Women and Development in Africa: How Gender Works 9781685850722

Citation preview

SECOND EDITION

Women and Development in Africa How Gender Works Michael Kevane

b o u l d e r l o n d o n

Published in the United States of America in 2014 by Lynne Rienner Publishers, Inc. 1800 30th Street, Boulder, Colorado 80301 www.rienner.com

and in the United Kingdom by Lynne Rienner Publishers, Inc. 3 Henrietta Street, Covent Garden, London WC2E 8LU

© 2014 by Lynne Rienner Publishers, Inc. All rights reserved Library of Congress Cataloging-in-Publication Data Kevane, Michael. Women and development in Africa : how gender works / Michael Kevane. — Second edition. p. cm. Includes bibliographical references and index. ISBN 978-1-58826-979-9 (hardcover : alk. paper) ISBN 978-1-58826-980-5 (pbk. : alk. paper) 1. Women in development—Africa. 2. Sex role—Africa. I. Title. HQ1240.5.A35K48 2014 305.42096—dc23 2014016145

British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library. Printed and bound in the United States of America

The paper used in this publication meets the requirements of the American National Standard for Permanence of Paper for Printed Library Materials Z39.48-1992.

5  4  3  2  1

Contents

List of Tables and Figures Preface

vii ix

1 Exploring Gender and Development

1

2 Women and Development in Africa

31

4 The Land Tenure Rights of African Women

87

3 Models and Methods

5 Constraints and Opportunities in the World of Work

61 115

6 The Marriage Market

143

8 Preferring Sons over Daughters

195

7 Bargaining Power at Home

9 The Gendering of Education

10 Getting Capital to Women

11 Emerging Movement For—and Against—Gender Equality References Index About the Book

171 217 239 265

285 315 329

v

Tables and Figures

Tables 1.1 Growth Rates and Levels of Income per Capita for African Countries 1.2 Various Indicators of Status of Women in Sub-Saharan Africa 2.1 Results of Regressions Explaining Growth, 1960–2009 3.1 The Difference-in-Difference Approach to Estimating the Effect of a Program 3.2 The Difference-in-Difference Approach: Hypothetical Numbers for a Teacher Absenteeism Experiment 3.3 Trust and Trustworthiness in Experimental Games in Africa 5.1 Average Time Allocation of Women in Béréba, Burkina Faso 5.2 Results of Estimations with Percentage Time Allocated to Activity as Dependent Variable, Burkina Faso 6.1 Median Age at First Marriage for Persons Aged 25–49, by Educational Attainment 6.2 Different Types of Marriage Sequences, Côte d’Ivoire 6.3 Percentage of Marriages That Are Monogamous or Polygynous 7.1 Who Has Final Say over Household Decisions? 7.2 Mean Yield, Area, and Labor Inputs per Plot by Gender of Cultivator, Burkina Faso 7.3 Estimates of the Determinants of Plot Output, Burkina Faso 7.4 Effects of Experimental Shocks on Expenditures, Transfers, and Savings 8.1 Sex Ratios (Female to Male) for Various African Countries, China, and India 8.2 Index of Female Mortality Disadvantage in Childhood 8.3 Estimating a Difference-in-Difference Model of Effects of Change in Price of Tea on Sex Ratios in China vii

9 21 50

71

72 80 131

134

146 148 163 173

184 186

192

197 199

207

viii

Tables and Figures

8.4 How the Number of Sisters Affects Nutritional Status, Ghana 8.5 Mean Budget Shares per Household and Outlay Equivalence Ratios for Adult Goods, Uganda 9.1 Ratio of Female-to-Male School Attendance of Young People in Households Included in the DHS Surveys, Various Years 9.2 Percentage of All Youth Aged 15–24 Who Are Literate, and Female-to-Male Ratio 11.1 Political Participation Gaps from Respondents to Afrobarometer Survey, 2008–2009 11.2 Women in Political Power in Africa Figures 1.1 Growth of Real GDP per Capita, 1960–2011 1.2 Indicators of Well-Being, 1975–2011 1.3 Indicators of Women’s Rights, 1980–2010 1.4 Commitment to Human Rights Treaties, 1960–2010 5.1 Fight or Cooperate? Prisoner’s Dilemma 5.2 Fight or Concede? War of Attrition 6.1 Supply and Demand in Marriage Market 9.1 Ratio of Female-to-Male School Enrollment for Various World Regions 9.2 Gendered Returns to Schooling 9.3 Peer Enrollment Affects Costs and Benefits of Schooling

210

212

220

222

271 280 10 13 24 28 139 141 157

221 224 228

Preface

he randomista movement in development economics, where randomized control trials (RCTs) are used to evaluate the impacts of policies and programs, has swept over much of economics in the years since the first edition of this book was published in 2004. Thus, this new edition integrates much research that uses randomized control trials to learn about how gender affects economic decisionmaking in African societies. In some chapters, such as those on intrahousehold bargaining and the gendered treatment of girls and boys, there are germane findings from RCTs. In others, such as the chapter on land tenure, randomized control trials have fairly little to say. But the field has been growing rapidly, and there are a number of large-scale trials being run that should yield considerable insight. This edition also introduces the method of instrumental variables (see the appendix to Chapter 2). Several important publications pertinent to gender issues have used the instrumental variables econometric strategy; thinking about causality has properly become central to applied economics. It is almost incredible, in retrospect, to think that my undergraduate economics education and even much of my graduate economics program never really dealt with problems of causality in understanding the social world. It was presumed that variables that “could be causal” probably were so, and issues of endogeneity and self-selection were pushed off as advanced topics for specialized courses. Even now one reads journal articles from the period before the mid-1990s and wonders why causality was not taken more seriously. The new edition features an updated introduction to the overall broad issue of gender and development, and the new final chapter incorporates current work by political scientists. Most of the other chapters have been extensively revised, and I have endeavored to include recent findings, but

T

ix

x

Preface

not at the expense of short-changing “classics” and other important work in the field. * * *

After the first edition of the book was published, I took a detour off the straight highway that is an academic career and started a small nonprofit called Friends of African Village Libraries (FAVL). Together with other dedicated colleagues and volunteers, and with some wonderful librarians in Burkina Faso, Ghana, Tanzania, and Uganda, I have tried to help share the joy of reading fiction and discovering knowledge with kids and adults in African villages. Since we started more than a decade ago, the libraries supported by FAVL have received hundreds of thousands of visits, and tens of thousands of books have been checked out. My involvement in promoting libraries and reading prompted me to start conducting research on these topics. As a result, I had opportunities to interview and observe readers and analyze survey questionnaires from villages in Burkina Faso and Ghana. The readers I spoke with have inspired me with their hope and ambition. These are girls and boys who live tough, frustrating, and often tragic lives. In addition to talking with young readers, I have read much of the young adult and children’s fiction produced in West Africa. Authors such as Fatou Keita, Marguerite Abouet, Ansomwin Ignace Hien, Nii Ayikwei Parkes, and Kwei Quartey, to name just a few, encourage this hope and ambition. This new edition is dedicated to all those confident young readers and ambitious authors. I hope that the gender-equal worlds that authors create in their fiction replace the gender-unequal reality that is all too discouraging. And I hope that young readers are the ones who will make that change happen! —M. K.

Women and Development in Africa

1 Exploring Gender and Development

This business of womanhood is a heavy burden. . . . How could it not be? Aren’t we the ones who bear children? When it is like that you can’t just decide today I want to do this, tomorrow I want to do that, the next day I want to be educated! When there are sacrifices to be made, you are the one who has to make them. And these things are not easy; you have to start learning them early, from a very early age. The earlier the better so that it is easy later on. Easy! As if it is ever easy. And these days it is worse, with the poverty of blackness on one side and the weight of womanhood on the other. Aiwa! —Tsitsi Dangarembga, Nervous Conditions

I always knew I was a businesswoman. When I was in high school, I used to sell illegal sweets. And I made money. I made a lot of money . . . I am very happy and proud. When I was young, they said: A woman is a woman—a man should take care of you. But women are actually contributing a lot more than men. We always find ourselves multitasking. —Lovin Kobusingya, Uganda, 2012

lizèta Ouédraogo is widely considered to be the richest woman in Burkina Faso, a small, landlocked country of 15 million people in West Africa. Burkina Faso is an oasis of slow and steady economic growth in war-torn West Africa, though the country is one of the poorest in the world. People eke out a living of less than $500 per year growing crops, such as sorghum, millet, and cotton, and raising livestock. Burkina Faso is in the Sahel zone of Africa, just south of the Sahara Desert, which also favors transhumant livestock production of sheep, cows, and camels. Herders move from place to place with their herds, earning their livelihoods by selling milk, and by occasionally selling livestock for meat and hides. For al-

A

1

2

Women and Development in Africa

most 25 years, Alizèta Ouédraogo had a monopoly on tanning hides in Burkina Faso. Her company, Tan-Aliz, grew steadily, eventually selling $10 million per year in hides to Italian luxury goods producers such as Gucci. Ouédraogo diversified into real estate development and infrastructure construction. That her daughter married the younger brother of President Blaise Compaoré, who has ruled from 1987 to the present, may have helped her obtain government procurement contracts. In 2011, she was elected president of the Chamber of Commerce of Burkina Faso. She succeeded Oumarou Kanazoé, Burkina Faso’s richest man, who had passed away the previous year. Alizèta Ouédraogo became the face of business for Burkina Faso. The success of Alizèta, as she is known in Burkina, has not been free of controversy. Tanning hides is a notoriously toxic industry, with lax oversight of disposal of chemicals used to treat the hides and other toxic wastes. Alizèta’s tanning factory, on the outskirts of Ouagadougou, has been the object of frequent protests. Workers at the factory have periodically gone on strike, claiming that the factory does not follow worker health, safety, compensation, and employment practice regulations. The press has hinted that her government-granted monopoly on tanning hides was not motivated by sound economic policy to temporarily protect an infant industry from regional competition, but rather so that her family (and hence the president’s) could benefit economically. Her construction company, according to rumors, benefited from a hasty and secretive privatization of a state-owned company. For some, Alizèta Ouédraogo’s rise, and the growth of her companies, encapsulate many of the controversies of African economic development: crony capitalism, corruption, environmental degradation, and disregard for workers. For others, her story is the shining example of the liberal way: self-made industrial powerhouse, capitalist with a human face (she is a well-known philanthropist and sponsors numerous charities), responsible citizen (being president of the Chamber of Commerce means she spends much time away from her business and represents one important sector of civil society). For still others, she represents a gender victory: a woman who succeeds and operates on equal terms with men. Whatever the evaluation of Alizèta Ouédraogo, the issues raised are central to development. First, development is about investment. The actors of development are the individuals who foster investment. Young people want to grow up in a society with expanding opportunities for realizing their ambitions and utilizing their talents. Opportunities expand when individuals, as owners of companies and directors of organizations, undertake to forego consumption today and make investments for the future. Societies differ in terms of how their social institutions encourage investment. In many African countries, illegitimate and unrepresentative government,

Exploring Gender and Development

3

powerful oligopolies, and mafia-like private militias and gangs thwart the best efforts of farmers, artisans, traders and small businesses, investors, and entrepreneurs. People and institutions use the power of the state, or sometimes private violence, for their own advantage. These “grabbing hands” are especially difficult to control in poor developing countries. A hard part of development studies is determining whether people like Alizèta Ouédraogo, and the social institutions she represents, are more like entrepreneurs or more like gangsters (Fisman and Miguel, 2009). Second, the development process balances benefits to current generations with costs to future generations. Most development is toxic to humans, to plants, and to animals. A parking lot enables humans to travel conveniently using automobiles, but nothing grows under asphalt, water rushes off and carries away leaking oil, and the heat of the sun is added to the atmosphere, thus warming our planet. A factory that tans sheep, goat, and cow hides is a parking lot on steroids. It is proper for public policy to regulate these negative externalities. Third, development may depend on using state policy to resolve certain well-understood problems with market economies, notably the tendencies of private investors to hesitate when many of the benefits of their investments spill over to others. That is, investment should be subsidized and guided by social institutions when there are significant positive complementarities. Granting a temporary monopoly of exporting hides to Tan-Aliz may not be a sign of corruption, but rather an indicator of good economic policy. Finally, and of particular importance for this book, development is a process that erodes social institutions of exclusion. The shackles of status are replaced by a new conception of persons as equal citizens under the law. Development is about expanding freedom and ending discriminatory institutions that restrict the freedom of some to realize their ambitions and talents (Nussbaum, 2001). Alizèta Ouédraogo, as a successful female entrepreneur in a society where women’s roles remain sharply defined and circumscribed, has to be considered in this light. Organization of the Book This book gives the reader the tools to understand the context and relevance of women in African economic development. Not just wealthy and politically connected entrepreneurs like Alizèta Ouédraogo, but also poor women who eke out a living hawking bananas on street corners. This introductory chapter defines some key terms—development, gender, gender equality— that are used throughout the book. The focus of the book will be on Africa south of the Sahara Desert, which means that the Mediterranean and Middle Eastern–oriented countries of Egypt, Libya, Algeria, Tunisia, and Morocco

4

Women and Development in Africa

are excluded. Henceforth, sub-Saharan Africa shall be shortened to just plain Africa, in accordance with common usage in the social sciences. Africa remains home to the poorest countries of the world, and social institutions that affect economic and development outcomes are profoundly gendered. The primary aim of this book is to trace the links among development, economic activity, growth, and gendered institutions. The approach adopted throughout emphasizes the feedback loops between the choices that individuals make and the social environment that structures those choices. Men and women make differing choices. These choices in turn constitute part of the gendering of economic activity. But the motivation for those choices is in part the differing opportunities that men and women experience. The structure of an economy—rights over property and persons, organization of market transfers of property and persons, and rules regarding nonmarket transfers of rights—is itself gendered. There is a pithy saying that economics is all about how people make choices, while sociology is all about how people do not actually have any choices to make. There is no need to reduce the social sciences to such gross stereotypes, but the saying does capture the essence of how to think about the feedback-style relationship between structure and agency. If Africans were rich, the analysis of the economic dimensions of gender would be less urgent. Unfortunately, African countries are the poorest of the world. Chapter 2 offers a broad overview of the problem of poor economic growth performance in Africa and tempers that assessment with more encouraging news of steady and sometimes dramatic improvement in many indicators of well-being, such as infant mortality and schooling attainment. The chapter introduces multiple regression analysis, a key tool of statistical analysis. The technique of multiple regression analysis is applied to examine correlates of economic growth, across the world and among African countries. Attempts to explain variation in development trajectories across countries reveal a major challenge: there can be little confidence that an estimated relationship constitutes a causal relationship. Regression analysis of country-level data suffers from the problem of endogeneity, where the outcomes to be explained are themselves explaining or causing other variables. The endogeneity problem has increasingly led social scientists to turn to experimental methods to try to measure the importance of gender in shaping economic outcomes. Chapter 3 introduces the method of randomized control trials (RCTs). The chapter reviews a number of representative studies using RCTs. Subsequent chapters note RCTs designed to address specific gender questions. Chapters 4–6 describe three structures that are deeply gendered in rural Africa: land tenure, labor allocation, and marriage markets. The system of land tenure (Chapter 4) is a basic structure of an agrarian economy. Tenure

Exploring Gender and Development

5

rules determine the allocation of rights to use and transfer land. Most African societies give women far fewer land-use rights than are given to men. Labor allocation is discussed in Chapter 5, which provides evidence of the gendering of labor markets, principally in the form of norms that enforce patterns of time allocation and occupational segregation by gender. Rural women typically have limited rights to choose how much time to spend on tasks of their choosing; they may be subordinate to the instructions of their husbands or fathers. Moreover, many African societies classify certain activities as “male” and others as “female.” In the past, it was rare for men or women to cross those occupational boundaries, but those boundaries are eroding rapidly as economies urbanize and diversify. Chapter 6 investigates marriage markets, which in many African societies are major economic structures determining the economic possibilities of women. Marriage traditions are constantly being reinvented and reinterpreted, but the marriage institutions of most social groups and countries remain lopsided in favor of men. Subsequent chapters look at important choices that men and women make in the context of the structures that they find themselves in as they enter adulthood. Chapter 7 discusses the organization of households. In rural Africa, many significant economic choices concerning production are made in the context of households. The vast majority of people in Africa do not work in factories or for large organizations. They work with other members of their household. The structure of household production is constituted by the choices made by household members. Chapter 8 focuses on one important household choice, investments in the health of children, while Chapter 9 reviews the economics of household investment in educating children. If investment choices in health and education are themselves gendered, then right away there is a feedback mechanism between structure and choice. Parents make different investments in girls than they do in boys, which affect the skills, outlooks, and opportunities that girls take with them into adulthood. Because of this, young women will make choices that are different from those of young men. People view the social patterns that emerge from these choices as part of the economic structure of their society, which in turn shapes the choices of the next generation. Over the past two decades, microfinance has come to be known as the single most important gendered development program promoted by governments and nonprofit organizations. Microfinance organizations tout their effects on women and claim that microfinance may have dramatic effects in reducing poverty. Chapter 10 addresses some of the many issues surrounding microfinance and shows how evaluations of impact suggest that the effects may be quite modest. Chapter 11 concludes the book with reflections on the gendering of politics in African societies. Political structures are enormously influential

6

Women and Development in Africa

in determining economic structures. African governments are increasingly mandating greater representation in political structures. It remains an open question whether these initiatives have large positive effects, and whether they will lead to more equal gender representation. Defining Development Development is a catchall word that refers to the process by which societies increase the possibilities that people have to realize their potential (Sen, 1999). A society develops when it expands the range of possibilities for people to be fulfilled. Children survive the many life-threatening illnesses of infancy and childhood. They spend their childhood learning, in nurturing educational settings, and enter adulthood with the capabilities and freedom to choose how to realize their potential. A developed society reduces poverty, so that people are less frequently faced with a lack of effective choices. People in a developed society seldom confront discrimination and social norms that curtail their choices. Humans, as social creatures, realize their potential when they are increasingly free to engage in the life of their community, however defined. Development is economic, political, and cultural, and is a process through which freedom in these domains is enhanced. Of course, development is relative. Most people, even those living in poor countries today, would curl their lips in distaste if they were confronted with the living conditions and lack of basic freedoms for people in the most developed societies of 1700. This gloss of development perhaps conceals the many controversies of development. For example, one corollary of development is degradation of the environment. As stewards of the planet, human aspirations for sustainability are often in conflict with aspirations for development. A related point is that development is a process of change, and persons who value tradition often find their sense of identity and well-being threatened by development. Development also often involves growing inequality in access to power. There is much to be learned from the radical critique of development (Escobar, 2011). This critique holds that the language and rhetoric of development obfuscate and mystify in order to forestall questioning of relations of power and oppression that exist around the world. At the risk of sounding blithe, however, this is not the place for an extended commentary on the controversies and rhetoric of development. Development involves many processes and many choices; there is a universe of ambiguity, nuance, and complexity to be explored (Watts, 1993). This book will be a lengthy treatment of just one aspect of development, the gender dimension, where the development process changes some things for the better and others for the worse. Africa at the beginning of the twenty-first century remains the least developed region of the world, in this sense of offering people possibilities to

Exploring Gender and Development

7

realize their potential. Proportionally more infants and children die in Africa than elsewhere. Children who survive are malnourished and have limited opportunities to go to school. African economies offer few opportunities for fulfilling and rewarding work. African societies shackle marginal groups with oppressive and degrading stereotypes. African political systems, though changing rapidly, deny many people essential freedoms. That Africa alone occupies the bottom rung of the development ladder is a recent phenomenon. Thirty years ago, Africans had plenty of company at the bottom. Most of the world’s poor, in fact, lived in China, India, and Bangladesh. But these countries have seen amazing growth in income per capita for several decades now. Hundreds of millions of people who were living on less than $1 a day in 1980 are now earning ten times that amount. The poor in these countries have also experienced significant improvements in living standards that have outpaced their increases in income. Many of these societies have also seen considerable improvements in freedoms to participate in cultural and political life. Unfortunately, most African countries have only seen growth in income since the early 2000s. Income declined or was stagnant for the previous twenty-five years. Moreover, recent African growth has been due to oil and mineral extraction, without sizable spillover benefits for the bulk of the poor. A recent wave of democratization has not been accompanied by deep changes in respect for human rights and opportunities for civic engagement. Africa has seen economic growth without much development. Growth in income, also commonly known as economic growth, is a key component, though by no means the only component, of development. Income growth is measured using statistics on national economic activity. Virtually all countries of the world regularly calculate the value of goods and services produced in the country in a year, known as gross domestic product (GDP). Most also report gross national income (GNI), which measures how much of domestic and foreign GDP accrues over a year as income generated by nationals of a country. GDP is increasingly diverging from GNI. Many Africans earn very high incomes when they are abroad, and much income produced in Africa takes the form of mineral and oil extraction by multinationals. Since a significant share of income earned abroad is remitted, and since much income earned by multinationals is repatriated, GNI might be a more suitable measure of well-being. GDP might, however, still be a good indicator of the future or potential earnings of Africans born in their countries. Another measure that might be considered is net national product, which subtracts the considerable depreciation of capital and infrastructure from the total product produced in a year. If depletion of natural resource assets were also subtracted from total production, by the argument that sales of existing assets are not really “goods produced,” then incomes would be lower (Winter-Nelson, 1995).

8

Women and Development in Africa

An issue arises when valuing GDP or GNI in a common currency such as the US dollar. To compare incomes earned in different countries, the local cash value of production and incomes measured in local currencies must be converted into a common currency. There are two main alternatives. The first is to use current market exchange rates. The second is to use purchasing power parity (PPP) exchange rates. Many economists favor the PPP rates. These are constructed by taking samples of the prices of goods and services that are purchased in most countries and calculating the exchange rate that would equalize the prices of a basket of these goods with the prices in the United States. Suppose it costs $10 in the United States to purchase a given amount of food and the same amount costs 20 rupees in India. It makes sense then to use a ratio of 2 rupees per dollar when comparing people’s incomes. The problem with market exchange rates is that they need not equalize purchasing power. Large deviations from purchasing power parity occur and persist, and market exchange rates often exhibit dramatic swings from one year to the next. Because of these swings, if market exchange rates were used to compare incomes in common currencies, then incomes also would appear subject to large swings. But income does not exhibit such large swings when measured in national currencies, so neither should it when using a common currency. PPP rates are not necessarily superior when it comes to comparing African incomes with those of wealthy, industrial countries. When PPP rates are used, African countries appear to have higher incomes. This is because many goods and services produced in African economies are laborintensive and not traded internationally (such as housing, haircuts, guard services, and food processing). Wage rates in African countries are extremely low, sometimes amounting to no more than US$0.25 per day in rural areas (using market exchange rates). Therefore, services are very, very cheap. When these services are priced at the same price as in industrial countries, they appear more valuable, and income appears higher. Table 1.1 presents the average rates of growth of GDP per capita for countries in sub-Saharan Africa. The table ranks African countries by their GNI per capita in 2011, converted to US dollars by the World Bank’s Atlas method, which is an exchange rate constructed as an index of recent averages of market exchange rates and inflation rates. The table also shows GDP per capita converted into US dollars using PPP rates. The PPP figures are about twice as high as the GNI levels. Differences are partly due to differences in PPP and Atlas methods, and partly due to the conceptual differences between GNI and GDP discussed earlier. The final columns show the population of each country and the cumulative percentage of all the population of Africa living in countries with lower average income than the income of that country. Thus, about 50 percent of the population in Africa in 2011 lived in countries with average GNI per capita lower than US$800 per year.

9 Table 1.1 Growth Rates and Levels of Income per Capita for African Countries Real GDP Real GDP GNI per Capita 2011 in US$ GDP per Capita per Capita per Capita 2011 in US$ Growth Rate, Growth Rate, Using Atlas Method Using PPP 1975–1999 2000–2011

Congo, Democratic Republic of 230 Burundi 240 Malawi 320 370 Liberia Ethiopia 380 390 Niger Madagascar 430 Guinea 440 Uganda 440 Eritrea 450 Togo 500 Central African Republic 510 Guinea-Bissau 510 Mozambique 510 Tanzania 570 Sierra Leone 580 Rwanda 600 Zimbabwe 650 Mali 660 Burkina Faso 670 Benin 750 Chad 770 Comoros 840 Kenya 860 Senegal 1,030 Mauritania 1,110 Cameroon 1,170 Côte d’Ivoire 1,220 Zambia 1,350 Lesotho 1,380 Nigeria 1,440 Sudan 1,500 Ghana 1,550 Congo, Republic of 2,550 Swaziland 2,860 Cabo Verde 3,830 Angola 4,580 Namibia 5,610 South Africa 7,610 Botswana 7,650 Mauritius 8,570 Gabon 10,040 Seychelles 12,260 Equatorial Guinea 13,560

364 483 660 560 971 674 843 921 1,165 488 906

943 965 882 1,380 1,171 1,167

–0.05 –0.01 0.00 –0.11 –0.01 –0.02 –0.02 0.00 0.02 0.06 –0.01

0.03 0.00 0.01 0.03 0.06 0.01 0.00 0.00 0.04 –0.03 0.00

Cumulative Percentage of Population Population of (millions) Sub-Saharan Africa 66 10 16 4 92 17 22 11 36 6 7

7.4 8.5 10.3 10.7 21.1 23.0 25.5 26.8 30.9 31.5 32.3

1,047 1,304 1,364 1,870 1,060 1,522 1,671 2,244 2,025 1,757 1,475 1,692 2,335 1,894 1,764

–0.02 0.01 0.01 0.00 –0.02 –0.01 0.00 0.00 0.01 0.01 0.00 –0.01 0.00 –0.01 –0.01 –0.01 –0.02 –0.02 0.02 –0.02 0.00 0.00

0.03 0.00 0.05 0.04 0.04 0.05 –0.05 0.02 0.03 0.01 0.06 –0.01 0.02 0.01 0.03 0.01 –0.01 0.03 0.03 0.06 0.04 0.04

5 2 25 48 6 11 14 15 16 10 12 1 43 14 4 22 20 14 2 169 37 25

36.0 35.3 35.1 41.4 42.1 43.3 44.9 46.6 48.4 49.5 50.9 51.0 55.9 57.4 57.8 60.3 62.5 64.1 64.3 83.3 87.5 90.3

26,057

0.08

0.07

1

100.0

3,814 4,522 4,245 5,262 6,520 9,860 14,109 13,056 13,811 23,416

0.00 0.03 0.05 –0.03 –0.01 –0.01 0.06 0.04 –0.01 0.03

0.02 0.01 0.06 0.08 0.03 0.02 0.03 0.03 0.00 0.02

4.3 1 0 21 2 51 2 1 2 0

90.8 91.0 91.0 93.3 93.6 99.4 99.6 99.7 99.9 99.9

Source: World Bank (2013). Note: The growth rates are calculated as regression coefficients against a time variable. Data are not available for Somalia, São Tomé and Principe, and Djibouti, nor is GDP per capita using purchasing power parity exchange rates available for Zimbabwe.

10

Women and Development in Africa

Figure 1.1 shows the evolution of a weighted average of per capita GDP levels for countries in sub-Saharan Africa. The GDP levels are weighted by population in each year. The solid black line, which displays very little variation, is this weighted average. For comparative purposes, the growth rates of India, China, and Malaysia are also presented. The populations of India and China each exceed that of the sum of countries in subSaharan Africa. Malaysia has been a star performer and is an example of the likely “upper bound” to economic growth. The graph covers the period

Figure 1.1 Growth of Real GDP Per Capita, 1960–2011 (constant 2000)

Source: World Bank (2012).

Exploring Gender and Development

11

1960–2011. African countries in 2011 were very far from the income levels of the rest of the world. They were even further from the income levels of rich countries, approximately US$40,000 per capita. Such gross inequality is an urgent problem. The table and figure make clear that growth performance in Africa was very bad until the late 1990s. Almost half of African countries experienced negative growth during the period 1975–1999. Indeed, most of the population of African countries had lower income in 1999 than in 1970. Only nine countries experienced growth over 2 percent per year during the 1975–1999 period, and, of these, three were small island nations (Cape Verde, Seychelles, and Mauritius) and another two were the small countries of Swaziland and Lesotho, entirely surrounded by South Africa. Uganda’s and Eritrea’s growth were recoveries from years of civil war—no small feat, but incomes remain dismally low. Equatorial Guinea’s growth has been due entirely to the recent discovery and exporting of offshore oil. The president and his family control the wealth. Even Botswana’s steady and rapid growth over the decades has been fueled by the diamond-mining sector and has slowed considerably, in part because of the HIV/AIDS (human immunodeficiency virus infection/acquired immunodeficiency syndrome) epidemic. The slow and often negative growth of the “lost 25 years” from 1975–1999 was especially serious because of the low levels of income that African countries had at independence. Income levels have barely budged and thus remain extremely low compared with the rest of the world. The upturn in growth since 2000 is very evident in the table and less evident in the figure (since a large percentage increase from a very small base is not very large compared to the income levels of Malaysia and China). The steady growth in Africa was not halted by the global economic recession of 2008–2010. There is considerable optimism about African economic growth at the time of this writing. Nevertheless, as shown in Figure 1.1, compared with growth in GDP per capita for India, China, and Malaysia, African countries remain woefully behind. Even continued high growth rates for decades will not enable them to converge with other developing countries. Some societies with low GDP per capita levels have very high levels of human development. They have good childhood nutrition and low mortality, high education levels, meaningful work, adequate standards of living, and equal access to primary and basic health care. Unfortunately, the evidence that African populations are disadvantaged on all dimensions, from GDP to malnutrition, comes through plainly in national statistics, field studies, and qualitative fieldwork. The Human Development Report, for example, published annually by the UN Development Programme, compiles many statistics indicative of the quality of life beyond the income measures of GDP and GNI (Klugman, 2011). The report calculates a Human Devel-

12

Women and Development in Africa

opment Index (HDI) that aggregates a number of indicators. African countries are at the bottom of the HDI, and virtually all of the bottom twenty countries are in Africa. Six of these indicators of well-being are plotted in Figure 1.2, for African countries averaged together, and again for India and China. There are many gaps in individual country coverage of these measures of well-being, so the average African performance reflects also an underlying variation in which countries reported in which years. Nevertheless, the trends are clear and robust to different ways of aggregating the African data. African countries as a whole have been improving at close to the same rate as India and China, but they remain stubbornly less well-off. Mortality of infants and children remains high. The death rate is almost 100 per 1,000 for infants and 150 per 1,000 for children under five in the bottom half of the group of African countries, and not much better even in the well-off countries. For schooling, primary completion and secondary enrollment are growing but are still far below the levels of India and China (which has 100 percent primary enrollment). Many children in Africa still lack opportunities to attend school simply because there is no school available. Moreover, the quality of education, essential to full participation in a world of “knowledge workers,” is vastly lower in African countries. Total fertility rate, defined by the World Bank as “the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with current age-specific fertility rates,” is an important indicator of the choices women have to pursue other opportunities besides child rearing. In all societies of the world, fertility rates fall as economies grow and development happens. Fertility rates in African countries have been falling, but remain much higher than those of India and China. Life expectancy in Africa has only recently begun to rise from the plateau of fifty years that resulted when rising deaths due to the HIV/AIDS epidemic cancelled improvements in life expectancy due to advances in public health infrastructure. For measuring development, income and physical well-being may not necessarily be the only measures or the best measures. Human flourishing may be more about mastering dance moves than listening to music on $200 headphones. Perhaps Africans do not value material possessions and instead spend time talking with neighbors and enjoying the arts. Perhaps some other measure of satisfaction or happiness would reveal Africa to be less poorly off. There is little evidence to buttress this viewpoint. Deaton (2010, p. 244) analyzed data from the Gallup World Poll of 2006 and concluded that the map of life satisfaction measures “looks similar to an income plot of the world: North America, Europe, Japan, Australasia, and Saudi Arabia are happy as well as rich, and the really unhappy places on the planet are in sub-Saharan Africa, plus Haiti and Cambodia.”

13 Figure 1.2 Indicators of Well-Being, 1975–2011 (unweighted average of available African countries)

Source: World Bank (2013).

14

Women and Development in Africa

Defining Gender Gender refers to a set of shared and evolving discursive habits that prescribe and proscribe behavior for persons in their social roles as men and women and that structure analysis and decisionmaking about the actions of others. These discursive habits may be invoked deliberately or they may emerge, unintended, from processes operating in the brains of people. In this sense, all societies of the world are known to be gendered. Gender is mental baggage, carried wherever one goes. Gender matters for practically everything people do; it is hard to think of any social or even individual action where gender might not be an important consideration in understanding action. The habits of gender include characterizations, inferences, expectations, and decision algorithms. The habits guide people when they think about how other people might react to proposed action. The habits are cognitive realizations that are habitual. People do not reconstruct gender from scratch as they ponder and evaluate a situation. Habitual, but not static; gender does change and evolve. Actions that take place in social settings, under the gaze of others, update gendered discursive habits. Gender shapes actions and is shaped by reasoned judgment about the actions of others; when habits and reasoned evaluation conflict, persons might experience cognitive dissonance, which generates a mental impetus to change habits. Shared discursive habits of gender overlap with discursive practices related to other important discursive constructs (ethnicity, religion, nationality). Ambiguity, contradiction, and conflict emerge across different categories of social construct. When at war, should a woman butcher an enemy soldier with her bayonet and shout in triumphant bloodlust? Gender in the social sciences is different from the biological categories of male and female. Biology typically focuses on how hormones such as testosterone affect physical and neurological processes (Udry, 1994). There is little doubt that hormones do affect neurological processes of cognition (Sapienza, Zingales, and Maestripieri, 2009; Zethraeus et al., 2009). But cognitive processes are so complex that it is unclear whether there is, at present, much to make of findings that simple, single-issue cognitive processes—such as solving a maze or choosing between $10 now and next week—are affected by hormones. Someday the accumulation of findings may offer useful insights into the underlying brain chemistry of gender and large-scale social patterns related to gender; at present they do not. Sociobiology maps biological differences onto behavior, arguing that biological differences create different incentives for specialization and other behavioral arrangements as humans coexist in groups. In general, the discursive habits discussed in this book are many steps removed from the basic practices and discursive habits of small hunter-gatherer societies that sociobiology might explain. Sociobiology argues that patterns and institu-

Exploring Gender and Development

15

tions that emerged in prehistory strongly persist into the present. But it has little to offer in terms of explaining how present-day institutions affect economic outcomes. For that, there is little need to understand the deep origins of gendered structures. The enormous variety and constant change of gender constructs around the world suggest that the only relevant fact to be explained by sociobiology is why virtually all agrarian and nonindustrial societies have been, persistently, patriarchal. But if present trends, and science fiction writers, are reasonable guides, it seems very likely that another thousand years will see many nonpatriarchal societies flourishing. Sociobiology may not have much to contribute if it devotes attention solely to explaining male dominance. While the biological division of humans into male and female accounts for much human behavior, there is more variance in the “sex” spectrum, and the complexity of hormone interactions means there is considerable room for more complex ranges of outcomes. Male and female do not completely capture the panoply of biological possibilities among humans, and male and female involve different biological characteristics over each individual life cycle (Kennelly, Merz, and Lorber, 2001). These issues are beyond the scope of this book. Gender is instantiated in the social structures that shape choices that individuals make. Social structures include legal rules, norms of behavior, ongoing institutions and hierarchies, patterns of competition in different markets, persisting prices of goods and services, shared expectations about the future, and mechanisms for nonmarket transfers of property, services, and resources. Structures fix the space of feasible action that persons might take. Gender becomes incorporated into and may be propagated by structures that take the forms of defined and named social institutions. The obstacles to changing institutions may be so overwhelming that people devote little attention to thinking about gender as a changeable discursive habit pertaining to the institution. The institution makes the habit automatic, and the institution is inhabited by people who are charged to, and given incentives to, reinforce the habits supported by the institution. The Catholic Church, for example, preached sharp distinctions for appropriate behavior for men and women. As an exemplar of these distinctions, the church forbade women from becoming priests and celebrating mass. Few people over the centuries have thought that expressing their discontent, or leaving the church, would change the institution in their lifetime. The gender strictures advocated by the church persisted and became viewed as natural and immutable. Similar discursive habits are made concrete by other religious institutions and systems of thought. Structures bound opportunities in a relational way. Each individual in a society occupies a different position in relation to a structure. Sometimes those positions are unique to each individual, as when a person contem-

16

Women and Development in Africa

plates who in his or her circle of friends might help in a time of need. In other settings, the relational spaces are common: all women are excluded from the priesthood in the Catholic Church. Sen (1981), in stressing the role of market structures in perpetrating famines, introduced the notion of structures as shaping the entitlement mappings of individuals. Food entitlements, how much food a person or family might command, determine whether someone lives or dies. More broadly, entitlements, opportunities, budget constraints, and feasible actions are all different terms for expressing how people might reasonably consider their actions to be constrained. The models that people construct to make sense of the world are themselves structures, if they are widely shared. If many people share the same model of how the economy or society works, then their behavior will be patterned in predictable ways. If everyone thinks the Catholic Church does not allow women into the priesthood, then no women apply, no male priests challenge the system, and the pope devotes little time to the issue. The structure persists, and people’s models are confirmed. If everyone thinks the price of gasoline will be similar to what it was yesterday, then no one changes their purchasing and supplying behavior to any large degree, and the price indeed turns out to be the same. In evolutionary biology, structure consists of the set of evolutionarily stable strategies, or modes of behavior. Because some genes are encoded to respond to social stimuli in certain ways, genes that are similarly encoded may do better than other kinds of genes. This is one way of thinking about the persistence of altruistic and social behavior in many species, including humans. Political processes are the metastructures of society. Political actions are often geared toward changing structures through deliberation, and political institutions are the structures that shape how structures can be changed in this way. Political structures are also about the way violence is controlled; this is of paramount importance in many African societies, as so many swing from periods of horrific violence to periods of relative calm with low-intensity violence in the background. Moreover, violence against women, whether condoned or ignored by the polity, forms an important backdrop to gender structures (Green, 1999). Sometimes structures shape choices by bundling rights, obligations, contingencies, and commodities. Choices that respect the bundling are treated as legitimate by other actors in society, while choices that unbundle the package are considered illegitimate. Sexual acts, for example, are often bundled with rights and obligations to care for children. It is not always easy to identify precisely what constitutes a social structure in any given society. The forms of structures vary. In some cases, structures are monolithic entities, institutionalized, codified, and crystal clear. In other cases, multiple structures exist and people choose and negotiate among structures to secure their own advantage. For many African so-

Exploring Gender and Development

17

cieties, structures are not written down. Regimes and rules are orally transmitted across generations and social groups, through recitation and collective memory. This is why “elders” are often so important in many African societies. Their expertise, derived from experience, makes them repositories of structure. Measuring Gender Inequality Few would dispute the assertion that the discursive habits of gender and the gendered social structures that exist in most societies favor men over women. This generalization holds for African countries and many other countries around the world, over much of history. Philosophers from antiquity in all major world traditions have remarked upon gender inequality. Folktales passed on through many oral cultures embed inequality in normative presentations of social relations. Gendered norms and institutions that are lopsided mean that women and girls are disadvantaged in life relative to men and boys. There are no debates over urgent problems of too many female presidents, too few male mathematicians, too many ethnic groups where women own all the land, too few men’s bathrooms in parliament, too many male victims of domestic violence, or too many women using household money for drink. The pervasiveness of female disadvantage is a commonplace and poorly understood feature of economic life. There are people who believe that comparing the well-being of boys and girls, or men and women, is like comparing papayas and lemons. One is sweet and the other sour, and the difference does not imply unequal welfare. Men and boys have their lives to lead and girls and women theirs. Nevertheless, girls and women in Africa face unequal chances for education, less inheritance and ownership of assets, discrimination in employment and occupations, violence at home and in public spaces, and limited political representation. These add up to unambiguously diminished welfare and capacity to fulfill life aspirations. The situation is more like comparing a rotten papaya with a lustrous lemon. Self-ownership is a basic structure of social life. In societies with slavery, self-ownership is obviously highly unequal. Self-ownership refers to whether a person owns her or his body and has control over her or his actions. Is a person an inferior entity in the social order, a legal minor? When women do not own themselves, the structure of self-ownership clearly trumps any other structures of social life for women. Societies around the world have systematically and extensively denied to women the right of self-ownership. Many African societies in the past have treated women as Coverture

18

Women and Development in Africa

legal minors unable to make independent decisions. Some continue to deny to women full rights of personhood. In the United States, the legal system whereby married women were considered to be minors was called coverture (after the French, femme couverte) and was in place in many states of the union until the early twentieth century. Hymowitz and Weissman (1978, p. 23) described the system as follows: Under coverture married women had no property and no money of their own. Even a married woman’s clothing and household goods belonged to her husband. A married woman’s dowry and any inheritance she might receive also belonged to him. In most instances he could sell her property without her consent. Nor did a married woman control her own wages. She could work all day as a servant, and her husband was legally free to take her earnings and allow her to starve. A married woman could not buy or sell anything without her husband’s permission. She could not make contracts, sue in court, or be sued.

A significant economic effect of coverture is likely to be the disincentive for women to carry out any extra effort at improving their livelihoods, if their husbands get to appropriate all of the benefit. Women are not the “residual claimants” of their efforts. A woman who wants to become an entrepreneur will think twice. Her husband will have first rights to all of her profits. There is no comprehensive data set measuring self-ownership by gender across countries. Yet, there is much evidence from anthropologists suggesting that coverture has been the norm in customary law for many African societies. Lack of control over self has been most evident in cases of forced marriage and bridewealth used to purchase a wife. There has been much discussion in anthropology as to whether societies with bridewealth ought to be thought of as having men purchase rights over women. That is the case for some, but not all, societies with bridewealth. Gray (1960, p. 36) summarized the evidence from a good number of societies of Central Africa: “Certain defined rights in a woman are regularly sold and purchased for a price, and therefore it is not absurd to consider her as property.” Gray’s interpretation was consonant with other contemporary reports. Colson quoted a woman as saying: “When we were young there was no argument. If a woman told her parents that her husband beat her, her mother would say, ‘you stay or we will beat you: we have already . . . accepted bridewealth for him. It is for you to return to him’” (1958, p. 183, cited on p. 119 of Chanock, 1985). As one male informant speaking about the colonial era in Tanzania put it to Lovett: “Once bridewealth was given for a woman, she was like a slave. She had no freedom because she could say nothing” (1996, p. 61). Even in Ghana, with matrilineal societies that gave women consider-

Exploring Gender and Development

19

ably more rights and privileges than many other societies, women were offered up as “pawns” for loans between men (Austin, 1993). The uncle or guardian of a woman would give her to a creditor as a kind of collateral. The creditor then had rights over the woman much as he might have rights over a piece of property. Sometimes the creditor could be her very own husband, lending money to the uncle (of his wife). In those cases, according to Grier’s (1992, p. 311) reading of the available historical record, a man was said to have “bought” his wife. The evidence suggested to Grier that during the World War I cocoa boom it was very common for men to obtain land for plantations by pawning female relatives. Girls were reduced to the status of indentured servants. Women could apparently buy themselves out of these sale marriages and indentureships. Gray’s own analysis of Sonjo bridewealth practices suggested as much: In most respects a broken betrothal is dealt with by the Sonjo in the same manner as a divorce. If a young man does not wish to marry his fiancée he sends her a broken twig, which signifies his decision. The girl is then free to accept another suitor, who has only to pay the fiancé the original bride-price in order to marry the girl. The girl herself can also break the engagement if she can find an alternative suitor who is willing to pay back the bride-price. If the fiancé finds another man who is willing to buy his marriage rights, the girl is obliged to accept the change unless she can find someone more to her liking before the time set for marriage. Whoever finally marries the girl must make the regular payment of ten goats to her father before the marriage, and must also pay a fee of seven goats to the village council, which is customary at every divorce and remarriage. (1960, pp. 41–42)

Cornwall (2001) offered a particularly poignant tale from southwestern Nigeria: Iya Safuratu was initially engaged to a man who turned out to be a thief. She needed a way to repay her dowry and had no means of her own. A wealthy man, an associate of her brother, approached her parents, who gladly accepted him as a husband for her. He repaid her dowry and they were married. He married a string of wives, but gave them nothing at all to feed themselves or their children. Each wife used her own business to keep her hearth-hold going. Iya Safuratu struggled to make ends meet. But she endured. At first she was the youngest of four wives, but her husband went on to marry another seven. She stuck it out. It was only when he told her one day that from then onward his latest wife would be doing everything that Iya Safuratu had been doing on the farm she decided she had to go. She went back to her parents and started trading, saving small amounts of money. And the day came, finally, when she could go to court and repay her dowry. All these years later, she proudly showed me a receipt dated 1963. (p. 76)

There is evidence that forced marriage and household servitude have

20

Women and Development in Africa

remained common. In Coastal Kenya, women in Mijikenda communities were not allowed to visit their natal families (Ciekawy, 1999). Hodgson (1996) offered a fascinating case study of a Maasai woman who resisted the authority of her biological father. The woman had been raised by her mother and stepfather since infancy. When she went to marry the man of her choosing, her father announced that he had already accepted bridewealth from a friend. Maasai elders concurred and obliged the woman to break her current engagement. She fought her elders and took her case to the national court system. A judge reprimanded the elders, who continued to insist that the woman did not have the right to choose her marriage partner. Another Maasai woman, Agnes Siyiankoi, also made headlines in the late 1990s for taking her husband to court for beating her. When asked why she had agreed to marry the man who had beaten her for more than a decade, Siyiankoi responded that her father would have cursed her if she had not done as he willed (Mwangi, 1997). These examples illustrate how a young woman becomes convinced that her marriage will be enforced against her will. She must either remain in the marriage or suffer physical punishment. Perhaps, over the years, she will resign herself to her situation, or rationalize it, and may in the end even participate in the same violent punishment of other young women who try to test the enforceability of these marriages. When analyzing gender inequality, economists often focus on a variety of outcomes that are indicative of gender differences in well-being or quality of life. Table 1.2 presents five of these indicators of well-being for African countries. The indicators are used by the authors of the Human Development Report to calculate a Gender Inequality Index (GII). The higher the GII, the worse is gender equality. The GII is explained as follows: “The index shows the loss in human development due to inequality between female and male achievements in these dimensions. It ranges from 0, which indicates that women and men fare equally, to 1, which indicates that women fare as poorly as possible in all measured dimensions.” The data in Table 1.2 come from the Human Development Report of 2013, which in turn compiles data from a variety of sources. (Some of the data are from years earlier than 2010, but all indicators are quite recent.) The first indicator, adolescent fertility, is the equivalent of teen pregnancy. One perhaps thinks that a young woman, in order to realize her potential, should not have a baby too early in life. Especially if, as shall be seen, responsibility for raising children is not shared equally between mother and father but rather often falls on the mother alone. African countries still have very high rates of teenagers having children. About one in The Gender Inequality Index (GII)

Exploring Gender and Development

21

ten young women will have a child before age nineteen. A related indicator is maternal mortality. Women in Africa very frequently die in childbirth. This high mortality is likely a correlate of broader gender inequality because maternal mortality is preventable at relatively low social cost through an effective delivery of basic antenatal and postnatal care (Liljestrand, Table 1.2 Various Indicators of Status of Women in Sub-Saharan Africa Maternal Population Adolescent Mortality Ratio with at Least Fertility Rate (deaths of Labor Force Secondary (births per 1,000 women per Participation Education women, 100,000 Rate (female/ (female/male aged 15–19) live births) male ratio) ratio)

Burundi Mauritius Rwanda Botswana Seychelles Ethiopia South Africa Comoros Sudan Zimbabwe Eritrea Togo Namibia São Tomé and Principe Lesotho Ghana The Gambia Swaziland Somalia Mauritania Gabon Senegal Guinea-Bissau Benin Kenya Central African Republic Sierra Leone Malawi Côte d’Ivoire Nigeria Congo, Republic of Equatorial Guinea Cameroon Burkina Faso Madagascar

21 32 36 44 48 48 50 51 53 53 54 54 54

800 60 340 160

350 300 280 730 570 240 300 200

55 61 62 67 68 68 71 81 90 96 97 98

70 620 350 360 320 1,000 510 230 370 790 350 360

113 115 115 117 123

560 240 690 300 240

99 104 106 106 111

890 890 460 400 630

1.02 0.58 1.01 0.88

0.87 0.72 0.44 0.40 0.93 0.89 0.99 0.84

0.57 0.80 0.93 0.87 0.62 0.49 0.36 0.87 0.75 0.87 0.86 0.86

0.85 0.96 1.04 0.64 0.76

0.94 0.87 0.83 0.86 0.94

0.57 0.85 0.93 0.95 1.01

0.95

0.70 0.79

0.34 0.97 1.11 0.74 0.54 1.08

0.39 1.55 0.42

0.44 0.48

0.39 0.47 0.51 0.46 0.90

0.61 0.28

Percentage Female in Parliament 35 19 52 8 44 26 41 3 24 18 22 11 25

18 26 8 8 22 14 19 17 42 10 8 10

13 13 22 11 7

10 10 14 15 16

HDR Gender Inequality Index 0.48 0.38 0.41 0.49

0.46

0.60 0.54

0.57 0.46 0.53 0.57 0.59 0.53

0.64 0.49 0.54

0.62 0.61

0.65 0.64 0.57 0.63 0.61

0.63 0.61

(continues)

22

Women and Development in Africa

Table 1.2 continued

Population Maternal Adolescent Mortality Ratio with at Least (deaths of Labor Force Secondary Fertility Rate (births per 1,000 women per Participation Education 100,000 Rate (female/ (female/male women, aged 15–19) live births) male ratio) ratio)

Liberia 123 124 Mozambique Uganda 126 129 Tanzania Guinea 134 Chad 138 Zambia 139 Angola 148 Mali 169 Congo, Democratic Republic of 171 Niger 194

770 490 310 460 610 1,100 440 450 540 540 590

0.90 1.04 0.96 0.98 0.84 0.80 0.86 0.82 0.53

0.97 0.44

0.40 0.25 0.96 0.61

0.58

1.23

0.30 0.33

Percentage Female in Parliament 12 39 35 36

13 12 38 10

8 13

HDR Gender Inequality Index 0.66 0.58 0.52 0.56

0.62

0.65

0.68 0.71

Source: Human Development Report (2013). According to the HDR, “the Gender Inequality Index relies on data from major publicly available databases, including the maternal mortality ratio from the United Nations Maternal Mortality Estimation Group (MMEIG), the WHO, UNICEF, UNFPA and the World Bank; adolescent fertility rates from the UN Department of Economic and Social Affairs’ World Population Prospects; educational attainment statistics from the UNESCO Institute for Statistics educational attainment tables and the Barro-Lee data sets; parliamentary representation from the International Parliamentary Union; and labour market participation from the International Labour Organization’s Key Indicators of the Labour Market (KILM), 7th Edition.” Note: Data are missing for selected variables for some countries.

1999; Winikoff, 1991). The fact that so many women die is a signal of misplaced and gendered public health priorities. Another indicator of inequality is labor-force participation. Strong norms in many societies proscribe women from working outside the home. Sociologists and anthropologists have often observed that women working together in factory situations outside the home develop and exercise more power when they return to their household in the evening. Men everywhere seem to resist this empowerment. African countries have relatively low female labor-force participation. Secondary and tertiary (university) school attainment is highly gendered around the world, though the gender gap has been falling rapidly (Sundstrom, 2004). Labor-force participation is not closely correlated with secondary school attainment in Africa, because so many people work in agriculture. However, education is correlated with high adolescent fertility, and the ratio of female to male secondary schooling is generally less than .50 for countries with high gender inequality indices and high fertility. One final measure included in the GII is the ratio of females to males in national parliaments. Many countries in Africa now have explicit gender quo-

Exploring Gender and Development

23

tas for parliamentary representation. Even so, the representation of women in parliaments is far from equal. The average ratio across countries is .24; there are four times as many men as women in the average parliament. The table presents the countries sorted according to the level of adolescent fertility, which has a correlation of .78 with the gender inequality index. Adolescent fertility is a good first indicator of gender inequality. The sorting makes clear that southern African countries, along with the Great Lakes countries of Rwanda, Burundi, and Uganda, have low GII scores, and thus relatively more gender equality. Three Sahelian countries (Chad, Niger, and Mali) and the Democratic Republic of Congo have very high GII scores (in the .70 range), indicating much gender inequality. Another indicator of gender inequality was compiled by Cingranelli and Richards (2010), who codified adherence by countries to a variety of human, civic, economic, and social rights, for all countries of the world. Their database includes several measures of women’s rights. These are averaged across African, Asian, and Latin American countries over the period 1980–2010. The trends for the three regions are presented in Figure 1.3. Women’s Economic Rights codifies on a scale from 0 to 3 whether the government, in law and practice, enforced a set of ten basic rights for women in the workforce (e.g., women have the right to work at night, in the police, free from sexual harassment). Women’s Political Rights codifies on a scale from 0 to 3 whether the government, in law and practice, enforced five basic rights for women in politics (e.g., right to vote, run for office, join political parties). Women’s Social Rights codifies on a scale from 0 to 3 whether the government, in law and practice, enforced a set of twelve basic rights for women in a variety of social roles (e.g., right to obtain a passport, equal inheritance, freedom from forced sterilization). Figure 1.3 includes, as a point of comparison with these women’s rights, Cingranelli and Richards’s Empowerment Rights Index, which is “an additive index constructed from the Foreign Movement, Domestic Movement, Freedom of Speech, Freedom of Assembly & Association, Workers’ Rights, Electoral Self-Determination, and Freedom of Religion indicators.” The index has a range from 0, indicating no government respect for the rights, to 14, indicating full government respect. The index does not take into account gender differences in these rights. Figure 1.3 shows that African countries were somewhat more gender unequal than Asia and Latin America. All three regions were quite unequal, with low scores for the three indicators of rights. There were improvements over time in women’s political rights, but not in economic and social rights. General human rights irrespective of gender, as captured in the Empowerment Rights Index, were much higher in Latin America than in Asia or Respect for Women’s Rights in Law and Practice

24

Figure 1.3 Indicators of Women’s Rights, 1980–2010

Source: Cingranelli and Richards (2010).

Exploring Gender and Development

25

Africa, though there appeared to be deterioration rather than improvement in all three regions. There have been a number of other efforts to capture these kinds of indicators (Crotty, 2009; Htun and Weldon, 2011). The Social Institutions and Gender Index (SIGI), available at genderindex.org, is based on the Gender, Institutions, and Development Database of the Organization for Economic Cooperation and Development (OECD). The SIGI aggregates a number of measures of institutional discrimination. Another measure of gender equality can be constructed by examining commitment to the international treaty known as the Convention for the Elimination of Discrimination Against Women (CEDAW) and other treaties that bind states to promote women’s equal rights. CEDAW sets standards for women’s rights in the political, cultural, economic, social, and family sectors, and calls for action to redress and arrest many forms of gender discrimination. Furthermore, the convention calls for specific actions to remedy discrimination. The UN General Assembly adopted CEDAW on December 18, 1979, and the convention entered into force on September 3, 1981, after it had been ratified by twenty states. During the following twenty-five years, virtually all states of the world ratified the treaty. The last states to ratify the convention, other than small island states, were Afghanistan, Syria, United Arab Emirates, Swaziland, and São Tomé and Principe in 2003 and 2004, Oman and Brunei in 2006, and Qatar in 2009. The only major countries not having ratified the treaty by 2013 were Iran, Sudan, Somalia, South Sudan, and the United States. South Sudan of course only recently became an independent country, and Somalia lacked an effective state, leaving the United States, Iran, and Sudan as the only countries deliberately not ratifying the treaty. There are three ways that ratification of an international treaty such as CEDAW directly impacts domestic practice of gender relations. First, CEDAW calls for changes in national constitutions when these are in contradiction to the treaty. Second, CEDAW calls for enactment of legislative and executive policies to bring the countries’ laws into conformity with the treaty. Third, the judicial system in a country might reinterpret previous law in the light of the treaty commitment. Two examples give a sense of the mixed effects of treaty ratification. In 1990, the Tanzanian High Court invalidated customary law that prevented women from inheriting clan land from their fathers (Coldham, 1991). The case was summarized nicely by the International Labour Office (2002): Adherence to International Treaties

Holaria Pastory brought a court challenge to the Haya customary law that prevented her from selling clan land. She had inherited land from her father, through his will, but when she tried to sell it her nephew applied to have the

26

Women and Development in Africa

sale voided. Tanzania’s Declaration of Customary Law clearly prohibited her sale of the land in s. 20 of its rules of inheritance. Pastory argued that this constraint on women’s property rights violated the Constitution. The court was faced with the difficulty of interpreting a constitutional guarantee of freedom from discrimination that did not make any specific reference to women. The court relied on the fact that the Tanzanian Government had ratified the Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW), as well as other international treaties and covenants, to find that women were constitutionally protected from discrimination. The court stated that “the principles enunciated in the above named documents are a standard below which any civilised nation will be ashamed to fall.” The High Court decided that the rules of inheritance in the Declaration of Customary Law were unconstitutional and contravened the international conventions which Tanzania had ratified. Thus, the rights and restrictions around the sale of clan land are the same for women and men.

An example from Zambia, however, illustrates the limited scope of CEDAW alone to effect change. A gender activist, Sara Longwe, sued the Lusaka InterContinental Hotel (Tamale, 2001). The hotel had a policy that unaccompanied women were not allowed onto the premises. The motive was to discourage prostitution. Longwe, however, saw the policy as being discriminatory: an unaccompanied woman was presumed guilty and denied entry into a “public space.” Longwe won the case, with the justices citing CEDAW in their decision. Unfortunately, the Longwe case seems not to have been precedent-setting, for a new Lusaka High Court judge dismissed a similar case brought by Elizabeth Mwanza in 1997. She, too, was evicted from a hotel for being unaccompanied by a man. Ratification of CEDAW by a state is not a sufficient measure of the commitment of a state to the evolving gender rights regime. CEDAW is somewhat remarkable because so many ratifiers entered strong reservations to their ratification. The “quality” of ratification has been quite different for different countries, perhaps more so than any other human rights treaty. Many predominantly Muslim countries ratified the treaty subject to major reservations that essentially rendered the treaty meaningless. Article 28 of the convention states that reservations may be made at the time of ratification or accession, although reservations that are incompatible with the convention are supposedly prohibited. About one-third of the countries that have ratified the convention have done so with reservations to one or more of the articles. Half of the states entering reservations submitted technical or symbolic reservations (some states reserved the right to continue male privilege in the exercise of royal and chiefly power). The other half of states entering reservations were excepting themselves from important substantive obligations of the treaty. A report prepared by the UN-CEDAW secretariat was scathing regarding many of the reservations of the predominantly Mus-

Exploring Gender and Development

27

lim states that constituted most of the states entering serious reservations (Committee on the Elimination of Discrimination Against Women, 1996). A cursory examination of the reservations of many of these states suggests that they are, perhaps, incompatible with the object and purpose of the convention. Kuwait reserved the right to continue to exclude women from voting. Saudi Arabia ratified the treaty in 2000, but subject to a general reservation that is worth quoting in full: “In case of contradiction between any term of the Convention and the norms of Islamic law, the Kingdom is not under obligation to observe the contradictory terms of the Convention.” Egypt’s reservation to article 16 on equality in marriage was that marriage law that ensured rough equality in outcomes was compatible with the convention, even though the treatment was unequal by gender (Jenefsky, 1991). The wife receives a payment upon divorce, the husband does not. Therefore, divorce could be harder to obtain for the wife than for the husband. Egypt also opted out of articles 2, 9, and 29. Figure 1.4 plots, for various regions and countries, the change over time in an indicator measure of commitment to the international women’s rights regime. The data cover the period 1960–2003 (by then, all African countries except Sudan and Somalia had ratified CEDAW). The indicator takes into account the quality of ratification of CEDAW and is higher for states ratifying without reservations, low for states with reservations that basically state the country will not actually be bound by any provision of the treaty that conflicts with existing domestic law, and intermediate for states with specific but substantive reservations. The more reservations, the lower the strength of ratification. The indicator variable also takes into account compliance with the reporting provisions of the treaty by coding completion of the regular reports that members of the treaty are required to submit to the UN-CEDAW committee. Finally, the commitment indicator is extended backward in time by coding earlier women’s rights treaties (only ratification is coded for these). CEDAW consolidated a smaller women’s rights regime that consisted of the Convention of the Nationality of Married Women, the Convention on Consent and Marriage Minimum Age, and the Convention on the Political Rights of Women. These conventions were opened for ratification during the 1950s and 1960s. For comparative purposes, another indicator of commitment to general human rights treaties is also constructed, by simply adding up instances of ratification over time for each country of ratification of nine treaties and protocols. The treaties are as follows: International Convention on the Elimination of all Forms of Racial Discrimination (CERD); International Covenant on Civil and Political Rights (CCPR) and its two optional protocols; International Covenant on Economic, Social, and Cultural Rights (CESCR); Convention Against Torture and other Cruel, Inhuman, or De-

28

Women and Development in Africa

Figure 1.4 Commitment to Human Rights Treaties, 1960–2010

Source: United Nations (2013).

grading Treatment or Punishment (CAT); and Convention on the Rights of the Child (CRC) and its two optional protocols. Figure 1.4 shows that African countries did not differ substantially from South Asia and the Middle East in ratification of general human rights treaties. For women’s rights, Africa has ratified earlier and more often and with fewer reservations than Middle Eastern and Asian countries. Within Africa, however, there was considerable variation. Sudan, for example, barely adhered to the earlier international regime of women’s rights treaties, while ratifying almost all general human rights treaties and protocols. Ghana and Malawi, meanwhile, were just about average in their ratification and commitment to treaties. Conclusion A successful businesswoman like Alizèta Ouédraogo exemplifies the messy and ultimately ambiguous evaluation of the development process. Every

Exploring Gender and Development

29

person, process, and action in the development process has a particular context and relationship to others. In order to reason more clearly about the development process, it is necessary to define concepts and measure aggregates that capture those concepts. This chapter has introduced some of the standard concepts and measures of development and gender. Subsequent chapters use these definitions and measures to set the context for more detailed exploration of specific aspects of gender in African economies.

2 Women and Development in Africa

Gender equality is a core development objective in its own right. It is also smart economics. Greater gender equality can enhance productivity, improve development outcomes for the next generation, and make institutions more representative. —World Bank, “Gender Equality and Development”

Social observers have long noted that the status of women and overall socio-economic development tend to go hand-in-hand. . . . In the poorest quartile of countries in 1990, only 5% of adult women had any secondary education, one-half of the level for men. In the richest quartile, on the other hand, 51% of adult women had at least some secondary education, 88% of the level for men. Other measures of gender inequality (in health or legal rights) paint a similar picture. In the poorest countries, women are particularly inadequately served in terms of education, health, or legal rights. What to make of this association is not obvious, however. It is possible that income affects gender inequality; that gender inequality affects growth and hence income; or both. Or, it may simply be that common underlying factors determine both income and gender inequality. —David Dollar and Roberta Gatti, Gender Inequality, Income, and Growth: Are Good Times Good for Women?

conomic growth happens when people and organizations (such as firms, churches, and governments) make investments by expending time, energy, and materials to build or invent things that will be durable and useful for producing goods and services in the future. Investments may be in the form of roads, office buildings, outhouses, ideas, new dance moves or singing styles, books, automobiles, computers, and baobab trees. An in-

E

31

32

Women and Development in Africa

vestment is a seed; when grown it yields fruit. By definition, an investment involves delaying consumption. The time, energy, and materials used to build things or refine an idea could have been used to enjoy pleasures of the present. A decision to make an investment depends on expectations about the future. These expectations include expectations of prices of goods and services, costs of making things, technology of production, tax rates, and property rights in ideas, buildings, land, and corporate entities. Taken together, expectations about the future enable someone who is making an investment decision to calculate the expected profitability, or return, to the investment. People and organizations rank investment projects according to expected returns and the riskiness of those returns. They invest in projects with better risk-return profiles. Gendered structures of society may determine expected profitability: if women are under coverture, so that profits from investments are appropriated by their husbands and males in their lineage, they may have limited incentives to invest. While investment depends on expectations, the level of investment also depends on the ability of the entity making an investment to borrow funds to finance the investment. The nature of investments is that people must work in the present and materials must be purchased in the present, while the returns of the investment accrue in the future. Even a lonely inventor trying to figure out a way to remove cotton seeds from the boll of cotton must eat while he or she tinkers. The gap between present expenditure and future returns must be bridged through a loan. The gap must be financed. Someone must be willing to accept a verbal promise or more commonly a piece of paper—a financial instrument—in return for his or her time or materials. The piece of paper credibly promises him or her goods and services in the future. Finance typically comes in the form of bank loans or in the form of selling shares in the profits of the future investment (i.e., selling equity or shares). In rural areas, microfinance organizations are increasingly important in financing small investment projects undertaken by women. The financial system then becomes a critical factor in affecting the level of investment. Around the world, there is considerable variation in the performance of financial systems. Many countries have repressed their financial systems. That is, regulations and barriers block the financial system from financing investments, and instead entrepreneurs must patiently save assets until they have enough to self-finance the investment. Financial systems are often strongly gendered. In many societies, women were under coverture and so precluded from entering into financial transactions. Until very recently, many banks refused to let married women open their own bank accounts. Banks refused to lend to women. Even infor-

Women and Development in Africa

33

mal women’s savings groups frequently had a man as “treasurer” because so few women could read and write. Investment activity also depends on mechanisms that coordinate expectations and enable collective decisionmaking. The profitability of many investments depends on the level of infrastructure provided by larger social entities. Governments, for example, often provide a large array of public goods. These goods are nonrival and nonexcludable: their capacity to provide services to individual users changes little as the volume of users rises, and it is difficult to prevent users from accessing the services. Some public goods are nonrival but excludable, and so are natural monopolies: only one firm will supply electricity or water to a town, because the cost of adding additional users once a network of wires or pipes has been laid is negligible. The profitability of many investments also depends on the complementary investments made by other firms. A developer who plans to build a small office tower in Lagos, Nigeria, knows that her investment will only pay off if businesses in Lagos are planning to expand their offices. Otherwise, commercial rents will fall and many new office towers will be vacant. Investors will undertake more activities when they are confident that others are confident. Since men have dominated the public sphere in so many African societies and governments, that public investment has often favored sectors important to men should come as no surprise. Maternity wards were often a very low priority in public health investments. Child-care centers, which would free recent mothers to work on income-generating activities or outside the home, have been practically nonexistent in rural Africa. In summary, investment depends on expectations of profitability, the financial system, and coordination or collective decisionmaking mechanisms. These determinants of aggregate investment are often gendered. The challenge for government officials who make policy, academics who write about policy, and citizens who have to choose among politicians and ideologies is to discern the set of policies that promote expectations of returns to investment, facilitate the financing of investments, and encourage the coordination and resolution of complementary decisions. The gender dimensions of these policies may be of vital importance. The analysis and choice of policies are also complex because the circumstances of each country differ for reasons of geography, history, and gender institutions. Some countries may have well-established property rights, and few entrepreneurs are concerned about their ability to enjoy the fruits of their investment; other countries have arbitrary and opaque systems for allocating and reallocating property, and so few entrepreneurs want to invest. Some countries in West Africa have thriving female traders who operate extensive financial networks; other countries exclude women from nearly all financial transactions.

34

Women and Development in Africa

The “Economic Method” Applied to Gender and Growth Exploring gender issues in development is an opportunity to master many analytical and empirical tools used in the social sciences. The tools applied in this book are used in all of the social sciences but are most closely associated with economics. Rather presumptuously, some economists call this particular set of analytical tools the “economic method.” This method involves model building and verification. The first step is a narrow, careful, and often mathematical exposition of theories, in the form of models of behavior. These theories typically start with assumptions about individual rather than group behavior. This “methodological individualism” is then coupled with logic to make inferences about the interactions among people. Assumptions, logic, and inference constitute a model. Logic is used to distinguish between causes and effects. Logical statements are often cast as syllogisms. An example that is common in economics goes as follows: People have a propensity to “truck and barter.” People also have different tastes and abilities. Therefore, any equal distribution of resources across society will quickly result in an unequal distribution. Stated another way, people are different, and because people like to trade, any initial equal distribution of resources will lead to situations where people can gain from trade. But trading means the original, equal distribution becomes unequal. The logic is impeccable. It is the often unstated implication of this syllogism that is troublesome: the implication that redistribution of resources is futile, since inequality will inevitably return. Does the logic say that redistribution is futile? No. Does it say that the same inequality will reappear? No. It says that some kind of inequality will likely reappear, but it is silent on the nature of that inequality. It is important to understand that not every implication that an author insinuates does indeed follow from the argument. Students in the social sciences must be alert to model overbite. There are syllogisms that concern gender and economic growth. One goes as follows: African economies are poor because of high levels of corruption; men are more corruptible than women; men dominate African governments; therefore the solution is to encourage and campaign for more representation of women in African governments. For those who think this argument a trifle glib, consider this passage from a book authored by the World Bank (2001, pp. 12–13): “Greater women’s rights and more equal participation in public life by women and men are associated with cleaner business and government and better governance. Where the influence of women in public life is greater, the level of corruption is lower . . . women can be an effective force for rule of law and good governance.” Take another syllogism: relative to men, women prefer that social spending be higher and more oriented toward the well-being of children; more social spending on local infrastructure, schooling, and antipoverty

Women and Development in Africa

35

programs is good for economic growth; thus, empowering women in the political process leads to larger allocations toward growth-enhancing government expenditures. Too obvious? Except for the growth-enhancing part of the story, this is a theory of “big government” in the United States. Lott and Kenny (1999, p. 1163), for example, found that the period of extension of the franchise to women in the United States “coincided with immediate increases in state government expenditures and revenue and more liberal voting patterns for federal representatives, and these effects continued growing over time as more women took advantage of the franchise.” Miller (2008) likewise found that after women obtained suffrage there were sharp increases in public health spending and sizable reductions in child mortality. Edlund and Pande (2002) found evidence that recently divorced women, desirous of higher levels of social spending, have increasingly voted for the Democratic Party in the United States. In the context of developing countries, Chattopadhyay and Duflo (2004) found that when the government of West Bengal required that village leadership positions be reserved for women, village councils indeed invested in different kinds of public goods. In another study from India, Clots-Figueras (2012) found that when female leaders won close elections against a male leader there were subsequent improvements in education attainment, compared with districts where the male leader won. Another link between gender and growth concerns fertility. A major problem in developing countries is that parents want children for old-age security. Child mortality rates are very high, so parents try to have many children. Parents do not take into account that their extra children may generate burdens for society as a whole that may slow economic growth and development. Crowded classrooms and primary care clinics translate to less education and health for all. Empowering women seems to be associated with lower fertility rates. If lower fertility rates generate more rapid economic growth, then the syllogism is complete. A good model tries to explain and predict the essence of some observed behavior. The economic method then applies statistical techniques to data in order to verify or refute models. At various points in the book, basic models used in the social sciences are introduced in nontechnical ways (i.e., no math). Sometimes graphs are used, as in the discussion of the supplyand-demand model used in economics, but more often the discussion can make use of simple numerical examples to illustrate the key concept, as with the Nash equilibrium approach to modeling strategic situations. The book also makes use, at various points, of regression analysis, the statistical technique used by most social scientists to analyze complex questions. The reader will find in the appendix to this chapter a nontechnical and intuitive explanation of regressions and hints about how to interpret regression results. There are many fine introductory textbooks on the subject, and the

36

Women and Development in Africa

more mathematically inclined are urged to seek them out (Angrist and Pischke, 2009; Wooldridge, 2008). It is worth reiterating that the tools used in this book do not belong to economics. John Nash, Nobel laureate and subject of the film A Beautiful Mind, whose work has become very important in the analysis of gender in the household, was a mathematician, not an economist. The approaches that he and others have developed are used in many disciplines. Hopefully, seeing the tools in action here, in an unfamiliar and compelling setting, may help students apply them to more familiar settings. No special training is required to understand the main ideas of these tools. That is why they are so powerful. Models of Gender and Growth There are good reasons to think that gender inequality might be important for economic growth. Gender inequality will affect expectations of profitability of investments across the economy. The performance of financial markets is clearly shaped by prevailing gender relations in African societies. Gendered patterns of political participation and representation affect collective decisionmaking entities, including the small collective entities known as households, where very local public goods are provided to members of a family. The previous section mentioned three models (in the form of syllogisms) that might explain and predict relationships between gender inequality and economic growth. There is a large literature exploring these and similar models (Agénor and Canuto, 2013; Doepke and Tertilt, 2009; Doepke, Tertilt, and Voena, 2012). This chapter cannot do justice to all of these models and settles instead for a modest cataloging in this section of a representative sample of the various models. The models are presented in a straightforward way, with only a few sentences sketching the hypothesized mechanisms at work. Of course, some of these hypotheses can be reformulated to yield the opposite prediction that more gender inequality might be associated with more rapid economic growth (sometimes by subtle changes in assumptions, sometimes by tortuous convolutions of logic . . .). Ultimately, particular hypotheses have to be validated or refuted by analysis of data. The inventory below is far from comprehensive and draws on similar surveys of the literature on gender and growth presented elsewhere (Blackden, Canagarajah, Klasen, and Lawson, 2006; Duflo, 2005). The three syllogisms above may be reframed as follows: • Gender inequality means that men who occupy public office are more likely to abuse the public trust for personal gain, and thus reduce the abili-

Women and Development in Africa

37

ties of government and other institutions to promote investment, growth, and development (Swamy et al., 2001). But see Esarey and Chirillo (2013) for some subtlety about the argument. • Gender inequality means that women, who appear to be more inclined to make public investments that benefit the future of their children, are excluded from the processes of determining these sorts of public and collective investments (Chattopadhyay and Duflo, 2004). • Gender inequality means that women have limited power to make choices between the quality and quantity of their children in the context of the household, which can be thought of as a small collective decisionmaking entity (Lagerlöf, 2003). Since women typically shoulder the costs of bearing and raising children, fertility is higher than would be optimal for more rapid economic growth and development. Another way of formulating this: gender inequality leads to overinvestment in high-risk fertility, because the burden of high rates of infant mortality and maternal mortality falls on women. There are numerous other models of particular links between gender inequality and growth. Here is a partial listing:

• Gender inequality, to the extent that it means women are closer to being under a variant of coverture, leads women to have few incentives to invest for the future. There is then less aggregate investment. • Gender inequality in education (or in the development of human talents through “learning on the job”) means that in aggregate, an economy with more equality in education will grow faster. This holds true if education exhibits diminishing returns, so that extra education yields smaller benefits, and if education is indeed central to innovation and complementary investments (Klasen, 2002). Lopsided investment in the education of boys means that extra, costly years of education yield small marginal benefits. Redirecting those education resources to girls would lead to faster economic growth. • Gender inequality means that men control household and corporate financial decisions, and they may be more inclined to value present consumption over savings and investment, so that economic growth will be slower (Rubalcava, Teruel, and Thomas, 2009). Moreover, long-term growth rates may be reduced as men choose short-term high growth that exhausts natural resources. • On a related note, gender inequality in financial decisions means that men may lean toward portfolios of investments that are poorly performing because of biases (such as holding on to losing investments); that is, in aggregate, funds flow into investment projects that yield low returns, and so economic growth is slower (Barber and Odean, 2001).

38

Women and Development in Africa

• Gender inequality in financial systems means that women are less able to finance profitable investments (Copestake, Bhalotra, and Johnson, 2001; Karlan and Zinman, 2011). • Gender inequality means there is a gap between male and female wages, and also differential labor supply elasticities (women are less responsive to increases in their wages because they are constrained to work at home in what some writers call the “care economy”); these labor market characteristics may mean there is less investment in export-oriented manufacturing, which appears to be associated with economic growth (Seguino, 2000, 2011). • Gender inequality is a broad form of income inequality, and this produces patterns of demand for consumption goods that reduce opportunities for scale economies and specialization, and so lower rates of economic growth (Seguino and Grown, 2006). • Growth depends crucially on the development of new ideas and innovations, and the production of cultural goods (which have high economies of scale), and gender inequality means women are excluded from this important sector of economic activity that depends, crucially, on human creativity that appears to be evenly distributed across men and women (Collins, 1992).

As these brief summaries suggest, there is mounting evidence affirming the very general hypothesis that gender inequality explains some of the variation in economic growth rates across countries. Before reviewing the evidence on the links between gender inequality and growth, some important preliminaries are required. The following sections review some of the important general determinants of economic growth, especially those relevant for African economic development. Following that, the technique of regression analysis is introduced, and then applied to the question of gender inequality and development. Regression analysis is used to estimate the likely importance of gender inequality in explaining some of the variation in income levels and economic growth in the past fifty years. An appendix to this chapter explains regression analysis in intuitive terms as an extension of cross-tabulation. Regression analysis and similar methods are ubiquitous in all the social sciences, and increasingly in business applications, as ways to determine how different factors affect outcomes of interest. Deep Structural Causes of Slow Growth in Africa There are many reasons why societies grow rich or remain poor. In terms of the basic theory presented in the beginning of the chapter, some societies do well in encouraging investors and innovators to increase the capital available for the future, in expanding the volume and terms of financing available, and

Women and Development in Africa

39

in crafting public and collective institutions that coordinate and resolve complementary decisions. Other societies do not perform so well. To explain the poor performance of African countries in comparative perspective, it is useful to distinguish between proximate reasons for growth or stagnation and deeper, structural factors that enable those proximate causes to be operative. A banking crisis disrupts the financial sector of a country and slows economic growth, but deeper, historical factors may have led the financial sector to be vulnerable to crisis in the first place. A large-scale civil war is inimical to growth, as is the destruction of the fisheries sector of an island economy due to a tsunami. But the causes of war and disaster are perhaps more important to explain. What structural factors make countries vulnerable to civil wars and economies vulnerable to destruction from tsunamis? Some countries invest in the institutions of peaceful resolution of conflict or in tsunami early-warning systems and disastermitigation and recovery processes. Others do not. There is no escaping the “laundry list” character of the causes of slow growth. This reflects the likely reality that there are indeed many interrelated causes of slow growth. A laundry list approach also underscores how some causes are cleaner in a social science sense than others. Social scientists often distinguish between causes that are exogenous to outcomes of interest and causes that are endogenous to outcomes of interest. Exogenous factors are ones determined by processes that are not related, in any fundamental way, to the outcome of interest. Endogenous factors may be partially caused by the outcome itself, or caused by other factors that also affect the outcome. For example, a political scientist might argue that the characteristics of anticolonial rebellions in Africa were important determinants of whether states were democratic or not after independence. Rural societies had complex entrenched hereditary hierarchies that did not foster civic virtues of citizenship and participation. Urban society tended to be more plural and rule oriented rather than social-status oriented. So urban anticolonial rebellions might have subsequently reinforced civic virtues, while rural rebellions fostered tendencies toward autocracy. The trouble with attributing this sort of causality to preexisting structure is that perhaps rebellions against colonial authorities snowballed and grew because of preexisting trends in society. Perhaps the colonial authorities encouraged indirect rule by local potentates, strengthening them enough so that when colonial rule weakened, the potentates were strong enough to rebel. Then the subsequent autocracy was not due to the rural rebellion, but rather the colonial policy of strengthening rural rulers. The urban or rural character of rebellion, in other words, is endogenous to social and political trends that may have continued well past the rebellion. Attributing the postrebellion trend to the rebellion itself, when the rebellion was caused by the same trend, would be a mistake.

40

Women and Development in Africa

With the distinction between exogenous and endogenous causal variables in mind, many authors would agree that there are several deep structural causes of relative stagnation in African countries. These deep structural causes are largely exogenous, in that they are immutable features of geography or social aspects of the distant past, not influenced by economic change in the twentieth century. In terms of geography, Sachs and coauthors (Bloom and Sachs, 1998; Gallup and Sachs, 2000; Sachs, 2012; Sachs, Mellinger, and Gallup, 2001) have been the leading proponents of the view that Africa’s low GDP and slow growth are attributable, in large part, to the preponderance of tropical countries and the prevalence of malaria and a number of other debilitating and often fatal diseases in the African tropics. Moreover, Africa’s lack of navigable rivers, and the lengthy distances from fertile and disease-free highland areas to coastlines, limited the extent of regional markets and thus investments in specialization and specialized tools. The tropics are also low in agricultural productivity. Rainfall patterns, soil, and sunlight combine to have less productive potential than temperate zones. Also relevant to the argument of the importance of geography is Diamond’s (1997) hypothesis that agricultural innovations are slow to diffuse when the geography of a continent is oriented on a north-south axis, so that there are many microclimates, as compared with an east-west (and temperate) axis, where innovations can diffuse across a wide swath of territory and population because of similarities of climatic zones. A related argument that gives prominence to geography is that Africa, as the cradle of humanity, appears to exhibit more cultural and linguistic diversity than other world regions. If heterogeneity impedes economic growth, then Africa might be expected to grow more slowly than other, more homogeneous regions of the world (Ahlerup and Olsson, 2012; Ashraf and Galor, 2007, 2011). Africa was subject to massive social disruption during the three centuries of the slave trade (from the 1500s to the 1800s). The trade involved the capture and transport across the Atlantic Ocean, the Sahara Desert, and the Indian Ocean of more than 12 million persons. Certainly, this monumental crime may have slowed investment and innovation (Fage, 1969; Lovejoy, 1989; Nunn, 2007a, 2007b; Rodney, 1972). Nunn and Wantchekon (2011), for example, argued that the slave trade involved much intraethnic and even intrafamily animus, as people tricked and betrayed each other under the perverse incentives of the trade. Strong social norms of distrusting people developed and persisted, hindering growth even in the present. At the end of the nineteenth century, European powers invaded and colonized African societies and then plundered those societies by appropriating fertile lands (as in Kenya) and minerals (as in Zambia and South Africa), and by using force and terror to compel African subjects to produce

Women and Development in Africa

41

crops for export (as with rubber plantations in the Congo, and cotton and groundnuts in some parts of West Africa). There has yet been no acceptable quantification of the value of the plundering. It is also difficult to determine how that plundering affected the livelihoods of those remaining over the years. Some cases are easier than others. For example, the expropriation of land by British in Kenya and forced relocation of Kikuyu (especially) farmers onto less fertile land clearly reduced economic growth for those populations. But the forced expropriation did not happen in a vacuum: the European settlers also brought with them many new farming techniques, organizational innovations, and infrastructure for exporting farm produce. For some of the displaced, the combination of effects may have made them better off: Wangari Muta Maathai, for example, the winner of the Nobel Peace Prize, describes in her 2006 biography how her father was the manager of a white settler farm and clearly achieved a level of prosperity that he might not have had as a small-scale farmer without the Europeans. Consider the case of cotton and groundnut exports in West Africa. While forced labor was used occasionally, it does not seem to have been overwhelming, and more often small farmers in West Africa were induced to plant cotton and groundnuts because the prices of the commodities made them very profitable. In these cases, there is no obvious reason why the introduction and spread of lucrative crops by the colonial authorities would lead to lower growth. Some argue that the new crops created a sort of trap. That is, while the crops were profitable in the short term, over the longer term they systemically led to a reduction in the profitability of future investments. One might imagine that as African farmers specialized in these crops, they lost the ability to innovate somehow, so that as the prices of cotton and groundnuts fell relative to costs (so that they became less profitable) overall options of the typical farmer were reduced. This seems implausible. When asking about the welfare impacts of complex events, such as colonialism, or introduction of cash crops (primarily grown for sale to the world market), one must devote time to considering the right counterfactual argument. Would someone like Wangari Maathai’s father have had independent access to new farming techniques had the British not settled in Kenya? Counterfactual scenarios—the what-might-have-been questions— are foggy notions. More apparent is that the colonial legacy disrupted the trajectory of societies to develop political institutions that promoted economic growth. In the contemporary world, polities are organized as states that monopolize violence within their borders, yet do not have to resort to violence because they are legitimate. States earn loyalty and acceptance when state actors respect self-imposed, constitutional rules of procedure. Citizens of states come to respect borders when these have emerged from long histories of

42

Women and Development in Africa

compromise. Effective, legitimate states then become important enablers of development and economic growth by investing in public goods, enforcing clarity in property rights, coordinating complementarities, and supervising financial systems. The colonial powers made it very difficult for African states to be successful after independence (Platteau, 2009). The arbitrary borders created and bequeathed by European powers bore little relation to the locally available infrastructure to “project” power over the entire state territory (Herbst, 2000). The colonial powers had at their disposal the enormous resources of the metropole, the colonial army, and an ability to coordinate with neighboring states often controlled by the same empire. After independence the resources and possibilities for coordination were gone. This meant that huge swaths of territory lay effectively outside of the states’ control; people in these hinterlands were subject to arbitrary policies, or even no policies. They had little ability or opportunity to act or be treated as citizens (Mamdani, 1996). African states have been consumed by the low-intensity rebellions and civil wars that resulted from this problem. Few cities or other potential growth areas have been able to free themselves from the insecurity of disruption. Recent evidence supports this view of the problem of borders, legitimacy, and ability to project power and so establish the rule of law. African boundaries contain more straight lines, more landlocked countries, and more partitioned ethnic groups than other borders on other continents. The consequences of this appear to be beyond doubt: poorer governance, more conflicts, and slower economic growth (Alesina, Easterly, and Matuszeski, 2011; Englebert, Tarango, and Carter, 2002; Michalopoulos and Papaioannou, 2011). The inability to control “the hinterlands” exacerbated the problem of unconstitutional rule. While newly independent states were left with paper constitutions, these reflected no real compromise among powerful entities. There was no constitution-making in the sense of different and important actors in society sitting down, bargaining, and agreeing on “rules of the game” for how political power and legal authority were to be acquired and exercised. The colonial regimes quite often had delegated important powers to “indirect rulers.” These chiefs and priests were given broad leeway to exercise authority, and their power could not be questioned by their constituents. (The colonial rulers, of course, kept the prerogative of appointing and dismissing these indirect rulers “at will.”) The colonial constitutions bequeathed to the new states typically gave no role to these indirect rulers. Chiefs and priests, powerful in “native” society, were excluded, and the Muslim societies of Sahelian Africa, with their extensive “brotherhood” organizations, were marginalized. What was worse, armies were created by the colonial authorities to

Women and Development in Africa

43

suppress domestic anticolonial rebellions and to aid the metropole in times of international conflict (hundreds of thousands of Africans fought in Europe during the two world wars). As long as soldiers were integrated into the broader colonial army, there was little opportunity for them to seize power. Once the colonial power was gone, however, officers quickly realized that new civilian elites were incapable of exercising control over the military. These civilians had never exercised power before, and they had few institutional ties to the armies. Coup d’état followed coup d’état, and the rise of tin-pot dictators and tyrants followed inexorably. As always, an exception demonstrates the reasonableness of the hypothesis that a legacy of colonialism was ineffective and illegitimate states across the continent. Botswana is the sole exception to the dismal economic performance of African countries. GDP and per capita income have been growing at a very rapid rate for almost fifty years. In fact, Botswana by some measures has had the highest growth rate in the world. Because of its exceptional growth, Botswana has attracted a lot of scholarly attention. No smoking gun has been discovered to solve the mystery. Reality turns out to be prosaic. Legitimate and effective political institutions capably managed a natural resource windfall. Botswana at independence in 1960 was very poor, certainly just as poor as any other African country. Fantastic diamond wealth was discovered, and that wealth was not squandered but rather wisely invested in infrastructure for cattle (the main economic activity of the population) and schooling. Macroeconomic policies have been good. There has been little inflation, and the central government has done a good job controlling fiscal deficits. The exchange rate has been kept in line with purchasing power in the major trading partners (chiefly by being kept formally pegged to the South African rand, by far the largest market for Botswana’s imports and exports). State-run enterprises, usually leading indicators of disastrous government policies of patronage, have been profitable and relatively uncorrupt. Policies were reasonable, it seems, because Botswana at independence maintained a set of political institutions that were widely perceived as legitimate. The political actors in those “indigenous” institutions did not have to compete with a new class of political actors to occupy the vacant state apparatus left by departing colonial administrators. Samatar (1999) and Acemoglu, Johnson, and Robinson (2003) argued that a constellation of good initial conditions enabled subsequent success. The most important of these initial conditions included the following: (1) an alignment of interests between political and economic elites (as Marx and Engels put it long ago, the state was simply a committee for managing the affairs of the bourgeoisie); (2) a lack of indirect rule, which meant that no separate dual authority was created; (3) strong local institutions of accountability persisted; (4) a single economic activity (cattle) meant that different groups could not be ex-

44

Women and Development in Africa

ploited for the benefit of other groups; and (5) ethnic difference was not as salient as in other countries. There are other deep structural causes for slow growth in Africa that predated the slave trade and colonialism. Many people in Africa are quick to point to what they call “African traditions” as sources of slow growth. One is a social structure that emphasizes strong solidarity among members of the same kin group. This solidarity, some argue, is often manifested as a kind of forced sharing. Confiscation of income by kin functions as a tax, then, and dampens incentives for investment (Haagsma and Mouche, 2012; Hoff and Sen, 2006; Platteau, 2009). Some authors see the enduring prevalence of witchcraft as a social mechanism for enforcement of norms of sharing (Van Binsbergen, 2001). Another so-called African tradition is excessive deference to authority, or respect for hierarchy. Finally, Africans are sometimes considered to be overly present-biased, and so discount the future benefits of investments. These various hypotheses have not been subject to much investigation and so should be seen as purely speculative. Furthermore, there are likely insurmountable methodological obstacles to the task of determining what part of these possible impediments to growth was influenced by the slave trade and colonialism (which may have induced present-bias and slavish respect for authority) and what part existed prior to extensive contact with Europe, the Middle East, and South Asia. Proximate Causes of Variation in Growth There is no great mystery to the proximate causes of sluggish growth in Africa. Conflicts have become more frequent, and economic policy was terrible for many decades, as regimes in control of the state practiced patronage and corruption, rewarding supporters and punishing opponents. The number and intensity of civil conflicts that have occurred in the continent since independence of most countries in the 1960s are tragic. The more important conflicts have been among the worst the world has seen since World War II. Civil conflict in Sudan has now lasted almost six decades and continues as low-level conflict in the south (despite the official secession of South Sudan) and in Darfur. Millions have died from the war and resulting destruction of medical and food infrastructure (Burr and Collins, 1995; Collins, 1983; Flint and De Waal, 2008). Somalia has been a “failed state” since the fall of the dictator Siad Barre in 1991. Horrific pictures of children with arms hacked off by rebel soldiers, living warnings of future terror, were staples of the decades-long civil conflicts of Liberia and Sierra Leone. Until the death of Jonas Savimbi in 2002, his rebel movement, the National Union for the Total Independence of Angola (UNITA), looked set to prolong indefinitely the civil war in Angola, where cities of tens of thousands of people had been besieged into starvation. Most re-

Women and Development in Africa

45

cently, much of Mali lay in the hands of Islamist and Taureg separatists, after the Malian army abandoned the fight and instead toppled the civilian president. It took a French military intervention in 2013 to restore territorial integrity. Rebels in Central African Republic drove within 100 kilometers of the capital city in 2013 before reaching a power-sharing agreement, which was promptly abrogated, and the president ousted. And in Democratic Republic of Congo, new rebel movements have regularly sprouted up along the increasingly unstable Rwanda border. On economic policy, African policymakers performed very poorly until the early 1990s (Collier and Gunning, 1999; Mohan et al., 2000; Sender, 1999; Sender and Smith, 1985; Tarp, 1993). During the 1970s, African polities had poor policy responses to the oil price increases, global stagflation, higher interest rate charges on foreign debt, and worsening commodity terms of trade. Oil importers went into long declines, often caused by macroeconomic mismanagement as they tried to use exchange rate controls and foreign aid to avoid the necessary adjustments to the higher price of oil and stagflation of the industrial world. The major African oil exporters, on the other hand, succumbed to the “Dutch Disease,” where the large influx of oil revenues quickly led to macroeconomic problems. Successive regimes in Nigeria, Africa’s most populous country, especially could not resist recycling oil revenues into poorly planned construction projects. The construction boom drove up the price of labor, encouraging rural to urban migration, and exacerbating a rapid decline in the agricultural sector. Corruption and rent-seeking followed, and Nigeria has seen economic decline for several decades. Many African countries have been caught in a vicious trap of economic decline leading to political crises and rebellion, leading to further economic decline. Poverty fuels conflict, especially when riches are within grasp. A number of countries have lucrative mining and oil enclaves. Whoever controls those enclaves can obtain considerable wealth. It is clear that this is a defining feature of many conflicts. Liberia and Sierra Leone have diamonds easily mined on the surface. Sudan has oil. Angola has a terrifying combination of oil and diamonds. When peasant populations are impoverished, it seems possible to mobilize large numbers of disillusioned young men with the promise of better lives through looting and conquest. The countryside provides recruiting grounds for rebel movements that aim to capture riches, all the while further impoverishing the countryside and setting in motion additional rounds of violence. Poor policy choices were rooted in struggles to obtain and maintain political power. Political scientists invented or resurrected all sorts of terms to describe the incentives for bad policies: rentier states, prebendal politics, neopatrimonialism, etc. Shortly after independence from colonial powers in the 1960s, most African states were taken over by military coups or single-

46

Women and Development in Africa

party strongmen. In much of Africa, the “night watchman” state of Adam Smith, ensuring peace and freedom for individuals to enter into contracts for the use of their property, mutated into a warlord state, an “open sore” on the African continent (Soyinka, 1996). This outcome was not necessary; in some parts of the world dictators have fostered decades of rapid economic growth. There was something different about African dictators during the 1970–1995 period, on the whole. Tyrants and tin-pot dictators had only a shaky hold on power; therefore, they were induced to plunder and distribute patronage. Plunder sustained and even improved the dictator’s position relative to his potential political rivals. The dictator got richer and more powerful, and everyone else got poorer. Policies that promoted economic growth might have made the dictator even better off, but they also would have made his rivals better off. Since control over political power depends on both absolute and relative standing of opponents, keeping them as far down as possible, and as far away from the dictator as possible, was likely the preferred course of action (Acemoglu and Robinson, 2000a; Acemoglu and Robinson, 2000b). Given the prevalence of conflict and poor economic policies, perhaps it is no wonder that African economies stagnated for the 1960–2000 period. It should be noted that on many development indicators such as schooling and mortality, African countries improved steadily even as GDP stagnated (Kenny, 2005; Sender, 1999). It should also be noted that the politics of the Cold War, Françafrique, and post–Cold War meant that the United States, France, the UK, and the Soviet Union (and international organizations such as the World Bank and the International Monetary Fund) aided and abetted conflict, corruption, and poor policies (Berger et al., 2012; Easterly, 2005). Economic growth from 2000 to 2010 was largely due to rapid growth in oil and mineral exports (Beny and Cook, 2009). Social scientists have been divided over whether the growth will exacerbate the “natural resource curse” or whether the post–Cold War political transitions of many African polities will lead to better governance and hence longer-term economic growth. Growth generated by the opening of trade in natural resources might be highly concentrated in the hands of a small elite, which then ignores the imperative of more equitable distribution and prevents the broader population from reaping any reward from their greater potential (Mehlum, Moene, and Torvik, 2006; Robinson, Torvik, and Verdier, 2006). Other changes in the global economy might also reinforce rather than remedy bad institutions. New technologies can be used for repression and predation, and not just production. Foreign aid can divert critical human talent away from investments with positive value and toward investments that reward corruption and bureaucratic aggrandizement. Globalization of technology can facilitate the lack of accountability of dictators by enabling them to control new opportunities for plunder and squirrel their booty

Women and Development in Africa

47

away in Switzerland. Old theories of growth assumed benevolent polities, even a benign neglect by people with the power to control societies through the use of violence. New theories recognize what Tornell and Lane (1999) called a “voracity effect,” in which windfall economic gains from technology or assistance could exacerbate predatory politics, making the country worse off. Regression Analysis of Gender Inequality and Growth In addressing the issue of whether and how gender inequality matters for growth, it is useful to imagine an experiment. Suppose that the Bill and Melinda Gates Foundation (the world’s largest foundation) secretly and randomly selected leaders of fifty countries and induced them, perhaps through promises of awards of $100 million to their families, to influence their countries to immediately change their gender policies to reverse all prior discrimination against women and proactively ensure that gender outcomes were equalized in a short period of time. Assuming such a secret agreement were feasible, and the leader’s newfound wealth did not itself affect economic growth (perhaps it would have to be kept in a European bank account for twenty years), one might then compare the outcomes between those induced to change and those that were not so induced. Such an experiment is of course unlikely. Rwandan president Paul Kagame, it should be said, has been exceptionally influential in promoting women in government, to the point that in 2010 a majority of members of the Rwandan parliament, ministers, judges, and certain other government positions were women. In the real world, leaders and societies design new structures of gender both to deliberately instigate change and to ratify and affirm evolving practices and understandings. Because gender equality both causes change and is caused by change, it is difficult to identify or estimate the magnitude of the causal effects of changes in gender equality. Gender equality generally is responding to other changes occurring in societies. These changes affect both gender equality and economic growth. Suppose, though, that the endogeneity of gender equality and the joint causation of growth and gender equality were ignored. The effects of gender equality could be estimated by running a regression explaining the variation in economic growth rates across countries. Variables measuring gender equality might explain part of the variation in growth rates. A regression enables the analyst to determine whether and to what degree a certain factor influences an outcome. The regression method measures the effect of interest by controlling for the effects of other explanatory factors. The appendix to this chapter explains regression analysis in more detail, emphasizing the intuition of how the technique “controls for” other variables and measures the effects of a single variable.

48

Women and Development in Africa

In this case of estimating the effects of gender equality on growth, the regression is slightly more complicated. Each country is observed repeatedly over the decades. These cases are not independent of each other, since each observation is the same country, so they have to be treated slightly differently than if they were like completely different entities. A set of advanced techniques, which will not be discussed here, can run regressions for this kind of data. For each country, the growth rate over a decade may be calculated. This is the outcome variable, and there are five decadal growth periods since 1960 for each country: 1960–1969, 1970–1979, 1980–1989, 1990– 1999, 2000–2009. With the ninety or so countries of the world for which data are available, this yields about 450 observations. This kind of regression is called a cross-country panel regression. The idea is to estimate the effects of different levels of the explanatory variables at the beginning of the decade (women’s rights, education inequality) on growth of GDP per capita for the decade that follows. Early efforts along these lines were featured prominently in the World Bank’s policy research report, Engendering Development (2001), and include the regression analysis of Dollar and Gatti (1999) and Klasen (1999). Unfortunately, as noted above and pointed out by numerous critics, this sort of cross-country regression is seriously flawed (Bandiera and Natraj, 2013). First, gender inequality and economic growth are likely mutually causative, so gender inequality is not exogenous. The coefficient estimated for gender inequality might reflect how economic growth affects inequality, rather than the other way around. Also possible is that gender equality and economic growth are jointly caused by a third variable, such as public health or public infrastructure. Second, most papers measure gender inequality using indicators of outcomes of gender inequality, rather than directly measuring the underlying gendered norms, institutions, and discourses that generate those outcomes. Since gendered outcomes may emerge from complex interactions of gendered institutions, finding correlations between inequality in outcomes and economic growth helps little in understanding the mechanisms that propel the relationship between inequality and growth. Bandiera and Natraj point to the example of slow economic growth associated with low female labor-force participation, which may be due to norms against work outside of the home for married women, or due to norms against schooling for young girls. The two norms might lead to the same outcome—low labor-force participation—but for different reasons and thus with different implications for policy. Blithely disregarding valid criticism is a bad habit that social scientists develop early in their careers. We follow suit. Table 2.1 presents results of regressions where an indicator of gender inequality explains GDP growth per capita. The measure of inequality is an index that summarizes the coun-

Women and Development in Africa

49

try’s commitment to early women’s rights treaties, to the Convention for the Elimination of Discrimination Against Women (CEDAW), and the strength of commitment to the gender rights regime put in place by the CEDAW (completing reports, abiding by protocols, not putting in reservations to articles of the treaty). The higher the index, the more the country is a “good world citizen” when it comes to gender rights. Columns (1) and (5) present the results for the entire sample of 138 countries around the world, while columns (4) and (8) present the results only including 44 countries in sub-Saharan Africa. Table 2.1 also estimates the same regression with two other explanatory variables. One measures the extent of “democraticness” of the country’s polity, as constructed by a major research initiative in political science, the POLITY IV project, that codes every country in the world on a scale of –10 (very autocratic) to 10 (very democratic). The results are in columns (2) and (6). The second is a general indicator of respect for human rights, not just women’s equality, measured by commitment to general international human rights treaties. The results are in columns (3) and (7). The dependent variable is the growth in real GDP per capita over the five- or ten-year interval (where 1960 is the starting year). In addition to the principal explanatory variables of gender inequality, democracy, and human rights, the regressions include a number of control variables. One is the initial level of real GDP per capita for each time period. Poorer countries might be expected to have higher growth rates than wealthier countries. They are expected to converge. Other controls are dummy variables for all the countries and the time periods. The regressions are “fixed effects” regressions. The coefficients for the index of commitment to women’s rights are statistically significant in both the five-year and ten-year panel. The coefficients for the degree of democratic governance are significant in the fiveyear panel but not in the ten-year panel. The coefficients on the index of commitment to general human rights are significant for both the five- and ten-year panels. The magnitudes of the coefficients are not large. If the explanatory variables were exogenous, the coefficients would represent the effects of a one-unit increase in the indices on the GDP growth rate. But a one-unit increase would be enormous; as Table 2.1 shows, the standard deviations of the indices are more in the .10 range. A two–standard deviation increase (of .20) would then lead to a change in GDP growth only one-fifth the estimated size of the coefficients (that is, around .006 for the case of the index of commitment to women’s rights in the five-year panel). So a very large increase in gender equality would only lead to a small increase in GDP growth. Of course, the index of commitment to women’s rights may not be exogenous; increases in GDP may lead government leaders to enact greater women’s rights. In that case, the causality would be reversed. There is, unfortunately, no satisfactory instrumental variable available that would

Table 2.1 Results of Regressions Explaining Growth, 1960–2009 (panels with 5-year and 10-year intervals)

Index of commitment to women’s rights

Degree of democratic governance

Index of commitment to human rights

Constant

Number of countries

Observations

Mean and Standard Deviation (1990) 0.12 (0.08) 0.01 (0.08)

0.03 (0.02)

Dependent Variable

Growth in GDP per Capita 5-year Intervals (1) (2) (3) (4) 0.03c (1.81)

0.04b (2.03)

–0.07 (–1.37)

0.14b (1.99)

Dependent Variable

Growth in GDP per Capita 10-year Intervals (5) (6) (7) (8) 0.04b (2.21)

0.02 (1.05)

0.03a (10.57)

0.03a (10.31)

0.03a (10.64)

0.02a (3.02)

0.03a (9.87)

0.03a (9.05)

1,241

1,013

1,241

366

1,241

1,013

138

137

138

44

138

137

–0.06 (–1.60)

0.13c (1.89)

0.03a (9.88)

0.02a (2.87)

1,241

366

138

44

Sources: World Bank (2013); United Nations (2013); Multilateral Treaties Deposited with the Secretary-General; Polity IV Data Set, www.systemicpeace.org/polity /polity4.htm. Notes: The mean growth rate in real per capita GDP over the entire sample period, whether 5-year intervals or 10-year intervals, was 2% for the entire sample, and 1% for the African countries. The regressions include fixed country and year effects and the initial level of GDP per capita for each time period for each country. Coefficients for these variables are not shown. The initial GDP per capita is negative (indicating convergence) but very small in size. Columns (4) and (8) restrict the sample to African countries. Absolute value of z-statistics in parentheses, errors clustered by country. a. significant at 1% level; b. significant at 5% level; c. significant at 10% level.

Women and Development in Africa

51

explain changes in women’s rights and not changes in GDP at the same time. The regression makes clear the limitations of attempting to infer, at the cross-country scale, how gender inequality affects GDP growth. Conclusion The regression analysis is hardly dispositive about the relationship between gender inequality and growth. As the remaining chapters in the book show, though, much microlevel evidence suggests that a causal relationship is very likely. This raises an interesting question: If gender inequality has been important for growth, why did so few early practitioners of the social sciences emphasize its importance? Ibn Khaldun, writing in North Africa in the 1300s, was an early, great theorist of the rise and fall of civilizations. Gender relations, apparently, played no role in his theories. Adam Smith, writing in Scotland in the 1700s, pointed out that “peace, easy taxes, and a tolerable administration of justice” was the basic recipe for rapid economic growth. But he had little to say about gender and economic growth; women were barely mentioned in The Wealth of Nations. To be fair, it appears that Smith was preparing a treatise that would have included an exegesis of the changing status of women and how economic growth would diminish the oppression found in so many societies, including his own Scotland in the eighteenth century (Dimand, Forget, and Nyland, 2004). If the greatest social scientists before the twentieth century ignored gender, perhaps it was because social scientists had their own gender-blinders on and could not see what in retrospect seems self-evident. Perhaps there was presumed to be little variation in gender inequality—women being oppressed everywhere—and so no reason to suppose it explained variation in growth. Finally, perhaps gender equality has only mattered for growth when a society has moved into a postindustrial era, when differences in physique matter little for economic activity, and fertility has fallen to a level where women have choices to make besides child rearing. Whatever the reasons for the previous neglect, gender now figures centrally in theories of economic growth. Appendix: Introduction to Regression Analysis Suppose that a researcher has measured the hourly wage rates earned by men and women working in urban Lusaka. On computing the average wage rates for the entire sample, the researcher notices that the rates for women are only half the rates for men. She reports this finding to the local newspaper, which promptly runs the screaming headline: “Discrimination Against Women in Lusaka Labor Markets!” The article recounts several anecdotes of women working in the same company receiving lower wages than men

52

Women and Development in Africa

and then suggests the statistical evidence confirms the presence of discrimination. Does it? Regression analysis is one of the social sciences’ most important tools for answering this and other basic questions. The problem with the researcher’s interpretation should be readily apparent. She may have jumped the gun in her eagerness to confirm her gut feeling that employers discriminated against women. She neglected to consider that the disparity in wages might be due to disparities in qualifications and experience rather than discrimination. Consider, for example, the possibility that discrimination did exist in the past but has disappeared in recent years. Older workers, with more experience and seniority and high wages, are all men, and younger workers are evenly divided between men and women, earning comparable wages. The researcher’s sample, however, contains a mix of old and young women, so that men, on average, appear to earn higher wages. Only the older men do, in fact. Regression analysis is the term used to encompass a variety of statistical techniques for analyzing and explaining variation in some outcome of interest. Regression analysis makes use of data. A data set consists of observations of different variables for a number of cases. A case may be a person, household, firm, region, or country at a certain point in time. A data set may consist of the education, age, and wage of a group of persons, or the GDP, interest rate, infant mortality rate, and average education level of a group of countries. In some ways, a regression may be thought of as a more elaborate form of cross-tabulation. A crosstab is a table where the average of an outcome variable is computed separately for different categories of one or more variables. The categorical variables are typically assumed to be exogenous. A crosstab might show the average growth rates of GDP per capita for African countries and non-African countries, further categorized by whether the states are viewed by their citizens as legitimate or not (determined by some measure concocted by political scientists). The crosstab might show that African countries with legitimate states had rates of economic growth just as high as legitimate states elsewhere, and that illegitimate states everywhere had lower rates of growth. This might be strong evidence that legitimacy of the state mattered for economic growth. In the Lusaka wage rate case, the relevant crosstab would have average wage rates computed for younger men and women and older men and women. The table would have four cells, with the cell for older women perhaps empty because there were no observations. The crosstab might make clear that the wages of older men were higher than the wages of younger workers, and that the wages of young men and women were roughly even. The crosstab might then suggest no contemporary discrimination. A crosstab controls for a factor (age or experience in the labor force)

Women and Development in Africa

53

that might explain some outcome (the wage rate). A regression, generally speaking, is a technique for controlling for certain factors when explaining the correlation between another factor and some outcome. In assessing the proposition that men earn more than women, the age or experience of men in the sample of workers needs to be controlled for. The essence of the regression method is to posit a linear relationship between the explanatory variables and the outcome variable and then calculate the “partial” correlations between each variable of interest and the outcome variable. These partial correlations are known as the regression coefficients, and they tell us how much a change in one variable, holding constant the other variables, changes the outcome variable (the ceteris paribus assumption of much economic reasoning). The workhorse method for calculating these coefficients is known as the method of ordinary least squares (OLS). An OLS regression is a method for estimating the parameters (β) of a linear relationship where one or more exogenous or independent variables (x) in the data set are explaining or causing an endogenous outcome variable (y), sometimes called the dependent variable. The method is to find values for the parameters that minimize the sum of squared deviations of the predicted value of the dependent variable yˆ from the actual value of y, using the relationship ˆy = βˆx. Suppose the relationship between some explanatory variables x1, x2, x3, and x4 and a dependent variable y were assumed to be linear. For every case, there is observed a set of x’s and a y. Every observation, moreover, is assumed to be observed or recorded with some small error. Imagine a term εi that measures the deviation of the actual observation for the case from what would have been predicted. This error term is assumed to be randomly distributed across the cases. For every case (identified by the subscript i), then, the following relationship holds: yi = β0 + β1x1i + β2x2i + β3x3i + β4x4i + εi.

The OLS method involves selecting values for the coefficients (β0 to β4) that minimize the following expression: Σi(yi – β0 – β1x1i – β2x2i – β3x3i – β4x4i)2,

the sum of the squared deviations. Suppose there were just one explanatory variable. There would be a scatter of points that are observed in the data. Each point would represent an observation, such as country or worker, and the levels of x and y associated with that observation. These might be wage rate and age of a worker, or the legitimacy of the state and the growth rate of GDP for that state. What line would best fit these data? One criterion for determining the best

54

Women and Development in Africa

fit is the least squares criterion. The line should minimize the sum of squared deviations of the actual value of y from that predicted by the regression line. That deviation is also known as the residual, so the method of least squares minimizes the sum of squared residuals. Fortunately for the student, any computer can now estimate a regression line using either statistical software (SPSS, Stata, SAS, or the open source statistical software R) or spreadsheet software (Excel). The data are entered in columns, with one column for each variable. Each row is an observation (e.g., a worker or country). The software produces the estimated values of the coefficients that minimize the sum of squared residuals. Informally, this is called “running” a regression, as in “I just ran a million regressions!” The first coefficient β0 is sometimes called the constant term or intercept term and is not usually of interest. The other coefficients are the partial correlations and are usually called the slope coefficients, or sometimes the estimated coefficients. The coefficients of continuous variables are interpretable as partial derivatives, using the language of calculus, indicating how much one variable changes when another variable changes. These partial derivatives have a special symbol, ∂, that is found every so often in this book. The expression ∂y/∂x stands for “how much y changes when x changes by one unit.” This is the estimated coefficient in a regression with y as the dependent variable and x as an independent variable. The coefficients are interpreted as marginal effects: how much the y variable might be expected to change with a one-unit change in the x variable, holding all other variables constant. Most data sets are actual samples of a population or are interpreted as samples (the set of countries that exist in the world is interpreted as a sample of the countries that have or might have existed). Moreover, the variables included in a data set are generally measured with error. For these reasons, regression analysis is an exercise in inference. One has imperfect information about the real world, and one would like to infer something about the real world from the sample of imperfectly observed variables. What determines how much confidence to have in the estimates of the coefficients? Statistical assumptions and probability theory are used to evaluate numbers called t-statistics (or z-statistics). These are reported by statistical software. There is one t-statistic for each estimated coefficient. The t-statistic can be used as an indicator of statistical confidence in the estimate of the coefficient. The general rule of thumb is that for an estimate to be worth talking about, the t-statistic should be greater than 1.96 in absolute value. The regression method will of necessity produce an estimate of every coefficient—that is, how much each explanatory variable affects the outcome. But some of these estimates will basically be arbitrary. Suppose there were only two cases observed. The line connecting the two would then represent the slope, or the estimated coefficient. The analyst would never want

Women and Development in Africa

55

to infer a population relationship, however, from a sample of two! The tstatistic on most coefficients will be very low when the coefficients are estimated with small samples. The t-statistic will also be less than 1.96 when there is little variation in the explanatory variable. For example, if we wanted to explain the frequency of indigestion in a population and included the number of eggs eaten per day, the regression software would generate an estimate of the coefficient. But this coefficient is likely to be meaningless. We are pretty sure that nobody in our sample eats enough eggs to get indigestion (at least compared with other sources of stomach indigestion). There is just not much variation in egg eating (presumably because people refrain from eating too many eggs!). The t-statistic for a coefficient increases when there is more variation in the explanatory variable whose marginal effect is estimated by the coefficient. The t-statistic for a coefficient will also depend on the overall explanatory power of the regression, and this is conventionally measured by the R2 statistic. This statistic indicates the fraction of the overall variation in the outcome variable that is explained by the variables included in the regression. The R2 statistic varies from a low of zero to a high of one. The better the regression overall, the more confidence one can have that the estimated coefficient on a particular variable is indeed worth considering. There is no rule of thumb about how high the R2 statistic should be. In time series data, where the observations are for different years or months, it is very common to have R2 statistics above .90. For cross-sectional data, such as samples of households, the R2 statistic may be as low as .05 and the regression still considered informative. Finally, the t-statistic’s value depends on the correlation of the explanatory variable with the other explanatory variables. If there were many explanatory variables that were closely correlated, and these were included in a regression, there would be no reason to have confidence in the estimated coefficient for any particular variable. If the variables were positively correlated, one should not think of an increase in one variable as having an effect on the outcome, controlling for the other variables, because the other variables always increase with increases in the variable under consideration. Two special kinds of variables may be included in regression analyses and are often important variables of concern to social scientists. One kind of variable is known as a dummy variable. This is a variable that takes on values of zero or one, according to some criterion. Zero could stand for men and one for women. Zero could stand for urban and one for rural. Combinations of dummy variables could represent more complex categories. A first dummy variable could take on value one for urban men and zero otherwise. A second could take on value one for urban women and zero otherwise. A third dummy variable could take on value one for rural men and zero otherwise.

56

Women and Development in Africa

There is no need to include a fourth dummy variable for rural women in the regression. Indeed, including such a fourth variable would make it impossible to compute the regression coefficients. The four dummy variables together perfectly explain the variation in wage rates. The deviations of observed outcomes from the predicted outcomes will add up to exactly zero if the coefficients on the dummy variables are simply the mean values of the outcome for those categories and the coefficients for other variables are zero. Therefore, the rule is to omit one of the dummy variables when these are grouping the data into different categories. The regression coefficient estimated for a dummy variable may be interpreted as the difference between the expected value of an observation in that category and the expected value of an observation from the excluded or omitted category, when other effects from other variables are controlled for. Returning to the example of wage rates and possible discrimination, it should now be clear that the coefficient of interest in a regression equation explaining variation in wage rates is the coefficient on a gender dummy variable, where the regression also includes a variable measuring age. The other special kind of variable is a composite variable created by multiplying one variable by a dummy variable. This is known as an interacted variable, and the coefficient measures how the partial correlation between the outcome and the continuous explanatory variable differs from one category to another. The categories are captured by the dummy variable. So, for instance, a regression equation may be: yi = β0 + β1x1i + β2x2i + β3Di + β4Dix2i + εi.

Now, β2 measures the effect of x2 on y for observations that fall into the zero category of the dummy variable, while β2 + β4 measures the effect of x2 on y for observations that fall into the other category of the dummy variable. In explaining differing wage rates, interaction terms could help control for possibly differing returns to age and experience for men and for women. An example may recapitulate and reinforce why statistical analysis is important. Suppose that a researcher has measured the incomes of men and women in households in a large number of villages in Zambia, and then also recorded the nutritional status of the boys and girls living in the household. The researcher categorizes the households into three groups—rich, middle, and poor—according to the total household income per person in the household. Then the researcher computes the average levels of nutritional status for each income group. It turns out that the nutritional status of children is actually lower in the richer groups. How could this finding be explained? It may be that the averages are being calculated without controlling for some other variable. For instance, it may be that in higher-income

Women and Development in Africa

57

households most of the income is earned by men. Women married to highearning men, say, most often do not work outside the home. When men earn most of the income in a household, they then get a disproportionate share of the power to make decisions in the household. Men do not care to spend income on children, perhaps, so the higher-income family spends less on the nutrition of children. The point is that OLS regression analysis is needed to control for both the level of overall household income and the share of that income controlled by the man and the share controlled by the woman. Only after relevant variables have been controlled for can the effects of changes in one variable be assessed. No discussion of regression analysis is complete without raising the issue of causality. If there is a strong correlation between two variables, a regression will estimate a statistically significant and sizable coefficient. But this correlation may not indicate a causal relationship. There are three other possibilities. First, both variables could be affected by a third variable that is omitted from the regression and so cannot be controlled for in the regression. Second, the causality could be reversed; the outcome variable in the regression might really be the causal variable. Third, there is often a feedback loop between explanatory and outcome variables. So x causes y, but y also causes x. The results of a regression, coefficients, t-statistics, and R2 cannot by themselves distinguish among these possibilities. Basically, only good reasoning can distinguish among them. The social scientist, when confronted with the results of a regression, must ask whether it is strongly likely or only weakly plausible that the hypothesized relationship between explanatory variable and outcome is actually causal. In some cases, the answer is easy enough to determine. Only a few moments of reflection are needed to see one of the three other possibilities. For example, the correlation between attendance at elite universities and high income later in life is quite likely a spurious correlation. People who go to elite universities are likely to come from social backgrounds (wealthy and well-educated parents) where they have acquired the skills, habits, and professional and personal networks that will make them successful in their careers regardless of which university they attend. The elite university may not contribute much to their income. Moreover, the causality might well be reversed: young adults intent on getting rich with highincome jobs may decide that the best way to do that is to attend an elite university. Students with other career and life projects may decide to forego the “rat race” and attend a quirkier or less stressful university. Clearly the coefficients of a regression can be interpreted as a kind of average causal effect in the context of a randomized control trial (RCT). The presumption of RCT analysis is that the treatment is indeed the cause of the outcome, since there is no omitted variable explaining both (since the treatment group was selected randomly), there is no feedback, and there is

58

Women and Development in Africa

no reverse causality. The one caveat is that the control group cannot always be interpreted as “what would have happened in the absence of treatment” since in many cases the control group may be somewhat affected by their knowledge that they are a control group. There may be resentment at not having been selected for the treatment, and this might influence behavior. Another method used to estimate causal effects when there is reason to believe there is reverse causality or spurious correlation is called the method of instrumental variables (Angrist, Imbens, and Rubin, 1996; Angrist and Krueger, 2001). In this method, the analyst finds a variable that is likely to be correlated with the troublesome explanatory variable (which is not really causal on its own) but not correlated in any causal way with the outcome variable of interest. That is, this other variable, called an instrumental variable (an IV), satisfies an exclusion restriction. It can explain the x variable, but not the y variable. Consider for example the problem of estimating how a wife’s bargaining power in the home affects child nutrition. There may be a strong correlation between measures of bargaining power and outcomes of child nutrition. Women with good measures of bargaining power have better nourished children. But this relationship could be spurious. Husbands who care about good child nutrition may give their wives considerable say in household decisionmaking. The wife’s empowerment is not causing the good child nutrition. Moreover, the causality could be reversed: when children are robust and healthy because by chance they are less susceptible to disease, the husband rewards the child’s mother by giving her more autonomy in household affairs. In this context, an experiment that changed bargaining power for a treatment group would enable the analyst to say that the effects of the change in bargaining power were causal. Alternatively, the analyst could find a variable that is correlated with the wife’s bargaining power but not directly causal of child nutrition. For example, it might be that women who attended primary school were mostly taught by male teachers, while some women had female teachers. The female teachers, who perhaps never talked about child nutrition, may have empowered their female students. Childhood exposure to a female teacher, then, might have changed adult empowerment and hence bargaining power of women, without directly causing changes in child nutrition practices. The variable could be used as an instrumental variable. In terms of the mechanics of IV estimation, basically the analyst first runs a regression with exposure to female teachers as the explanatory variable and the measure of contemporary bargaining power as the outcome variable. The predicted values from this regression can be interpreted as what bargaining power would have been in the absence of the feedback loop from child nutrition to bargaining power. Then the pre-

Women and Development in Africa

59

dicted values of bargaining power are used in the second-stage regression explaining child nutrition status. The estimated coefficients can now be interpreted as causal effects. There is of course much more subtlety and technique in the actual implementation and interpretation of the method of instrumental variables. But the idea is straightforward, and has been a major advance in the social sciences.

3 Models and Methods

Butterfly Burning [by Yvonne Vera] captures a transitional moment in Zimbabwe’s history and uses it to explore possibilities for national and individual self-creation. Set in the emerging “modern” city of Bulawayo, it tells the love story of Phephelaphi, a young woman born and bred in the city, and Fumbatha, an older man of fifty with an inherited legacy of resistance. Their intense relationship survives until Phephelaphi seeks to extend herself by applying to train as a nurse. Her quest sets her in collision with Fumbatha, especially when she aborts a pregnancy in order to free herself to enter college. When she becomes pregnant for the second time, Phephelaphi sets herself on fire and burns in her own storm. —Nana Wilson-Tagoe, Africa after Gender?

he quest for freedom ends in self-immolation; such is the cautionary metaphor of Yvonne Vera’s novel Butterfly Burning. The exacting portrayal of the mind of Phephelaphi is one way to learn about the world— about the consequences of choices, and about how choices are changed when circumstances change. This way of learning about the world differs from the approach of the social sciences. Economists, political scientists, sociologists, and even anthropologists have borrowed the techniques of the natural sciences: learning about how the world works by deducing falsifiable predictions from theory, designing controlled experiments, and measuring outcomes to see whether the theory is verified or falsified. To take a banal example from the natural sciences, a theory might be that sunlight affects how humidity levels produce oxidation of iron (or, does rust never sleep?). A simple experiment will not illuminate the underlying chemical reactions, but will answer the question; further experiments might be de-

T

61

62

Women and Development in Africa

signed to determine exactly what occurs at the molecular and atomic level in the process of rusting with and without sunlight. Social scientists are increasingly adopting the experimental method to understand human behavior. Theories, often framed as models about human intentions, frame expectations about behavior. A model suggests how people might respond to a change in circumstances. A treatment or intervention of interest is then applied to a population in a systematic way, with variation controlled by the experimenter. This chapter introduces basic models and methods used in the social sciences. The first part of the chapter summarizes the basic theoretical approaches that are used by many social scientists. The second part of the chapter reviews recent efforts of social scientists to measure questions pertinent to economic development (and gender and gendered structures) by using two related experimental methods: randomized control trials and field experiments. Randomized control trials (RCTs) are methods for evaluating the impact of programs or policies (Duflo, Glennerster, and Kremer, 2008). In its simplest form, an intervention is implemented (an education program, for example) for one group and not implemented for the other group. The experimenter randomly selects both groups from the population of interest. The groups are thus reasonably identical on average. The first group is called the treatment group and the second the control group. If the sample groups chosen are sufficiently large and random selection of the groups was performed carefully, then differences in outcomes from the intervention can be attributed to the program. RCTs identify causal effects of programs and policies and are thus especially important in development studies, where researchers, government policymakers, and nongovernmental organizations (NGOs) want to learn which programs and policies will improve well-being. In the typical RCT dealing with gender, the treatment and control groups are blocked by gender. The difference between the male treatment group and the female treatment group, compared with the difference between the male and female control groups, can be interpreted as the measure of how much gender structures affect program outcomes. Field experiments are experiments conducted outside of a laboratory, in settings that are comfortable or normal to the population under study (Duflo, 2006; Levitt and List, 2009). In a typical field experiment, the experimenter invites people in a community to complete tasks or make choices. There are real consequences to the activities, in the sense that performance generates rewards according to a protocol understood by the sample population participating in the study. The activities are artificial, but not hypothetical. The label of field experiment arose to distinguish these sorts of experiments from laboratory experiments typically conducted on a university campus using university students as subjects.

Models and Methods

63

Field experiments have been used to measure differences in fundamental preferences across genders. These experiments create artificial settings designed to give incentives for people to reveal their preferences or attitudes over choices that are important for economic development (Cardenas and Carpenter, 2008). Six preferences have received much attention: (1) the rate at which future money is personally discounted (how much a person values cash today versus cash in the future); (2) the degree of aversion to risk (how much a person is willing to give up in order to have a certain amount of cash compared with a lottery ticket that represents risky possibilities of getting different amounts of cash); (3) cooperativeness for public benefit (how much a person is willing to contribute in order to benefit a group, even if the person is not the direct beneficiary); (4) trust and trustworthiness (how much money a person is willing to send to another, trusting that the other person will return the favor); (5) altruism, fairness, and egalitarianism, all concepts related to how empathy translates into specific choices; and (6) competitiveness (how much a person enjoys or avoids situations that are competitive, controlling for the likelihood of winning or losing the competition). Many people believe these preferences vary systematically between men and women and are thus responsible for many patterns of outcomes. For example, many people believe that in advanced economies there are few women chief executive officers (CEOs) of major corporations because women are less competitive than men. Field experiments are also used extensively to study how people respond to different framing contexts. For example, do people perform better or worse when they have to multitask? When they are given messages about ethnic or gender stereotypes? When they receive different forms of extrinsic variation, or are reminded of intrinsic motivation? The labeling of an evaluation or experiment as an RCT or a field experiment is somewhat arbitrary. There are many situations where both labels are applicable. An RCT is an experimental situation intended to evaluate an actual or prospective program or policy. The goal of an RCT is to learn about the program in order to replicate on a larger scale. RCTs are instruments toward improving policy. A field experiment is a more artificial setting that often does not correspond to any actual program or policy. The goal of a field experiment is to learn about human behavior. Some field experiments have immediate implications for policy and programs; others are more like basic research, testing abstract theories and offering insight and understanding with less clear practical value. A Model of Human Behavior Suppose that an economist conducting a randomized control trial in rural Kenya offered households the opportunity to purchase bed nets treated with

64

Women and Development in Africa

insecticides that would prevent mosquitos from biting people while they are sleeping. Mosquitos transmit malaria, a devastating disease, and avoiding being bitten while sleeping greatly reduces the chances of contracting malaria. The economist might randomly change the amount of the subsidy for the bed nets. Some might be free, others might be offered at extremely low prices, and others at some price closer to the cost of production. The experiment might reveal that households accept and use bed nets when the price is zero, but refuse to purchase or use bed nets when the price increases. Such a finding must be explained. A zero price causes families to obtain many more bed nets than a small positive price. Why does a small positive price have such a large effect on behavior? Such questions are answered by models of human intentionality, or, more simply, of decisionmaking. There are many such models. Some hyperrational models have humans applying the foresight of chess grand masters to every action they might take, including how such actions might affect their own “future selves.” Psychologically based models have every decision compartmentalized into different modules of the brain: the same underlying decision, if framed in different ways, generates completely different choices. Simpler models, standard in much of economics, assume that people make choices to maximize their satisfaction or happiness, subject to the structural constraints they face and subject to the cognitive biases and impairments that affect abilities to evaluate alternatives. The measure of satisfaction or happiness is often labeled the “utility” received by a person from consuming or doing something, following the usage of Jeremy Bentham, the nineteenth-century British philosopher who is credited with founding the school of thought known as utilitarianism. A utility function maps actions taken and goods and services consumed by an individual to an indicator of the individual’s utility. The utility function can be thought of as representing a ranking of all possible outcomes of the conceivable actions. Budget and time constraints delimit the feasible range of utility. Social structures are important in determining these constraints, but of course individual, personal characteristics also matter. Structures typically determine the resources available to a person, their budget constraint, and their available labor, a time constraint. Given constraints, the simple maximization model predicts that people choose the highest-ranking achievable outcome. Why, after all, would they ever choose a lower-ranking achievable outcome? This model, suitably expanded to include cognitive biases and complex preferences, is the filter through which researchers design and interpret the results of many RCTs and field experiments. That is, most experiments explore decisions amenable to analysis by the maximization model. Social science researchers design experiments to observe how behavior is affected by

Models and Methods

65

changes in incentives or changes in the context of decisions. Predictions generated by maximization models are compared with outcomes of experiments. Out of the back-and-forth between theorizing and experimentation emerges nuanced and more complete understanding of behavior. That said, the vehemence of debate between “rational choice” adherents and “behaviorists” should not be underestimated. For some social scientists, any suggestion that behavior can be explained by decisions that people might, after reasoned discussion, agree were not good decisions for them is not an explanation at all. Rational choice proponents, on the other hand, have a hard time explaining why people prone to diabetes keep eating donuts. The maximization model illuminates several reasons why the choices that people and firms make might be gendered. The choices men and women face may be different; the choices people make may be about how to treat men and women; and men and women may make decisions differently when faced with the same choice and context. Choices might be gendered because the trade-offs or alternatives presented to a decisionmaker might be gendered. Structures determine the trade-offs that people have to confront. Two kinds of trade-offs might result from gendered social structures. One is where men and women have to make similar decisions, but the specific choices and opportunities available to them are different. The other is where men or women, as individuals or as members of institutions, have to make choices about the allocation of resources to other people, especially children, who are gendered. As parents and as members of institutions, people must choose how to allocate resources across gender. The outcome of their resource allocations will depend on how social structures condition the status of dependents, clients, or subjects. Another reason why choices might be gendered is that men and women may comprehend and respond to stimuli and opportunities in different ways. Part of the difference may be biological. The brains of men and women may have evolved differently. Another part of the difference may be due to nurture. Human thinking is conditioned by the social setting in which a child is raised. Nurturing shapes the thoughts and preferences of individuals. After years of wearing gendered clothes as a child, pants for boys and dresses for girls, an adult may retain preferences for those gendered clothes as an adult. Habit is rewarded by pleasure; breaking with habit is punished with stress. Evolution and childhood socialization into gendered roles persist in the form of gendered adult preferences and ways of thinking. Stigler and Becker (1977) tried to discount this second approach to understanding decisionmaking, as part of a very general argument about the usefulness of disciplinary boundaries. They suggested that a useful analytical approach for the social sciences (of economics and sociology, as opposed to psychology) was to follow the maxim “de gustibus non est dis-

66

Women and Development in Africa

putandum” (of tastes we do not dispute). They argued that it was certainly true that much social behavior could be explained as a matter of tastes. But for analytical approaches outside psychology, a useful and complementary method, in their view, was to proceed as if tastes or preferences were not the issue and instead to seek explanations in other social variables (such as prices, technologies, and social structures). Behavioral economists successfully challenged this position in the 1990s (Camerer, 1999; Rabin, 2002). These researchers argued that the dispassionate disdain for tastes of the de gustibus approach led many economists and sociologists astray. Too much research effort was wasted trying to come up with tortured explanations of human behavior as resulting from something other than the “humanness” of humans. The classic example of this comes from predicting how humans respond to very simple strategic settings. For example, in the “ultimatum game,” an experimenter offers two subjects the following proposition. They are offered $20, on condition that one proposes a split of the $20 and the other agrees. The experimenter chooses who gets to propose and who agrees. The subjects might be anonymous, and not even in the same room. The remorseless logic of maximization tells the proposer to offer to keep $19.99, and the respondent to accept. A penny, after all, is more than no pennies. Nothing of the kind happens in laboratory and field experiments. Enormous fractions of populations all over the world agree on fifty-fifty splits. Only theories that incorporate and model tastes for fairness or altruism can explain this behavior. These tastes vary across individuals and populations, and are demonstrably gendered (Eckel and Grossman, 1998; Kamas, Baum, and Preston, 2008). Behavioral experiments have also made clear that people have no single organizing “rationality” to decisionmaking. The behavioral economics critique of the simple maximization model goes well beyond demonstrating obvious cognitive biases and relevance of preferences. It invites more careful attention to the brain as a biological mechanism making choices. Field (2001) suggested that human brains evolved with distinct kinds of hardwired programs in their brains. One is the maximization program, which Field called the “foraging” algorithm. Other programs kick in and direct behavior as circumstances dictate. These modular algorithms lead to behavior quite different from that predicted by a foraging algorithm. One discussed by Field is the “cheater detection module,” whereby in many social situations people become obsessed with determining whether a social rule has been violated. To the outside observer, these people are irrational. And they are! Their brains cannot help it. Quite possibly many of the gender differences observed in human behavior are due to these kinds of modular responses to the same stimuli. Another challenge to the de gustibus approach came from a group of economists and social scientists who might be best described by the appel-

Models and Methods

67

lation “culture vultures,” coined many years ago by Homans (1983). These scholars, from a variety of disciplines, had increasingly become dissatisfied with the idea that their disciplinary boundaries precluded them from studying cultural variation in preferences (Harrison and Huntington, 2000; Rao and Walton, 2004). Culture, in their view, had profound effects on politics and economics and should not be something relegated to psychologists and anthropologists. Careful studies have identified the enduring role of culture, by showing how cultural traits that emerged long ago, or in social situations removed from present situations, persist and influence actions in the present. For example, Miguel, Saiegh, and Satyanath (2008) demonstrated that levels of civil conflict in home countries predicted levels of violence on the soccer pitch in European football. Players from countries with more civil conflict were more likely to get yellow and red cards. Similarly, Fisman and Miguel (2007) revealed that diplomats coming from countries that were more corrupt were more likely to be scofflaws regarding New York parking regulations when they were posted to the UN. Diplomats from countries with low levels of corruption rarely had outstanding parking violations. Fernández and Fogli (2006) and Fernández (2011) showed that descendants of immigrants followed fertility and work choices that were shaped by social structures back in the home country that their parents had emigrated from. Returning to the general point, the simple maximization model is a building block of more complex models that realistically incorporate behavioral realities and cultural preferences. It emphasizes how a choice to do one thing, given a constraint (culture, cognition, budget, prices), implies that something else cannot be done. Not doing the other thing is the opportunity cost of doing the chosen action. The opportunity cost of purchasing and consuming an extra unit of the good with low marginal utility, relative to its price, is not consuming the other good. The other good generates more utility per dollar spent. The model predicts that a person allocates his or her income so that the opportunity costs of different actions are equal. The perceptions of opportunity costs are, of course, subject to behavioral biases and influenced by culturally shaped preferences. Firms, farmers, and traders are often modeled as maximizing an even simpler function, a profit function free from behavioral biases and cultural encumbrances. Profits are the difference between revenues and costs. Typically, one thinks that profits rise as more product is produced and sold, usually at a decreasing rate. As a firm expands production, it uses the inputs less and less efficiently. Management may get top-heavy, or machines may wear out faster. A simple expression for profit is that it is the value of production minus the costs of inputs. The costs of production are simply the wage times the amount of labor hired. The optimal amount of labor to hire is where the extra revenue from producing and selling the additional units

68

Women and Development in Africa

produced by another unit of labor input, which is the price of the product times the marginal product of using more inputs, is equal to the marginal cost of production, here simply the wage. The marginal product indicates how much a change in labor affects production. With a normal relationship between inputs and production, there will be only one level of labor that maximizes profits. Utility and profit maximization models, and clear articulation of relevant behavioral biases and cultural influences, are useful for a number of reasons. Just the simple maximization model alone has been very productive in this regard. Many problems, such as determining the optimal allocation of a financial portfolio or the optimal mix of crops and planting strategies in high-risk environments, can benefit from careful specification of objectives and use of mathematical techniques to derive solutions. Indeed, application of these techniques has been one of the driving forces behind the increased productivity of agriculture around the world. More deliberate maximization may be very rewarding. In addition, maximization models may help researchers explain people’s decisions. The maximization model does not preclude ignorance, mistakes, and bad judgment. Nor does it preclude choices made under the influence of addictive necessity. It does emphasize how unlikely it is that an outside observer to a social situation, especially one involving economic decisions, will be able to easily identify choices that are making people worse off. The spirit of maximization models is that people are probably making the best choices they can, given the information and possibilities that they have. If an outside observer sees a person doing something that seems strange, the person probably has a good reason for doing it. A classic case of the importance of this spirit concerns discussions about the future of peasant farmers (Schultz, 1964). For many, peasants are tradition-bound, ignorant, and superstitious. There is little hope of bringing them into the world of “modern” agriculture, in this view. To put land to more productive use, peasants should be removed. The maximization approach suggests more careful investigation. Peasants are poor, but their behavior exhibits many of the traits predicted by simple maximization models. Give them more opportunities, this approach says, and you will be surprised. Thinking about people as maximizers weeds out bad thinking and bad writing (the two grow together). Consider the following sentence, from an actual paper submitted to an academic journal, but typical of bad thinking: “Because farmers are poor, they are forced to grow cash crops, rather than food crops, and this makes them poorer.” Interpretation of this sentence hinges on the sense of the much-abused word “forced.” Was force actually used? The context makes clear it was not. Farmers had choices about whether to grow cash crops or food crops. In this case, the context makes clear that the sentence has little meaning. The choice to grow cash crops

Models and Methods

69

must make the farmer less poor. Otherwise, the farmer would be choosing to be poorer. The maximization model suggests caution when evaluating assertions about how choices have made people worse off. Circumstances make people find themselves in dire straits; choices make them better off. Maximization models are also useful in figuring out and predicting complex responses to stimuli. For instance, the utility model enables us to see that most changes in economic environments have both “substitution effects” and “income effects.” The substitution effect captures how a change alters the opportunity costs of different actions and consequently shifts behavior toward the action with the now lower opportunity cost. The income effect captures how changes make some people better off and others worse off. When the price of a product goes down, a consumer is conceptually wealthier; he or she can purchase everything he or she did before, with income left over. The income effect captures how a person’s spending changes. These effects are much in evidence in gendered projects, where the interventions that change the circumstances of choices regarding girls and boys, or choices made by women or men, might have income and substitution effects that go in opposite directions. An example may clarify this idea. In the development economics literature, considerable attention has been devoted of late to the desirability of empowering women through microcredit programs. The credit programs intend to lower the cost of capital for women and thus increase women’s incomes and give them more bargaining power at home, and women then will be able to direct more household resources toward their children. Utility-maximization models (and some common sense) alerted researchers to the possibility that promoting women’s microenterprises might have an unintended side effect. The income effect works in the right direction of helping children, but the substitution effect works in the wrong direction. The woman’s time working is now more valuable since she has more capital and skills. The opportunity cost of her time at home with her children is higher (she now earns more outside the home than she did before). She may earn a higher income but spend less time at home caring for the children. Examples of Randomized Control Trials: Bed Nets and Fertilizer Many RCTs shed light on gender-related aspects of the simple maximization model. This section reviews recent RCTs to promote use of bed nets against malaria and fertilizer to increase crop productivity. The next session reviews field experiments designed to measure trust and contributions to public goods. The brief reviews serve to introduce the experimental method, explore some important development issues, and elaborate on the discussion of the simple maximization model.

70

Women and Development in Africa

It turns out that there do not seem to be important gendered dimensions to decisions, taken within household settings, to acquire and use bed nets and fertilizer, at least for the small samples represented by the studies. The studies do, however, shed light on the importance of developing appropriate maximization models. Studies of bed net adoption and use have focused on how households allocate goods within the household; that is, when a household is maximizing, what is the weight of emphasis placed on the interests, preferences, or well-being of different household members (i.e., children, spouses, grandparents, and men versus women)? The study of fertilizer adoption and use reviewed here raises interesting questions about how people choose between the present and the future, especially when they have partial awareness that their future selves may not make the choices that their past selves wanted them to make. It should be noted that most RCTs cannot distinguish what the exact mechanism is for the difference in gendered outcomes. When an RCT finds that effects of interventions differ by gender, it may be the case that biology, and not gendered structures, is responsible for the outcomes. An education program to help men and women build fences on their farms might have no impact on female farmers. The reason for the nonoutcome may have nothing to do with gendered social structures, and might instead simply be the differences in the physical strength needed to pound fence posts into the hard ground of rural farms. More generally, it may be that men and women are biologically predisposed to act in different ways in many important social and environmental settings, and likely respond to social change (new programs, new relationships) in different ways. Skeptics who doubt the influence of gendered social structures on behavior instead emphasize that the mental processes of men and women differ for reasons due to evolutionary biology, and these differences explain gendered outcomes. The RCTs reviewed below are unlikely candidates for this sort of explanation. There is little reason to think that fundamental biologically determined predispositions are so strong as to overwhelm intentional deliberation. For the field experiments discussed later, it is perhaps more likely that emotions—induced by hormonal changes in response to the situations of the experiment—are generated differently according to gender and influence the choices men and women make. Before discussing the two case RCTs, a short digression on estimating the impact of interventions is in order. Suppose an organization was running a randomized control trial to see how absenteeism by teachers in school was affected by a program to pay them a bonus if they took pictures of themselves on tamperproof cell phones that dated the pictures. A treatment group of teachers would be paid a bonus for each class day they took a picture of themselves in the classroom with kids. A control group would not get the cameras or bonus plan. Actual absenteeism might be recorded

Models and Methods

71

by inspectors (who were always truthful) randomly visiting schools. Very generally, we assume that the level of absenteeism for the teachers that were in the control group, before the program was implemented, is α0. This is the “pre, non-camera” absenteeism rate, where pre stands for before the program was implemented. We represent the absenteeism rate for the treatment group of teachers, again before the program, as α0 + α1, so that α1 represents any differences in absenteeism caused, perhaps, by the treatment group being different from the control. If the randomization protocol was followed and the sample size was large, this difference should normally not be statistically significant. For the control group, after the program, their absenteeism becomes α0 + β0, where β0 is the change that happens over time, other than the change due to the program, which by assumption does not affect the control group. But the control group might know the treatment group is in the program, and so the mere knowledge that there is increased surveillance might prompt behavioral change. This change is assumed to be common to both treatment and control. For the treatment group, their change over time is β0 + β1. The first term reflects what would have happened to them without being in the program, while the second reflects the specific effects of their inclusion in the program. So their absenteeism after the change will be α0 + α1 + β0 + β1. These absenteeism rates, before and after, and for control and treatment groups, can be better visualized as a table. Table 3.1 represents them, and adds the differences between rows and columns, and then the difference in the differences. This final value, the difference in the differences, can be thought of as the true impact of the program. A numerical illustration might be helpful. Suppose the inspectors found that before the program, the treatment teachers (who got the cameras) were absent 25 percent of the time. The control teachers were absent 24 percent of the time. After the program had been running for six months, the treatment teachers were found to be absent 19 percent of the time, and the control teachers were absent 22 percent of the time. Table 3.2 shows what the best estimate of the effect of the program would be. The table calculates the “difference in difference” as –0.04; that is, absenteeism went down by four per-

Table 3.1 The Difference-in-Difference Approach to Estimating the Effect of a Program Pre Post Difference

Control α0 α0 + β0 β0

Treatment

α0 + α1 α0 + α1 + β0 + β1 β0 + β1

Difference α1 α1 + β1 β1

Note: Pre refers to value of average level of outcome before program; post refers to outcome after program.

72

Women and Development in Africa

Table 3.2 The Difference-in-Difference Approach: Hypothetical Numbers for a Teacher Absenteeism Experiment Pre Post Difference

Control 0.24 0.22 –0.02

Treatment 0.25 0.19 –0.06

Difference 0.01 –0.03 –0.04

Note: Pre refers to value of average level of outcome before program; post refers to outcome after program.

centage points due to the program. More generally speaking, the table makes clear that only using data comparing a treatment group to a control group after a program is likely to be a biased estimate of the impact of a program, and that comparing just the treatment group before with the treatment group after is also likely to be a biased estimate of the impact of a program. When it comes to protecting their children from the debilitating and often deadly disease of malaria, male and female parents might be expected to make similar choices when confronted with similar contexts for their decisionmaking. Randomized control trials conducted in numerous African countries suggest, indeed, that there are few substantial differences in behavior. Malaria is a parasite that causes great damage to the health of humans and is one of the leading causes of death in sub-Saharan Africa. The parasite has a complex life cycle. A sporozoite is transmitted from mosquito to human. There it reproduces and transforms itself, and other mosquitoes take up gametes from the blood of an infected person. The parasite occupies and destroys red blood cells, and thus causes anemia, fever, vomiting, and possibly death. Malaria kills approximately 600,000 children each year, mostly in Africa. One of the most effective preventive remedies against being bitten by malarial mosquitoes is to sleep under a bed net that has been treated with insecticide to repel and kill mosquitoes. For almost two decades, insecticide-treated bed nets have been manufactured and distributed in African countries. Two developments greatly expanded the reach of these efforts. In the early 1990s, Sumitomo Chemical Corporation of Japan developed a process to manufacture permethrin-treated polyethylene fiber that could be used to make bed nets. Later in the 1990s, the World Health Organization and major donor foundations began to subsidize antimalarial bed nets on a large scale. Recent analyses of malaria prevalence suggest that bed nets and other public health measures have led to substantial reductions in the incidence of malaria (Ceesay et al., 2010; O’Meara et al., 2010). As distribution of bed nets expanded, debates erupted over whether the Antimalaria Insecticide-Treated Bed Nets

Models and Methods

73

better long-term strategy was to give bed nets away for free or to charge some amount for each bed net. Proponents of fees for bed nets argued that purchased bed nets would be valued and thus used more than free bed nets, and that fees from sales would enable more bed nets to be produced and thus reach more people affected by malaria (Minakawa et al., 2008). Opponents argued that even low prices for bed nets would dissuade most of the population at risk from malaria from obtaining bed nets (Cohen and Dupas, 2010; Guyatt, Ochola, and Snow, 2002). Regardless of whether bed nets were free or purchased, another debate concerned whether to target distribution channels to men or to women (Mujinja, Makwaya, and Sauerhborn, 2005). Would it be more likely that bed nets would be used for children if women were targeted to make decisions about obtaining or purchasing bed nets? Hoffman (2008, 2009) reported on an experiment designed to measure the effects of varying the gender (and position in the household) of recipients of antimalarial bed nets. She estimated the effects of distributing bed nets to men rather than to women. The RCT was conducted in rural southwestern Uganda in a very poor region. Households consumed less than $1 per day per person and adults had about four years of schooling. More than 80 percent of people in the households suffered from malaria every year. Hoffmann’s sample consisted of 131 households that had a child up to age five living in the household. Half of the households received bed nets for free, and half received cash equivalent to the number of bed nets they would have received (according to the number of people in the household). The husband or wife was randomly selected to participate, and those who received bed nets could resell them for cash while those who received cash could purchase bed nets. All households except one, apparently, kept their bed nets if they were received in kind, while about 15% of the households receiving cash decided not to purchase any bed nets. Surprisingly, young children under five were the least likely group to be found using bed nets when the household was subsequently visited in the evening for a check on usage. The person who obtained the net was more likely to be using the net. Men and women were both likely to keep nets; gender differences were minor. The randomized control trial, albeit for a very small sample, provided quite strong evidence that perhaps the gender of the adult recipient of bed nets was not very important in adoption and usage of bed nets. Maize, cotton, sesame, groundnuts, sorghum, and millet are crops grown by hundreds of millions of small farmers in Africa. Crop yields, measured as how many tons farmers harvest per hectare, have stagnated throughout the continent. One reason is that African farmers use relatively little fertilizer. Promoting Fertilizer Use

74

Women and Development in Africa

Elsewhere in the world, crop breeders have developed varieties of crops that respond well to fertilizer. Fertilizer refers to a variety of compounds, typically combinations of nitrogen, phosphorus, and potassium. Fertilizers can be both organic (manure and guano) and inorganic (produced through a variety of chemical processes). Two centuries of experimentation on public research stations and farmer’s fields have demonstrated the profitability of applying fertilizer, provided other conditions such as soil quality, slope of the field, rainfall patterns, and the value of the crop are appropriate. One is hard-pressed to think why the decision over how much fertilizer to purchase and apply on a field should be gendered, unless gendered structures were affecting the context of fertilizer choice, or unless the trade-off between risk and return was evaluated differently according to gender. There is, in fact, mixed evidence on the question of systematic gender differences in application of fertilizer (Peterman, Behrman, and Quisumbing, 2010). Analysis by Udry (1996), using data from Burkina Faso, suggested that women who farmed their own fields separately from household fields managed by their husbands used far less manure on their fields. Kilic, PalaciosLopez, and Goldstein (2013) decomposed the sizable gender gap in input and productivity for a large sample of Malawian farmers. They found that much of the difference was due to what they called an “endowment” effect, with female farmers having fewer resources for farming than male farmers. That is, if female farmers had the same resources as male farmers, they would apply similar levels of inputs. Chirwa et al. (2010), likewise analyzed results from a large-scale distribution of fertilizer vouchers in Malawi and found efficient intrahousehold distribution of extra subsidized fertilizer across fields of men and women. Duflo, Kremer, and Robinson (2008, 2011) examined the return to fertilizer and the effects of changing incentives to acquire fertilizer in western Kenya. The two papers offer considerable insight into the decisionmaking process (the maximization model) of farmers. First, the authors reported that replicable trials on the fields of farmers demonstrated that fertilizer application of a half teaspoon per plant as top dressing (applied when maize plants were about a month old) could be profitable. That is, there were amounts of fertilizer that when applied in a timely way would generate substantial increases in yields of maize. At prevailing prices of maize and fertilizer, then, fertilizer seemed like a profitable, though risky, investment for a typical farmer. At the very least, they suggested that purchases of about 50 kg, costing around $20, would generate net returns of about $5. Therefore, there was a 25 percent return over the growing season, yielding about a 50 percent annualized return. Since farmers only earned about $1 per day (the local wage), this would amount to about one week’s worth of income. Not a huge amount, but not trivial either. It should be noted that their calculations of the overall benefits of in-

Models and Methods

75

vesting in fertilizer did not take into account the riskiness of the investment. The benefits of fertilizer use depend on rainfall, pests, and the human element (whether weeding was done in a timely manner, since fertilizer also encourages weed growth). There is work to be done to determine the full distribution of returns to applying fertilizer on maize, in order to calculate the “risk-return” frontier (Alem et al., 2010). Moreover, it is imperative to remark that Duflo, Kremer, and Robinson found that the government-recommended application of fertilizer with hybrid seed likely yielded negative returns (costs were larger than benefits) for most farmers. So many farmers had good reasons to be skeptical about the benefits of fertilizer and cautious in purchasing and applying fertilizer. The interesting part of the work of Duflo, Kremer, and Robinson was that they determined that by and large farmers agreed on the profitability of modest amounts of fertilizer, and indeed many farmers used fertilizer. However, a substantial fraction of farmers, more than 50 percent, used no fertilizer at all. This nonusage by almost half of farmers appears to be difficult to explain. Farmers seem to be in agreement that small purchases and application of fertilizer would generate positive returns, even including the transaction costs of going to the nearby town to purchase the fertilizer. If farmers were profit-maximizers, they either had severe constraints on their ability to finance the investment or were very averse to the risk of a bad season or illness. (That is, as discussed in the previous chapter, if the financial system performed very poorly, then profitable investments would not be made, and incomes would grow slowly.) Duflo, Kremer, and Robinson believed that financial market performance could not be the whole story, since farmers did have cash flow that would have enabled them to purchase fertilizer. After the harvest sales of their crops, months before the next season, farmers had cash. They could purchase fertilizer. Duflo, Kremer, and Robinson hypothesized that farmers in western Kenya were similar to people everywhere and had a hard time convincing themselves to make the commitment to postpone consumption. They were maximizers, but they were slightly irrational maximizers, in that they consistently underestimated the ability of their future selves to make the right decisions. They procrastinated, and knew that they were likely to procrastinate, but still did not control their procrastination. As the researchers put the behavioral hypothesis (p. 2352): Going to the store, buying fertilizer, and perhaps deciding what type of fertilizer to use and how much to buy involve a utility cost. Even if this cost is small, so long as farmers discount future utility, even farmers who plan to use fertilizer will choose to defer incurring the cost until the last moment possible, if they expect they will purchase the fertilizer later. However, farmers who end up being impatient in the last period in which buying is possible will then fail to invest in fertilizer altogether.

76

Women and Development in Africa

In order to confirm whether this behavioral trait was a significant explanation of the low take-up of fertilizer, the authors designed an RCT for about 1,000 farmers in the Busia region. Working with a local nongovernmental organization, International Child Support (ICS), they randomly selected farmers to participate in a program they designed called Savings and Fertilizer Initiative (SAFI). The program was very simple: field staff from ICS visited the farmers shortly after the harvest and offered them the opportunity to purchase a voucher for fertilizer for the next planting season. The idea was that a farmer with no problems of control would refuse and just save and invest his or her money until the time of applying fertilizer was at hand, while farmers who knew they had some self-control problems would appreciate the commitment device, and so would purchase the vouchers (which cost the same as fertilizer). To sweeten the incentives to purchase vouchers, the ICS staff promised free delivery, thus saving, on average, perhaps an hour of walking to and from town. The program generated substantial increases in fertilizer use. Compared with the control group, who did not have access to the program, many more farmers in the SAFI program used fertilizer (45 percent compared with 34 percent in the control group). There were more complex programs offered in subsequent seasons, which confirmed that time-consistency problems were central to explaining why more than 50 percent of farmers did not use fertilizer, even though fertilizer use was profitable. There were, however, no gender differences at all. Once again, the prior belief that farmers would differ greatly, by gender, in their responses to new opportunities would seem not to have been apposite. Field Experiments to Measure Economic Preferences The novelist Chinua Achebe was fond of the African proverb “Until lions start writing down their own stories, the hunters will always be the heroes.” So it has been with development studies, where researchers from wealthy countries have often ascribed virtues to the societies they regard as their own. The West, they have said, rose above “the rest” because it deserved to be great. Virtues of patience, trust, altruism, empathy, tolerance, fairness, egalitarianism, creativity, and willingness to contribute to public goods were all abundant in the West (Clark, 2008; Landes, 1999). They were central to development. They made the West rich, while the rest were left behind. This triumphalist retrospection blocks other possibilities. Other preferences may also have been key to economic growth: being greedy and selfish, perhaps, or more competitive, or more or less averse to risk. Even the West-was-best scholars appear to disagree over the exact list of important preferences and which direction those preferences should have leaned.

Models and Methods

77

Would a mistrusting, self-interested society grow more rapidly than an altruistic, cooperative society? There are numerous models where preferences may influence the aggregate rate of growth. Depending on the underlying economic structure and patterns of preferences, evolutionary models of preferences also generate predictions about which sorts of preferences might be more frequent in a population. The definitions of preferences are also often vague. What does a preference for fairness mean, exactly? Thus, the bold representation of whole societies as sharing a set of virtues has given way to a more nuanced and careful attempt to measure economically relevant preferences among groups of people. This literature is in its infancy, however, and there is very little knowledge on how stable economic preferences are over time, what the mechanisms are that lead to change, and what external stimuli induce change in preferences. Some field experiments are deeply flawed in this regard, as when experimenters compare behavior from a handful of groups and assume that cultural differences account for variation across group behavior. The sample sizes in effect are sometimes as small as two (one group framed to make culture salient, the other not; one group from one culture and another group from another culture). Since the experimental conditions cannot be identical (different times, different people in the group), any variation (a loud cough) also explains the difference across the groups. A standard laboratory and field experiment called the investment game or trust game is frequently used to measure trust and trustworthiness. Persons recruited to participate in the game (in the experimental literature they are known as subjects) are told that they will be given an amount of money, and part of this money can be transferred to another player, called a receiver. The other player is usually anonymous to the first player. When the sender transfers money, it is multiplied (doubled or tripled, typically). The receiver chooses whether to send some of the multiplied money back to the sender. The sender “trusts” the receiver to send back some money. If there were little trust, the sender might not send anything to the receiver. The receiver demonstrates that he or she is trustworthy, in the sense that returning money to the sender yields no payoff other than a feeling of having merited the trust of the first player. The trust game is played for real money, with rewards varying according to the budget of the experimenter, who also determines the subject pool. When comparing outcomes of the trust game across social groups, it is important to clarify that referring to the outcomes of the game as trust and trustworthiness is shorthand that may be misleading. That is because there are other explanations for variation in average behavior across social groups. A sender in the trust game weighs a number of considerations in his Trust and Trustworthiness

78

Women and Development in Africa

or her maximization problem. There may be a “warm glow” feeling from doing the morally right thing or even the efficient thing (sending money increases the total available to the players as a group). Trust games confound kindness (or altruism) with trust and trustworthiness. Many subjects may experience a good feeling when they act according to a norm of being trusting, regardless of whether they actually trust the person to reciprocate (Ashraf, Bohnet, and Piankov, 2006). Moreover, even if the average level of trustworthiness is high, there may be considerable variation across individuals in a social group, and so the transaction is risky. Two social groups with the same level of average trustworthiness and the same variance in trustworthiness, but with different levels of aversion to risk, will have senders sending different amounts. Indeed, both Schechter (2007) and Barr (2003) have noted that the amount sent in the standard trust game should usually be interpreted as both trust and willingness to undertake risky behavior. They have suggested that when explaining cross-sectional variation in trust, choices in experimental games that elicit risk-taking behavior (or other measures) should be included as controls. There may also be an “aversion to betrayal” as a motivation for sending less. If people in one social group feel instances of betrayal (where the receiver does not reciprocate) more keenly than those of another group, they will send less money, even though levels of trust are the same. Similarly, focal points or norms shared by individuals in a social group may vary across groups. These focal points determine the fair or right level of reciprocity to be displayed by receivers when senders have trusted them. In one society, the norm might be to respond to trust by “doubling down” and being very generous in response; in another society, the norm might be to respond to trust by treating it as a gift, and little reciprocity is expected. The same level of trust that the person will reciprocate according to the local norm will lead to very different measured responses, if the norms differ. Framing the trust game appears to matter a great deal for outcomes. Framing refers to the presentation, information, explanation, examples, experimental cues, and other aspects of the experimental setting. Framing is sometimes chosen deliberately by the experimenter, but also can be unintentional or understood in retrospect. Cronk (2007) found that when Maasai players in a standard trust game were “framed” with a rhetorically powerful, locally well-known, and complex concept called osotua, their transfers were significantly different from those of a control group. Combining preexisting differences and experimentally induced framing, Castro (2008) found quite different outcomes between Italian and English players in experimental prosocial games, and found that knowledge of the nationality of other players changed behavior. Cassar, d’Adda, and Grosjean (2013) found that framing trust games with prior games that manipulate an experimental

Models and Methods

79

“quality of market institutions” generated less trust when institutions were of lower quality. One popular technique to lower the costs to the experimenter of conducting trust games is to have players engage in multiple rounds and variants of the trust game, and then randomly select only one variant to be paid. Since players do not know which game will be selected, they still have incentives to play carefully. Of course, the more games that are played, the lower the chances any single game will be selected, and the less attention or thought players might devote to each game. Johnson and Mislin (2011) found that the randomization of rewards induced players to be more conservative. The more games that were played, the more conservatively players played, choosing sure payoffs over possibly risky payoffs where they depended on the receiver. The trust game has been played in hundreds of social settings around the world, and several meta-analyses have been published. Johnson and Mislin (2011, p. 875) reported on a meta-analysis of 162 trust games and concluded that “the large and negative coefficient on the Africa dummy [in regressions explaining trust and trustworthiness] is one of our more robust results.” In the African studies included in the meta-analysis, respondents on average sent less (trusted less) and returned less (were less trustworthy) than social groups from other regions. Table 3.3 lists various studies and their findings. African participants in the experiments may have differed on other dimensions (e.g., wealth, income, education, framing, context), and so the results must be taken as suggestive rather than strong evidence of differential trust. Nevertheless, African populations also typically have responded to standard survey questions in ways that suggest they are less trusting and less trustworthy. A number of studies set in Africa focus on gender differences in the trust game. Barr (2003) reported on twenty-four sessions of the trust game run in different communities in Zimbabwe. Each of 280 participants was endowed with 20 Zimbabwean dollars (approximately US$.80 at the time), equivalent to one or two days’ wages. The modal amount sent by the sender was Zim$10, and the mean was Zim$8.58. The amount sent was tripled, and the receiver then decided how much to send back. The mean proportion of the original amount sent that was sent back was Zim$1.28 (that is, about 30 percent more). Trust in the trustworthiness of the receiver paid off quite nicely for these communities. About half of the participants were female. Barr found that there was no difference between men and women for the initial amounts sent, but women receivers sent back less than male receivers. Greig and Bohnet (2008) similarly reported mixed evidence of gender differences in the trust game, as played by 268 men and women in the Kwa Reuben area of Embakasi Slum in Nairobi, Kenya. Both the 134 senders

Table 3.3 Trust and Trustworthiness in Experimental Games in Africa Author(s) of Study

Ashraf et al. Barra Burns Carter and Castillo Cronk Danielson and Holm Ensminger Etang et al. Etang et al. Greig and Bohnet Haile et al. Hargreaves-Heap et al. Holm and Danielson Jack Mosley and Verschoor Vollan

Year of Publication 2006 2003 2006 2011 2007 2007 2000 2010 2011 2008 2008 2009 2005 2009 2003 2007

Country

South Africa Zimbabwe South Africa South Africa Kenya Tanzania Kenya Cameroon Cameroon Kenya South Africa Uganda Tanzania Kenya Uganda South Africa

Sample Size 128 280 308 283 50 110 262 40 280 268 56 60 200 64 186 218

Source: Author calculations based on Johnson and Mislin (2011), appendix B. Note: a. Study by Barr not included in Johnson and Mislin (2011).

Amount Given to Sender (US$) 120.37 0.80 8.31 2.77 3.39 17.01 3.91 3.98 3.15 1.72 5.41 7.49 10.21 3.44 6.89 1.99

Average Percentage Sent (trust) 43 43 33 53 38 56 44 75 69 30 55 33 53 42 49 30

Average Percentage Returned (trustworthiness) 27 43 23 24 33 46 18 41 48 41 28 35 37 51 33 23

Models and Methods

81

and the 134 receivers were endowed with 50 Kenyan shillings (about $.65), corresponding to about one-third of a day’s wage. The amount sent by the senders was doubled. The amounts sent were extremely low by standards of the trust game: Ksh15, or 30 percent of the endowment was sent, and receivers returned Ksh12, or about 82 percent of the amount sent, and 41 percent of the amount received. The trustworthiness of the Kenyan slum residents was far lower than that of the Zimbabwean villagers (though many other factors, such as the framing or context, could have accounted for the differences). There was no statistically significant difference between the amounts sent by men and by women. Gender differences were larger for the decisions to return amounts sent. Greig and Bohnet had some sessions with same-sex and some with mixed-sex senders and receivers. Senders knew, then, the gender of receivers and receivers knew the gender of senders. The patterns were quite strong: “Female first movers were rewarded with more trustworthiness than male first movers by both male and female second movers . . . and female second movers returned smaller fractions than male second movers.” Greig and Bohnet concluded that people were more likely to display balanced reciprocity, returning the same amount the sender had sent, when the sender was a woman. This could have been due to a sense that since women on average were poorer than men in the slum area, they were needier and more deserving of being treated fairly. Most of the population in Africa lives in villages of a few thousand persons. In these communities, people see each other every day. Isolated in forests, savannas, and deserts, far from tarmac roads, they have to work together to produce local public goods. They have to be willing to contribute their own labor and resources to make their community a better place. There is, of course, always a temptation to take a free ride, to let others do the work of building a community and providing basic infrastructure and services (ensuring public order, guaranteeing local property rights, respecting promises and contracts, digging and cleaning wells, clearing paths and roads, hosting visiting politicians and dignitaries, celebrating community events). The willingness of people to contribute to public goods is hard to observe, but experimental methods have developed a standard public goods game. Individuals are invited to participate in an experiment. They are then told that they have been assigned to groups. Different experiments reveal different pieces of information about their groups, but for most games group members do not know each other. All they know is that their group is composed of people from their community. All members are then told that they are allocated an amount of money, say the local equivalent of US$20. They have a choice. They can keep the entire $20, or they can contribute part of the $20 to a group fund. Everything that group members allocate to Willingness to Contribute to Public Goods

82

Women and Development in Africa

the group fund will be multiplied by a number, typically two or three, and then divided by the number of group members (which must be larger than the multiplication factor). If there are six members, and each donates $10, the resulting $60 is multiplied by three, to yield $180, and then divided by six, to yield $30 for each member. The temptation for each member is to not contribute to the group fund. Note that unlike the trust game, there is no trust involved in the decision. Regardless of what a person thinks the others will do, a person’s best strategy is to not contribute anything. One hopes, however, that others will be public-spirited, and if one is public-spirited oneself then contributions are made. The game brings together, in a perhaps artificial way, the conflicting desires of personal gain and helping others. Public goods experiments conducted by Barr and Kinsey (2002) illustrated some of the complexity of gender and cooperation in an African setting. The research was conducted in Zimbabwe in 2000 with 308 individuals distributed across thirteen villages. Incomes in the area were lower than US$1 per day. In each village a group of about twenty-five to thirty men and women sat outside, some feet apart from each other. They played four rounds of the public goods game, with initial endowments of Zim$100, equivalent to about US$2. The first two rounds were private, with subjects circling on paper their contribution levels. The following two rounds were public, with individuals standing up and saying out loud what their contribution levels were. They did not, however, know who the members of their group were. After the third round, participants had a chance to publicly comment on the contributions of others. A research assistant would indicate each player in turn and say, “Player number . . . , Mr./Mrs. . . . , contributed $. . . . Does anyone have anything to say about that?” The criticisms or comments made were coded according to who made them and to whom they were directed. Interestingly, the results concerning whether men or women were more willing to contribute to the public goods fund were mixed. For each individual round of the public goods game, the difference between the mean contribution level of men and the mean level of women was not statistically significant. The standard deviation of contribution levels was very high; the average contribution levels were about Zim$50, or about half of the endowment, but the standard deviation was 30. Barr and Kinsey then used regression analysis to control for a variety of socioeconomic variables measured for each participant. Female participants were much younger, and male participants were from households with higher incomes. So these differences needed to be controlled for in the analysis. The regression analysis also pooled the four rounds together, in effect increasing the effective sample size. Still, women did not appear to have robustly contributed more. Some of the regressions indicated that the female dummy variable was statistically significant, but others did not. The effect was on the order of a 20 percent increase in contributions by women.

Models and Methods

83

Analysis of the effects of the public commentary on contribution levels, after the third round, likewise did not find significant gender differences in how people played the public goods game. Barr and Kinsey were interested in whether there were gender differences in responses to social or shame sanctions imposed by others, effectiveness of verbal sanctioning, the likelihood of not being sanctioned even when behaving in an antisocial manner, and the likelihood of sanctioning others. These hypotheses were tested by examining the regression coefficients on equations specifying, for example, how the number of times subjects were censured in the discussion affected their contribution level in the fourth round. It turned out that there were no strong gender differences except that men appeared to be less effective in changing behavior of those they criticized; it seemed that women’s words cut deeper. Greig and Bohnet (2009) conducted public goods games in the Kwa Reuben area of Embakasi Slum in Nairobi, Kenya. Some of the sessions were all female or all male, and others were mixed. Men and women contributed similar amounts in the same-sex sessions, but women contributed significantly less in mixed-group sessions. Greig and Bohnet had asked participants to report what they expected other subjects to contribute. Women had lower expectations than men did in mixed-sex groups. This expectation of low contributions of others, then, seemed a likely explanation for women’s lower contributions. The authors concluded, “These findings are consistent with social identity theory: lower status social groups (women) are more sensitive to social categories and changes in the situation because their fates are determined by their ability to adapt to situations controlled by the more powerful social group (men).” Further Reflections on RCT and Field Experiments The preceding examples introduce some key aspects of RCTs and field experiments and demonstrate why they can be effective tools for learning about human behavior and evaluating the impacts of programs and policies to promote development. Nevertheless, it should be remembered that RCT cannot be used to answer many questions of interest. For example, there is no RCT possible to shed light on the previous chapter’s discussion of the relationship between indicators of gender equality and economic growth at the country level. No experimenter can randomly assign different levels of gender equality to different countries. No experimenter can take millions of people in a society and assign them different gender roles to play. Likewise, field experiments cannot answer questions that are often important and interesting: When confronted with life-threatening choices, such as a famine, will men and women have different risk preferences? Will women be willing to risk death in order to have better lives for their

84

Women and Development in Africa

children? These choices are not amenable to study through RCT or field experiments. There are also a number of well-known methodological criticisms and problems of RCTs and field experiments; no social scientist argues that they are foolproof. It is important to remember that these criticisms typically apply to any other method of evaluation or interpretation of human behavior, and advocates of experimentation do not exclude or dismiss other sources of information about effects of programs and policies or behavior (introspection and theorizing, anecdotes, participant-observation, observational surveys of nonrandomly selected groups). The literature on these critiques of RCT is extensive, and only a few of the critiques are mentioned here (Barrett and Carter, 2010; Dahl Rasmussen, Malchow-Møller, and Barnebeck Andersen, 2011; Ehmke and Shogren, 2010; Ravallion, 2012). One first broad category of criticism deals with what is often called the “internal validity” of an experiment. That is, are there reasons why the experiment and analysis do not actually measure the hypothesized effects? Straightforward challenges to internal validity concern various biases in the data. For example, survey workers might be biased to show the program has a large effect, and so they deliberately, or subconsciously, skew answers to questions. They might be less willing to track down respondents who they think viewed the program unfavorably and have dropped out of the program. Another bias is that effects arise from the experimental setting itself rather than from the treatment. An RCT evaluation cannot usually distinguish between the effects of a treatment and the effects of being assigned to a treatment. Being treated with a program or policy is akin to being named, and the simple fact of being named differently may generate different outcomes for humans. Psychological experiments demonstrate that once groups of humans are named, they begin to experience behavior associated with sentiments of group solidarity. A double-blind RCT where the subjects did not know what “treatment” they were receiving would then be preferable, but then the real-world situation where a program or policy was not transparent would seem to be very artificial. Related issues about knowledge of the treatment assigned concern response bias, a problem common to any evaluation. If people know what the treatment is, and begin to reflect on and discuss what they think the evaluators would like to hear about the impact of the program, they may bias their answers and behavior in one direction or another. The “Hawthorne effect” familiar to sociologists, after a series of studies conducted at the Hawthorne Works factory in Illinois in the 1920s, suggests that workers (and people generally) typically respond positively to any sort of change when they know that an experimenter has chosen the change and that they are being monitored. A second category of criticism of RCT groups together arguments about what social scientists term the “external validity” of an experiment.

Models and Methods

85

In order to have random groups selected from within a population, and have the groups be representative of the population, and also be able to measure and record the effects of the treatments, the usual RCT involves a delimited geographic area and a small sample (often of several hundred persons). Only replications of the same RCT for different populations can enable a researcher to be confident that the effects of a program or policy would be generally applicable to a large population of a country or region. A third category of criticism of RCTs concerns the ethics of random assignment into programs. An important difference between human and animal experiments is that social scientists working in universities are under ethical obligations to have subjects give informed consent before participating in experiments. Gone are the days when a sociologist could hire actors to pretend a murder was being committed in order to observe people’s reactions. The basic ethical precept often clashes, however, with a fundamental assumption of any experiment, that the control group measures what would have happened in the absence of an experiment. With informed consent, the control group knows about the experiment. Their behavior may be influenced by this knowledge. Moreover, many RCTs deliberately randomize treatments, even when researchers could at fairly low cost determine which persons or households are more likely to benefit from a program. In real programs, people often have to opt in to a program. But in many RCTs, the treatment group includes people who likely would not have opted in to receive the program. To be fair to RCT proponents, many of these criticisms apply to other forms of statistical and anthropological investigations. Response-bias, ethics, and external validity are endemic in the social sciences. No single approach can satisfy all methodological concerns.

4 The Land Tenure Rights of African Women

We have found that the majority of widows suffer attempts by neighbours or relatives to grab their land—and in a non-industrialised country with no urban jobs, having nowhere to farm means starving, children having no education and the sick going untreated. Almost all divorcees and separated women are denied land by their own brothers, and live, often with children to support, in misery and destitution. Land grabbing is always aggressive and unimaginably abusive, and sometimes violent. —Judy Adoko and Simon Levine, “Land and Equity Movement in Uganda,” The Guardian, March 26, 2009

A woman has to leave her lineage land behind when she marries. . . . She is no longer considered to be part of her own family. . . . In the case of a man, however . . . even if he leaves town or the village . . . [he] is still the heir of the land. —Woman farmer in Ho district, Ghana

ore than 500 million people in sub-Saharan Africa live in rural villages, growing crops and raising livestock. Like farmers everywhere, rural people in Africa have intense and sometimes mystical ties to their land. Daily conversations revolve around how land is used and who gets to use land. The constellation of these conversations generates the norms, rules, and discourses about land in a community. This is what is meant by the term land tenure. The word tenure is derived from the French word tenir (to hold), so land tenure refers to “who gets to hold the land.” For agrarian societies, land tenure rules are among the most important social institutions determining economic activity and individual well-being. In South and Central America, land has been held by large landowners in

M

87

88

Women and Development in Africa

the latifundia system put in place by the conquistadors. Landless workers and sharecroppers turn over a considerable portion of their production to the landowner. This dispossession has led to cycles of revolt and oppression. In Europe and the United States, the enclosure movement of the 1800s privatized common forests and grazing areas, and rural land came to be rented, mortgaged, and sold anonymously in a market system of private property. In most African countries, formal land legislation had only modest influence on local land tenure rules, with the notable exception of countries such as Kenya and Zimbabwe where white settlers dispossessed black owners during the colonial period. Several factors, however, meant that by 2000 land tenure became an important national issue, affecting village land tenure. Population growth and rising prices of agricultural commodities roused interest in securing access to land (Becker, 2013; Ubink, 2008; Yaro, 2012). Discourses about land were deployed in electoral campaigns, as politicians were attracted to an easy equation of promising land rights to supporters and dispossession to opponents (Boone, 2009, 2013). Ethnic groups and lineages, and their leaders, also attempted to consolidate ownership. New residents in villages, meanwhile, have been trying to promote local discourses of “hospitality” that strengthened their rights to borrow land from traditional owners. At the other end of the spectrum of land users, well-capitalized agribusiness firms have been lobbying for sweeping changes in national land tenure, while simultaneously striking deals with customary authorities to obtain use rights to large tracts of farmland (Amanor, 2012; Borras et al., 2011; Koopman and Faye, 2012). In the typical African setting, the vocabulary of local land tenure, the actual relationships of people to the land, and the processes of dispute resolution are profoundly gendered. There are, moreover, gendered valences for the crops that might be grown on the land. The actual relationships of people to land include women only tangentially; in the vast majority of ethnic groups men control land. The processes for adjudicating disputes rarely involve women in positions of authority. Given women’s marginalized position in land tenure, a vocabulary of “grabbing” has developed in the popular imagination. Relatives “grab” land from widows when their husbands die (Peterman, 2012), and agribusinesses “grab” common village land used by women to collect fuelwood, wild plants, and tree products (Kachika, 2010). Stylized View of Land Tenure for Men and Women in Africa A young man, growing up on the homestead of his father, might ask for a small field to cultivate on his own account. His father might give him a plot Land Tenure for Men

The Land Tenure Rights of African Women

89

of land, but he might also refuse, saying that his son should continue to work on the family farm. The young man might then approach another villager, asking for a field. A neighbor might give him an abandoned field. The village chief, or perhaps an “earth priest” or a lineage elder, might give him a larger field, especially if the young man is married. This larger field may be out in the forest, and the young man might call upon his friends to help clear the trees and scrub. After harvest, the young man might give back a portion of the crop as compensation to the ancestors or as a token rent out of respect. Slowly, and depending on the social status of the young man, the field that he farms comes to be “his.” He owns the land, though not in the Western sense of freehold tenure. He cannot sell the land. He may be enjoined from transferring the land to another person, even a villager. There may be restrictions on planting trees. He may have to allow pastoralists to graze their cattle, sheep, and camels on stubble. When the young man’s father dies, the sons of the family might continue to cultivate the father’s farm. After some time they will divide up the fields among themselves. The man now has use rights over a substantial number of fields. As his life cycle progresses, he may take advantage of a nascent rental market and lease out his extra land to other villagers or new settlers. He might buck village rules and sell some land down by a stream for a town merchant to establish an irrigated garden. At some point in his life, his rights over land will be challenged, and he will have to go to the chief, earth priest, or lineage head to have the matter resolved. He will call upon his friends and allies in the village for a show of solidarity. Elders friendly to his family will come forward to testify on his behalf. Magic might be summoned. He might consider approaching the district administrator. Very rarely, he might go to a town court. If the young man lives in Kenya, Uganda, or on an irrigation scheme, or in one of many official resettlement communities throughout the continent, he may possess a land title issued by a government agency. The district office might have a land register, with his father’s name listed as owner. Official documents and local testimony, with complex negotiations and exchanges of favors in the background, including outright bribes, are ingredients in the outcome of the dispute. Quite often, in the Africa of the twenty-first century, outcomes of disputes are not written down, and so possibilities for renegotiations open up within a few years as circumstances change. The stylized description of land tenure presented above covers a lot of the basics of land tenure in Africa (Bassett and Crummey, 1993; Berry, 1993; Downs and Reyna, 1988; Jacob, 2007; Lavigne-Delville et al., 2002; Peters, 2009). Land is not usually titled, and only a few countries have conducted cadastres. Rental markets are common, though rental prices are more customary than market-determined. Sales markets are inactive and

90

Women and Development in Africa

often explicitly prohibited by both local and national laws. Africans gain, maintain, or lose access to land through multiple paths: as members in social groups or networks, through labor and investment, and sometimes through purchase. Deliberately, the emphasis in the preceding paragraphs was on men. A young woman in the typical African village has a very different story. Her father might give her a very small field to cultivate, and so might her husband when she marries. Nobody else will give her a field. Any field she farms never becomes “hers.” She can neither rent nor sell the land she cultivates. As a pithy saying has it, “Women are owners of crops and not owners of fields.” There is, of course, more complexity to the land tenure status of African women than this basic description. Most societies fall into one of six systems, detailed in the remainder of this section. The first two systems described are ones in which women have rights to land that allow them to alienate (sell) or allocate land. These are the exceptions. More significant and widespread are systems where women have few rights. In these systems, women’s access to land comes through their social ties to kin and husbands. Land Tenure for Women

In some regions, Islamic norms of inheritance have codified women’s rights to inherit and transfer land. The usual Islamic inheritance is for daughters to inherit shares half the size of their brothers’, the reasoning being that the daughter is provided for by her husband, while the son has to provide for his wife. Regions where Islamic norms hold sway include Somalia, Eritrea, Sudan, coastal areas of East Africa, and northern regions of many Sahelian countries (Chad, Niger, Mali, Senegal). Northern Nigeria also has a strong tradition of applying Islamic law, known as sharia, to land matters. Caplan’s (1975) study of inheritance rights in a Swahili village on Mafia Island off the Tanzania coast is a good example of how Islamic law can provide women with strong rights. Women on the island have rights to both bushland and higher-value coconut land. A large and dispersed descent group owns the land, but this ownership amounts to no more than an ability to exercise veto power over sales of land. Rights that are valuable are those held in coconut trees. On the death of her father, a daughter inherits a share of the plantation equal to one-half of her brother’s. Husbands and wives also inherit from each other; a woman receives one-eighth of her husband’s property, while a husband may receive a quarter of his wife’s property. Women are more likely to inherit coconut trees from their husbands than vice versa because they tend to live longer and marry younger. Valuable coAreas Influenced by Islamic Inheritance Laws

The Land Tenure Rights of African Women

91

conut groves are also presented to women as marriage payment or compensation on divorce. Daughters retain rights to their groves even when they have married out of their kin group. The power of Islamic norms of inheritance can also be seen in the Gezira area of Sudan, site of one of the largest irrigation schemes in the world (Barnett, 1977). During the establishment of the scheme in the 1920s, and throughout subsequent extensions, land was confiscated under national domain laws. The scheme and extensions then redistributed land in the form of semipermanent tenancies (a right to cultivate the irrigated field for an indefinite period, revocable by the schemes). The schemes gave these tenancies only to men. Over time, though, women regained their access to land through Islamic inheritance rules. Bernal (1988) found that in one irrigation scheme, about 15 percent of tenancies had reverted to women in just a few decades. Being Muslim, of course, does not mean that a society will adhere to either the spirit or the letter of orthodox Islamic inheritance laws. Most Muslim societies in West Africa do not follow principles of Islamic inheritance when allocating land. Hill (1970) found that among the Muslim Hausa, “when a father dies, then, his manured farms are usually divided between his sons. . . . Lip service only is paid to the rules of Muslim inheritance. . . . In so far as daughters receive shares, they often proceed to sell them to a brother” (p. 148). Holý (1974) describes how the Berti, a Muslim society in western Sudan, prevented women from inheriting valuable land planted with gum arabic trees. These general patterns of limited inheritance and vulnerability as widows are common throughout the Islamic-influenced regions of Sahelian Africa (Monimart and Tan, 2011; Van de Walle, 2013; Lambert and Rossi, 2014). There is some evidence that the high fertility rates of the Sahel may, in fact, be due to vulnerability in landownership, despite the formal protections of sharia law. Women seek security for their old age and likely widowhood (because they marry young) by having more children. There are many different family systems in Africa. When people say who their family is, they may refer to the blood relatives of their mother. Such a society is said to be matrilineal. If they speak of their family as being the blood relatives of their father, it is patrilineal. Most people in Western societies think of both sides as their family; they practice bilateral descent. Family and descent matter in part because families often own property collectively. Being a member of a family gives one rights to the land of the family. Even when property is not owned collectively, families matter because they may have rules of inheritance that specify how property is transmitted across the generations. Many societies prohibit property from being Matrilineal Areas Sometimes Give Strong Rights to Women

92

Women and Development in Africa

inherited outside of the family. Only members of the matrilineage, for example, may inherit the property of a deceased member of the matriline. Patrilineages often restrict landownership to men in the patriline. Matrilineages sometimes have men inheriting, sometimes women. Central Africa has a large number of matrilineal societies that privilege women as owners of land. These have been well described by anthropologists. Grundfest Schoepf (1985) writes that among the Lemba of Zaire, “Men say: we live at the homes of the women; the land here belongs to women; women have a say in everything we do; we listen to the opinions of women; and so on” (p. 5). Women live in their natal villages after marriage, allow their husbands to use their land, and pass the land on to their children. In nearby southern Malawi, Davison (1988) notes, “the matrilineal banja household centers on the woman who has primary rights to land through her mbumba lineage” (p. 184). The residents of Zomba District in Malawi also have the same practice of women owning the land and husbands coming to live with them (Hirschmann and Vaughan, 1984). Muntemba (1982) similarly finds that among the matrilineal Tonga and Lenje of Zambia, women who were married and lived in their husbands’ villages retained access to land belonging to their maternal relatives’ villages. They could inherit and own property. Poewe (1981) describes the Luapula of Zambia where women had exceptional control over land. Keller, Phiri, and Milimo (1990) summarize the early anthropological literature for the Bemba, Tonga, and Lozi, which illustrates how some of the women’s land could be passed on to their heirs, while some was controlled by their husbands. A final example comes from the Uluguru Mountains in Tanzania, where van Donge (1993) found that women were just as likely as men to own land and participate in disputes over land—on their own behalf or on behalf of their matriline. Similar rights are sometimes found outside of Central Africa. Women in West Africa often have strong customary claims on land they have cleared for low-lying rice swamps. In some parts of Gambia, women had the right to pass this land on to their daughters and considered themselves its owners (Dey, 1981). In southwestern Burkina Faso, Goin women owned and inherited rice swamps (Dacher and Lallemand, 1992), and in Sierra Leone women could appeal to their natal kin for rice fields for their exclusive use (Leach, 1992). Although the Islamic system and some matrilineal societies give women strong rights, it is far more common to find that women’s rights are limited and secondary, as described in the beginning of the chapter. In West Africa, where women usually have very limited rights to cultivate on their own account, growing land scarcity and concentration of ownership are shrinking Sahelian West Africa

The Land Tenure Rights of African Women

93

their allotments (Roos and Gladwin, 2000; Stone and Stone, 2000). Women gain land chiefly through marriage. A newly married woman will be granted land from her husband’s lineage. These use rights are precarious and contingent. On divorce, widowhood, or relocation, women generally lose these rights. Single younger women rarely have rights in land. Widows may have rights to small plots in their natal community. Typically, neither the local rules of tenure of particular ethnic groups nor the state’s rather unsystematic interventions give women standing as holders of rights to land. The major published interpretation of “customary law” of the Mossi, the majority ethnic group of Burkina Faso, stated quite clearly that women never have a permanent land right (Pageard, 1969, pp. 230, 248, 256). For the Mossi and most other ethnic groups of Burkina Faso, the only conceivable case of a woman having direct control over land was a widow who might assume responsibility for the land of her deceased husband on a temporary basis, while her own status is sorted out. She may be married to a male relative of her husband under the levirate system. To add insult to injury, most studies from Burkina Faso also have found that personal fields granted to women by their husbands, as part of the marriage contract, are invariably small and often of low fertility (Kevane and Gray, 1999). Another frequently cited system in which a woman obtains use rights to land upon marriage is the East African “house-property complex.” A man allocates to each of his wives cattle, farmland, and homestead land. While cattle would invariably remain under the control of the husband, wives would frequently be the “managers” of the land allocated to them, with much control over day-to-day decisionmaking. In the house-property complex, a husband cannot alienate the land without the wife’s permission. She would insist on her “veto” right, or at least on her right to subsequent compensation, since her children—and not the children of any co-wife or the members of any lineage—would inherit the land on the husband’s (their father’s) death. Her children would be providing her old-age security. Some key references are Gluckman (1950), writing about the Zulu in South Africa, and Gray and Gulliver (1964), Moore (1986), Håkansson (1994), Davison (1988), Oboler (1985), von Bulow (1992), Okeyo (1980), and Karp (1978), writing on Kenya and Tanzania. The system extends into some ethnic groups in Uganda as well (Khadiagala, 2002). Oboler (1986) observed that widows in Nandi society, though they had few options for remarriage, were better off than women from other Kenyan groups because of the house-property system. At the time of marriage, a woman received a marriage settlement in the form of her husband’s property that would belong to her descendants, even if she were subsequently East Africa: The House-Property System

94

Women and Development in Africa

divorced (as long as she had sons). Oboler further suggested that “house property means that a widow can have secure property rights without maintaining an ongoing relationship with any members of her husband’s family.” These secure rights are temporary and contingent in an unusual way. Instead of depending on the goodwill of a husband, a woman depends on having sons, and then depends on the goodwill of her son. If a widow has no sons, then her deceased husband’s patrilineal kin inherit his land and cattle (LeVine and LeVine, 1966). As Glazier (1985) put it, “Inheritance is thus patrilineal, although rights to property are passed on through women who, paradoxically . . . cannot own productive property outright” (p. 111). Her sons will claim the land when they are older, and they will have the power to allocate land to their wives. Not all groups in East Africa have even the limited rights of the houseproperty system. Moore (1986) found that women had no independent rights to land in the Chagga areas of Mount Kilimanjaro in Tanzania. Smith and Stevens (1988) described another Tanzanian group, the Haya, located in a coffee-producing area near Mount Kilimanjaro. Women appeared to be systematically excluded by local men and the state from independent access to land. For example, female household heads who were divorced could not use the land from the clan of their ex-husband—land they may have been farming and continued to farm—to plant coffee. Women seemed never to inherit land from their fathers, and their status when widowed was precarious. Southern Africa presents a complex picture in terms of tenure. Large-scale confiscation of land by settlers disrupted any “traditional” tenure systems that may have existed. Numerous legal scholars have shown how the traditional land laws codified during the colonial period were in fact products of collaborations between self-interested chiefs and administrators (Chanock, 1985). Several studies have characterized the complex semiformal systems of customary tenure practiced in so-called communal areas of southern Africa (Goebel, 1999, 2005, 2012). Fortmann and Nabane (1992), for example, described tenure in the Shona and Ndebele areas of Zimbabwe in which local authorities adjudicated according to local rather than national rules. The system they described was much like the West African system. Women received land through membership in patrilineages, according to their status as wives or as daughters. Daughters obtained land from male lineage heads, who in turn obtained land from chiefs and headmen. Daughters’ production on their fields would be exchanged for materials for marriage, their trousseau. Married women obtained land rights from their husbands. These rights were maintained as long as the marriage endured. Southern Africa

The Land Tenure Rights of African Women

95

Cheater (1982) painted a different picture, highlighting the central role of female tenure in discussing the land question in Zimbabwe. She showed that a high proportion of women had usufruct rights, and argued that the model of traditional tenure in communal and freehold areas was falsified by on-the-ground reality. In the standard, accepted model, women had no direct rights, “only temporary usufruct within the lineage system through their husbands or male patrikin” (Cheater, 1990, p. 191). In reality, Cheater argued, “women as affines would appear to be using their importance as farm workers to acquire . . . land in their own right” (1982, pp. 84–85). Her data from the freehold area of Msengezi showed that the number of women holding usufruct rights increased substantially from 1970 to 1980. Additionally, she contended that the rights of widows were considerably stronger than elsewhere. Though not recognized in either customary or formal law, in practice a widow was usually guaranteed rights to manage her husband’s land until she chose to relinquish control to the heirs. Finally, Cheater presented a most interesting newspaper account of a land dispute involving a man who had purchased land from a woman and the seller herself. The woman claimed she had not intended to sell, and that the land had been inherited from her mother-in-law. About this Cheater could only exclaim in wonder: “Legitimation of the rules, in the reported views of civil servants, comes from legislation, not tradition; the ‘illegality’ of purchasing land by men is trumped by the ownership of land by women, and its transmission—well beyond any rules of the family or matri-estate . . . between female affines” (1990, p. 195). In a system where women supposedly had little control over land, this example showed that they were not only owning land, they were selling it, and, even stranger, they were inheriting it from their mothers-in-law. Possibly one should take a very humble attitude about the state of knowledge of women’s land rights in southern Africa. This ambiguity and complexity were echoed by Paradza (2011), who intensively studied two villages in Zimbabwe. Paradza noted that the significant socioeconomic upheavals of Zimbabwe during the 1990s had created high levels of migration and mortality. Land, whose owners or users had departed, was constantly contested. This matongo (vacant) land was an important resource for single women farmers. More than a century ago, farmers in Ghana, Côte-d’Ivoire, and Nigeria responded to the enormously lucrative opportunity to grow cocoa and coffee trees in the moist hillsides of the region. Vast areas of forest were transformed by migrant farmers into complex agro-forestry systems of cocoa and maize cropping. Land tenure arrangements governing these new fields were complex. Hill (1963) found that patrilineal migrants in Ghana organized land purchases through “companies” that were fronts for comparaCocoa Areas of West Africa

96

Women and Development in Africa

tively individualized ownership rights. Matrilineal migrants, on the other hand, more typically obtained land in the name of the matriarchy. As for women in the company areas, Hill (1963, p. 117) assessed the situation in the following manner: It is certainly clear that many of [the women] suffer from many disabilities in the extent of their individual control. It is unusual for a woman to be a company member in her own right, and it is usually only those fathers who have no sons who contemplate buying land for their daughters. A woman who possesses the status of an original farmer has as much control over her property as a man would have, and her children may inherit. But a great proportion of the women farmers in whose names company farms have been registered by the Ministry of Agriculture are daughters who are representing their non-resident brothers, or wives who have been granted usufructuary rights by their husband—the farms reverting to their husbands’ sons on their deaths. Occasionally daughters as well as sons inherit on the death of their father, but it seems that such daughters are nearly always unmarried.

For matrilineal areas there has been more disagreement. Berry, for instance, commented that “the proportion of tree-crop owners who are women is relatively low. . . . The chief exception is in areas of Ghana where descent is reckoned matrilineally. . . . In some communities, fifty percent or more of the women own tree crops” (1988, p. 6). Mikell, on the contrary, found that despite the expectations that “matrilineal systems were more liberal in granting rights to females,” there was little difference in practice between matrilineal and patrilineal groups in giving women access to land to engage in cocoa farming (1984, p. 195). Women received few rights in both groups. In both patrilineal and matrilineal systems, women have owned considerably less area than men. Oppong, Okali, and Houghton (1975) found that in cocoa communities in southern Ghana in the early 1970s, women owned cocoa farms quite frequently, though men held double the area held by women. Mikell (1989) reported that women in the Sunyani District constituted approximately one-eighth of cocoa farmers, and again the sizes of women’s cocoa farms were much smaller than those of men; the median size of women’s farms was roughly 8 acres, much smaller than men’s median 160 acres. Even these relatively insignificant and secondary rights have been eroded over time through an inexorable logic of male domination. Tambiah observed that “when ‘permanent’ cash crops like cocoa or tea have transformed the land on which they grow into patrimonial properties capable of being owned and transferred by owners by sale or inheritance, there too men have usually exercised the rights of ownership and excluded women (even where so-called matrilineal systems of kinship have prevailed)” (1989, p. 416). Asare (1995) found that among the matrilineal Akan, the

The Land Tenure Rights of African Women

97

dominant ethnic group of the cocoa area in Ghana, women fared poorly. Male relatives in the matriliny were merciless in evicting women who had helped their husbands establish farms. Mikell (1984) developed a careful analysis of this erosion of women’s rights. Brong women in Ghana were successful cocoa farmers during the first half of the century, even though their farms were smaller than those of their male counterparts. But by the 1960s, women had retreated from the market, and during the 1970s the only women with cocoa farms were elderly women who had obtained their farms during the earlier period. Mikell observed that in Brong areas women were less interested in working on their own fields because their own matriline would prevent their daughters from inheriting. This loss of rights occurred even as men increasingly gained the right to designate an inheritor. Mikell’s interpretation of this decline was that there was an apparent loss in a struggle over “meanings”: Land did not exist as a sex-linked good before 1900, and the stool [symbol of chieftaincy] was the custodian of it. When land did begin to generate produce and raw materials intended for international markets, males tended to acquire it and pass it on to males. The females who began to acquire and control farm land after 1920 considered it their property (thus female property) and desired to pass it on to daughters and sisters . . . they did not list “brother” as the desired inheritor and seldom listed “son.” Yet it is clear that the transmission of cocoa farms to female offspring was not taking place. (1984, p. 209)

Men increasingly were successful in defining land as “self-acquired” and outside the purview of the lineage. Matrilineages found unpersuasive the same argument when presented by women, that land they farmed was “female property.” Men escaped the claims of the corporate body, while women found themselves more incorporated (Grier, 1992). Men seemed to adopt the view that their old-age security was better served by having their sons inherit rather than their lineage group. Cocoa farms purchased by Brong males were inherited by their sons rather than going to the matrilineage, or abusua. Women likewise seemed to want to pass land on to their sisters or daughters. Instead of arguing that these farms were self-acquired, they argued that they were female property, like movable property such as cattle or money that had always been inherited by women from their mothers or sisters. The women failed. Women’s farms increasingly went directly to the abusua and generally went to males. Women, Mikell argued, became less likely to invest in cocoa farms because their farms were not “treated as female property but . . . as abusua property transferred to men following the death of their sisters” (1984, p. 212). Elsewhere, Mikell (1989, p. 125) argued that in the late 1960s, “the court was silent on the issue of whether female generated property belongs

98

Women and Development in Africa

to the lineage and therefore to abusua-controlled positions, or to individual female heirs.” Okali (1983) confirmed this interpretation of the decline or erosion in the content of status of women’s tenure. Court cases were favoring the rights of men to pass their farms on to their sons, and disfavoring the rights of women to exercise independent rights over their own land. Bukh’s (1979) study suggested that the same process has been at work among patrilineal Ewe cocoa farmers. The patrilineage used to provide all women with a parcel of land to cultivate, but women were increasingly having to ask permission from men to use their fallow land, or go about “begging” land, borrowing land from men who did not belong to the women’s lineage. Duncan (2010) confirmed that the patterns described did not change much in subsequent decades, noting moreover that “contract marriages” had increased in frequency. These were temporary seasonal marriages; the two parties exchanged land and labor for the duration of a cocoa season, and then terminated the relationship at the end of the season. Land tenure is a complex amalgam of formal and informal institutions. Legal pluralism is the norm. Tripp (2004), to take one example, summarized the complex tenure situation in Uganda, where there were four different and overlapping tenure regimes: (1) freehold, with registration through government and full ownership rights; (2) leasehold tenure, with considerable constraints on land use; (3) mailo land tenure, which originated in the Land Law of 1908 and gave freehold leasehold titles in perpetuity for large tracts of land to the Baganda king, his family, clan chiefs, and about 4,000 individuals; and (4) customary tenure, with a variety of rules and arbitrated by clan leaders. Needless to say, the situation was not a healthy land tenure policy situation. Each of the four different tenure systems had its own rules concerning transfers and inheritance, often applicable to the same piece of land, and with disputes adjudicated in multiple venues with different actors, rules of procedure, and principles of justice. Writing about land in Uganda, Rugadya (2010, p. 1) noted that “in all land tenure systems women are excluded from owning land, and only retain secondary rights.” Local land tenure has remained an oral affair, with few societies having written down rules for inheritance, transfer, and rights of possession. Land tenure rules have persisted as the oral tradition of elders and shared discourses of members of ethnic groups. These discourses comprise a vocabulary for describing the proper relationships of people to land, a shared understanding of what kinds of land relationships exist in practice, and processes for adjudicating disputes. Recapitulation

The Land Tenure Rights of African Women

99

Change in Land Tenure Systems The discussion of cocoa farms exemplifies how land tenure is constantly evolving, especially as population densities rise and the products of the soil become relatively more or less valuable. Two broad trends are clearly visible on the African tenure landscape. First, property rights to land are becoming more individualized, and a market for land rental and sales is emerging in many regions (Kevane, 1997; Lund, 2001; Otsuka and Place, 2014; Platteau, 1996). Ironically, this is often not in areas that have seen expensive projects to develop cadastres and maintain title registers. Indeed, many researchers conclude that poorly run titling projects impede, rather than facilitate, the emergence of individual property rights. The titling process repeatedly seems to generate insecurity rather than tenure security (Dickerman and Barnes, 1989; Firmin-Sellers, 2000). Private landholding and individual transfer are instead surfacing in areas untouched by the heavy hand of centralized schemes to transform tenure. The spontaneous emergence of new tenure rights may be desirable, in that the process embeds the new property rights in the existing web of social relations and thus carries more legitimacy. However, new privatization of land may involve stark expropriation of marginal or secondary holders of land rights. The second broad trend is partly a reaction to the pressures toward individualization of tenure and has been accelerated by government support for large-scale acquisition of land by corporate agribusiness. In many areas, local jural groups are reasserting and strengthening their rights to communal ownership of land. Traditional chiefs and earth priests recognize their powerful positions as arbiters of land conflicts within and between groups, and so form an influential body of citizens interested in maintaining notions of collective ownership (Berry, 2000; Toulmin and Quan, 2000). If land becomes individualized, the role of chiefs and earth priests becomes an anachronism. Within the push and play of these forces surrounding individualization and communalization, women as holders of land rights are subject to a different set of pressures and strategies. The discussion of change in the cocoa areas of West Africa suggested several processes by which structures of land tenure changed for women. One process is that in which men and women negotiate definitions and meanings that are applied to land tenure. As the balance of power shifts from women to men, or as circumstances change, definitions and meanings are manipulated to favor one group over the other. Crops that previously “belonged” to women are redefined as household (read, male) crops. Several authors have studied the effect of irrigated rice cultivation on women’s access to land in Gambia, where this process of redefining tenure was particularly stark. In Gambia, farmers recognize both common and individual

100

Women and Development in Africa

land rights. Women historically controlled rice fields that they cleared with their own labor. Their rights to this land were well defined: they controlled the production from this land, but more significantly they controlled the right to transfer land, which they generally did, to their daughters. The studies of different irrigation schemes illustrate how women’s access to land changed when the irrigation projects changed the landscape. Brautigam (1992) argued that because the rice land was cleared and developed by malecentered development projects, local men were able to claim land as their personal plots. Some of these plots were categorized as household property that came under the control of male household heads. Inputs and mechanized services were allocated overwhelmingly to men. Carney (1988) described how women’s rights to irrigated rice land evolved in the Jahaly Pacharr irrigation project, which was targeted to benefit women farmers ignored by previous rice projects. About 13 percent of the irrigated rice land was registered in women’s names, but Carney noted that “even though the land was registered in women’s names, none of the pump-irrigated plots were considered their individual (kamanyango) fields. Irrigated plots throughout the project were designated by both men and women as maruo, or compound land” (1988, p. 71). Women’s access to land was reshaped by redefining the meanings of the categories by which their access traditionally had been allocated. The new projects allocated land to women, but that did not mean that women had the power to control the land. Irrigated crops turned out to be different kinds of crops, and control over land was linked to the crop cultivated, rather than to a spatial concept of ownership. A second process that alters the terms of access to land rights is one in which the terms of marriage contracts are changed. Consider a stylized model of marriage, developed by Akerlof, Yellen, and Katz (1996), which may be adapted to a sub-Saharan setting. In this model, men and women decide whether or not to agree to marry. Assume that a woman can only marry once, but she can make offers of marriage to single men until an offer is accepted. She would like to marry a man who will promise to give her part of his land. Only men can own land, and husbands decide whether women can manage fields on their own. The woman is like a sharecropper, in some sense, and has only so many chances to find a landlord with whom to begin a long-term relationship. The problem is that some men would do better if they farmed their fields on their own and did not give any land to their wives. So the woman may want to extract a promise from the potential husband that he will indeed give her a plot of land. The man, of course, would much rather not make such a promise: Who knows whether his wife will turn out to be a good farmer or not? Perhaps the land will be wasted and lie underutilized. When asking for a promise to give land, the woman risks that the man will simply refuse and go try his luck in the marriage “market” with another potential partner. Assume that promises

The Land Tenure Rights of African Women

101

are enforceable—the man cannot renege on the promise. The woman has to decide whether to request such a promise and risk refusal by the man. The woman’s best strategy depends on what she thinks other women are doing. If most other women demand promises to receive land, then the woman will also make such a demand. If the man considering the offer refuses and goes off to entertain offers from other women, he will find they are all making the same demand. He may as well have accepted the initial offer. Expecting this to be the case, he will indeed accept a demand for land at marriage. When many women do not demand promises, this undermines the willingness of men to make promises and marry, since they revise their expectations about the likelihood that they will meet other women who do not extract promises of receiving land. The woman has the first decision of whether to insist that the man must give her land if they marry. If the woman insists that he give her land, and he agrees to be married under this condition, the man’s income is possibly affected by the cost of not allocating land optimally, but rather having the wife control some significant amount of the land. If the woman does not insist, the man of course marries (women are assumed identical in all other respects). He can do no better by waiting. The woman gets a benefit from marriage, but this comes at a cost. The cost of marriage is being tied down, as it were, as the wife (with children) and foregoing opportunities for entrepreneurship. Women have many income-generating activities (selling food and drink in the marketplace, for instance, rather than preparing it for a husband and children), and some of the more lucrative activities involve travel, which may be proscribed after marriage and children. When the man gives the woman land, part of this cost is offset by the advantage of having an opportunity to farm independently. There are then a number of possible sequences of actions. The question to ask is whether one set of strategies for men and women is mutually consistent, and so constitutes an equilibrium. That is, given that the actions of others determine the probability of getting married if a current “match” does not result in marriage, will men and women settle on a particular sequence? The discussion above suggested that the sequence of play that has the women insisting on a promise of land, and the men granting the promise, can be an equilibrium in the marriage market. In this formulation, a norm is an equilibrium in which people are observed following similar strategies because they assume that other people will follow the same strategy. The equilibrium norm may change if any of the payoffs change. Suppose, for instance, that the initial equilibrium is the one where men and women marry with promises that men give out land. Suppose further that women have different costs of foregoing entrepreneurship. Some women have no talent or interest in working outside the home; others do. It may then arise that increasing commercialization lowers, for entrepreneurial

102

Women and Development in Africa

women, the opportunity cost of being married. They no longer have to travel to be entrepreneurs because there are increasing opportunities in their own villages or towns. This leads them to be willing to marry without promises of land. Other women, who may have no intention of being entrepreneurs, find that if they continue to insist on promises of land, they are more often rejected in the marriage market. If the entrepreneurial women constitute a significant fraction of the population, then all women will be willing to marry without promises of land. A change that affects a fraction of women—those entrepreneurs who could take advantage of new commercial opportunities—potentially impoverishes all women. This model is instructive for understanding the Gusii case in Kenya (Håkansson, 1986). At one time a woman marrying was entitled to ownership or at least clear control over house property. Over the 1970s and 1980s, however, men increasingly refused to honor this right; they insisted on marrying outside the traditional customs. Men seemed able to negotiate more flexibility in the matter of disposing of their land. The new “elopement” marriage came with no implicit right to property. Elopement meant that until bridewealth was paid, the marriage had no legal status and women had no rights to remain and cultivate the land of their husbands. As the men said, women could be “chased away” if they did not work hard. As more and more men demanded this advantageous form of marriage, women either reluctantly agreed or else refused marriage propositions in the hopes of finally attracting a suitor willing to marry under the old terms. As more men demanded the advantageous marriage, the chance of finding other men fell, and women came to accept the new terms. This, of course, made men even less willing to offer the old terms; they knew they could easily find a woman willing to marry the new way. While in the 1960s only 26 percent of women eloped, in the 1980s Håkansson found that 87 percent had eloped. Andre and Platteau found a similar process under way in a Rwandan village in the decade before 1994, where “roughly two-third of the couples in N. have been married without inkwano [customary payment], and the proportion is obviously much higher among young couples” (1998, pp. 32– 33). The effect was to give women little bargaining power in the marriage, and even less with respect to their own lineage should they divorce. The model and experiences of Kenya and Rwanda suggest how processes can be self-reinforcing and can snowball. As many men begin to marry in more informal ways, without promising to give land, a “tipping point” may be reached. Even relatively unattractive men (in terms of their role as potential marriage partners) can refuse to marry formally and thus not grant their wives land. Unmarried women will accept these terms because their prospects of obtaining better terms in the marriage market are dimmed.

The Land Tenure Rights of African Women

103

These “thick” descriptions of gendered local processes that have resulted in changes in land tenure have not been the subject of much rigorous quantification. The importance and extent of the processes remain somewhat speculative. Land Tenure Policy In most African countries, there has been perennial interest in reforming tenure to strengthen the institutions that secure individual and communal land rights (Cotula, 2012; Deininger and Feder, 2009; Platteau, 2000; Shipton, 1988). Three arguments are offered for tenure reform. First, informal tenure may result in an excess of land disputatiousness, and the constant contestation over rights reduces the incentives of farmers to invest in improving their farmland. Why invest a lot if a politically connected neighbor can lay claim to your land? Tenure insecurity deters investments in land improvement, such as building terraces, planting trees, and applying manure. Second, banks may be reluctant to fund investments in land if the land cannot collateralize the loan. There may be too little lending for investment in agriculture because farmers do not have secure title. Third, tenure insecurity might lead farmers to continue farming extensively, instead of intensively, because they are reluctant to rent out land to others for fear they might lose their rights, as the person actually farming the land might be gaining ownership rights. This insecurity often affects women more than men because women are typically secondary rights holders. Despite the obvious logic that insecure land tenure might be a costly institutional problem, there have been relatively few papers measuring just how costly insecurity might be. Deininger and Castagnini (2006) reported on a survey that directly addressed the issue. They presented evidence from Uganda that confirmed the negative relationship between tenure security and agricultural productivity. They noted that in Uganda land had increased in value due to population growth, new agricultural technologies, greater opportunities for trade, and increases in agricultural commodity prices. One might expect institutions to have adapted or evolved to more efficiently enable the allocation of land to users who would be able to utilize land more effectively. But Ugandan land tenure, after decades of civil conflict, appeared to be dysfunctional. The government bureaus enacted by the comprehensive Land Act of 1998 had been perennially underfunded, so preexisting (but abrogated) customary institutions still held considerable influence in rural areas. There was an expressed demand for titling, but the state has had little capacity and willingness to implement titling procedures. Possibly policymakers also feared that ethnic conflict, and conflict between herders and farmers, might escalate with large-scale programs to change land tenure.

104

Women and Development in Africa

Deininger and Castagnini interviewed 430 households in five districts, with a retrospective household history and local history. They found that only about 10 percent of respondents were involved in land rental markets. Few appeared to know about the Land Act of 1998. About 30 percent of respondents said they had lost land in conflicts. Insecurity dampened the incentives that landowners had to rent out their land, or even sell their land, to more productive farm managers. Deininger and Castagnini calculated that the mean value of output of land without conflict was US$200 per acre, while with conflict it was only US$90. Rather than a slam dunk, however, indicating that conflict reduces land value by half, the calculation provides an excellent illustration of a crosstab being wrongly interpreted as indicative of causality. For Deininger and Castagnini it was self-evident that causality ran from the conflict characteristic of the field to its value. But the causality could easily have run the other way: perhaps land generally was prone to conflicts, and in order to reduce conflict some fields were planted in more valuable tree and perennial crops as a way of strengthening tenure rights. Planting tree crops doubled land value, and conflict played only a small role in determining land value. Both interpretations are plausible, but rather than prioritizing tenure reform, one suggests the prioritization of conflict resolution, while the other suggests the prioritization of promoting tree and perennial investments. In any case, the data analysis of Deininger and Castagnini also suggested that women were more likely to suffer from conflicts and lose land value due to tenure conflicts. Moreover, they found that widows and single women bore a disproportionate burden of tenure insecurity. These women were much more likely to be involved in land conflicts than men. Given that the negative effects of tenure insecurity are likely generalizable across the continent, gender reform of land tenure may be seen as both an equity and an efficiency issue. Women ought to have equal rights to secure ownership, and providing those rights would benefit society at large because it would lead to better allocations of land and labor and investment. Until the late 1990s, efforts to reform land tenure at the national level overlooked the importance of women in agricultural systems, acting as if only male farmers mattered for land tenure (Goebel, 2005; Jacobs, 2009). That changed as organized women’s groups and funders of reform efforts such as the World Bank gave more priority to gender concerns. There were numerous attempts made by national governments to redress the perceived trend toward growing gender inequality in land tenure (Giovarelli, 2005; Higgins and Fenrich, 2012). These efforts involved greater attention to strengthening and enforcing joint ownership of land for married couples, registration of land title to include the signature of the spouse, and clarification of the rights of women under communal land tenure (Razavi, 2007; Tsikata, 2003).

The Land Tenure Rights of African Women

105

These reform attempts have been controversial. Kapur (2011) expressed skepticism over various attempts by governments and nongovernmental organizations to reform tenure at the national level. In her view, land tenure was ultimately implemented at the local level. Local authorities continued to have gendered perspectives on land. Legal reforms were then nothing more than cheap talk. This nihilistic view about the relevance of national policy might be contrasted with the view of Aldashev, Platteau, and Wahhaj (2011). They considered a model where there were two kinds of households, ordinary community members and marginalized community members (e.g., women). A customary authority influenced the setting of a social norm that affected marginalized members of the community. If the norm were onerous, and marginalized members had an alternative, they might leave the community. Their departure weakened the community and the customary authority. Therefore, as the formal legal regime increasingly offered a viable alternative to subjection to customary authority, the latter was given incentives to soften the force of local social norms that may formerly have been highly discriminatory. Aldashev, Platteau, and Wahhaj cited as examples the evolution of women’s rights to land in Senegal, where formal law had strengthened women’s rights, and Ghana, where reforms of laws of intestate succession also favored women and led customary inheritance to allocate more to wives of deceased husbands. These legal reforms, the authors argued, induced custom to change in the right direction of more equality. Evaluations of Land Tenure Initiatives The impetus toward implementing national land tenure policy may be misplaced. A voluminous literature suggests that many reform efforts have had limited effects, and deleterious unintended consequences such as increased insecurity of property rights are held to be common (Peters and Kambewa, 2007; Place, 2009). Much of this literature has relied on statistical methods that have serious flaws. Surprisingly little is known, in the sense of being derived from state-of-the-art empirical studies, about the cost-effectiveness and distributional effects of reforms of land tenure. Conning and Deb (2007, p. 2) noted that, at the time of their review of evaluations of land tenure reform programs, “Although tens of billions of dollars have been invested into land property reform projects, to date there is not, to the best of our knowledge, one successfully completed impact evaluation study of a land reform intervention using a rigorous study design built into the program design where comprehensive measurement and appropriate, modern statistical methods were used.” Similarly, Gignoux, Macours, and WrenLewis (2013), in their review of the challenges of impact evaluation studies dealing with land tenure, noted that “while . . . donors have invested in ti-

106

Women and Development in Africa

tling and other land administration programs in Latin America for a relatively long time, rigorous quantitative evidence on the impact of such interventions is rare, in particular for rural areas.” The two review papers identified a host of conceptual, statistical, and implementation challenges to evaluating land tenure reforms. Debate and experimentation on land tenure matters have given more emphasis to investigating effective mechanisms to promote gender-oriented reform of land tenure. One reform has been to promote and enforce forms of joint conjugal ownership over land upon marriage, with provisions for spousal consent over land use and transfer, and provisions for division of land upon the termination of marriage through death or divorce. Sometimes this reform amounts to a change in national land legislation, with little effect on the ground. Several countries included in their land legislation of the late 1990s and 2000s requirements that both spouses had to consent to transactions of land. The parliaments of Uganda and Ghana had considered legislation intended to strengthen joint ownership, but ultimately did not approve the legislation. Opponents in Uganda argued that improved rights for women would undermine cohesion among clans (Deininger and Ali, 2008). In both cases, it appeared that the ruling political parties perceived the issue as divisive, with negative consequences for the majority in subsequent elections. That is, they perceived that more equality for women in marriage would generate a backlash. It appeared likely, however, that judiciaries in both countries would continue to strengthen marital rights through their decisions that become part of the common law corpus (Sam, 2012). The legal identification of who is a spouse remained a key issue in debates over conjugal rights. Rugadya (2010, p. 32) quoted Mabuya Mubarak, Uganda’s minister of Gender, Labour, and Social Development, as observing: Joint Conjugal Ownership

So far, the most powerful clause for women’s land rights is the “consent of spouse” in terms of transactions on land where the family derives subsistence. This clause however only talks about a spouse in the context of the law (a legally married spouse) yet majority of women are in unions whose legality is not recognized hence being left out.

A large-scale land registration program in Ethiopia provided for joint registration and followed through with the bureaucratic structures on the ground that would deliver joint titles. A nationally representative largescale survey revealed that the program had a mixed impact (Deininger et al., 2008). The certification program was envisaged as rapid and low-cost, in the sense that the goal was to deliver certificates to as many uncontested plots as possible. Indeed, apparently over 20 million certificates were deliv-

The Land Tenure Rights of African Women

107

ered in two years from 2003 to 2005. The process was that a local committee for land use and administration was formed in each locality, supposedly by popular vote. The committee had to include at least one woman. (The survey found, however, that 80 percent of committees did not, in fact, include a woman!) The committee then issued certificates of land registration after public meetings. Again, many of the committees apparently abnegated their responsibility to issue certificates jointly to husbands and wives. In the Oromia region, for example, 58 percent of certificates were issued only in the husband’s name, even though all were required to be issued jointly. Further analysis of the Ethiopian program suggested that female-headed households benefited from being able to access the land rental market more easily (Holden, Deininger, and Ghebru, 2011). Women without access to family labor or cash to pay hired laborers thus benefit from renting out their excess landholdings, while those in a position to farm at a larger scale benefit from being able to rent land from others. Ali, Deininger, and Goldstein (2011) summarized a similar land tenure regularization program undertaken by the government of Rwanda. Rwanda in 2010 was one of the highest-density countries in the world, with 384 persons per square kilometer. The great majority of the population earned income directly or indirectly related to agriculture. The average household owned only .72 hectare. Moreover, land was highly fragmented, with the average household having its land divided into four plots. The issue of land was critical: population was growing, and pressure on land contributed to the genocide of 1994. The government responded with initiatives concerning land, including the Matrimonial Regimes, Liberties and Succession Law of 1999, the Land Policy document of 2004, and the Organic Land Law of 2005. These laws and policy directives regularized land tenure under a single statutory system. All land was deemed to be owned by the state, but users had ninety-nine-year transferable usufruct rights. The smallest unit of land recognized by the law was 1 hectare. The idea of legislating a minimum size was to prevent fragmentation, though the experience of many countries suggests these kinds of mandates are easily evaded through informal property arrangements that subsequently have to be recognized in the judiciary. The initiatives further provided for compulsory registration of property and of transfers of land. Customary law was no longer recognized. In terms of gender, the legal changes in Rwanda strongly protected property rights by women vis-à-vis the land held by their husbands provided they were in a legally registered marriage. Moreover, land was subject to the provisions of family law that specified equal shares for the marriage partners. If a husband wanted to sell land or obtain a loan and use the land as collateral, his wife’s formal consent had to be obtained and recorded. Daughters obtained equal rights to inherit their parents’ property.

108

Women and Development in Africa

As numerous critics pointed out, the various pieces of legislation and policy did not protect the property rights of women in traditional, customary marriages not recognized by the state. In their paper, Ali, Deininger, and Goldstein evaluated a pilot program of the land tenure regularization undertaken in 2007. The program intended to demarcate and adjudicate almost 15,000 parcels covering 3,500 hectares held by approximately 3,500 households. The pilot program chose to implement the program in areas where there were many former refugees, soldiers, and female-headed households. The project identified plots on aerial photos. If a plot of land was not subject to any disputes, verified by neighbors, then the program entered the plot in a registry book with the names of owners and the heirs who would have claims to the land. The program issued a land certificate upon payment of a fee. In order to evaluate the program, the authors conducted a short survey of 3,500 households on both sides of four pilot area borders in 2010, two and a half years after the government implemented the pilot. They assumed that households close to the boundary were similar, so those not in the pilot program constituted a valid control group. That is, the pilot project boundary was essentially arbitrary, just like a random selection, for households close to the boundary. As the authors put it: If land registration were voluntary, we would be concerned about endogeneity of outcomes such as investment and land market activity. However, tenure regularization was compulsory and covered all private land in the cell. In other words, plots were registered up to a cell’s administrative boundaries, but not beyond the cell border. The spatial discontinuity generated by this allocation rule thus can be used to allow identification of program effects by comparing individuals who live within a band on either side of the border. (2011, pp. 10–11)

The only difference between households on either side of the border is that on one side their plots are registered, and on the other side they are not. The survey collected information on household demographics, housing, assets, participation in the credit market, participation in the registration program, and knowledge of the law. Regarding farmland, questions included characteristics of each field, investments, inheritance dynamics, and participation in land sales and rental markets. The authors estimated a regression equation: Yph = α + β1Tph + β2Xph + β3Zh + εph

where Y was the outcome of interest for parcel p in household h, and explanatory variables included T, a dummy variable indicator for participation in the land tenure regularization program; X, a set of parcel characteristics;

The Land Tenure Rights of African Women

109

and Z, a set of household characteristics. The main variables of interest were the household variables representing the gender of the household head (22 percent of households were headed by women). The regression results suggested that the intervention led to a large increase in investment in soil conservation, with an even larger increase for female-headed households. As they noted, “For male-headed households, investments in soil and water conservation (which are critical given the intensive use of land in Rwanda) increase by approximately 9 percent, but for female-headed households, these investments increase by 18 percent.” Moreover, the likelihood that married women claimed to have ownership over land increased dramatically because of the registration program. Respondents in the pilot program zone were also much more likely to indicate that daughters would inherit land. The evaluation did identify some key areas for improvement. It was clear that women in traditional marriages (such as polygynous marriages) did not enjoy the same benefits as women in the monogamous marriages that were formally registered. The fee structure for obtaining land certificates also seemed to err too far on the cost-recovery side, ignoring the externalities of widespread access to title (reduced conflict, better land management, more transfers to efficient users of land). Likewise, the program in its early years appeared to give short shrift to the easily anticipated problem of managing the land registry as households conducted transactions and transfers (especially through inheritance). It was still too early to tell whether Rwanda’s ambitious program might not end up the same way as numerous other programs: outdated land registries and corrupt bureaucracies creating even more ambiguity in land tenure than before. Legal changes and tenure regularization processes can only be effective if people are apprised of changes in rights and opportunities and processes for exercising rights. Effects of land legislation depend on knowledge. Deininger, Ali, and Yamano (2008) found that only about one-third of respondents in a large-scale survey in Uganda were aware of the provisions of the 1998 Land Act, including provisions that granted more control to women over land transfers. Santos, Fletschner, and Daconto (2012) reported on a community awareness project undertaken in conjunction with Rwanda’s land regularization effort. The project delivered in 2011 a series of public dialogue and awareness-raising programs in Musanze District in Northern Province. The target communities were both local officials and the broad public affected by the legal changes. Prior to the public discussions and conversations, a variety of media was used to raise awareness of the coming program in the target districts. A radio skit that was played is worth reproducing in full as it gives a clear idea of the nitty-gritty of land issues involved (Daconto et al., 2011). As can be seen, the issues were local in flavor, requiring atten-

110

Women and Development in Africa

tion to nuance, history, and context that are difficult to reproduce in national land legislation. The following is the text of the radio message that the project produced and broadcasted.

[Beginning] A child singing in the background BROTHER: Where is your mother child? Oh, there she is. What do you want from me woman? Do you see, I put in boundaries around my land, and then you pull them out? How dare you do such a thing? Don’t you want to quarrel with me woman? (With angry voice) SISTER: What are you talking about? I don’t feel shy to pull them out because this is the land I received from back ago, and here I have documents from the Court to prove it. Try to beat me again . . . (With confidence) BROTHER: Who told you that a woman can own land? SISTER: Oh my God! I have rights to pull out the boundaries since everybody knows that I received one land as Ingarigari from my father, and in addition you know well that I received as well those three acres (ibikebo bitatu) in Gacyamo valley from my Grandfather . . . this is why I presented my land as collaterals to the bank so that I would receive a loan. BROTHER: This is what I don’t like. You need to grow up woman— know that you don’t have any rights to using my father’s land starting from now. (With angry voice) SISTER: Things have changed given that I have authentic documents proving ownership—you will never pull out anything from me again. BROTHER: Shameless! You dare insulting me, let me come and show you on how to beat a woman . . . (With furious voice) NEIGHBOUR: Please calm down and don’t quarrel over large lands as you are brothers and sisters—What if you had a small land like mine, won’t you kill yourselves over it?—try to resolve the problem peacefully. If LTR [land tenure regularization] could come to the Paysannat, it could resolve these family disputes including yours by means of confirming the boundaries of each land plot to the right owner in the community. Let’s hope competent authorities will come soon to start the LTR. SISTER: I know! Even my brother has some land he is using which belongs to some people in Kigali. Now the case has been presented in Court by landowners claiming their property. NEIGHBOUR: Please understand each other and try again to resolve the matter peacefully. BROTHER: Neighbour, you should have let me beat her since this lady is stubborn! You know I am a man in the family . . . (With embarrassed voice)

In order to study the impact of powerful messages such as these, intended to raise awareness and stimulate open debate, the researchers se-

The Land Tenure Rights of African Women

111

lected a random sample of 355 landowning households, including monogamous households under various forms of marriage, polygamous households, and households headed by widows. One-third of the households were in villages outside of the program area and constituted a control group. Note that if there were strong correlations within villages concerning land issues, perhaps due to the influence of local customary authorities, the sample in some sense was only four sites, with only one outside the program area. The effective sample size was then quite small. In the language of statistics, the idiosyncratic errors for the household observations may have been closely correlated within each of the four regions. Overall, the researchers reported that awareness of and participation in tenure regularization were extremely high, generally approaching 90 percent of the sample. Socioeconomic privileges of wealth, gender, and ethnicity seem to have not played much influence in participation. Households were questioned about their knowledge of legal changes, and responses varied from very high knowledge of most changes in legal rights to widespread ignorance of the provision that only plots of 1 hectare or more could be registered and so land could not be subdivided below the 1 hectare limit. The researchers estimated regressions with outcomes of the awareness campaign explained by the treatment of the campaign and controlling for other variables describing household status (e.g., age, size of landholdings, literacy, gender). The awareness campaign generated a paradox, however, in that wives were apparently less likely to be listed as having title in those villages, compared with the control group. The campaign was more successful in changing men’s and women’s perceptions of tenure security, with these rising substantially compared with the control group. Finally, the awareness campaign significantly affected responses to questions about whether sons and daughters were likely to inherit. The research revealed a problematic ambiguity about the legal changes and their promulgation through awareness campaigns. Households were unclear about whether wives in polygamous households could share title with their husbands. Their marriages were not recognized by the state, which only recognized monogamous unions. Since husbands wanted to clarify inheritance for their children in ways that would not lead to ambiguity about which children of which mothers would inherit, the husbands ended up in many cases having land registered in the name of their wife, rather than jointly. This outcome undermined, in local discourses, the land registration, since the understanding was that the husband remained the owner, and the registration title was simply a document to satisfy the state. Down the road, one could imagine situations (divorce, death of the wife) where courts would likely have to make accommodation for local practice rather than strictly adhere to the letter of the land registration law.

112

Women and Development in Africa

Lineage and ethnic groups across the continent lay claims to land, and the failures of individual titling led many policymakers to explore granting of communal tenure certificates. The idea is that corporate groups, with legitimate leaders, well-understood processes for resolving disputes, and clear and inclusive membership, would function like clubs. That is, they would share an interest in managing land effectively and would be able to work out arrangements among members that would be mutually beneficial. They would reduce the likelihood and costs of tenure insecurity. The push to devolve authority to quasi-formal groups was attractive, not least because the simple devolution of powers has no costs for the national government. Francophone West Africa, with a long tradition of supporting community land control (where the concept of terroir is very prevalent), was fertile ground for this approach (Bassett, Blanc-Pamard, and Boutrais, 2007). Skeptics have noted that corporate groups in Africa, as elsewhere, exhibit considerable variation. Some are authoritarian, with leaders using violence, threats of sorcery, and discourses of exclusion to maintain their power. Many are dysfunctional, with competition among interest groups within the corporate group undermining sentiments of legitimacy and solidarity. Nearly all patrilineal groups exclude women from leadership or even voice in corporate affairs. Given the reality of variation within and across regions, the potential for abuses of corporate titling, or deleterious negative consequences such as increases in conflict and insecurity, were quite possible outcomes. Berry (2009), for example, noted that customary authorities in the Ashanti region of Ghana were quick to take advantage, for their own benefit, of programs and processes to return ownership and control of land to local jurisdictions. Peters (2009), similarly, questioned the trend toward promotion of customary ownership by corporate groups. Tripp (2004) reviewed land tenure in Africa in the 2000s, noting that there was increasing policy interest in recognizing customary tenure. The World Bank held that communal tenure would reduce conflict and be a more efficacious route to privatization. Nongovernmental organizations such as Oxfam were viewing communal tenure as a mechanism to strengthen local communities, thus promoting grassroots development. But at the same time, women’s movements were agitating to abolish customary tenure. As Tripp put it: Formalized Communal Ownership

Women’s movements have been particularly concerned that heightened protection of customary land tenure arrangements has taken place in a context where the customary and religious laws and practices that have been retained have selectively preserved those elements that subordinate women. These arrangements have included customary divorce and inheritance practices, keeping women as minors (e.g., Swaziland, Lesotho, Zimbabwe), bridewealth,

The Land Tenure Rights of African Women

widow inheritance (levirate), dehumanizing rituals pertaining to widows, early childhood marriage, polygamy, and female genital cutting. (p. 3)

113

Goldstein et al. (2013) presented preliminary results of a randomized land tenure intervention in Benin that addressed some of the effects of granting corporate tenure. The Plan Foncier Rural (PFR) program, funded by the Millennium Challenge Corporation of the United States, organized information campaigns about land rights, surveyed land, developed land-use plans, adjudicated disputes, and issued land-use certificates. Villages participating in the project were randomly selected in 2008 from among those that applied. A survey of about 5,500 households in 2011, from 193 treatment villages and 98 control villages, measured a variety of agricultural and landuse outcomes. Interestingly, the initial results were that the program actually increased people’s fear that their fields might be used or claimed by someone else. Perhaps because of that, there was a small increase in trees planted (planting trees is a way of reducing insecurity) and decrease in land loaned out to others (people perhaps feared they might not recover a field loaned out). There were few other changes, and no increase in average productivity. There appeared to be, however, an increase in trust in the village land committee, which was heartening. There was evidence that the program appeared to be reducing conflict (perhaps because of awareness that there was greater scrutiny surrounding land issues). When the data were disaggregated by gender, women were somewhat more likely after the program to obtain land and to increase their inputs, fertilizer in particular. Overall, then, the program seemed to be having more impact on women than on men, though the results were very modest. Knight et al. (2012) reported on a substantial initiative to experiment with methods to protect community rights to land. They selected twenty communities in Mozambique, eighteen communities in Uganda, and twenty communities in Liberia to implement community demarcation and title processes that were established in each country’s land legislation. The communities were then randomly assigned to four different treatments. One group received monthly legal education sessions; a second group received the legal education and also regular visits by trained paralegals; a third group received regular assistance from lawyers; and a fourth group was the control group and simply received paper copies of relevant manuals and legislation. The communities were encouraged and supported in their efforts to establish a community land committee, demarcate community land, develop procedures for local adjudications, develop local land-use zoning plans, elect a legal local land authority, and proceed with steps to formalize community control vis-à-vis the national government. Although the sample sizes were too small to draw robust conclusions, the findings, after two years, were suggestive. First, the process for estab-

114

Women and Development in Africa

lishing formal community rights was very slow and drawn out. None of the communities had established legal rights by the end of the second year of the project. Nevertheless, the authors documented that the process, in Liberia and Uganda, but not Mozambique, had high levels of participation and numerous positive effects in improving local civic accountability. The authors found that the education sessions with paralegal support appeared to be more cost-effective than the other interventions. Finally, the report found considerable evidence that a participatory process, if explicit about the need to incorporate women’s land rights and supported by paralegals, could shift attitudes and community practices toward land tenure practices more favorable to women. The process in Mozambique, however, appeared to have had the opposite effect, undermining women’s rights. Conclusion Land tenure in rural Africa remains a difficult subject matter for social scientists because there are few large-scale, nationally representative surveys. Moreover, it is difficult to conduct randomized control trials on a large scale. So the literature is permeated with extrapolations from small-scale studies. A well-placed anecdote can be enough to shape someone’s opinion about a $100 million program to reform tenure! Nevertheless, it does appear clear that recent innovations in tenure programs, such as joint registration and communal tenure, suffer from many of the same implementation problems as earlier efforts to regularize individual tenure. Meanwhile, local land tenure systems evolve. Women remain at the margins of both formal tenure programs and local change.

5 Constraints and Opportunities in the World of Work

Some girl children start crying to head-load water as early as 4 years . . . the girl child naturally knows how to centrally head-load the bucket while the boy head-loads poorly so that often the water spills. —Ghanaian mother, 2010

Nobody in my family approved of my choice to be a motorcyclist. In fact, they accused me of being a prostitute because I was joining a “male” job. It was hurtful and discouraging but I decided to go with it anyway. —Nadine Mukantwali, Kigali, Rwanda, 2013

In our village, like the villages around here, the boys work with their fathers and the girls with their mothers. That is why each one learns the jobs of his or her sex. And the father tells his boys what to do, and a mother commands the girls. . . . Before, a woman only planted and harvested, and the man did the weeding and plowing. But now women do all the work except that at harvest they are told to step aside. —Elderly Bwa woman in village of Béréba in Burkina Faso, 1996

t is often taken as a truism that African women do all of the work, at home and in agriculture. A commonly repeated statistic is that women “produce nearly 90 percent of food on the continent” (Sibanda, 2011). First-time visitors to African cities and countrysides constantly remark on the preponderance of young men “sitting about” while women are fetching firewood, cooking, pounding grain, and carrying out innumerable other chores. Truisms, like statistics and personal observations, can sometimes be quite misleading. Surveys of the allocation of time suggest that while

I

115

116

Women and Development in Africa

women in Africa work more than men, on average, the numbers of hours worked are closer than many think. Some studies find the extra work burden of women to be on the order of an hour a day. When they work in agriculture, very often women and men are doing similar activities in the fields (Ancey 1983; Stone, Stone, and Netting 1995). Moreover, simple comparisons of hours worked may not correlate well with indicators of well-being and perceptions of well-being. As Doss (2010) observed, the claim about inequity in hours of work obscures the complex underlying reality, which is that women’s labor in agriculture cannot be neatly separated from their other time uses; neither can it be separated from men’s labor; nor can women’s labor in agriculture be understood properly without also understanding their differential access to land, capital, assets, human capital, and other productive resources. (p. 20)

More interesting than simply counting hours is to try to understand the determinants of the allocation of time, and especially the extent to which people are constrained in their choices over how to allocate their labor. Freedom to choose what kind of work to do, and for whom, is a basic human right, and should be considered one of the goals of development. Freedom in the allocation of labor time also underlies a presumption that a labor market, free from government regulation or discriminatory employment practices, is a desirable social institution for allocating labor. It should be remembered that labor markets have not always existed. For centuries, people in many parts of the world were not free to sell their labor. They were slaves or serfs, with their labor tightly controlled by masters and lords; they belonged to castes required to work in certain tasks, such as blacksmiths, weavers, singers, or animal herders; they were members of communities where labor was supposed to be an expression of solidarity, freely shared; or they lived in socialist societies, where bureaucrats evaluated abilities and assigned tasks accordingly. For people in these societies, dickering over an appropriate wage for an afternoon’s work was neither a right nor a choice. The social and institutional restrictions on freedom to work may on occasion be preferable to a formal freedom that is undercut by practical limitations. As Doss noted, differential access to other factors of production may render a formal freedom meaningless. If a small group of landlords control all the land, and are backed by the police powers of the state to enforce their restrictions, then people may have limited freedom to choose their work. The costs of leaving a home and moving to live among strangers may be very high. Most African societies in the twenty-first century are admixtures of free labor markets and social norms restricting choices. Few people think they can completely control the choices of others. Rare is the chief or gov-

Constraints and Opportunities in the World of Work

117

ernment official who can bully his population into forced labor; rare is the father who thinks he can make his sons work for him until he gives permission to marry; rare is the mother who thinks her daughter will accept being a beast of burden after age eighteen. West African villagers are no longer outraged when a member of a griot caste (of blacksmiths, weavers, and musicians) becomes a government official. Descendants of former slaves employ the descendants of their former masters. This is not to claim there are no pockets and incidents of indentured servitude, child labor, quasi-slavery, and other forms of coercion. Clearly there are, but these are the exceptions rather than the rule. Evidence of Restrictions in the Allocation of Labor in Africa Despite this movement toward greater freedom in labor choices, most observers still find that women’s labor remains restricted in numerous ways in many parts of sub-Saharan Africa. The restrictions are less draconian than the gendering of land rights described in Chapter 4. Chiefs, earth priests, and other authority figures of village society have little to say on the gendering of work. Neither, typically, does the state. Instead, the gendered structuring of work seems to be sustained by informal norms operating at the level of households and communities (Boudet, Petesch, and Turk, 2013). These norms are enforced via internalization and sanctions. Children grow up learning the norms. Were they to violate the norms, they would feel psychological distress from cognitive dissonance. Their identity, as a person who does or does not engage in a certain kind of work, would be challenged by violating the norm. Moreover, they might be subject to ridicule, scorn, and ostracism by other members of the community. They might even be physically intimidated. The norms are similar and related to notions of coverture, and vary widely across ethnic groups, with some groups considering women to be under the tutelage of their husbands and thus incapable of exercising choice over the disposition of their talents and labor. Restrictions on work come in two varieties: restrictions on occupations and activities (what a woman can do as work) and restrictions on the allocation of time (for whom and where a woman can work). Examples of these restrictions abound in African settings. Occupations and activities are strongly gendered in the largely informal economies of Africa. Discourses categorizing occupations and activities as male or female often present the gendering of roles as something given by nature or religion, inherent in the qualities of being male or female, and so there is little room for contesting the norms. Market activities are often Restrictions on Occupations and Activities

118

Women and Development in Africa

gendered. In rural Sudan, for example, several case studies have shown how Islamist discourses of the military government restricted, through new gender norms, livelihood choices in village marketplaces (Kevane and Gray, 1995; Willemse, 2007). Many societies sanction women who engage in intervillage, cross-border, or long-distance trade. Women engaging in this trade have been subject to verbal and even physical abuse, and are sometimes accused of being prostitutes (Hodgson, 1996). There are numerous other examples of the gendering of economic activities. Werthmann (2009) presented a nuanced interpretation of gender roles in the artisanal mining sector of Burkina Faso. Other sectors that have been the object of study include beer-brewing (Green, 1999), weaving (Ajayi, 2009), blacksmithing (MacGaffey, 2009), and singing and playing music (Belcher, 2004). LaViolette (1995) traced some of the many gender identity issues associated with pottery-making in the city of Jenne in Mali. Lyons and Freeman (2009) found that female potters in the Tigray region of Ethiopia resisted marginalized identities as tainted persons. The sorting of men and women into gendered occupations is very strong. Quite possibly, the sorting has emerged from discriminatory norms rather than from the unconstrained choices of men and women. Occupational restrictions and discrimination in urban areas often have their origins in the array of colonially mandated controls over women’s occupations and mobility. These controls regulated both rural and urban areas. In colonial Mozambique, for example, virtually all garment workers were men. Well into independence, women constituted only a tiny fraction of the urban, industrial workforce (Sheldon, 1991). Patterns of occupational segregation in the urban labor force continue (Kobou, 2000). The illegality of prostitution itself constitutes a norm restraining women’s activities with economic implications. Social histories of East Africa, following the lead of the seminal paper by Bujra (1977), have convincingly demonstrated the importance of prostitution in enabling women to accumulate capital and establish themselves in the informal economy (Barnes, 1992; White, 1990). Selling sexual favors to men proved more rewarding than being “owned” by a husband. Monopolization of the labor of wives has been reported for many other African societies and takes the form of requirements that wives work on “family” fields controlled by men. Roberts surveyed the literature and remarks that “the claims of a male head of household or husband over his wives’ labor services are considerable and are not reciprocal. The extent of wives’ obligations to provide labor to their husbands is a major constraint upon the development of their own-account enterprises” (1988, p. 104). Women in many societies are prohibited from working on the fields of Restrictions on Working Outside the Home

Constraints and Opportunities in the World of Work

119

other men. They cannot participate in an open, competitive labor market. Often, women are further prohibited from working on their own account as traders or processors of food to be sold in the marketplace. They are expected to minimize their involvement in market activities (Bernard 1966, p. 140; Broekhuyse and Allen 1988, p. 337). Schroeder documented how women who work in personal small irrigated market gardens may be “demonized . . . as bad wives” (1996, p. 72). Jones (1986) and Carney and Watts (1991) likewise described the pressures on women to work in their husbands’ irrigated rice fields. Other researchers found that women are allowed to carry out market activities but not to hire labor. They are restricted to using their own labor, so their enterprises cannot expand. As Guyer put it with regard to the West African Yoruba and Beti pre-cocoa economies, “women were ultimately subordinate to men in both systems in that men could mobilize women’s labour but the reverse was not usually true. Women did not have the authority to recruit male labour for their own purposes except through their husbands” (1980, p. 364). Related to restrictions to work for husbands, but not exactly, are requirements that married women work in the household compound and not work in public or in particular work for wages for non-kin. Restrictions on women’s ability to work outside the home are familiar to development economists who work in South Asia, where the prohibition of work outside the home is known as purdah. Cain, Khanam, and Nahar argued that these norms actually entail specific geographic restrictions: “The norms of purdah influence the distance a woman would be willing to travel to work, the distance a husband would permit his wife to travel, a woman’s willingness to work for a stranger, and the receptivity of potential employers . . . the psychic and tangible costs of the job search rise quickly when a woman leaves the confines of her ‘circle’” (1979, p. 428). Ramamurthy commented on how these norms were sustained among men of the Kamma ethnic group in India: “Patriarchal domination by the Kamma men limits women to work within the household. . . . Kamma men are highly critical of the local men and criticize them for being ‘lazy’ and letting their womenfolk work in the fields; they question the very manhood of these men” (1993, p. 198). The norms are not just applied to rural women. Using an RCT design, Field, Jayachandran, and Pande (2010) found that caste norms limiting women’s activities significantly reduced the benefits of a training program for microentrepreneurs in the Indian city of Ahmedabad. In Africa, the seclusion of wives has been most studied in northern Nigeria and Niger, where the Hausa word kulle refers to the “locking” or “tying” of wives (Miles, 1994, p. 260). The problem there is more complex than in the Indian case. Restricted women are often very active participants in local markets, using what Hill (1969) referred to as a “honeycomb” network of small children to carry out their trade. Thus, seclusion may be de-

120

Women and Development in Africa

sired by both men and women and is sometimes characterized as a luxury good available only to the wealthy. Are These Really Restrictions? The experience of purdah suggests an application of the Nash equilibrium concept, discussed in the appendix to this chapter. For a norm to be constraining, it must mean that people would rather do other things. How are norms that restrict the allocation of labor sustained, if many people have an interest in violating the norm? Suppose the men of a village got together and colluded. They might act as a cartel and all agree not to hire each other’s wives. The idea of a game situation being repeated over time is at the heart of how social scientists understand collusive behavior. For example, the theory of cartels, such as the Organization of Petroleum Exporting Countries (OPEC), explores the question of how a cartel survives when each cartel member has an interest in cheating. Models of cartels suggest that a cartel may survive when the short-term benefit from cheating on the cartel may be lower than the longer-term benefits from abiding by the cartel rules. Models of cartels have been adapted to models of how village labor markets work and how discriminatory collusion can be sustained (Akerlof, 1980; Bardhan, 1979; Osmani, 1990). If men colluded to not hire each other’s wives, then there would be no competitive labor market for women who wanted to work. Instead, each woman would have to bargain individually with her husband. Recall that men own almost all of the land, the most productive asset. Men are then in very strong bargaining positions vis-à-vis their wives and can offer them low wages. The collusive norm clearly favors the men at the expense of their wives. But some of the husbands might want to break the norm. They can offer low wages to the oppressed wives of the other husbands and reap a windfall of low-cost labor. Some husbands may own more land than others and not be able to rent out the land because of tenure insecurity (they fear that the other men might claim the land as their own if they rent it for several years). They would like the wives of other men to be able to work for wages in their fields. Also, husbands who have little land might hope that their wives could go out into the labor market and earn higher wages to supplement the family income. What constrains these potential “cheaters” of the norm from realizing their interests? The husbands are constrained by their fear of the consequences. If the situation is repeated over time, then each husband calculates his short-term benefit from cheating on the norm A Model of Collusion

Constraints and Opportunities in the World of Work

121

and weighs that against the long-term loss from the breakdown of the collusive equilibrium. If other men have strategies that tell them to also cheat on the norm whenever they detect cheating by one husband, then the one potential cheater might refrain from hiring the wives of other men. He fears the breakdown of the norm that favors men, and himself, in the long run more than he favors his short-term gain. The collusion is sustained; it is a Nash equilibrium. The long-term benefit of obtaining the labor of the wife at a lowered wage outweighs the short-term advantage of hiring more labor at the low wage. The intuition captured by the model was recorded by Bassett, writing about cotton cultivation in Côte d’Ivoire: One man told a story that spoke to the origins of these gender struggles taking place at the cotton market. He said that one day during the 1993/1994 market period, he was at the market selling his wife’s cotton. When his turn came to weigh and load her cotton, a number of men spoke up and said that their wives were not growing cotton so they were not going to help him lift his wife’s bales onto the scale and load them into the truck. They departed from the marketplace. This left the remaining men in the socially awkward position of being viewed by other men as “working for their wives.” My informant expressed his discomfort at being publicly humiliated for not behaving in ways that conformed to ideals of male dominance and power. . . . Like other men in the village, he too feared that if his wives gained some economic autonomy they would no longer respect him and refuse to work in household fields. After this incident, the men who had been assisting their wives decided that they would no longer support their efforts to grow cotton. (2001, p. 365)

A concrete example from the anthropological literature may illustrate how difficult it is to analyze collusive equilibria in the absence of a fully fleshed-out model. Researchers writing about the Kofyar people of Nigeria asked, “Why, given the high demands on labor among the Kofyar, are men not further monopolizing female labor for household production to the exclusion of women’s ability to produce independently?” (Stone, Stone, and Netting, 1995, p. 181 [emphasis added]). That is, why did men not prevent their wives from working outside the household and then, once they had to work in the household, make them work for the men, rather than for themselves? Stone, Stone, and Netting answered their question by saying that men did not monopolize the labor of women because “women’s labor and reproductive power are too valuable to the system as a whole, and to their households most particularly, for men to risk alienating them” (p. 181). They painted a picture of a dynamic and growing agricultural economy in the Kofyar region that was consistent with their claim that the labor of women was valuable. Apparently, a restrictive norm was too costly to sustain. The labor of women was valuable, and men would be tempted to hire

122

Women and Development in Africa

female labor and break a collusive norm. The incentives were such that the Nash equilibrium could not be sustained. One implication that could have been explored was that in other, less productive areas women would indeed have been constrained to work for their husbands. Collusion need not be the only explanation for occupational segregation. A fascinating theoretical paper by Francois (1998) argued that occupational segregation by gender in industrial economies could occur as a Nash equilibrium. The gendered outcomes emerge as a pattern of uncoordinated behavior generated by shared expectations rather than as the result of collusion. In the model, production takes place under two different processes. One process requires motivation because worker effort and care cannot be monitored; the other process is easily monitored. The workforce of the first process is motivated when the wage is at a premium; then workers will work hard for fear of being fired and having to return to the low wages paid to workers on the other process. Francois noted that the premium to workers in the first process could be partially composed of extra bargaining power within the household. When one member has a premium-earning job and the other does not, the first member is in a better bargaining position. The consequences of this, if recognized by employers, follow immediately. If employers know that other employers only hire men for premium-paying jobs, then they also will only hire men. If they hired a woman, chances would be high that her husband would also have a premium-earning job, and so her premium would be “nullified”—her husband would be able to match her bargaining position. The gender nature of the equilibrium is arbitrary, in the sense that another equilibrium with all employers hiring women rather than men also exists. But it draws attention to how labor market outcomes may interact with bargaining going on within the household. Two other hypotheses regarding time allocation eschew discussion of social norms, restrictions, and constraints, focusing instead on common patterns of outcomes across similar households. These two hypotheses share the assumption that women’s economic activity is being determined through a process of decisionmaking within the household. This decisionmaking takes place in the context of a gendered distribution of property (men own most of the assets) and a gendered biological and learned comparative advantage (girls are trained in home work). A short digression on the concept of comparative advantage is in order. The concept is usually seen in the context of international trade, where it is used to suggest that barriers to trade reduce the welfare of the citizens of a country. This is because barriers to trade block cheap imports from entering the country. Workers are then encouraged to move into the protected industry. But by definition, the extra domestic production of the protected good Unconstrained Maximization and Household Bargaining

Constraints and Opportunities in the World of Work

123

uses up more resources per unit of production than would have been used up if the previous production pattern had continued and the good had been imported. That is what it means for the foreign good to be cheaper, after all. The opportunity cost of encouraging domestic production in one sector is foregoing production of other goods and services in other sectors. This same logic can be applied to the domestic situation of a husband and wife. The two partners in marriage have to determine an allocation of their time between home activities and income-generating activities. Suppose each has ten hours to work in the day, and the husband can earn $1 per hour outside the home while the wife can earn only half that, $0.50 per hour. In the home, the wife can make five units of home cooking, say, in an hour, while the man can only make four units of cooking. Confining ourselves to this simple problem of allocating time across activities, what is the efficient allocation of their labor? Clearly, they should not both be doing both activities. If both the man and the woman were doing both, there would be room for them to switch activities (the man doing some more work outside the home and the woman doing more work in the home) so that they both would have more money and more cooked meals. The idea of comparative advantage, as usually presented, ignores the quite reasonable observation that productivity in tasks changes as a person engages in a task. Women may be unproductive in the marketplace, and only earn $0.50 per hour, because they have little experience in the marketplace. Men are poor housekeepers because they have never swept a room. Comparative advantage changes over time, and so decisionmakers need to take into account the longer-term effects of their decisions in the present. It may be quite reasonable, and profitable, to ignore the short-term logic of comparative advantage in order to reap a longer-term gain. As men and women in households decide how to efficiently allocate their time, in both the short and long term, and spend their incomes, the decisionmaking may be amicable. Everyone in the household tries to do what is best for the household. This is the so-called unitary household. Or the decisionmaking might be acrimonious, with parents, children, and grandparents engaged in struggles over control of the resources of the household. They cooperate when it is in their interest, and they do not cooperate when it is in their interest. This is the bargaining household. In the unitary household, the allocation of women’s labor is determined by competitive labor markets. Husbands and wives are partners in a production and consumption enterprise. A woman’s property, capital, outside income, or social standing has no particular bearing on outcomes or welfare. (Large enough changes in capital or outside income may, however, make the existing marriage inefficient and lead to its dissolution in favor of remarriage with more appropriate partners.) The household’s property is pooled and joint welfare is maximized; a marriage of a wealthy man and

124

Women and Development in Africa

poor woman will be indistinguishable from a couple consisting of a poor man and rich woman, in terms of economic outcomes. The power that is derived from ownership over property is assumed to be uninteresting, since whoever has property acts benevolently rather than malevolently. The bargaining household model posits that women’s economic activity is influenced by the distribution of economic power within the household, as well as the household’s interaction with markets. Members of households, in this view, bargain among themselves for the fruits of their labors. Some of the goods they consume are public within the household, and some of the goods they produce require joint production. As with all bargaining situations, if men and women do not have full information about their partners, and do not have third parties who can monitor and enforce contracts, then the outcomes of bargaining may be inefficient. (This idea of bargaining outcomes being inefficient is easily understood when one thinks about how costly strikes—from auto factories to baseball stadiums—lead to lose-lose situations where owners and workers both are worse off.) More important, the distribution of property within the household will affect economic decisions and welfare. A poorer woman might be more likely to act in her husband’s interest, for fear of the large reduction in welfare that might occur with divorce or with a breakdown in cooperation. Other characterizations of household decisionmaking fall somewhere in between the unitary model, where the household can be treated as a single actor, and the bargaining model, where the household consists of individuals who selfishly come together to enjoy potentially shared benefits and who stay together because of the high cost of exit from the relationship. (There will be more discussion of alternative theories of the household in Chapter 7.) Regardless of whether the household is amicable or acrimonious, then, it may be the case that the resulting pattern of time allocation rarely has women working in the market, or her market activity is secondary to her work at home. The more efficient bargaining and market institutions are, the more likely this outcome is, according to this approach. If women’s comparative advantage really were in home activities or as unskilled labor (and not as managers or entrepreneurs), then, with efficient mechanisms for bargaining, women’s labor would be allocated to those activities. While some bargaining models contradict the oft-held presumption that social institutions lead to optimal allocations of resources, they are nevertheless firmly in a tradition that de-emphasizes the importance of social norms and customs in explaining behavior. There are thus various approaches to thinking about how labor markets structure the allocation of women’s time. One approach considers how men Implications for Policy and Activism

Constraints and Opportunities in the World of Work

125

exercise extra-economic power over women. The hypothesis is that men have an extra source of economic power stemming from their implicit collusion embodied in social norms and custom. The theory of cartels and selfenforcing agreements suggests that men, as a group, may extract a surplus from women by turning the terms of trade against them or by restricting their participation in markets. This social norm approach contrasts with other approaches where gender matters at the individual level only. These approaches see labor market outcomes as resulting from biology (e.g., specialization and buildup of human capital induced by childbearing roles, musculature, or psychological propensities) or historical contingencies (women own less property than men). Becker (1981), for example, argued that the gender division of labor between home and market was the result of specialization in sector-specific human capital brought about by the principle of comparative advantage and constructed on sexual lines because only women bear children. Should enforcement, legislation, and promotion of women’s empowerment be viewed as having high long-term payoffs because they undermine a cartel that distorts the price allocation mechanism and reallocates income in an unfair manner? Or should general poverty-alleviation programs that have high economic returns be promoted in a gender-blind manner? The questions have been asked for several decades. Betty Friedan argued in a Newsweek article published in conjunction with the UN World Conference on Women in Beijing that “sexual politics—reifying women’s oppression and victimization by men—had come to dominate women’s studies and feminist thought” (1995, p. 30). The rhetoric of sexual politics ought to have been toned down, according to Friedan, because it was drawing attention away from recognizing that bargaining among unequal household members was becoming more important than collusion. “The basis of women’s empowerment is economic,” Friedan wrote, reversing the causality of much previous feminist analysis, whereby a lack of empowerment made women poor. Indeed, empowerment used to be an end in itself. In developing countries, female literacy, consciousness-raising, group associations, and promotion of female solidarity were high on government and donor agendas. In Friedan’s view, the resources and energy devoted to programs to empower women were reducing the more significant welfare gains to be had from gender-blind projects of poverty alleviation, such as work programs and asset redistribution. There is no empirical basis for arguing that empowerment is secondary to economic improvements. Discriminatory norms exist and are influential. It is possible that resources directed to empowerment programs may have larger effects on a sensitive local equilibrium of implicit male collusion, in terms of improving women’s welfare, than do resources directed to marginally altering bargaining positions within the household. If male discrimina-

126

Women and Development in Africa

tion were recast in the light of cartels—that is, if it were also recast as an economic issue—then the case for working on the empowerment side would be strengthened. The durability and strength of cartels depends on the ability of a cartel to clearly identify and sanction members. Identity enables groups to self-regulate their actions. The principle applies equally to a group suffering the effects of power. What are resisting and undermining a cartel through exercising countervailing power about, if not strengthening a sense of identity, of solidarity, of empowerment? There is a real issue here. While all three approaches are valid and account for some part of how labor markets and time allocation work, policymakers have to decide what characterization is most appropriate for shaping public policy. If there are social constraints on the actions of women, then programs to empower women may have welfare-improving effects for women even if they appear to imply very low individual net benefits from improved intrahousehold bargaining—or even imply individual net costs (such as being required to attend frequent group meetings). It is possible that programs that change norms generate externalities that escape individual cost-benefit analysis. The enforcement, legislation, and promotion of women’s empowerment could be viewed as having high long-term payoffs because empowerment undermines social norms that reallocate income in an unfair and inefficient manner. If, on the other hand, norms have already been eroded by market mechanisms, then the general poverty-alleviation programs that have high economic returns should be promoted in a genderblind manner. Empirical studies of time allocation are vital in this regard. The typical study that finds that women’s time allocation responds to market incentives can be misleading (Khandker, 1988; Skoufias, 1993). These findings are in line with Friedan’s message that the gender problem is one of low endowments for women. The results are thus construed as casting doubt on the role of patriarchy or other discriminatory norms as explaining women’s behavior. Though not usually the intent of the authors, these papers may be interpreted as lending an empirically based rationale to the twin forces of cultural relativism and antifeminist backlash, which have for some decades now eroded the consensus around empowerment as policy and politics. Relativism says that programs of empowerment cannot be justified by appealing to norms about the roles and rights of women because there are no universal norms in this regard. The backlash approach says that, in practice, programs to empower generate considerable inefficiency and may even be paradoxically disempowering over the long run (because by privileging women they dull the incentives for women to invest in their own human capital). The backlash approach also objects to empowerment programs as affirmative remedies for past discrimination because these violate core ideals of gender-blind public policy. Much more empirical work is needed

Constraints and Opportunities in the World of Work

127

before commentators and policymakers can have a serious debate about the magnitudes of these mechanisms and policies. A Case Study for Distinguishing Hypotheses The three alternative theories for explaining time allocation illustrate an important feature of much social science research: many social facts are consistent with different explanations. A finding that women’s activities follow certain patterns, and that people articulate norms that prescribe these patterns, does not clinch the case that there are indeed norms that restrict behavior. Other processes may be at work. A structure that was thought to result from strong social norms may turn out to be no more than a pattern of activity common to many individuals. Some light may be shed on the difficult empirical question of distinguishing among various hypotheses by looking at the allocation of women’s labor time in Béréba, a medium-sized village of southwestern Burkina Faso. The evidence from this village suggests that ethnicity was a consistently significant determinant of variation in the allocation of time. One reason why ethnicity was so important may be that ethnic groups differed in the norms and attitudes regarding women’s work. In particular, one ethnic group, the Mossi, appeared to have women working less time at home and more time in the fields of their husbands, compared with another ethnic group, the Bwa. Mossi women also appeared to be less responsive to incentives, a finding consistent with the proposition that they were more constrained. The anthropological literature also suggests that Mossi norms regarding gender and rights in marriage were less “free” compared with those of other ethnic groups. Men and women in Bwa and Mossi societies in the 1990s organized their lives in ways similar to other West African societies. Forced marriages were common. Polygyny was common. Men and women had separate budgets and activities. Men owned most of the property and wealth; women often claimed that they were poorly compensated for their contributions to household production. But there was a widespread understanding in Burkina Faso that there were significant differences in gender relations between the Bwa and the Mossi. The two groups had very different histories. The Mossi were one of West Africa’s more famous kingdoms, having survived repeated attacks and threats for more than 500 years. The Mossi kingdom was rigidly hierarchical, with local rulers, called naba, having considerable control over the local population. Part of this control was expressed in the pugsuiré, whereby the chief of a village would decide on the marriage partners of Mossi and Bwa in Burkina Faso

128

Women and Development in Africa

young girls for the village as a whole (Labouret, 1940, p. 110; Skinner, 1964). While the pugsuiré was rarely observed in the postcolonial era, a second institution of marriage in Mossi society, the levirate, where a son inherited “rights” to his father’s wives (except for his own mother), remained quite prevalent. This form of the levirate was somewhat different from the more usual form, in which widows are inherited by the brothers of the deceased husband, rather than by his sons. The transfer of authority from deceased father to son was usually accompanied with an option for the widow to remain alone, or perhaps to return to her natal village (Lallemand, 1977). The Bwa, on the other hand, have been considered to be a premier example of a “stateless” or acephalous society. The major ethnographic work on the Bwa, by Capron (1973), and the classic “ethnographic novel” of Nazi Boni (1962), Le crépuscule des temps anciens, go to some length to emphasize the egalitarian and decentralized nature of village life. Boni is quite explicit in extending this notion to gender relations; the central plot of the novel is a love affair and marriage between a young man and young woman; the parents almost seem to be absent. An old institution that perhaps is indicative of gender relations among the Bwa was the fairly prevalent practice of woman-woman marriage (Taraoré, 1941). An older woman, typically with no children or whose children have left, could use her income generated from her own activities to marry a younger woman, known as a yarohan. The yarohan was free to sleep with men of her choosing, and the children belonged to the older woman. While there has been no study of the symbolic roles the yarohan plays in Bwa society, it is conceivable that she and her “husband-wife” stood as representations of a relative lack of dependence of women on men. A woman could acquire rights to children independently of a man. Divorce is one of the primary indicators of social norms regarding gender relations, and in this instance the two societies also seemed to be quite different. Most commentators on Mossi marriage have noted that divorce was not readily granted to a wife wishing to leave her husband for another man. Skinner (1964, p. 84) observed that district chiefs would order a woman back to her first husband, except in the case of cruelty. Capron and Kohler (1978, p. 206) found that 90 percent of Mossi marriages were stable, only ending with the death of one of the partners. Their survey was administered to a large-scale sample in the Mossi-dominated areas, and they commented that this stability could be explicitly contrasted with the instability of neighboring ethnic groups. Retel-Laurentin (1973, p. 294) confirmed this marital instability of neighboring groups, finding that less than half the marriages in her sample of 545 Bwa women in the Houndé region were stable; the other half ended in de facto divorce, with the woman typically leaving her husband for another man. (The incidence of a woman returning to her natal family was

Constraints and Opportunities in the World of Work

129

low.) She presented overwhelming ethnographic evidence to support the notion that women were able to ignore the supposed indissolubility of the marriage alliance created between the husband and woman’s family through his performance of work and giving of other marriage gifts. Capron (1973, p. 85) argued that the Bwa social sanctions that might have previously enforced a marriage transaction—derived from the fact that a marriage was considered a transaction between two lineages inhabiting, almost always, different villages—had basically disappeared. In some ways this situation was paradoxical. Mossi marriages were typically made with minimal gifts going from the husband’s side to the bride’s family, while Bwa marriages had more onerous demands on the groom to work in the fields of his bride’s father. Social sanctions against leaving an arranged marriage (which most marriages were until recent decades) enforced this stability; the power of father and chief (whether village, patrilineage, or local “clan”) was still strong. There was no doubt that these social sanctions were waning, under economic pressure from continued migration of young men to Côte d’Ivoire and to urban areas (who could then take escaped lovers with them) and under administrative pressure as successive regimes continued to push for institutionalization and preference for formal monogamous marriage. By Bwa tradition, according to Capron (1981, p. 98), women were not free to choose how to allocate their time and so had to work for their husbands. The word used to refer to women, selero, was glossed by Capron as one “who offers their work.” The man was the “owner” of the work. Capron, in a footnote, further glossed selero as carrying with it the connotation of working “under constraint.” He added, “In the household, the woman cannot freely decide on where she allocates her labor power. She is selero, the one who offers her labor to others. The man—and here it is understood as the collection of all the men in the household . . . the producers of the household—is seso, the owner to the labor power of the woman.” Capron argued that in practice things were not always so rigid. At the time of his research, women did little work in agriculture, concentrating more on processing karité butter (a kind of vegetable oil). This processing and most other work were conducted in the household as a group under the direction of the senior woman of the household. Capron suggested that women exercised considerable freedom and control over their activities (p. 101). In fact, he argued, a man who called his wife selero would find himself condemned by the public rumor of other villagers (p. 104). Times had changed. Anthropological accounts of time-allocation norms in Mossi villages are rare. Lallemand (1977, pp. 61, 71), who spent considerable time on fieldwork in a Mossi village, was clear in using the word constraint to describe women’s time allocation toward the “collective” fields of the household managed by the husband. Rohatynskyj observed that work burdens of

130

Women and Development in Africa

Mossi women were very high and concluded, “The absolute nature of demands on the part of the husband’s patrilineage and the tenuousness of the bond [to the husband and his lineage], as sensed by women, lead to the overemphasis on hard work and obedience as the basis of women’s virtue” (1988, p. 538). She suggested that women who do not follow the wishes of their husbands risked becoming labeled witches and subject to ostracism. It should be kept in mind that generalizing about social norms of marriage and divorce is an old question in West African studies, and a delicate one, for one of the primary areas of contention between the colonial authorities and their defeated “native subjects” was the institution of marriage and the native oppression of women that was a justification for colonial rule (Labouret, 1940; Skinner, 1964). The village of Béréba, roughly 100 kilometers northeast of Bobo-Dioulasso (the regional capital and second-largest city in Burkina Faso), lay in 1995 in the heart of the country’s cotton zone, a dynamic and expanding sector of the national economy. The village was very well developed in terms of using animal traction, fertilizer, pesticides, and improved cotton and maize seed. There were few tractors, however, and many families still cultivated large areas by hand. Oxen were readily rented, though the price was high. Béréba was a divided village, like so many others in the Sahel, with Bwa and Mossi vying for power and privilege. In 1995, two-thirds of the 121 households were Bwa, one-third Mossi. There was a scattering of Dafing, Dioula, and Samo. The great majority of the Bwa adhered to the local Do religion, while the overwhelming majority of Mossi and Dafing were Muslim. Only a small minority were Christian. Women’s economic activities in Béréba could be summarized in terms of three broad categories: working in their husbands’ fields, working at a “personal” activity that generated income for themselves, and working at home. Table 5.1 presents the averages of the proportion of weekdays and market days (Fridays and Mondays) devoted to these activities in 1995, according to whether the period was an agricultural peak period of intense planting or weeding. The data come from weekly interviews conducted fourteen times from July to December of 1995, recording the major two activities of each of the previous seven days. The data are described in more detail in Kevane and Wydick (2001). The data reveal a relatively minor contribution of women to their husbands’ fields, but with significant differences across the two ethnic groups. Overall, women worked on the fields of their husbands slightly more than one day per week. This average varied over the season; when agricultural operations were at their peak times in terms of seasonal importance (plowing and planting during the initial early rains, and later during the harvest), Mossi and Bwa Women in Béréba

Constraints and Opportunities in the World of Work

131

Table 5.1 Average Time Allocation of Women in Béréba, Burkina Faso Mossi Market day off-peak Market day peak Weekday off-peak Weekday peak Bwa Market day off-peak Market day peak Weekday off-peak Weekday peak

Home

Husband

Income

0.35 0.33 0.32 0.29

0.11 0.14 0.12 0.35

0.55 0.53 0.56 0.35

0.51 0.50 0.46 0.42

0.04 0.16 0.10 0.29

0.45 0.34 0.44 0.29

Source: Author survey, Burkina Faso, 1995; cited in Kevane and Wydick (2001).

women spent more time in their own and their husbands’ fields. Mossi women worked in their husbands’ fields slightly more often on weekdays and much more so on off-peak market days. Bwa women were less likely to work in their husbands’ fields, especially older Bwa women, whose work in husbands’ fields dropped by half compared with younger women. Many of the older Bwa women in the sample were widows, while there were only three widows in the Mossi sample, and this may have explained their continued high participation. (Since the Mossi were a migrant community, widows were less likely to have migrated to Béréba in the first place, and if their husband died in Béréba they were likely to move back to their natal village.) Men sometimes made conventional payments for this work in their fields, in addition to meeting the wife’s consumption requirements. For cotton fields, the sums varied from 5,000 CFA to 20,000 CFA (the exchange rate was roughly 500 CFA [Communauté Financière d’Afrique franc] per US dollar). Grain was put into household granaries in most Mossi households. The Bwa conventionally divided the harvest into two parts: large cobs were put in the men’s granaries, small cobs into the women’s. Norms about rights to use and sell this grain varied from household to household. In the past, women’s time working on their husbands’ fields was strongly conditioned by norms against women’s use of the plow, so that their labor was limited to planting and harvesting. In the interviews, people would still often speak as if women did not plow, but many did. Women spent from one-third to one-half of their days on income-generating activities. This was divided among a number of activities. One was cultivating small groundnut, sorghum, or maize fields. The fields rarely exceeded one hectare, and the modal field was one-quarter hectare. Most Mossi women had personal fields; only a few older Bwa cultivated on their own. In fact, the only Bwa woman to work a significant amount of time

132

Women and Development in Africa

(about 30 percent of weekdays) on a personal field was an older woman whose husband was sick (and died) during the study period. Toward the end of the season, she spent most of the week helping relatives with their cotton harvests. Another personal activity was selling home-processed foodstuffs, such as soumbala (a seed fermented to make a condiment for sauce) and karité (also known as shea butter), and making fried cowpea or flour fritters in the market. Two Mossi women illustrated this choice: they spent almost all of their time in the market, seven days a week. One, Ouédraogo Marie, sold sorghum fritters every day in the market, from morning to afternoon. The other, Sana Assiatou, operated a restaurant selling rice and sauce. Marie gave birth to a girl during the study, and Assiatou was breastfeeding her newborn daughter. Neither of their husbands cultivated. Marie’s husband was a livestock trader and butcher; Assiatou’s husband sold plastic dishware in Béréba and surrounding markets. A third activity was brewing and selling sorghum beer, dolo. In Béréba, only Bwa women made dolo, though many Mossi women had made it in their natal villages, and two Mossi women were regular drinkers. Bwa women spent slightly more time on income-generating activities than did Mossi women during the week, but Mossi women were more likely to spend market days in the market selling processed foods. Only one woman in the village worked for wages. She helped people in their fields, helped make karité butter, and helped with laundry. Several men in the village worked as day or contract laborers; the labor market was thin because land was easily accessible and income levels reasonably high so that people were not constrained, in general, to earn cash on a day-to-day basis. Home time also varied from one-third to one-half of total time. This time was for cooking meals and drawing water, tending to sick children, assisting in childbirth, and carrying out the numerous other tasks associated with home life. Home time also included specific reports of leisure and travel time; that is, it was “non-income-generating” time. The age distribution of children was a significant determinant of time at home. Some young women stayed at home almost all the time. Among the Bwa, for example, Koura Marie was a new bride and did not work in the small field her husband cultivated or in his father’s field. Among the Mossi, several young women worked with or for their husbands on market days and spent weekdays at home, since their husbands did not farm. Another woman was pregnant, and her co-wives had many small children, so she stayed home to care for the infants while they worked. But many older women also stayed at home for varying reasons. In terms of ethnic differences in the averages, Bwa women spent somewhat less time in their husbands’ fields, considerably less time in the market,

Constraints and Opportunities in the World of Work

133

and almost no time in their own fields. They did spend more time at home. While Bwa women had more education than Mossi women, this reflected in part the much higher levels of education of the Bwa in general, for boys and girls. Polygyny was more common among younger Mossi women than among younger Bwa women, but more common among older Bwa women than among older Mossi women. Other notable differences between the two groups are seen in, among other factors, whether a woman’s father held a salaried position, whether her father had spent time in the army, degree of family wealth, size of area being cultivated, and number of oxen owned. An index of housing quality was higher for younger Bwa women than for younger Mossi women, while higher for older Mossi women. Table 5.2 reports the results from a set of regressions with the percentage of time devoted to home, income-generating activity, and husband’s field as dependent variables. The explanatory variables included are a subset of the variables described above. There are a number of results to note. First, the presence of an older girl in the household greatly reduces the time a woman spends at home, enabling her to work on her husband’s fields and on her own income-generating activities. The presence of older boys induced women to spend more time at home; perhaps she had to spend more time cooking for the boy and his friends. For the Mossi, a woman with an older boy was considerably less likely to spend time in the market. Second, as might be expected, widows were less likely to work on the “family” field (of the male head of household with which they were affiliated). Third, none of the personal status variables of the woman (whether her father was a regular salaried employee, or in the army, or a chief, and whether she had many brothers) affected the allocation of time to any large degree. Fourth, the time of the season, whether peak or off-peak, and whether it was a market day or a regular weekday, mattered quite a lot. Women worked more in their husbands’ fields during weekdays of the peak period. Fifth, and most important, the husband having more oxen seemed to influence the pattern of time allocation for the Bwa, but not for the Mossi. The responsiveness of Bwa women to their husbands having more oxen was quite strong, but the coefficient was almost negligible for the Mossi. This is consistent with the idea, developed above, that there were restrictions on the time allocation of Mossi women, compared with Bwa women. When an ethnic group places greater restrictions on a woman’s ability to work as she pleases, the woman’s decision to work in her husband’s field varies less with his stock of oxen (or farm capital) than it would for women who were unrestricted. If both groups of women were equally free to choose their own activities, then the similarities in the economic environments confronting them (living in the same village) would suggest that their responsiveness to their husbands’ capital should be similar. And yet they were clearly different.

134

Women and Development in Africa

Table 5.2 Results of Estimations with Percentage Time Allocated to Activity as Dependent Variable, Burkina Faso Time at Home (OLS)

Constant Number of children 0–2 Number of children 2–10 Number of girls 11–20 Number of boys 11–20 Year of birth Whether widow Whether have co-wife Years schooling of woman Index of housing quality Number of husband’s oxen Oxen*Oxen Bwa? Oxen*Bwa Oxen*Oxen*Bwa Years schooling of siblings Whether father was a chief Number of brothers Whether father salaried Whether father in army Weekday peak? Market day peak? Weekday off-peak? R-square Log-likelihood N=

t-stat

β

0.26 –0.05 0.01 –0.06 0.03 0.00 0.03 0.02 0.01 –0.02 0.06 –0.01 0.25 –0.10 0.01 0.00 –0.03 0.00 0.02 –0.09 –0.07 0.00 –0.04

Time in Market (OLS)

3.22 –1.67a 0.62 –4.07c 2.54b 0.77 0.80 0.66 0.38 –1.65a 2.66c –2.55b 5.73c –3.40c 3.09c 0.47 –0.67 0.23 0.56 –2.45b –2.26b –0.13 –1.20 c

0.21 380

t-stat

β

0.71 0.02 –0.01 0.03 –0.03 0.00 0.02 –0.02 0.01 0.02 –0.06 0.01 –0.14 0.03 0.00 0.00 0.03 0.00 –0.06 0.03 –0.19 –0.09 –0.01

Time on Husband’s Field (Tobit)

8.67 0.84 –0.56 2.16b –1.97b –1.68a 0.48 –0.64 0.44 2.00b –2.59c 2.15b –3.07c 0.87 –0.50 0.25 0.69 0.67 –1.53 0.72 –5.99c –2.78c –0.22 c

0.22 380

Marginal Effect –0.03 0.02 0.00 0.02 –0.01 0.00 –0.06 0.00 –0.01 0.00 0.01 0.00 –0.18 0.11 –0.02 0.00 0.00 –0.01 0.03 0.06 0.24 0.11 0.06

z-stat

–0.47 1.29 0.28 2.28b –0.59 1.06 –2.53b 0.00 –1.01 –0.30 0.44 –0.07 –6.03c 5.44c –5.32c –1.46 0.00 –1.52 1.15 2.27b 11.42c 5.24c 2.89c

0.30 380

Source: Reprinted with permission from Kevane and Wydick (2001), table 3, p. 126. © 2001 by Blackwell Publishing. Notes: OLS = ordinary least squares. a. Significant at 10% level; b. significant at 5% level; c. significant at 1% level.

Of course, it must be remembered that this result is suggestive and not at all definitive, because the husband’s ownership of oxen was not exogenous. It could be that a husband would acquire more oxen if he had a wife who was very willing to work cooperatively with him on the family field. While few farmers would explicitly give credence to such a possibility, it might indeed have been an underlying process at work. The ideal confirmation of the importance of restrictive social norms would come from a randomized control trial, but distribution of hundreds of oxen, randomly, would be a very expensive experiment. There is a small body of literature that obtains similar findings.

Constraints and Opportunities in the World of Work

135

Nikièma, Haddad, and Potvin (2008) confirmed the anthropological accounts suggesting that the Bwa women would have higher responsiveness to changing incentives than Mossi women. Mabsout and Van Staveren (2010) used methods similar to those of the Bwa and Mossi study described above to find differences in responsiveness for a sample of farmers in Ethiopia. Roos and Gladwin (2000) traced the very different gender relations in two groups subject to the same pressures of increased commercialization in a study of farming in Cameroon. An interesting paper by Lilja and Sanders (1998) took a different tack. They measured the returns to labor for women from working on their own personal plots compared with working in family fields controlled by their husbands. In their sample area of southern Mali, time in family fields increased as cotton cultivation was extended. The authors calculated that when women went to work in family fields, they were losing income. They appeared to be constrained to do that work, however. Industrialization and Social Mechanisms to Control Labor The industrial revolution that began in England in the eighteenth century, and started transforming the global economy at the end of the nineteenth century, was characterized by factory employment. People left farms and pastures and began working with machinery. The factory setting was loud, dangerous, monotonous, and disciplined. Expensive and complex machinery required constant attention if it was not to break. The movements of workers were controlled. But an unexpected social phenomenon arose. As they worked together under tight supervision, and saw the enormous profits created by the application of machinery to production, workers in factories developed a sense of solidarity. They attempted to collectively demand a greater share of profits. They perceived factory owners to be vulnerable to slowdowns and strikes. Thus began the era of confrontation between capital and labor that was so vividly and exhaustively analyzed by Karl Marx. As factories spread around the world, factory owners developed techniques to enforce the discipline and submission of workers and to reduce the likelihood of labor actions. One technique adopted by many factory owners was to hire young women. In many societies, young women were the least empowered social group. When a factory manager recruited young women, he was often counting on them being more compliant. When they did object to onerous work practices and low pay, their protests were less likely to succeed. Their bargaining power was low. Young women had little experience in bargaining, their expectations were modest, and they had worse outside options than did people in other social groups. Perhaps employers have preferred young unmarried women because in many societies young women have been regarded by social institutions of power (govern-

136

Women and Development in Africa

ments, village leaders, heads of families—i.e., older men and women) as persons without full rights in society, and hence without rights to complain of mistreatment. Young women are “owned” by their fathers. Young women, to put it even more bluntly, may be thought of as livestock, to be used in whatever work earns the most relative to the alternatives, and without any regard for their own “agency” (i.e., what they—the young women—experience or desire). When donkeys are used for transport and then switched to pulling a plow, no one asks whether the donkey is happy with the arrangement. A factory is simply another opportunity for mutual (factory owner and household head) enrichment, at the expense of the young women. As more and more young women worked in factories, the relationships between factory managers and young female employees created new gendered practices and norms. These innovations then had effects on nonfactory domains. Factory work altered relations between young women and their parents, between young women and men in the marriage market, and among young women themselves. As young women grew older, working in factories, some of the emerging practices and norms were passed on to the next generations. In some respects and in many countries, the movement toward gender equality originated on factory floors, especially in garment sweatshops. Daughters began to assert their full rights to earn their own living, independent of their fathers, and when they married they refused to remain completely subordinate to their husbands. Globalization of factory production accelerated rapidly in the 1980s and 1990s—in Asia as a result of increasing export orientation of economic policy, and in Latin America (particularly Mexico) with the signing of freetrade agreements with the United States. Many factory assembly plants have been located in rural areas, and the primary labor force is often younger unmarried women. Employers and governments (which are the entities that permit or encourage this rural factory development) encourage this development of the labor force. It appears that in the present era of globalization, young women go to work in factories over the objections of their parents. The factory offers an opportunity for young women to break the bonds of subservience that tie them to agricultural and domestic work at home. A factory is a new opportunity for mutual (factory owner and young women) enrichment, at the expense of the parents. Diane Wolf (1994) was one of the first sociologists to explore the working conditions and feelings of young women working in factories in the new era of globalized factory production. Her study area in rural Indonesia was small (and, it should be added, possibly unrepresentative). Indonesia in the 1960s had experienced a century of repressive Dutch colonial rule, a brief conquest by Japan prior to World War II, a short but bitter resistance to Dutch attempts at recolonization, and fifteen years of divisive

Constraints and Opportunities in the World of Work

137

internal political strife after independence. The countryside was very poor, and analysts were skeptical about any hope of breaking a process of “agricultural involution,” where farms grew ever smaller in size and more and more labor was applied to maintain pitifully low levels of income. There seemed to be little likelihood of rapid economic growth. In 1965, independence leader Sukarno was overthrown by Suharto, hundreds of thousands of Community Party sympathizers were killed, and the archipelago seemed destined for decades of political instability and economic stagnation. But Suharto proved masterful in creating a single-party state that focused on economic development. His regime was fortunate that oil reserves were discovered and agricultural scientists around the world discovered techniques that could bring about large productivity improvements in crops well suited to Indonesia’s climate. Suharto’s economic team was widely credited as being very pragmatic in pursuing solid macroeconomic policies. By the 1980s, the Suharto regime had begun aggressively encouraging foreign investment in factories. Until the financial crisis and political crisis of 1997, Indonesia experienced tremendous economic growth and transformation. During the early 1980s, Wolf spent a year living and conducting research in three village communities where factories had just been established. She interviewed factory managers, observed factory operations, and completed formal survey questionnaires and unstructured participant observation with young women and their families. Factory managers confirmed, in discussions with Wolf, that they deliberately sought to create a docile workforce by hiring young, unmarried women. Wolf recounts how a factory guard told young women to make sure they said they were single when applying for jobs; married women would not be hired. Wolf recounted powerful stories of the interactions between factory managers and female employees, describing how the workers were intimidated. Female workers were not naïve. They were aware of the intimidation tactics and reflected on how their socialization made it harder for them to secure better working conditions and higher pay. Some observers of the globalization of industrialization that took place in the 1980s had asserted that factory owners benefited from a kind of subsidy from farm families. This prompted debate over the household and individual decisions to work in factories. What, exactly, was the strategic basis for decisions to abandon work on family farm or informal enterprises and take up factory work? There were a great many hypotheses. One possibility, of course, was that factory work offered a higher income than self-employment. Another was that while factory remuneration might be lower at first, the career ladder would eventually lead to higher incomes. A third was that the total package of factory employment had nonpecuniary benefits that made factory work preferable to self-employment. But many observers thought that factory work paid less than feasible alternatives. Why did people take factory

138

Women and Development in Africa

jobs, in that case? One idea was that factory work was a strategic diversification of the household, to have a household member earn a lower income but less variable return. Another was that factory work emerged out of intrahousehold bargaining, as young women saw the work as a way to control their earnings rather than be “last in line” claimants to household profits. Wolf’s research among Javanese factory workers came down firmly on the side of the intrahousehold strategy, but with a twist that made the question more complex. In early stages of their factory careers, daughters contributed little to the budgets of their parents, and indeed drew on their parents for support. But once their factory careers were well established, daughters very often channeled large lump-sum payments back home, for the benefit of their parents. Wolf’s work drew out the complexity of the social settings she examined. She was honest about her initial presumption that factories were exploiting docile, subservient female labor. By the end of her fieldwork, she was more inclined to evaluate factory work as offering many of the young women escape from oppressive patriarchal homes. Wolf and others initiated a research agenda that has explored the complexities of the experiences of women working in factories, not just in Indonesia but around the world. Subsequent research has focused, for example, on the role of violence. Silvey (2003) discussed the killing of Marsinah, a young female Indonesian factory worker, in the course of a strike in 1993, as an example of how factory owners and state security forces use violence to “manufacture docility” among female factory workers. Ciudad Juárez in Mexico has also been the subject of numerous studies exploring the problem of extraordinary violence against women that appears to be tightly linked to conditions of factory work (Livingston, 2004; Staudt, 2008). Africa, as a world region that has been very late to industrialize, has seen little research on the sociology and economics of contemporary female factory workers. This remains an area for considerable and important future research. Appendix: Concept of the Nash Equilibrium Many economic settings are characterized by anonymous and atomistic interaction, as in the supply-and-demand model, where individual actions have a negligible effect on the economic environment. Other settings are strategic settings, where the action one takes is influenced by and in turn influences the actions of others. Everyone is familiar intuitively with these types of settings. As in chess, one finds oneself thinking, “If I do this, then she will do that, but if she thinks I will do that, she might do something else altogether, and besides, what do I think she will think that I will think.” The theory of how to analyze these settings is game theory, and strategic thinking is the essence of game-theoretic situations.

Constraints and Opportunities in the World of Work

139

John Nash (1951) developed an influential equilibrium concept for analyzing these types of settings. He proposed that when each person’s strategy was his best strategy, given his expectations of everyone else’s strategy, and when each person’s expectation of the other’s best strategy was indeed that person’s best strategy, then no one would want to change his strategy, and so the set of strategies having this property would be the equilibrium of the game. No other strategies would be the outcome because by definition those other strategies would not be best responses to others’ best strategies. An example should help clarify the concept. Figure 5.1 describes the payoffs to men and women in a situation of conflict. Perhaps the men and women have scheduled a meeting where they are to decide on a rule over how labor is allocated, or crops divided. Men and women have to decide whether to fight angrily at the meeting or to engage in reasonable, cooperative dialogue. When they fight, they waste energy on a long, loud, angry meeting that keeps them from working the next day. People start yelling, hoping that their anger might intimidate the other side. Of course, if the other side is conciliatory and ready to cooperate, they might give in more speedily when faced with an angry crowd of the other gender. The payoffs from the interactions of men and women are as indicated in the chart. The numbers represent the total value to men and women from the various outcomes. If women choose to fight and men to cooperate, then the women get

Figure 5.1 Fight or Cooperate? Prisoner’s Dilemma Men (payoffs in lower right corner)

Fight

–1

Cooperate

Women (payoffs in upper left corner)

Fight

9

Cooperate

–2

–1 5

–2

9

5

140

Women and Development in Africa

9 and the men get –2. The units by which value is measured are not themselves so important in this context. If the situation is reversed, with women cooperating and men fighting, then the women get –2 and the men get 9. What is the Nash equilibrium? Consider men and women in turn. If the men think the women are going to cooperate, then the men should fight and get 9. If the men think the women are going to fight, the men should fight, too, and get –1 instead of –2. So the men should fight no matter what they think the women will do. The same logic applies to the women, and so the Nash equilibrium has both of them fighting. Of course, they are then both made worse off; joint cooperation is the better outcome. Both women and men cooperating maximizes their individual and the joint well-being. They do not waste their time and energy being angry. However, neither side can credibly commit to cooperate. Perhaps there is no mechanism in local society that can enforce a promise not to fight. If one side cooperates, the other will fight and grab all of the benefits from cooperation. Suppose the game were repeated over time. Every year the same game is played. What should men do? Fight, or cooperate? Now the game is more interesting, and there is the possibility of a self-enforcing good equilibrium, where both cooperate. Why? Because a Nash equilibrium could be sustained by each having a strategy of cooperating as long as the other side also cooperates, and fighting for the rest of the time if the other side fights. If the other side has this strategy, then it may be in the interest of the first to also abide by the strategy. Not necessarily, though. Each side will weigh the benefits of “cheating” against the costs of living for the remainder of the time in the fight-fight equilibrium. Consider Figure 5.2, which shows a version of the game described in Figure 5.1, now with slightly different numbers. There are two possible equilibria. If the men think the women are going to concede, they should fight, and if the women think the men are going to fight, they will indeed concede. The reverse also holds. There is no way to know which equilibrium will hold in practice. When this game is repeated over time, the optimal strategies for the parties may be somewhat different. One way to think about this is to suppose that what people have are strategies about when to concede. That is, about when to stop fighting and begin cooperating. In this strategy, they fight up to a certain point in time, at which point, even if the other party has not conceded, they cooperate. Suppose men had reputations of being always and everywhere willing to fight. Women should then cooperate immediately. Suppose women had reputations of being always and everywhere willing to fight. Men, then, should cooperate immediately. Now suppose women in society were not sure what the men were like: they thought there was a 90 percent chance that men were the kind who would cooperate and only a 10 percent chance that they would fight. Now, the correct strategy for women might be to fight for a certain number of periods,

Constraints and Opportunities in the World of Work

141

Figure 5.2 Fight or Concede? War of Attrition Men (payoffs in lower right corner)

Fight

–10

Concede

Women (payoffs in upper left corner)

Fight

9

Concede

0

–10 5

0

9

5

and then either concede or continue fighting. (The number of periods depends on the exact numbers that characterize payoffs and impatience.) After three periods, say, where the men fought every period instead of cooperating, the chances that the men really were the fighting “type” would be greater. It would no longer pay to keep fighting them and sustaining losses. Of course, the men have similar problems with assessing the real nature of women’s character and deciding how many periods to fight. This kind of repeated game and resulting equilibrium is very interesting because it illustrates how it is possible to model a difficult real-world problem in which people fight for a finite period of time and then cooperate, without there having been any change in the underlying game. This accords well with the intuition about “wars of attrition.” Some wars and conflicts suddenly end on terms remarkably similar to terms that were proposed many periods earlier (Alesina and Drazen, 1991; Bulow and Klemperer, 1999). Why were so many resources and opportunities wasted? Presumably, the different sides to the conflicts were learning about each other and updating their beliefs about the character and motivations of their opponents. These illustrations of strategic situations suggest some of the richness and importance of this area of study for the economics of gender. One can think of objections to the Nash equilibrium approach, however, especially when applying it to understand the real world. Perhaps people do not know much about the other people playing and so have no ability to know what

142

Women and Development in Africa

their best strategy is. Perhaps people do not have the mental capacity to actually compute best strategies. Perhaps people have an inherent dislike for strategizing in this way. Perhaps strategies in this period depend on strategies that are expected in future periods and so extend down through infinity and may not be calculable. Also, there is no guarantee to the analyst that every strategic situation has a Nash equilibrium. Finally, many plausible situations will have more than one Nash equilibrium. An ordinary person might have no basis for choosing the more likely outcome.

6 The Marriage Market

It is springtime in Zimbabwe, when hearts turn to romance and pockets empty for roora, the word for bride price in the local Shona language. . . . Some people are requesting cellphones, second-hand cars or even canisters of gasoline, in this era of chronic fuel shortages and deepening poverty, to sweeten the deal. . . . Local newspapers report that some prominent families are charging thousands of dollars at a time when the average private-sector employee earns only about $1,800 a year. —Rachel L. Swarns, New York Times, October 3, 2001

President Omar Hassan al-Bashir has urged Sudanese men to take more than one wife in order to double the country’s population of 30 million. The Sudanese should ignore international family planning policies, Bashir said in a speech to the ruling National Congress Party, shown on state television Tuesday night. He said Sudan needed more people for development, since it is Africa’s biggest country and rich in resources. “We should achieve this aim by having many wives,’’ Bashir said. —Reuters, August 15, 2001

arriage is like a groundnut: you have to crack it open to see what is inside. So goes an Akan proverb. The wisdom applies equally to the social science of marriage institutions. Consider, for example, the forms of marriage. Why do some societies prohibit polygyny? Why do some villages mandate that residents marry partners from other villages? Another set of questions focuses on the causes of cross-cultural and temporal variation in the terms and incidence of marriages. Why was bride-price rising in Zimbabwe in 2001? Why are couples across Africa increasingly avoiding traditional marriage and simply cohabiting? Still another set of questions con-

M

143

144

Women and Development in Africa

cerns variation in the terms and incidence of marriage across different categories of persons. In some societies, the parents of girls who have experienced menarche at early ages receive greater transfers of bridewealth from the groom. Taller young men and women often receive more favorable terms in the marriage market. Social scientists also study the process of identity created and made salient through marriage. Marriages create social groups of affines (people related through marriage ties). There may be significant effects on well-being and production from having social groups formed through marriage alliances, as opposed to other social groups. A final set of questions addresses government policy that permits, prohibits, and regulates the terms of marriages. For example, many African governments have recently moved to try to curb the incidence of polygyny by offering a “menu” of official marriages that includes an exclusively monogamous official marriage. Change in the content and enforcement of divorce laws presents another avenue by which governments influence marriage. Addressing issues of marriage in Africa requires attention to the work of anthropologists. Anthropology has consistently focused on the economics of marriage; indeed, much anthropological work on marriage reads like economics in its application of statistics to test the implications of models of individual behavior. In anthropology, this economic approach competes with an alternative “symbolic” approach that sees marriage as a structured ritual where cultural discourses are negotiated. This alternative asserts that the structures that shape marriage, and payments such as bridewealth, are more like theaters than markets. Marriage is a cultural event, a staging of ritual performances. There is certainly truth in this assertion for many times and places. The brief news item from Sudan exemplifies how states sometimes attempt to turn marriage into a political statement, in this case by linking polygyny with “patriotic” outcomes. Both analytical points of view are worth exploring, and in fact their divergence may be thought of as the difference between assuming competition in an anonymous marriage market and assuming strategic interaction in the political-economy domain. The economics approach treats individuals as “structure-takers,” choosing and bargaining for the best deal within a set structure. The cultural or symbolic approach treats individuals as having considerable power over others, with every marriage act laden with meaning that affects other people’s preferences and decisions, including marriage structures. In explaining marriage, varied and changing moral standards regarding sexuality are often more relevant than changes in economic and social conditions. Prudery, squeamishness, priggishness: these are powerful socially created emotions that matter a lot for marriage. These psychological aspects of marriage are fascinating, but are left for another occasion.

The Marriage Market

145

Patterns of Marriage in Africa Hertrich (2007) summarized the demography of marriage in Africa at the end of the twentieth century. She noted five characteristic features that distinguished Africa from other world regions. First, nearly everyone married and so single-person households were rare. Second, women married young, and men married when older, leading to large age differences between spouses. Third, lineage groups treated marriage as an important quasi-contractual event for the corporate group. Marriage transactions had complex terms and conditions and often took place over many years and involved numerous actors. Fourth, widows remarried quickly and often to brothers or other male relatives of their deceased husbands. Fifth, men and women entered into polygynous marriages with high frequency. These five features of African marriage suggested, for Hertrich and many other researchers, that the social institution of marriage reinforced gender inequality. Women entered into marriage at young ages; into relationships with older, more worldly and mature men; under circumstances often chosen by lineage elders; and constantly under the threat of the husband taking a second or third wife should he be dissatisfied. Women’s adolescence, where they are no longer under the full authority of their parents, nor under the authority of their husband, has been relatively short in Africa. Consequently, women have had considerably less freedom for their life projects, outside of raising children and caring for the household. While African marriage patterns have been quite different from other regions of the world for decades, the patterns are changing in the same direction as other world regions: more men and women are delaying marriage, or not getting married, or are simply cohabiting; age differences have been shrinking; men and women have been increasingly choosing to marry for love, eschewing lineage pressures to enter into alliances preferred by lineage elders; widows have been more resistant to leviratic remarriage; and polygyny has become less common in urban areas. These broad generalizations conceal much regional and local variation. Lesthaeghe, Kaufman, and Meekers (1989) and Kaufmann and Meekers (1992) described some of the different marriage zones in sub-Saharan Africa, and these differences have persisted. Table 6.1 shows the median age at first marriage for women and men aged 25–49 sampled in the Demographic and Health Surveys (DHS) for a variety of countries where surveys were conducted after 2000. Women married very young in the Sahel and West Africa; the median age for women with no education was only 16.5, which means that 50 percent of women were marrying at younger ages. The median age in Niger was 15.4. Men in the Sahel who had no education married at median age 23.7, yielding a gap in ages of almost seven years. In the Sahel countries, at the time of the surveys, more than 50 percent of most

146

Women and Development in Africa

Table 6.1 Median Age at First Marriage for Persons Aged 25–49, by Educational Attainment Survey Year

Sahel average Niger 2006 2004 Chad 2006 Mali 2010 Burkina Faso Mauritania 2000–2001 Senegal 2010–2011 East average Ethiopia 2011 Eritrea 2002 West average Guinea 2005 Sierra Leone 2008 Liberia 2007 Nigeria 2008 Ghana 2008 Benin 2006 Central average Uganda 2011 Tanzania 2010 Kenya 2008–2009 Burundi 2010 Rwanda 2010 Congo, Democratic Republic of 2007 Cameroon 2011 Gabon 2000 Congo, Republic of 2005 South average Malawi 2010 Mozambique 2003 Zambia 2007 Zimbabwe 2010–2011 Madagascar 2008–2009 Lesotho 2009 Swaziland 2006–2007 Namibia 2006–2007

Women

No Education Primary 16.5 15.4 15.7 16.4 17.6 16.2 17.9 16.7 15.9 17.4 17.1 16.0 16.4 17.8 15.5 18.6 18.1 18.0 16.9 17.7 17.5 19.9 20.1

18.1 15.8 18.0

18.3 18.8 17.1 17.2 17.3 17.7 17.4 17.8 22.5 23.7

18.1 16.5 16.7 17.1 18.4 18.5 21.5 17.8 17.5 18.1 18.2 17.0 17.8 18.0 18.3 18.3 19.6 18.9 17.4 18.8 18.9 20.2 21.4

17.8 18.1 18.1

19.0 18.7 17.5 17.5 17.5 18.0 18.5 18.7 23.1

Secondary or Higher 20.8 21.5 18.5 20.3 22.2 21.6

23.1 23.3 22.8 21.5 19.1 21.4 20.6 23.3 21.3 23.5 21.8 20.8 23.1 22.4 24.1

20.0 21.8 21.0

21.2 21.1 20.8 21.4 20.9 20.7 20.6 21.9

No Education 23.7 22.4 22.5 24.9 24.8

Men Primary 22.9 23.1 22.7

21.9 21.9

23.1 23.1

24.1 23.2 23.3 24.9 23.5 23.5 22.8 23.2 22.7 23.5 24.1

24.5 23.1

23.8

24.7 23.2 23.1 22.9

22.5 24.1 21.5 24.5

24.2

24.7 24.3 23.6 21.7 23.9 24.0 23.4 24.6

23.0

23.6

24.6 22.4 21.8

22.3 23.5 22.1

Secondary or Higher 24.5

24.5

24.4 24.4 24.3 24.5

24.8

23.7 24.6 24.5 24.7

Source: ICF International (2012). Measure DHS STATcompiler, http://www.statcompiler.com (accessed June 26, 2013).

populations had not been to school. Ages were higher and gaps smaller for other regions. Southern African countries had the highest median ages at first marriage and the smallest gaps, and education levels were quite high, so the low median age for those with no education represented only a small percentage of the population.

The Marriage Market

147

Young women have also been less likely to be married in southern Africa. In the Sahel region and coastal West Africa, by contrast, marriage is nearly universal among young women. Central African countries have been somewhat in between. Chad, Niger, Mali, and Burkina Faso have had very high rates of marriage of young women and among the highest age gaps, close to seven years. None of the high-marriage-rate countries in Africa have had small age differences. Botswana has had the lowest marriage rate but a fairly high age gap, suggesting that those few women marrying young (less than 30 percent) have married in more “traditional” marriages. Finally, it is also true that in Africa (as in most of the world), there are many more single widows than single widowers, reflecting the common pattern that men die younger than women (leaving many widows) and older women are less likely to remarry (widows remain widows). Some countries, like Burkina Faso, have extensive incidence of the levirate, where a widow remarries someone in the lineage of her husband. Dimensions of Marriage Going beyond the demographic measures of marriage and attempting to quantify the economic character of marriage is problematic. Around the world, marriage institutions comprise both contractual and status considerations. Understanding institutions requires imposing structure on what is clearly a multidimensional and constantly changing and contested set of rules and practices. Marriage contracts typically encompass multiple resource flows, which may begin even before the official marriage ceremony, may be contingent and so never observed, and may also encompass difficult-to-observe behaviors, such as sexual practices. The characterization of payments made at marriages has caused sharp disagreements even among anthropology field-workers studying the same ethnic group (Chanock, 1985). Experts and locals seem to disagree with startling frequency. Another, and related, problem is that marriage practices vary greatly across societies and regions, as well as among people within a single region. Table 6.2 reports the findings of a careful study of actual marriage practices in Côte d’Ivoire. A majority of marriages did not involve a standard ceremony and payments leading to cohabitation and sexual relations. Many couples simply cohabited, while others had ceremonies performed long after cohabitation and sexual relations. One way to think about the status dimension to marriage is to note a peculiar property of marriage as a transaction: the state or local jural group is a kind of participant in the marriage contract, along with the bride and groom and their families. (A “jural group” is a society that makes rules that are enforced internally; the rules are usually not written, and enforcement is not by the state but rather within the group.) Marriage is not just a private

148

Women and Development in Africa

Table 6.2 Different Types of Marriage Sequences, Côte d’Ivoire Phase 1 Event

Cohabitation, ceremony, and sexual relations Ceremony

Ceremony

Ceremony and sexual relations

Sexual relations

Sexual relations

Cohabitation and sexual relations

Cohabitation and sexual relations Sexual relations

Sexual relations

Total

Phase 2 Event

Cohabitation and sexual relations

Phase 3 Event

Number

Percentage

468

9.6

2,247

46.3

Sexual relations

Cohabitation

40

0.8

Ceremony

Cohabitation

101

2.1

673

13.9

304

6.3

257

5.3

4,855

100

Cohabitation Cohabitation

Ceremony Cohabitation

Cohabitation and ceremony

Ceremony

Source: Reprinted from Meekers (1992), table 2, p. 68.

79

405

281

1.6 8.3

5.8

contract enforced by the state or jural group but rather an occasion during which persons enter into contracts with the state or jural group itself. Marriages that do not include this element of contracting with the state—marriages that are not “official”—are not achievable as private contracts. This can be seen most clearly when it comes to rights over children. In a common form of official marriage, for example, a man acquires rights over his wife’s children. The rights are granted and guaranteed by the state, and the state often reserves some residual power to terminate those rights. Private contracts to acquire rights over children, however, are almost never enforceable; a private person cannot usually contract away rights over children. Without being married, men typically cannot have enforceable parental rights to children, no matter how much they might be willing to pay. Similarly, in many countries, a couple that marries needs the permission of the state to divorce. Imagine an employment contract where you could not leave your employer unless a state official granted approval. Marriages may specify a status for the husband or wife that cannot be contracted away. On the contract side, marriages may specify actions to occur for the duration of the marriage. Parties to the marriage may negotiate the terms of

The Marriage Market

149

these actions. The freedom of a woman to visit her relatives, for example, or the husband’s obligation to give his wife a plot of land, may be negotiated. Bearing this distinction between contract terms and status in mind, the marriage institutions of most societies bundle together in one transaction three separate transactions that various parties want to engage in: (1) the establishment of a household, with expectation of consequent benefits; (2) the creation and transfer of status-contingent rights and responsibilities; and (3) the creation of prestige. Not every African society makes marriage so consequential. James (1970, p. 75) noted that in eastern Sudan, “the Uduk practise an extremely ‘free’ system, in which it is individuals who make up their minds to marry, rather than their families who arrange the liaison; and no material or legal pledges are made between families in connection with marriage. . . . Either the man, or the woman, may decide to break off the relationship at any time; there is no formal divorce.” According to James’s informants, children belonged to the matriline, and so it was untenable that legal rights to people, whether woman or potential children, could be the object of a transaction. Marriage simply acknowledged a sexual relationship and was accompanied by only “a few well-defined but short-term obligations.” There was no elaborate marriage ceremony. A marriage contract is in many regards like a partnership agreement that facilitates establishing a household that makes investments, produces services, and generates children. Services range from the provision of housing to the provision of insurance. These services, produced in the household, are not like ordinary, private goods. They usually have moderate degrees of economies of scale; cooked food and shelter are more cheaply provided for persons living together than for persons living apart. Household services have moderate degrees of nonrivalry. Many people can enjoy them at the same time without infringing on the use or enjoyment of other household members. The household also serves as a kind of local, open-access storehouse for all kinds of goods. Pots and pans, books, lanterns, bedding for guests, toys, and educational materials for children—all are available to household members when they might need them. Sharing those items among a limited number of persons greatly enhances welfare, as any one person is unlikely to need them all the time, or at the same time as the other household members. Marriage contracts commit parties to a continuing economic relationship, which enables investment in specialization and self-enforcing mutual insurance. Individuals are typically reluctant to specialize or insure others unless they are sure that the relationship will be long-lasting; an agreement to mutually insure each other against calamity would be of little use if you Marriage as an Enforceable Partnership Contract

150

Women and Development in Africa

suspected that your partner might renege on his or her commitment whenever it suited. Marriage, then, is a more binding commitment to share the benefits of specialization and insurance. Households are also efficient places to raise children. Child care is typically cheaper to provide in a group setting. As any parent knows, the marginal cost of a second child is considerably lower than the marginal cost of the first child. More important, long-term marriage minimizes the likelihood of the possible adverse effects on children of frequent change in adult caregivers. A parent wishing to ensure a meaningful investment in his or her children will likely seek a partner willing to enter a longterm relationship. So there is a straightforward economic reason to think of marriage as a contract for the establishment of a household. This view of marriage has much to commend it. For one thing, it explains the decline in marriage in advanced industrial economies. Modern appliances have reduced greatly the costs of single-person households. Day-care centers have emerged to provide economies of scale in child care, as wealthy societies can afford special controlled environments that are safer for children and have plenty of toys and play structures to keep them busy with less supervision. Marriage is not only about the establishment of a household. If it were, there would not be so much uniformity in household structures. Around the world, households have seemed to consist disproportionately of monogamous, heterosexual couples (Laslett and Wall, 1972). Marriage contracts among larger groups of men and women are rare. It is not difficult to imagine, even in a very poor country, many of the services mentioned earlier being provided by a variety of alternative groupings, ranging from communal kitchens to nunneries to dormitories. One aspect of marriage in particular is not like a partnership contract. Marriage contracts almost always speak to the question of rights and obligations regarding sexual activity; in fact, most contracts are quite clear (explicitly or implicitly) about some degree of exclusivity. For example, in many states of the United States, partners have had exclusive rights to one another’s sexual activity (Cohen, 1987, p. 271; Siegel, 1992; Weinstein, 1986). Marriage is a transaction involving the creation and transformation of status-contingent rights. Rights are properties of persons that are not easily alienable by the person to whom the right appertains. Status-contingent rights are rights that a polity does not see as universal but rather as appertaining to persons of well-defined social status. Marriage typically generates two sorts of rights for marriage partners: rights they will have over children and rights over the sexual services of each other. In addition, for most African societies, the rights of a father over his daughter are transCreation and Transformation of Rights

The Marriage Market

151

formed into rights of a husband and his kin over the wife. The importance of rights is particularly salient in most African societies, for the simple reason that one of the parties to the marriage, the young girl, typically has little autonomy at the time of marriage. For the young woman, marriage is about creation and transfer of her rights, and less about her voluntary entry into a contract to create a household. Besteman (1995, p. 202), writing on Somalia, observed: Women’s economic independence was thus controlled by early marriage and the denial of access to land. Women’s labor was controlled by those who held land. Prior to marriage, a woman’s labor was at the service of her father; following marriage, her labor was at the service of her husband. Between marriages, her labor may be her own, but she usually had to exchange it for access to land or food from her brothers.

Marriage also generates rights that children who emerge from the marriage will have over their mother and father and the families of their parents. An interesting example of this, though not from Africa, is the Hindu coparcenary system in India, whereby male children at birth became joint owners of the family estate. African marriages are primarily geared toward establishing parental and kin group rights over children. This contrasts with the United States, where rights over children have only recently become the central element in marriages (Carbone, 2000). Children are valuable both intrinsically and for the resources they generate when they are adults. There has been a long debate, it should be noted, over whether this latter value is in fact substantial in Africa. When children grow up, how much do they actually transfer to parents? Does the transfer value exceed the costs of rearing the child? The measurement exercise is complicated because if children are an important form of old-age “insurance,” then many parents may not “need” their children (Hoddinott, 1992). No transfer might be observed, even though there may have been an understanding that it would have been forthcoming. Parental authority and rights over children are central features of the economic organization of any society. Marriage can be thought of as the social occasion where well-defined property rights over children are reassigned. (There is considerable variety among ethnic groups over who retains rights over children born out of wedlock and how those are transformed at marriage.) Rights over fertility or children are often tied to bridewealth payments. When bridewealth is paid, the husband typically takes full control rights over children. Sometimes the rights over children are the primary purpose of bridewealth, in situations where a man has paid bridewealth and the wife then runs off with a lover, for instance. If the woman had children, they stay with the husband and the lover does not have to restitute the bridewealth paid by the husband. If the woman

152

Women and Development in Africa

had no children, then the lover will have to restitute the husband for the bridewealth he paid. Parties to a marriage commit themselves to sharing in rights over children and to binding themselves not to alienate those rights. Three caveats should be mentioned. While children cannot be “sold,” they can be “rented,” and the issue of child fostering has been a perennial concern in the social science of African childhoods (Akresh, 2009; Bledsoe, 1990). Moreover, because children will grow up to become adults, a tension exists between rights of parents over children and the ability of the child-cumadult to exercise his or her will independently. It is hard even to imagine a society where parents maintained their strong rights over a child’s behavior beyond adulthood. The reality of the transition from childhood to adulthood is that rights of parents over children diminish sharply at certain welldefined social occasions observed as rites of passage. Woman-woman marriage, a common institutional arrangement involving from 5 to 10 percent of the female population in many regions, underscores the importance of rights transfer as an element of marriage. An older, perhaps wealthier, woman marries a younger woman so that the older woman might own the children (Amadiume, 1987; Cadigan, 1998; Greene, 1998; Herskovits, 1937; Taraoré, 1941). The older woman becomes the “female husband,” and has the same rights over children as a man might; indeed, she might be considered to have male gender. Håkansson (1986, p. 16) described another variant: The Gusii practice a form of marriage where women with only daughters can take the bridewealth received for a daughter and use it to obtain a wife for a fictitious son. The children of such a wife become the paying woman’s grand-children. This form of marriage is increasing in Gusii land due to the growth in the number of single mothers who see it as an alternative way of obtaining economic security and social acceptance.

If a norm determines which activities generate prestige, then people will respond strategically to that norm (Cole, Mailath, and Postlewaite, 1992; Frank, 1985). The basis for prestige varies from society to society, but it typically involves wealth, education, and social class or caste. In marriage, the party with higher prestige lends its prestige to the party with lower prestige, or the prestige of both parties may be reinforced. Thinking about prestige may resolve a conundrum in the analysis of marriage: Why is dowry—payments that the bride’s family makes to the groom’s family upon marriage—so prevalent in India, whereas bridewealth is so prevalent in Africa? If one thinks that the underlying economic conditions of peasant society in India are fairly similar to the underlying conditions of peasant life in many parts of Africa, one might wonder why the two Marriages Change the Allocation of Prestige

The Marriage Market

153

areas should have such different patterns of marriage. One of three possibilities might be the case, if the underlying economic fundamental conditions are indeed similar: (1) either money and people are very fungible, so that the quality-adjusted prices are roughly comparable (i.e., dowry is accompanied by many more rights for the woman and obligations for men in India, so that the “net transaction” is the same); (2) somehow in a social situation where marriage payments signal information about family quality to others, societies can get “stuck” in equilibria where deviations from payment norms signal disadvantageous characteristics, and therefore families continue to adhere to the norm; or (3) exogenous “culture” in India prescribes that to earn prestige for a family, a daughter must marry up the caste ladder. This third possibility is analyzed by Anderson (2003), who focused more narrowly on the question of dowry inflation. Anderson wanted to explain why dowry had risen dramatically over the past five decades in India and Pakistan. If anything, one might think the reverse should have been happening if, as seems likely, men were the relative beneficiaries of the slow but steady growth in income. As men desire women in the marriage market, the higher their incomes, the more women, or women’s services, they might be willing to acquire. Anderson suggested that a contrary process might be at work, in which rising and more dispersed income levels led successful families within each caste to be more willing to pay dowries to buy into higher castes. The ratcheting nature of prestige has seemed to dominate other effects in some African contexts. Masquelier (2004) described for a town in Niger the common practice of indirect dowry, whereby grooms paid bridewealth to the mother of the bride, who then added her own savings to endow her daughter with a large trousseau, with an expensive bed as centerpiece. Over the years the value of the dowry has risen rapidly, as the public display of bed, pots, pans, mattresses, and other household items leads to a competition over prestige. Most interestingly, the indirect dowry persisted despite a vigorous “reforming” Islamic movement led by males who tried to limit bridewealth and dowry displays. Ensminger and Knight (1997) also followed this line of thinking, explaining how indirect dowry gradually replaced bridewealth in the Orma community of northeastern Kenya. According to them, wealthy fathers of daughters desisted from requesting bridewealth in order to make alliances with other wealthy families. Payments among the Orma elite declined relative to those of poorer classes. Models of Marriage To summarize, marriage is an occasion where rights, services, status, and prestige are transacted between bride and groom and their respective fami-

154

Women and Development in Africa

lies. Marriage in many societies has numerous economic dimensions. The new couple often merges their previously separate finances. They move into a shared residence. They cook meals together and perhaps sleep in a shared bed. They raise children (and livestock and pets) jointly. The decision is not one to be taken lightly, and the rights and obligations of each partner to provide for and to cherish the partner are often spelled out by social norms, laws, and implicit understandings. A conjugal contract sets the terms, or price, of a marriage. The task now is to develop a model of this complex and varied institution. The literature on the economics of marriage began with the very simplest workhorse model of market transactions, the supply-and-demand model. This approach to marriage was pioneered by Gary Becker (1981) in his well-known Treatise on the Family. Very quickly, however, modelers became aware of the limitations of the supply-and-demand model. Marriage, as a dynamic phenomenon, necessarily involved considerations of forward-looking behavior, and that involved modeling the expectations that men and women would have about what the marriage market would look like in the future. Just as the foreign exchange market is best modeled through arbitrage or interest parity conditions rather than supply and demand, so too the marriage market is better modeled by arbitrage-style or search considerations where brides and grooms are “just indifferent” about marrying in the present or waiting and considering other suitors. The supply-and-demand model for understanding markets is a basic tool of the social sciences. This model is useful for thinking about situations in which large numbers of people are transacting some commodity or service. The model is very common in economics, where it is a staple of introductory courses. It is also used implicitly in many other disciplines. The phenomenon of globalization, much discussed today, is partly about the increasing spread of market transactions. Goods and services that were formerly allocated via nonmarket mechanisms are increasingly being allocated via market mechanisms. So understanding how markets function is an essential beginning for any discussion of structure in society. Market structures thrive where people have individual property rights and are free to transfer those rights to people in return for a pecuniary consideration. The usual supply-and-demand analysis asks how many exchanges or transactions are made in a given time period and at what terms. The most important term of the transaction is the price at which a piece of property, an asset, a good, or a service is exchanged. The model of supply and demand supposes that the price fluctuates until all the willing buyers, at that price, are buying from willing sellers, at that price. If the price were any other price, some willing buyers or sellers would not The Supply-and-Demand Model

The Marriage Market

155

find anyone to transact with. That other price would not be an equilibrium price. The concept of equilibrium in a social situation is an important element of supply-and-demand analysis. People interact in markets to pursue their interests, however broadly or narrowly defined. What is the outcome of their interactions? Just by using the phrase “the outcome,” we are already halfway to the concept of equilibrium. In real life, people interact continuously and simultaneously. The planet does not sleep at the same time, and neither do markets. Every moment there is an outcome of the interaction of people in markets. Which outcome is the outcome? The equilibrium approach to the outcomes is to suppose something like this: the outcome is a kind of average outcome of all the outcomes over a reasonably short time period and is the outcome that people, again on average, actually think will be the outcome. Equilibrium is a situation in which what people think is going to happen (according to the models they have in their heads about the way social interactions work) does happen. The models are validated. Suppose what happens is not what people had thought. Then, equilibrium reasoning goes, people might change their actions, expectations, and models. The state of disappointment would not persist, because people would change their behavior. The reality of the world must be a continual process of learning and change, with some people satisfied and others disappointed. No social scientist believes that anything is really “in equilibrium” all of the time. Applying these ideas to the marriage market, the first question to ask is how to measure the terms or the price in the marriage market, and what is being transacted. Following Friedman’s (1990) formulation of the supplyand-demand marriage model, suppose that everyone in a society has some idea of the bundle of rights, services, status, and prestige transacted in a standard marriage contract. People in the marriage market are transacting or exchanging this standard bundle that is like a very complex good or service. The “price” of the transaction has to be approached from the viewpoint of either the bride or the groom or their families, depending on who has authority to organize the marriage. From the point of view of either side, a monetary value can be assigned to deviations from the standard contract. Not all deviations in the various dimensions to the contract will be fungible, in this sense, but for now the assumption is useful. This monetary value of deviations from the standard contract is the price in the marriage market, and that price fluctuates so that the supply of brides available in the market is equal to the demand for brides. This is equivalent to the supply of grooms being equal to the demand for grooms. The price may be interpreted as the difference between the value, to a woman, of a marriage contract that implies a given bundle of rights, obligations, and payments and the value of the standard contract.

156

Women and Development in Africa

There is an analogy to be made with computers. For the past several decades, the price of a given amount of “computing power” has fallen every year, even though the price of “computers” has risen or stayed the same. The commodity being transacted in the computer market is constantly changing. A computer is a bundle of hardware and software services and guarantees about future performance and compatibility. In the marriage market, a contract that specifies that women should work longer than they are normally expected to implies a lower price for the marriage services of the woman; the man gets more of a woman in the marriage. If the contract specifies that a woman does less work, or is freer to do as she likes, compared with the ordinary contract, then the man is paying more than he would ordinarily. Following convention, a positive price is “bridewealth,” which involves net payments from a groom to the bride or bride’s family (Haanstad and Borgerhoff Mulder, 1996). A negative payment is “dowry,” which the bride’s side pays the groom’s side (always comparing payments to the standard contract). A positive price can then be interpreted as a situation in which the implicit contract is more favorable to the woman’s family (e.g., it may involve the man turning over assets to his wife, when he ordinarily might not be obligated to do that). This view helps us understand that since marriage involves complicated transfers of rights, the money trading hands during the marriage ceremony may bear little relation to the true price of a marriage. For instance, if a larger sum of money is paid at the time of marriage from the groom to the bride, but the man now has fewer obligations to his wife, the price of a wife may be falling rather than rising. For empirical work this poses a problem, especially if marriages involve payment flows from both bride to groom and groom to bride, and involve payments or obligations that are contingent, and so more like insurance obligations. In order to find out whether brides are really more expensive, researchers must collect a wide variety of data on marriage transactions and the provisions of the implicit contract. With this understanding of the price and the bundle of services being transacted, the idea of equilibrium in the supply-and-demand model can be illustrated with a graph as in Figure 6.1. The horizontal axis measures the number of brides “supplying” marriage bundles to grooms, while the vertical axis measures the price as defined earlier. A mirror-image of the graph has as horizontal axis the number of grooms supplying marriage bundles to brides. There is no need, however, to show the two graphs. The only reason to privilege the graph with brides on the x-axis is that in many societies a groom can “demand” more than one bride, while almost no societies permit brides to marry more than one groom.

The Marriage Market

157

Figure 6.1 Supply and Demand in the Marriage Market

The graph assumes three things: demand behavior of grooms is such that the quantity of wives demanded declines when the price rises; the quantity of brides supplied (by women) rises when the price rises; and the price adjusts in the marketplace until the supply offered is equal to the demand. The supply curve is described by S (p, Z) and the demand curve by D (p, W), where p is the price (measured on the y-axis) and Z and W are other independent factors that determine supply and demand, respectively. These independent, or exogenous, factors may overlap and include such things as economic opportunities for women and men, opportunities from migration, and legal age at marriage. The equilibrium price p* in the market must be such that the quantity supplied equals the quantity demanded, or S (p*, Z) = D (p*, W). The price level is endogenous; it is determined by the other factors Z and W. The model may be used to answer questions about what happens in the marriage market when there are changes in the exogenous variables. The idea that social states are in equilibrium enables us to apply one of the most powerful tools in the social sciences. This is the method of seeing how an equilibrium changes when the initial situation is altered in some way. Typically, analysis is restricted to simple changes, where the initial situation is

158

Women and Development in Africa

altered along only one dimension. The idea is to see how an exogenous variable (the one changing first) affects the other, endogenous, variable (the one changed by the first). An exogenous variable is determined outside the model and is said to be independent. Endogenous variables are determined, according to the model, by exogenous variables. They are the dependent variables. This method is known in economics as the method of comparative statics and as the method of applying ceteris paribus, or “holding other factors constant.” As Figure 6.1 shows, the equilibrium differs for two different values of Z. Let Z1 be the initial level of opportunities for girls to pursue schooling in secondary school. Z 1 represents a low level of opportunities, and many young women are willing to marry at a young age. Let Z 2 represent an increase in opportunities for girls for secondary schooling, so that fewer are willing to enter the marriage market at any given price. The result is that the equilibrium price with Z 2 will be higher. Brides will be better off. It is useful to understand the process whereby the price adjusts to clear the market. If young women were more able to pursue secondary education, but men still thought they needed to pay the old equilibrium price (when opportunities were Z1), then there would be a shortage of women. Many men would find that they were not able to get married at what they thought were the going terms of marriage. So those men unable to marry might offer women a better marriage contract. Some might be willing to let their young brides continue their education and not expect them to do so many household chores, for example. Others might simply offer larger bridewealth to compensate the bride’s family and induce them to marry her off instead of allowing her to pursue schooling. Aggregating across all the dimensions of change, the supply-and-demand model suggests that the implicit equilibrium price would rise from p 1 to p2 when opportunities for women improved. Figure 6.1 can be used to examine a different comparative static. Suppose that population grows so that the younger cohorts become progressively larger than older cohorts. Further suppose that the society in question has a firm rule that men are only to marry younger women. Then, each year the cohort of women in the marriage market grows more rapidly than the cohort of men; the supply of wives shifts out. Z 2 in Figure 6.1 now represents a smaller relative cohort of brides, and Z1 a larger relative cohort of brides. What happens? Since more women are available, at any price, than before, the equilibrium price will fall, from p2 to p1. This represents a decrease in bridewealth or, if the price becomes negative, an increase in dowry. This “marriage squeeze” hypothesis is a leading contender for understanding the much-discussed problem of dowry inflation in India (Anderson, 2007; Bhat and Halli, 1999; Edlund, 2000; Rao, 2000). Some commentators see increases in bridewealth in African societies.

The Marriage Market

159

Why does the “marriage squeeze” logic not apply? Populations were growing rapidly in most African countries, at least until the AIDS crisis. African men also marry younger women, as a rule. So bridewealth should be falling, and even turning into dowry, by this model. Something else must be changing, increasing the demand for wives or reducing the supply of wives, or else the model is not right. Suppose the economy is changing, for example. If men’s incomes are rising, then they might demand more wives; the demand curve shifts to the right and the price rises. But suppose the fathers’ incomes are rising and their daughters are “normal goods.” In this case, they might be less willing to part with their daughters. They might enjoy the company of their daughters or might want them to complete school. So the supply of brides shifts to the left, and the price rises. In both cases, prices are rising, but for different reasons. Edlund and Lagerlöf (2006) developed a theory of the dynamics and long-term growth implications of differences in who gets to decide about marriage (that is, parental consent versus the bride’s consent). Still, the supply-and-demand model is useful as a first step. Marriage payments and the incidence of marriage do fluctuate over time in ways consistent with hypotheses from this supply-and-demand model. Borgerhoff Mulder (1996), for example, suggested that in Kenya, Kipsigi males’ interest in paying bridewealth to secure wives diminished over time as the price of maize fell and the returns to child and female farm labor declined. La Mattina (2012) found that in Rwandan districts that were severely affected by the genocide of 1994, in the sense that the number of men killed was sizable, leading to a change in the sex ratio of females to males, women had little choice but to accept marriages with more domestic violence. In effect, the price for women in the marriage market worsened. Håkansson (1986) reported that marriages in rural Kenya, as a publicly recognized institution, were clearly on the decline. The percentage of marriages following the norm of full marriage payments before cohabitation declined dramatically over the period 1957–1983. In the beginning of that period, 100 percent of marriages among Gusii women had the bridewealth paid completely by the third year of the marriage process. By the early 1980s, less than half of marriages had the bridewealth fully paid by the third year. Likewise, in the early 1960s about two-thirds of marriages had the bridewealth fully paid immediately or prior to cohabitation. By the 1980s, more than 85 percent of couples were cohabiting before bridewealth was paid. The implied price of marriage (the value to the women) was declining. Håkansson and others have attributed this decline in the incidence of marriage to shifting opportunities for men and women as Kenyan society became more fluid and urbanized and as agricultural productivity growth stagnated. In particular, Håkansson argued that women’s value in agricultural production declined rapidly during the period.

160

Women and Development in Africa

The supply-and-demand model represents the demands for and supplies of husbands as if husbands were homogeneous. Men and women differ greatly in many dimensions, from tastes to habits to looks; any marriage market will have implicit prices reflecting that variation in quality. The model could be amended to include some room for bargaining when men and women meet. The price in the supply-and-demand model is the normal price for a standard marriage with standard partners. A special person might command more than the normal price. This is analogous to modeling the labor market in terms of supply and demand, knowing that individual workers and employers bargain over many particulars of a labor relationship. Variation in marriage across individuals and households seems to accord with commonsense hypotheses. Jacoby (1995) found that men with greater wealth had more wives and that, conditional on wealth, men with more productive farms had more wives, as did taller men. Borgerhoff Mulder (1996, 1988) took a more Darwinian perspective and found evidence consistent with the hypothesis that men in Kipsigi communities in Kenya treated marriage as a mechanism to maximize reproductive success (most tellingly, she found that young age at menarche led to earlier marriage and higher bridewealth). Cronk (1991) had an interesting discovery in this regard. He found that wealthier men had more wives, but they also had fewer children per wife. This raised the interesting possibility that while women’s maximum fertility was not being attained, male fertility in a given time frame might improve reproductive fitness. That is, perhaps reproductive fitness was being maximized by having many children at the same time, rather than many children spaced over many years. Variation in Quality

A second generation of models of marriage has arisen over the past decade. These models focus more on the process and the timing of marriage decisions. On the process side, they draw inspiration from insights from matching models, which typically assume that people search for partners, make tentative contracts, break contracts, and search again, until everyone is satisfied with his or her partner, in the sense that no one prefers to reenter the marriage market, search for a partner, and convince that person to marry. Cigno’s (1991) remarkable book on the economics of the family contains a very nice introduction to matching models. In matching models, the main feature of the marriage market is the variation in quality, and transaction costs of “shopping” or “searching” in the marriage market are high. The process by which people meet is important. On the dynamic side, the new models are more realistic about how people can control the timing of marriage. If the market is not favorable for marrying this year, a person might wait until the following year. AnDynamic Models of Marriage

The Marriage Market

161

derson (2007) showed how this linking of the marriage market across time suggested that the “marriage squeeze” argument in its simple form could not explain rising dowry. Recall that the “marriage squeeze” argument was that population growth meant that over time the cohort of young women would be larger than the cohort of older men ready for marriage. So each period there would be more young women ready to marry relative to men, compared with the previous period. The price would fall as the woman or her family transferred value to the man, compared with the standard contract. (This is the same as saying that dowry is rising.) But Anderson noted that this pricing pattern over time produced a contradiction. In societies where all women married (as in India, but also Sahelian Africa), and where delaying marriage was costly for women (she might have fewer children for old-age security), a woman who waited to be married in the following period would have to get a better deal in the marriage. Otherwise, why wait? Now, if dowry were rising, since the value from marriage would be lower for women in the following period, all women would want to get married early. But there are not enough men, with monogamous marriage, for all women to marry. Some women have to wait. If women were competing among themselves for who would marry early and who would marry later, the marriage prices would adjust so that women were just indifferent between marrying and waiting. If they were not indifferent, all women would want to marry and some mechanism other than the price mechanism would be “allocating” brides to grooms. Since waiting is costly, the terms of marriage in the following period must be better for some women to be indifferent. Dowry must be falling. “It would seem,” Anderson pithily concluded, “that population change is not a promising explanation of increasing dowry payments.” Anderson demonstrated how expectations about the future play a role in determining the equilibrium outcome in the present. This implies that different expectations may create different outcomes. That is, there may be multiple equilibria and tipping points in marriage markets. Recall the modified model of Akerlof, Yellen, and Katz (1996). They supposed that men and women decide whether or not to agree to marry in a monogamous marriage market. The marriage contract might be multidimensional, including a bridewealth payment and also a commitment to allow the woman to farm land that belongs to the husband and his lineage. The value of the commitment depends on the prospects for women to earn incomes outside of agriculture. If these prospects are high, then women value less the promise by men to give them land. Since women are different, at any point in time a woman has to assess how important the promise of land is for other women. If her expectation is that women in the present or future will value land less, then she might be induced to accept a marriage proposal that is relatively unfavorable in the present, because for her the terms of marriage will

162

Women and Development in Africa

decline anyway. The terms of marriage will worsen for her in the future; more men will be married in the present and not wait to get married since they will find women who do not want land. But since the marriage market is competitive, if enough women think this, the marriage contract might indeed generate low value. The outcome might be consistent even without a rise in income opportunities, because these outside income opportunities might themselves depend on the status of women as consumers and producers. If income opportunities outside of agriculture depend on economies of scale or scope, determined by how many women are active in the nonagricultural economy, and too many women get married and agree to work for their husbands, then indeed there are few outside opportunities to earn income. The expectation becomes self-fulfilling. Polygyny Polygyny is a subject that is guaranteed to generate a heated discussion. Many Christians abhor the idea of polygyny even though it was common in biblical times. Islam limits the number of wives in a polygynous union to four and counsels that a man ought not to have multiple wives if he cannot expect to treat them equally. When asked their opinions about polygyny, African women are ambivalent (Meekers and Franklin, 1995). Some women think that the right of a man to remarry without the consent of the first wife is awful; other women appreciate the companionship that a cowife might bring to the household. Note that the term polygamy refers generally to multiple partners in marriage, polygyny to multiple wives, and polyandry to multiple husbands. Polyandry is quite rare, existing as a social institution mainly in small societies in the Himalayas. The incidence of polygyny is high in African countries compared with the rest of the world (Adams and Mburugu, 1994; White, 1988; White and Burton, 1988). Table 6.3 presents data on the incidence of polygyny for African countries having DHS surveys after 2000. Southern African countries have rates of only about 13 percent; Central African countries have incidence of polygyny of approximately 17 percent; while the rates in West Africa and the Sahel region are about 35 percent, with some countries, such as Burkina Faso, Benin, and Guinea, having rates exceeding 40 percent. Interestingly, male responses to questions about polygyny generate lower rates than female responses, suggesting some significant response bias. Perhaps male polygynists are somewhat ashamed of their marital status, and play down their multiple unions when confronted with outsiders from officialdom who tend to be monogamous. The incidence is lower for moreeducated men and women, but even among this group it is common for men to have socially recognized mistresses, the informal equivalent of a polygynous union.

The Marriage Market

163

Table 6.3 Percentage of Marriages That Are Monogamous or Polygynous Survey Year

West average Benin 2006 Ghana 2008 Guinea 2005 Liberia 2007 Nigeria 2008 Sierra Leone 2008 Sahel average Burkina Faso 2010 Chad 2004 Mali 2006 Mauritania 2000–2001 Niger 2006 Senegal 2010–2011 Central average Burundi 2010 Cameroon 2011 Congo, Republic of 2005 Congo, Democratic Republic of 2007 Gabon 2000 Kenya 2008–2009 Rwanda 2010 Tanzania 2010 Uganda 2011 South average Madagascar 2008–2009 Malawi 2010 Mozambique 2003 Namibia 2006–2007 Swaziland 2006–2007 Zambia 2007 Zimbabwe 2010–2011 East average Eritrea 2002 Ethiopia 2011

Women’s Responses

Men’s Responses

Monogamous

Polygynous

Monogamous

Polygynous

81

15

92

8

65 56 80 47 78 67 61 66 58 61 60 88 64 65 80 93 69

71 78 85 91 77 72 80 92 85 68 81 66 85 84 90 91 89

33 43 19 53 16 33 37 34 42 39 39 11 36 35 17 6 25

21 21 13 8 21 25 13 3 14 24 6 18 14 11 10 9 11

79 66 90 63 92 81 80 80 75 77 72 94 78 83 88 97 84

20 28 10 37 8 19 20 20 25 23 28 6 22 17 10 3 16

73 88 93

18 12 7

95 98 92

5 1 8

90

97 94 92 95 94 0 94

10

3 5 8

3

6

Source: ICF International (2012). Measure DHS STATcompiler, www.statcompiler.com (accessed June 26, 2013).

Becker (1981) used the supply-and-demand model to study the question of whether banning polygyny was in the interest of women. He developed the paradoxical (for his US audience) proposition that polygyny would typically be good for women. The supply-and-demand model’s definition of the price of marriage as the value of deviations from the standard marriage contract has to grapple with the problem of polygyny. When a person buys one radio, and then a second radio, the vendor of the second radio does not care that the person

164

Women and Development in Africa

already has one radio. But a wife offering marriage services to her husband may care a great deal about whether there is a second wife. Becker’s analysis assumed this problem away and instead supposed that women were indeed like anonymous vendors of services. What the man did “outside” of the marriage contract had no bearing on the value of the first wife’s terms of marriage. Women do not care whether they are sharing a man with another woman, as long as the terms of the marriage are the same as if she were monogamous. The woman only cares about the income, benefits, and duties that she herself receives or has to perform, and her welfare is not affected by what her husband does on his own time, as it were. The assumption is most unreasonable, but for Becker it neatly guaranteed his provocative result. The assumption in Figure 6.1 was that men were demanding a single wife. Suppose now that men were allowed to marry more than once. What would be the effect? In terms of the model, there would be a shift to the right of the demand curve for wives. If the price represents deviations from a standard monogamous contract, then the supply of women stays exactly the same (she only cares about what she gets in the marriage contract). The new equilibrium will have a higher price. Women who would have been happy to be married at the old price now receive a windfall (or perhaps their fathers do). Women who might not have gotten married because the price was too low now are willing to get married because the price is higher. As Friedman (1990) put it, there is nothing really different here from a rule that said people could only have one car. Relax the rule, and the suppliers of cars (or wives) are better off. Anderson and Tollison (1998) indeed argued that banning of polygyny in Utah by the Edmunds Act of 1882 was partly influenced by the fear men in the eastern United States had of women getting better terms out of marriages if they had the option of moving out West to Utah and entering polygynous marriage. Given the heated political debates over banning polygyny, it is somewhat surprising that the economics literature has not yet developed any convincing theoretical arguments against polygyny. If polygyny is so favorable to women, why has it often been banned? As with other arguments justifying government regulation, there would have to be some nonpecuniary negative externality arising from polygyny. The state of being married (polygynously) or unmarried might have some effect that the person paying the price of getting married does not take into account. The assumption by Becker that women’s welfare does not depend on what their husbands are doing on their own time is unrealistic. The husband does not take into account the change in valuation of the first wife of the marriage to the second wife. One could imagine also that widespread polygyny entails either a substantial increase in age at marriage for many men, or lowering of age at marriage for girls, or else entails positive and

The Marriage Market

165

significant probability of nonmarriage for many men. If there are positive externalities from marriage, then “spreading women around” so that every male has a reasonable expectation of marriage after adolescence could increase overall welfare, even while making worse off some women who would have joined polygynous households. Tertilt (2005) used a dynamic model of marriage to emphasize a different point about polygyny. She noted that raising a family requires resources and thus competes with other forms of investment that a person might make. If parents receive the bridewealth offered by men to marry women, then allowing polygyny raises the price of marriage (the value that parents of women receive). Structural dissimilarity between sons and daughters means that parents receive benefits from daughters, while sons have to work on their own to marry someone else’s daughter. The implications of this are that parents are induced, by the prospect of bridewealth, to have more children than they might otherwise have. Fertility is higher. Basically, parents want to have more children because they can “sell” the daughters in the marriage market. This means that investment (in improved agricultural technology, for instance) is lower, and so overall growth is lower. Banning polygyny, by making marriage cheaper for young men, thus induces parents to care less about having many children and care more about investment. An economy might move to a higher-growth path. As Schoellman and Tertilt (2006) point out, however, there would be significant negative redistribution effects from such a ban. Adult men who “owned” their daughters would be the losers. They would likely strongly oppose, and act to undercut or ignore, such a ban. Fenske (2012) conducted an extensive analysis of the correlates of polygyny in Africa using the available set of DHS surveys. The ninety surveys contained observations on 494,157 women. The data revealed several interesting findings that were not well-known. Polygyny was declining, although the incidence still remained higher than in any other world region. The most important correlate of the rate of decline was reduction in the mortality of children. As child mortality dropped, polygyny became less common. The correlation is not equivalent to causation, as there is some evidence that children in polygynous families have poorer health. But Fenske also examined a World Health Organization (WHO) program in Uganda to eradicate malaria in the Kigezi region in 1960. Since the reduction in child mortality was exogenous, the resulting sizable drop in polygyny arguably was a causal effect (which could have been mediated by changing incomes or levels of education). Perhaps some part of the broader correlation, then, also reflected a casual effect. Contemporary opportunities for schooling of girls, contemporary economic outcomes, and disruptions such as civil wars appeared to have insignificant or very low correlations with the incidence of polygyny. Spatial variation in polygyny rates did seem to vary substantially

166

Women and Development in Africa

with regional experiences of exposure to slave raiding and colonial education and religious institutions. (Dalton and Leung [2014] find that there is a strong correlation between slave raiding and subsequent rates of polygyny, and they use an instrumental variables strategy to determine that the relationship was causal.) Historical factors, then, seemed to matter more than contemporary economic changes, except for child mortality. The findings suggest that the underlying determinants of polygyny are poorly understood. Government Policy Toward Marriage African governments have tried on numerous occasions to regulate bridewealth payments, including Niger in 1975 (Masquelier, 2004), Cameroon during the early colonial period (Guyer, 1986, p. 206), and Kenya during the colonial period (Shadle, 2000, pp. 168–174). The supplyand-demand model predicts that there will then be an excess demand for brides (or, equivalently, an excess supply of men), and some mechanism will have to ration the brides. Now that men do not have to pay bridewealth, even the poorest man wants to have multiple wives. But there are not enough brides to go around. The men who are rationed out of the market will be unhappy and will seek ways to pay women to choose them rather than other men. A black market for bridewealth payments might arise, with hidden gold jewelry exchanged at wedding ceremonies. The policy is unlikely to be successful. James (1970) discussed an interesting policy innovation along these lines, mandated in 1963 in Uduk areas of eastern Sudan by the local omda, a kind of district commissioner in the “traditional” hierarchy of native administration. The omda ordered that marriages had to be validated by bridewealth. Marriages were too unstable. Presumably the omda was spending much time listening to marital disputes. Despite the very few conditions of the existing contract, noted earlier, there still seems to have been plenty to fight over. The reasoning behind making people pay to get married must have been that men would be less willing to forfeit the “bond” posted. The omda fixed some prices for bridewealth and told people they had to show him official receipts if they wanted him to hear cases of marriage disputes. James discusses how men and women rejected the policy and refused to pay, or when they did pay, the bridewealth would be returned some time later. The true price of marriage was zero, and fixing a high price simply led to “price evasion.” Another important area of intervention in marriage laws concerns divorce. Much ink has been spilled in the United States and other industrial countries over whether the switch to no-fault, unilateral divorce changed behavior. Changing the divorce law implies a reallocation of property rights regarding who can decide when to terminate marriage. Under fault law, the

The Marriage Market

167

spouse most wanting to leave has to buy out the other. Under no-fault law, the spouse least wanting to leave has to pay the other to stay. Either way, divorce should only happen when efficient (when the sum of benefits in marriage is less than the sum of benefits if separate), and so a change in property rights should have no effect on divorce rates. All no-fault law does is leave women (who more often want to stay, perhaps) with lower settlements; their husbands no longer have to buy them off. For the economics student, this is a variant of the Coase theorem (Coase, 1988; Usher, 1998). There is a lively controversy as to whether divorce patterns actually changed with changes in divorce rules in industrialized countries where data exist to carry out proper analyses (Allen, 1992; Chiappori, Fortin, and Lacroix, 2002; Dnes, 1998; Friedberg, 1998), and a related controversy is whether the switch led to different patterns of labor market participation. When divorce was contested in the United States, women had substantial negotiating power and could get remunerated for “investment” in marriage. Under no-fault divorce, courts decide settlements and do not consider compensation for investment. More important, Parkman (1992) suggests that courts do not compensate for human capital foregone (i.e., opportunity cost, in addition to investment). Because of this, women, especially those with high opportunity costs to foregoing maintenance of market-oriented human capital, have more incentives to work outside the home to maintain marketability under no-fault. Political processes do more than simply distribute property rights. They also restrict and encourage the form of economic transactions. For this reason, there can be no presumption that marriage policy is neutral in affecting outcomes in terms of household division of labor and welfare. The Beckerian or Coasian view of marriage has tended toward this neutrality position. Fafchamps (1997, p. 20), for example, contended that marriage payments, “at the onset of the union itself, [make] it difficult for an equity-minded policy maker to durably affect the intrahousehold distribution of welfare.” The argument is this: Because only net bridewealth matters, and the components of bridewealth that are negotiated are fungible, a policy change that alters some aspect of the marriage transaction simply spills over into other dimensions. The net bridewealth will be the same. But this cannot be a general result, for by forbidding contracting on certain dimensions, the range of potential contracts that marriage partners might make changes, with consequent welfare effects. Changes in policy might unwittingly lead to opposite outcomes from those intended. Government intervention that forbids contracting on certain dimensions or that changes rights assigned as part of marriage may thus have effects on behavior. Assume, for example, that women have investment projects that yield returns in the future. Women contract for marriage in the present, and in marriage money can be exchanged as bridewealth or dowry. But one

168

Women and Development in Africa

thing cannot be negotiated: residual claimancy in the second period. Residual claimancy determines who gets the surplus or product from some effort, above and beyond contractually stipulated payments. A worker who receives a wage is not the residual claimant to any discovery or extra effort that she might put in; her employer is the residual claimant. Under systems like that of coverture, the husband, through marriage, becomes the residual claimant for all the wife’s activities. The state enforces this rule; it says it will not enforce contracts between husband and wife. The woman will not make the investment if the husband has no ability to assure her that she will reap the benefits from her investment. How can he promise not to take the money away when she does earn more? What is his “commitment device”? Without some enforcement mechanism, she will not make the investment. Changing the policy, then, affects the investment behavior; the policy is not neutral. The problem is twofold: men cannot commit to compensating their wives on the basis of investments wives made before their marriage, and women upon marriage cease to own their activities. The neutrality-of-government-policy proposition assumes that activities in marriage are not subject to moral hazard (where the husband promises something but then acts in a different way later on). But there is moral hazard in conjugal interaction, so marriage contracts need multiple dimensions to reduce inefficiencies of moral hazard. This perspective bears some similarity to the discussion by Geddes and Zak (2002) on the “rule of one-third,” whereby a widow inherits one-third of the husband’s estate should he die intestate. Geddes and Zak interpret the rule as an informal institution (enforced by the grandfather paterfamilias) having the effect of ensuring investment in children. Another example of government action having effects is in the area of remedy for breach of promise. For many societies, sexual relations outside marriage remain a legally enforceable breach of the marriage contract. In some places, adultery is a criminal offense. Chanock (1985, p. 179) notes that in 1944 in Zambia the Ngoni Native Authority established a “right to damages in the event of adultery,” and this “new payment grew in both value and significance.” Lovett (1996) reports how the colonial authorities sided with male elders in western Tanzania, imposing fines and even imprisonment for women committing adultery. Finally, in an interesting study, Brinig (1990) discusses how the elimination of breach of promise as an actionable tort offense in the United States encouraged the diamond engagement ring to emerge as a commitment device. There may be similar commitment devices in many African societies, but this remains an understudied area. In the analysis of government policy, one might assume that only one variable, the policy, will change. However, even here a problem arises, because changes in government policy might be caused by other forces. They

The Marriage Market

169

are themselves endogenous. For example, suppose an African government changes divorce laws to make divorce easier for women. After the implementation of the new laws, fewer women get married in the “marriage market.” Did the divorce law somehow cause the decrease in marriage rates, or did legislators change the laws in recognition or anticipation of a social phenomenon that was already leading to a decline in marriage? Causality is difficult to determine when a purported causal factor may itself have been caused by another factor through a related process.

7 Bargaining Power at Home

When I first arrived at no. 23’s home she was dressed up as a man “for a joke” as it was Easter Day, and the charade pungently emphasised the unusual nature of their marital relationship. She drank beer, sat and joked with us in a quite uncustomary manner. . . . There were separate bank accounts, separate sectors of the firm in the management of each, and a sense of independence rather than interdependence. —A. F. Robertson, “Bugerere—A County Case History”

He took you with him and you sit in his home and bear him children. You fall ill, go to the dispensary, and would have to use your own money to care for yourself? He’s your owner, he’s the one who must care for you! —Mossi woman in a Burkina Faso focus group

hen people marry in Africa, as elsewhere, they create a household. The household acts jointly, much like a corporation. It has a joint identity that may be thought of as separate from the individual identities of the husband and wife. The interactions of husband and wife (and eventually children and other members of the household) shape the character of the household. In some places, marriage contracts are like implicit employment contracts. They specify a comprehensive range of what might be labeled work practices and performances. Once the marriage happens, the household operates on a kind of autopilot, with husband and wife fulfilling their duties. More typically, the implicit contracts that structure marriages are open-ended and vague. The character of the household will exhibit considerable variety, each one happy or unhappy in its own way. To take a simple and common feature of household behavior, consider

W

171

172

Women and Development in Africa

how food is shared at the evening meal. Certainly no marriage contract ever specified that the husband should first eat the larger and tastier pieces of meat that the wife had cooked. Nor would a contract prescribe penalties for a wife who, while cooking, gulped down the best morsels and left for her husband the more gristly pieces. Negotiation and conflict over food are thus a staple of life in poor households. Husband and wife must come to some reasonable accommodation that will vary from household to household depending on the metabolism, work habits, and preferences of the household members. Some households might have an altruistic rule for sharing food with the one who is perceived as most in need; other households might take a more consequentialist view and allocate food to the one who “converts” food energy into purchasing power most efficiently, never mind that others might starve; still other households might devolve into a situation where members fend for themselves, thus foregoing the benefits of the economies of scale in cooking. Careful research suggests that intrahousehold allocations of food among households in Africa may be very complex (Villa, Barrett, and Just, 2011a, 2011b). There are several reasons why husband and wife are left on their own to work out the terms of their relationship. First, marriages involve extremely personal interactions of eating, child-rearing, and sexuality, and people find it distasteful to discuss and write down expectations of behavior. While a retail store may mandate that clerks are to smile at customers, most people would heap scorn on a marriage contract that contained similar clauses detailing interactions among spouses and offspring. Second, life in a household involves repeated interaction aimed at producing a shared common good. When disputes arise, judges may, rightly or wrongly, decide that the scope of their authority should stop at the front door of the home. The marriage institution may be structured so that individual actions are not litigable, and third-party enforcement is only to be invoked at dissolution of the marriage. Third, in most societies the institution of marriage has increasingly become tied to the notion of romantic love, which is somewhat antithetical to anticipating and negotiating future household behavior. When love is eternal, nobody wants to talk about who will sweep the kitchen clean every morning. Table 7.1 gives an idea of the variation across the continent in the way members of households interact. The Demographic and Health Surveys in the early 2000s asked female respondents to indicate who had “final say” over certain common decisions taken in the context of household. The table presents the responses for three of the decisions (other decisions are correlated with these three): own health care (that is, the woman’s), large household purchases, and visits to family or relatives by the woman respondent. In the Sahel region and West Africa, two-thirds of women remarked that their husbands had the final say over these important decisions. There was

Table 7.1 Who Has Final Say over Household Decisions? (percentage of women respondents indicating self or husband)

Sahel average Senegal Mali Burkina Faso West average Ghana Benin Guinea Nigeria Central average Uganda Tanzania Kenya Rwanda Cameroon South average Mozambique Lesotho Zambia Madagascar Malawi East average Eritrea Ethiopia

Year 2005 2001 2003

2003 2001 2005 2003

2000–2001 2004–2005 2003 2005 2004 2003 2004 2001–2002 2003–2004 2004 2002 2005

Own Health Care

Self 11 14 11 9 21 37 20 16 13 34 44 43 40 26 18 31 41 37 30 28 18 47 80 15

Husband 72 67 74 75 55 35 61 51 73 43 38 39 43 38 58 41 32 44 47 12 70 21 9 33

Large Household Purchases Self 9 8 11 10 14 21 15 13 7 13 11 11 12 20 13 10 5 14 11 14 6 18 23 12

Husband 68 60 73 70 57 41 62 48 78 55 60 61 61 40 54 52 58 48 62 15 80 40 37 42

Self 16 15 18 16 18 21 17 15 18 21 25 13 23 24 22 18 19 24 17 13 19 26 41 10

Husband 57 49 62 62 45 34 45 41 60 39 42 47 39 23 44 33 32 31 56 9 38 20 20 21

173

Source: ICF International (2012). Measure DHS STATcompiler, http://www.statcompiler.com (accessed June 26, 2013). Note: Excluded categories include whether decisions are made jointly or with others or do not know.

Visits to Family or Relatives

174

Women and Development in Africa

little pretense of equality in decisionmaking authority within the household. In Central Africa, approximately half of respondents indicated that their husbands took the final decision. In southern and eastern Africa, husbands were far less likely to have the final say over decisions affecting the wellbeing of their wives. The choices that people make within households after marriage deals have been struck are shaped by gendered social norms that prescribe and proscribe. A relevant question is the degree of influence of social norms about household behavior. Are particular norms guiding most households through a wide range of behavior? If patterns of behavior are common to most households, then deviations may be perceived as violations of normal behavior rather than as eccentricities. The departures from the norm etch, in the minds of observers, what normative behavior should be. Focus groups and anthropological-style participant observation techniques can shed much light on how social norms affect household behavior. Nikièma, Haddad, and Potvin (2008) report on discourses and norms in Burkina Faso relating to decisions about seeking health-care treatment upon illness. Social norms prescribe that a woman display behavior of acquiescence toward her husband and his family. Correct behavior then results in a quid pro quo, an obligation by the husband to provide for health care. Generalizing about households and measuring the influence of social norms are important for a number of reasons. First, to predict the effects of any change that affects households, analysts need to have a reasonable model of how households operate on average. This is especially true for the economic analysis of policies that are designed to affect men or women in different ways. Second, and somewhat more selfishly, we might want to understand households in order to test whether our own experience in households is the norm. Everyone grows up in a household of some form or the other, is exposed to other households of friends and neighbors, and absorbs the rhetoric and iconography of households from various media. Are these experiences and images shared? Are there other “models” of the household besides the one we know? Characterizing the Household: The Unitary Model In economics, the first theories of household behavior emerged from a specific problem. Measurement of expenditure patterns is important for the purpose of estimating elasticities of demand. When the prices of goods and services change, the demand for them will change according to the elasticity of demand. When incomes rise or fall, so too will the demands for goods and services. (Interestingly enough, this is one of the few activities that economists perform “for the market.” Corporations have an enormous interest in estimating demand elasticities as they plan investment decisions and

Bargaining Power at Home

175

marketing campaigns.) The economists estimating price and income elasticities of demand realized early on that they had a problem. Survey researchers were collecting data on household expenditure patterns, rather than individual expenditure patterns. So many expenditures were at the level of the household—from housing to cars to food—that it was often convenient to ask about all purchases at the level of the household. The household was treated as the decisionmaker. With all the data at the household level yet a strong sense that decisions were being made through a process of interaction between individuals, analysts were challenged to demonstrate how their estimated elasticities could have predictive value, since different changes in prices and incomes might lead to changes in the relationships among the individuals within the household. Thus, early economic theories of the household explored the conditions under which the household behaved as if it were an individual. Samuelson (1956) argued that this might happen if household members had homothetic preferences. Having homothetic preferences means that the share of expenditures on a particular item does not depend on the size of the budget. If the household consists of a group of people with different budgets, who also have homothetic preferences, then the expenditure effects of changes in prices or income will not depend on the distribution of resources within the household. The household acts as if it were a single individual. Obviously, as individual budgets shift within the household, individuals get more or less goods and services. This artificial model of household behavior simply underscored how unlikely it would be that household behavior resembled individual behavior. New tools of analysis were required. Despite this admonition by Samuelson, economics research on households perhaps unwisely followed Becker’s (1981) lead in developing a different “unitary” model of the household. There are now many versions of this model. Some simply assume that the household members act in concert because they are linked together by ties of love and altruism. Everyone sits around and discusses what is best for the household, much as an individual decides whether it is better to satisfy the cravings of his tongue or the warnings of his digestive tract. Other versions of the unitary model are more elaborate and offer more explicit modeling of the ties of empathy among family members. In Becker’s (1981) version, the household contains an altruist who is much wealthier than the other household members. Suppose a two-person household has one member (the husband, say) who is comparatively more wealthy than the other member (the wife), and the husband’s utility depends on consumption of some amount of household goods, and also on the wife’s level of utility. Higher utility for the woman means higher utility for the man. The man’s altruism is said to be “effective” if the equilibrium lev-

176

Women and Development in Africa

els of consumption for household commodities were different than they would be if the utility of the wife did not enter into his utility function. If the husband were effectively altruistic and spent some of his income on his wife, then he would be making a transfer to his wife. We might then imagine the husband choosing the size of the transfer to maximize his utility, which depends on the wife’s utility. When deciding on a transfer, he will take the wife’s income into account, so the relevant budget constraint is the family budget constraint, not just his own. If his wife’s earnings are relatively high, he does not have to make as big a transfer. The upshot of this is that even if the wife knows that the husband will be redistributing income to her, she will nevertheless have incentives to maximize the family income. If she did not, she would still receive a share of the man’s income, but now from a smaller “pie.” Becker developed the model of the family in the context of the interactions of parent and child, and the insight became known as the “rotten kid” theorem. The “rotten kid” would not take an action that raised his or her own income if it lowered total family income, because the altruistic chef de famille would just confiscate the kid’s gain by giving her a lower transfer. The members of the family, then, even if self-interested, act as if they cared about joint family well-being. In this way, Becker provided a clean justification for thinking of the household as a unitary household, grounded in a social reality where households were formed by combining individuals with highly unequal endowments. In African societies, in particular, men control most assets, and so it is not implausible to model the husband as an altruistic dictator. The empirical challenge is to determine whether indeed the inequality in the household renders the husband effectively altruistic, if in fact the typical husband is altruistic with respect to his wife or wives. Characterizing the Household: Bargaining Models The main alternative to the unitary model is known as the bargaining model (Agarwal, 1997). Members of households are in situations where on occasion they might cooperate for the common good of household members, but they are also in situations where they promote their own interests. Interactions occur in a context where the household is expected to have a long life, and where people know their preferences might change (i.e., an attractive man eventually might become the failed provider for one’s children). Husband and wife then may be constantly elbowing each other gently as they pursue personal and shared interests. The study of bargaining goes beyond its applications to households and is worth developing more generally. The usual supply-and-demand model of market interaction is applicable to situations where there are large num-

Bargaining Power at Home

177

bers of people transacting in marketplaces where prices are known to everyone and sellers are not discriminating among buyers. Many other interactions take place in strategic situations, in the sense that the actions one person takes depend on his or her expectation of the actions of others. When a star soccer player sits down with the team owner to negotiate over her salary, the two do not simply look at a posted price for a player of her caliber. Instead, the player and owner bargain. Offers and counteroffers are made, and a salary is finally agreed on. Sometimes the bargaining may go on for a long time, to the extent that the player misses some games. Clearly, supply-and-demand analysis will yield only partial insight into the outcomes that might result from such interactions. Nash (1950, 1953) derived one of the first models of this two-person interaction. He supposed that any predicted outcome of a bargaining situation should satisfy certain properties. For instance, the outcome of their bargaining should not be inefficient, in the sense that the bargainers should not agree to do something if there were an alternative available that made one of them better off without making the other worse off. That latter alternative is clearly preferable (by definition) and should be the outcome of the bargaining. Another property Nash assumed was that only people’s rankings over alternatives mattered, not some unknown cardinal measure of how much people actually valued the alternatives. That is, the utility functions of the bargainers should only matter in how they ranked alternatives, and not in how much abstract utility was enjoyed by the person. Nash also assumed that bargaining “ability” should be the same for the two bargainers. The last, and more controversial, assumption was that the outcome should be independent of irrelevant alternatives. Outcomes that would not be chosen by the bargainers should not affect the bargaining outcome. That is, in a situation where two people are bargaining, if an alternative that they did not choose in the first instance were changed, then, if they also do not choose that alternative in the second instance, the outcome that they do choose should be the same as in the first instance. Alternatively, suppose that options that will not be chosen were added to a bargaining situation; then again the resulting choice should still be the same as before. With these assumptions, Nash proved that the resulting outcome of bargaining would be such that the product of the gains of the two players would be maximized. The gains represent the increase in utility over a fallback position, or threat point, that describes the utility of each person should bargaining fail. (Just how to think about threat points and fallback positions is specified very clearly in Chiu and Yang [1999].) Consider, for example, the simple problem of bargaining over the distribution of a cake. If the bargainers agree, they each receive a share of the cake. If they disagree, they each receive nothing. So their fallback positions are zero. Then,

178

Women and Development in Africa

the Nash bargaining solution is a share of the cake, s, that maximizes the product of the gains: (s – 0) (1 – s – 0), or (s) (1 – s). Clearly, the share that maximizes this expression is 1/2: the bargainers would split the cake evenly. Suppose one bargainer had a fallback position, such as being able to get a cupcake worth 1/4 of a cake should the bargaining fail. Then the share that maximizes the product of the gains would be the share that maximized (s – 1/4) (1 – s – 0). This share would be 5/8; the person with the cupcake fallback now gets more than 1/2. This model can be applied to the context of the household. Suppose a husband and wife were bargaining over how much money the wife should get if she worked on the husband’s farm. The husband owns the land. A farm production function summarizes how output depends on the labor supplied by his wife. That is, the husband gets zero if the wife does not work, and gets F(l) if she agrees to work. If working, she must work a fixed amount of time, l. If she does not work, she can go off and work elsewhere and earn an amount of money If. If the wife does not work for the husband, then the husband gets income Im. The husband values income according to the utility function V(), while the wife has utility function U(). So the Nash bargaining solution is a value for w, the wage paid to the wife by the husband, which maximizes [U(w) – U(I f )] [V(F(l) – w) – V(Im)].

The solution, w*, depends on the levels of the threat points. The higher I f is, the higher will be w*, and the higher Im is, the lower will be w*. Measuring these threat points will be a major issue in empirical applications of bargaining models. The Nash bargaining solution is just the beginning in terms of modeling the bargaining household. All of the fascinating and complex results of game theory apply once one begins to go beyond the simple cake division problem. For instance, the repeated, dynamic nature of household interaction means that agents need to strategize about the longer-term effects of their short-term choices. Basu (2006) shows that this strategic dimension greatly complicates the static analysis. In addition, households almost by definition are about mixing common interest with self-interest. The maximization problem of a household member, as Becker (1981) understood, would not be the same as that of an individual in an anonymous marketplace. This is even truer for parenting decisions regarding children. Helping children in the transition to adult independence, where social norms mean that the socially approved manners and opportunities of young men and women will be quite different, means that parental behavior will likely be gendered. The self-interest of the child may not always mesh, however, with the altruism of the parent.

Bargaining Power at Home

179

Distinguishing Models: Looking at Household Expenditures Suppose men typically care less than women about children’s nutritional status. They tend to spend more of their income on luxury items for themselves (such as cigarettes and alcohol). Women, on the other hand, care about their children and spend relatively more of their income on good food and health care. This certainly sounds like a plausible characterization of men and women in many societies over much of history. How would the characterization be verified, if men and women always lived in households and expenditures of people were recorded as household aggregates rather than at the level of the individual? Moreover, for every anecdote that a historian or anthropologist produced about some society, another anecdote could be found singing the praises of a wonderful man who cared about his children while his wife drank and cursed in the local tavern. This is the empirical problem of characterizing households in a nutshell. Theories of household behavior have very different predictions about what empirical investigation would find. The Nash bargaining approach to interactions between two persons suggests that threat points, or fallback positions, influence the outcomes of the bargaining. The discussion above suggested that households are good places to think about two-person bargaining, since they are usually composed of a husband and wife who have responsibility over a large range of economic actions. The model proposes that expenditures on male-oriented luxury items (cigarettes, say) should rise when the man’s income or assets increase. Likewise, expenditures on nutrition for children might be more influenced by increases in women’s income or assets. When looking at the income effects on expenditures, then, the regression estimates using data from a household survey might show that the partial derivatives of how income of men and women affects expenditures might vary according to the following: ∂xc

∂Ym

>

∂xc

∂Yf

while

∂xn

∂Ym