men: Understanding the Sources of the Gender Gap,” American Journal of Political Science 43(3) (1999):  864–​887; Torben Iversen and Frances McCall Rosenbluth, “The Political Economy of Gender:  Explaining Cross-​National Variation in the Gender Division of Labor and the Gender Voting Gap,” American Journal of Political Science 50(1) (2006): 1–​18; and Lena Edlund and Rohini Pande, “Why Have Women Become Left-​Wing? The Political Gender Gap and the Decline in Marriage,” Quarterly Journal of Economics 117(3) (2002): 917–​961. 14 In recent years, the exit polling has been carried out by Edison Research for the National Election Pool, a consortium of ABC News, the Associated Press, CBS News, CNN, Fox News, and NBC News. The 2016 estimates are based on interviews with 24,537 voters leaving 350 voting locations; 4,398 additional interviews were conducted over the telephone with early and absentee voters. 15 The estimate of the gender gap from the Pew survey is based on “verified” voters, accomplished by matching Pew’s nationally representative sample of adult Americans with official voter files. 16 Research that finds the presence of a woman on the ballot to increase turnout among women voters includes: Nancy Burns, Kay Lehman Schlozman, and Sidney Verba, The Private Roots of Public Action (Cambridge, MA:  Harvard University Press, 2001); Beth Reingold and Jessica Harrell, “The Impact of Descriptive Representation on Women’s Political Engagement: Does Party Matter?” Political Research Quarterly 63(2) (2010):  280–​ 294; and Lonna Rae Atkeson, “Not All Cues Are Created Equal:  The Conditional Effect of Female Candidates on Political Engagement,” Journal of Politics 65(4) (2003): 1040–​1061. For research that reports no such effect, see David E. Broockman, “Do Female Politicians Empower Women to Vote or Run for Office? A Regression Discontinuity Design,” Electoral Studies 30 (2013): 1–​15; and Kathleen Dolan, “Symbolic Mobilization? The Impact of Candidate Sex in American Elections,” American Politics Research 34(6) (2006): 687–​704. 17 We rely here on classic studies of political participation: Sidney Verba, Kay Schlozman, and Henry Brady, Voice and Equality (Cambridge, MA: Harvard University Press, 1995); Steven J. Rosenstone and John Mark Hansen, Mobilization, Participation, and Democracy in America (New York: Longman, 2003); and Raymond E. Wolfinger and Steven J. Rosenstone, Who Votes? (New Haven, CT: Yale University Press, 1980). 18 In equation form, our model can be written: Turnoutt = β0 + β1Gendert + β2Educationt + β3Incomet + β4Aget + β5[Aget > 65] + β6 Racet + ε

Categorical Politics in Action  175 19 On common fate, see Donald T. Campbell, “Common Fate, Similarity, and Other Indices of the Status of Aggregates of Persons as Social Entities,” Behavioral Science 3(1) (1958): 14–​25; on the expressive foundations of group identification, see Henri Tajfel, Human Groups and Social Categories (New  York:  Cambridge University Press, 1981). 20 On this point see: Nancy Burns, Ashley E. Jardina, Donald R. Kinder, and Molly E.  Reynolds, “The Politics of Gender,” in New Directions in Public Opinion, ed. Adam Berinsky (New York: Routledge, 2016); and Nicholas J. G. Winter, Dangerous Frames:  How Ideas about Race and Gender Shape Public Opinion (Chicago, IL: University of Chicago Press, 2008). 21 We would have preferred to rely on the American National Election Study for this purpose. Alas, the 2016 edition of the ANES failed to include measures of gender group identification. In the four studies we analyze, the measurement of gender group identification varies only slightly. In three, women were asked how much they thought their own fate was tied up with the fate of women in general; how often they felt pride over the accomplishments of women; and how often they felt angry about the way women were treated. The 2008 CCAP includes just the latter two questions. 22 Figure 7.3 is based on logit regression estimates generated by the following equation: Vote = β0 + β1Gender Group Identification + β2Party Identification + β3 Belief in Equal Opportunity + β4 Belief in Limited Government + β5Education + β6 Region + β7Race + ε 23 All variables in the estimation are coded from zero to 1. 24 Gordon W.  Allport, “Attitudes,” in Handbook of Social Psychology, ed. Carl A. Murchison (New York: Russell & Russell, 1935, p. 806). 25 On welfare, see Martin Gilens, Why Americans Hate Welfare (Chicago, IL: University of Chicago Press, 1999); on immigration, Jack Citrin, Beth Reingold, and Donald P. Green, “American Identity and the Politics of Ethnic Change,” Journal of Politics 52(4) (1990): 1124–​1154; and on voting against Kennedy in the 1960 presidential election, Philip E. Converse, “Religion and Politics: The 1960 Election,” in Elections and the Political Order, eds. Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes (New York: John Wiley, 1966). 26 Janet T. Spencer and Robert Helmreich, “The Attitudes toward Women Scale: An Objective Instrument to Measure Attitudes toward the Rights and Roles of Women in Contemporary Society,” JSAS Catalog of Selected Documents in Psychology 2(66) (1972). 27 For a brief history of this period from the point of view of gender politics, see Christina Wolbrecht, The Politics of Women’s Rights: Parties, Positions, and Change (Princeton, NJ: Princeton University Press, 2000) and Jane J. Mansbridge, Why We Lost the ERA (Chicago, IL: University of Chicago Press, 1986). 28 The Modern Sexism scale was developed in parallel to what scholars were describing as a similar shift in racial attitudes from blatant to subtle in the post Civil Rights era. For an account of this shift, see Donald R. Kinder and Lynn M. Sanders, Divided by Color (Chicago, IL: University of Chicago Press, 1996). On modern sexism, see Janet K. Swim, Kathryn J. Aikin, Wayne S. Hall, and Barbara A. Hunter, “Sexism and Racism: Old-​Fashioned and Modern Prejudices,” Journal of Personality and Social Psychology 68(2) (1995):  199–​214; Janet K.  Swim and Laurie L.  Cohen, “Overt, Covert, and Subtle Sexism,” Psychology of Women Quarterly 21(1) (1997): 103–​118.

176  Donald Kinder, Molly E. Reynolds, and Nancy Burns Conceptually and empirically, modern sexism and what Glick and Fiske call “hostile sexism” are interchangeable. See Peter Glick and Susan T. Fiske, “The Ambivalent Sexism Inventory:  Differentiating Hostile and Benevolent Sexism,” Journal of Personality and Social Psychology 70(3) (1996): 491–​512. 29 Figure 7.4 is based on logit regression estimates generated by the following equation: Vote  =  β0 + β1Women’s Place + β2Modern Sexism + β3Party Identification + β4Racial Resentment β5Belief in Limited Government + β6 Belief in Equal Opportunity + β7Religion + β8Education + β9Region + ε 30 On the gender gap in voting in presidential elections in the decades immediately following passage of the 19th Amendment, see J.  Kevin Corder and Christina Wolbrecht, Counting Women’s Ballots (New York: Cambridge University Press, 2016). 31 There are already many scholarly papers written on the 2016 election; no doubt many more are on the way. Two particularly good treatments of gender and 2016 are: Nicholas Valentino, Carly Wayne, and Marzia Oceno, “Mobilizing Sexism: The Interaction of Emotion and Gender Attitudes in the 2016 U.S. Presidential Election,” Public Opinion Quarterly 82(S1) (2018):  213–​ 235; and Nicholas J.  G. Winter, “Ambivalent Sexism and Election 2016,” Paper prepared for presentation at the annual meeting of the American Political Science Association, Boston, August 29 to September 2, 2018. 32 Data on the record number of female candidates: http://​cawp.rutgers.edu/​footnotes/​ women-​candidates-​election-​2018-​5-​key-​data-​points-​midway-​through-​primaries. 33 On comparing gender and race, see Nancy E.  Burns and Donald R.  Kinder, “Categorical Politics: Gender, Race, and Public Opinion,” in New Directions in Public Opinion, ed. Adam Berinsky (New York: Routledge, 2012); Donald R. Kinder and Allison Dale-​Riddle, The End of Race? Obama, 2008, and Racial Politics in America (New Haven, CT: Yale University Press, 2012); and Nicholas J. G. Winter, Dangerous Frames:  How Ideas about Race and Gender Shape Public Opinion (Chicago, IL: University of Chicago Press, 2008). 34 Kinder and Dale-​Riddle (2012); Michael Tesler and David O. Sears, Obama’s Race (Chicago, IL: University of Chicago Press, 2010). 35 For an introduction into the vast literature on racial inequality, see Reynolds Farley and Walter R.  Allen, The Color Line and the Quality of Life in America (New  York:  Russell Sage Foundation, 1987); Gerald David Jaynes and Robin M.  Williams, Jr., A Common Destiny. Blacks and American Society (Washington, DC: National Academy Press, 1989); Melvin L. Oliver and Thomas Shapiro, Black Wealth/​White Wealth (New York and London: Routledge, 1997). 36 On the spatial distribution of men and women in society, see Goffman and Elanor E.  Maccoby, The Two Sexes:  Growing Up Apart, Coming Together (Cambridge, MA: Belknap Press, 1998). 37 On racial segregation, see Douglas S.  Massey and Nancy A.  Denton, American Apartheid:  Segregation and the Making of the Underclass (Cambridge:  Harvard University Press, 1993).

Chapter 8

Worldview Politics Marc Hetherington

Among the first things a student learns when taking a course about the American electorate is pretty shocking. A high percentage of us don’t care very much about politics and a truly staggering percentage knows precious little about it. It is hard to imagine these things could possibly be true in one of the world’s great democracies, but it is true. The traditional explanation scholars have given for the American public’s disinterest in politics is that it is not central enough to most people’s lives for them to make a big investment in learning the ins and outs.1 Americans have other things to do –​jobs to work, families to care for, sports to watch, and movies to attend. Although it does seem like citizens should, in a perfect world, pay attention to how they are being governed, it is understandable why they don’t. Why would people learn much about something they don’t care about? It follows from this lack of interest that the heat generated by American political conflict has been more like a toasty day at the beach than a rolling boil on the stove. In general, political conflict in America has been less intense than in other parts of the world. How passionate can people get about something that they don’t care much about? The traditional story about the American electorate no longer seems quite right. As anyone who has been around deeply polarizing political arguments at the dinner table during the holidays can attest, it sometimes feels like Americans care too much rather than not enough about politics these days. Many more people are much more passionate about politics than they were 20 or 30 years ago. That is not to say that Americans are becoming political experts. Indeed, despite the exploding availability of information, little evidence suggests that people are any better informed than they were 75 years ago when political surveys came into being. A recent survey, for example, showed that only about a quarter could name all three branches of government.2 But Americans have developed much stronger feelings when it comes to politics. Shanto Iyengar’s work in Chapter 4 of this volume tells the full story of what he and other scholars call affective polarization. Unfortunately, that passion is driven by anger rather than affection. Perhaps the key change in American public opinion over the past generation is that partisans of one side

178  Marc Hetherington

have developed more strongly negative feelings about the other party than ever before. It is not an exaggeration to say that most Democrats these days hate the Republican Party and most Republicans hate the Democratic Party. Majorities of both think that members of the other party are close-​minded, and that those who identify with the other side make them “afraid,” “angry” and “frustrated.” Worse still, nearly half of both Republicans and Democrats believe the other party’s ideas are so problematic that they threaten the country’s well-​being.3 Partisans’ surge in negative feelings about their opponents has not been accompanied by a similar surge in positive feelings about their own party. In fact, Republicans, have developed less favorable feelings about their own party of late, which probably helps to explain why an outsider like Donald Trump won the GOP nomination for president in 2016.4 In short, the present political moment is best understood in terms of increasingly negative feelings and emotions, and the often problematic consequences these feelings and emotions have. This chapter will explain why and why now.

What Divides Us Matters Readers might guess that negative feelings about the political system would cause people to disengage from it. Some probably have. With all the entertainment opportunities available these days, it has only become easier to tune politics out altogether.5 But that is not the general story. In fact, people care a lot more about who wins elections than they used to. Back in the 1960s and 1970s, only 55 to 65 percent of Americans said they cared who won the presidential election. That percentage has cleared 80 percent in every election after the one in 2000. The same pattern holds for congressional elections. In the late 1970s and early 1980s, only about half of Americans cared who won. These days the percentage is above two-​thirds.6 Indeed, for all the handwringing the media often engage in about low voter participation, some recent elections have set modern day turnout records. The 2008 presidential election attracted a higher percentage of participants than any election after 18-​to 20-​year-​olds were granted the right to vote. Participation in the 2018 midterm election was even more impressive, drawing about 50 percent of the eligible electorate to the polls, the highest percentage for a midterm election since the 1910s. For comparison’s sake, midterm turnout rates over the last 40 years or so have been mired in the high 30s or low 40s.7 Of course, anyone who has ever experienced anger and hate in other parts of their lives knows that these emotions can motivate a lot of action. Voting seems to be one of them. Knowing that politics today is driven by negative emotions is helpful to some extent, but it leaves a puzzle. Why do so many Americans all of a sudden seem to care so much about something they typically haven’t cared much about? The answer is to be found in what divides the parties. Political scientists refer to the division between the parties as a cleavage line.8 That line is different now

Worldview Politics  179

than it was before. Theoretically, the cleavage that divides the parties from each other could be driven by disagreements about an infinite number of things. During the Civil War period, it was what to do about slavery. Often the divide is driven by events that simply cannot be ignored by politicians. Should the impact of climate change worsen in the future, for example, the central cleavage dividing Republicans from Democrats might be organized around a set of proposals that deal with how best to combat it. In general, parties are going to disagree about issues. That is what parties do. But the key to understanding how intense political conflict becomes is identifying what, specifically, is the dominant difference between the two major parties. Beginning in the 1930s, the major division between Republicans and Democrats, conservatives and liberals was how big the government should be. This became the cleavage line because of the Great Depression, the worst economic calamity in the past 100 years. In the last years of Herbert Hoover’s administration, the stock market crashed, economic growth plummeted, and unemployment reached 25  percent. Democrats under Franklin Roosevelt’s leadership championed an expanded role for the federal government –​using it to create more programs and spend more money –​to bring the country out of the doldrums. The Republicans of the time weren’t opposed to government involvement, but they preferred for government to do less and private enterprise to do more. As a consequence, how much government ought to be involved in the economy and what role it should play in alleviating human suffering separated the parties for decades to follow. Democrats argued government needed to do more to overcome economic inequities, while the Republicans argued government should do less. Importantly, neither side argued that government should do nothing at all. This dividing line is relatively complicated and, when not in the throes of an economic depression, beyond the personal experience of most people. The lion’s share of Americans don’t have a strong sense of what government does in the first place, much less whether it should do more of it or less of it.9 Those who spend a fair amount of time thinking about politics might develop something of a philosophy of when government ought to intervene in the economic marketplace and when it ought not, but not many people spend enough time to develop a philosophy of governing.10 Because the issues were remote and required a fair amount of information to understand them, the intensity of feelings that most Americans had about politics during the New Deal era was not especially strong. More recently, the dividing line between the parties has changed, and this is why feelings have grown so intense of late. As the issue agenda began to evolve in the 1960s to include matters that hadn’t divided the parties before –​ such as civil rights for African-​Americans, women’s equality, various matters that brought religion into the political sphere, and, much later, terrorism –​the cleavage line between the parties evolved along with it. Now it is organized around Americans’ fundamental worldviews.11 By worldview, I mean “someone’s

180  Marc Hetherington

deeply ingrained beliefs about the nature of the world and the priorities a good society should have” (x–​xi).12 Rather than a philosophy about governing, it is more a philosophy about life. It isn’t complicated. Everyone has one. And, most important, people feel passionately about it. At the core of the worldviews people adopt is how dangerous they perceive the world to be. Is it fraught with danger, or is it safe to explore? Some tend to think it is more like a playground. Others tend to think it is more like a snake pit. Because fear is perhaps humans’ most primal instinct, it has a unique capacity to organize people’s reaction to the world around them. People have always had differing worldviews, probably going back to cavepeople times. Some wanted to explore more, and some wanted to play it safe. And the country has always had political parties. Why, then, haven’t we always had the kind of angry polarization that we are experiencing today? The reason is that the parties did not used to appeal to these worldviews like they do today. Over the past couple decades, worldview has become neatly mapped onto party conflict. Consider some striking evidence from the 2016 election. Douglas Rivers, among the best-​respected survey researchers in the business, asked Americans to choose which of the following two statements jibed better with how they understood the world. 1) Our lives are threatened by terrorists, criminals, and immigrants and our priority should be to protect ourselves. 2) It’s a big, beautiful world, mostly full of good people and we must find a way to embrace each other and not allow ourselves to become isolated. About 80 percent of Donald Trump supporters chose the first statement while about 80  percent of Hillary Clinton supporters chose the second. Attitudinal differences of this size almost never exist in public opinion research. These results clearly demonstrate that the two sides do not see the world the same way.13

Why Worldview Politics is Inherently Polarizing This survey result is not an anomaly. It results from decades of rhetoric from Republicans and Democrats painting differing pictures of the world around us. Donald Trump’s speeches during the campaign suggested a particularly perilous view. He regularly would read, of all things, a poem to his throng of supporters. Titled “The Snake,” and written by Oscar Brown, Jr., it is based on one of Aesop’s fables. Trump delivered the poem to illuminate the dangers of allowing immigrants and refugees to enter the United States. In the poem, a gentle-​hearted woman nurses a sick, freezing snake back to health. In return for her good deed, the snake administers to her a fatal bite rather than thanking her for her kindness. After she cries out in horror at the betrayal, the snake replies “Oh shut up, silly woman,” with a grin. “You knew damned well I was a snake before you took me in.” The moral of the story is that people can never be too

Worldview Politics  181

careful and must never let down their guard. This is common-​sense wisdom to people on one side of the worldview divide. Outsiders might end up hurting us. We lend a hand to refugees at our own peril. This style of politics is inherently polarizing because not everyone sees the metaphor of the snake the same way. Those who don’t perceive the world as a dangerous snake pit interpret Trump’s use of the poem as a disgusting incitement of prejudice. Immigrants and refugees to them aren’t snakes at all, but rather people in desperate need of help. Americans shouldn’t judge them because of where they come from, what language they speak, or what god they worship. The point here is not to suggest one side of the divide is right and one is wrong. The key is to show how incompatible the views are of those on opposite sides of this divide. Consider how those on opposite sides of this worldview divide responded to the September 11 terrorist attacks, an event that appears to have been central to fixing the worldview divide between the parties in place. An example of an interaction between a father and son that took place after the 2004 election illustrates this nicely. The son, a liberal legal services lawyer, and his father, a conservative veteran of the Marine Corps, sit on opposite sides of the worldview divide. September 11 had happened recently enough that it still occupied Americans’ thoughts, but it was far enough in the past that most were no longer in the grip of palpable fear. The son and his father found themselves talking about the trade-​offs between protecting civil liberties and protecting the security of Americans. The legal services lawyer predictably argued for the critical importance of maintaining 1st Amendment and 4th Amendment protections even in the face of the potentially grave threats posed by world terrorism. Yet the father, the Marine Corps veteran, saw the situation differently. To him, those high-​minded principles were all well and good most of the time, but they wouldn’t matter very much to Americans if they were dead from a terrorist attack. Talk about the irreconcilability of worldview conflict.14 Just like in “The Snake,” the son and his father differed in how dangerous they thought the world was and, hence, what the real problem was when it came to dealing with terrorism. Because the father thought the threat from terrorists remained intense, he believed it foolhardy to risk underestimating the potential damage. That is what people did in a September 10th, 2001 world, and the costs were catastrophic. Leaders couldn’t afford to let their guard down again. The danger to him was the terrorists. In contrast, the son did not perceive the threat posed by terrorists to be grave. It had been years since the attacks on New York and Washington occurred, and he didn’t think it likely that something like it would happen again. His worldview, then, allowed him to see value in things like protecting civil liberties rather than just maximizing safety. As such, the problem for him was different. The extra-​Constitutional steps taken by the Bush administration in the wake of 9/​11 to maximize safety were, in his mind, the real threat to the nation. Ideals were critical, not physical security.

182  Marc Hetherington

This is the nub of the problem when it comes to the worldview divide. Those on opposite sides see the problems facing the country as different. One side thinks it is maintaining safety and security. If you see the world like that, anyone who doesn’t see it that way is a potential danger to the nation’s well-​ being. Because those on the other side of the worldview divide worry less about safety and security, the problem for them is the potentially problematic steps leaders might take in the name of safety and security. If you see the world like that, anyone who doesn’t is actually the real danger to the nation’s well-​being. The opposites simply don’t get each other. Their worldviews are incompatible when they are applied to political issues. As the Republican and Democratic party bases have become increasingly homogeneous by worldview, politics has taken on an angrier and darker cast.

Where Worldviews Come From An intriguing part of this story is how people come to adopt different worldviews. A  relatively new approach to political research, which Gonzalez et  al. examine in Chapter  10 of this volume, uncovers differences between liberals and conservatives that run more than skin deep. Indeed, it appears that those on opposite sides of the worldview spectrum may be getting quite different signals from their bodies and minds about the world around them. Many of these differences seem to lead them to have different reactions before their conscious thoughts even begin. Let me be clear from the outset that these studies tend to involve small groups of subjects that are not representative of the country as a whole. Putting too much stock in any single study might be risky. But the studies taken together tell the same basic story, which provides more confidence in what they suggest. Conservatives are warier than liberals. Because these studies are a little unconventional, they require a bit of extra explanation. In all these studies, researchers ask subjects a battery of questions to measure how conservative or liberal they are in advance of measuring their responses to various stimuli applied in the laboratory. In that sense, people have revealed themselves to be more or less conservative before any scientific poking and prodding takes place. One of the tests involves what conservatives and liberals tend to fix their gaze on when given the choice between looking at things that are more and less dangerous. Specifically, researchers provide people a palette of objects, instructing them that they can choose to look at any or all of them. Eye-​tracking devices measure how long subjects dwell on each of the objects. The result is striking –​conservatives tend to spend much more time looking at potentially dangerous objects than liberals do.15 Another of these tests involves how easily startled people are. When we find ourselves in a dangerous situation, our nervous system reacts automatically to help us to get out of harm’s way. In that sense, the startle reflex is an important device for keeping human beings safe. Everyone startles when something

Worldview Politics  183

unexpected happens. Think about what happens when a car cuts you off on the highway or when a character in a scary movie pops up out of nowhere. It turns out, however, that some people’s response is stronger than others. Scholars measure the strength of a person’s startle reflex by counting the number of times he or she blinks after a startling event occurs. To do that, they outfit subjects with headphones that, at a random moment, blow a loud noise into their ears. Then, the researchers use something called a blink amplitude meter to count how often subjects blink in response. It turns out that conservatives blink much more than liberals do after being startled.16 Another reflex designed by nature to keep humans safe is the gag reflex. It keeps people from ingesting things that might prove harmful to their bodies. If you have ever taken a whiff of a container of milk well after its expiration date, you’ve almost surely experienced the gag reflex. The body recoils from the potential threat, and it feels like you might vomit. Like with the startle reflex, it turns out that some people have stronger gag reflexes than others. Researchers measure this by accounting for how much people’s hands sweat after seeing disgusting images, such as a person eating worms or a badly soiled toilet. None of these images are political in any way. They are just objectively gross. As the reader might have guessed by now, the results showed that conservatives had significantly sweatier hands than liberals did in response to the disgusting pictures.17 For those who find these types of studies unconvincing, survey research also shows that disgust and political orientation are correlated. Rather than a small group of subjects in a college town, this work involves a thousand or more people responding to questions in political surveys. In addition to answering questions about whether they are conservative or liberal, pro-​immigration or anti-​immigration, and so forth, they also report how disgusted they are by various things in life, such as smelling spoiled milk, encountering vomit on the street, or having their soup stirred by a flyswatter. The evidence is consistent with the poking and prodding done in the lab. Almost everyone expresses disgust to some degree, but conservatives express more disgust, on average, than liberals do. Hence it should not be surprising that those who express more unease about using a public restroom are also the same people who tend to be less supportive of immigration.18 These physiological reactions seem to contribute to physiological differences in liberals’ and conservatives’ brains. Studies in both the United States and the United Kingdom demonstrate how the amygdala, which is the part of the brain that governs fight or flight reactions, tends to be larger among conservatives than liberals.19 This makes sense given the fact that conservatives’ startle and gag reflexes are sharper and that they are more on watch for dangerous things. Brain areas that are used a lot tend to grow in size while those used less often tend to shrink. The point of describing all this research is to suggest that liberals and conservatives might react very differently to the same situations. And one

184  Marc Hetherington

area where those differences in reaction are most likely to occur is when it comes to situations that could be considered dangerous. Because perceptions of danger and fear are automatic and not the result of a reasoned and deliberative process, it means that the first instincts of those on opposite sides of the worldview divide will be different. No wonder it is hard for those on one side of a divide like this to understand how the other side can possibly see the world as they do. People adopt different habits of the mind to cope with the degree to which they believe danger lurks everywhere. One example is the development of personality characteristics. Most personality psychologists believe that personality can be best considered in five domains –​Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. There is now significant agreement among political scientists that openness and conscientiousness are strongly associated with people’s political orientations.20 Those who are more open to new experience tend to be much more likely to be liberals. Those who are more conscientious tend to be much more likely to be conservatives. What it means to be open to experience is reasonably self-​explanatory. Those who are open are more inclined to want to try out new things, to be around people who might be different. They are fine with changing up how things have been done in the past. Edgy art exhibits and Ethiopian food might be especially interesting to someone who is open to new experiences (and isn’t from Ethiopia). One can easily see how seeing the world as a safe place might allow someone to adopt a personality characteristic like this. On the other end of this continuum are those who are more closed to new experiences. They’d rather stick with the tried and true. New things might be great, but, to those who are not especially open to new experience, it is not worth the risk of finding out for sure. If your view is that the world is dangerous, being more closed is a reasonable adaptation. Conscientiousness is another habit of the mind that divides the political left from the political right. This is a tendency toward dependability and dutifulness. Those who are high in conscientiousness value things like order and organization. If you experience people who are always on time for meetings and their rooms or offices are spotless, they probably score high in conscientiousness. It also turns out that they tend to be conservative. People who value order and organization tend to embrace hierarchies and traditions that keep things predictable. They embrace tightknit communities of people who are likeminded and, probably, look and act much the same. In a dangerous world, a strong community is important because trustworthy allies can help keep dangers at bay. Liberals, in contrast, are less interested in conventions and traditions. Indeed, they might perceive traditions as problematic because they don’t allow people to express their individuality. Perceiving the world as less dangerous provides that luxury. Worldviews are like personality characteristics in that they are habits of the mind that help people navigate the world without having to think very much

Worldview Politics  185

about it. We’ve adopted two labels for opposite ends of the worldview divide –​ fixed and fluid. The term “fixed” describes people who are warier of social and cultural changes and hence more set in their ways, more suspicious of outsiders, and more comfortable with the familiar and the predictable. People we call “fluid,” on the other hand, support changing social and cultural norms, are excited by things that are new and novel, and are open to, and welcoming of, people who look and sound different (pp. xii).21 The mantra to the “fixed” is that the world is a dangerous place. Because they view our hold on the world as tenuous, it is important to follow directions. Relying on established hierarchies and pathways provide order to a potentially chaotic world. Traditions are useful, too, because they have proved their value over time. They also make the world more predictable, which is a comfort when you fear something bad could happen. As such, the fixed are wary of social change. Such change might be fine, but why risk it? If it ain’t broke, don’t fix it. Finally, the fixed believe that it is important to be vigilant. When it comes to physical safety, it is critical to keep “us” safe from “them.” The mantra for those with “fluid” worldviews is that the world is a playground to explore. It is safe to follow your own path rather than follow directions. For the fluid, traditions go by another, less flattering, name –​conformity. Perhaps even more problematic, unfair hierarchies that advantage men over women, Whites over minorities, old over young, straight over gay, and so forth remain frozen in place unless they are challenged. Social change, then, is a must to overcome old injustices. It is fine if people take an unconventional approach to life even if it is different from others in their community. To the fluid, the individual is more important than the community. The latter should adapt to allow for individual self-​expression. The next step for a survey researcher is to get people to reveal something as abstract as their worldview. As strange as it might sound, scholars have found that the best device is to ask people what qualities they think are most important for children to have.22 You read that right –​desirable qualities in children. Starting in the early 1990s, the American National Election Study began to ask four questions. The preamble to the questions is the following: Although there are a number of qualities that people feel that children should have, every person thinks that some are more important than others. I am going to read you pairs of desirable qualities. Please tell me which one you think is more important for a child to have. The four pairs of choices are 1) respect for elders versus independence, 2) curiosity versus good manners, 3)  obedience versus self-​reliance, and 4)  being considerate or being well-​behaved. Those who choose respect for elders, good

186  Marc Hetherington

manners, obedience, and being well-​behaved reveal themselves as having the most fixed worldviews. Those who choose independence, curiosity, self-​ reliance, and being considerate reveal themselves as having the most fluid worldviews. We refer to those who have some mixture of the two different worldviews as being “mixed.”

How Different Worldviews Became Affixed to the Parties A worldview isn’t a political identity. It is a way of understanding the nature of the world. Is it safe or dangerous? Should we protect traditional ways of doing things or is it safe to challenge traditions? Partisanship isn’t a worldview but is an identity with a political group. What it has meant to be a Republican or Democrat has changed a lot over the years. Party identity is not a philosophy about life. Today’s political acrimony results from Americans’ worldviews becoming married to their partisanship. Figure 8.1 illustrates this key change in American politics. When the ANES first asked the worldview questions back in 1992, no association existed between whether people were fixed or fluid and whether they were Republicans or Democrats. Among those with a fixed worldview, about 50  percent identified as Democrats, and about 40  percent identified as Republicans. The same was true of those with a fluid worldview  –​about 50  percent said they were Democrats while about 40  percent said they were Republicans. By 2016, however, worldview and party were strongly associated with each other. Among those with fixed worldviews, 60  percent identified 80 70

Percent

60 50

Fluid

40

Fixed

30 20 10

Figure 8.1 Worldview and Partisanship, 1992 versus 2016

ep ub lic

an

t

1992

R

nd en de pe

In

D em oc ra t

an ep ub lic R

nd en t de pe

In

D

em

oc ra t

0

2016

Worldview Politics  187

as Republicans and less than a quarter identified as Democrats. Among the fluid, 70  percent identified as Democrats and a paltry 20  percent said they were Republicans. That is a giant change in a relatively short period of time. Americans have sorted their worldviews into the political parties whose ideas represent them best. The marriage between worldview and party happened because a cascade of new issues subsumed the old size of government conflict that once divided the parties, creating a new party divide. There is no reason to think that worldview ought to have anything to do with whether people wanted government to spend more or less money, which is evident in the absence of an association between party and worldview in 1992. But as race, feminism, law and order, religion, gun rights, sexual orientation, the proper way to combat terrorism, and immigration took center stage, things changed. All these issues have one thing in common; people’s preferences on all of them are informed by their worldviews. In decades past, these issues tended to divide the parties internally, if they were salient at all. Take civil rights for African-​Americans as an example. When the landmark civil rights laws passed in the 1960s, Democrats and Republicans were on both sides of the issue. The Democratic President, Lyndon Johnson, and other liberal Democrats were central to the passage of these bills, but the most vociferous opposition came also from Democrats, southerners who were mostly abject racists at that time. Indeed, a higher percentage of Republican lawmakers voted for the Civil Rights Act of 1964 and the Voting Rights Act of 1965 than Democrats.23 The parties’ positions on racial issues evolved over time, with Democrats consistently taking the pro-​African American position on these matters and Republicans taking the opposite. Indeed, nothing is more important to the evolution of the South from an overwhelmingly pro-​ Democratic region in the mid-​twentieth century to an overwhelmingly pro-​ Republican region today than the parties taking clear and opposite positions on race-​based issues. Equal rights for women did not emerge as a partisan issue until the late 1970s and early 1980s. On abortion rights –​perhaps the key women’s rights issue of the time –​the parties’ positions in the 1976 presidential election were opposite what they are today. The Democrat, Jimmy Carter, was the pro-​life candidate and Gerald Ford, the Republican, was the pro-​choice candidate. Ronald Reagan, a dynamic Republican elected in a landslide in 1980, pulled the party toward the traditional side on abortion, other matters involving women’s equality like the Equal Rights Amendment, and on a range of religious issues as well. Democrats drifted in the opposite direction on all these issues. As a result, traditional fixed worldview Democrats in the 1980s started to vote for Republicans during this era, even if they didn’t change their party identity yet. Similarly, fluid worldview Republicans likely began to wonder whether their party was sufficiently modern in its thinking. Homosexuality burst on the scene as a political issue in the 1980s, with the AIDS crisis. The sad fact is that neither side of the party divide showed much

188  Marc Hetherington

sympathy for the LGBT community during this period despite the tragic loss of life caused by this killer disease. Some members of Congress who represented cities where LGBT people disproportionately lived supported their cause, but that was about it.24 Gay rights took on a more partisan cast in the 1990s, after Bill Clinton forwarded a policy to allow gay people to serve in the military. Republicans, already having taken the more “traditional” position when it came to race and gender equality, found it quite natural to fall onto the same side when it came to sexual orientation in the 1990s and later in the early 2000s when same sex marriage became a salient issue. Democrats who had embraced the social change inherent in the civil rights and gender equality movements found it natural to embrace the same principles when it came to sexual orientation. Immigration is a fourth issue area over which the parties’ stances used to differ from what they are today. Although immigration was not an especially salient issue for most of the late twentieth century, many Republicans of that time were open to a more generous approach while many Democrats wanted more limits. This is because the New Deal party cleavage featured Republicans supporting the interests of business while Democrats supported labor. Because businesses benefit from lower labor costs, more immigration can be a boon to them. With more people in the workforce, wages go down as more people compete to fill a fixed number of jobs. Because labor unions have an interest in keeping wages high, they would rather limit the number of potential workers coming into the country who are willing to work for less money. Ronald Reagan and George H.W. Bush –​both Republican presidents –​signed major immigration bills that increased the number of immigrants in the country. With the Democrats holding majorities in at least one house of Congress during this entire period, at least some congressional Democrats needed to support these measures for them to become law. In that sense, immigration was not an especially partisan issue in the late twentieth century, and, to the extent that it was, Republicans were often more sympathetic. Clearly times have changed. Today, immigration is a deeply partisan issue because it fits the current worldview cleavage so well. Consistent with the notion that it is a dangerous world out there, most Republican leaders have taken a strongly anti-​immigration stance since the early 2000s. Among the last immigration moderates in the Republican party was George W. Bush, who attempted to pass a comprehensive immigration reform plan during his second term as president. The legislation would have provided a path to citizenship to some people who were in the country without documentation in addition to providing more money for border security. His own party killed those efforts in Congress. Democrats for their part have become increasingly supportive of immigration. Among their key voting constituencies are Hispanic-​Americans and Asian-​Americans. Often immigrants themselves, members of these groups tend to favor a more open immigration system. This impulse toward inclusion also sounds good to the ear of Whites with fluid worldviews because they are less worried about the dangers new arrivals might bring and because they are

Worldview Politics  189

more open to increasing diversity in general. The key point is that Democratic and Republican leaders have taken clear and opposing positions on this issue as well. The last issue area I’ll discuss is how best to cope with threats from outside the country. Specifically, how to deal with terrorism and whether it’s best to use military force are two areas where the parties have taken clear and opposing positions. In both cases they are consistent with the fixed–​fluid worldview divide. Republicans have tended to adopt a more no-​holds-​barred approach to combatting terrorism. During the George W.  Bush presidency, they championed a range of policies that promised more security even though it sometimes meant Americans trading some civil liberties in return. More controversially, Bush allowed the use of “enhanced interrogation techniques” against terrorist suspects, which critics suggested amounted to torture. Donald Trump took this stance a bit further, explicitly embracing torture and whatever else might be necessary to keep Americans safe. When it comes to using military assets, Republicans from George W. Bush to Trump have been willing to deploy them more often and more unilaterally. Democrats, in contrast, have been less willing to deploy the use of military force and, when they have, the support of global partners has been more important to their approach.

Evidence that Worldview Organizes Opinion on the New Partisan Cleavage At this point, I’ve described the rise of a new set of issues dividing Republicans from Democrats. Many of these issues used to divide the parties internally back when the dominant cleavage was the size and scope of government. Now race, gender, immigration, religion, and terrorism divide one party from the other. They are also all hot button issues that people care deeply about. But what do race and immigration have to do with gender equality? What do all the domestic issues, in general, have to do with keeping Americans secure from terrorist threats? Our research shows that Americans’ preferences on all these issues turn on the same thing –​their worldviews.25 People at opposite ends of the worldview spectrum differ by 40, 50, and sometimes 60 percentage points on them. This is why Americans’ worldviews now map so neatly onto their party identifications. A few examples might be useful to show the connection that exists between people’s worldviews and all these matters. Let’s first explore Whites’ attitudes about African-​Americans. Since the 1980s, scholars have been asking Americans a battery of questions designed to tap their level of “racial resentment.” This is not old-​fashioned racism. Instead it seeks to understand why people think African-​ Americans are generally not as well off as Whites. Are structural factors mostly to blame, or does the fault lie mostly with African-​Americans themselves? As Figure  8.2 makes clear, those at opposite ends of the worldview divide have different ideas about the causes. When asked whether they agreed or disagreed

190  Marc Hetherington

100 90 80

Percent

70 60

Fixed

50

Fluid

40 30 20 10 0 Disagree Blacks have gotten less than they deserve

Agree

Disagree

Blacks only need Discrimination makes it to try harder to be as difficult for Blacks to move up well off as Whites

Figure 8.2 Worldview and Racial Attitudes

that “Blacks only needed to try harder to be as well off as Whites,” about 60 percent of the fixed agreed. Only 20 percent of the fluid did. When asked whether they agreed or disagreed that “Discrimination makes it difficult for Blacks to move up,” nearly 70 percent of the fixed disagreed while only about 20 percent of the fluid did. These are truly massive differences in opinion. We see similar sized fault lines between the fixed and fluid on gender equality. Some results appear in Figure  8.3. When asked whether “women should return to their traditional roles,” nearly 90  percent of the fluid disagreed. Only about half of the fixed did. When asked whether it is “better if the man achieves and the woman stays at home,” the divide was about the same, with 90 percent of the fluid disagreeing and a shade below 50 percent of the fixed taking that position. The same sized chasm between the fixed and fluid turns up when it comes to LGBT issues. The groups differ in their support for same sex marriage and same sex adoption by more than 40 percentage points. When asked whether they were “frustrated” by the notion of transgendered bathrooms, nearly 60 percent of the fixed said “extremely” while only about 10 percent of the fluid did. Attitudes about immigrants and immigration are also strongly related to worldview. Figure 8.4 shows the details. Among those with fixed worldviews, nearly 70 percent believe “immigrants threaten traditional American customs.” Only about 15 percent of the fluid believes that is the case. About 70 percent of the fixed also say they are “bothered when they encounter someone who speaks little English” compared with about a quarter of those with fluid worldviews.

100 90 80

Percent

70 60 Fixed

50

Fluid

40 30 20 10 0 Disagree Women should return to traditional roles

Disagree Disagree Disagree Harrassment Better if man Women equality = complaints achieves special favors cause problems woman stays home

Figure 8.3 Worldview and Gender Attitudes

100 90 80 Percent agree

70 60

Fixed

50

Fluid

40 30 20 10 0 Believe immigrants challenge traditional American customs

Bothered when encounter someone who speaks little English

Figure 8.4 Worldview and Immigration

Build a border wall

192  Marc Hetherington

It is not surprising, then, that support for building a wall on the U.S.–​Mexico border has the support of about 70 percent of those with fixed worldviews and only about 20 percent of those with fluid ones. Finally, attitudes about military force differ by a lot as well. When asked, for instance, whether they favored the use of force over diplomacy, 60 percent of the fixed said yes, while only a quarter of the fluid did. Although space prohibits providing further examples, it is also the case that worldview is strongly related to a range of other political matters, from the proper way to interpret the Constitution to gun rights to support for democracy itself.26 The reason is because people’s worldviews inform the degree to which they perceive danger around them. Physical threats will be especially concerning, but social change is troubling, too. Existing traditions and hierarchies have maintained order for millennia. Racial and gender equality threaten those hierarchies. LGBT people challenge those traditions. If, in contrast, people think the world is safe, they won’t see refugees and immigrants as trying to infiltrate the country to do harm. Instead they see them as needing help. They are unlikely to perceive identity groups vying for equality as threats. Instead they’ll tend to see old traditions and hierarchies as threats, themselves, because they perpetuate discrimination against groups that have been lower in the social hierarchy in the past for what they see as no good reason.

Worldview and the Nonpolitical World As though the political differences aren’t enough between the fixed and the fluid, they are only the beginning of the story. Worldviews don’t just affect people’s choices about politics. They affect their life choices as well. As a result, people on opposite sides of the worldview divide tend to live, go to school, work, and worship apart from each other.27 This wouldn’t make much difference politically except for the fact that people with fixed worldviews tend to be Republicans and those with fluid worldviews tend to be Democrats. That means Republicans and Democrats will have little contact with each other, which will make it nearly impossible for them to overcome the negative, sometimes prejudiced, feelings that they have for each other. In generations past, Republicans and Democrats were relatively likely to live among each other. One way to get a sense of how things have changed is to examine how counties vote. If one party or the other won by huge margins, it would suggest there was a disproportionate number of either Republicans or Democrats living in that country. Before worldview began to divide Republicans from Democrats, counties were much more evenly balanced than they are today. In 1976, for example, Jimmy Carter defeated Gerald Ford in a very close election. Less than 25 percent of Americans lived in so-​called landslide counties, in which either Ford or Carter won at least 60  percent of the vote.28 Compare that with 2012, an election that produced a similar margin of

Worldview Politics  193

victory as in 1976. When Barack Obama defeated Mitt Romney, over half of Americans lived in such counties. The point here is that far more people than before live in counties that are homogeneous by party. Because politics isn’t especially important to most Americans, it is unlikely people are moving to places so they can live closer to people of the same partisan stripe. It is a pretty rare conversation between a prospective homebuyer and a real estate agent that starts with something like “Only show me houses near where the Republicans live.” Rather, the sorting process that has inarguably taken place more likely results from the fact that Republicans and Democrats have different worldviews today whereas the parties were not sorted by worldview a generation ago. Fluid worldview people, with their openness to new experiences tend to favor cities, with their diversity, unpredictability, cool ethnic restaurants, and modern approach to life. Fixed worldview people are unlikely to find that environment especially attractive. More traditional, homogeneous areas that they consider safe to live in would appeal to their more conscientious and wary nature. This has always been the case for people with different worldviews. The difference is that, today, worldview is now connected to people’s party identification, so now partisans are less likely to live near each other than before. They are also less likely to go to school together. Because they don’t live near each other, they won’t tend to attend the same elementary and high schools. But the same is also true for college. That is because people with fixed worldviews tend to go to school for much less time than do those with fluid worldviews. Based on data from 2016, for example, 51 percent of the fixed had a high school diploma or less, compared with only about a quarter of the fluid. Forty-​two percent of the fluid had four-​year college degrees, with 17 percent also earning post-​graduate degrees. Among the fixed, those percentages are 18 and 6 percent, respectively.29 Success in college and especially graduate school is often all about challenging old ideas and traditional conventions. Those with fluid worldviews will, of course, find this much more attractive than those with fixed worldviews. It is noteworthy, too, that as the parties have become sorted by worldview, the relationship between education and party identification has changed. During the New Deal era, lower education voters were more likely to be Democrats and high education voters were more likely to be Republicans. The opposite is true now. Worldview is also strongly related to people’s religious commitments. Tradition-​minded fixed worldview people are both more likely to identify with a church than fluid worldview people and more likely to attend church. In addition, the fixed are far more likely than the fluid to pray regularly. As for religious denomination, the fixed most often identify as either Protestant or Roman Catholic. Those with a fixed worldview are especially likely to identify with an evangelical religious tradition and report having had a born-​again experience. In contrast, the most common answer to the religious denomination question among the fluid was “nothing in particular.” The combined

194  Marc Hetherington

percentage of “agnostic” and “atheist” outpolled the combination of Protestant and Roman Catholic. Clearly they are more likely to go to brunch on Sundays than they are to go to religious services. Although these differences might not seem especially relevant, they could be. Among the most concerning developments that have occurred in American politics is that those on opposite sides of the party divide seem to be willing to discriminate against each other, a tendency we see in a number of recent studies. In one, partisans were asked whether they would be unhappy if one of their children married someone from the opposite political party. Scholars had actually done this before back in 1960, and found that a vanishingly small percentage found this notion discomfiting at all, with the numbers in the low single digits. In the 2010s, however, about a third of Democrats expressed unhappiness at the notion of partisan inter-​marriage, and nearly half of Republicans did.30 A  follow up article found more evidence of partisan discrimination. When presented with the opportunity in a survey experiment to give a scholarship to a student of either their own or the opposite party, partisans were much more likely to choose the student who shared their partisanship, even if that student had an objectively worse academic record than one who identified with the other party.31 The latter study also turned up some racial discrimination in awarding these scholarships, but the partisan discrimination was much stronger than the racial discrimination. We know from the literature on prejudice that the best way to overcome it is for people to come into contact with each other more often.32 When people are prejudiced against certain groups, the members of that group become caricatures. Getting to know those from the discriminated against group can help people to overcome that kind of stereotyping. In that sense, racial segregation can be seen as a tool that helped maintain prejudiced attitudes against African-​Americans. When Whites and Blacks began to have more opportunities to spend time with each other, however, those of different races had more opportunity to see the other as actual people rather than as gross stereotypes. Given that partisans are today much less likely than before to encounter each other in their everyday lives, it serves to exacerbate the partisan prejudice present in American politics.

And Then There is the Truly Strange Stuff A final consequence of a party system divided by worldview is that many personal tastes that Democrats and Republicans have differ from each other. Specifically, they differ in what they like to eat, where they shop, what they drink, the pets they prefer, and so many more things. I am not arguing here that people have these tastes because they are Democrats or Republicans. In other words, I  do not think that people’s partisanship causes them to prefer cats to dogs and hybrid vehicles over pickup trucks. Rather pet ownership, vehicle preferences, and so much more are associated today with people’s

Worldview Politics  195

partisanship. This is because Americans’ worldviews inform both their political predispositions and their nonpolitical tastes. Even if the relationships I describe below are not causal, the fact that they exist is still important. Such differences in personal taste provide those on opposite sides of the party divide yet another reason to think they have nothing in common with their opponents. Not only do they differ on political issues like immigration and gay rights, they differ on the things that people really care about, like the shows they watch, the music they listen to, and the books they read. Let’s start with food and drink. A couple of studies have found that Democrats tend to favor vegetables more than Republicans and that Republicans are more likely to be meat-eaters than Democrats.33 When it comes to cuisine type, Democrats are more likely go for ethnic food while Republicans prefer to eat American.34 Democrats are more likely than Republicans to frequent independent restaurants while the Republicans are more likely than Democrats to go to chain restaurants. Vice President Mike Pence’s trip to New York City provides a telling anecdote. He and his wife dined at Applebee’s rather than one of the city’s world-​famous restaurants, remarking happily afterward that it was just like the one that they usually frequent in their native Indiana. Beer preferences differ by party, too. Republicans have a penchant for light beer, in particular. Coors Light, Bud Light, Miller Light, these macro-​brands are all consumed disproportionately by Republicans. Democrats tend to go for more niche breweries that make beers darker in color and stronger in taste. Indeed studies turn up a huge micro-​brew skew with Samuel Adams the only craft brew maker that Republicans consume more than Democrats.35 Perhaps this is because it is also the oldest of the craft breweries, which would square with the fixed’s preference for things that are tried and true. In general, more hops means disproportionate consumption by Democrats. Although the reasons seem less likely related to differing worldviews, Republicans and Democrats also tend toward different types of liquor. Republicans favor brown ones, such as bourbon and scotch, while Democrats favor light ones, like gin and vodka.36 Music provides yet another fault line between the parties, further driving a wedge between partisans of opposite stripes. Facebook carried out a study that examined people who “liked” a political candidate of one party and also a musical act. Those who gave a like to Republican candidates also tended to like country music artists, while those who gave a like to a Democratic candidate tended to like just about any music other than country. It didn’t matter whether it was new country or old country  –​George Strait, Lady Antebellum, Blake Shelton, etc.  –​Republicans liked them disproportionately. The acts with big Democratic skews went across the range of music from reggae star Bob Marley, to classic rockers The Beatles, to pop artists like Michael Jackson and Beyoncé.37 All this suggests that, if Republicans and Democrats got together for dinner, they’d have a hard time agreeing on what to eat, what to drink, or what kind of music to listen to. Because of the worldview divide, partisans are now worlds apart on a personal level, too.

196  Marc Hetherington

The title of my recent book with Jonathan Weiler is Prius or Pickup? which suggests that another of the ways Republicans and Democrats differ is in their vehicle choices. The data on this are somewhat patchy, although personal experience certainly suggests it is true. The notion of a partisan vehicle divide first dawned on me about ten years ago when I used to pick up my son from elementary school in the Republican suburbs outside Nashville. From my Subaru sedan, I often could not see the front of the school because behemoth trucks and sport utility vehicles surrounded me. An online repair service actually did a study of what it termed unusually popular cars in congressional districts that were represented by Republicans or Democrats. In the blue congressional districts, the most unusually popular car was, of course, the Toyota Prius. Almost all the cars in the blue district top ten were small and foreign made. That was in stark contrast to the red districts where American made trucks and SUVs dominated the top ten.38 This makes sense from a worldview perspective. If you view the world as a dangerous place, what better hedge against potential pitfalls than driving a giant vehicle that can take on just about anything short of a tractor trailer? If you think the world is safe, you can focus on things other than safety like gas mileage and protecting the environment. More than one study has also demonstrated a relatively robust relationship between party identification and what type of pets people own. In fact, one of these studies broke the map of the United States into states that were disproportionately dog states and those that were disproportionately cat states. It is not a perfect fit, but that map bears much more than a faint resemblance to the Electoral College maps that voters have produced since 2000.39 The northeast, upper Midwest, and Pacific Coast tend to be cat country, which is also where Democrats dominate in presidential elections. The South and Great Plains tend to prefer dogs to cats and Republicans to Democrats for president. Of course, plenty of Democrats have dogs, but studies suggest that they might value different things in their dogs. Democrats appear to like smaller dogs and Republicans bigger ones. Again the relationship to danger is not hard to see. Neither a cat nor a small dog is going to keep you and your family safe from an intruder. A big dog with a bigger bark might. These associations only scratch the surface. Partisans also differ in their tastes in television shows, movies, books, and much more.40 These partisan differences would not have existed a generation ago because the party system was not divided by worldview. Because it is today, these worldview differences manifest as party differences, too. As fun as these relationships are to explore, it is important for readers to understand that the correlations between party and personal tastes are far from perfect. There are plenty of Democrats who drive trucks. Lots of Republicans like ethnic food. More important than the actual behavior of Republicans and Democrats is the pictures partisans have in mind about the other side. The fact that these differences exist between the parties, on average, provides those on

Worldview Politics  197

opposite sides with yet one more reason to think their opponents simply aren’t like them. They’re more like aliens.

Consequences of Worldview Politics Although the consequences are many, I will focus here on one, in particular, and that is increasing partisan hatred. Having made reference to it early in the chapter, let me now provide some details. Since the 1970s, the American National Election Study has been asking Americans how they feel about various people and groups in society. Specifically, they ask people to rate them on a feeling thermometer bounded between 0 and 100. If people absolutely love a group, they are instructed to rate it at 100. If they absolutely hate it, they are supposed to rate it at 0.  If their feelings are neutral, they are told to give it a 50. They can also choose any number in between. The Democratic Party and the Republican Party are among the groups the ANES has asked about. In Figure 8.5, I’ve arrayed the average score that partisans have given their own party and the other party, broken down by presidential administration going back to the Carter years. The results are striking. The first thing that is important to note is that in this period of intense party polarization, partisans do not like their own party any more than they did back in the 1970s. In fact, the Republicans’ rating of the Republican Party in 2016 was slightly lower than in previous years. In other words, polarized politics has done nothing to increase people’s identification with their own party. Rather, its effect has been to diminish their feelings about the other party. Back when Carter and Reagan served as president, partisans didn’t much dislike the other party. The average scores tended to be around 50 degrees, the

Average feeling thermometer scores

80 70 60 Dem-Dem

50

Dem-Rep

40

Rep-Rep

30

Rep-Dem

20 10 0 Carter

Reagan

Bush I

Clinton

Bush II

Obama

Figure 8.5 Party Feeling Thermometers, Carter Through Obama

198  Marc Hetherington

neutral point. That gradually started to change during the 1990s, with both sides becoming a little less positive about the other. But those other party averages have nosedived in the 2000s. Importantly, it was at this juncture that worldview also began to shape party identification profoundly.41 By 2016, the average rating that Republicans gave the Democratic Party was 27 degrees, and the average rating that Democrats gave the Republican Party was 25. Those temperatures sound cold, and they really are in a relative sense. For example, the average score that Republicans generally give “Atheists” is a little above 30 degrees. Put another way, Republicans dislike the Democratic Party much more than they dislike Atheists. Similarly, there is no group that Democrats rate more negatively than the Republican Party. Some preliminary data that I collected in 2018 suggests that the negative feelings have grown even more intense since Trump’s election. Partisan hatred is having pernicious effects on the political system. One is what seems like an increased tendency for partisans to believe what they want to believe rather than what is really true. Anyone who has ever hated someone knows that you are willing to believe just about anything that casts a person you hate in a negative light. Why do so many Republicans continue to believe that Barack Obama was born in Kenya even after he produced his long form birth certificate? Why do so many Democrats think the American economy, with the strongest growth and lowest unemployment in decades, is not performing well? When you really hate your opponent the tendency to engage in partisan motivated reasoning increases. It is critical for partisans to continue to see their side as the good guys and the other side as the bad guys even if the facts might not always support it. Along these same lines, perhaps the most disconcerting feature of American politics today is that partisans of one stripe don’t seem to be rooting for the country to succeed when the other party is in power. Rooting for failure is a natural reaction when rival groups compete. In baseball, for example, Boston Red Sox fans root for the New York Yankees to fail. When it comes to college basketball, University of North Carolina fans hope that Duke loses every game. While that approach to sports is fine, it is not so clear that this is especially appropriate when it comes to politics. Although it would be hard to get people to reveal it in a survey, the perception was certainly widespread that Republicans were not hoping for a spectacular recovery from the Great Recession of 2008–​2009. A spectacular recovery would have been good news for Barack Obama and the Democratic Congress. Similarly, it often feels like Democrats are faintly disappointed when a new economic report comes out showing high growth and low unemployment now that Donald Trump is president. A good economy is good news for Trump. Political failure, however, means very real people suffer as a consequence. For example, fewer will have jobs or health care. Their lives will be harder as a result. Worldview-​divided politics helps create this unhealthy climate. As he faced a country divided by Civil War, Abraham Lincoln said, “A house divided against itself cannot stand.” American might not be on the brink of a

Worldview Politics  199

Civil War, but the logic still holds. When those on opposite sides of the partisan divide are not pulling together on the same rope, the country will be the worse for it. Worldview politics brings out partisans’ worst instincts.

Notes 1 Downs, Anthony. An Economic Theory of Democracy. New  York:  Harper, 1957; Campbell, Angus, Philip Converse, Warren Miller, and Donald Stokes. The American Voter. New York: Wiley, 1960. 2 The survey was done by the Annenberg Public Policy Center. Details can be found at www.cnn.com/​2017/​09/​13/​politics/​poll-​constitution/​index.html. 3 These results come from a 2016 Pew Foundation survey. Details can be found at www.people-​press.org/​2016/​06/​22/​partisanship-​and-​political-​animosity-​in-​2016/​. 4 Hetherington, Marc, and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 5 Prior, Markus. Post-​broadcast Democracy:  How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. Cambridge: Cambridge University Press, 2007. 6 These survey results can be accessed at https://​electionstudies.org/​resources/​anes-​ guide/​. 7 All these results come from Michael McDonald’s work on the Voting Eligible Population, which can be found at www.electproject.org/​home/​voter-​turnout/​ voter-​turnout-​data. 8 Schattschneider, Elmer E.  The Semi-​Sovereign People. New  York:  Holt, Rhinehart and Winston. 1960. 9 Carpini, Michael X.  Delli and Scott Keeter. What Americans Know About Politics and Why it Matters. London: Yale University Press, 1996. 10 Converse, Philip E. “The Nature of Belief Systems in Mass Publics (1964).” Critical Review 18, no. 1–​3 (2006): 1–​74. 11 Hetherington, Marc J.  and Jonathan D.  Weiler. Authoritarianism and Polarization in American Politics. Cambridge: Cambridge University Press, 2009; Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 12 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 13 Rivers, Douglas, “What the Hell Happened? The Perils of Polling in the 2016 US Elections,” presentation, Harvard University, Cambridge, MA, April 24, 2017. 14 In the interest of full disclosure, the characters in this story are my brother and father. 15 Hibbing, John R., Kevin B.  Smith, and John R.  Alford. Predisposed:  Liberals, Conservatives, and the Biology of Political Differences. London: Routledge, 2013. 16 Douglas R.  Oxley, Kevin B.  Smith, John R.  Alford, Matthew V.  Hibbing, Jennifer L. Miller, Mario Scalora, Peter K. Hatemi, and John R. Hibbing “Political Attitudes Vary with Physiological Traits,” Science 19 September 2008: 1667–​1670

200  Marc Hetherington 17 Smith, Kevin et al., “Disgust Sensitivity and Neurophysiology of Left–Right Political Orientations,” PLOS ONE, 6, no. 10:e25552. 18 Kam, Cindy D., and Beth A.  Estes. “Disgust Sensitivity and Public Demand for Protection.” The Journal of Politics 78, no. 2 (2016): 481–​496. 19 Goldhill, Olivia, “The Shape of Your Brain Influences Your Political Opinions,” Quartz, March 28, 2018; Rayota, Kanai et al. “Political Orientations Are Correlated with Brain Structure in Young Adults,” Current Biology, no. 8 (2011): 677–​680. 20 Gerber, Alan S., Gregory A. Huber, David Doherty, Conor M. Dowling, and Shang E. Ha. “Personality and Political Issues,” The American Political Science Review 104, no. 1 (2010): 111–​133. 21 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 22 Feldman, Stanley. “Enforcing Social Conformity:  A Theory of Authoritarianism,” Political Psychology 24, 1(2003): 41–​74; Stenner, Karen. The Authoritarian Dynamic. New York: Cambridge University Press, 2005. 23 Harry Enten provides an interesting treatment of this moment in time in The Guardian, www.theguardian.com/​commentisfree/​2013/​aug/​28/​republicans-​party​of-​civil-​rights. 24 Garretson, Jeremiah. The Path to Gay Rights:  How Activism and Coming Out Changed Public Opinion. New York: NYU Press, 2018. 25 Hetherington, Marc J. and Jonathan D. Weiler. Authoritarianism and Polarization in American Politics. Cambridge: Cambridge University Press, 2009; Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA: Houghton Mifflin, 2018. 26 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 27 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 28 Bishop, Bill, and Robert Cushing. “The Big Sort: Migration, Community, and Politics in the United States of ‘Those People,’ ” in Red, Blue and Purple America: The Future of Election Demographics, edited by Ruy Teixeira. Washington, DC:  Brookings Institution Press, 2008, pp. 50–​75. 29 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 30 Iyengar, Shanto, Gaurav Sood, and Yphtach Lelkes. “Affect, Not Ideology:  A Social Identity Perspective on Polarization,” Public Opinion Quarterly 76, no.  3 (2012): 405–​431. 31 Iyengar, Shanto and Sean J. Westwood. “Fear and Loathing Across Party Lines: New Evidence on Group Polarization,” American Journal of Political Science 59, no.  3 (2015): 690–​707. 32 Putnam, Robert D. Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster, 2000. 33 Wilson, Chris. Time. “Do You Eat Like a Republican or a Democrat?,” July 18, 2016.

Worldview Politics  201 34 Hibbing, John R., Kevin B.  Smith, and John R.  Alford. Predisposed:  Liberals, Conservatives, and the Biology of Political Differences. London: Routledge, 2013. 35 Chris Mooney provides a write-​up here www.salon.com/​2013/​02/​27/​conservatives_​ and_​lilberals_​drink_​different_​beer_​partner/​ 36 Will Rabbe provides a write-​ up here, www.msnbc.com/​hardball/​democrat​or-​republican-​look-​their-​drink. 37 Becca Stanek provides a write-​up of the Facebook data here https://​mic.com/​articles/​102556/​5-​charts-​show-​how-​you-​can-​tell-​democrats-​from-​republicans-​on-​ facebook#.wRz0XB6JZ. 38 The write-​up from the YourMechanic study can be found here, www.yourmechanic. com/​article/​red-​car-​blue-​car-​do-​political-​views-​predict-​car-​preferences. 39 One can view this very curious map here http://​mentalfloss.com/​article/​58161/​ do-​you-​live-​cat-​state-​or-​dog-​state. 40 Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA:  Houghton Mifflin, 2018. 41 Hetherington, Marc J.  and Jonathan D.  Weiler. Authoritarianism and polarization in American politics. Cambridge University Press, 2009; Hetherington, Marc and Jonathan Weiler. Prius Or Pickup? How the Answers to Four Simple Questions Explain America’s Great Divide. Boston, MA: Houghton Mifflin, 2018.

Chapter 9

The Emotional Foundations of Democratic Citizenship Ted Brader and Erin Cikanek

In a democracy, citizens exercise power over government by choosing leaders in elections and, on some occasions, expressing their preferences on matters of policy directly. The quality of democracy therefore depends on not only how well leaders and political institutions respond to the needs and desires of the people, but also how effectively citizens participate in the process of self-​government. A  growing body of evidence testifies to the important role emotions play in shaping the ability and motivation of citizens to take part in politics. Fear, anger, enthusiasm, and other emotions affect public opinion by altering whether and how citizens pay attention, learn new information, think through their decisions, and act on their opinions. This chapter highlights some of what we’ve learned about the emotional foundations of democratic citizenship.

Emotions and the Performance of Democratic Citizenship What does democracy require of citizens? The ideal citizen has been envisioned as embodying some mix of the following virtues: vigilant, active, independent, open-​minded, informed, thoughtful, tolerant, loyal, and courageous.1 A substantial portion of public opinion scholarship over the past 80 years has focused, implicitly or explicitly, on how well citizens live up to these expectations. The conclusions have been decidedly mixed, with enough evidence to sustain the views of both optimists and pessimists. On balance, the tone tips toward the negative, ranging from alarm and disappointment, on one side, to contentment (things are “good enough”) on the other side. We see such mixed results across many aspects of democratic citizenship. Americans pay only scattered attention to politics, though they occasionally become much more engaged for a limited time or on a particular issue. Their knowledge about public affairs is spotty at best, but many seem capable of learning what they need to know under the right conditions. Opinions on political candidates and policy matters typically exhibit an extraordinary—​to some observers, a disturbing—​level of loyalty to one’s social group, nationality,

Emotional Foundations  203

and especially political party. Yet this loyalty is tempered by some measure of responsiveness to changing circumstances, such as when economic conditions worsen or cherished policies are threatened. Where participation is concerned, some Americans are habitually active while others appear withdrawn from politics entirely, but these general tendencies obscure tremendous fluctuations in participation over time and across situations. Whether they conclude that the democratic glass is half full or half empty, researchers have not been content to leave it at that. They have sought to identify features of both individuals and the political environment that help explain the quality of citizen performance. Some of that research examines the motivations that can cause a person to react differently across situations or cause different people to respond differently to the same situation. Emotions turn out to be key motivational forces, capable of (re-​)directing human decision-​making and behavior. However, until recently emotions received scant attention from public opinion scholars. The past two decades have witnessed a surge in research, suggesting that emotions are indeed a potent force guiding the formation of both opinions and decisions to take political action.

What Are Emotions? What we call emotions are, in fact, a complex “syndrome” of reactions to our circumstances that include electrochemical processes in the brain, changes in autonomic and motor systems (e.g., breathing, heart rate, muscle tension, facial expressions), and behavioral impulses. In our everyday lives, we know emotions primarily by our experience of the feeling states that accompany them—​our awareness of what it feels like to be angry, overjoyed, sad, or frightened—​even though emotions occur all of the time without rising to the level of conscious awareness. In this chapter, we discuss the political implications of a number of discrete emotional reactions such as anger, enthusiasm, fear, pride, shame, guilt, and disgust. So, what is the function of emotions? What purpose do they serve? Emotions are motivational impulses. Our brains monitor the world around us for changes that have relevance for our goals and well-​being.2 These changes are perceived and become signals—​appraisals of our circumstances—​that trigger an emotional reaction, which is reflected in a distinct pattern of changes in our thinking and behavior. Scholars believe emotions evolved as a way to allow reasonably efficient, differentiated responses to the sorts of situations humans (and many other animals) tend to encounter repeatedly. Thus, emotions enable us to adapt our thinking and bodies rapidly to meet the needs of particular situations. Consider examples for three emotions. When we make progress toward our goals, we feel enthusiastic and energized to keep doing what we’ve been doing. When something appears that threatens our well-​being, we feel afraid, become more alert and focused on the danger, and shift from reliable habits to active reconsideration of our options. When someone puts obstacles between us and

204  Ted Brader and Erin Cikanek Table 9.1 Some Appraisals and Action Tendencies of Common Emotions Emotion

Appraisals

Action Tendencies

enthusiasm

progress toward goals expectations for reward met/​exceeded threat to well-​being uncertainty about outcomes lack of control

continue pursuing goals less effortful thinking greater confidence escape/​prevent  danger attention on potential danger more effortful thinking avoid risks more open to compromise overcome/​remove obstacle punish offender less effortful thinking take risks less open to compromise withdraw more effortful thinking expel contaminant avoid contact harsh moral judgments expressive displays maintain/​enforce standards hide (the failure) maintain/​enforce standards repair harm done by action

fear/​anxiety

anger

obstacles blocking goals undeserved harm inflicted by others high degree of control

sadness

failure to achieve goals loss of something valued noxious contaminant physical/​moral impurity

disgust pride shame

succeed by own effort/​skill meet/​exceed social standards fail to meet others’ expectations

guilt

recognize own action as wrong

something we want (and feel entitled to), we feel angry and are emboldened to challenge and even punish those who stand in our way. Table 9.1 summarizes key appraisals linked to these and other emotions as well as some of the patterns of thinking and behavior (“action tendencies”) motivated by the emotions. These three emotions, perhaps because they are such common parts of our experiences, have been studied the most, including by scholars of public opinion. The present chapter, therefore, also focuses heavily on these emotions. But other emotions also likely hold relevance for public opinion:  sadness, shame, guilt, pride, and disgust. Sadness, for example, stems from our failure to achieve goals and especially the loss of something valued. It often causes us to withdraw and spend more time reflecting on the details of a situation. Disgust is a reaction to the presence of noxious conditions (e.g., rotting food, bodily excretions) and seems to have evolved to encompass reactions to both physical and moral “impurities.” It generates a strong desire to purge the offending substance and thereby purify one’s environment. Pride and shame are social emotions that arise when we contemplate how others judge us and how well we live up to the standards of our community. Both emotions motivate us to conform better to the expectations of our group, though pride makes us want to put ourselves on display, while shame makes us want to hide ourselves or

Emotional Foundations  205

our actions away. Guilt is also a social emotion that stems from the recognition that our actions are morally wrong (e.g., we hurt others who did not deserve it, or took something that does not belong to us); it differs from shame principally in that our actions, not ourselves, are deemed to be bad and we are motivated to make amends by repairing the harm that we’ve done. With this general understanding of emotions in mind, researchers have begun to study their role in public opinion and democratic citizenship. For starters, the preceding insights offer valuable clues about how members of the public will react to particular events, messages, and political figures. If emotions are tied to personal relevance, then we should expect to see stronger emotional reactions to political topics from those who are attentive to and care about politics. We do. People who know more about politics, those who express greater interest in politics, and those who identify most strongly with political parties are all more likely than their fellow citizens to report strong emotional reactions to political candidates. Emotions not only vary in predictable ways across individuals, but also over time and across diverse situations. For example, not all candidates or elections generate reactions of equal intensity or type. The American National Election Study (ANES) has been asking representative samples of eligible voters about their emotional reactions to presidential candidates since 1980. Figures  9.1a and 9.1b show the average levels of anger, fear, and enthusiasm toward the Democratic and Republican presidential nominees, respectively, in each year. It is immediately apparent that emotions rise and fall from one election to the next. A  closer look reveals a pattern. Emotions are often noticeably stronger toward incumbent presidents (Democrats in 1980, 1996, and 2012, Republicans in 1984, 1992, and 2004). Incumbents hold greater relevance and have accumulated a long record of actions and inactions capable of provoking strong feelings. We thus see emotional peaks in reelection years, especially for retrospective emotions like anger and pride (a component, along with hope, of the enthusiasm measure in ANES). This is born out as well in laboratory experiments conducted by a team of scientists at Dartmouth College in the 1980s. They studied individuals’ reactions to the emotional facial expressions of political leaders and discovered, among myriad other insights, that the strength of citizens’ emotional responses increased with the status and familiarity of the leader (reactions to incumbents > reactions to current candidates > reactions to past candidates/​losers).3 There are two exceptions to the incumbency pattern in the ANES time trends. In 2008, Democratic nominee Barack Obama provoked historically high reports of enthusiasm. This may owe something to the central campaign theme of “hope,” the second component in the enthusiasm measure, or to the pride and hope he engendered by his historical candidacy as the first non-​white presidential nominee of a major party. The even bigger exception comes in the open-​seat contest to succeed President Obama. The 2016 presidential contest, unusual in many respects, generated the strongest emotional reactions overall

0.8

Percent expressing emotion

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1980

1984

1988

1992 Anger

1996

2000

Enthusiasm

2004

2008

2012

2016

Fear

Figure 9.1A Emotional Reactions to Democratic Presidential Nominees, 1980–​2016

0.8

Percent expressing emotion

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1980

1984

1988

1992 Anger

1996

2000

Enthusiasm

2004

2008

2012

2016

Fear

Figure 9.1B Emotional Reactions to Republican Presidential Nominees, 1980–​2016

Emotional Foundations  207

in the entire ANES time series. Voters reported elevated levels of fear, anger, and enthusiasm toward Hillary Clinton, and record-​setting levels of fear and anger toward Donald Trump. Why? It is impossible to say for certain with these data alone. However, both candidates had spent considerable time in the public eye long before their nomination and, as public figures, had developed polarizing personal reputations that transcended typical partisanship. Whatever the root causes, the sense among many during the 2016 presidential election that emotions were running extremely high is confirmed by the historical data. One difficulty in sorting out exactly why citizens feel as they do is that researchers often ask very general questions about emotions toward politicians, institutions (e.g., the government), or issues (e.g., immigration). Politicians and governments do many things, and issues have many distinct facets, any of which may have provoked the reactions individuals report. To learn more about the sources of public emotions, political scientists need to ask more detailed questions and design studies that can isolate features of a politically relevant event, message, person, or organization. Drawing on the insights about appraisals discussed earlier, we are conducting research to explain better which situations give rise to particular emotions.4 “Bad news” or unpleasant circumstances can provoke a range of reactions—​fear, anger, sadness, disgust. But not all bad news is the same. Sometimes we encounter threats to our lives or well-​being; as noted above, these should provoke fear. If somebody is to blame for a harm, an injustice, or an obstacle in our path, we should feel anger. And so on. To study this further, we presented people with a number of scenarios they might read about in the news and asked how they would react. For example, we told them about “road rationing” policy proposals in some metropolitan areas that would restrict the number of days each week that drivers could use local roadways. What reaction(s) should we expect? Such policies are rare at present and propose to curtail broadly used freedoms, therefore reactions should be largely negative. Because most Americans take for granted their ability to drive cars when and where they like, we specifically anticipated a good deal of anger. Cars are not merely a convenience, however. Many people depend on cars to get to work, making the proposed limits a potential threat to their livelihood and source of anxiety. Still others rely on other forms of transportation or are less mobile; they may not be bothered by these proposals. Figure 9.2 shows the results. Although anger was the most common response to road rationing in our survey, reactions overall were a roughly even mix of anger, worry, and no feelings at all. Figure 9.2 also shows three of the other scenarios from this study. A policy proposal to tax Americans based on the extent of their Internet usage (e.g., rates would be especially high for users who stream video content) generated anger among the vast majority of respondents. This is consistent with the fact that the tax would impose a burden or obstacle on (mostly) recreational activities enjoyed by increasing numbers of Americans, but without putting many lives or livelihoods at risk (i.e., by contrast, we might expect a rise in gasoline taxes

208  Ted Brader and Erin Cikanek

Road Rationing

Internet usage tax

Asteroid risks

Women’s history museum

0

25

50

75

100

Percent reporting as dominant emotion Angry

Worried

Proud

None

Figure 9.2 News Stories Elicit Distinct Patterns of Emotional Reactions

to generate more anxiety). In another scenario, we suggested that scientists had discovered that the chances of a large asteroid striking the Earth in the next 20 to 30 years were much higher than previously thought. The most common response was fear. This too is in line with expectations, given that the situation combines uncertainty, potentially catastrophic harm, and a lack of control. Finally, we asked people about a congressional bill that would create a new monument and museum in Washington, DC, celebrating the contribution of women to American society and history. It is not obvious that this plan would pose much of a harm or burden to anyone, instead it invites positive feelings (“celebrating”) tied to the achievements of large numbers of Americans. Of course some people are apt to feel more connected to or invested in this issue than others. Thus, we expected and found predominantly feelings of pride (and admiration) or no feelings at all, but very little fear or anger. Scenarios like these illustrate how we can predict the emotional reactions of the public to policies and information, by focusing on key facets of their

Emotional Foundations  209

relevance for individuals (i.e., on the most likely appraisals). As in all things, our predictions may sometimes be off the mark. When they are, it will usefully signal that we misjudged what the information means to other people.

Vigilance and (Selective) Learning about Politics Emotions help explain why the public pays attention to some issues and events more than others. Because emotions arise from circumstances that are relevant to people, just about any emotion may predict interest in a subject. But we expect fear or anxiety to trigger greater attentiveness beyond what routinely interests a person, because it signals the need to deal with a potential threat to a person’s safety or well-​being.5 Moreover, fear does not just make a person more generally attentive, but directs attention selectively toward the threat and ways of removing it. This heightened attention, in turn, creates conditions favorable for increased learning by citizens.

Political Attention and Engagement Citizens pay more attention to politics during election campaigns, though their attention waxes and wanes from one election to the next and even over the course of a single campaign. Emotions contribute to some of these short-​term fluctuations. Drawing heavily on data from the ANES surveys, Marcus and colleagues have conducted extensive research into the impact of emotions on electoral behavior.6 Even controlling for long-​term levels of engagement with politics, when citizens feel enthusiasm or anxiety (fear) about candidates, they report greater interest in the campaign, care more deeply about the election outcome, and spend more time watching and reading the news.7 A more nuanced picture of the short-​term effects of these emotions emerges from “panel” surveys in which individuals are re-​interviewed at multiple points during the campaign. Such surveys can help researchers isolate changes in emotions, opinions, and behavior. In this case, the researchers found that enthusiasm produced short-​ term boosts in citizens’ reported levels of interest and caring, while anxiety provoked greater attention to almost all sources of campaign news coverage.8 Other studies also support the notion that negative emotions have a positive impact on interest and attention to politics. Anxiety, for example, awakens greater interest in election campaigns, particularly among citizens high in political efficacy—​that is, those who feel competent to participate in politics.9 This mirrors decades of research on the effectiveness of fear appeals in public health and safety campaigns (e.g., targeting seat belt usage, smoking cessation, dental hygiene), showing that fear appeals are more effective when people perceive the recommended actions as likely to work and themselves as capable of carrying out the recommendations.10 In a study conducted during the build-​up to the 2003 Iraq War, Americans who felt anxiety and anger toward terrorists and

210  Ted Brader and Erin Cikanek

Saddam Hussein spent more time thinking about the war, talking about the war, and consuming national news.11 The impact of anxiety on thinking and talking, however, was much larger than the impact of anger. In a departure from the results discussed so far, a study of the 1996 presidential election found that citizens who felt hopeful about the candidates were more likely to watch TV coverage of the summer party conventions and the fall campaign.12 The researchers distinguished hope from enthusiasm, as well as from anxiety and anger, in their analyses. Although self-​reported feelings of hope and enthusiasm often are so highly correlated as to be statistically indistinguishable, appraisal theories of emotion emphasize their conceptual distinctiveness: hope is oriented toward future and uncertain outcomes, something it shares with fear.13 What is the source of these feelings? The emotions driving interest and attention in the preceding studies presumably are responses to events, messages, and other features of the political world, most of which citizens encounter through mass media. Scholars have made use of experiments to learn more about how such mass-​mediated communications trigger emotions and in turn shape public opinion. For example, two companion experiments carried out during a gubernatorial primary election examined how political ads use imagery and music to trigger emotions and thereby influence viewers.14 Ads eliciting enthusiasm stoked viewer interest in the campaign, increasing expressed levels of interest by over 12 percentage points after exposure to a single ad embedded in a news program.15 Ads eliciting fear or anxiety increased the desire to contact campaigns for more information and to watch political news. As Figure 9.3 shows, fear cues caused the percentage of citizens wanting to contact campaigns to rise from 3 percent to 17 percent and the percentage wishing to watch more political news from 41  percent to 60  percent.16 Enthusiasm-​inducing images and music did spark interest in the news, but had no effect on the desire to contact campaigns. Thus, the impact of emotional advertising appeals parallels the effects of feelings toward candidates as uncovered in survey studies: enthusiasm boosts general interest, while fear principally stimulates attention to sources of new information.17 It is well documented that news coverage affects which social problems and policy issues the public sees as most important.18 The roots of this “agenda setting” effect are often thought to lie in the way news stories make some issues more cognitively accessible (i.e., more readily called to mind) than other issues.19 Although beliefs about issue importance are not precisely the same as individual attentiveness, they are about what issues should receive public attention. Recent research suggests that emotions play a key part in the agenda setting effect. Specifically, news stories alter perceptions of issue importance by either arousing or abating negative emotions, depending on whether the issue is portrayed as getting worse or better, respectively.20 The experimental evidence confirms that, to change perceptions of importance, it is not enough for new stories to make the issue more accessible in the minds of the audience, they must also influence

Emotional Foundations  211

emotions. Fear and sadness in particular—​as opposed to anger or any of several positive emotions—​mediate these shifts in public priorities.

Political Information-seeking

Percentage seeking or recalling information

Emotions influence not only citizens’ attention and desire for information, but also their actual information-​seeking behavior. In the advertising study mentioned earlier, viewers exposed to fear-​eliciting campaign ads more closely scrutinized and recalled information from subsequent news stories.21 After watching the news broadcast in which the ad appeared, subjects had an opportunity to list as many topics from the news program as they could recall. Fear cues seemed to improve subjects’ recall specifically of topics that aired after the political ad and that were related to the themes contained in the ad. This suggests that fear cues caused subjects to keep their eyes out for relevant information even after the ad was finished; indeed, as Figure 9.3 indicates, recall of subsequent relevant news stories jumped from 13 percent to 38 percent when fearful images and music were present in the ad. In contrast, enthusiasm cues produced a 15 percentage point drop in the recall of those same news stories. Similarly, in a different experimental study, anxiety aroused by news stories about immigration caused citizens to request additional information from the government and advocacy groups.22

100 90 80 70 60 50 40 30 20 10 0 Desire to watch more political news

Desire to contact Seek more campaigns for information related more information to issues in the ad

Negative ad w/o fear cues

Recall subsequent news stories relevant to the ad

Negative ad w/fear cues

Figure 9.3 Fear Cues in Political Advertising Trigger Greater Attention and Informationseeking for Relevant Information

212  Ted Brader and Erin Cikanek

Researchers have turned recently to examining how people search for information on the Internet. One such study finds that campaign-​related (and experimentally induced) fear consistently makes voters more inclined to pay attention to the candidates and debates, but that enthusiasm and anger also increase those inclinations at least some of the time.23 When actual informationseeking is monitored, however, anger causes voters to spend less time searching candidate websites and less time on each page clicked. In contrast to this, anxiety provoked by threatening election news stories causes voters to seek out a broader range of information by visiting more unique web pages. Fear therefore seems to play a strong role in motivating citizens to actively seek out more information. This inquisitiveness appears to be broad and yet still selective or targeted in particular ways. Scholars have long argued that people have a tendency to engage in selective exposure to information that confirms their existing point of view and, while this tendency does not always prevail, in politics partisans often exhibit this sort of biased information search.24 Yet studies suggest that anxiety can disrupt this tendency while anger may reinforce it. Researchers compared the online behavior of individuals who got angry or anxious in response to threatening news stories about campus affirmative action policies.25 Anxious citizens sought out more information that was challenging to their own views, but those who were angry avoided web pages that challenged their position. Another study found that anxiety made citizens more likely to visit an opposing candidate’s website and indeed to visit the sites of both candidates in an election.26 Anxiety generated such a balanced search, however, only when people were aware the information might be useful in the future (e.g., in defending their views to others). While anxiety tends to broaden the focus of citizens’ attention in terms of agreement with a person’s political views, it also narrows the focus of their attention to information relevant to the potential threat. In experimental studies, fear ads prompted viewers to request more information about the campaign and about the threatening issues raised in the ad.27 After viewing both news and ads, subjects were invited to list issues they would like to hear more about from journalists and politicians. When subjects saw a relatively unemotional ad, nearly 36 percent of them listed issues that had been raised in the political ads. This number jumped to over 52 percent among those who saw the identical ad messages except with fearful imagery and music (see Figure 9.4). Enthusiasm-​eliciting cues however had no discernable impact on these topical “wish lists.” Similarly, fear appeals increased the desire for political news, as opposed to other news content, and focused viewers’ attention on subsequent news stories that were relevant to the issues raised in the ad, but not on irrelevant news stories.28 Other studies confirm this narrowing effect of fear and its potential for creating a vicious cycle in which fear drives citizens to consume information that further stokes those fears, rather than toward information that might reassure them. For example, fears about increasing immigration caused citizens to

Emotional Foundations  213

request information from government and anti-​immigrant organizations, more so than from pro-​immigrant or academic sources.29 In another major study, anxious citizens were more likely to read online news stories about immigration, especially negative stories, than about other topics.30 They also better remembered threatening news stories and agreed more with the arguments in those stories. This research highlights a danger posed by anxiety—​or what we have to fear from fear itself, namely that citizens may enter a vicious cycle in which anxiety is triggered, spurring an over-​focus on threatening information, which in turns fuels further anxiety.31 As a result, anxious citizens are prone to overestimating the dangers they confront and to taking too little notice of reassuring signals.

Political Learning Does all of this increased interest, attention, and information-seeking affect what citizens learn about candidates and issues? Few studies to date examine the ultimate impact of emotions on political learning. Evidence suggests that fear can precipitate increases in citizen knowledge during elections. Citizens who are particularly anxious possess more information about the candidates and know better their relative policy positions.32 In addition, over the course of a campaign, anxious citizens’ knowledge of candidate positions improves much faster than that of either enthusiastic or unemotional citizens.33 Emotional appeals in political advertising do not universally enhance candidate name recognition, but when fear appeals target a viewer’s preferred candidate, that viewer has an easier time recalling the names of both candidates.34 Similarly, when threatening campaign information induces anxiety, as opposed to other emotions (i.e., anger, enthusiasm, disgust), citizens not only read a broader array of candidate web pages, they also display better knowledge of these stories on a follow-​up quiz.35 Some studies, however, arrive at more mixed conclusions about the ­relationship between anxiety and learning. One such study tracks subjects through a simulated, interactive campaign on computers. Anxious voters better learned their preferred candidate’s policy views under highly threatening conditions (when subjects encountered lots of unexpected information), but this learning did not extend to learning other candidates’ positions.36 Other scholars have warned that anxiety and stress can impair learning even as those emotions cause people to devote more attention to what worries them.37 Consistent with this, in surveys conducted between 2001 and 2003, researchers found that anxiety about terrorism, Saddam Hussein, and a possible war in Iraq led Americans to spend more time thinking about Iraq, yet worsened their knowledge of basic facts about the country.38 Such discrepancies suggest we should be cautious about assuming that fear-​driven increases in attention and information-seeking will lead automatically to gains in relevant public knowledge.

214  Ted Brader and Erin Cikanek

Emotions and Information in Contemporary Politics Recent work on emotions and learning in contemporary politics may further support those who feel alarm and disappointment regarding the capacity of citizens to participate in democracy. Emotions that spur greater action, such as anger, may also adversely impact learning and the types of information that people either search out or provide. Contemporary work on anger and anxiety shows that issue attitudes and ideological biases may persist when people are searching for more information on a subject. Anger appears to influence the relationship between issue attitudes and ideological biases in such a way that when people search for new information, they are not concerned with the accuracy of the information.39 Rather, people are looking for more evidence that bolsters their point of view so that they can further argue against information with which they disagree. In this way, anger appears to further bias information searches toward a person’s predispositions and beliefs. If anxiety causes people to search for more information, one might hope that the search is motivated by a desire for a more accurate understanding of the situation. However, it may be the case that anxiety simply alters searching behavior but not other behaviors that may be important to incorporate into a search. Research respondents were presented with ads about a pending nuclear attack, with the ad either containing vivid imagery—​moving, color images—​or non-​vivid black and white images that lack movement.40 Calfano and Kruse find negative ads with colorful, moving images, mediated by anxiety, cause people to consume more information (e.g., news stories and congressional testimony) about the impending threat. Yet the same ads did not cause respondents to become more interested in checking on the accuracy of the information, even when that option was readily provided to them. Researchers have also investigated the role of disgust in information-seeking behavior. Levels of disgust and anxiety were manipulated by telling subjects about the symptoms of a fake illness and its likelihood of spreading.41 Subjects exposed to the more “disgusting” version of the illness were more likely to remember the disgust-​inducing symptoms. However, these disgusted individuals were no more or even less likely to recall other potentially useful information about the illness—​additional symptoms, transmission methods, availability of a cure. In keeping with earlier research, anxiety caused subjects to request more information and to report a stronger desire to look up information and discuss the illness with friends. These search-​enhancing effects of anxiety were short-​circuited, however, under conditions of heightened disgust. Whereas anxiety prompted engagement, disgust triggered avoidance of important details about the illness. Many see deliberation as a central part of democracy, emphasizing the ability and value of citizens in working together to tackle societal problems. Deliberation ideally involves the open and equal sharing of information.

Emotional Foundations  215

However recent work suggests that emotions are critical for whether social inequality is reproduced or overcome in deliberative settings.42 Based on 150 interviews of people who had attended recently a conference organized by communities, political parties, companies, or similar organizations, participants were asked in-​depth about their experiences and the emotions that they felt. While many people had similar emotions, when circumstances were characterized by disappointment and shame, those with lower status had to do more “emotional work”—​attempting to manage (i.e., change or overcome) their emotions in order to participate. They worried more about their reactions in these situations, thus stifling their own participation and making the deliberation less democratic by holding back information in order to deal with their negative emotions. This did not seem to be a strategy for those with higher social status, nor did they perform as much work around the emotions they were feeling. In this way, certain negative emotions may influence the decision to provide others with information, not just the motivation to search for information.

Summary: Emotions, Attention, and Learning Emotions influence the political engagement and attentiveness of citizens. Fear, in particular, powerfully shapes whether and when citizens adopt a more vigilant posture. Fear redirects the focus of citizens’ attention as well as their beliefs about which issues merit public attention more generally. Fear motivates citizens to seek out new information from a broader, more balanced array of political viewpoints, yet focus more narrowly on potential threats. This targeted vigilance can be beneficial, helping people to learn information that is useful in reassessing, reducing, or escaping a potential threat. But it can also produce distortions in public attention and knowledge, to the extent citizens focus too much on threatening issues and neglect non-​threatening information that might ease their anxieties. Similarly, fear-​induced vigilance will not always give rise to better-​informed citizens, especially if citizens are routinely exposed to information that is unhelpful, error-​filled, or misleading. Other emotions—​anger, disgust, shame—​can at times boost attention but often in ways that are highly selective and that lessen the likelihood citizens will become better informed.

Rigidity versus Responsiveness in Opinion Formation In addition to influencing attention and learning, emotions affect the way people make decisions and form opinions. Moreover, they do so in at least two ways. They can directly influence citizens’ evaluations, with positive feelings leading to more positive judgments and negative feelings to negative judgments. Emotions can also influence opinions in a second, more indirect manner, by changing the way that citizens arrive at their views. Because these

216  Ted Brader and Erin Cikanek

indirect effects are less obvious and speak to deeper questions of how people make up their minds, we devote special attention to them in this chapter.

Direct Influence on Opinions Let us begin by briefly considering the direct impact of emotions on public opinion. That emotions exert such influence is not surprising since opinions usually involve “affect”—​that is, liking or disliking. Affect of this sort is an emotional phenomenon much the same as the discrete emotions (fear, enthusiasm, etc.) we’ve been discussing. If a person makes you feel angry, afraid, sad, or disgusted, you will be inclined not to like that person as much. If a group or a policy makes you feel proud, hopeful, and enthusiastic, then you will be more apt to like it. Many studies, including some of the earliest research on emotions and public opinion, indeed find that emotions have a strong impact on evaluations of government and political leaders, over and above any influence of cognitive beliefs or judgments.43 For example, feelings of enthusiasm about a presidential candidate or even about his policies feed directly into greater support and a higher likelihood of voting for the candidate.44 The same pattern emerges for negative feelings except in the opposite direction; reactions such as fear, anger, disgust, drive voters away from candidates. New research argues that feelings of contempt—​arising from an appraisal that someone is untrustworthy and motivating rejection—​damage support for candidates in a way that is particularly difficult to dispel or counter, compared to anger or other negative feelings.45 Citizens also draw, consciously or not, on their emotions about the country when expressing opinions: people who experience positive feelings about the United States express more trust in other people, more trust in the government, and more support for policies like the North American Free Trade Agreement.46 Even when unrelated events put people in a good or bad mood (e.g., the smell of freshly baked cookies, a rainy day), these feelings can cause more positive or negative evaluations of candidates, as long as the individual does come to recognize the true source of her feelings.47

The Process of Opinion Formation and Decision-​m aking Emotions also influence public opinion indirectly, altering the process by which citizens make decisions. People don’t invest equal effort in making every decision. Sometimes they think through an issue carefully, taking the time and effort to sift through the available information and weigh arguments for or against each option. Other times they make a quick judgment based on simple cues or decision rules, such as a voter who decides to mark the ballot for any candidates who are women or for all Republicans. The first mode of decision-​making is called “systematic,” “central,” or “effortful” processing; the second is called “heuristic” or “peripheral” processing.48 There is mounting

Emotional Foundations  217

evidence that a person’s emotional state affects whether she engages in more or less effortful thinking and therefore what sorts of considerations play a role in decision-​making. Fear and sadness both appear to trigger more effortful thinking, while enthusiasm and anger encourage more peripheral or habitual thought processes. Fear breaks citizens out of their habitual modes of thinking and gets them to reconsider their choices in light of the situation. As a result, it opens the door to persuasion. Fear does not guarantee a change of mind, but does prompt “second thoughts” about the decision. A  number of studies find evidence consistent with these expectations. In elections over the past 30 years, voters anxious about presidential candidates have been less likely to rely on partisan loyalties or ideological affinities, but more apt to make decisions on the basis of assessments of candidates’ issue positions and leadership qualities.49 Similarly, fear elicited by campaign ads causes voters to place greater weight on the advertising message; fearful ads are thus more persuasive.50 Voters fearful about the threat of terrorism cast votes less according to partisanship and more according to evaluations of candidates’ leadership qualities.51 Such findings are not confined to electoral settings and extend to multiple domains of public opinion. For example, Americans anxious about the 1991 Gulf War were more likely to change their political judgments in light of their assessment of how well the war had gone.52 Thus, to the extent they felt anxiety, Americans were more likely to set aside preconceptions and update their opinions about whether the U.S.  did “the right thing” by going to war and, further, to update their support for President George H.W. Bush more generally, based on the outcome of the war. Under more controlled experimental conditions, researchers found that as Republicans grew more anxious they shed their party loyalties to express greater approval of the Democrat, President Bill Clinton, even though anxiety was induced subliminally with apolitical images (e.g., snakes, skulls).53 Anxiety also influences political tolerance judgments:  anxious citizens were found to be more responsive to persuasive pro-​or anti-​free speech arguments they had recently read.54 These examples illustrate anxiety’s role in loosening the hold of predispositions, while facilitating the incorporation of new information and changes of opinion. These sorts of indirect effects on opinion formation, however, are not confined to anxiety. Enthusiasm elicited by campaign ads, in contrast to fear, causes voters to place extra emphasis on their initial candidate preferences and to express greater certainty about their choice.55 Figure 9.4 illustrates the contrasting ways these two emotions affect the role of political predispositions in voter decision-​making.56 In experimental research on emotional advertising appeals, many subjects expressed a clear fondness for one candidate over the other when they began the study.57 When asked near the end of the study which way they planned to vote, their original loyalties or “predispositions” predicted quite well their final voting decision. However, the strength of this relationship between predispositions and voting choice depends dramatically on which

218  Ted Brader and Erin Cikanek

100

Percentage voting for the sponsor

90 80 70

Enthusiasm cues

60 50 40 Fear cues

30 20 10 0 Predisposition to support opponent

No predisposition

Predisposition to support sponsor

Figure 9.4 Fear Cues in Political Advertising Weaken the Impact of Predispositions on Voting Decisions, while Enthusiasm Cues Strengthen the Impact of Predispositions.

emotions, if any, are elicited by the political ad they viewed. The correspondence between predispositions and voting decision is much stronger following exposure to enthusiasm appeals, and notably weaker following exposure to fear appeals. We encounter a similar pattern in a study of how the framing of policy options affects preferences for risky policies. It is well established that people are more willing to undertake risks when information is framed in terms of potential losses, while they prefer less risky options when information is framed in terms of gains.58 But responsiveness to these frames turns out to depend on an individual’s emotional state.59 Whereas anxiety increases responsiveness to the framing of options, both enthusiasm and anger decrease responsiveness. Another study contrasts the effects of anger and sadness on opinion formation.60 Sadness led citizens to engage in more effortful thinking when asked whether people should receive public welfare assistance, while angry citizens did not engage in more effortful thinking. This resulted in divergent preferences, with sad citizens favoring more public assistance and angry citizens favoring less.

Summary: Emotions on Opinion Formation Emotions exert considerable influence over the opinions of citizens. In some cases, this impact is simple and direct. At times, people who are in a bad mood

Emotional Foundations  219

or experiencing negative emotions such as anger, fear, or sadness, express more negative opinions and people experiencing positive emotions express more positive opinions. This can occur regardless of whether their emotional state is related to the substance of the opinion. But the impact of emotions goes deeper, to the very process of opinion formation. Some emotions, like fear and sadness, cause citizens to reconsider their political predispositions, expend more effort thinking through the decision, and display greater responsiveness to new information and the details of the situation. Other emotions, like enthusiasm and anger, reinforce the automatic tendency of citizens to spend little time reflecting on their opinions and to stick more closely to their political habits and loyalties. Does this mean fear and sadness improve democratic citizenship, and anger and enthusiasm diminish it? Perhaps. Certainly many would consider it a good thing if public opinion were based on a more thoughtful review of the available information and arguments. But fear and sadness do not guarantee well-​reasoned decisions or good outcomes. As we have seen, detachment from political habits and a focus on details can render citizens more susceptible to persuasive advertising and framing effects, thus making political manipulation easier. Emotions clearly affect how people perform their functions as democratic citizens, but the normative desirability of the outcomes depends on more than the emotion or the thought process it provokes.

Turning Opinions into Political Action Emotions motivate action and push it in particular directions. This doesn’t mean people reflexively act on the emotional impulse, merely that action is more likely because the impulse exists. Emotions often elevate levels of physiological arousal and thus physically prepare people to take actions, though there are also emotional states that reduce arousal and the inclination to act (e.g., sadness, serenity). Emotions also motivate people to act in ways that are specific to a particular emotion, what psychologists call the “action tendencies” of the emotion in question.61 For example, anger creates an impulse to confront and fight, fear an impulse to escape, and disgust an impulse to avoid and purge (see Table 9.1). In considering the political consequences of these action tendencies, it is useful to remember that democratic politics is primarily a realm of collective action. Citizens do not often take direct action to deal with political issues. They instead express preferences over (a) policy actions to be taken on their behalf by the government, (b) the selection of leaders entrusted to take these actions, and (c)  policy actions determined directly from a collective voting process (e.g., ballot initiatives). Most forms of democratic political action thus occur through fairly institutionalized channels:  voting, working on a campaign, donating money, contacting government officials. As a result, we expect

220  Ted Brader and Erin Cikanek

emotions to affect overall levels of political participation, spurring citizens to greater political involvement or prompting them to withdraw from politics, commensurate with their level of physical arousal. However, direct personal action to satisfy more specific emotional impulses (to flee, fight, hide, help, etc.) is likely to be limited in the political realm by the lack of opportunities for meaningful actions of this sort. Nonetheless, these impulses may affect citizens’ preferences for collective policy actions; the impulse to fight, for example, may lead to greater support for aggressive policies. We therefore expect emotions to affect public opinion in distinct ways that reflect the action tendencies of the emotion in question.

Political Participation Not surprisingly, “high arousal” emotions tend to be associated with more political action. This general relationship between emotions and participation has illuminated both persistent differences among individuals and short-​term reactions to specific situations. Recall that people respond emotionally when circumstances hold relevance for them. Put differently, people respond emotionally when they care about what is happening. Therefore, some scholars look to see if individuals report ever having emotional reactions—​positive or negative—​to a particular issue, as a technique for measuring those individuals’ level of conviction.62 Citizens who report stronger feelings about policies are more likely to take political action, above and beyond any intellectual investment reflected by their knowledge of the issue. Other scholars are concerned with the participatory impact of short-​term emotional reactions that from time to time can flare up and fade away for anyone, regardless of a person’s long-​term interest in policies or politics. Both citizens who feel enthusiastic about presidential candidates and those who feel anxious are more likely to participate in election campaigns beyond simply voting.63 Similarly, campaign ads can increase the desire to volunteer and vote during elections by eliciting either enthusiasm or fear.64 Such effects occur outside of elections as well: for example, anxiety triggered by news stories on immigration provokes more Americans to contact their members of Congress to advocate for reducing the number of immigrants to the U.S.65 One recent study compares the impact of three emotions—​fear, anger, and enthusiasm—​on political participation and notes important distinctions among them.66 Fear has the most variable effects; in some cases it motivates voters, but other times has no effect or even suppresses their participation levels. Anger, in contrast, has an especially powerful impact on political mobilization, but its impact depends more heavily on the person possessing the sorts of resources (e.g., education, income, social ties, experience) that enable participation.67 Absent such resources, citizens are less likely to act on their anger in politically consequential ways. Data from the 2008 presidential election in the U.S. provide an example of the direct and distinctive impact of emotions on political participation. In a

Emotional Foundations  221

national survey (conducted with Nicholas Valentino), we asked Americans how they felt about the way things were going in the country. Given the unpopularity of the incumbent president and the onset of a global financial crisis, most Americans felt a good deal more fear and anger than enthusiasm in the fall of 2008. We also asked during the campaign how likely they were to contact a government official and, right after the election, how often they had engaged in a range of different political activities. Table 9.2 shows how enthusiasm, fear, and anger about the way things are going in the country affected the propensity to participate in these various ways, controlling for many other attributes known to predict participation.68 Anger had the clearest, most consistent, and most powerful impact on political participation in 2008, exerting a strong positive effect for eight of the ten activities. In most cases, people who were extremely angry scored somewhere between 10 to 25 percentage points higher on the frequency-​of-​participation scale. The impact of enthusiasm was considerably spottier: it was positively related to several forms of participation, though the evidence was not always reliable enough to give us full confidence that these were genuine effects (in statistical parlance, they did not quite reach accepted levels of “statistical significance”). The results for fear, in contrast, point rather consistently in a negative direction, though these too fall shy of significance in all but one case. Table 9.2 The Impact of Emotions on Political Participation in the 2008 Election Political Participation

Vote Contact Government Officials Discuss Politics w/​Friends & Family Argue about Politics Take Part in a Protest Attend Political Events for a Candidate Display a Campaign Sign Volunteer for a Campaign Donate Money to a Candidate or Party Sign a Petition

Emotional Reactions to the Way Things Are Going in the Country Enthusiasm

Fear

Anger

−​ 4% + 2% + 11% −​ 2% 0% −​ 7% + 11% −​ 6% + 8% + 10%

−​ 6% −​ 23%** 0% −​ 4% −​ 5% −​ 3% −​ 8% −​ 11% −​ 5% −​ 10%

−​ 2% + 24%** + 17%* + 22%* + 7%* 0% + 18%* + 12%* + 14%* + 16%*

Note: The table shows the estimated effect of each emotion on various forms of political participation, expressed as a percentage-point difference in the overall participation scale between those experiencing no emotion and those experiencing intense emotion (i.e., “not at all angry” vs. “extremely angry”). Participation questions asked respondents, whether they voted as well as how likely they were to contact government officials or how often they had engaged in any of the other activities on a four-​, five-​, or eight-​point scale. Analyses are based on ordinary least squares regression and included statistical controls for gender, age, education, income, race, church attendance, residency, strength of partisanship, internal political efficacy, general interest in politics, and political knowledge (N = 460 to 568). Data are from a representative national survey carried out in the U.S. during the 2008 presidential election. Asterisks indicate levels of statistical significance: ** p < 0.01, * p