32204 world development report 2006 Equity and Development world development report 2006 Equity and Development world development report 2006 Equity and Development A copublication of The World Bank and Oxford University Press ©2005 The International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org E-mail: feedback@worldbank.org All rights reserved 1 2 3 4 08 07 06 05 A copublication of The World Bank and Oxford University Press. Oxford University Press 198 Madison Avenue New York NY 10016 This volume is a product of the staff of the International Bank for Reconstruction and Devel- opment / The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any ter- ritory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone: 978-750-8400; fax: 978-750-4470; Internet: www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street, NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. ISBN-10: 0-8213-6249-6 ISBN-13: 978-0-8213-6249-5 ISSN: 0163-5085 eISBN: 0-8213-6250-X DOI: 10.1596/978-0-8213-6249-5 Cover image: Dream of a Sunday Afternoon in Alameda Park 1947­48 (fresco) by Diego Rivera. The mural is located in Museo Mural Diego Rivera, Mexico City. Reproduction authorized by the Instituto Nacional de Bellas Artes y Literatura­Mexico; Copyright © Photograph by Francisco Kochen. Library of Congress Cataloging-in-Publication Data has been applied for. Contents Foreword xi Acknowledgments xiii Abbreviations and Data Notes xiv Overview 1 Inequity within and across nations 4 Why does equity matter for development? 7 Leveling the economic and political playing fields 9 1 Introduction 18 Equity and inequality of opportunity: the basic concepts 18 Inequality traps 20 A brief preview of the Report 23 Part I Inequity within and across countries 25 focus 1 on Palanpur 26 2 Inequity within countries: individuals and groups 28 Inequalities in health 29 Inequalities in education 34 Economic inequalities 36 The relationship between group differences and inequality 43 Agency and equity: inequalities of power 48 The inequality trap for women 51 3 Equity from a global perspective 55 Examples and concepts 55 Global inequalities in health 56 Global inequalities in education 60 Global inequalities in income and expenditure 62 v vi CONTENTS Global inequalities in power 66 A glimpse of the future 68 focus 2 on empowerment 70 Part II Why does equity matter? 73 4 Equity and well-being 76 Ethical and philosophical approaches to equity 76 Equity and legal institutions 78 People prefer fairness 80 Income inequality and poverty reduction 84 5 Inequality and investment 89 Markets, wealth, status, and investment behavior 89 The evidence on underinvestment 96 Inequalities and investment 101 focus 3 on Spain 106 6 Equity, institutions, and the development process 107 The distribution of power and institutional quality: circles vicious and virtuous 107 Institutions and political inequality matter for development: historical evidence 109 Institutions and political inequality matter for development: contemporary evidence 113 Transitions to more equitable institutions 118 Conclusion 124 focus 4 on Indonesia 126 Part III Leveling the economic and political playing fields 129 7 Human capacities 132 Early childhood development: a better start in life 132 Basic education: expanding opportunities to learn 135 Toward better health for all 141 Social protection: managing risk and providing social assistance 148 Summary 155 Contents vii 8 Justice, land, and infrastructure 156 Building equitable justice systems 156 Toward greater equity in access to land 162 Providing infrastructure equitably 168 Summary 175 focus 5 on taxation 176 9 Markets and the macroeconomy 178 How markets relate to equity 178 Achieving equity and efficiency in financial markets 179 Achieving equity and efficiency in labor markets 185 Product markets and trade reform 193 Macroeconomic management and equity 198 focus 6 on regional inequality 204 10 Achieving greater global equity 206 Making global markets work more equitably 207 Providing development assistance to help build endowments 218 Transitions to greater equity 221 Summary 223 focus 7 on drug access 224 Epilogue 226 Bibliographic note 231 Endnotes 233 References 247 Selected Indicators 275 Measuring Equity 277 Selected world development indicators 289 Index 309 viii CONTENTS Boxes 2.1 Unequal opportunities persist across generations 7.8 Better maternal health in Malaysia and Sri Lanka 144 in Brazil 29 7.9 Mobilizing support for universal coverage 2.2 Unequal assets, unequal opportunities: AIDS orphans in in Thailand 146 Southern Africa 33 7.10 Public works programs: key issues 152 2.3 Health improvements and greater health equity 7.11 Africa's orphans and public action 155 in Peru 34 8.1 Increasing legal literacy and public awareness: 2.4 Child test scores in Ecuador: the role of wealth, parental "My Rights" on Armenian public television 157 education, and place of residence 35 8.2 Affirmative action in India and the United States 158 2.5 Beware of intercountry comparisons of inequality! 38 8.3 State frameworks and customary institutions 2.6 Revisiting the Kuznets hypothesis for economic growth and in South Africa 160 inequality 44 8.4 The impact of legal aid in Ecuador 160 2.7 Inequitable agencies and institutions in Pakistan 48 8.5 Bogota, Colombia: civic culture program 161 2.8 Legacies of discrimination and the reproduction 8.6 Land reform in South Africa: picking up steam 164 of inequalities and poverty among the Batwa in Uganda 49 8.7 Clarifying how customary rights fit with formal systems 166 2.9 Sex ratios and "missing women" 51 8.8 Land and output tax combinations 168 3.1 Three competing concepts of inequality: global, international, and intercountry 57 8.9 Lagging infrastructure in Africa 170 4.1 A simple representation of different concepts 8.10 The distributional impact of infrastructure privatization in of equity 78 Latin America: a mixed bag 171 4.2 Capuchin monkeys don't like inequity either . . . 82 8.11 The pro-poor agenda for urban water in Senegal 173 4.3 Worker perceptions of unfairness, product quality, and 8.12 Addressing accountability and transparency in consumer safety 83 telecommunications in Brazil and Peru 174 6.1 Banking in the nineteenth century, Mexico and the United 9.1 Markets and development: policy, equity, and social welfare States 109 in China 180 6.2 Growth with poor institutions does not last 113 9.2 Too much and too little regulation: Russia before and after the transition 182 6.3 Polarization, conflict, and growth 118 9.3 Organizing in the informal economy 190 6.4 Aiding equitable growth in early modern Britain: the role of the Poor Laws 120 9.4 Employment protection legislation 191 7.1 ECD programs are an essential ingredient for the 9.5 Two cases of labor market reform: One comprehensive, attainment of education for all 134 one partial 192 7.2 School fees--an instrument of exclusion 9.6 Did the Russian 1998 crisis have equitable or accountability? 137 consequences? 201 7.3 Desegregating Roma schools in Bulgaria: 10.1 International law, globalization, and equity 207 the Vidin model 138 10.2 Making migrant worker schemes more development 7.4 Remedying education: the Balsakhi program friendly 210 in India 140 10.3 Cotton subsidies are huge--and tenacious 212 7.5 School vouchers: efficient and equitable? 141 10.4 Will improved working conditions in Cambodia's textile 7.6 Working with mothers to treat malaria 142 industry survive the end of the quota system? 214 7.7 Poor people and ethnic minorities receive 10.5 Expanding access to antiretroviral drugs lower-quality care 143 in South Africa 215 Figures 1 Wealth matters for the immunization of children 5 4 A long-run diverging trend in income inequality begins to 2 Opportunities are determined early 6 reverse because of growth in China and India 7 3 Life expectancy improved and became more equal--until 5 Children's performance differs when their caste is made the onset of the AIDS crisis 6 salient 8 Contents ix 6 Catching up through early interventions 11 3.11 Absolute poverty declined globally, but not in 7 Better to be close to economic opportunities 15 every region 66 1.1 The interaction of political, economic, and sociocultural 3.12 There is no one-to-one relationship between voice and inequalities 20 income 67 2.1 Infant mortality varies across countries but also by mother's 4.1 The distribution of observed offers in ultimatum education within countries 30 games 80 2.2 Stunting levels of children born in rural versus urban areas 4.2 Views on inequality from the World Values Survey 84 are far from the same 30 4.3 Growth is the key to poverty reduction . . . 85 2.3 Access to childhood immunization services depends on 4.4 . . . and, on average, growth is distribution-neutral 85 parents' economic status 31 4.5 The national growth incidence curves for Tunisia 2.4 Stunting and underweight in Cambodia 33 1980­1995 and Senegal 1994­2001 86 2.5 Education levels vary across countries, but they also depend 4.6 Greater inequality reduces the power of growth to reduce on gender of household head 36 poverty 87 2.6 Education levels vary by country and between rural and 5.1 In rural Kerala and Tamil Nadu, the rich access most of the urban sectors 37 credit and pay relatively low rates 90 2.7 The share of inequality in years of schooling attributable 5.2 Children's performance differs when their caste is made to differences between males and females has been public 96 declining 37 5.3 Returns to capital vary with firm size: evidence from small 2.8 Market capitalization controlled by the top 10 families in Mexican firms 97 selected countries, 1996 38 5.4 Inefficient allocation of resources; the example of the 2.9 Africa and Latin America have the world's highest levels of Gounders vs. the outsiders 98 inequality 39 5.5 Average returns for switching to pineapples as an intercrop 2.10 Between-group inequality decompositions: social group of can exceed 1,200 percent 98 the household head 40 5.6 Profit-wealth ratios are highest for the smallest 2.11 Between-group inequality decompositions: education of the farms 99 household head 41 6.1 Countries with more secure property rights have higher 2.12 Location, education, and social groups can make a average incomes 108 difference: regressions of total inequality on shares of 6.2 Low population density in 1500 is associated with a lower between-group inequality of different household risk of expropriation today 110 characteristics 43 6.3 Worse environments for European settlers are associated 2.13 Women work longer hours than do men 53 with worse institutions today 110 3.1 Vanishing twin peaks in life expectancy at birth 58 6.4 A worse environment for settlers is associated with fewer 3.2 Life expectancy is highly correlated with income, constraints on the executive at independence 111 particularly in poor countries 59 6.5 Constraints on the executive are greater in Mauritius than 3.3 The distribution of years of schooling improved greatly in in Guyana 116 the second half of the twentieth century 60 6.6 GDP per capita is rising in Mauritius, not 3.4 Mean years of schooling increased while inequality declined in Guyana 116 across birth cohorts 60 6.7 Inequality in Britain began to fall around 1870 121 3.5 Gender disparities in years of schooling declined but 7.1 Children from better-off households have a big edge in remained significant in some regions 61 cognitive abilities by age three 133 3.6 Incomes range broadly across countries 7.2 Early childhood interventions are good investments 133 and individuals 62 7.3 Catching up through early intervention 134 3.7 Since 1950, intercountry inequality increased while 7.4 Boosting enrollments is not enough to overcome the international inequality declined 63 learning gap 139 3.8 Unlike relative inequality, absolute inequality has been 7.5 Almost all countries spend more on social insurance than steadily increasing 63 on social assistance (percent of GDP) 149 3.9 The inequality decline between countries was neutralized by 8.1 Unequal initial land distributions go together with slower increases within countries 64 economic growth 163 3.10 Inequality between countries became much more 8.2 Title to land increases investment and access to important over the long run 65 credit 165 x CONTENTS 8.3 Poor families did not benefit from an expansion of access in 9.4 It's better for household welfare to be close to economic Africa 172 opportunities 195 8.4 Poorer households have lower-quality water and pay more 9.5 Weaker institutions are associated with macroeconomic in Niger 173 volatility and crises 199 9.1 Poland's stock market started slowly but then surpassed the 9.6 Labor shares fall during crises and don't fully recover Czech Republic's 185 afterward 200 9.2 Patterns of employment and unemployment vary widely 9.7 In Argentina, the wealthy had a way out during across African countries 187 the crisis 201 9.3 Different labor market institutional setups can yield equally 10.1 Wage differentials are substantially larger today than at the good productivity growth paths: Scandinavia versus the end of the nineteenth century 208 United States 188 10.2 More subsidies than aid 220 Tables 2.1 Decomposition of inequality between and within 9.2 Financial policy and institutions are often captured by the communities 42 few: case study evidence 181 2.2 Percentage of women who have ever experienced physical or 9.3 Fiscal costs of selected banking crises 200 sexual violence by an intimate partner 54 10.1 ODA as a share of GNI, 2002, 2003, and simulation 3.1 Increases in life expectancy at birth slowed down for 2006 220 dramatically in the 1990s 58 A1 Poverty 278 3.2 Mean years of schooling increased continuously while A2 Income/consumption inequality measures 280 inequality declined 61 A3 Health 282 3.3 Mobility matrix in absolute country per capita incomes, A4 Education 284 1980 to 2002 66 Classification of economies by region and income 291 5.1 The effect of income shocks on consumption, Côte d'Ivoire 92 1 Key indicators of development 292 5.2 Farm size productivity differences, selected countries 99 2 Poverty and income distribution 294 7.1 Examples of social protection programs 150 3 Economic activity 296 7.2 Targeting performance of conditional 4 Trade, aid, and finance 298 transfer schemes 153 5 Key indicators for other economies 300 9.1 Two pathologies in the interaction between equity and growth 179 Foreword Poverty reduction comes about through individuals, families and communities taking advan- tage of the opportunities available to them by working, investing and innovating to better their lives. But we live in a world of extraordinary inequalities in opportunity, both within and across countries. Even the basic opportunity for life itself is disparately distributed: whereas less than half of one percent of children born in Sweden die before their first birthday, this is the case for close to 15% of all children born in Mozambique. Within El Salvador, the infant mortality rate is 2% for children of educated mothers, but 10% for those whose mothers have no schooling. In Eritrea, immunization coverage is close to 100% for children in the richest fifth of the popu- lation, but only 50% for the bottom fifth. These children can not be blamed for the circumstances into which they are born, yet their lives--and their ability to contribute to the development of their nations--are powerfully shaped by them. That is why the World Development Report 2006, the twenty-eighth in this annual series, looks at the role of equity in the process of development. Equity is defined in terms of two basic principles. The first is equal opportunities: that a person's life achievements should be determined primarily by his or her talents and efforts, rather than by pre-determined circumstances such as race, gender, social or family background. The second principle is the avoidance of deprivation in outcomes, particularly in health, education and consumption levels. For many if not most people, equity is of intrinsic importance as a development goal in its own right. But this report goes further, by presenting persuasive evidence that a broad sharing of economic and political opportunities is also instrumental for economic growth and devel- opment. This is for economic reasons, because greater equity can lead to a fuller and more effi- cient use of a nation's resources. It is also for political and institutional reasons: excessive inequalities in power and influence can lead to political, social and economic institutions that are less conducive to long-term growth. Few today's prosperous societies, if any, developed by excluding the majority of their people from economic and political opportunities. The implication of this message for the work of the World Bank and others in the develop- ment community is that a focus on equity should be a central concern in the design and imple- mentation of policy for development and growth. This insight needs to be integrated into both analytical and operational work on core areas of development design, including the role and functioning of markets. Public action should seek to expand the opportunity sets of those who, in the absence of policy interventions, have the least resources, voice and capabilities. It should do so in a manner that respects and enhances individual freedoms, as well as the role of markets in allocating resources. Equity in the international arena is also a central concern, and can play a powerful comple- mentary role to domestic action. In a globally interconnected world, leveling the international playing fields, both economically and politically, will help domestic efforts to combine equity with efficiency and growth. In my view, the evidence that equity and economic efficiency as well as growth are comple- mentary in the long run helps to integrate the main two components of the World Bank's poverty reduction strategy. The focus on broadening opportunities strongly supports the first xi xii WORLD DEVELOPMENT REPORT 2005 pillar of the Bank's development strategy, namely enhancing the investment climate for every- one. Together with the interdependence between the economic and political dimensions of development it also reinforces the importance of empowerment. This report shows that the two pillars are not independent from each other in supporting development, but instead are intricately linked with one another. It is my hope that this report will have a real influence in the way that we and our development partners understand, design and implement development policies. Paul D. Wolfowitz President The World Bank Acknowledgments This Report has been prepared by a core team led by Francisco H.G. Ferreira and Michael Wal- ton, and comprising Tamar Manuelyan Atinc, Abhijit Banerjee, Peter Lanjouw, Marta Menén- dez, Berk Özler, Giovanna Prennushi,Vijayendra Rao, James Robinson, and Michael Woolcock. Important additional contributions were made by Anthony Bebbington, Stijn Claessens, Mar- garet Ellen Grosh, Karla Hoff, Jean O. Lanjouw, Xubei Lou, Ana Revenga, Caroline Sage, Mark Sundberg, and Peter Timmer. The team was assisted by Maria Caridad Araujo, Andrew Beath, Ximena del Carpio, Celine Ferre, Thomas Haven, Claudio E. Montenegro, and Jeffery C. Tan- ner. The work was conducted under the general guidance of François Bourguignon. Extensive and excellent advice was received from Anthony B.Atkinson,Angus Deaton, Naila Kabeer, Martin Ravallion, and Amartya Sen, to whom the team is grateful without implication. Many others inside and outside the World Bank also provided helpful comments; their names are listed in the Bibliographical Note. The Development Data Group contributed to the data appendix and was responsible for the Selected World Development Indicators. Much of the background research was supported by a multidonor programmatic trust fund, the Knowledge for Change Program, funded by Canada, the European Community, Finland, Norway, Sweden, Switzerland, and the United Kingdom. The team undertook a wide range of consultations for this Report, which included work- shops in Amsterdam, Beirut, Berlin, Cairo, Dakar, Geneva, Helsinki, Hyderabad, London, Milan, Nairobi, New Delhi, Oslo, Ottawa, Paris, Rio de Janeiro, Stockholm, Tokyo, Venice, and Washington, D.C.; videoconferences with sites in Bogota, Buenos Aires, Mexico City, and Tokyo; and an on-line discussion of the draft Report. The team wishes to thank participants in these workshops, videoconferences, and discussions, which included researchers, government officials, and staff of nongovernmental and private-sector organizations. Rebecca Sugui served as executive assistant to the team, Ofelia Valladolid as program assis- tant, Madhur Arora and Jason Victor as team assistants. Evangeline Santo Domingo served as resource management assistant. Bruce Ross-Larson was the principal editor. Book design, editing, and production were coordinated by the World Bank's Office of the Publisher under the supervision of Susan Graham and Monika Lynde. xiii Abbreviations and Data Notes Abbreviations The following abbreviations are used in this Report: AA Affirmative action NGO Nongovernmental organization AIDS Acquired immune deficiency syndrome ODA Official development assistance CCP Chinese Communist Party OECD Organisation for Economic Co-operation and DAC Development Assistance Committee Development DHS Demographic and Health Survey PPA Participatory Poverty Assessment ECD Early child development PPP Purchasing-power parity EPL Employment protection legislation PROMESA Promoción y Mejoramiento de la Salud FDI Foreign direct investment SMEs Small and medium enterprises GDP Gross domestic product TAC Treatment Action Campaign GHG Greenhouse gas TIMSS Third International Mathematics and Science GNI Gross national income Study HIPC Heavily Indebted Poor Countries TRIPs Trade-related aspects of intellectual property HIV Human immunodeficiency virus rights ICOR Incremental Capital-Output Ratio U.N. United Nations ICRISAT International Crop Research Institute UNCTAD United Nations Conference on Trade and in the Semi-Arid Tropics Development IDA International Development Association UNDP United Nations Development Programme ILO International Labour Organization UNAIDS Joint United Nations Programme on HIV/AIDS IMF International Monetary Fund UNICEF United Nations International Children's IMS Intercontinental Marketing Services Emergency Fund KDP Kecamatan Development Project VAT Value added tax MDG Millennium Development Goals WHO World Health Organization MMM Movement Militant Mauricien WTO World Trade Organization MSF Médecins Sans Frontières WWII World War II NAFTA North American Free Trade Agreement Data notes The use of the word countries to refer to economies implies The countries included in regional and income groupings in no judgment by the World Bank about the legal or other status this Report are listed in the Classification of Economies table of a territory. The term developing countries includes low- and at the beginning of the Selected World Development Indica- middle-income economies and thus may include economies in tors. Income classifications are based on GNP per capita; transition from central planning, as a matter of convenience. thresholds for income classifications in this edition may be The term advanced countries may be used as a matter of conve- found in the Introduction to Selected World Development nience to denote high-income economies. Indicators. Group averages reported in the figures and tables Dollar figures are current U.S. dollars, unless otherwise are unweighted averages of the countries in the group, unless specified. Billion means 1,000 million; trillion means 1,000 bil- noted to the contrary. lion. xiv Overview Consider two South African children born on great business idea (such as an innovation the same day in 2000. Nthabiseng is black, to increase agricultural production), she born to a poor family in a rural area in the would find it much harder to persuade a Eastern Cape province, about 700 kilometers bank to lend her money at a reasonable from Cape Town. Her mother had no formal interest rate. Pieter, having a similarly schooling. Pieter is white, born to a wealthy bright idea (say, on how to design an family in Cape Town. His mother completed improved version of promising software), a college education at the nearby prestigious would likely find it easier to obtain credit, Stellenbosch University. with both a college diploma and quite pos- On the day of their birth, Nthabiseng and sibly some collateral. With the transition to Pieter could hardly be held responsible for democracy in South Africa, Nthabiseng is their family circumstances: their race, their able to vote and thus indirectly shape the parents' income and education, their urban policy of her government, something or rural location, or indeed their sex. Yet denied to blacks under apartheid. But the statistics suggest that those predetermined legacy of apartheid's unequal opportunities background variables will make a major dif- and political power will remain for some ference for the lives they lead. Nthabiseng time to come. It is a long road from such a has a 7.2 percent chance of dying in the first (fundamental) political change to changes year of her life, more than twice Pieter's 3 in economic and social conditions. percent. Pieter can look forward to 68 years As striking as the differences in life of life, Nthabiseng to 50. Pieter can expect to chances are between Pieter and Nthabiseng complete 12 years of formal schooling, in South Africa, they are dwarfed by the Nthabiseng less than 1 year.1 Nthabiseng is disparities between average South Africans likely to be considerably poorer than Pieter and citizens of more developed countries. throughout her life.2 Growing up, she is less Consider the cards dealt to Sven--born on likely to have access to clean water and sani- that same day to an average Swedish tation, or to good schools. So the opportuni- household. His chances of dying in the ties these two children face to reach their full first year of life are very small (0.3 percent) human potential are vastly different from and he can expect to live to the age of 80, 12 the outset, through no fault of their own. years longer than Pieter, and 30 years more Such disparities in opportunity translate than Nthabiseng. He is likely to complete into different abilities to contribute to 11.4 years of schooling--5 years more than South Africa's development. Nthabiseng's the average South African. These differences health at birth may have been poorer, owing in the quantity of schooling are com- to the poorer nutrition of her mother dur- pounded by differences in quality: in the ing her pregnancy. By virtue of their gender eighth grade, Sven can expect to obtain a socialization, their geographic location, and score of 500 on an internationally compara- their access to schools, Pieter is much more ble math test, while the average South likely to acquire an education that will African student will get a score of only enable him to put his innate talents to full 264--more than two standard deviations use. Even if at age 25, and despite the odds, below the Organisation for Economic Co- Nthabiseng manages to come up with a operation and Development (OECD) median. 1 2 WORLD DEVELOPMENT REPORT 2006 Nthabiseng most likely will never reach that The second set of reasons why equity and grade and so will not take the test.3 long-term prosperity can be complementary These differences in life chances across arises from the fact that high levels of nationality, race, gender, and social groups economic and political inequality tend to will strike many readers as fundamentally lead to economic institutions and social unfair. They are also likely to lead to wasted arrangements that systematically favor the human potential and thus to missed devel- interests of those with more influence. Such opment opportunities. That is why World inequitable institutions can generate eco- Development Report 2006 analyzes the rela- nomic costs. When personal and property tionship between equity and development. rights are enforced only selectively, when By equity we mean that individuals budgetary allocations benefit mainly the should have equal opportunities to pursue a politically influential, and when the distri- life of their choosing and be spared from bution of public services favors the wealthy, extreme deprivation in outcomes. The main both middle and poorer groups end up with message is that equity is complementary, in unexploited talent. Society, as a whole, is some fundamental respects, to the pursuit then likely to be more inefficient and to miss of long-term prosperity. Institutions and out on opportunities for innovation and policies that promote a level playing field-- investment. At the global level, when devel- where all members of society have similar oping countries have little or no voice in chances to become socially active, politically global governance, the rules can be inappro- influential, and economically productive-- priate and costly for poorer countries. contribute to sustainable growth and devel- These adverse effects of unequal opportu- opment. Greater equity is thus doubly good nities and political power on development are for poverty reduction: through potential all the more damaging because economic, beneficial effects on aggregate long-run political,and social inequalities tend to repro- development and through greater opportu- duce themselves over time and across genera- nities for poorer groups within any society. tions. We call such phenomena "inequality The complementarities between equity traps." Disadvantaged children from families and prosperity arise for two broad sets of at the bottom of the wealth distribution do reasons. First, there are many market fail- not have the same opportunities as children ures in developing countries, notably in the from wealthier families to receive quality markets for credit, insurance, land, and education. So these disadvantaged children human capital. As a result, resources may can expect to earn less as adults. Because the not flow where returns are highest. For poor have less voice in the political process, example, some highly capable children, like they--like their parents--will be less able to Nthabiseng, may fail to complete primary influence spending decisions to improve schooling, while others, who are less able, public schools for their children. And the may finish university. Farmers may work cycle of underachievement continues. harder on plots they own than on those The distribution of wealth is closely corre- they sharecrop. Some efficient developing- lated with social distinctions that stratify peo- country producers of agricultural com- ple, communities, and nations into groups modities and textiles are shut out of some that dominate and those that are dominated. OECD markets, and poor unskilled workers These patterns of domination persist because have highly restricted opportunities to economic and social differences are rein- migrate to work in richer countries. forced by the overt and covert use of power. When markets are missing or imperfect, Elites protect their interests in subtle ways, by the distributions of wealth and power affect exclusionary practices in marriage and kin- the allocation of investment opportunities. ship systems, for instance, and in ways that Correcting the market failures is the ideal are less subtle, such as aggressive political response; where this is not feasible, or far manipulation or the explicit use of violence. too costly, some forms of redistribution-- Such overlapping political, social, cultural, of access to services, assets, or political and economic inequalities stifle mobility. influence--can increase economic efficiency. They are hard to break because they are so Overview 3 closely tied to the ordinary business of life. on property rights for all; and greater fair- They are perpetuated by the elite, and often ness in markets. But policies to level the eco- internalized by the marginalized or oppressed nomic playing field face big challenges. groups, making it difficult for the poor to There is unequal capacity to influence the find their way out of poverty. Inequality traps policy agenda: the interests of the disenfran- can thus be rather stable, tending to persist chised may never be voiced or represented. over generations. And when policies challenge privileges, The report documents the persistence of powerful groups may seek to block reforms. these inequality traps by highlighting the Thus, equitable policies are more likely to be interaction between different forms of in- successful when leveling the economic play- equality.It presents evidence that the inequal- ing field is accompanied by similar efforts to ity of opportunity that arises is wasteful and level the domestic political playing field and inimical to sustainable development and introduce greater fairness in global gover- poverty reduction. It also derives policy nance. implications that center on the broad concept Third, there may be various short-run, of leveling the playing field--both politically policy-level tradeoffs between equity and effi- and economically and in the domestic and ciency. These are well recognized and exten- the global arenas. If the opportunities faced sively documented. The point is that the by children like Nthabiseng are so much (often implicit) cost-benefit calculus that more limited than those faced by children like policymakers use to assess the merits of vari- Pieter or Sven, and if this hurts development ous policies too often ignores the long-term, progress in the aggregate, then public action hard-to-measure but real benefits of greater has a legitimate role in seeking to broaden the equity. Greater equity implies more efficient opportunities of those who face the most economic functioning, reduced conflict, limited choices. greater trust, and better institutions, with Three considerations are important at dynamic benefits for investment and growth. the outset. First, while more even playing To the extent that such benefits are ignored, fields are likely to lead to lower observed policymakers may end up choosing too little inequalities in educational attainment, equity. health status, and incomes, the policy aim is By the same token, however, those inter- not equality in outcomes. Indeed, even with ested in greater equity must not ignore the genuine equality of opportunities, one would short-term tradeoffs. If individual incen- always expect to observe some differences in tives are blunted by income redistribution outcomes owing to differences in preferences, schemes that tax investment and produc- talents, effort, and luck.4 This is consistent tion too steeply, the result will be less inno- with the important role of income differences vation, less investment, and less growth. in providing incentives to invest in education The history of the twentieth century is lit- and physical capital, to work, and to take tered with examples of ill-designed policies risks. Of course outcomes matter, but we are pursued in the name of equity that seriously concerned with them mainly for their influ- harmed--rather than spurred--growth ence on absolute deprivation and their role in processes by ignoring individual incentives. shaping opportunities. A balance must be sought, taking into Second, a concern with equality of oppor- account both the immediate costs to indi- tunity implies that public action should focus vidual incentives and the long-term benefits on the distributions of assets, economic of cohesive societies, with inclusive institu- opportunities, and political voice, rather than tions and broad opportunities. directly on inequality in incomes. Policies can While careful assessment of policy design contribute to the move from an "inequality in local contexts is always important, equity trap"to a virtuous circle of equity and growth considerations need to be brought squarely by leveling the playing field--through greater into the center of both diagnosis and policy. investment in the human resources of the This is not intended as a new framework. It poorest; greater and more equal access to means integrating and extending existing public services and information; guarantees frameworks: equity is central both to the 4 WORLD DEVELOPMENT REPORT 2006 investment environment and to the agenda Within-country inequities of empowerment, working through the have many dimensions impact on institutions and specific policy Direct quantification of inequality of designs. Some may value equity for its own opportunity is difficult, but one analysis of sake, others primarily for its instrumental Brazil provides an illustration (chapter 2). role in reducing absolute poverty, the World Earnings inequality in 1996 was divided Bank's mission. into one share attributable to four predeter- This report recognizes the intrinsic value mined circumstances that lie beyond the of equity but aims primarily to document control of individuals--race, region of how a focus on equity matters for long-run birth, parental education, and paternal development. It has three parts. occupation at birth--and a residual share. · Part I considers the evidence on inequal- These four circumstances account for ity of opportunity, within and across around one-quarter of overall differences in countries. Some attempts to quantify earnings between workers. Arguably, other inequality of opportunity are reviewed determinants of opportunity are equally but, more generally, we rely on evidence predetermined at birth but not included in of highly unequal outcomes across this set--for example, gender, family groups defined by predetermined cir- wealth, or the quality of primary schools. cumstances--such as gender, race, fam- Because such variables are not included in ily background, or country of birth--as the inequality "decomposition," the results markers for unequal opportunities. here can be seen as lower-bound estimates of inequality of opportunity in Brazil. · Part II asks why equity matters. It dis- Unfortunately, predetermined (and thus cusses the two channels of impact (the morally irrelevant) circumstances deter- effects of unequal opportunities when mine much more than just future earnings. markets are imperfect, and the conse- Education and health are of intrinsic value quences of inequity for the quality of and affect the capacity of individuals to institutions a society develops) as well as engage in economic, social, and political intrinsic motives. life. Yet children face substantially different · Part III asks how public action can level opportunities to learn and to lead healthy the political and economic playing fields. lives in almost all populations, depending In the domestic arena, it makes the case on asset ownership, geographic location, or for investing in people, expanding access parental education, among others. Consider to justice, land, and infrastructure, and how access to a basic package of immuniza- promoting fairness in markets. In the tion services differs for the rich and the international arena, it considers leveling poor across countries (figure 1). the playing field in the functioning of There is substantial inequality in access global markets and the rules that govern between, for example, Egypt, where almost them--and the complementary provi- everyone is covered (on the left), and Chad, sion of aid to help poor countries and where more than 40 percent of children are poor people build greater endowments. excluded (on the right). Yet the disparities The remainder of this overview provides a can be as large within some countries as they summary of the principal findings. are across all nations in the sample. In Eritrea, for instance, the richest fifth enjoys almost complete coverage, but almost half of Inequity within all children in the poorest fifth are excluded. and across nations Significant gender differences also per- From an equity perspective, the distribu- sist in many parts of the world. In parts of tion of opportunities matters more than East and South Asia, notably in certain areas the distribution of outcomes. But opportu- in rural China and northwest India, the nities, which are potentials rather than opportunity to life itself can depend on one actuals, are harder to observe and measure single predetermined characteristic: sex. than outcomes. These regions have significantly more boy Overview 5 Figure 1 Wealth matters for the immunization of children Percentage not covered 70 60 50 40 Poorest 30 20 10 0 Wealthiest i a (*) bia (*) ala (*) (*) en ali Egypt anda Peru oros Faso Haiti ublic M Africa KenyaM alaw Brazilbia Benin India Togo bodia NigerChad Bolivia Yem eroon Eritrea Rw VietnamTurkey orocco Ghana biqueGuinea Com Uganda auritani EthiopiaRep Pakistan JordanColom Zam Paraguay GuatemTanzaniaIndonesiaenistan M Philippines Cam M adagascar ozam Cam South Bangladesh Burkina M M Kazakhstan Turkm African Central Source: Authors' own calculations from Demographic Health Survey (DHS) data Note: * indicates that the poorest quintile have higher access to childhood immunization services than the wealthiest quintile. The continuous orange line represents the overall percentage of children without access to a basic immunization package in each country, while the endpoints indicate the percentages for the top and the bottom quintile of the asset ownership distribution. infants than girls, in part because of sex- groups can become internalized into behav- selective abortion and differential care after iors, aspirations, and preferences that also birth. And in many (though not all) parts of perpetuate inequalities. the world, more boys than girls attend Inequalities of opportunity are also school. The hundreds of millions of dis- transmitted across generations. The chil- abled children across the developing world dren of poorer and lower-status parents also face very different opportunities than face inferior chances in education, health, their able-bodied peers. incomes, and status. This starts early. In These inequities are usually associated Ecuador, three-year-old children from all with differences in an individual's"agency"-- socioeconomic groups have similar test the socioeconomically, culturally, and polit- scores for vocabulary recognition and are ically determined ability to shape the world close to a standard international reference around oneself. Such differences create population. But by the time they are five, all biases in the institutions and rules in favor have faltered relative to the international of more powerful and privileged groups. reference group, except for those in the This is seen in realities as diverse as the low richest groups and with the highest levels of chances for mobility of scheduled castes in parental education (figure 2). Such pro- a village in rural India and the frequent nounced differences in vocabulary recogni- episodes of discrimination against the tion between children whose parents had 0 Quichua people in Ecuador. Persistent dif- to 5 years of schooling and those whose ferences in power and status between parents had 12 or more years are likely to 6 WORLD DEVELOPMENT REPORT 2006 Figure 2 Opportunities are determined early ity, undernutrition, and the earnings a child Cognitive development for children ages three to five in Ecuador differs markedly across family can generate by working instead of study- backgrounds ing, many children leave school early. The Wealthiest and poorest quartiles Maternal education average person born between 1975 and Median score Median score 1979 in Sub-Saharan Africa has only 5.4 110 110 years of schooling. In South Asia, the figure rises to 6.3 years; in OECD countries, it is 100 Wealthiest 25% 100 12 or more years 13.4 years. 90 90 With such differences in education and health, compounded by large disparities in 80 80 access to infrastructure and other public 0­5 years Poorest 25th% services, it is not surprising that opportuni- 70 70 ties for the consumption of private goods 60 60 differ vastly between rich and poor coun- 40 50 60 70 40 50 60 70 tries. Mean annual consumption expendi- Age in months Age in months tures range from Purchasing Power Parity Source: Paxson and Schady (2005). (PPP) $279 in Nigeria to PPP $17,232 in Note: Median values of the test of vocabulary recognition (TVIP) score (a measure of vocabulary recognition in Spanish, standardized against an international norm) are plotted against the child's age in months. The medians by exact month Luxembourg. This means that the average of age were smoothed by estimating fan regressions of the median score on age (in months), using a bandwidth of 3. citizen in Luxembourg enjoys monetary resources 62 times higher than the average Nigerian. While the average Nigerian may carry over to their performance once they find it difficult to afford adequately nutri- enter primary school, and will likely persist tious meals every day, the average citizen of thereafter. Intergenerational immobility is Luxembourg need not worry too much also observed in rich countries: new evi- about buying the latest generation cell dence from the United States (where the phone on the market. Because of the much myth of equal opportunity is strong) finds greater restrictions on the movement of high levels of persistence of socioeconomic people between countries than within status across generations: recent estimates countries, these inequalities in outcomes suggest that it would take five generations among countries are likely to be much more for a family that earned half the national closely associated with inequalities in average income to reach the average.5 opportunities than within countries. Immobility is particularly pronounced for Global inequality trends have varied. low-income African Americans. Between 1960 and 1980 there was a pro- nounced decline in the inequality in life Global inequities are massive expectancy across countries, driven by Figure 3 Life expectancy improved and If unequal opportunities are large within major increases in the poorest countries in became more equal--until the onset many countries, they are truly staggering on the world (figure 3). This welcome develop- of the AIDS crisis a global scale. Chapter 3 shows that cross- ment was due to the global spread of health Population-weighted international country differences begin with the opportu- technology and to major public health distribution of life expectancy, 1960­2000 nity for life itself: while 7 of every 1,000 efforts in some of the world's highest mor- 0.05 American babies die in the first year of their tality areas. Since 1990, however, HIV/AIDS lives, 126 of every 1,000 Malian babies do. (predominantly in many African countries) 1990 Babies who survive, not only in Mali but in and a rise in mortality rates in transition 1960 2000 much of Africa and in the poorer countries economies (largely in Eastern Europe and of Asia and Latin America, are at much Central Asia) have set back some of the ear- greater nutritional risk than their counter- lier gains. Because of the AIDS crisis, life parts in rich countries. And if they go to expectancy at birth has fallen dramatically school--more than 400 million adults in in some of the world's poorest countries, developing countries never did--their sharply increasing the differences between 0 27 39 51 63 75 87 schools are substantially worse than those them and richer societies. Years attended by children in Europe, Japan, or Inequality in access to schooling has also Source: Schady (2005). the United States. Given lower school qual- been falling around the world, within most Overview 7 Figure 4 A long-run diverging trend in income Why does equity matter inequality begins to reverse because of growth in China and India for development? Why do these persistent inequalities-- Mean log deviation both within and across countries--matter? 1 The first reason is that the interconnec- 0.83 0.8 tions and resilience of these inequalities 0.69 Global inequality imply that some groups have consistently 0.6 Within-country inequality inferior opportunities--economic, social, 0.50 0.42 and political--than their fellow citizens. 0.4 0.36 Most people feel that such egregious dis- 0.37 0.33 0.33 parities violate a sense of fairness, particu- 0.2 larly when the individuals affected can do 0.05 Between-country inequality little about them (chapter 4). This is con- 0 sistent with the teachings of much politi- 1820 1850 1870 1890 1910 1929 1950 1960 1970 1980 1992 cal philosophy and with the international Source: Authors' manipulation of data from Bourguignon and Mor- risson (2002). system of human rights. The core moral and ethical teachings of the world's lead- ing religions include a concern for equity, countries as well as across them, as average although many have also been sources of schooling levels rise in the vast majority of inequities and historically have been linked countries. This too is a welcome develop- to unequal power structures. There is also ment, although concerns over the quality of experimental evidence suggesting that schooling provide reasons for guarding many--but not all--people behave in against complacency. ways consistent with a concern for fair- While our primary concern is with ness, in addition to caring about how they inequality of opportunities, the large dif- fare individually. ferences in income or consumption across Important as these intrinsic reasons are countries surely affect the life chances faced for caring about inequality of opportunities by children born today in those different and unfair processes, the primary focus of nations. Trends in life expectancy at birth this report is on the instrumental relation- and years of schooling were converging, at ship between equity and development, with least until 1990, but a different picture particular emphasis on two channels: the emerges for income and consumption. effects of unequal opportunities when mar- While the recent trends depend greatly on kets are imperfect, and the consequences of the specific concept chosen (discussed in inequity for the quality of institutions a great detail in chapter 3), global income society develops.6 inequality has steadily increased over the long run until the onset of rapid economic growth in China and India in the 1980s With imperfect markets, inequalities in (figure 4). power and wealth translate into unequal It is possible to decompose total opportunities, leading to wasted productive inequality across individuals in the world potential and to an inefficient allocation of into differences among countries and dif- resources. Markets often work imperfectly ferences within countries. Between-coun- in many countries, whether because of try differences were relatively small early in intrinsic failures--such as those associated the nineteenth century, but they came to with asymmetric information--or because account for a larger part of total inequality of policy-imposed distortions. Microeco- toward the end of the twentieth century. If nomic case studies suggest that an ineffi- China and India are excluded, global cient allocation of resources across produc- inequalities have continued to rise, owing tive alternatives is often associated with to the continuing divergence between differences in wealth or status (chapter 5). most other low-income countries and rich If capital markets worked perfectly, there countries. would be no relation between investment 8 WORLD DEVELOPMENT REPORT 2006 and the distribution of wealth: anyone with behalf of their children and because the a profitable investment opportunity would expected returns to investment are influenced be able to either borrow money to finance by location, contacts, and discrimination-- it, or to sell equity in a firm set up to under- on grounds of gender, caste, religion, or take it. But capital markets in just about race. Discrimination and stereotyping-- every country (developed and developing) mechanisms for the reproduction of inequal- are very far from perfect: credit is rationed ity between groups--have been found to across prospective clients, and interest rates lower the self-esteem, effort, and perform- differ considerably across borrowers, and ance of individuals in the groups discrimi- between lenders and borrowers, in ways nated against. This reduces their potential for that cannot be linked to default risk or individual growth and their ability to con- other economic factors affecting expected tribute to the economy. returns to lenders. For example, interest Striking evidence of the impact of rates decline with loan size in Kerala and stereotyping on performance comes from a Tamil Nadu in India, and across trading recent experiment in India. Children from groups in Kenya and Zimbabwe, in ways not different castes were asked to complete sim- explained by risk differences.7 In Mexico, ple exercises, such as solving a maze, with returns to capital are much higher for the real monetary incentives contingent on per- smallest informal sector firms than for formance. The key result of the experiment larger ones. is that low-caste children perform on par Land markets also have imperfections with high-caste children when their caste is associated with a lack of clear titling, histo- not publicly announced by the experi- ries of concentrated land ownership, and menter but significantly worse when it is imperfect rental markets. In Ghana, lower made public (figure 5). If a similar inhibi- security of tenure among women leads to tion of talent occurs in the real world, this an inefficiently low frequency of land fal- implies a loss of potential output owing to lowing and, hence, to progressive declines social stereotyping. in land productivity. The market for human capital is also Economic and political inequalities are imperfect, because parents make decisions on associated with impaired institutional development. The second channel through Figure 5 Children's performance differs when their caste is made salient which inequity affects long-run processes of development is the shaping of economic Average number of mazes solved, by caste, and political institutions (chapter 6). Insti- in five experimental treatments 8 tutions determine the incentives and con- Piece rate Tournament straints people face and provide the context in which markets function. Different sets of 6 institutions are the outcome of complex historical processes that reflect the interests 4 and structure of political influence of different individuals and groups in a soci- ety. From this perspective, market im- 2 High caste perfections may arise not by accident but Low caste because they distribute income or power in 0 particular ways. In this view, there will be Caste Caste Caste Caste Caste social conflict over the institutions of not announced not announced announced society and incentives for people who con- announced announced and segregated trol power to shape institutions in ways that benefit them. Source: Hoff and Pandey (2004). Note: The figure depicts the number of mazes correctly completed The central argument here is that by low-caste and high-caste children from a set of Indian villages in a number of different experiments. The difference between the unequal power leads to the formation of first two and the last three columns refers to payouts: whether institutions that perpetuate inequalities in children are paid per correct maze completed (piece rate) or only if they complete the most mazes (tournament). power, status, and wealth--and that typi- Overview 9 cally are also bad for the investment, inno- inequality is objectionable on both intrinsic vation, and risk-taking that underpin and instrumental grounds. It contributes to long-term growth. Good economic insti- economic inefficiency, political conflict, and tutions are equitable in a fundamental institutional frailty. What are the implica- way: to prosper, a society must create tions for policy, and do they give rise to an incentives for the vast majority of the agenda that is different from the poverty population to invest and innovate. But reduction agenda already embraced by the such an equitable set of economic institu- World Bank, other multilateral institutions, tions can emerge only when the distribu- and many governments? tion of power is not highly unequal and in We argue that an equity lens enhances situations in which there are constraints the poverty reduction agenda. The poor on the exercise of power by officeholders. generally have less voice, less income, and Basic patterns in cross-country data and less access to services than most other peo- historical narratives support the view that ple. When societies become more equitable countries moving onto institutional paths in ways that lead to greater opportunities for that promoted sustained prosperity did so all, the poor stand to benefit from a "double because the balance of political influence dividend." First, expanded opportunities and power became more equitable. benefit the poor directly, through greater One example comes from a comparison participation in the development process. of the early institutions and of the long- Second, the development process itself may term development paths of European become more successful and resilient as colonies in North and South America. The greater equity leads to better institutions, abundance of unskilled labor prevalent in more effective conflict management, and a the South American colonies--where either better use of all potential resources in soci- native Americans or imported African ety, including those of the poor. Resulting slaves were available in large numbers-- increases in economic growth rates in poor combined with the technology of mining countries will, in turn, contribute to a and large plantation agriculture to provide reduction in global inequities. the economic base for hierarchical and One manifestation of the greater partic- extractive societies, in which land owner- ipation of the poor in economic growth is ship and political power were highly con- the fact that the growth elasticity of centrated. In North America, by contrast, poverty reduction falls with greater income similar attempts to introduce hierarchical inequality. In other words, the impact of structures were foiled by the scarcity of (the same amount of) growth on poverty labor--except where agro-climatic condi- reduction is significantly greater when ini- tions made slavery economically feasible, tial income inequality is lower. On average, such as in the southern region of the United for countries with low levels of income States. Competition for free labor in the inequality, a 1 percentage point growth in northern areas of North America led to the mean incomes leads to about a 4 percent- development of less unequal land owner- age-point reduction in the incidence of $1 ship patterns, a faster expansion of the fran- per day poverty. That power falls to close to chise, and rapid increases in literacy and zero in countries with high income basic education. The resulting economic inequality.8 Policies that lead to greater and political institutions persisted over equity thus lead to lower poverty--directly time, with positive consequences for long- through expanding the opportunities of run economic development. the poor and indirectly through higher lev- els of sustained development. An equity lens adds three new--or at Leveling the economic least often neglected--perspectives to devel- and political playing fields opment policymaking: So a portion of the economic and political inequalities we observe around the world is · First, the best policies for poverty attributable to unequal opportunities. This reduction could involve redistributions 10 WORLD DEVELOPMENT REPORT 2006 of influence, advantage, or subsidies The analysis of development experience away from dominant groups. Highly clearly shows the centrality of overall politi- unequally distributed wealth associated cal conditions--supporting the emphasis with unduly concentrated political on governance and empowerment in recent power can prevent institutions from years. However, it is neither the mandate enforcing broad-based personal and nor the comparative advantage of the property rights, and lead to skewed pro- World Bank to engage in advice on issues of visioning of services and functioning of political design. In turning to policy impli- markets. This is unlikely to change cations, we focus instead on the core areas unless voice and influence, and public of development policy, while recognizing resources, shift away from the domi- that policy design needs to take account of nant group toward those with fewer the broader social and political context, and opportunities.9 that accountability mechanisms influence development effectiveness. · Second, while such equity-enhancing Because economic policies are deter- redistributions (of power, or access to mined within a sociopolitical reality, how government spending and markets) can policies are designed, introduced, or often be efficiency-increasing, possible reformed matters as much as which specific tradeoffs need to be assessed in the policies are proposed. Policy reforms that design of policy. At some point, higher result in losses for a particular group will be tax rates to finance spending on more resisted by that group. If the group is pow- schools for the poorest will create so erful, it will usually subvert the reform. The much disincentive to effort or invest- sustainability of reforms, therefore, may ment (depending on how the taxes are depend on making information about its raised), that one should stop raising distributional consequences publicly avail- them. When making a policy choice able and, perhaps, forming coalitions of along such tradeoffs, the full value of the middle and poorer groups that stand to benefits from equity enhancement gain from them to "empower," directly or should be considered. If greater spend- indirectly, relatively disadvantaged mem- ing on schools for lower-caste children bers of society. means that, over the long term, stereo- How policies are implemented has a typing will decline in society, with atten- technical aspect as well. Just as we empha- dant increases in performance that are size that the full long-term benefits of redis- additional to the specific gains from tributions need to be taken into account greater schooling today, these gains when making policy choices, so must all should not be ignored. their costs. A focus on equity does not · Third, the dichotomy between policies change the facts that asset expropriations-- for growth and policies specifically even in instances of historical grievances-- aimed at equity is false. The distribution may have adverse consequences for subse- of opportunities and the growth process quent investment, that high marginal tax are jointly determined. Policies that affect rates create disincentives to work, or that one will affect the other. This does not inflationary financing of budget deficits mean that each policy needs to take tends to lead to regressive implicit taxation, equity into account individually: for economic disorganization, and reduced example, the best way to deal with investment and growth. In short, a focus on inequitable effects of a particular trade equity must not be an excuse for poor eco- reform is not always through fine-tuning nomic policy. trade policy itself (which might make it The report discusses the role of public more susceptible to capture) but through action in leveling the economic and politi- complementary policies for safety nets, cal playing field under four main headings. labor mobility, and education. The over- Three of the headings concern domestic all package and the fairness of the under- policies: investing in human capacities; lying process are what matter. expanding access to justice, land, and infra- Overview 11 structure; and promoting fairness in mar- Figure 6 Catching up through early interventions kets. The fourth turns to policies for greater Development quotient global equity, in terms of access to markets, 110 resource flows, and governance. Children of normal Stimulation and nutritional supplement Throughout the discussion, the report height 105 weighs a desire to be specific and practical Stimulation against the fact that the best specific policy 100 Nutritional supplement mix is a function of country context. The educational challenges facing Sudan are 95 different from those facing Egypt. The Control group 90 optimal sequencing of reform in the public sectors of Latvia and Bolivia are unlikely to 85 be the same. The capacity for implement- Baseline 6 12 18 24 ing health finance reform in China and Months Lesotho are also different. So the detailed, Source: Grantham-McGregor and others (1991). Note: The development quotient is an index of progress on four behavioral and cognitive indicators of childhood specific policy advice always needs to be development. Number of months refers to the time after entry into the program--generally at an age of nine months. developed at the country--or even subna- tional--level. Everything that is said below therefore retains some level of generality with children who started life at a normal and should be interpreted accordingly, and height (figure 6). This illustrates how deci- cautiously. sive and well-designed public action can substantially reduce the opportunity gaps Human capacities between those least privileged and the soci- Early childhood development. In many etal norm. Investing in the neediest people developing countries, the actions of the early in their childhoods can help level the state in providing services magnify--rather playing field. than attenuate--inequalities at birth. A guiding principle is to shape public action Schooling. The process continues through- so that the acquisition of human capacities out the school system. Actions to equalize is not driven by circumstances of their opportunities in formal education need to birth, although it can reflect people's prefer- ensure that all children acquire at least a ences, tastes, and talents. basic level of skills necessary to participate Because differences in cognitive devel- in society and in today's global economy. opment start to widen from a very early Even in such middle-income countries as age (see figure 2), early childhood develop- Colombia, Morocco, and the Philippines, ment initiatives can be central to more most children completing basic education equal opportunities. Evidence supports the lack an adequate level of achievement, as view that investing in early childhood has measured by internationally comparable large impacts on children's health and test scores (chapters 2 and 7). readiness to learn and can bring important Access to schooling matters--especially economic returns later in life--often in very poor countries--but, in many coun- greater than investments in formal educa- tries, it is only a small part of the problem. tion and training. Greater access needs to be complemented by An experiment in Jamaica focused on supply-side policies (to raise quality) and undersized children (ages 9 to 24 months) demand-side policies (to correct for the pos- and found that they suffered from lower sibility that parents may underinvest in the levels of cognitive development than those education of their children for various rea- of normal height. Nutritional supplements sons). There are no magic bullets for this, but and a program of regular exposure to men- increasing teachers' incentives, enhancing tal stimulation, helped offset this disadvan- the basic quality of the school's physical tage. After 24 months, kids who received infrastructure, and researching and imple- both better nutrition and more stimulation menting teaching methods to increase the had virtually caught up developmentally learning performance of students who do 12 WORLD DEVELOPMENT REPORT 2006 not do well when left to their own devices are ple with a safety net. In addition to ill health, some of the suggestions on the supply side. macroeconomic crises, industrial restructur- On the demand side, there is now a con- ing, weather, and natural disasters can con- siderable body of evidence showing that strain investment and innovation. The poor, scholarships conditional on attendance have with the lowest capacity to manage shocks, significant impacts. Such transfers work in generally are the least well covered by risk- countries from Bangladesh to Brazil, with the management structures, although in most impacts often greater for girls. There are also countries many among the non-poor risk promising approaches to bring in excluded falling into poverty. Broader social protec- groups--as in the Vidin model of reaching tion systems can help prevent today's Roma in Bulgaria--and to bring up those left inequalities--sometimes generated by bad behind through remedial education--as in luck--from becoming entrenched and lead- the Balsakhi program using young women as ing to tomorrow's inequities. Just as safety para-teachers in 20 cities in India. As argued nets can spur households to engage in riskier in World Development Report 2004, devel- activities that can yield higher returns, they oping the accountability of schools and can also help complement reforms that pro- teachers to students, parents, and the broader duce losers. community can help ensure effective service Safety nets typically target three groups: provider behavior. the working poor, people viewed as unable to work or for whom work is undesirable, and Health. Two areas stand out in reducing special vulnerable groups. If safety nets are inequity and tackling economic distortions designed in a manner appropriate to the local in the provision of health services. First, realities on the ground in each country, indi- there are many cases when the benefits spill vidual targeted interventions in these three over beyond the direct beneficiary in a categories can be combined to provide an range of areas of service provision: for effectively universal public insurance system. immunization, for water and sanitation, In such a system, each household that suffers and for information on hygiene and child a negative shock, and falls below some prede- care. Public assurance of provisioning termined threshold of living standards, makes sense in these areas. Demand-side would qualify for some form of state support. subsidies to provide incentives for maternal and child health increase use, offsetting Taxes for equity. Successful interventions to possible information problems as in Mex- level the playing field require adequate ico's Oportunidades program. resources. The main aim of good tax policy is Second, insurance markets for cata- to mobilize sufficient funding, while distort- strophic health problems are beset with fail- ing incentives and compromising growth as ures. (Here "catastrophic" is in relation to the little as possible. Because taxes impose effi- capacity of the household to deal with the ciency costs by altering individual choices direct costs and the loss in earnings.) The tra- between labor and leisure and consumption ditional supply-side model of relying on pub- and savings, most developing countries are lic hospitals works badly, especially for poor likely to be best served by avoiding high mar- and excluded groups.What can work better is ginal taxes on income and relying on a broad public provisioning or regulation that pro- base, especially for taxes on consumption. vides some insurance for all. Examples Public spending should play the primary role include risk pooling in Colombia, health in actively furthering equity. Nevertheless, cards in Indonesia and Vietnam, and Thai- there is some scope for making the overall tax land's "30-baht" universal coverage scheme. system moderately progressive without large As with education, these interventions need efficiency costs. Societies that desire such an to be combined with incentives for providers outcome can consider simple exemptions for to be responsive to all groups. basic foodstuffs, and an expanded role for property taxation, for example. Risk management. Social protection sys- While the capacity of the tax administra- tems shape opportunities by providing peo- tion and the structure of the economy influ- Overview 13 ence the ability to raise revenues, the quality Equity in laws and fairness in their imple- of institutions and the nature of the social mentation involve striking a balance between compact are also critical. When citizens can strengthening the independence of justice rely on services actually being provided, they systems and increasing accountability-- likely are more willing to be taxed. Con- especially to counter the risk that the pow- versely, a corrupt or kleptocratic state engen- erful and wealthy might corrupt, influence, ders little citizen trust in authority and little or ignore the law. Measures to make the incentive to cooperate. As a general rule, a legal system more accessible--mobile more legitimate and representative state may courts, legal aid, and working with custom- be a prerequisite to an adequate tax system, ary institutions--all help reduce the barri- even as the notion of adequacy varies from ers that excluded groups face. Customary country to country. institutions raise complex issues and may incorporate inequities (for example, with Justice, land, and infrastructure respect to gender), but they are too impor- The development of human capacities will tant to be ignored. South Africa is an exam- not broaden opportunities if some people ple of a country that is pursuing a policy face unfair returns on those capacities and that balances recognition of customary unequal protection of their rights, and have practices with the rights and responsibilities unequal access to complementary factors of in state law. production. Toward greater equity in access to land. Building equitable justice systems. Justice Broader access to land does not necessarily systems can do much to level the playing have to come through ownership (chapter 8). field in the political, economic, and socio- Instead, improving the functioning of land cultural domains, but they can also rein- markets and providing greater security of force existing inequalities. The report pays tenure for poorer groups may be a more attention to both codified law and the ways fruitful area for policy--as in rural Thailand in which the law is applied and enforced in and in urban Peru. Redistributive land practice. Legal institutions can uphold the reform can make sense in some circum- political rights of citizens and curb the cap- stances in which land inequalities are extreme ture of the state by the elite. They can equal- and the institutional context allows for ize economic opportunities by protecting designs that effectively redistribute land to property rights for all and ensuring nondis- smaller farms and support this with comple- crimination in the market. They underpin mentary services, without large transitional and reflect the rules of the game in society costs. But this can be difficult, and tradeoffs and thus are central to fair process--and to may be large when property rights have a the broad-based property rights and un- high degree of legitimacy. biased dispute resolution mechanisms so Expropriating land (with compensation) important for investment. is probably the most disruptive redistribution The law can also accelerate shifts in instrument. Divesting state lands and recu- norms, and justice systems can serve as a perating illegal settlements, possibly in progressive force for change in the social exchange for titling a portion of the settle- domain by challenging inequitable prac- ment, may be two cost-effective alternatives. tices. For example, the U.S. Civil Rights Act Market or community-based approaches that of 1964 and Medicare in 1965 enforced the allow community members to obtain subsi- desegregation of hospitals and led to large dized credit for land rentals or purchases reductions in infant mortality for African according to the willing-buyer-willing-seller Americans. Affirmative action programs principle, as in Brazil and South Africa, have also been shown to reduce group- appear promising. A land tax can be a useful based differences in earnings and educa- complement, generating revenues to pur- tion. But they can become politically chase land to redistribute or encouraging entrenched and limited to helping the bet- redistribution by disproportionately taxing ter-off among disadvantaged groups. large or underused plots. 14 WORLD DEVELOPMENT REPORT 2006 Providing infrastructure equitably. Access to Markets and the macroeconomy infrastructure--roads, electricity, water, Markets are central to shaping the potential sanitation, telecoms--is typically highly un- for people to convert their assets into out- equal across groups. For many people in comes. When market transactions are influ- developing countries, lack of access to enced by the wealth or status of participants, affordable infrastructure services means liv- they are both inequitable and inefficient-- ing in isolation from markets and services and can also influence the incentives for and having intermittent or no supply of different groups to expand their assets power or water for productive activities and (chapter 9). daily existence. This often results in a signifi- cant curtailment of economic opportunities. Financial markets. Captured banking sys- While the public sector will in many tems exchange favors: market power is pro- cases remain the main source of funds for tected for a few large banks, which then lend infrastructure investments aimed at broad- favorably to a few selected enterprises, which ening opportunities for those who have the may not be those with the highest expected fewest, the efficiency of the private sector risk-adjusted returns. This may lie behind a can also be harnessed. Although utility pri- cross-country association between greater vatizations have often been attacked for financial depth and lower income inequal- having unequal effects, the evidence indi- ity. Achieving more equal access to finance cates a more complex reality. Privatizations by broadening financial systems thus can in Latin America typically led to expansions help productive firms that were previously in access to services, particularly in electric- beyond the reach of formal finance. ity and telecommunications. In some cases, These relations are only suggestive, how- however, postprivatization increases in ever, so the report draws on case studies prices more than outweighed gains from from middle-income economies, such as quality and coverage, leading to widespread the Republic of Korea, Malaysia, Mexico, popular discontent. and the Russian Federation, and poorer Privatizations are therefore a classic case economies, including Indonesia and Pak- of a policy that may or may not make sense, istan, to provide more concrete evidence. depending on the local context. If the public These studies suggest an apparent paradox. system is highly corrupt or inefficient, and Societies with extensive inequalities in one expects postprivatization regulatory power and wealth, weak institutions, and capacity to be adequate, it can be a useful controlled financial systems typically suffer tool. In other cases, poorly designed privati- from narrow financial sectors that are ori- zations may be captured, transferring pub- ented to the influential and hide weak asset lic assets, at excessively low prices, into pri- quality. Opening the financial system would vate hands. seem to be an obvious solution. Liberaliza- Experience suggests that whether infra- tion, however, has also often been captured structure services are provided by private by the influential or wealthy, in countries operators or public utilities seems less ranging from Mexico (in the early 1990s) to important for equity than the structure of transition economies such as the Czech incentives facing providers and how Republic and Russia. accountable these providers are to the gen- Gradual deepening and broadening thus eral public. We argue that policymakers can needs to be combined with stronger horizon- improve the equitable provision of infra- tal accountability (in regulatory structures), structure services by focusing on expand- greater openness to societal accountability, ing affordable access for poor people and and, where feasible, external commitment poor areas--which often means working devices (such as the entry of Central Euro- with informal providers and targeting pean and Baltic states to the European subsidies--and strengthening the gover- Union). Programs targeted to the poor-- nance of the sector through the greater such as microcredit schemes--can help but accountability of providers and the are no substitute for the overall broadening of stronger voice of beneficiaries. access. Overview 15 Labor markets. Leveling the playing field in Figure 7 Better to be close to economic opportunities labor markets consists of seeking the right Changes in household welfare in Mexico, following trade liberalization in the 1990s (country-specific) balance between flexibility and protection to provide more equal access Changes in to equal employment conditions to as many household welfare workers as possible. Many countries have > 5 percent fairly extensive regulations and provisions for 4­5 percent 2­4 percent formal sector workers, and far fewer for "out- 0­2 percent siders"in the unregulated (and often less safe) informal sector. There is usually a degree of voluntary movement between the sectors, and great diversity within the informal sector itself, ranging from microentrepreneurs and some of the self-employed with incomes above formal sector workers to many with much worse employment conditions. This mix leads to inadequate protection for poorer workers, while regulations for formal workers can reduce the flexibility of employment and often are a poor deal for the workers them- selves, such as when job-related social secu- Source: Nicita (2004). rity systems are inefficient. Two broad labor market approaches are relevant for equity. First, interventions in Product markets. There is substantial het- the labor market should ensure effective erogeneity in the effects of opening a coun- application of the core labor standards try's product markets to trade, at least in the across the whole market, implying no slave short to medium term. This can be due to or indentured work, no dangerous forms of geographic location, as illustrated by the child labor, and no discrimination. Workers varying impact of trade liberalization in should be free to assemble and form associ- Mexico (figure 7). This illustrates the ations, and their unions should be free to importance of interactions between domes- have an active role in bargaining. Second, in tic product markets and patterns of infra- all areas the policy mix needs to be assessed structure provision. There are also often in ways that balance protection (for all strong interactions with skills in the labor workers) with allowances for the restructur- market. In many countries, opening to ing so central to dynamic growth and trade (often coinciding with opening to for- employment creation. eign direct investment) has been associated Worker security is often provided by vari- with rising inequality in earnings in the past ous excessively stringent forms of employ- two decades. This is especially so for mid- ment protection legislation, which, in general dle-income countries, notably in Latin make it costly to hire and, in some cases, America. Opening to trade often boosts the make it even costlier to hire unskilled, young, premium on skills as firms modernize their and female workers--exactly those the laws production processes (skill-biased technical seek to protect. For many countries, less dis- change, in the jargon of economists). This is tortionary and more inclusive policy alterna- bad for equity if the institutional context tives are available, which may make the play- restricts the capacity of workers to shift into ing field more even in labor markets. These new work--or limits future cohorts' access alternatives include unemployment insur- to education. ance schemes (more likely in middle-income countries) and low-wage employment Macroeconomic stability. This report argues schemes (ideally with an employment guar- that there are two-way relationships between antee), which can be applied successfully even inequitable institutions and macroeconomic in poor countries or states. crises, with mostly bad effects for equity and 16 WORLD DEVELOPMENT REPORT 2006 long-run growth.Weak and captured institu- countries. Indeed, most policy advice given to tions are associated with a greater propensity poor countries over the last several decades-- for countries to experience macroeconomic including that by the World Bank--has crises. When crises occur, they can be costly emphasized the advantages of participating for the poor, who have weaker instruments in the global economy. But global markets are to manage shocks. In addition, crisis resolu- far from equitable, and the rules governing tion is often regressive, through a variety of their functioning have a disproportionately mechanisms (most of them not captured in negative effect on developing countries traditional household survey instruments): (chapter 10). These rules are the outcome of declines in the labor share, at least for formal complex negotiating processes in which workers; capital gains for those who get their developing countries have less voice. More- money out; and fiscal workouts that bail out over, even if markets worked equitably, the influential at substantial cost. Such unequal endowments would limit the ability bailouts must be paid for through some of poor countries to benefit from global combination of higher taxes and lower opportunities. Leveling the global economic spending. Because taxes are typically propor- and political playing fields thus requires more tional and spending is often progressive at equitable rules for the functioning of global the margin (notably in Latin America), the markets, more effective participation of poor cost of bailouts is borne disproportionately countries in global rule-setting processes, and by poorer groups. High inflation has also more actions to help build and maintain the been found to be both bad for growth and endowments of poor countries and poor regressive in its impact. people. A concern for equity would lead, in gen- The report documents some of the many eral, to a highly prudent stance on macroeco- inequities in the functioning of global mar- nomic management and financial regulation. kets for labor, goods, ideas, and capital. Populist macroeconomic policy, sooner or Unskilled workers from poor countries, who later, is bad for equity and bad for growth. could earn higher returns in rich countries, Policy design can increase equity through the face great hurdles in migrating. Developing- pursuit of countercyclical fiscal policy, build- country producers face obstacles in selling ing safety nets before a crisis, reducing risky agricultural products, manufactured goods, lending, and supporting only smaller deposi- and services in developed countries. Patent tors in bailouts. But, as in other policy areas, protection restricts access to innovations these responses need to be underpinned by (particularly drugs) for poor countries, while institutional designs that combine greater new research is strongly oriented to the institutional freedom from political influence diseases of richer societies. Rich-country (such as independent central banks and auto- investors often get better deals in debt crises. nomous financial regulatory agencies) with In most cases, more equitable rules would greater information and debate in society. bring benefits to developed- and developing- country citizens. Benefits vary across markets The global arena and countries, with those from greater legal One predetermined circumstance that most migration likely to be greatest (and to accrue powerfully determines a person's opportuni- directly to migrants) and those from trade ties for leading a healthy and productive life is likely to accrue mostly to middle-income his or her country of birth. Global inequities rather than the least developed countries. are massive. Reducing them will depend pri- The report discusses options to reduce marily on domestic policies in poor countries inequities in the functioning of global mar- through their impact on growth and develop- kets, including the following: allowing greater ment. But global action can change external temporary migration into OECD countries, conditions and affect the impact of domestic achieving ambitious trade liberalization policies. In this sense, global and domestic under the Doha Round, allowing poor coun- actions are complementary. tries to use generic drugs, and developing We live in an integrated world in which financial standards more appropriate to people, goods, ideas, and capital flow across developing countries. Overview 17 The international laws that govern global devoting 0.7 percent of gross national income markets are the product of complex negotia- to aid. Larger volumes of aid will only help, tions. In some cases, as for human rights however, if aid is effective in alleviating con- covenants, the processes generating the laws straints and spurring development in the are perceived to be fair. In other cases, recipient countries. Greater effectiveness can processes and outcomes are perceived as be achieved by emphasizing results, moving unfair, even though the formal regulations away from ex ante conditionality, and pro- are equitable. Within the World Trade Orga- gressively shifting design and management nization (WTO), for example, each country from donors to recipients. Aid should not be has a vote and each can block proceedings. undermined by debt, for debt reduction that Even so, WTO processes are at times per- is not financed by additional resources can ceived as unfair because of the underlying actually undercut effective aid programs. power imbalance between strong commercial Innovative mechanisms to expand develop- interests and the public interest, in both ment assistance should be explored, includ- developed and developing countries. These ing global taxes and private contributions. imbalances manifest themselves, for instance, in the number of staff employed in Geneva Equity and development by different WTO members. More effective Bringing equity to the center of development representation of poor countries in global builds on and integrates the major emphases institutions would help improve processes in development thinking of the past 10 to 20 and may lead to more equitable rules. years--on markets, on human development, The impact of reducing imperfections in on governance, and on empowerment. It is global markets varies by country. The larger noteworthy that this year equity is the focus of and fast-growing developing countries stand both this World Development Report and the to benefit significantly from freer global Human Development Report of the United trade, migration, and capital flows, helping Nations Development Programme. The plea them sustain fast growth (while equitable for a more level playing field in both the pol- domestic policies both help underpin long- itics and the economies of developing coun- run growth and the broad internal sharing of tries serves to integrate the World Bank's this growth). Countries left behind in the twin pillars of building an institutional cli- global economy stand to benefit much less mate conducive to investment and empow- from global markets in the short run and will ering the poor. By ensuring that institutions continue to rely on aid. For them, global enforce personal, political, and property action that helps compensate for unequal rights for all, including those currently endowments is truly essential.Action to build excluded, countries will be able to draw on endowments is primarily domestic, through much larger pools of investors and innova- public investments in human development, tors, and be much more effective in provid- infrastructure and governance structures. But ing services to all their citizens. Greater global action can support domestic policies equity can, over the long term, underpin through resource transfers in the form of aid, faster growth. This can be helped by greater which is not offset by debt repayments, and fairness in the global arena, not least through investments in global public goods, particu- the international community's meeting its larly global commons. commitments made at Monterrey. Faster Aid levels need to be bolstered in line with growth and human development in poorer the commitments rich countries made at the countries are essential to reducing global 2002 Monterrey Conference and concrete inequity and to reaching the Millennium plans should be made to reach the target of Development Goals. Introduction 1 Nthabiseng and Pieter--the hypothetical Equity and inequality of South African children who opened the opportunity: the basic concepts c h a p t e r report's overview--are not unusual exam- What is equity? As with any normative con- ples of people who face highly disparate cept,the word"equity"means different things initial opportunities. A girl born to a lower- to different people. It is a difficult concept, caste family of nine in the slums of Dhaka with a history of different interpretations, has vastly different opportunities from a varying by country and academic discipline. boy born to well-educated and affluent Economists link equity to questions of distri- parents in the well-heeled neighborhoods. bution. Lawyers tend to think of principles An AIDS orphan in rural Zimbabwe is meant to correct the strict application of the almost certain to have fewer chances and law, which may lead to an outcome judged to choices in life than a compatriot born to be unfair in specific circumstances. Philoso- healthy and well-educated parents in phers have produced the most headway in the Harare. Those differences are even greater thinking about equity. Indeed, the attributes across borders: an average Swiss, American, that would characterize a just and fair society or Japanese child born at the same instant lie at the foundation of Western political phi- as Nthabiseng will have incomparably losophy, from Plato's Republic and Aristotle's superior life chances. Politics onward. Equity is also central to most Such staggering inequalities in opportu- of the world's great religions, including Bud- nity are intrinsically objectionable, and dhism, Christianity, Hinduism, Islam, and almost every culture, religion, and philo- Judaism, as well as to most other faith tradi- sophical tradition has developed argu- tions. More recently, social choice theory, and ments and beliefs that place great value on the closely related domain of welfare eco- equity for its own sake. In addition, Part II nomics, have been concerned with the aggre- of this report will argue that we now have gation of preferences into some form of considerable evidence that equity is also "social optimum." instrumental to the pursuit of long-term Summarizing such long-standing and prosperity in aggregate terms for society as nuanced characterizations is perilous, but a whole. But before one can describe the common denominator of these many inequity, or assess its impact on growth and different views is that equity relates to fair- development, a clear definition of the term ness, whether locally in families and com- is needed. munities, or globally across nations. We do This introductory chapter presents our not dwell on the different approaches to working definition of equity and briefly dis- equity here, but we do elaborate on them in cusses its main component--equality of chapter 4, which reviews various categories opportunity. It then turns from our central of evidence in support of the intrinsic normative concepts to one of the report's importance of equity. For this report, we key positive concepts: inequality traps. An think of equity as being defined in terms of inequality trap encapsulates the mutually two basic principles: reinforcing nature of various inequalities, which leads to their persistence and to an · Equal opportunity. The outcome of a inferior development trajectory. person's life, in its many dimensions, 18 Introduction 19 should reflect mostly his or her efforts connect to the rest of the world, the quality and talents, not his or her background. of the services available, the way institutions Predetermined circumstances--gender, treat them--relate to one another? And how race, place of birth, family origins--and do these factors vary across groups? Such an the social groups a person is born into approach would require a focus not only on should not help determine whether peo- the dispersion of univariate distributions ple succeed economically, socially, and (such as income inequality or life expectancy) politically.1 but also on the correlations among them · Avoidance of absolute deprivation. An (how do health outcomes vary across socio- aversion to extreme poverty, or indeed a economic groups?). This is the approach Rawlsian2 form of inequality aversion in taken in most of chapter 2, which summa- the space of outcomes, suggests that rizes information on inequalities (with em- societies may decide to intervene to pro- phasis on the plural) in the various building tect the livelihoods of its neediest mem- blocks of opportunity and on their interre- bers (below some absolute threshold of lationships. need) even if the equal opportunity In taking this route, the report recognizes principle has been upheld. The road that predetermined circumstances, or mem- from opportunities to outcomes can be bership in prespecified groups, affect oppor- tortuous. Outcomes may be low because tunities in two ways: of bad luck, or even because of a person's · The circumstances of one's birth affect the own failings. Societies may decide, for endowments one starts with, including all insurance or for compassion, that its kinds of private assets, such as physical members will not be allowed to starve, wealth (including land and financial even if they enjoyed their fair share of assets), family background (the human, the opportunity pie, but things some- social, and cultural capital of one's par- how turned out badly for them. ents), and access to public services and The equal opportunity principle is con- infrastructure (sometimes referred to as ceptually simple: circumstances at birth geographic capital). should not matter for a person's chances in · Group membership and initial circum- life. But to measure inequality of opportu- stances also affect how one is treated by nities is much harder. Chapter 2 briefly dis- the institutions with which one must cusses one approach, which decomposes interact. Two individuals may both live in observed income inequality into one part areas where formal labor markets exist, that can be attributed, in a statistical sense, where courts are agile, and where a police to predetermined circumstances--such as force is present. But if these two (other- race, place of birth, and parental back- wise identical) people, because of their ground--and one part that cannot. The gender, race, religion, sexual orientation, first component captures a lower bound political beliefs, residential address, or value for the opportunity share of income any other morally irrelevant reason, are or earnings inequality. But it is generally differently rewarded for the same work in very difficult to measure things like family the labor market, are discriminated background precisely: years of schooling against by the court of law, or are treated and broad occupational categories are with bias by the police force, then the imperfect proxies for a family's endow- rules are not being applied fairly. There- ments of human, physical, and social capi- fore, these two people do not have the tal. same opportunity sets. Equity also A superior approach would be to capture requires fairness in processes. the inherently multidimensional and group- based nature of inequality of opportunity. Endowments that are less unequal, How do the factors that determine a per- processes that are fair, and protection from son's chances in life--the access to health deprivation are not always mutually con- and educational opportunities, the ability to sistent. At the policy level, there may be 20 WORLD DEVELOPMENT REPORT 2006 tradeoffs among them. Indeed, some policies Figure 1.1 The interaction of political, economic, and or institutions developed to further one of sociocultural inequalities the principles may compromise the other. Political For example, a policy of affirmative inequalities action that seeks to correct past inequities in the access to educational opportunities Institutions for one group--to equalize endowments-- Sociocultural Economic may imply that individuals of greater merit inequalities inequalities (but from another group) are excluded, cre- ating unfair processes. For another exam- ple, the taxes needed to raise government revenues to make transfers to poor individ- uals (desirable to avoid deprivation) expro- priate some fruits of the efforts of hard- The interaction of political, economic, working men and women. This might be and sociocultural inequalities shapes the seen as violating property rights or the institutions and rules in all societies. The rights to appropriate the fruits of one's own way these institutions function affects peo- labor, again creating unfair processes. ple's opportunities and their ability to invest Whenever such tradeoffs exist--which is and prosper. Unequal economic opportuni- most of the time--no textbook policy pre- ties lead to unequal outcomes and reinforce scription can be provided. Each society must unequal political power. Unequal power decide the relative weights it ascribes to each shapes institutions and policies that tend to of the principles of equity and to the effi- foster the persistence of the initial condi- cient expansion of total production (or tions (figure 1.1). other aggregate). This report will not pre- Consider the status of women in patriar- scribe what is equitable for any society. That chal societies. Women are often denied prop- is a prerogative of its members to be under- erty and inheritance rights. They also have taken through decision-making processes their freedom of movement restricted by they regard as fair. social norms that create separate "inside" and "outside" spheres of activity for women and Inequality traps men. These social inequalities have economic If people care about equity, and if political consequences: girls are less likely to be sent to systems aggregate people's views into social school; women are less likely to work outside preferences, why don't the distributions we the home; women generally earn less than observe represent optimal choices? Why do men. This reduces the options for women inequalities of opportunity persist, if they outside marriage and increases their eco- are both unfair and inimical to long-term nomic dependence on men. The inequalities prosperity? And how do these inequalities also have political consequences: women are reproduce themselves? The short answer is less likely to participate in important deci- that political systems do not always assign sions within and outside the home. equal weights to everyone's preferences. These unequal social and economic Policies and institutions do not arise from a structures tend to be readily reproduced. If benign social planner who aims to maxi- a woman has not been educated and has mize the present value of social welfare. grown up to believe that "good, decent" They are the outcomes of political economy women abide by existing social norms, she processes in which different groups seek to is likely to transmit this belief to her daugh- protect their own interests. Some groups ters and to enforce such behavior among have more power than others, and their her daughters-in-law. An inequality trap views prevail. When the interests of domi- may thus prevent generations of women nant groups are aligned with broader col- from getting educated, restrict their partici- lective goals, these decisions are for the pation in the labor market, and reduce their common good. When they are not, the out- ability to make free, informed choices and comes need be neither fair nor efficient. to realize their potential as individuals. This Introduction 21 reinforces gender differences in power that networks that maintain economic rank. Rich tend to persist over time. parents can use their social connections to Similarly, the unequal distribution of ensure that their child gets into a good power between the rich and the poor-- school, or they can call a few good friends to between dominant and subordinate groups-- make sure that their son gets a good job. Con- helps the rich maintain control over re- versely, poor parents are more subject to sources. Consider an agricultural laborer chance. Connections open doors and reduce working for a powerful landlord. Illiteracy constraints. and malnourishment may prevent him from Social networks are closely allied with cul- breaking out of the cycle of poverty. But he is ture. (By "culture" we mean aspects of life also likely to be heavily indebted to his that deal with relationships among individu- employer, which puts him under the land- als within groups, among groups, and lord's control. Even if laws were in place that between ideas and perspectives). Subordinate would allow him to challenge his landlord's groups may face adverse "terms of recogni- dictates, being illiterate, he would find it diffi- tion,"the framework within which they nego- cult to navigate the political and judicial insti- tiate their interactions with other social tutions that might help him assert his rights. groups.6 One obvious expression is explicit In many parts of the world, this distance discrimination that can lead to an explicit between landlords and laborers is com- denial of opportunities and to a rational pounded by entrenched social structures: choice to invest less at the margin. landlords typically belong to a dominant But the process may also be less overt. A group defined by race or caste, tenants and person born into a low social class or a laborers to a subordinate group. Because socially excluded group may adopt the domi- members of these groups often face severe nant group's value system.7 Religious beliefs constraints from social norms against inter- may propel this: women may take on gen- marrying, group-based inequalities are per- dered beliefs about their economic and social petuated across generations. role, and low castes may absorb the upper Poor individuals in geographically isolated castes' view of their "inferior" status. In regions and racial and ethnic minorities also schools, a stigmatized group may face a have less political power and less voice in "stereotype threat," adopting the dominant many countries. This affects their ability to group's view of their ability to perform in propose and implement policies that would cognitive tests or in occupations historically reduce their disadvantage, even if such poli- controlled by dominant groups.8 This can cies might be growth-enhancing for the coun- affect a discriminated group's "capacity to try.3 The correlations between the unequal aspire."9 It also implies that"voice,"the capac- distribution of assets, opportunities, and ity of individuals to influence the decisions political power give rise to a circular flow of that shape their lives, is also unequally dis- mutually reinforcing patterns of inequality. tributed and that"effort"and"ability"are not Such a flow, and its associated feedback loops, necessarily exogenous (predetermined).10 help inequalities persist over long periods-- The existence of these inequality traps-- even if they are inefficient and deemed unfair with mutually reinforcing inequalities in the by a majority of the population.4 economic, political, social, and cultural Economic and political inequalities are domains--has two main implications for this themselves embedded in unequal social and analysis. The first implication is that, because cultural institutions.5 The social networks of market failures and of the ways in which that the poor have access to are substantially institutions evolve, inequality traps can affect different from those the rich can tap into. For not only the distribution but also the aggre- instance, a poor person's social network may gate dynamics of growth and development. be geared primarily toward survival, with This in turn means that, in the long run, limited access to networks that would link equity and efficiency may be complements, him or her to better jobs and opportunities. not substitutes.11 The rich, by contrast, are bequeathed with Capital, land, and labor markets in devel- much more economically productive social oping countries are imperfect. Informational 22 WORLD DEVELOPMENT REPORT 2006 asymmetries and contract enforcement prob- one segment of the labor market, such as lems imply that some people with good management, and if this brings a new pool project ideas (and thus a potentially high of talent into that segment, efficiency is marginal product of capital) end up con- likely to increase even in the short run.14 In strained in their access to capital. This, even other cases, however, expanding the oppor- as other people earn a lower return on their tunity sets for the disadvantaged may (more abundant) capital. In agriculture, require more costly redistribution. To land market failures mean that some farm- finance better-quality schooling for those ers exert too little effort on some plots who have the least educated parents, and (where they are sharecropping), and too who attend the worst schools, it may be much effort on other plots (which they necessary to raise taxes on other people. own).12 Investment in human capital can The basic economic insight that such taxa- also be allocated inefficiently, because of tion distorts incentives remains valid. Such intrahousehold disputes, because credit- policies should be implemented only to the constrained households lack the resources extent that the (present) value of the long- to keep their children healthy and in school, run benefits of greater equity exceed the or because discrimination in the labor mar- efficiency costs of funding them.15 ket reduces the expected returns to school- The point is that some of these long- ing for some groups. What do such diverse term benefits of pursuing greater equity are market failures have in common? They ignored in the conceptual calculus of policy cause differences in initial endowments-- design. The fact that better-schooled chil- such as family wealth, race, or gender--to dren who are poor and from a racial minor- make investment less efficient. ity will be more productive is usually taken There also are political and institutional into account. But the fact that they may reasons why equity and efficiency are long- acquire greater political voice and help term complements. Markets are not the make social institutions more inclusive-- only institutions in society. The function- which, in turn, may increase the stake of ing of states, legal systems, and regulatory that group in society, potentially leading agencies--indeed, of all the institutions to greater trust, less conflict, and more that assign and enforce property rights and investment--may not be. To the extent that mediate conflicts among citizens--is influ- such indirect (but important) benefits of enced by the distribution of political power equity-enhancing policies are ignored, too (or influence, or voice) in society. Unequal few of them are pursued--even assuming a distributions of control over resources and purely benevolent government. of political influence perpetuate institu- By placing equity and fairness as central tions that protect the interests of the most elements of an efficient development strat- powerful, sometimes to the detriment of egy, developing countries will be better able the personal and property rights of others.13 to reach sustainable growth and develop- Those whose rights are not protected ment trajectories. Such equitable growth have little incentive to invest, perpetuating paths are likely to lead to faster reductions poverty and reproducing inequality. Con- in the many dimensions of poverty, the cen- versely, good institutions that protect and tral objective of development everywhere.16 enforce personal and property rights for all The second implication of the existence citizens have led to higher sustained eco- of inequality traps is that no real-life policy nomic growth and long-term prosperity. or institution is entirely exogenous: no Equity can, once again, help societies grow existing organization or application of a and develop. policy idea has been implemented on a This does not mean, of course, that purely technocratic basis. All policies and efficiency-equity tradeoffs have somehow institutions exist because the political sys- been abolished. In some cases, equity tem has brought them into being or allowed enhancements bring immediate--as well as them to survive. The political system long-run--benefits for efficiency. If we reflects the distribution of power and voice reduce discrimination against women in attained at a particular time and place. This Introduction 23 distribution is, in turn, influenced by the A brief preview of the Report distribution of wealth, income, and other Part I summarizes evidence on inequity assets and outcomes in that society. Such within and across countries. Part II asks why "circular causality" for wealth, income, equity matters for development, both intrin- social and cultural capital, and power, sically and instrumentally. Part III turns to mediated through institutions, evolves the policy implications. If unequal opportu- throughout time and history. nities and absolute deprivation are inimical Acknowledging history and social and to long-term prosperity--as well as intrinsi- political institutions is crucial to avoid pol- cally objectionable--there is scope for policy icy mistakes. But a fatalistic view of the and institutional reform aimed at leveling the world is not only wrong, but also counter- economic and political playing fields. productive. To propose policies without An equity lens and the focus on leveling understanding history, or the specific con- the playing field add three basic points. First, text for developing these policies, often redistributions from richer and more power- leads to failure. But this acknowledgment is ful groups to poorer groups that face more not equivalent to the view that no policies limited opportunities are sometimes neces- should be suggested at all. Such a view fails sary and should be pursued. Second, when to recognize how purposeful social and considering policy tradeoffs between equity political action can achieve significant pol- and efficiency, the full long-term benefits of icy and institutional changes--and would equity--including on the development of result in fatalistic inaction. better and more inclusive institutions--need History is not endlessly repetitive and, as to be taken into account. Third, all categories this report documents, many countries have of economic policy--macro and micro-- taken on the challenge of breaking inequality have effects on both efficiency (and growth) traps with some success. Groups have also and equity (and distribution). Because our changed their circumstances or changed ultimate goal is the reduction of poverty social and political institutions. Consider through the equitable pursuit of prosperity, the civil rights movement in the United the policy suggestions in these chapters are States, the democratic overthrow of consistent with good poverty-reduction poli- apartheid in South Africa, the more partici- cies, which the World Bank has been advocat- patory budgeting practices in some Brazil- ing since at least the publication of the World ian cities, and the reforms in access to land, Development Report 1990.17 These sugges- education, and local government in the tions are also in line with the 2000 World Indian state of Kerala. The challenge for Development Report's three pillars of oppor- policy is to ask when and how such changes tunity, empowerment, and security.18 can be supported. Inequity within and across countries I P A R T f o c u s 1 o n Palanpur Inequality traps stifle economic development in a north Indian village Villagers differ markedly from one another in the opportunities they have to improve their welfare and in their abilities to use the assets and endowments available to them. Mirrored in village economic and social institutions--and in the politi- cal processes for seeking change--these deep-seated inequalities have prevented the village from improving human devel- opment and accelerating economic growth. T he village of Palanpur, in the north Caste Gender Indian state of Uttar Pradesh, has Caste in Palanpur defines opportunities Gender inequalities in Palanpur are pro- been the subject of intensive study and determines the activities villagers pur- nounced. In 1993 there were 84 females for by a group of development economists sue, even independent of occupation, edu- every 100 males, strikingly lower than in between the late 1950s and early 1990s.1 cation, and other standard household char- most parts of the world (where the ratio is Researchers visited the village repeatedly acteristics. The three largest castes in usually greater than one). Child mortality and collected detailed quantitative and Palanpur are Thakurs, Muraos, and Jatabs. rates are much higher among girls than qualitative information. While a single vil- At the top is a martial caste known as among boys. As the researchers reported, lage study covering a specific period of time the Thakurs, which accounted for about a "We witnessed several cases of infant girls cannot be used to draw inferences about quarter of the population in 1993. Thakurs who were allowed to wither away and die in development in rural India as a whole, it are disproportionately represented in jobs circumstances that would undoubtedly does provide a distinct window into the such as the army and police that accord well have prompted more energetic action in the kind of processes that can shape growth with their martial past. They are typically case of a male child."2 and equity over time. averse to wage employment in the village, Young girls leave their village to join The study documents modest economic because this would place them in a subordi- their husband's family. Marriage is "the gift progress over time with slow growth in per nate position. Alert to nonfarm employ- of a daughter." In the new household, the capita incomes and some declines in income ment opportunities outside the village, they girl is acutely vulnerable with no income- poverty. But alongside this sluggish growth is are well placed to take advantage of them, earning opportunities, no property, no pos- evidence of stagnation and even deteriora- thanks to stronger information and social sibility of returning home permanently. tion along other dimensions of well-being. networks. Giving birth to a child improves her sta- Different groups of villagers, defined by Just below the Thakurs is a cultivating tus--particularly if it's a boy. But family such predetermined characteristics as caste caste, the Muraos, also accounting for a planning practices are limited, leading to or gender, face radically different opportuni- quarter of the population. Muraos are tra- high fertility rates and short birth-spacing. ties for economic and social mobility. Their ditional cultivators who have continued to Repeated pregnancies take an enormous economic endowments differ markedly, as specialize in agriculture. Very hardworking, toll on women's general health and put do their education, health, occupational they have seen a rapid rise in wealth and their lives at risk at the time of delivery. Old mobility, and capacity to influence and economic status in the village. While they age is strongly associated with widowhood, shape social and political institutions in the may still not enjoy the same social status as in part because of the typically large age dif- village. Disadvantage in one dimension of Thakurs, they have become more prosper- ference between husbands and wives. To opportunity is generally reinforced by disad- ous and now challenge the previously survive, widows depend overwhelmingly vantage in others, combined in a way that unquestioned political and economic dom- on adult sons. perpetuate the stark inequities over genera- inance of the Thakurs. The participation of women in the labor tions. At the bottom are the scheduled castes force in Palanpur is extremely low. Of 313 These deep-seated inequalities of known as Jatabs, accounting for 12 per- women age 15 or older in 1993, only 14 had opportunity shape, and are shaped by, mar- cent of the population. Traditionally anything other than domestic work as their ket imperfections in the village, resulting in "untouchable" leather workers who now primary or secondary occupation. This low suboptimal investments and impeding engage primarily in agricultural wage female participation in the labor force and growth. Inequalities are also mirrored in labor, Jatabs have not seen any of the society, more generally, has extensive conse- village institutions. State and central gov- social mobility of the Muraos. They quences. For example, the survival disad- ernment policies that were introduced in remain a caste apart, with little or no land, vantage of girls compared with boys tends the village were inevitably filtered through a poor education, and little access to non- to narrow only when adult women have highly unequal distribution of power and farm employment outside the village. wider opportunities for gainful employ- influence. Rather than stimulating broad Despite some slight improvement over the ment. Similarly, the virtual exclusion of economic and social progress, public policy years, Jatabs continue to endure many women from most representative institu- has simply reproduced the prevailing pat- forms of discrimination, including that tions in Palanpur has limited the focus and terns of inequality. from government officials. quality of local politics and public action. Focus on Palanpur 27 Schooling often less strenuous and demanding than in the Jatabs, are the principal targets of fraud- Inequalities in education are wide, declining agriculture. Access to nonfarm jobs is far ulent accounting practices that have only slowly. In the late 1950s, just under 20 from equal, however. Workers who wish to resulted in a dizzying accumulation of percent of males age seven or older, and only obtain a regular job generally have to pay debts and dramatically raised the cost of 1 percent of females, were literate. By 1993, bribes and, more important, get a recom- borrowing for such households. Those male literacy had risen to 37 percent and mendation or introduction from a friend or without access to cheap formal credit have female literacy to just below 10 percent. Yet relative. Such rationing by personal con- to fall back on private moneylenders, at education is clearly of great value in Palan- tacts and influence implies that people with high interest rates. pur. Years of schooling strongly increase the low social status tend to be at a disadvan- likelihood that an individual will find tage in the competition for nonfarm jobs, Collective inaction employment in a regular job outside the vil- even for given education levels, skills, and The different bases of social division in lage. Among farmers, too, direct observation endowments. Palanpur have led to multiple solidarities strongly suggests that better-educated farm- The least advantaged segments of the and oppositions. The village society is ers in Palanpur have been crucial in techno- labor force in Palanpur are highly repre- highly fragmented, with few solid rallying logical innovation and diffusion. sented in agricultural wage labor. Casual points for collective action, whether coop- The perceived value of female education wage labor in agriculture can be described erative or adversarial. The limited reach of is quite different from that for boys, because as a "last-resort" occupation, one taken up collective action, in turn, is responsible for girls are expected to spend most of their by those who have no significant alterna- some of the most serious failures of its adult life in domestic work. Although there tive. Agricultural wage rates have risen over development. For example, the village is good evidence of the benefits of educa- time, but slowly, and there are prolonged assembly (panchayat) is constituted every tion in domestic activities, it is not clear periods of seasonal unemployment. few years, but it rarely meets. In 1984 it was that the effects of maternal literacy on child Econometric analysis indicates that-- made obligatory that at least one woman health, for example, are recognized. Even if controlling for a large number of house- participant be selected, but in Palanpur she benefits are correctly perceived, they might hold characteristics (caste, demographic is never consulted and has never attended not be of direct interest to the parents, characteristics, education, land, and so any panchayat meetings. All decisions and because daughters are "transferred" from on)--the probability of engagement in responsibilities are effectively taken by a vil- the village when they marry. Those who agricultural labor is 50 to 60 percent higher lage headman, who has always come from bear the costs of female education thus for households that had engaged in this one of the privileged groups. There also is share little in the benefits. occupation a decade earlier. Occupational ample scope for self-serving patronage and The upper-caste Thakurs have a view inequalities thus result in income inequal- fraud. Modern arrangements (elections, (adopted by many others) that education is ity, and they persist over long periods. reserved seats for low castes, and women on not important or even suitable for the lower the panchayat) have not profoundly altered castes. Blatant forms of discrimination Incomes, assets, and liabilities the elitist and nonparticipatory character of against children from disadvantaged castes Per capita incomes in Palanpur have grown local politics in the village. have disappeared from the schooling sys- at around 2 percent a year between 1957­8 The dominance of privileged groups tem, but subtler forms of discrimination and 1983­4 and income poverty fell from over collective institutions has had far- have remained--for example, the high- around 47 to 34 percent during this period. reaching consequences. Between the late caste teacher considered any form of con- Incomes in the village are distributed about 1950s and early 1990s, no fewer than 18 tact with Jatab children as "repulsive," as unequally as they are in India as a whole, types of government-provided programs which likely affected his or her rapport with and income inequality has remained rela- were introduced to the village: a public them and probably discouraged their atten- tively stable over time. works road-building program, free school- dance. An assessment of economic inequalities ing, free basic health care, old-age pensions, based on wealth provides a different pic- a fair-price shop, a farmer's cooperative, Work ture. Ownership of durables has expanded, and so on. Most of them remained non- Occupational divisions in Palanpur have and the value of land and other productive functional, particularly when there was a widened as the village has shifted from an assets has grown, implying a significant rise redistributive component. Only programs overwhelmingly agricultural economy to in gross wealth. But there has also been a that enjoyed strong backing from the polit- one in which nonagricultural activities have dramatic and uneven expansion of liabili- ically advantaged in the village were come to account for 30 to 40 percent of vil- ties. Inequality in the distribution of net allowed to succeed. The authors of the lage income. In 1957­58 some 13 villagers wealth has widened in Palanpur from a study conclude, "There is little prospect of (of 528) were employed in regular or semi- Gini of around 0.46 in 1962­3 to a conserv- major improvement in the orientation and regular nonfarm jobs. By 1993, this number atively estimated 0.55 in 1990. achievements of government intervention had increased more than four times to 57 Many of the liabilities come from pub- without a significant change in the balance jobs (the total population had only dou- licly provided and subsidized credit sources of political power, both at the state and at bled). that have expanded sharply over time, but the local level."3 Outside jobs are associated with higher that have been associated with pervasive and more stable incomes, and the work is corruption. Disadvantaged groups, such as Source: Drèze, Lanjouw, and Sharma (1998). Inequity within countries: individuals and groups 2 Across the world, individuals and groups country but also to compare across coun- face highly unequal opportunities to better tries, we use group definitions of broad c h a p t e r themselves economically and socially. relevance. Inequalities, as such, might not be of particu- Although economic inequalities are lar concern if outcomes varied for reasons clearly part of the story, this chapter goes that had to do mainly with individual efforts. beyond incomes to emphasize inequalities But, taking our cue from the first chapter, we in key dimensions of opportunity, such as are concerned here with systematic differ- health, education, and the freedom and ences in opportunities for individuals and capacity of people to participate in and groups who differ only in skin color, caste, shape society. There is a special concern gender, or place of residence, predetermined with inequalities that tend to perpetuate characteristics that can be argued to be differences across individuals and groups "morally irrelevant."As illustrated in focus 1, over time, within and across generations. on the Indian village of Palanpur, when such These result in "inequality traps" that are inequalities of opportunity are pronounced, pervasive in many countries. Such inequal- they are often reproduced over time and not ity traps reinforce our concern with equity only affect welfare directly but also act to sti- on intrinsic grounds, but they can also be fle human development and economic particularly detrimental to the development growth. process, because they act to curtail eco- On the basis of what predetermined nomic dynamism. characteristics should groups be defined A key objective here is to show how such that we would not want to see sys- inequalities combine, interact, and are tematic differences in their opportunities? reproduced through interlinked economic, Clearly there is no single answer. Roemer political, and sociocultural processes. Indi- (1998) argues that society has to make this viduals and groups differ markedly in their choice through some kind of ethical and power to influence these processes; indeed, political process. The circumstances could they differ even in their capacity to aspire include social origin variables outside an to such influence. The report emphasizes individual's control, such as sex, race, eth- that such "agency" is a dimension of nicity, caste, parental education and occu- opportunity, alongside education, health, pation, wealth, or place of birth. Cogneau and wealth. And inequalities of agency are (2005) notes that a society's choice of central in explaining how inequalities of circumstances establishes a direct link opportunity are transmitted over time between equality of opportunities and the (box 2.1). intergenerational transmission of out- This chapter presents evidence of a comes. In this chapter, we are largely com- high degree of inequality of opportunity pelled to let data availability dictate the in many developing countries--inequali- group definitions we consider. We can ties manifest in a variety of dimensions, thus present only a partial, and often rudi- such as health, education, and income. It mentary, picture of the full range of then focuses on the specific dimension of inequity that might exist in a country. inequality of power, or agency. Through- Because we wish not only to look within a out the chapter, we emphasize that 28 Inequity within countries: individuals and groups 29 inequalities in different dimensions can interact with, and reinforce, one another B O X 2 . 1 Unequal opportunities persist across generations over time. To highlight these connections, in Brazil we end by focusing on the specific case of gender inequity. As a prelude to the themes in this chapter, cohorts. Of the four, family background was we describe one attempt to quantify the most important. level and persistence of inequalities of This distribution of certain opportuni- Inequalities in health opportunity in Brazil, based on nationally ties and outcomes has persisted across gen- representative household survey data. Brazil erations.When the authors estimated Alongside the intrinsic importance of was selected for a reason.With a Gini coeffi- econometrically the relationship between health as a dimension of welfare, poor cient of per capita incomes just below 0.6 schooling and race, region of origin, health can directly influence an individ- and persistent over time, it is generally per- parental education, and father's occupation, ceived to be one of the world's most only the coefficient on parental education ual's opportunities--his or her earnings unequal countries.* seems to have fallen across cohorts. In other capacity, performance at school, ability to Brazil's main household survey, the words, race, region of origin, and father's care for children, participation in commu- Pesquisa Nacional por Amostra de Domicilios occupation continue to predict an individ- nity activities, and so on. This important (PNAD), included in 1996 a set of ual's education level. And even for educa- supplemental questions on the parents of tion, mechanisms are at work to reproduce instrumental function of health implies respondents.This permitted an analysis of schooling levels across generations, espe- that inequalities in health often translate the intergenerational persistence in cially at the lower end of the distribution. into inequalities in other dimensions of inequalities. Using four circumstance vari- Brazil underscores the need to look at a welfare. And these inequalities are repro- ables (parental schooling, father's occupa- range of outcomes (of which incomes are tion, race, and region of birth), Bourguignon, only one,with education,health and services duced over time. We focus here on chil- Ferreira, and Menendez (2005) investigated also of great concern).It also underscores the dren, while recognizing that differences in how inequalities of opportunity generate need to look at a range of processes--of social status, wealth, and health also mat- inequality in current earnings across differ- which income and economic wealth-based ent cohorts of adult individuals. Applying a mechanisms form only part,and for which ter for adults. conceptual framework closely related to group-based interactions are as central as Demographic and Health Survey (DHS) that in chapter 1, they decomposed earn- household and individual conditions,behav- data indicate that health status varies ings inequality into a lower bound compo- iors,and characteristics. sharply across population groups. To what nent attributable to the inequality of oppor- Source: Bourguignon, Ferreira, and Menendez tunity--to the effect of the four observed extent does it vary across population groups (2005). circumstance variables--and a residual * The perception of particularly high inequality defined by characteristics that are predeter- component, which would account for per- in Brazil may to some extent be a result of the mined and arguably have no moral rele- sonal effort, luck, measurement error, transi- way income is measured there. Alternative approaches to measuring inequality, based on vance? We draw on DHS data from 60 tory income, and other unobservable char- other welfare indicators, indicate that Brazil acteristics.They found that the four countries to examine how the health of may be less of an outlier in Latin America than variables accounted for more than a fifth of previously believed. See box 2.5 and also De children varies across population groups the total earnings inequality within gender Ferranti and others (2004). defined by mother's education, rural or urban residence, and parent's economic sta- tus, proxied by an index of household own- ership of consumer durables. (We look fur- below 50 (estimates from 1996), the rate ther at cross-country differences in health for children whose mothers have not been in chapter 3.) educated is roughly twice as high. Further analysis, not reported here, indicates that Infant mortality. For these countries, infant mortality rates are also sharply differ- infant mortality rates vary markedly-- entiated across population groups defined from a low of around 25 per 1,000 live by rural-urban residence and economic sta- births in Colombia and Jordan, to more tus, proxied by asset ownership. than 125 in Mali, Niger, and Mozambique (figure 2.1). But even where overall infant Stunting. Another dimension of health, mortality rates are high, the figures for extreme stunting (with height-for-age children whose mothers have a secondary below three standard deviations from the education or higher are dramatically reference population), also varies markedly lower. The risk of death among children across countries. Overall rates are as high with well-educated mothers in Mali, for as 30 percent in Pakistan and the Rep- example, is about the same as that for the ublic of Yemen, but negligible in Trinidad average child in Indonesia. And while the and Tobago and very low in Jordan, Arme- overall infant mortality rate in Brazil lies nia, Brazil, and Kazakhstan (figure 2.2). Figure 2.1 Infant mortality varies across countries but also by mother's education within countries Infant mortality rate per 1,000 live births 160 140 No education 120 100 80 60 40 Secondary or higher 20 0 e i bia ana ala en bia ali Peru oros Haiti ublic FasoChad'IvoireMalawEthiopiaRwanda M Africa BrazilTurkey EgyptTunisiababw India BoliviaEritreaSudanNepal eroonTogo bodia Niger bique UgandaYem Benin Guinea d Colom JordanLankaVietnam Sri Paraguay Thailand GhanaoroccoSenegalKenyaNigeriaSalvadorenistan M Com Zam Pakistan Rep Philippines Botsw Nicaragua GuatemIndonesia Zim Cam El Cam adagascar ozam South Bangladesh Turkm M Burkina Côte M African Central Source: Authors' calculations from Demographic Health Survey (DHS) data. Note: The continuous dark line represents the mean infant mortality rate in each country, while the endpoints of the whiskers indicate the infant mortality rates by different levels of mother's education. Figure 2.2 Stunting levels of children born in rural versus urban areas are far from the same Percentage of children severely stunted (z-score < 3) 35 30 25 20 Rural 15 10 5 Urban 0 e i (*) (*) bia ali ala bia en enia ibia M Brazil Egypt Haiti Benin Faso anda bodia Chad alaw JordanArm Lanka Tunisia Turkey Gabon enistan orocco Ghana Bolivia babw NepalNigeriaZam M Yem Ethiopia Colom Sri Nam M Senegal Guinea Ugandad'IvoireauritaniaTanzania Rw Pakistan Tobago RepublicParaguay Nicaragua Zim Côte M Cam Bangladesh Guatem and Burkina Kazakhstan Turkm inican Dom Trinidad Source: Authors' calculations from Demographic Health Survey (DHS) data. Note: The continuous dark line represents the percentage of severely stunted children in each country, while the endpoints of the whiskers indicate the percentages for urban and rural areas. * Indicates stunting level in urban areas are higher than in rural areas. 30 Inequity within countries: individuals and groups 31 Figure 2.3 Access to childhood immunization services depends on parents' economic status Percentage not covered 70 60 50 40 Poorest 30 20 10 0 Wealthiest i a (*) bia (*) ala (*) (*) en ali Egypt anda Peru oros Faso Haiti ublic M Africa KenyaM alaw Brazilbia Benin India Togo bodia NigerChad Bolivia Yem eroon Eritrea Rw VietnamTurkey orocco Ghana biqueGuinea Com Uganda auritani EthiopiaRep Pakistan JordanColom Zam Paraguay GuatemTanzaniaIndonesiaenistan M Philippines Cam M adagascar ozam Cam South Bangladesh Burkina M M Kazakhstan Turkm African Central Source: Authors' calculations from Demographic Health Survey (DHS) data. Note: The continuous dark line represents the percentage of children without access to a basic immunization package in each country, while the endpoints of the whiskers indicate the percentages for the top and the bottom quintile of the asset ownership distribution. * Indicates that the poorest quintile have higher access to childhood immunization services than the wealthiest quintile. The difference between children born in access to health services, proxied here as rural and urban areas can be dramatic, having received at least one of three key particularly at higher overall stunting lev- childhood vaccinations--bacille Calmette- els. In Guatemala, stunting rates for chil- Guérin; diptheria, pertussis, and tetanus; or dren in urban areas are around 10 percent, measles (figure 2.3). This is so even in but in rural areas they are as much as three countries where the overall percentage of times higher. Children in Guatemala clearly children without any coverage is as high as have no choice in deciding whether they 40 percent. Conversely, children whose par- are born in the countryside or the city, but ents are in the bottom quintile are much their opportunities to achieve good health more likely to lack access to such basic are clearly much less assured in rural than health care. In Morocco, where roughly 5 in urban areas. As for infant mortality percent of children have not received even rates, stunting among children is also one of these three vaccinations, the propor- sharply differentiated by mother's educa- tion for children in the poorest quintile is tion and household economic status. well above 15 percent. Access to immunization. Children born in High-impact health services. The World families whose asset ownership places them Bank (2003j), drawing on DHS data from in the top quintile of the distribution of 30 low- and middle-income countries, finds economic status have a high probability of that the poor are considerably less likely 32 WORLD DEVELOPMENT REPORT 2006 than the non-poor to have access to high- the main inequalities that affect health impact health services, such as skilled deliv- may not be in the income space. He cites ery care, antenatal care, and complemen- examples of other key dimensions of tary feeding. Similarly, Wodon (2005) draws inequality: land ownership, women's on household survey data from 15 African agency (health and fertility in India), and countries to indicate that, while virtually all democratic rights (in England in the urban households are within one hour's 1870s and in the U.S. South in the 1960s). travel time to a health center, the propor- In general, an individual's rank in the rele- tion in rural areas is generally only around vant hierarchy has been found to be im- half, and as low as 35­38 percent in Niger portant to health in animal and human and Ethiopia. experiments. Repeated stress associated with insults and the lack of control that Disability. Data from a number of coun- comes from low rank has a well-developed tries suggest that disabled people are biochemical basis.3 much more likely to be poor. Hoogeveen The consequences of poor health are (2003) reports that in Uganda the proba- reflected in education achievements, eco- bility of poverty for urban dwellers living nomic prosperity, and future generations. in a household with a disabled head is 38 Consider the plight of AIDS orphans in percent higher than for those who live in a southern Africa, the stark inequalities of household with an able-bodied head. The opportunity they face, and the possible role Serbian Poverty Reduction Strategy for public action (box 2.2). reports that 70 percent of disabled people DHS data (figures 2.1­2.3) provide are unemployed. In a study drawing on detailed insights into the relationship 10 household surveys in eight countries, between inequalities in health and some self-reported disability was found to be key circumstance variables. But they are more correlated with nonattendance at not particularly well suited to capturing school than other characteristics, includ- the contribution of detailed spatial factors, ing gender or rural residence.1 Sen (2004) such as place of birth, in overall inequality, emphasizes that the disabled face not only because of the limited sample size. In one an "earnings handicap," associated with a attempt to get around this problem, child lower probability of employment and height in Cambodia was estimated at the lower compensation for their work, but commune level based on a statistical pro- also a "conversion handicap." By this he cedure to combine DHS data with popula- means that a physically disabled person tion census data.4 The study documents requires more income than an able- considerable heterogeneity across Cambo- bodied person to achieve the same living dia's more than 1,600 communes in the standard. prevalence of stunting and being under- weight among children under the age of Social inequalities damaging health. Not five (figure 2.4). The analysis provides only are health outcomes correlated with clear evidence that in Cambodia a child's inequalities in other dimensions, but such opportunities for good health have a social inequalities can be argued to be strong spatial dimension to them. Yet detrimental to individual health out- clearly, no child is able to determine in comes.2 In his comprehensive review of which locality he or she is born. the literature, Deaton (2003) argues that, while it is certainly plausible that various Trends inequalities (such as those in power) cause Average health in most countries improved bad health, it is not clear that inequality of in the twentieth century (chapter 3). income is the main culprit. He provides Deaton (2004) documents that improve- evidence suggesting that, after controlling ments in health are likely to have accom- for an individual's income, income in- panied economic growth, but he also equality at the group level does not matter emphasizes the globalization of knowl- independently for individual health. Thus, edge, facilitated by local political, eco- Inequity within countries: individuals and groups 33 B O X 2 . 2 Unequal assets, unequal opportunities: AIDS orphans in Southern Africa It is hard to imagine people with fewer assets, savings,or other valuables.From an education in their most formative years and the paucity of through absolutely no fault of their own,than standpoint,it is vital to keep children in school, opportunities available to them thereafter. AIDS orphans.Left to fend for themselves on the where the acquisition of even basic skills can death of one or both parents from a progressively give them some viable prospect of being able to Avoiding infection debilitating,heavily stigmatized,and costly-to- move out of poverty.Where a child is the head of The most immediate priority,however,is ensuring treat disease,their plight would be of concern a household and perhaps its sole income earner, that AIDS orphans do not themselves become even if they numbered but a few.In southern however,the pressures to drop out of school are infected with the disease,thereby increasing the Africa,however,the United Nations Children's Fund enormous.Numerous studies document signifi- likelihood that they will perpetuate the cycle.AIDS (UNICEF) estimates that there were 12.3 million cantly higher dropout rates of AIDS orphans.In orphans face precisely such a risk,however, AIDS orphans in 2003,a veritable demographic Kenya,one extreme example found that"52 per- because the stigma of HIV/AIDS means that people group in their own right.By 2010,UNICEF projects cent of the children orphaned by AIDS were not often assume that the children of parents who died that there will be 1.5 million AIDS orphans in in school,compared to 2 percent of non- from AIDS must be infected,shunning,shaming,or South Africa;by 2014,1 million in Zambia. orphans"(UNAIDS 2002,135). exploiting them accordingly.Some AIDS orphans An entire generation of Africans is emerging Attending school is also important from a have even been denied access to schools and who will have been raised, if they are lucky, by civic perspective: it socializes children into the health clinics because of the fear their very grandparents or extended family members norms and mores of society,and gives them the presence generates.Children grieving the loss of a (themselves likely to be impoverished, confidence and capacity to participate more parent are also vulnerable to the sexual predations overwhelmed, and suffering from the disease). fully in it.Without such socialization,vulnerable of those putatively claiming to offer them comfort. At worst, they will grow up in child-headed young children are easy targets for those offer- Indeed,the desperation and apparent hopeless- households or in situations in which their basic ing them security and status through member- ness of their circumstances--all the more so if it rights to food, clothing, shelter, and adequate ship in a street gang,criminal network,or militia coincides with a natural disaster such as drought-- care are routinely denied. movement.If AIDS orphans continue to stay can drive AIDS orphans into prostitution. from school at their current rate,comments one The plight of AIDS orphans provides a Wills and schooling senior U.N.official,"you will have a society where graphic illustration of how cycles of disadvantage Beginning to overcome the huge disadvantages kids haven't been to school and therefore can't can perpetuate themselves,and how social isola- that AIDS orphans start life with requires special fulfill even basic jobs...a society where a large tion and exclusion (especially at a young age) can attention on numerous fronts (box 7.11 consid- proportion can have antisocial instincts because preclude the acquisition of assets and undermine ers a variety of policy options).From a legal their lives have been so hard.You [will] have a the capacity to sustain participation in the insti- standpoint,parents who know their death is generation of children who will be more vulnera- tutions that provide the best path out of poverty. imminent and who have young children need to ble to exploitation and to disease because they Sources: Avert.org (2004) http://www.avert.org/ be encouraged (even if they are illiterate) to pre- won't have the same sense of self-worth"(cited aidsorphans.htm. Accessed December 14, 2004. pare enforceable wills that will protect the inher- in Fleshman 2001,1).Such children face the dis- Fleshman (2001). Hargreaves and Glynn (2002), itance rights of their children to ensure that sur- mal prospect of failing to accumulate assets Lewis (2003), UNAIDS (2002), UNICEF (2003), USAID, viving adults do not just forcibly take their land, because of the extreme burdens thrust on them UNAIDS, and UNICEF (2004). nomic, and educational conditions. In the Figure 2.4 Stunting and underweight in Cambodia 1980s and 1990s, however, progress slowed--a result of the worldwide HIV/ AIDS epidemic and rises in cardiovascular mortality in Eastern Europe and former Soviet Union countries. How have inequalities in health evolved within countries? Data from DHS provide some clues. For a subset of countries, mul- tiple rounds of DHS data are available to document changes in infant mortality over time. Of some 36 "spells" of health change Stunting and underweight that could be identified, roughly 25 corre- (number of communes) sponded to improved health outcomes in Stunting low/underweight low (260) the form of lower infant mortality rates. Stunting low/underweight high (189) Stunting high/underweight low (365) Although overall health improved in these Stunting high/underweight high (780) 25 cases, the gaps between urban and rural Incomplete data areas, between groups defined by mother's 0 100 200 education, and between groups defined by kilometers durable asset ownership did not univer- Source: Fujii (2005). sally decline alongside the overall declines 34 WORLD DEVELOPMENT REPORT 2006 1990s, the overall decline in the infant B O X 2 . 3 Health improvements and greater health equity mortality rate to 7.9 in 1994 was accompa- in Peru nied by an increase in the ratio of black to white infant mortality rates from 1.6 in Paxson and Schady (2004),drawing on mul- Infant mortality rates in Peru varied 1950 to 2.2 in 1991. Inequality in health tiple rounds of DHS data,document the markedly with the education level of the does not inevitably fall as overall health declining infant mortality rate in Peru mother in late 1970s and the 1980s (see fig- between the late 1970s and late 1990s.A ure below). During the economic crisis, improves, but such a virtuous process is general downward trend exhibited a sharp increases in mortality were largest among possible (box 2.3). setback during the major economic crisis infants born to women with less education. between 1988 and 1992,but resumed after After the crisis, the gap between infant mor- Inequalities in education the crisis.The downward trend remained tality rates associated with different mater- evident even after adjusting for age of nal education levels declined steadily, sug- Education is of great intrinsic importance mother,recall period,education,and urban gesting an overall decline in inequality in when assessing inequalities of opportunity. status--indicating that the overall trend mortality alongside the decline in overall decline in infant mortality was not attributa- mortality rates. It is also an important determinant of indi- ble only to general improvements in educa- There is some support for the view that viduals' income, health (and that of their tion,an aging population,or urbanization.In changes in the amount and composition of children), and capacity to interact and com- addition,the fact that infant mortality rose public expenditures on social programs municate with others. Inequalities in educa- sharply around 1990,even after these drove these improvements. Real total adjustments,supports the notion that the expenditures increased two and a half times tion thus contribute to inequalities in other decrease in household income and the col- between 1991 and 2000, and such public important dimensions of well-being. lapse of public expenditures on health as a spending did not bypass the poor. Measuring inequality in education is result of the crisis were important. not easy. Census and survey data in most Adjusted infant mortality rates by maternal education countries can generally yield statistics on, Probability of dying in the first year for example, years of schooling. But such 0.15 information does not capture well the quality of education and how that might vary across individuals. Nor is it easy to 0.12 compare years of schooling across coun- tries, because those years might mean something quite different from country to 0.09 country. Years of maternal education 0.06 Test results. Despite the measurement diffi- 10th percentile or 0 year 25th percentile or 2 years culties, there is considerable evidence of 50th percentile or 5 years inequalities of opportunity in education in 0.03 75th percentile or 10 years the developing world. Consider the differ- 90th percentile or 11 years ences in test performance among Ecuado- rian children ages three to six years across 0.00 population groups defined by parental 1978 1982 1986 1990 1994 1998 education, region of residence, and wealth Source: Paxson and Schady (2004). (box 2.4). Test results among very young children capture well the inequality in opportunity in education, but such data are not readily in infant mortality.5 The improvements in available for large numbers of developing health were not necessarily shared across all countries. So we look instead at the percent- groups in the population. age of household heads with no education As Cornia and Menchini (2005) note, by gender and by urban-rural residence. mortality differentials across groups tend to narrow with an improvement of the Male and female household heads. The average only if policies focus explicitly on overall percentage of household heads equity. Without such a focus, improve- without any education varies dramatically ments in the average may not translate to across our sample of 60-odd countries declining group differences. For example, (figure 2.5). In the high-income countries, in the United States between the 1950s and the percentage rates are negligible. But at Inequity within countries: individuals and groups 35 the other extreme, in Burkina Faso and Mali, for example, the overall percentage is B O X 2 . 4 Child test scores in Ecuador: the role of wealth, more than 80 percent. What is similarly parental education, and place of residence striking is that, in most countries, the likeli- hood that the household head is uneducated That education achievements vary Wealthiest and poorest quartiles markedly by population groups--and that is dramatically higher than average when Median score this can have profound implications--is she is a woman. In the Laos People's Demo- brought out forcefully in a recent study by 110 cratic Republic, for example, although the Paxson and Schady (2005).They show that overall percentage of household heads with cognitive development of Ecuadorian chil- 100 Wealthiest 25% dren ages three to six years, as measured by no education is about 20 percent, the rate is a test of vocabulary recognition (TVIP), 90 closer to 70 percent for female household varies significantly depending on the heads. wealth of their household, their place of res- 80 idence, the education of their mother, and that of their father.The extent to which Poorest 25th% Rural and urban household heads. Simi- 70 these circumstance variables are associated lar patterns can be observed for rural and with performance on cognitive tests is typi- 60 urban areas (figure 2.6). In general, house- cally more pronounced for the older 40 50 60 70 children in their sample. hold heads are far more likely to have no Age in months These socioeconomic characteristics education when they are based in rural are significantly associated with cognitive Maternal education areas than in urban areas. Even in coun- development even after controlling for Median score tries where the overall percentage without child health and home environment.The 110 researchers point to the striking evidence education is very high, the rate in urban that, in Ecuador, the youngest children, areas can be dramatically lower. For exam- 100 irrespective of wealth quintile or educa- 12 or more years ple, in Burundi, the percentage of house- tion of their parents, perform broadly as 90 hold heads with no education in urban well as their comparators. But as children in Ecuador get older, their cognitive devel- areas compares with the national average opment, relative to this benchmark, falters 80 0­5 years in Mexico, the Dominican Republic, and significantly. Only children in the top half Brazil. of the wealth distribution and with highly 70 educated parents maintain their perform- ance relative to their comparators. By the 60 Access to teachers. A recent study of primary time they are six years old, most children 40 50 60 70 schools and health clinics in Bangladesh, in the sample are so far behind in their Age in months Ecuador, India, Indonesia, Peru, and Uganda cognitive development that it is uncertain whether and how they could ever catch has identified teacher absenteeism as an up. important, common, problem. The study Source: Paxson and Schady (2005). found that higher income areas generally have lower teacher absentee rates than poorer areas.6 It also found that higher paid teachers, generally more educated and expe- cation and to assess how much overall rienced, appear equally or more likely to be inequality of education can be attributed to absent than contract or less remunerated mean differences between "morally irrele- instructors, perhaps because these instruc- vant" groups. Araujo, Ferreira, and Schady tors sense a lower risk of being fired for their (2004) find that the inequality of adult absence. And although salaries in rural areas education, measured by years of schooling were often higher than in urban areas, for 124 countries, can be pronounced. They teacher attendance in these areas was typi- also find that it is strongly (and inversely) cally lower than in urban areas. In most sur- correlated with mean years of schooling veyed countries, the quality of infrastructure across countries.7 and the frequency of monitoring appeared to The data assembled by these authors also contribute to lower absenteeism. indicate that the inequality of education for specific subgroups of the population can Trends change. While female schooling achieve- Another way to assess inequalities of ments relative to male achievements were opportunity in education is to calculate an dramatically lower among the oldest cohorts, overall index of inequality for years of edu- particularly in Sub-Saharan Africa, South 36 WORLD DEVELOPMENT REPORT 2006 Figure 2.5 Education levels vary across countries, but they also depend on gender of household head Percentage of household heads with no education 100 80 Female 60 40 20 Male 0 n a (*) (*) (*) (*) IsraelChile Peru Lanka Italy Rica HaitiTimor Benin Maliso Finland Germany KingdomCanada Turkey Sweden NorwayStates Brazil PDRPoland JordanKenya Pakistan te Yemend'IvoireGhana Guinea Sri VietnamAlbania EcuadorPanamaBolivia Fa Tajikistan Romania MoldovaParaguay Thailand ColombianezuelaMexico Salvador Republic Lao Tanzania Burundi Morocco Ethiopia KyrgyzstanJamaica Kazakhsta Georgia Costa Indonesia Ve El HondurasNicaraguaGuatemala Cambodia Bangladesh Mauritania United Herzegovin de Madagascar Cameroon Cô United and Burkina Dominican Bosnia Bolivariana ública Rep Source: Authors' calculations from household survey data. Note: The continuous dark line represents the percentage of household heads with no education in each country, while the endpoints of the whiskers indicate the percentages for male and female-headed households. * Indicates that female-headed households have higher average levels of education than male-headed households. Asia, and to a lesser extent the Middle East individual over goods and services that can and North Africa, these disparities are be purchased in the market and that con- noticeably lower for the younger cohorts, tribute directly to well-being. It is clear too, particularly in Sub-Saharan Africa (figure that individuals' economic status can deter- 2.7). Additionally, disparities in years of mine and shape in many ways the opportu- schooling between urban and rural areas nities they face to improve their situations. have been falling in some regions, most Economic well-being can also contribute to strikingly in the Middle East and North improved education outcomes and better Africa and in Eastern Europe and Central health care. In turn, good health and good Asia. But in Sub-Saharan Africa there has education are typically important determi- been little, if any, change. The (urban-rural) nants of economic status. between-group contribution to inequality An ideal measure of economic well- in this region has hovered at around 30 per- being for assessing inequality will capture cent across all the cohorts examined. an individual's long-term economic status. But it is difficult to produce such a compre- Economic inequalities hensive indicator accurately. In practice, it is An individual's consumption, his or her common to work with measures of current income, or his or her wealth have all been income or consumption compiled from used as indicators of the command of an household survey data. While consumption Inequity within countries: individuals and groups 37 Figure 2.6 Education levels vary by country and between rural and urban sectors Percentage of household heads with no education 100 80 60 Rural 40 20 Urban 0 n a so es(*) Chile Peru Italy Rica HaitiTimor Benin MaliFa Finland Brazil PDRPoland JordanKenya Pakistan te Yemend'IvoireGhana Guinea Germany Swedengdom(*)Canada Stat Jamaica RomaniaIsrael(*) Moldova VietnamAlbania EcuadorPanamaBolivia Tajikistan Paraguay Lanka(*) ThailandGeorgia Burundi Ethiopia Indonesia ColombianezuelaMexico Salvador Republic Lao Tanzania Morocco Kin Norway(*)Kyrgyzstan Cambodia Cameroon Kazakhsta Ve Sri Costa Turkey(*) El HondurasNicaraguaGuatemala Herzegovin Mauritania de Madagascar Cô Bangladesh Burkina United and United Dominican Bosnia Bolivariana ública Rep Source: Authors' calculations from household survey data. Note: The continuous dark line represents the percentage of household heads with no education in each country, while the endpoints of the whiskers indicate the percentages for urban and rural households. * Indicates that rural households have higher average levels of education than urban households. and income inequality are expected to cor- Figure 2.7 The share of inequality in years of schooling attributable to differences between relate reasonably well with long-term well- males and females has been declining being, it is unclear exactly how well they Between-group contribution to total inequality actually do. And different measures of eco- (proportion) nomic welfare--based on income, con- 0.4 sumption, or wealth--can yield quite dif- Sub-Saharan Africa ferent assessments of inequality (see also Middle East and 0.3 North Africa box 2.5). For example, Sudjana and Mishra Europe and (2004), drawing on evidence produced by 0.2 Central Asia South Asia Claessens, Djankov, and Lang (2000), argue that wealth inequality in Indonesia is far more concentrated than suggested by com- 0.1 Other parable figures based on consumption (fig- East Asia ure 2.8). In 1996 more than 57 percent of Latin America and Caribbean the stock market capitalization in Indonesia 0.0 was controlled by 10 families. This is in 1935­9 1940­4 1945­9 1950­4 1955­9 1960­4 1965­9 1970­4 1975­9 stark contrast to neighboring countries, Year of birth such as Singapore and Malaysia, but it is Source: Araujo, Ferreira, and Schady (2004). 38 WORLD DEVELOPMENT REPORT 2006 B O X 2 . 5 Beware of intercountry comparisons of inequality! Because countries differ in their data collection · The degree of disaggregation of consump- Inequality: summary measures in a selection systems, cross-country data on economic tion items. of countries: consumption versus income inequality are generally based on a variety of · The methods for imputation of housing and Gini coefficient indicators that are treated interchangeably.The durables consumption. Year Consumption Income lack of a uniform basis for measuring economic Similarly, income inequality can vary depending inequality in different countries has serious on whether income-- Panama 1997 0.468 0.621 implications for comparability. One of the main sources of noncomparabil- Brazil 1996 0.497 0.596 · Is intended to capture pre- or post-tax income, ity of inequality is that some countries use · Includes actual and implicit transfers, and Thailand 2000 0.428 0.523 household income as indicator of well-being · Refers to full income or earnings only. Nicaragua 1998 0.417 0.534 while others use consumption expenditures (Atkinson and Brandolini 2001).These two indi- Additional factors confounding comparabil- Peru 1994 0.446 0.523 cators capture different aspects of economic ity include differences in survey nonresponse Morocco 1998 0.390 0.586 welfare, with the former perhaps seen better as rates across countries (which are likely to affect Vietnam 1998 0.362 0.489 a measure of welfare opportunity and the latter measured inequality--see Korinek, Mistiaen, as a measure of welfare achievement. In most and Ravallion forthcoming). Differences across Nepal 1996 0.366 0.513 countries, measured inequality based on countries in the availability of spatial price Albania 1996 0.252 0.392 income is higher than if it is based on consump- indexes can also affect conclusions.Thomas Bulgaria 1995 0.274 0.392 tion. But this is not inevitable, and the degree to (1987) demonstrates that adjusting for spatial which the two indicators disagree varies from price variation can affect conclusions about the Russian country to country (see table to the right). degree of income or consumption inequality. Federation 1997 0.474 0.478 The problem of comparability is not Across countries there tends to be little unifor- Bangladesh 2000 0.334 0.392 confined to the choice of welfare indicator. An mity in whether, and how, spatial price variation important but underappreciated additional is accommodated. Source: Authors'creation. issue is that, even for a given indicator, its defini- Cross-country datasets on economic tion varies considerably across countries and inequality generally incorporate some attempts even within countries over time. Consumption to improve comparability, but they typically fall inequality based on different definitions of con- far short of achieving strict comparability.With- sumption can vary markedly, and will depend out a concerted effort to harmonize data collec- on a variety of factors, including the following: tion across countries, it is unlikely that such global databases can be relied on to provide · The length of the recall period over which more than a tentative picture of differences in consumption is recorded. inequality across countries. Figure 2.8 Market capitalization only marginally higher than the figure for picture of how economic inequality is dis- controlled by the top 10 families in the Philippines. More generally, Davies and tributed across countries. The highest levels selected countries, 1996 Shorrocks (2005) report estimates pub- of recorded inequality occur in Africa, the Indonesia 57.7% lished by Merrill Lynch and Forbes that second highest in Latin America. But in- Philippines 52.5% some 20 percent of the world's millionaires equality measures for Latin America come Thailand 46.2% come from the developing world. Similarly, largely from income data, while those in Hong Kong, China 32.1% Morck, Stangeland, and Yeung (2000) find other regions, such as South Asia, come Republic of Korea 26.8% a higher ratio of billionaire wealth to gross mainly from consumption data. As box 2.5 Singapore 26.6% domestic product (GDP) in Latin America illustrated, income data tend to produce Malaysia 24.8% and the Caribbean, and East Asia, but not higher measured inequality. Within regions, Taiwan, China 18.4% India and South Africa (see chapters 6 and the data suggest that inequality can vary Japan 2.4% 9 for further discussion). These figures markedly between countries: consumption Source: Claessens, Djankov, and Lang (2000). imply that the distribution of wealth may, inequality in South Africa is extremely high, on average, be more concentrated in devel- while in Mauritius it is lower even than in oping countries than in the developed. OECD countries. When wealth is associated with political How much overall economic inequality influence, such inequalities also translate within countries is attributable to differ- into political capture and can provide a ences across population groups? Unlike window on this added dimension of health and education inequalities, the sys- opportunity. tematic decomposition of income inequal- Bearing in mind the warnings offered in ity by population groups has long been sub- box 2.5, figure 2.9 provides an approximate ject to analysis in the economics literature. Figure 2.9 Africa and Latin America have the world's highest levels of inequality Income and expenditure Gini coefficients Sub-Saharan Africa Europe and Central Asia Azerbaijan Mauritius Based on consumption Based on consumption Hungary Ethiopia Based on income Bosnia and Herzegovina Based on income Niger Czech Republic Tanzania Slovakia Benin Ukraine Mauritania Armenia Guinea Uzbekistan Ghana Serbia Mali Romania Mozambique Bulgaria Senegal Belarus Angola Slovenia Burkina Faso Kyrgyzstan Nigeria Kazakhstan Burundi Poland Cameroon Albania Uganda Croatia Kenya Russia Côte d'Ivoire Tajikistan Madagascar Lithuania Zimbabwe Moldova Gambia, The Latvia Malawi FYR Macedonia Zambia Turkey South Africa Estonia Lesotho Georgia Swaziland Turkmenistan Central African Republic Botswana Namibia High-Income Economies Taiwan, China Japan Finland Latin America and the Caribbean Sweden Peru Belgium Trinidad and Tobago Denmark Nicaragua Norway Jamaica Germany R.B. de Venezuela Austria Uruguay Netherlands St. Lucia Luxembourg Guyana France Costa Rica Switzerland Guatemala Italy Panama Ireland Dominican Republic Australia Mexico Canada El Salvador Spain Honduras United Kingdom Argentina Greece Chile New Zealand Ecuador United States Colombia Portugal Paraguay Hong Kong, China Bolivia Brazil Middle East and North Africa Yemen Egypt East Asia and Pacific Israel Algeria Vietnam Jordan Rep. of Korea Morocco Indonesia Tunisia Mongolia Iran Lao PDR Thailand South Asia Cambodia Philippines Pakistan Singapore Bangladesh China India Papua New Guinea Nepal Malaysia Sri Lanka 0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8 Source: Authors' calculations from household survey data. 39 40 WORLD DEVELOPMENT REPORT 2006 These decomposition exercises seek to measured levels of inequality. In that sense, understand what share of inequality can be some of the difficulties with cross-country attributed to differences between groups comparisons described in box 2.5 are atten- and what to inequality within groups. There uated by subgroup decompositions. are several attractions to studying certain population groups in this way and to com- Between-group shares paring findings across countries. of total inequality Our interest here is to define groups by While the "between-group" share of overall circumstances we might consider "morally inequality is an appealing indicator of the irrelevant," thereby gaining a window on salience of differences across groups in the the importance of inequality of opportu- overall assessment of inequality, there are nity in the economic sphere. Additionally, concerns about its interpretation.8 In par- decomposition results generally are far less ticular, empirical measures of between- sensitive to differences in definitions of group shares are generally found to be underlying welfare indicators than are quite low (see figures 2.10 and 2.11).9 The conventional presentation of between- group inequality is relative to total inequal- Figure 2.10 Between-group inequality decompositions: social group of the household head ity. Elbers and others (2005), however, note Sub-Saharan Africa that total inequality can be viewed as the South Africa between-group inequality that would be Madagascar Benin observed if every household in the popula- Côte d'Ivoire tion constituted a separate group. Clearly, Niger Guinea against such a benchmark, one would East Asia and Pacific rarely observe a high share of between- Vietnam group inequality. Europe and Central Asia Elbers and his colleagues propose an Kyrgyzstan Romania alternative, comparing the actual between- High-Income Economies group inequality with the maximum possi- United States ble inequality that would be obtained by Germany France keeping the number of groups and their sizes Luxembourg at actual levels. For example, an assessment United Kingdom Canada of the contribution of gender differences to Belgium Conventional inequality compares actual between-gender Switzerland Feasible Australia inequality with the hypothetical between- Ireland gender inequality that would be obtained by Norway Sweden sorting the income distribution so that all Austria males appeared at one end of the distribu- Finland Latin America and the Caribbean tion and all females at the other. This ratio Paraguay provides a measure of how far actual Guatemala between-group inequality lies below the Bolivia Panama maximum between-group inequality that is Peru Brazil feasible given the existing configuration of Guyana groups. Nicaragua St. Lucia Economic inequality can be decom- Middle East and North Africa posed in a large sample of countries based Israel on several population breakdowns, two of Jordan which are presented in figures 2.10 and South Asia 2.11: social group and education of house- Nepal Sri Lanka hold head. Such decompositions can follow Bangladesh the conventional decomposition method- 0 0.1 0.2 0.3 0.4 0.5 0.6 ology, complemented by the Elbers and Proportion others (2005) measure of feasible group Source: Authors' calculations from household survey data. decomposition. Inequity within countries: individuals and groups 41 Figure 2.11 Between-group inequality decompositions: education of the household head Sub-Saharan Africa High-Income Economies Cameroon Luxembourg Madagascar France Senegal Taiwan, China Burundi United States Uganda Ireland Côte d'Ivoire United Kingdom Benin Italy Mali Germany Guinea Finland Kenya Netherlands Switzerland Tanzania Sweden Burkina Faso Conventional Norway Conventional Nigeria Feasible Canada Feasible Mauritania Austria Ethiopia Australia Niger Belgium East Asia and the Pacific Latin America and the Caribbean Thailand Guatemala Indonesia Brazil Papua New Guinea Panama East Timor Peru Philippines Nicaragua Lao PDR Argentina Vietnam Chile Europe and Central Asia Paraguay Mexico Lithuania Colombia Romania Ecuador Serbia Jamaica Poland Haiti Turkey Bolivia Hungary Honduras Albania Costa Rica Macedonia Uruguay Estonia Dominican Republic Kyrgyzstan El Salvador Georgia R.B. de Venezuela Moldova Trinidad Tobago Bosnia & Herzegovina St. Lucia Russia Guyana Kazakhstan Suriname Ukraine Middle East and North Africa Armenia Tajikistan Morocco Azerbaijan Jordan Israel 0.00 0.15 0.30 0.45 Yemen Proportion South Asia Bangladesh Sri Lanka Pakistan Nepal 0.00 0.15 0.30 0.45 Proportion Source: Authors' calculations from household survey data. Different population breakdowns con- high. For example, in Paraguay, when tribute to differing extents to overall inequality is decomposed between groups inequality. In general, the conventional cal- by language spoken at home, the conven- culation of the between-group contribu- tional between-group share is approxi- tion points to a fairly low share attributable mately 30 percent (figure 2.10). And to between-group differences. But in some when inequality is decomposed for five countries even the conventional share is broad education groups in Guatemala, 42 WORLD DEVELOPMENT REPORT 2006 the between-group contribution is above ties account for a larger share of total 40 percent (figure 2.11). inequality as the number of localities In most countries, the between-group increases. The analysis confirms that for share is noticeably higher for decomposi- some countries the spatial dimension of tions based on the alternative,"feasible" cal- inequality is of considerable importance. culation. Based on this approach, observed This conclusion carries over even more between-group differences are indeed sub- powerfully at the global level, where the stantial in many countries--for the group between-country contribution to global definitions here. To the extent that these cir- inequality is dramatic (chapter 3). cumstances are judged "morally irrelevant," Other studies and methodologies cor- the findings suggest that in economic life, roborate the finding that spatial differences just as in health and education, a substantial within countries are important. Using farm- portion of observed inequality in many household data for rural China, Jalan and developing countries can be linked to Ravallion (1997) identify "spatial poverty inequalities of opportunity. traps," where poorer areas have lower provi- sions of essential public goods (such as Spatial differences roads) and, as a result, households in the As with inequalities in health, conventional area experience lower productivity on their survey data cannot say much about the con- investments. Various studies find spatial tribution of finely detailed spatial hetero- effects on living standards, even after con- geneity to overall inequality--because of the trolling for nongeographic household char- limited sample size. In an exercise analogous acteristics. Ravallion and Wodon (1999) to that for health in Cambodia (figure 2.4), a demonstrate that place of residence is an variety of studies have applied statistical important determinant of poverty in techniques to combine survey data with Bangladesh. They also note that important population census data to produce tentative spatial differences can be discerned even estimates of inequality at the community within urban areas--households in the dis- and district levels. Elbers and others (2004) trict of Dhaka are markedly better off than document the contribution to overall esti- their counterparts in other urban districts. mated inequality of differences in mean Many studies suggest that spatial differ- consumption for subdistricts in Ecuador, ences in incomes are driven by policy. In Madagascar, and Mozambique. They China, Kanbur and Zhang (2001) find a demonstrate that the between-subdistrict measurable polarization between inland and contribution to total estimated inequality coastal regions where factors unrelated to ranges from a low of 22 percent in Mozam- physical geography--development of heavy bique to more than 40 percent in Ecuador industry in certain provinces, trade open- (table 2.1). Based on a similar approach, ness, and government investment in coastal World Bank (2004e) reports between-com- regions--are associated with widening mune differences in Morocco, accounting interregional inequality. Escobal and Torero for 40 percent of overall estimated con- (2003) compare coastal Peru with the high- sumption inequality. The general impres- lands and find that average per capita expen- sion is that spatial differences across locali- ditures vary markedly and that this variance is associated with fewer and weaker infra- structure services in the highlands. Table 2.1 Decomposition of inequality between and within communities The role of infrastructure is thus central. Level of Number of Within-group inequality Between-group inequality Although it is not disputed that physical geog- decomposition communities (percent) (percent) raphy can also influence poverty directly, the Ecuador 1,579 58.8 41.2 association between geographic variation in Madagascar 1,248 74.6 25.4 poverty and geographic variation in infra- Mozambique 424 78.0 22.0 structure access is typically strong. Accord- Source: Elbers and others (2004). ingly, it is argued that the influence of Note: Our communities in Ecuador are zonas in urban areas and parroquias in rural areas. Communities in Madagas- regional geographic location on inequality car are firiasana (communes) and in Mozambique they are administrative posts. The decompositions are performed using the conventional methodology. will diminish as access to transport and Inequity within countries: individuals and groups 43 communications services improve; being Figure 2.12 Location, education, and social groups can make a difference: regressions of total geographically isolated will matter less inequality on shares of between-group inequality of different household characteristics because infrastructure improvements will Urban-rural Social group of the household head help compensate for distance.10 Overall inequality Overall inequality The relationship between group differences and inequality As is clear from the discussion here, our interest in the contribution of group differ- ences to total inequality extends beyond normative considerations of fairness and Between-group share Between-group share justice. Differences between groups are also thought to explain overall inequality out- Occupation of the household head Education of the household head Overall inequality Overall inequality comes, particularly the reproduction of inequalities over time. The basic idea is that between-group differences in income inequality, for example, will tend also to be mirrored in between-group differences in health and education inequalities--and in the agency of groups in influencing their circumstances (see below). These group dif- ferences will then reinforce one another. Between-group share Between-group share Group differences in education, for exam- Source: Authors' calculations from household survey data. ple, will translate into differences in Note: Regressions include as controls (X) regional area dummies and a welfare measure (Y/C) dummy. The shares of the between component of inequality across gender and age of the household head, and regions within the country incomes and in political voice and partici- were not significant. pation. These inequalities will, in turn, affect health inequalities between groups, which are passed on to education inequali- One interpretation of these findings is ties and so on. "Inequality traps" are the that between-group differences account for, result. A corollary of this idea is that efforts and possibly explain, a non-negligible por- to moderate overall inequality levels might tion of overall inequality. This is consistent require a focus on reducing between-group with the broader theme of this report: that differences. group differences reinforce one another It is difficult to systematically document and in this way contribute to the replica- this instrumental role of group differences. tion of inequality over time. But these sim- Figure 2.12 illustrates one attempt. Overall ple correlations, while suggestive, could inequality is correlated with the between- also be pointing to other processes and on group share for the sample of countries in their own cannot exclude other competing figures 2.10 and 2.11, controlling for region explanations. and whether the underlying welfare indica- tor is income or consumption. Nothing in Inequality and growth, economic the mechanics of the calculation forces structure, and trade overall inequality to be correlated with the Systematic exploration of the impact of share attributable to between-group differ- between-group shares on overall inequality ences. Yet, for this sample of countries, has not, to date, been a major topic of higher overall inequality is associated with a empirical investigation. A longer-standing larger between-group share of overall question in economics has been how inequality, which is attributable to the inequality evolves with economic growth rural-urban breakdown, to differences more generally. Pioneering work by Kuznets across social groups, to differences in edu- in the 1950s launched an enormous cation, and (weakly) to differences in broad amount of empirical work on this question, occupation class of the household head.11 stimulating much debate. There is still no 44 WORLD DEVELOPMENT REPORT 2006 B O X 2 . 6 Revisiting the Kuznets hypothesis for economic growth and inequality The starting point of the literature linking eco- from a long-run series of inequality indicators composition, and period of observation. See, nomic development and income inequality for England, Germany, and the United States, among others, Bourguignon and Morrisson dates to the well-known works of two Nobel and from a single observation in time for three (1989), Fields and Jakubson (1994), Deininger Prize winners,W. Arthur Lewis (1954) and Simon developing countries--India, Ceylon (Sri Lanka and Squire (1998), and Bruno, Ravallion, and Kuznets (1955). Lewis, in his classic 1954 article today), and Puerto Rico.These were the data Squire (1998). Bruno, Ravallion, and Squire "Economic Development with Unlimited Sup- available at that time, and Kuznets was well (1998), while drawing in part on cross-country plies of Labor,"developed a theoretical model in aware of the limitations of the empirical backing data, also analyzed one country--India--for which growth and accumulation in a dual econ- of his argument, in his own words, on"5 percent which relatively long time-series data had omy would start in the modern industrial sector, of empirical information and 95 percent specu- become available, and again found no sign that where capitalists would hire at a given wage lation, some of it possibly tainted by wishful growth increased inequality. and reinvest a share of their profits.The number thinking." Why the Kuznets curve does not hold in prac- of traditional agricultural laborers willing to Kuznets based his speculation primarily on tice probably has to do with the fact that devel- move to this high-productivity, high-wage sec- longitudinal data and called for in-depth case oping countries do not generally satisfy the tor was assumed to be unlimited. In this process studies of the economic growth of nations. But assumptions on migration processes and sectoral of development, and as long as these assump- many subsequent studies simply used development underlying the Kuznets hypothesis. tions would prevail, inequality in the aggregate cross-country data (often of not par- To explain international differences in inequality distribution of income would increase as aver- ticularly high quality) and reduced-form models of incomes,it is important that the link between age incomes rose.There would be a turning to explore and support the hypothesis of an economic inequalities and other factors,such as point after which inequality would fall again as inevitable tradeoff between development and economic dualism,land,education,and regional the surplus labor phase ends and the dualistic equality.The Kuznets curve became one of the differences,be more carefully analyzed. economy becomes a single-sector, fully industri- most quoted stylized facts of the study of No straightforward relationship between alized economy. income distribution for nearly four decades. income and inequality Although Kuznets did not explicitly model To conclude, there is today something of a con- the intersectoral shifts of population as part of Cross-country data can be misleading sensus that no straightforward relation between the development process, he did build on them for dynamic processes income and inequality can be established. As to articulate his basic idea of an inverted-U rela- With the development of much larger data sets, argued by Kanbur (2000) in his exhaustive tionship between economic growth and income such as the Deininger and Squire (1996) interna- review of the Kuznets curve literature in the inequality (the"Kuznets curve"). In his presiden- tional inequality database (following on from Handbook of Income Distribution: "it seems to us tial address at the Annual Meeting of the Ameri- Fields 1989), empirical"tests"of the Kuznets far better to focus directly on policies, or combi- can Economic Association in 1954, he hypothe- curve were widely conducted. But it has become nation of policies, which will generate growth sized that in the process of growth and understood that the use of cross-country data without adverse distributional effects, rather industrialization, inequality would first increase, to analyze what are essentially dynamic than rely on the existence or nonexistence of an because of the shift from agriculture and the processes can be strongly misleading. Moreover, aggregative, reduced form relationship between countryside to industry and the city, and then numerous studies have shown that the per capita income and inequality." decrease as returns across sectors equalized.The evidence in favor of the Kuznets curve is not at data Kuznets used to make this statement came all robust to econometric specifications, sample Source: Authors'creation. consensus on a systematic relationship A large body of literature has also between the long-term growth processes of explored the relationship between trade industrialization and urbanization--and openness and inequality but has not reached overall inequality (box 2.6). a consensus. For example, Dollar and Kraay Cross-country studies have also analyzed (2002) and Dollar and Kraay (2004) find no the relationship between inequality and eco- effect of trade openness on inequality, but nomic structure. Bourguignon and Morri- Lundberg and Squire (2003) do find such son (1990), for example, argue that"develop- an effect. Ravallion (2001) and Milanovic ing countries which are comparatively (2002) report that at low incomes openness endowed with mineral resources and land may be inequality-increasing, but that this (climate) tend to be less egalitarian than oth- effect reverses at higher incomes. ers, although the effect of the agricultural comparative advantage may be offset by the Trends distribution of land." They also find that the The discussion above highlights the many labor productivity difference between agri- mechanisms for hypothesizing how aggre- culture and the rest of the economy is a pow- gate economic growth, and the evolution of erful explanatory factor for differences in different sectors of the economy, can influ- income inequality in a number of developing ence economic inequality. Popular lines of countries in the 1970s and 1980s.12 argument have emphasized Lewis-Kuznets Inequity within countries: individuals and groups 45 type processes, the race between relative sup- round--Hong Kong (China), Republic of ply and demand for skills along with house- Korea, Malaysia, Singapore, and Taiwan hold adjustments to participation, educa- (China). A complete picture of the factors tion, and fertility; the transitions from behind this process is as yet unclear. controlled to market-oriented economic Although it is likely that at least part of the systems; and various forms of power and story is linked to intersectoral transfers, as bargaining-related views of the world. In emphasized by Lewis (box 2.6), Ravallion the end, and perhaps not surprisingly, it is and Chen (2004) indicate that inequality in difficult to identify a single overarching China grew fastest during periods when explanation. Until recently, this did not economic growth and poverty reduction seem to matter much because there was a were slow. They argue that China provides general perception that inequality does not little support for the view that rising vary markedly over short periods.13 In ear- inequality is inevitable with rapid economic lier studies, few countries having data on growth and poverty reduction. inequality over multiple time periods indi- Third, South Asia has generally been cated sharp changes. perceived as a region with relatively low inequality. This probably is due, in part, to For countries and regions. Empirical investi- inequality being measured by consump- gation of how inequality evolves in a country tion. In this region, too, the prevailing view is subject to concerns similar to those for has been that inequality changes little over comparisons of levels (see box 2.5). But there time. But the stylized fact of low and stable is a growing sense that the impression of sta- inequality in South Asia has also been chal- ble, unchanging income inequality may well lenged. In India, the largest country in the be misleading. A few recent examples of region, some uncertainty remains over how changing inequality bear mentioning. First, inequality has evolved, because of well- careful work by Atkinson (2003) has docu- publicized issues concerning data compa- mented the evolution of inequality in OECD rability over time.14 The best available esti- countries during the second half the twenti- mates suggest that inequality in India has eth century. He finds that inequality in the been rising, but with no solid assessment of United States has been rising steadily since by how much.15 the early 1970s (after seeing little change, and In Bangladesh, Nepal, and Sri Lanka, possibly some decline, in the preceding however, recent and reliable data show very decades) and has risen dramatically in the large increases of inequality in the late United Kingdom since 1980. Between 1984 1980s and 1990s. In Bangladesh, income and 1990, the Gini coefficient in the United inequality (as opposed to consumption Kingdom rose by 10 percentage points (but inequality) has been documented to have then did not increase further)--an unprece- risen from a Gini of 0.30 to 0.41 between dented increase over such a short time. Else- 1991 and 2000.16 In Sri Lanka, the increase where in the OECD, inequality changes have in consumption inequality has been very been less marked. But to the extent that the similar, from 0.32 to 0.40 between 1990 and early and middle decades of the twentieth 2002.17 And, in Nepal, the Planning Com- century were associated with declining mission has produced estimates suggesting inequality in these countries, this trend that consumption inequality rose from 0.34 seems to have halted by the century's later to 0.39 between 1995­6 and 2003­4.18 Only decades. in Pakistan is the evolution of inequality Second, inequality in China was not clear, because of difficulties with data markedly higher at the end of the 1990s comparability. than it had been in the early part of the In other regions of the world, the recent 1980s. In general, the recent evidence in picture on inequality trends is more diffi- East Asia suggests that inequality has risen cult to summarize. For Latin America, De faster in the second round of high growth Ferranti and others (2004) indicate that Asian economies--such as China and Viet- inequality increased in most countries, by a nam--than had been observed in the first sizable margin, during the "lost decade" of 46 WORLD DEVELOPMENT REPORT 2006 the 1980s. But during the 1990s, inequality income inequality will thus differ across continued to rise in only about half of the countries. countries in the region, and less rapidly. The authors note that, in Argentina, inequality Across generations. Our assessment will also has risen sharply in the growth period and depend on the degree to which inequalities during the crisis years. In Brazil and Mex- are transmitted across generations. The ico, the 1990s witnessed some small study of intergenerational transmission of declines. In Eastern Europe and Central welfare is not straightforward, because of Asia, changes in inequality during the early the scarcity of datasets containing informa- 1990s, associated with the transition to the tion on various generations of adults in the market economy, have been difficult to doc- same family. Data from long panels are rare, ument systematically because of data prob- and questions about family background of lems, according to World Bank (2000c). individuals are not always asked in surveys Between 1998 and 2003, consumption (the Brazil data described in box 2.1 are a inequality declined in the former Soviet rare exception). Information about educa- Union countries (with the exception of tion or occupation for various generations Georgia and Tajikistan), while there was no can be captured relatively easily in recall clear trend in eastern and southern Euro- questionnaires. But information about pean countries (World Bank, 2005a). In other dimensions, such as the incomes, Africa and the Middle East, it is difficult to earnings, or even health status of earlier point to broad trends, largely because of generations, is not easily remembered by concerns with data comparability over individuals (not least because they often time. change during a lifetime). The scarcity of To what extent does our examination of intergenerational data is particularly strik- levels and trends in income inequality bear ing in developing countries. Even though on the themes of this report? This report is the persistence of inequalities across gener- most concerned about changes in inequali- ations is often thought to be much more ties in incomes, and other specific dimen- acute in developing countries, studies on sions, if these dimensions are associated intergenerational mobility in the develop- with changes in underlying inequalities of ing world remain few and far between. opportunities. Rising income inequality in Even when the data exist, differences in Russia during the 1990s, for example, is of methodologies and data often limit the concern precisely because of its strong asso- scope for comparisons across countries. ciation with rising political influence and The most widespread measure of intergen- state capture. erational mobility in the economics litera- But this is not inevitably the case. A ture is the intergenerational earnings elas- recent study of income distribution dynam- ticity, or the elasticity of sons' earnings ics in six East Asian and Latin American with the earnings of their parents. This countries by Bourguignon, Ferreira, and measure generally comes from a log-linear Lustig (2005) decomposes income distri- regression of sons' earnings (although it bution dynamics into the underlying driv- could also be income or years of schooling) ing forces. They show that complex and on fathers' observed earnings (or its pre- country-specific interactions between dicted value using such other information powerful underlying social and economic as education or occupation). The closer the phenomena imply that distributional expe- elasticity is to zero, the more mobile the riences must be assessed country by country. society is supposed to be. This elasticity has For example, improvements in education been widely used in the U.S. literature, (equalizing opportunities) may be associ- where longitudinal data are relatively ated in one case with falling income inequal- abundant. And for comparability, it has ity--Brazil or Taiwan, China--but in also been calculated in most other coun- another country with rising inequality-- tries' recent studies.19 Indonesia or Mexico. Our assessment of Until recently, estimates of the intergener- the equity implications of changes in ational elasticity of earnings were thought to Inequity within countries: individuals and groups 47 be around 0.4 in the United States, suggest- tinuing research (on international evi- ing a reasonably mobile society in incomes.20 dence of intergenerational mobility) will More recently, however, Mazumder (2005) improve our understanding of why the uses new data and recent econometric tech- intergenerational transmission of eco- niques to correct for transitory fluctuations nomic status is strong in some countries in earnings--he shows that the previous esti- and weak in others."23 mates of intergenerational elasticity were The intergenerational transmission mech- biased downward by about 30 percent. He anisms of inequalities will differ across coun- argues that the true estimate is somewhere tries and within countries across different around 0.6 for the United States. population groups. As described above, Mazumder (2005) points to rather low lev- An intergenerational elasticity of 0.6 com- pared to 0.4 paints a dramatically different els of intergenerational mobility in the picture of mobility in American society. For United States. He also highlights an impor- example, it implies that a family whose earn- tant racial dimension to this limited mobil- ings are half the national average would ity and finds evidence of substantial immo- require five generations instead of three bility at the ends of the distribution. He before it substantially closed the gap. Obvi- shows that of the individuals whose fathers ously a difference of two generations, or were in the bottom decile of the earnings about fifty years, is quite substantial and sug- distribution, 50 percent will be below the gests the need to examine policies that foster thirtieth percentile and 80 percent below greater mobility.21 the sixtieth percentile. He finds the evidence In parallel analyses, estimates of inter- to be consistent with the hypothesis that generational mobility in Canada, Finland, such immobility "might be due to the or Sweden, among others, have tended to inability of families to invest in their chil- report elasticities closer to 0.2 or lower, sug- dren's human capital due to the lack of gesting that these societies are considerably resources." By contrast, more than 50 per- more mobile than the United States. A rela- cent of the individuals whose parents were tively early study of mobility in the United in the top decile will remain above the Kingdom (Atkinson, Maynard, and Trinder eightieth percentile and two-thirds will be 1983) reports an elasticity of 0.43, while a above the median. more recent study by Dearden, Machin, and In another U.S. study, Hertz (2005) con- Reed (1997) estimates an elasticity of 0.57. firms the findings of Mazumder (and oth- These studies indicate that people in the ers) on the size of the intergenerational United Kingdom are about as mobile as elasticity. He then shows evidence that it is those in the United States. Because of the largely driven by the especially low rate of data limitations, only a few exceptional mobility of black families from the bottom studies on intergenerational earnings elas- of the income distribution. While only 17 ticities for less-developed countries have percent of whites born to the bottom decile been carried out. These provide evidence of of family income remain there as adults, the relatively low mobility.22 corresponding figure is 42 percent for In another literature review of cross- blacks. He also finds that "rags-to-riches" country differences in intergenerational transitions from the bottom quartile to the earnings mobility, Solon (2002) asks top were less than half as likely for black as whether there is any link between cross- for white families. He further provides evi- sectional inequality within a generation dence that the black-white mobility gap is and the intergenerational transmission of not "appreciably altered by controlling for inequality. Although there is greater cross- parents' years of schooling." Last, he pro- sectional inequality in the United States vides evidence that the incomes of black and the United Kingdom than in Sweden children are unresponsive to small changes or Finland, Canada also has relatively high in parents' incomes at the bottom of the inequality. The evidence needed to pro- distribution. vide a clear answer to this question is To recap, summary measures suggest that therefore still fragmentary, and only "con- even in such developed countries as the 48 WORLD DEVELOPMENT REPORT 2006 United States and United Kingdom there is poverty the inequalities in capacity to forge rather limited intergenerational mobility the institution or society can be as impor- across generations. Research in these coun- tant as inequalities in health, income, and tries has highlighted important hetero- education.24 geneities in the patterns of reproduction of A recent study of inequalities in gover- different inequalities across populations nance in four slums of Delhi found that groups. For most developing countries, rela- access to formal government by slum tively little is known about intergenerational dwellers is more available to the better off income mobility. But given the acute group- and to those who have good contact net- based inequalities in many developing coun- works.25 Community leaders in these slums tries, there appears to be little basis for facilitate access primarily to their caste expecting much intergenerational mobility. members, and slum dwellers are more likely to delegate custodianship of their interests to better-educated community leaders. The Agency and equity: study concludes that because access to inequalities of power bureaucracy and political representation for The foregoing discussion has raised explic- slum dwellers in Delhi is largely the pre- itly the question of how inequalities are serve of the better off and better connected, determined and reproduced. It has pointed decisions of formal policymakers do not to the potentially important role of group seek to represent slum interests as a whole, differences in this process. This focus on producing interventions that do not target process and the factors that account for the those in most need. The lack of broadly dis- persistence of inequality over time puts the tributed "voice" thus results in patterns of spotlight on how much inequality is rooted resource allocation, and income generation, in deeper institutions in society--institu- that are far from egalitarian. tions of governance, access to land, control The nature of this unequal capacity can of labor, market regulation. Chapter 6 deals be captured through the sociological con- with the emergence and effects of such cept of agency. Agency refers to people's institutions in more detail. Here we turn to capacity to transform or reproduce such different kinds of evidence--and tradi- societal institutions. Some of this capacity is tions of analysis--to discuss the unequal conscious--for example, when interest capacity of people to influence the form groups lobby for a change in land tenure taken by these institutions and the conse- legislation, or when women refuse to accept quences of unequal institutions for contin- laws around marriage that systematically uing inequality in such capacities. For disadvantage them. Some of it is uncon- B O X 2 . 7 Inequitable agencies and institutions in Pakistan A recently completed Human Development the extremely poor paid a bribe, while only 4.3 most immediate representatives of the formal Report for Pakistan provides rich documenta- percent of the non-poor had to do so. In urban justice system), who are involved in only 1 per- tion of the skewed distributional impact of cor- areas, the extremely poor paid on average 8,700 cent of their disputes but nearly 5 percent of the ruption (United Nations Development rupees in bribes, while the non-poor paid only disputes of the non-poor.The poor perceive that Programme 2003).The report notes that corrup- 1,200 rupees. the police will be slow and inefficient in tion raises the costs of getting things done--for Similar patterns emerge for mediating dis- handling their cases, and they frequently experi- setting up a new business, for crossing borders, putes.The extremely poor not only pay a higher ence outright harassment and intimidation. for obtaining a driver's license. In Pakistan these price to seek a resolution than the non-poor, but Even to register a case of kidnapping with the costs fall most heavily on those least able to also they are less likely to receive a satisfactory police requires paying a bribe. In these afford them: the poor. According to the Pakistan outcome (38.5 percent versus 80.8 percent). situations, it is hardly surprising that the poor Human Development Report, 16.7 percent of Indeed, the fee the extremely poor must pay is find it more expedient to take the law into their the extremely poor reported paying a bribe to often higher than their annual household own hands, creating in many urban areas a host run their business enterprise, handing over an income, leaving many to choose to suffer the of new problems related to gang violence and average of 6,800 rupees. Only 6.7 percent of the consequences of a dispute even when they are vigilantism. non-poor paid a bribe, of 9,300 rupees. In rural clearly in the right. In addition, the extremely Source: United Nations Development Programme areas, the contrast is even starker: 20 percent of poor receive less assistance from the police (the (2003). Inequity within countries: individuals and groups 49 B O X 2 . 8 Legacies of discrimination and the reproduction of inequalities and poverty among the Batwa in Uganda The Batwa,who are described in many parts as of power differences among Batwa and other respond in ways that--however rational and self- pygmy peoples,live in Eastern Uganda,eastern affected groups,nor consider Batwa preferences. protecting--often reproduce the extent to which Democratic Republic of Congo,and Rwanda. All communities were viewed as uniform,a prac- they are excluded.The same PPA reports some Batwa have been subject to negative stereotypes tice that the authorities later recognized"did not Batwa children saying that they did not attend since at least 1751,when Edward Tyson take into account Batwa realities and left them school because it was so unfriendly to them. concluded that pygmies were not human but with nothing"(Zaninka (2003),170). When asked what they wanted to do upon com- rather apes or monkeys.They suffer multiple Non-Batwa locals have resisted efforts to pro- pleting school,one child replied that she wished asset depletion and wide ranging forms of dis- vide more appropriate compensation to the to be"a cleaner."Discrimination and prejudice crimination,a situation that public actions have Batwa.A Participatory Poverty Assessment (PPA) diminish the capacity to aspire to and imagine a at times made worse.Though longstanding forest highlighted persistent discrimination,describing different future. dwellers,the British sought to expel them to cre- the Batwa as a"group of people who are Repudiation and discrimination can also ate forest reserves in the 1930s.In 1991,the despised"and who"have no means of production lead the Batwa to self-exclude from the public Uganda National Park authorities increased such as land,credit and training.They are sphere.The PPA notes that no Batwa attended efforts to enforce this exclusion from forest areas. regarded by other ethnic groups in Kisoro as a PPA exercises. Non-Batwa locals explained: Although the World Bank--which was funding people with no rights."This leads to everyday and "Batwa would never come to such meetings, so some of the park authorities'work--required that institutionalized forms of exclusion,with the there is no point in mobilizing them." the government assess the impact on indigenous Batwa suffering discrimination in access to both communities and follow defined compensation public spaces and services.While some Batwa Source: Moncrieffe (2005), citing Participatory procedures,these did not take sufficient account respond to this by organizing themselves,others Poverty Assessment reports. scious--for example, when people engage the very norms that frame their social lives. in land transactions without questioning Being so routinely treated with contempt by them, they reproduce the institutions of government officials, employers, and fellow land tenure and the markets in land. When citizens--and encountering such enormous a disadvantaged group accepts its disadvan- obstacles to advancement--means that tage as "taken for granted," the effect is to excluded groups can, over time, come to allow the continuing existence of the rela- subscribe to norms about themselves and tionships that create such disadvantage. their situation "whose social effect is to fur- The internalization of disadvantage ther diminish their dignity, exacerbate their leads to pernicious forms of agency that inequality, and deepen their lack of access perpetuate inequalities. From inequalities to material goods and services."28 in agency come inequalities in power, voice, In these circumstances, the poor are not and self-confidence--a major part of our only persistently and overtly discriminated story (box 2.7). Inequalities of agency are as against. Their problems are further com- much products of dominant institutions as pounded and consolidated by their apparent sources of those institutional arrangements. complicity in it, their revealed "adaptive Maintaining these arrangements both preference"29 for menial occupations and reflects and produces the distribution of ascription to norms and subservient behav- power among people. As for health, educa- iors that only legitimize and perpetuate their tion, and income, though, this distribution powerlessness. Dire material circumstances, can change--and it has. Indeed, it has often rational expectations about their limited changed in relation to changes in these prospects for upward mobility, and strong other distributions. beliefs about the legitimacy and immutabil- ity of their situation conspire to create a Internalization of disadvantage vicious circle from which it may be very dif- and inequalities of agency ficult for the poor to escape (see box 2.8).30 Recent work on urban slum dwellers in Inequality traps may cause crime and India26 (and elsewhere)27 suggests that a key violence. First, people who perceive their form of powerlessness for the poor involves poverty as permanent may be driven by living with "negative terms of recognition." hostile impulses rather than rational pur- This concept highlights the conditions and suit of their interests. Second, sensitivity constraints under which the poor negotiate to inequality, especially by those feeling 50 WORLD DEVELOPMENT REPORT 2006 trapped at the bottom, may lead to higher- positions of elites, and the effectiveness of risk tactics like crime, when the expected governments to implement approved courses payoffs from socially legitimate activities of action. The capacity of subordinate groups are poor. Third, people may be particularly is influenced by their "economic" capital-- sensitive to group-based inequalities. If, for their education and economic resources-- example, racial heterogeneity and income their"capacity to aspire,"and the closely asso- inequality are correlated and consolidate ciated capacity to organize.33 status distinctions in a society, this could In Indonesia, the Kecamatan Develop- spell potential for violence. Finally, as Mer- ment Project (KDP) illustrates change ton (1938) elegantly states, occurring through action from above and below: it aims to improve the terms of recog- . . . when a system of cultural values empha- sizes, virtually above all else, certain common nition and the political agency of marginal symbols of success for the population at large groups, and to create new institutions for while its social structure rigorously restricts greater agency to lead to material changes in or completely eliminates access to approved patterns of public investment. Consistent modes of acquiring these symbols for a con- with the ongoing process of democratization siderable part of the same population, . . . anti- in Indonesia, the source of change comes social behavior ensues on a considerable from public policy rather than nongovern- scale. (italics reflect original emphasis) mental action, allowing the project to oper- A lack of upward mobility in a society, com- ate on a large scale (see also focus 4 for exam- bined with a high premium on economic ples of change occurring at the local level). affluence, results in anomie--a breakdown of A recent study34 of the efficacy of the standards and values.31 KDP on challenging and changing the terms of recognition of participants sug- Changing between-group inequalities gests that it does provide villagers with a set of agency and institutional power of deliberative routines for more equitably Inequality of agency often leads to institu- managing the conflicts it inevitably trig- tions that reproduce such inequality. But gers.35 These routines introduce marginal these relationships are not immutable. actors to more equitable spaces of engage- There are ample cases in which interven- ment with more organized and influential tions--by civil society, reformist public actors. But building this conflict manage- officials, external actors, religious institu- ment capacity among marginal groups tions, and others--have given more self- depends on more than just forging collabo- confidence and assertiveness to disadvantaged rative routines. It also requires a set of groups, worked against the internalization of rules--defined by the KDP--that limit the disadvantage, and created new channels for unfair exercise of power by dominant excluded groups to exercise voice with greater groups. With the KDP cultivating collabo- effect. These changes improve the terms of ration and tangible points of political recognition for the powerless: they become power for marginalized groups, the results recognized by more powerful groups who include a well-functioning school or med- otherwise would not acknowledge them at all, ical clinic but equally important a style of leading to empowerment of disadvantaged group (re)definition and defense. groups in economic, social, and political Changes in the agency of indigenous realms. peoples in Ecuador since the 1960s provide Empowerment can occur in many ways.32 another example in which mobilization Change typically occurs through the interac- from below came to change national and tion between the opportunities for action local structures. These changes are clear at created by dominant political structures and both local and national levels. In the 1960s the capacity of poorer or middle groups to in the Andean province of Chimborazo, engage. The "political opportunity struc- the indigenous Quichua people suffered ture"--that shapes the possibilities for multiple deprivations. They were subject action--is itself a function of the openness to everyday forms of violence and to dom- of political institution, the coherence and ination and racism in their interactions Inequity within countries: individuals and groups 51 with other ethnic groups and with author- ities.36 Power was concentrated in the tri- B O X 2 . 9 Sex ratios and "missing women" umvirate of landowner, priest, and local Gender inequity causes many societies to Juvenile (newborn to four years old) sex ratios in government authority. Much indigenous display some preference for male children. China, the Republic of Korea, India, and Punjab labor was tied to large rural estates on But the"son preference"is strong enough and Haryana, 1950­2000 which labor relations were sometimes vio- to result in substantial excess female child lent and returns to labor manifestly unfair. mortality in parts of East and South Asia-- China leading to the phenomenon of what Life expectancy was short, alcoholism 1953 Amartya Sen calls"missing women."(Sen, 1964 severe, and children's access to education 1990). In China and India the practice of 1982 1990 and health acutely constrained. female infanticide was noted at least a 1995 At the start of this twenty-first century, century ago, and in the Republic of Korea 2000 and India high juvenile sex ratios (the pro- indigenous people now occupy several Republic of Korea portion of male to female children below 1949 county mayorships and have a majority of the age of 4) have been documented since 1960 councilors in several counties. The provin- the first modern censuses were taken. By 1970 contrast there seems to be little son pref- 1980 cial prefect is also Quichua. Similarly at a 1990 erence in Southeast Asia or in most other national level, former leaders of national 1995 parts of the developing world. 2000 indigenous people's organizations are now The reasons for this seem to stem from India ministers. And the national Confederation rigid patrilineal inheritance systems.While 1951 most societies deny women inheritance 1961 of Indigenous Nationalities of Ecuador has rights, in other parts of the world there is 1971 control of the directorate of bilingual edu- 1981 some flexibility in these rules. In peasant 1991 cation, the indigenous development coun- Europe and Japan, for instance, women 2001 cil, and the office of indigenous health. It could inherit land if their parents had no Punjab and Haryana (India) sons. Despite egalitarian laws, customary also played a big part in negotiating and 1951 practices in China, the Republic of Korea, 1961 administering a World Bank and Interna- and northwest India permit a man, if he 1971 1981 tional Fund for Agricultural Develop- does not have sons, to adopt one from 1991 ment­supported national Program for the other male kin. In the past, it would also 2001 have been possible to take another wife. Development of Indigenous and Afro- The driving motivation is to use whatever 1.00 1.05 1.10 1.15 1.20 1.25 Ecuadorian Peoples. By any calculation, means possible to continue the male fam- power relationships have changed in ily line.Thus, girl children are undervalued. risen. In the Republic of Korea stark declines Ecuador, becoming more equitable, with During pregnancy, sex-selection may have become only apparent in the last lead to aborting female fetuses, reflected in decade--perhaps because of indigenous people participating more sex ratios at birth that are more masculine improvements in labor market opportuni- completely (and more equitably) in their than the biological rate of 105 boys for ties for women. India, as a whole, does not society. every 100 girls. Sex-selection can also hap- have juvenile sex ratios that are far different pen through infanticide, although the data from many other parts of the world. But The inequality trap for women make it difficult to distinguish between northwest India has seen some particularly selective abortion and infanticide.The third, worrying trends, with sex ratios sharply ris- Unequal opportunities in health, education, and most common, mechanism is the neg- ing between 1981 and 2001, much attribut- economic welfare, and political agency can lect and other practices that result in higher able to the higher incidence of sex- mortality rates for girls than boys during selection in abortion. Other parts of India, be readily observed in most developing early childhood.37 especially the south, have more equitable countries. The preceding sections have In China, intense efforts by the govern- labor markets and fewer restrictions on emphasized that these different manifesta- ment resulted in a brief improvement in the women's mobility and inheritance. tions of inequity are not generally inde- sex ratio during 1953­64 (see figure to the right). But since the 1980s it has steadily Source: Das Gupta and others (2003). pendent from one another and that this interdependence can replicate inequalities over time. This interrelationship can be vividly illustrated by examining the nature sions in the home, affecting investments in and implications of the inequality that traps children and household welfare (box 2.9). many women in developing countries. Gender inequity is the archetypical Men and women around the world have "inequality trap." Most societies have norms starkly different access to assets and opportu- that preserve the prevalent social order, delin- nities, reinforced by unequal norms and eating different roles and spheres of influence social structures, perpetuating gender differ- for men and women. The male sphere is typi- ences over centuries. Gender inequity directly cally outside the home in market work and affects the well-being of women and deci- social interactions that enhance the family's 52 WORLD DEVELOPMENT REPORT 2006 status and power.The female sphere is usually In addition to being denied inheritance inside the home--looking after household and property rights, women in many soci- work, rearing children, and contributing to eties face restrictions on their mobility. For the stability of the household. So, women's example, in the state of Uttar Pradesh in activities serve primarily as inputs into the northern India close to 80 percent of household's collective well-being, while men women require their husband's permission are ostensibly at its center--its breadwinners to visit a health center, and 60 percent have and its link to the larger world where eco- to seek permission before stepping outside nomic and social status are determined. their house.43 These mobility restrictions Marriage and kinship systems preserve may be socially imposed, as with gunghat these structures of patriarchy. Most societies among Hindus--or have religious sanc- are "patrilocal," with women moving from tions, as with purdah among Muslims. Such their parents to their husband's home after practices are not just socially enforced, they marriage. Marriage can therefore be thought can be internalized by women who treat of as a framework that serves to exchange them as marks of honorable behavior. These women between households, and marriage norms are transmitted by parents to their decisions are made with a view toward ensur- children, ensuring their continuity over gen- ing that this exchange of women promises the erations; in many societies, they are enforced maximum gain to both households. The by older women in the community.44 man's household is the point of reference-- Restrictions on mobility and rules of kin- while the woman is simply an input into the ship and inheritance help shape social per- processes for households controlled by men ceptions about women's roles. If women are to generate economic and social returns.38 socially and economically directed to focus Inheritance tends to be consistent with their attention and energy on activities in the this pattern. Most societies are not just home, this is not just what men expect of patrilocal--they are also patrilineal, with them--it is also what other women expect of inheritance and property rights primarily them. In much of the developing world, passed on to men. The majority of countries, women's participation in the labor market is outside of Europe and Central Asia and Latin more a function of adversity than active America and Caribbean, restrict inheritance choice--because husbands cannot earn an rights to women.39 Some countries have leg- adequate income or because of an unantici- islation that guarantees equality in inheri- pated shock, such as a child's illness. tance laws. But these laws often are not Bangladeshi women described it this way, enforced, and real authority over decisions "Men work to support their families, women on inheritance rests in the hands of village work because of need."45 Women around the elders and chiefs, who follow customary world participate in a fair amount of market- practices that discriminate against women. based activity for a wage,but they have to con- Most countries that have unequal inheri- tinue to perform most household chores (fig- tance laws also have unequal property rights ure 2.13). They thus face a time squeeze, regimes.40 Indeed, the vast majority of land spending more time at work, both in and out owners are men.41 Many societies compound of the home, than men do. this by denying women the right to divorce. Because social and economic factors This inequality in property rights regimes determine women's life chances more in mar- persists even in countries where agricultural riage than in labor markets,parents invest less production depends heavily on women's in their human capital. Throughout the labor, such as many in Sub-Saharan Africa. developing world, women are much less In Cameroon, women make up more than likely to be enrolled in secondary school or 51 percent of the population and do more university than men.46 So, they typically work than 75 percent of the agricultural work, but in less lucrative occupations. Moreover, labor they are estimated to hold fewer than 10 per- markets may themselves be discriminatory, cent of all land certificates.42 So, if women paying women less than men for the same work on farms, they are usually working on work. For these reasons, even when women farms owned by men. participate in the labor market, they earn less Inequity within countries: individuals and groups 53 than men. Low earnings are a further disin- Figure 2.13 Women work longer hours than do men centive for women to enter the labor market, Women Men perpetuating traditional social roles. Nonmarket Market Market Nonmarket Inequality in the home activity activity activity activity For a long time, economists did not ade- Australia quately recognize that gender inequity has an Austria impact in the home, and models of the Bangladesh Canada household assumed that decisions were taken Colombia by one person--with no room for different Denmark choices across spouses. The consequence of Finland this world view is not just academic. It sug- France gests, for instance, that policy interventions Germany that attempt to alleviate poverty should not Guatemala Indonesia bother with targeting by gender--or suggests Israel that taxes on a household will not affect the Italy allocation of resources within it. Kenya, rural Economists now question this view, Kenya, urban developing models of household decision Nepal, rural making that allow for inequality between Nepal, urban Netherlands spouses. The new models start with the Norway assumption that households are efficient, in Philippines the sense that they make decisions that max- United Kingdom imize the use of the household's resources. United States With this assumption, the models show that R.B. de Venezuela a spouse's share in household resources is 400 0 400 determined by two factors. The first is the Minutes a day fallback option for the spouse in the event of Source: United Nations Development Programme (1995). Note: Data refers to rural Bangladesh in 1990, urban Colombia 1983, rural Guatemala 1977, urban Indonesia 1992, rural divorce--laws of inheritance, property, and Kenya 1988, urban Kenya 1986, rural Nepal 1978, urban Nepal 1978, rural Philippines 1975­77, urban Venezuela 1983, divorce would matter here. Second is the rel- Australia 1992, Austria 1992, Canada 1992, Denmark 1987, Finland 1987-88, France 1985­86, Germany 1991­92, Israel 1991­92, Italy 1988­89, the Netherlands 1987, Norway 1990­91, the United Kingdom 1985, and the United States 1985. ative size of the spouse's contribution to the household's income, which is determined by benefit more than when men are better off. their opportunities in the labor market.47 If The most obvious way to explain bargaining husbands and wives have different prefer- and sharing is to assume that women intrin- ences, an increase in a woman's outside sically care more about children than men do, options or in her labor market opportuni- but this risks being tautological. ties should reflect consumption choices Perhaps the explanation can benefit from more in line with her preferences. understanding that social and economic dif- Econometric work confirms that an in- ferences outside the household can matter crease in a woman's relative worth and an not only for determining bargaining power improvement in her fallback options have but also for socially determined perceptions effects on consumption patterns.48 The of what men and women consider impor- health of Brazilian children improves when tant. If men and women occupy different additional nonlabor income is in the hands of "outside"and"inside"spheres of influence, it women.49 In the United Kingdom, when leg- seems to make sense that improvements in islation ensured that child support payments women's incomes would have a greater were made directly to mothers, expenditures impact on investments in the household. on children's clothing tended to rise.50 In Improvements in the income of men, by Bangladesh and South Africa, women bring- contrast, are more likely to result in socializ- ing more assets into the marriage increase ing activities outside the home and in pur- household expenditures on children's educa- chases that reflect social status. tion.51 The patterns seem to indicate that, Another consequence of this separation when women are better off, children seem to between inside and outside roles is that 54 WORLD DEVELOPMENT REPORT 2006 Table 2.2 Percentage of women who have ever tic violence is that it allows husbands to experienced physical or sexual violence by an institute a regime of terror to control their intimate partner wives' behavior. In India, Bloch and Rao Physical violence Sexual violence (2002) find that husbands systematically use Bangladesh, rural 42 violence as a means of extracting a larger Brazil, urban 27 10 dowry from their wives. This "instrumental" Ethiopia, rural 49 59 use of violence has widespread acceptance Namibia, urban 31 17 among both men and women. Surveys have Peru, rural 62 47 found that large percentages of respondents Samoa 41 20 in developing countries report that men Serbia and have the right to beat their wives when they Montenegro 23 6 answer back or disobey them.52 Tanzania, urban 33 23 Gender inequity is thus the result of an Thailand, rural 34 29 overlapping set of economic, social, cultural, and political inequalities that reinforce each Source: Unpublished data from the WHO Multi-Country Study on Women's Health and Domestic Violence Against Women obtained other.They cause women to have less access to from a presentation by Claudia Garcia-Moreno at the World Bank's property rights, wealth, and education--and Conference on Gender-Based Violence. The final published com- parative report is forthcoming. limit their access to labor markets and to Note: Data refer to different time periods. Brazil, Peru, and Thailand refer to 2000. Reference period for Bangladesh, Ethiopia, Namibia, spheres of activity outside the home. This, in Samoa, Serbia and Montenegro, and Tanzania are unknown. turn, constrains their ability to influence household decisions. Also limiting this influ- inequalities in the home are also manifested ence are asymmetries of information in the in differences in access to information, household and the use of violence to control which can be used to manipulate intra- women's behavior.All of this maintains a clear household bargaining. In an ethnographic demarcation between the roles of women and study of Bangladeshi garment workers, men, readily reproduced across generations. Kabeer (1997) found that men and women There are some signs that changes in tried to control information about their labor markets and interventions by the state incomes from their spouses so that they can break this inequality trap. The develop- could make purchases without consulting ment of the garment industry in Bangladesh them. Women may also hesitate to share has resulted in a sharp and visible increase in information with their husbands, or to col- women's access to a lucrative labor market, laborate efficiently in farming their plots of expanding their ability to influence house- land, to retain control over their property. In hold choices.53 Higher wages for women studying the farms owned by men and those seem to compensate for restrictive practices, owned by women in Ghana, Udry (1996) such as purdah, by reducing limits on found, keeping other things constant, that women's physical mobility, and increasing women-owned farms were less productive their say in household decision making.54 than those owned by men. When wives and Globalization has expanded opportunities husbands are not sharing information, or for women in Mumbai and increased their manipulating the flow of information, they access to schooling.55 A comparative study of clearly are not using their resources opti- the Philippines, Sumatra, and Ghana found mally. In other words, intrahousehold that patterns of land inheritance and invest- behavior is not efficient--contradicting an ments in schooling have became more egali- important assumption in economic models. tarian because of changes in labor market The widespread domestic violence in the opportunities for women.56 And although family is another type of inefficiency. Recent China, Republic of Korea, and India started World Health Organization (WHO) data out with similar discriminatory social struc- show that both physical and sexual violence tures, intervention by the state has improved are widespread in diverse parts of the world gender equity much more in China than in (table 2.2). An important reason for domes- Republic of Korea or India.57 Equity from a global perspective 3 In examining the inequality of opportuni- ties to live a free, healthy, and fulfilled life. ties within countries, the previous chapter As Angus Deaton writes, c h a p t e r emphasized people's "predetermined cir- cumstances," or life chances beyond their We are living with appalling inequalities, in control, as distinct from their "efforts" and which the poor of the world die of AIDS, and, "talents" as individuals. One of these cir- more broadly, where poor people around the world die of diseases that are readily prevent- cumstances is a person's place of birth. In able elsewhere, including in the first-world many countries, access to basic public hospitals and clinics that serve the rich in health services, for example, is significantly poor countries.1 lower in rural areas than in urban areas. That can mean much for surviving the first In 2000 the life expectancy of a child born year of life--the infant mortality rate in Rio in Sierra Leone (37 years) or Botswana (39 de Janeiro was 3.3 percent in 1996, less than years) was less than half that for a child half the 7.4 percent in northeast Brazil. born in the United States (77 years).2 The But, just as being born in a village or a average educational attainment (uncondi- city is one circumstance that should be tional on quality of schooling) of an indi- irrelevant to a person's chances in life, being vidual born in a Sub-Saharan country born in a specific country is another. Why is between 1975 and 1979 is less than 6 years, it objectionable for, say, Turkish women to but more than 12 years in OECD coun- have inferior opportunities and outcomes tries. Inequalities in income are also high compared with Turkish men, but not so among individuals in different parts of the objectionable if the comparison is between world.3 Turkish men and English women? After all, How do we view large average improve- in many dimensions of well-being, major ments in the world, set against this picture differences in opportunities and outcomes of unacceptable inequalities between coun- exist between citizens of different countries, tries? Sen (2001) describes the current state in some cases differences larger than those of the world while making the case for a between various groups within countries. fairer distribution of the fruits of globaliza- This chapter tries to answer two questions. tion: "Even though the world is incompara- First, how much does one's country of birth bly richer than ever before, ours is also a determine one's opportunities in life? Second, world of extraordinary deprivation and does one's country of birth mean less for life staggering inequality." He argues that chances today than in the near or distant whether there have been some gains for all past? To answer these questions, we discuss is not as important as whether the distribu- inequalities in health, education, income, and tion of gains has been fair. Inequalities in power in the global arena. We show that the affluence--and in political, social, and eco- inequalities between countries are staggering nomic power among countries--are cen- despite some improvements over time. tral to the debate on globalization. As long as the sharing of potential gains from glob- Examples and concepts alization is viewed as unfair by many, the There is no doubt that we live in a world inequalities described in this chapter will with massive inequalities in the opportuni- be deemed unacceptable. This, despite the 55 56 WORLD DEVELOPMENT REPORT 2006 fact that absolute poverty has declined in The debate on inequalities in various the last two decades--though by no means dimensions of well-being and their relation uniformly. to globalization rages on as you read this To put global inequalities in well-being in report.4 It is indeed harder to assert whether perspective, it helps to examine two coun- inequalities increased or decreased over tries at opposite ends of the spectrum-- time. Various questions have to be answered Mali, one of the world's poorest countries, first: inequality of what, over which time and the United States, one of the richest. A period, using which concept of inequality? baby born in Mali in 2001 had an approxi- While there is some evidence of convergence mately 13 percent chance of dying before in opportunities in health and education reaching age one, with this chance declining and some divergence in incomes (or at least only slightly (to 9 percent) even if the baby lack of convergence), these results cannot be were born to a family in the top quintile of stated without many qualifications and the asset distribution. By contrast, a baby caveats. Box 3.1 introduces some underlying born in the United States the same year had concepts that need to be clarified. a less than 1 percent chance of dying in its first year. The picture for under-five mortal- Global inequalities in health ity is even more egregious: 24 percent of The unweighted and weighted international children will not reach age five in Mali, distributions of life expectancy at birth compared with less than 1 percent of Amer- (ignoring the distribution of life expectancy ican children. Even a child born into the at birth within countries) both show a clear richest quintile in Mali is more than 16 "twin-peakedness" in 1960.7 Data show that times likely to die before age five than an 50 countries had life expectancies between average American child. 35 and 45 years, 41 countries had life The picture does not improve for educa- expectancies between 65 and 75 years, and tion. The average American born between there was relatively little mass in the middle 1975 and 1979 has completed more than 14 of the distribution. years of schooling (roughly the same for By 1980 the left-hand mode of the distri- men and women, and in urban and rural bution had decreased considerably in size. areas), while the average school attainment The distributions began to look more for the same cohort in Mali is less than two right-skewed, unimodal, especially in the years, with women's attainment less than weighted international distribution: 73 half that for men, and virtually zero in rural countries had a life expectancy between 65 areas. If one considers the quality of the and 75, compared with 31 countries between education received, the inequalities in learn- 55 and 65, and 35 countries between 45 and ing achievement are possibly much larger. 55. But by 2000 the two modes become evi- It is not surprising, then, that many citi- dent once again, especially in the unweighted zens of Mali, having survived immense distribution, although there is more mass in hardships as children and without much the right mode of the distribution. education, can barely eke out a living as In 1980, the average life expectancy in adults, on average living on less than $2 a four regions--Middle East and North Africa, day ($54 a month) in 1994. By comparison, East Asia (excluding China and Japan), the average American earned $1,185 a South Asia, and Sub-Saharan Africa--was month, more than 20 times that for the below the world average.8 Between 1980 and average Malian. 2000, rapid increases in life expectancy in the While there is probably some consensus first three of these regions were globally that inequalities in health, education, inequality-reducing, while the decline of life income, and voice are large globally, there is expectancy in Sub-Saharan Africa in the much less agreement on whether things 1990s boosted inequality by stretching the have been getting better or worse. Is one's bottom tail of the distribution. By 2000, country of origin more or less pertinent only South Asia and Sub-Saharan Africa today to the life chances that she faces at were below the world average, with the dif- birth than it was 20, 50, or 200 years ago? ference in life expectancy at birth between Equity from a global perspective 57 B O X 3 . 1 Three competing concepts of inequality: global, international, and intercountry On the welfare gains from globalization, the tional inequality, the inequality in the distribu- the division of the gains from globalization. two sides of the debate often make tion of all of the world's citizens, but with each The measures most widely quoted by the crit- statements that are diametrically opposed, all person assigned the mean income of his/her ics of globalization treat each country as one the while examining the same data.While country instead of his/her own income. Global observation, while decompositions of world there are some differences in and problems inequality is calculated by simply adding inter- inequality into between-country and within- with data, the wide discrepancy in views on national inequality to within-country inequality. country components described above give the topic seems to stem from the fact that the people equal weight, whether they live in Intercountry inequality: each country has two sides do not share the same values about China or Chad. one representative at its mean income what constitutes a just distribution of the Note that in the globalization debate, the These two concepts, however, are not enough to gains from globalization. choice of the measure of inequality can also settle the debate.Think of the following state- Considered here are three different depend on the question one is trying to ment in support of the argument that inequality concepts of inequality, drawing from Milanovic answer. If one is interested in the impact of in the world has been increasing:"The GDP per (2005) and Ravallion (2004a). Both authors, and some"globalizing"policies on growth or distrib- capita of the richest country in the world was the globalization debate in general, discuss utional outcomes at the country level, it might about 9 times that of the poorest around 1870 these"competing concepts"in the domain of be preferable to use a measure of intercountry compared with 45 times by 1990."6 Notice that incomes. But these concepts can be extended to inequality. while this statement seems to be referring to other dimensions, such as health and education something akin to international inequality, there Why use international inequality-- (especially for inequality between countries). is a subtle but very important difference: the as we do in this report? The conclusions one would draw in each of size of the richest or the poorest country plays Alternatively, if we are trying to determine these dimensions of well-being then depend on no role in this statement.The statement remains whether world poverty or inequality decreased the concept of inequality adopted. It is impossi- the same whether the richest country is Palau as a result of"globalizing"policies, then we ble for the two sides to communicate without and the poorest country is Jamaica, or whether might be more inclined to examine measures of first making these concepts clear. they are China and India. international inequality. Has global income inequality increased or This is why a third concept is needed. In this not? Before we can answer this, we have to No right or wrong choice concept, all countries of the world (instead of all define what we mean by global inequality and Arguments can be made in favor of each of citizens) line up together, and each of them is how that differs from what we will call interna- these two concepts when assessing trends in assigned her mean income.We will call the tional and intercountry inequality. inequality between countries.This choice is not inequality in this distribution (of roughly 200 or a matter of what is right or wrong. When it Global inequality: forget country so countries of the world) intercountry inequal- comes to judging inequality, intelligent people boundaries, each person has his ity. Milanovic (2005) refers to our intercountry, can disagree about whether countries or peo- or her real income international, and global inequality concepts as ple should be weighted equally--something Global inequality is easy to define: simply forget Concept 1, Concept 2, and Concept 3 inequality, that Ravallion (2004a) argues in detail.The countries, line up all citizens of the world, and respectively (see figure below). point: the judgments (or the questions of inter- calculate the inequality in the distribution of Why use intercountry inequality est) that affect the choice of the inequality con- their real incomes, adjusted for purchasing The implicit value judgment in using inter- cept employed in empirical work matter greatly power parity.5 The global inequality measures country inequality instead of international to the assessment one can make about the dis- that belong to the general entropy class, such as inequality is that countries, not people, should tributive justice of current globalization a mean log deviation or Theil's index, can be get equal weight in assessing the fairness of processes. neatly decomposed into inequality attributable to inequalities between persons within each country and the mean differences of income between countries (Shorrocks 1980). Three concepts of inequality illustrated Within-country inequality is what the overall inequality in the world would be if there were no differences in mean consumption across Intercountry inequality: countries but each country had its actual Three countries and three representatives inequality level. Between-country inequality can with mean incomes (height) be interpreted as measuring what the level of inequality in the world would be if everyone within each country had the same (the country- International inequality: average) consumption level.Total inequality in Entire population included, the world is the sum of these two parts, and the but with mean incomes ratios of the respective parts to total inequality provide a measure of the percentage contribu- tion of between-country and within-country inequality to total inequality. International inequality: each person has Global inequality: his or her country's mean income All individuals with their actual income Throughout the rest of the report, we will refer to this between-country inequality as interna- Sources: Milanovic (2005) and Ravallion (2004a). 58 WORLD DEVELOPMENT REPORT 2006 Figure 3.1 Vanishing twin peaks in life expectancy at birth 1960 1980 2000 Estimated density, unweighted 0.05 0.05 0.05 0.04 0.04 0.04 0.03 0.03 0.03 0.02 0.02 0.02 0.01 0.01 0.01 0.00 0.00 0.00 29 35 41 47 53 59 65 71 77 83 89 29 35 41 47 53 59 65 71 77 83 89 29 35 41 47 53 59 65 71 77 83 89 µ = 50.2 µ = 62.3 µ = 66.4 1960 1980 2000 Estimated density, weighted 0.05 0.05 0.05 0.04 0.04 0.04 0.03 0.03 0.03 0.02 0.02 0.02 0.01 0.01 0.01 0.00 0.00 0.00 27 33 39 45 51 57 63 69 75 81 87 27 33 39 45 51 57 63 69 75 81 87 27 33 39 45 51 57 63 69 75 81 87 µ = 53.4 µ = 61.0 µ = 64.8 Source: Schady (2005). these two regions having increased from 5.8 consistent improvements in life expectancy years to 15.6. Between-country inequality at birth came to a screeching halt in the declined until the early 1990s and then 1990s (table 3.1). Between-country inequal- increased back to its 1980 level by 2000. The ity among developing countries is as high as large decline in life expectancy at birth in it has ever been since 1960. Sub-Saharan Africa more than offset the So, there is some convergence in life inequality-reducing effect of growth in expectancy at birth over a long period, South Asia in the 1990s. although there are significant losses in the Over a longer period (1820­1992) Bour- 1990s in Sub-Saharan Africa, mainly caused guignon, Levin, and Rosenblatt (2004a) by AIDS, and in some European and Central show tremendous gains in life expectancy at Asian countries.9 With the developed coun- birth (rising from approximately 27 years to tries reaching a biological limit at the top of 61 years), unequally distributed at first, the distribution and many regions catching then equalizing in three waves between late up to them, the inequality of life expectancy nineteenth century and 1990. Decades of in the world will become more a function of changes in health and population growth in Sub-Saharan Africa--barring a major health Table 3.1 Increases in life expectancy at birth slowed down dramatically in the 1990s catastrophe elsewhere in the world. (We 1960 1970 1980 1990 2000 revisit this issue at the end of the chapter.) But for now there remain two worlds with Mean 53.4 57.4 61.0 64.0 64.8 significantly different life expectancies: the Coefficient of variation 0.233 0.203 0.183 0.173 0.194 gap in life expectancy between Sub-Saharan Theil-T 0.027 0.021 0.017 0.016 0.020 Africa and Europe and North America in Theil-L 0.028 0.022 0.018 0.017 0.021 2000 is higher than it was in 1950.10 Source: Schady (2005). Health outcomes of even the rich citi- Note: Theil-L and Theil-T are two inequality measures that belong to the general entropy class, with parameters 0 and 1, respectively (unweighted). zens in poor countries remain well below Equity from a global perspective 59 the average in OECD countries. For exam- Figure 3.2 Life expectancy is highly correlated with income, particularly in poor countries ple, for all countries with average per capita Life expectancy, 2000 GDP below the $2 a day threshold, the child 85 mortality rate of the richest 20 percent of Japan Mexico Argentina Spain the population is more than 10 times the OECD average of six years.11 While this is 75 Germany United States China Rep. of Korea also certainly the case with many other Brazil indicators, it is difficult to make statements 65 Russia about how large differences between coun- Pakistan tries are in comparison with inequalities India 55 within countries. Unlike the income Gabon inequality literature, there are no accepted South Africa practices for decomposing inequalities in 45 Namibia health into within-country and between- Botswana country components.12 35 Consider, however, a simple thought 0 10,000 20,000 30,000 experiment. In the 45 developing countries GDP per capita, 2000, current PPP $ for which a Demographic Health Survey Source: Deaton (2004). was available in 2000, 4.9 million infant Note: The curve is nonparametrically fitted, weighted by population. The figure plots country life expectancy (using circles whose size is proportional to population) against GDP per capita in purchasing power parity (PPP) dollars at the deaths could be prevented by bringing their turn of the twenty-first century. infant mortality levels to the OECD aver- age. But if one eliminated the infant mortal- ity gap between the rich and the poor within each of the same countries by lower- ing the infant mortality rate for everyone to the adoption of even inexpensive tech- the level of the top decile, 3.1 million infant niques, adequate nutrition, and water and deaths could be prevented.13 While the sanitation infrastructure. Life expectancy average infant mortality rate for the rich in increases steeply with income among the these poor countries is almost five times poorest countries (figure 3.2).15 larger than the OECD average, it seems But differences in income growth ex- that eliminating within-country differences plain less than a sixth of the variation in between the rich and the poor (by improv- improvements in life expectancy at birth. ing the health of the poor), at least in this More important determinants are clean particular case, would get us about two- water, health systems, demand for ade- thirds of the way to the number of total pre- quately operated and equipped health sys- ventable deaths (by moving everyone to the tems, and basic sanitary knowledge, the lat- OECD average). ter two having much to do with education, So, while large differences in health out- particularly women's education.16 comes remain between countries and within While life expectancy at birth continued them, it is not possible to make definitive to increase, and the infant and child mortal- statements about the relative weight of ity rates declined, the last decade of the these components in global health inequali- twentieth century has seen a divergence ties. One can say, however, that there is no between rich and poor countries.17 The dif- clear presumption that inequalities between ficulties faced by Europe and Central Asia countries dwarf those within them. This countries during transition, and the spread finding, as we will see later in this chapter, of HIV/AIDS and civil conflicts, were major stands in sharp contrast to that in incomes factors in this, but they are not solely but is congruent with that in education. responsible.18 Cornia and Menchini (2005) While technical change in private and cite changes in health spending, public public health knowledge may be more health programs, and the structure and sta- important to account for the overall better bility of households as possible reasons for health,14 income may be important in the the slowdown in health progress in devel- poorest countries, through its impact on oping countries. 60 WORLD DEVELOPMENT REPORT 2006 Figure 3.3 The distribution of years of schooling improved greatly in the second half of the twentieth century 1960 1980 2000 Estimated density, unweighted 0.20 0.20 0.20 0.15 0.15 0.15 0.10 0.10 0.10 0.05 0.05 0.05 0.00 0.00 0.00 0 1 3 5 7 9 11 13 0 1 3 5 7 9 11 13 0 1 3 5 7 9 11 13 µ = 3.3 µ = 4.6 µ = 6.3 1960 1980 2000 Estimated density, weighted 0.20 0.20 0.20 0.15 0.15 0.15 0.10 0.10 0.10 0.05 0.05 0.05 0.00 0.00 0.00 0 1 3 5 7 9 11 13 0 1 3 5 7 9 11 13 0 1 3 5 7 9 11 13 µ = 4.4 µ = 5.6 µ = 6.6 Source: Schady (2005). Global inequalities in education Many of these countries saw large in- The unweighted distribution of adult creases in enrollments in the 1960s, 1970s, school attainment was clearly left-skewed and 1980s. As these younger cohorts aged, in 1960 (figure 3.3). This reflects the fact the mean years of schooling in their coun- that many countries, particularly in Africa tries increased, and the skewness in the and Asia, had mean years of schooling close unweighted international distribution of to zero. The weighted distribution, like that schooling disappeared. Note that the bi- for life expectancy at birth, was bimodal, modal distribution of schooling across per- with one peak around two years of mean sons (weighted by population) persisted schooling and a smaller peak around eight until the 1990s and then gave way to a uni- years. modal distribution only by 2000. By any measure the international distri- Figure 3.4 Mean years of schooling increased while inequality declined across birth cohorts bution of years of schooling has undergone dramatic changes between 1960 and 2000. Inequality in years of schooling Mean years of schooling As mean levels have risen, inequality has 1.0 9 Inequality in years of schooling fallen, decade after decade (figure 3.4). The 0.8 8 mean years of educational attainment for the world almost doubled from 3.4 to 6.3 0.6 7 (table 3.2). Sub-Saharan Africa, the Middle East and North Africa, and South Asia 0.4 6 started with high inequalities (not shown Mean years of schooling 0.2 5 here) and reduced them over time--the Middle East and North Africa region was 0.0 4 particularly successful. Latin America and 1935­9 1940­4 1945­9 1950­4 1955­9 1960­4 1965­9 1970­4 1975­9 the Caribbean and East Asia also had some Year of birth inequalities, which they essentially elimi- Source: Araujo, Ferreira, and Schady (2004). nated for their youngest cohorts. Despite the Note: Inequality in years of schooling is measured using GE (0.5), that is, the general entropy class inequality measure with an inequality aversion parameter of 0.5. progress, mean levels of educational attain- Equity from a global perspective 61 ment in Sub-Saharan Africa and South Asia Table 3.2 Mean years of schooling increased continuously while inequality declined remain low even for the youngest cohorts. 1960 1970 1980 1990 2000 While significant disparities remain in Mean 3.38 3.82 4.67 5.55 6.30 educational attainment across countries Coefficient of variation 0.739 0.705 0.612 0.518 0.461 despite evidence of significant catch-up by Theil-T 0.281 0.259 0.195 0.143 0.115 poorer countries in the past half century, Theil-L 0.392 0.365 0.250 0.179 0.144 there is also large variation within countries (chapter 2). In fact, less than 20 percent of Source: Schady (2005). Note: Theil-L and Theil-T are two inequality measures that belong to the general entropy class, with parameters 0 and the inequality in educational attainment 1, respectively (unweighted). between adults born between 1935 and 1979 is attributable to that between coun- Figure 3.5 Gender disparities in years of schooling declined tries, a share that has been steadily declining but remained significant in some regions over time. While both inequality within and between countries is declining, the rate of Male to female schooling ratio convergence in country means has been OECD faster. East Asia Europe and Central Asia The story remains the same when decom- South Asia posing inequality in educational attainment Latin America and the Caribbean into inequalities between men and women. Middle East and North Africa Sub-Saharan Africa Roughly a quarter of global inequality in edu- 4 cational attainment is attributable to differ- ences between men and women, but this gap is again declining over time, from 31 percent in the oldest cohort in our sample, to 16 per- 3 cent in the youngest. But there are large dif- ferences in this convergence by region (figure 3.5).While Latin America and the Caribbean, 2 East Asia, and Europe and Central Asia seem to have reached gender parity in education, along with other developed countries, the 1 progress in South Asia, Sub-Saharan Africa, and the Middle East and North Africa has 1935­9 1940­4 1945­9 1950­4 1955­9 1960­4 1965­9 1970­4 1975­9 been slower. Women still lag far behind men Year of birth in educational attainment. Source: Araujo, Ferreira, and Schady (2004). It should not be assumed that high attainment necessarily implies high achieve- ment, and vice versa. An analysis of the rela- of Australia, Finland, and Ireland are much tionship between attainment (measured by higher than their rankings in attainment. the percentage of 25 to 34 year olds with Achievement differences between devel- upper-secondary education) and achieve- oping countries and OECD countries re- ment (measured by reading proficiency of main strikingly large. Using internationally 15 year olds) in 27 OECD countries (plus comparable assessments of reading, mathe- Brazil) shows a rank correlation coefficient matics, and science, Pritchett (2004b) between these two variables of 0.57. It is shows that developing countries do not just clear that the rankings of countries accord- constitute the lower tail of the learning dis- ing to these two indicators are not the same. tribution, but that most actually do far worse The Republic of Korea and Japan (at the top than the poorest performing OECD coun- of the OECD distribution) and Mexico, Por- tries. For example, children in Argentina, tugal, and Turkey (at the bottom) have simi- Mexico, and Chile perform about two lar ranks for both attainment and achieve- (OECD) standard deviations below chil- ment. But the Czech Republic, Norway, and dren in Greece--one of the poorest per- the United States do worse in achievement forming countries in the OECD. In reading than attainment. And achievement rankings competence (based on PISA 2001), the 62 WORLD DEVELOPMENT REPORT 2006 average Indonesian student performed at between 1997 and 2002, as well as the dis- the level of a French student at the seventh persion of those incomes within each coun- percentile. Considering children who have try (figure 3.6). Large differences across never attended school, those who enrolled countries and across people within coun- but dropped out, and those who completed tries are striking. For example, an indi- grade nine but whose test scores remain vidual in the tenth percentile in the U.S. more than one standard deviation below distribution enjoys a level of income higher the OECD mean in mathematics, Pritchett than an individual earning the mean finds that 96 percent of 15 to 19 year olds in income in Brazil or Argentina.20 While a Morocco lack achievement in "adequate Chinese individual living in a rural area has learning."19 a mean income similar to an average Cam- bodian, an urban Chinese enjoys a similar Global inequalities income to an average Brazilian.21 A South in income and expenditure African at the bottom of the income distri- The answers to basic questions--such as bution in her country earns as much as the whether income inequality has been increas- average individual in Mali while a South ing or decreasing--depends, among other African at the ninetieth percentile of that things, on which concept of inequality is income distribution enjoys a standard of under the microscope: intercountry inequal- living (in income) comparable to that of a ity (in the distribution of unweighted coun- median Irish individual. try means), international inequality (in the The difference in the evolution of inter- distribution of country means weighted by country (unweighted) and international their population size), or global inequality (weighted) inequality between 1950 and (in the distribution of individual incomes). 2000--borrowing from Milanovic (2005), We start the discussion by presenting the who calls this the "mother of all inequality median and mean incomes of selected disputes"--could hardly be more dramatic countries by region for a range of years (figure 3.7). When countries are the unit of Figure 3.6 Incomes range broadly across countries and individuals Mali Sub-Saharan Africa Ethiopia South Africa Cambodia India, rural South Asia Pakistan India, urban China, rural East Asia Indonesia China, urban Nicaragua Latin America and Argentina the Caribbean Brazil Yemen Middle East and Morocco North Africa Israel Russia Europe and Albania Central Asia Poland Ireland OECD countries Denmark United States 0 250 500 750 1000 1250 1500 1750 2000 2250 2500 Income per capita (1993 PPP $) Source: Authors' calculations. Note: Years range from 1997 to 2002 as measured by adjusted (1993 PPP $) monthly per capita income (blue box) or consumption (orange box). The lowest point of each line represents the income level at the tenth percentile, followed by that at the median, the mean (the two edges of each box), and the ninetieth percentile (top of each line). Equity from a global perspective 63 Figure 3.7 Since 1950, intercountry inequality Figure 3.8 Unlike relative inequality, absolute increased, while international inequality declined inequality has been steadily increasing Gini index Indexes, 1970 = 100 0.6 International inequality Kolm index (0.3) Absolute Kolm index (3.0) (weighted) measures Absolute Gini index Gini index International inequality Theil index Relative without China and India Mean logarithmic measures 0.5 deviation 250 225 Intercountry inequality (unweighted) 200 0.4 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 175 Source: Milanovic (2005). 150 observation, (intercountry) inequality has 125 been undeniably increasing, especially since the 1980s. But international inequality has 100 been steadily declining, thanks mostly to the income growth in some populous coun- 75 tries, mainly China and India. Note that 1970 1975 1980 1985 1990 1995 2000 intercountry inequality and international Source: Atkinson and Brandolini (2004). inequality without China and India track each other quite closely from 1980 onward, coinciding with the period of rapid growth absolute terms or relative terms. There is no in these two countries, the slower average economic theory that tells us that inequality growth in other developing countries, and is relative, not absolute. Again, as with inter- the declines in measured output in Eastern country and international inequality, it is Europe and former Soviet Union countries. not that one concept is right and the other If Luxembourg and Nicaragua, at oppo- one wrong. Nor are they two ways of meas- site ends of the world income distribution, uring the same thing. Instead, they are two grew at the same annual rate of 2 percent different concepts. The revealed preferences per capita a year for the next 25 years, the for one concept over another reflect implicit per capita yearly incomes in Luxembourg value judgments about what constitutes a would increase from $17,228 (PPP-adjusted) fair division of the gains from growth. to $28,264, an increase of more than Those judgments need to be brought into $10,000 dollars. That of Nicaragua, by con- the open and critically scrutinized before trast, would increase by a mere $375, from one can take a well-considered position in $573 to $940, during the same period. this debate. Atkinson and Brandolini (2004) note that An examination of international inequal- "with annual per capita growth rates of 5 ity using absolute rather than relative meas- percent in China and 2 percent in the ures of inequality reveals a steady increase United States, the absolute income gap over the long run, as well as in recent between the two countries would widen for decades--this latter finding contrasts with a further 41 years before starting to narrow, relative international inequality trends. Atkin- to finally disappear in 72 years." son and Brandolini (2004) find that absolute The evaluative judgments drawn about indexes of inequality, such as the Absolute the distributional changes associated with Gini and the Kolm Index22 (with various globalization may depend crucially on parameters of inequality aversion), have been whether one thinks about inequality in increasing steadily since 1970 (figure 3.8).23 64 WORLD DEVELOPMENT REPORT 2006 What happened to global inequality in Most of the world's income inequality the past 20 years or so has been the subject can be explained by the differences in coun- of fierce debate in the context of globaliza- try means--that is, by international (or tion and is perhaps the hardest question to between-country) inequality. Our estimates answer. Some authors24 claim that global show that the share of global inequality, inequality increased slightly, while others25 which can be attributed to inequality argue that they have declined. between countries, declined steadily from Examining global inequality requires 78 percent around 1988 to 74 percent knowledge of the distribution of inequal- around 1993 and to 67 percent by around ity within each country. Household sur- 2000. With global inequality staying roughly veys that collect such data are a relatively the same during this period, within-group new phenomenon, having become more inequality increased at a somewhat steady common since the 1980s even in develop- pace (figure 3.9). These results are consis- ing countries. So, if we want to know tent with the evidence (in chapter 2) of about the distribution of income for increasing inequality within countries in everyone in the world, we are confined to many parts of the world, including a much shorter time period. We have Bangladesh, China, the United Kingdom, selected three waves, similar to those used and the United States. by Milanovic (2005): 1986­1990, 1991­96, The between-country share of global and 1997­current. inequality is also consistent with Milanovic Global inequality (measured by the (2005), who puts this figure at about 71 per- mean log deviation) did not change signif- cent in 1998. It is possible that the icantly over this period, although there is a Milanovic figures overestimate between- slight decrease between 1993 and 2000 country inequality because he assigns all (figure 3.9). The mean log deviation for households in a decile the same income the world would have increased without instead of estimating a Lorenz curve (for China and India, consistent with the con- percentiles). Our results use slightly sensus in the literature that international improved data from Milanovic in three inequality declined in this period thanks aspects. First, for many countries, we calcu- largely to these two countries. But if global late our welfare measures using raw data at inequality stayed roughly the same while the household level, while Milanovic (and international inequality declined, inequal- many others) use grouped data. Second, we ity within countries must have increased incorporate more recent data for the cur- by approximately the same amount--a rent period, possibly providing an improve- subject that we discuss below. ment in data quality, especially for Eastern European countries. Third, for the coun- tries with grouped data, we estimate Lorenz Figure 3.9 The inequality decline between countries was neutralized by increases within countries curves instead of assigning everyone in the group with the same income.26 That most Mean log deviation Percent of the global inequality in incomes is 1.15 Global inequality Between-country explained by between-country inequality (left axis) share (right axis) seems to be a robust finding in the litera- 100 0.90 0.87 ture, in stark contrast with the picture in 0.84 0.82 health and education. 74 78 75 67 Over a much longer period (1820­1992) 0.65 Bourguignon and Morrisson (2002) esti- 0.65 0.64 50 mate that global inequality has been Between-country 0.55 inequality (left axis) Within-country steadily increasing, because of a rapid 0.40 inequality (left axis) 25 increase in international inequality until 0.27 0.22 World War II, and then to smaller increases 0.19 0.15 0 in both within-country and international 1988 1990 1992 1994 1996 1998 2000 inequality between 1970 and 1992 (figure Source: Authors' calculations. 3.10).27 They also argue that international Equity from a global perspective 65 inequality was essentially negligible at the Figure 3.10 Inequality between countries became turn of the nineteenth century (accounting much more important over the long run for roughly 12 percent of global inequality), Mean log deviation but that it increased very rapidly until 1 World War II, and then continued to 0.83 increase, but at a much slower pace. Within 0.8 0.69 Global inequality country inequality, however, reached its peak around 1910 and declined dramati- 0.6 Within-country inequality 0.50 cally between the two world wars (mainly 0.42 0.4 0.36 because of equalizing forces in the now- 0.37 0.33 developed countries), and started creeping 0.33 0.2 back up only since the 1970s. The combined 0.05 Between-country inequality effect of these changes is an increase in the 0 share of international inequality from 1820 1850 1870 1890 1910 1929 1950 1960 1970 1980 1992 roughly 10 percent in 1820 to more than 60 Source: Authors' manipulation of data from Bourguignon and Mor- percent by 1992. risson (2002). In summary, while the world got richer, income inequality--relative and absolute, ance) between 1960 and 1990; however, international and global--increased tremen- there also were large differences by region. dously over a long period of time Inequality between the countries of OECD (1820­1992). But the story is less clear-cut (and the rest of Europe including Turkey) for a more recent time frame. In the decreased by 50 percent during this post­World War II era, intercountry inequal- period, at the end accounting for only one- ity (unweighted) has continued to increase third of total inequality. During the same while international inequality (weighted for period, international inequality in Sub- population) declined. International in- Saharan Africa nearly doubled, causing its equality declined in the final decades of the share in total inequality to increase from twentieth century, because the inequality- 20 percent to 36 percent. In both Sub- reducing effects of income growth in Saharan Africa and Latin America and the China and South Asia more than offset the Caribbean, overall inequality levels remain inequality-boosting effects of continued high, while high-income countries show steady income growth in the now-developed signs of convergence. countries and the declining incomes in Sub- One can also examine inequality trends Saharan Africa. by focusing on the mobility of countries Pritchett (1997), examining the period rather than by taking an anonymous between 1870 and 1990, argues that while approach to inequality comparisons. Poor there was convergence of incomes for countries' mobility from the bottom has today's developed countries (what Maddi- been limited in the past 25 years. With the son 1995 calls the "advanced capitalist" exception of China, the six countries that countries), the growth rates between devel- occupied the bottom decile (population- oped and developing economies show con- weighted) in 1980--all in Sub-Saharan siderable divergence. He provides evidence Africa--had no growth worth noting.29 that "the growth rates of developed coun- While there is significant upward mobil- tries are bunched in a narrow group, while ity between 1980 and 2002--the 97.08 per- those of less developed countries are all cent entry in the first row of table 3.3 is over with some in explosive growth and China--there is also troubling stagnation others in implosive decline."28 and downward mobility. Note that approxi- Further evidence of convergence among mately 8 percent of each of the second and rich countries and divergence between rich third income ranges fell into the bottom and poor countries comes from Schultz range over these two decades. "It is clear (1998), who estimates that international that no Pareto improvement has taken place inequality accounted for about two-thirds in the world between 1980 and 2002, which of total inequality (measured by log vari- leaves room for different value judgments 66 WORLD DEVELOPMENT REPORT 2006 Table 3.3 Mobility matrix in absolute country per capita incomes, 1980 to 2002 While some populous countries, almost Income in 2002 exclusively in Asia, such as Bangladesh, Income in 1980 <710 711­1,100 1,101­2,890 2,891­10,000 10,001> China, India, and Pakistan, made signifi- cant headway against extreme poverty, <710 1.28% 1.64% 0.00% 97.08% 0.00% almost all increases in extreme poverty-- 711­1,100 8.23% 3.89% 87.88% 0.00% 0.00% especially in countries with high initial 1,101­2,890 8.09% 0.56% 59.08% 32.28% 0.00% headcount rates--took place in Sub-Saha- 2,891­10,000 0.00% 0.00% 0.98% 90.84% 8.17% ran Africa.32 Among the larger countries 10,001> 0.00% 0.00% 0.00% 3.99% 96.01% with rising headcount rates are Nigeria, Source: Bourguignon, Levin, and Rosenblatt (2004a). South Africa, and Tanzania. Note: Incomes are per capita (constant PPP dollars). If the poverty trends discussed here con- tinue, the Millennium Development Goal of Figure 3.11 Absolute poverty declined globally, but not in every region halving the proportion of people living on Headcount ($1 per day) in 2001 less than $1 a day will be met. But only East and South Asia will reach this goal. We can- AFR not be satisfied if this were to happen. Other 80 EAP NGA things equal, we would prefer to see the ECA ZAM poverty rate falling at the same pace in all LAC MDG countries. Currently, hundreds of millions 60 CAF MLI MENA NER BDI ZWE of people in numerous developing countries SAR MWI BF A TZA lack the opportunity to avoid hunger, poor LSO R W A 40 health, and low access to vital services, such GHA NIC MOZ KHM IND as education and clean water.33 NAM KEN BGD CIV GAM ETH MON 20 Global inequalities in power CMR MAU VEN BWA GTM CHN UZB YEM LAO MYS ZAF MEX One of the main arguments in the conclud- PHL SEN P AK RUS COL BRA THA ARG UKR TUR ing chapter of this report is that the rules VNM 0 SLE EGY and processes in global markets can be LKA JOR MAR unfair to developing countries. A country's 0 20 40 60 80 power in decision making in multilateral Headcount ($1 per day) in 1981 banks is usually correlated with its eco- Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp). nomic strength. Even when each country has equal representation in an international body, such as the United Nations system or the World Trade Organization (WTO), pow- about the evolution of world welfare, inequality, and relative poverty."30 Milanovic erful forces can chisel away at developing- country interests (through separate bilat- (2005) also point out the "downward mobil- eral agreements, for example). And the ity"of many countries in the past 40 years or capacity of developing countries to make so. Those who do not share the view that informed decisions can be limited. inequality between countries fell in the past 20 to 25 years--that is, those who take the Poor countries lack the financial and human "unweighted" view of the world--may have capital resources that would allow them to be such mobility concerns in mind. equal participants in the international bodies Absolute poverty rates have declined in in which decisions are taken that affect them and, beyond that, in setting the rules under the past 20 years or so, and a variety of which the international system operates.34 studies have confirmed this trend (figure 3.11).31 Overall, while there are roughly In the International Bank for Recon- 400 million fewer people who live on less struction and Development (IBRD)--the than a $1 a day in 2001 than there were in market-lending arm of the World Bank--a 1981, the number of poor people in Sub- country's voting power depends on the per- Saharan Africa almost doubled, from centage of IBRD shares it holds. The largest approximately 160 million to 313 million. shareholders are the United States with 16.4 Equity from a global perspective 67 percent of the vote, Japan with 7.9 percent, Figure 3.12 There is no one-to-one relationship Germany with 4.5 percent, France and the between voice and income United Kingdom with 4.3 percent. Each has a Denmark representative on the Board of Directors. By Ireland contrast, all Sub-Saharan countries together United States have two representatives and 5.2 percent of Poland the vote. China and India both have 2.8 South Africa percent of the vote.35 Country influence in Israel setting the agenda for the institutions is not India limited to board membership.A 1998 study by Brazil Filmer and others (1998) shows that roughly Mail two-thirds of the senior management­level Argentina positions at the World Bank are occupied by Nicaragua citizens of Part I (mainly OECD) countries, Albania although these countries account for less Morocco than one-fifth of the global population and Indonesia a smaller share of the number of member Russia countries. Cambodia At the WTO, each member country has Yemen Voice and one vote. Moreover, because decisions are accountability Pakistan by consensus, each country effectively has Income category Ethiopia veto power. So the WTO is, at least on 0 20 40 60 80 100 paper, perhaps the most democratic of international organizations. In practice, the Source: Kaufmann, Kraay, and Mastruzzi (2004). Note: "Voice and accountability" refers broadly to the extent to ability of countries to influence the agenda which citizens have freedom of expression, a free press, and open and decisions depends crucially on their elections based on a statistical compilation of responses on the quality of governance given by a large number of enterprise, citi- capacity to be present, to follow negotia- zen, and expert survey respondents in industrial and developing tions, to be informed, and to understand countries, as reported by a number of survey institutes, think tanks, nongovernmental organizations, and international organizations. fully the impact of the complex issues at Countries' relative positions on these indicators are subject to margins of error that are clearly indicated. Consequently, precise hand. A rough indicator of a country's country rankings should not be inferred from these data. capacity is the size of its representation in Geneva. A study by Blackhurst, Lyakurwa, and Oyejide (2000) found that only 8 of the ative assessment to date, integrate data col- 38 Sub-Saharan countries had close to five lected by 25 separate sources constructed by (the WTO average) resident delegates listed 18 (commercial and advocacy) organiza- in the WTO directory. Worse, 19 of the 38 tions. The authors used the data to provide countries--half of the Sub-Saharan WTO a common empirical basis to assess the rela- membership--had no delegate resident in tive differences among countries of the Geneva. Only Nigeria had a delegation that quality of their "governance." deals solely with the WTO.36 Figure 3.12 summarizes information on Even when country representation in the "voice and accountability," which refers international arena is considered adequate, broadly to the extent to which citizens have it is debatable whether the representatives freedom of expression, a free press, and of some countries are fully accountable to open elections, using standardized meas- their citizens. There are considerable differ- ures for selected countries (the same ones as ences among countries in the extent to in figure 3.6). The upper bar for each coun- which their political and legal institutions try represents the country's percentile rank provide citizens with fair, transparent, and in the "voice and accountability" distribu- inclusive environments to enhance and tion with the intersecting black rule line leverage their assets. While there are numer- representing the confidence interval. The ous problems with trying to measure such lower bar is the average percentile score for things, Kaufmann, Kraay, and Mastruzzi the income category to which the country (2004), in the most comprehensive compar- belongs.37 The top of the "voice" rankings is 68 WORLD DEVELOPMENT REPORT 2006 filled with wealthy countries, such as Den- What explains the convergence in health mark, the United States, Ireland, and Israel. and education and the lack of it in incomes? The voice ranking of such countries as Deaton (2004) points out that, while gains South Africa, Poland, and especially Mali in income were undoubtedly important for and India exceed their ranks in incomes. improving nutrition and funding better The opposite is true for China, Ethiopia, water and sanitation schemes, some coun- Pakistan, and the Russian Federation. Cam- tries made progress in reducing child mor- bodia and the Republic of Yemen are both tality even in the absence of economic quite poor and rank low in freedom of growth. These improvements came from expression. It is clear that there is no one- the globalization of knowledge, facilitated to-one relationship between citizens' voice by local political, economic, and education and average income at the country level. conditions. A possible explanation for the disconnect between the convergence in A glimpse of the future education and the divergence in incomes is Despite improvements over time, inequali- that education is not translating into ties among countries in various dimensions human capital and that the rise in per remain unacceptably high. Each year 10.5 worker schooling explains only a small part million child deaths are preventable in the of the growth in output per worker.39 sense that these children would not have We have seen that the story of income died if they had been born in rich coun- inequality in the world has been a story of tries.38 The mean educational attainment falling international inequalities and ris- level for adults born in 1975­79 in Sub- ing within-country inequality. For global in- Saharan Africa remains at 5.4 years, com- equality, these two effects are offsetting, and pared with 10.1 years in Latin America and the conclusion drawn depends on knowing Caribbean and 13.4 years in developed which effect dominated.The decline in inter- countries. Developing countries also face national inequality is largely due to fast massive challenges in influencing the global income growth in China and South Asia.40 rules and processes that determine out- But as China and South Asia catch up to the comes, which matter greatly to the well- world average, their equalizing effect will being of their citizens. diminish. And if they continue to develop at International inequalities in educational similar rates to that in the past two decades, attainment have been steadily declining. the effect of their growth will increase inter- This is also true in health--one's country national inequality.41 Without the offsetting of birth 50 years ago was much more perti- effect of declining international inequality, nent to survival than it is today. In this global inequality would also be on the rise sense, opportunities across countries are again unless inequality within countries equalizing. But improvements in life starts to decline and Sub-Saharan African expectancy at birth have reversed since the economies begin to experience healthy early 1990s, because of the devastating growth. This suggests that the future of effects of HIV/AIDS and the difficult cir- world income inequality will increasingly cumstances facing citizens of some transi- be a function of economic growth in Africa tion economies. The world distribution of (and some other low-income countries incomes, by contrast, was becoming secu- under stress), especially if the population larly more unequal from the early nine- growth rates in Africa remain above the teenth century until about the end of World world average. That both population War II. Since the war, international in- growth and economic growth in Africa have equality between countries has decreased been stunted by the AIDS tragedy is doubly immensely, because of the fast growth in disturbing. China and India in more recent times, and On whether today's poor countries with global inequality has leveled off. Because stagnant economies will take off, some China and India are only two countries, researchers are optimistic. Lucas (2003) intercountry inequality in incomes has con- suggests that the countries that have not yet tinued to increase. joined the industrial revolution (which he Equity from a global perspective 69 attributes to socialist planning, lawlessness, Thus, millions of babies are being infected and corruption) will become the miracle at birth, which is mostly preventable with economies of the future. He reckons that proper interventions. Life expectancy in the growth rates in these catch-up countries Africa would not improve much, and cer- may be quite high and that they will also go tainly not soon, if these assumed improve- through a similar demographic transition ments in HIV/AIDS prevalence rates do not experienced by today's developed countries. materialize. The world population will stop rising and Because South Asia has almost caught up world production growth will stabilize until to the world average in life expectancy, Sub- all countries, economically, start resembling Saharan Africa will be the only region signif- countries like the United States, thanks to icantly affecting health inequalities between free trade and the diffusion of technology. countries, barring a major catastrophe else- Pritchett--who calls this idea "advantage where.46 So, improvements in life expectancy to backwardness"42--remains more cau- in Sub-Saharan Africa are the key to future tious. Conceding that such rapid gains in declines in international health inequalities. productivity are a possibility, he argues that Chapter 2 documented within-country "the cases in which backward countries, and inequalities in health opportunities for chil- especially the most backward of countries, dren born to poor or rich parents, educated actually gain significantly on the leader are or uneducated mothers, in rural or urban historically rare."43 He observes that there areas, and so on. Steep gradients in health are also forces for "implosive" declines in opportunities and outcomes exist along these countries, suggesting that backward- these dimensions in many countries.A confi- ness may also carry "severe disadvantages." dent assessment of past and future trends in On health in Sub-Saharan Africa, the health inequality awaits future research. UN Population Division projects that life If the trends that brought about the expectancy at birth in Africa will decline catching up of many poor countries outside over the next 5 to 10 years and then start of Africa continue in health, education, and climbing again, reaching 65 years by about incomes, the biggest challenges will remain 2050.44 These projections assume that in Africa and some poor countries in other HIV/AIDS prevalence rates in Africa will regions. Growth with equity needs to be peak sometime before 2010 and then revived in stagnating economies around the decline over the next decades. But the Joint world, and the AIDS tragedy (along with United Nations Programme on HIV/AIDS the folly of preventable diseases) needs to estimated that 43 percent of pregnant be addressed urgently, especially in Sub- females in 2000 were HIV positive in Saharan Africa. These remain the biggest Botswana and 19 percent in South Africa.45 global challenges in development today. f o c u s 2 o n empowerment Popular participation and equitable transitions at the local level Promoting equity through public action requires changes in the existing configurations of power and influence. Because established institutions privilege certain interests and marginalize others, making governance institutions more democratic and more equity-enhancing calls for reforms that increase the possibilities for effective participation by tradi- tionally marginalized groups. L ocal government is a critical domain These two initiatives have much in com- representation in the population.3 Moreover, for the exercise of democratic rights mon. They were both conceived as direct both these initiatives have created a new and for making effective public and conscious efforts to break with the cadre of grassroots politicians who either did choices. But several factors have conspired elite-dominated and clientelistic politics of not exist before (delegates in Brazil) or who against good governance, democracy, and local government by promoting redistribu- previously had no powers (the 14,000 equity at the local level in much of the tive policies through broad popular partici- elected panchayat councilors in Kerala). The developing world. The social and economic pation. Thus, they both shifted the political local public sphere--the sine qua non of any power of local elites has often translated opportunity structure and involved action vibrant democracy--has become more into disproportionate influence over the to strengthen the agency of subordinate extensive, more inclusive, and more mean- political process, and top-down, insulated, groups.2 Both have, in effect, comple- ingful. and nontransparent decision-making mented representative forms of democracy structures have made it difficult for ordi- with participatory forms of democracy by Extending democracy nary citizens to have voice.1 Democratic opening institutions to the direct engage- These initiatives have been marked by the deepening in the developing world often ment of civil society. And both have extension of democracy, specifically public begins with the democratization of local strengthened public authority and public decision making in arenas of authority pre- government, and that is precisely what two action by increasing both the depth and viously dominated by private and state participatory governance initiatives--in the scope of democratic decision making. elites. Municipal budgets in Brazil have long Indian state of Kerala and in a variety of been the preserve of oligarchic parties and municipalities in Brazil--have tried to do. Deepening democracy narrow sectoral interests. Panchayats in In 1996 the state government of Kerala The evidence shows that these initiatives Kerala have long been little more than pas- launched what is widely viewed to be the have deepened democracy, expanding the sive recipients of top-down projects most ambitious initiative for democratic range of social actors participating in the designed and delivered by state bureaucra- decentralization in India: the People's Cam- political arena. In Porto Alegre, an esti- cies. In both cases, citizens now have a voice paign of Decentralized Planning. The gov- mated 100,000 adults have participated at in determining how public resources are ernment not only devolved significant some point in the budgeting assemblies. allocated. In the most successful participa- resources and authority to Kerala's 1,214 Other cities that have adopted some form tory budgeting cases, the entire budget of panchayats (village councils) and munici- of the process have also experienced active the municipality is discussed and approved palities, but it also promoted direct citizen participation, including municipalities by delegates acting on priorities established participation by mandating village assem- without established civil societies. In by neighborhood assemblies, with citizens blies and citizen committees to plan and Kerala, nearly one in four households deliberating on capital and operational seg- budget local development expenditures. attended village assemblies in the first two ments of the budget. In Brazil, the city of Porto Alegre years of the campaign, and despite rou- In some municipalities, direct partici- launched a participatory budgeting initia- tinization of the process in subsequent pation has been extended to thematic tive in 1990 that has since been copied in at years, these assemblies continue to draw areas, such as economic development, least 400 municipalities throughout the large numbers. Hundreds of thousands of public transportation, education, social country. The process begins with neighbor- citizens have undergone training in plan- services, and urban planning. In Kerala, hood assemblies in which citizens deliber- ning and budgeting, and the committees panchayats have been given authority for ate and set budgeting priorities, and ends that actually design and budget specific up to 35 percent of the development with a citywide budget formulated by dele- projects have been composed primarily of budget, a fivefold increase in their gates directly elected by neighborhood civil society actors. resources base. Panchayats have ranked, assemblies. The success in Porto Alegre has A redesign of institutional incentives and designed, and implemented hundreds of seen its steady diffusion, with at least 100 new mobilizational efforts saw women projects a year across all development sec- municipalities, including São Paulo, imple- account for 40 percent of the participants in tors. These have included housing for the menting variations of participatory budget- village assemblies (a level otherwise unheard poor, small-scale irrigation, local roads ing in 1996­2000, and some estimated 250 of in India) and the participation rate of dal- and infrastructure, agricultural projects, municipalities in 2000­04. its (scheduled castes) has exceeded their support services in health and education, Focus on empowerment 71 and a range of projects specifically tar- the economy is otherwise growing, the capacity to realize their interests and aspira- geted at women and dalits. most marginalized social groups--widows, tions that would have been denied them slum dwellers, sex workers, the very poor-- otherwise.7 Young, female, often illegal Enhancing equity may continue to be excluded. What can be immigrants, contractually bound to work These initiatives have generally had equity- done? The challenge is greater when the long hours for ruthless bosses, and facing enhancing effects. In Porto Alegre, the best extent of commitment from above and sure rejection by their families if they man- known and most documented case, there is mobilization from below is less than in aged to escape and return home, the sex clear evidence that expenditures on poorer these two cases. workers had virtually no capacity to exer- areas of the city increased steadily with the As discussed in chapter 2, the most mar- cise their voices and realize their interests. introduction of participatory budgeting. In ginal groups are often stuck in more severe Persistent efforts to organize the women other large cities with participatory budget- forms of an "inequality trap"--a situation into a union, however, eventually gave them ing, such as Belém and Belo Horizonte, characterized by dire material circum- the confidence and competence to bring expenditures have also targeted the poor. A stances, rational expectations about limited about a change in condom use by clients. statistical analysis of all Brazilian munici- mobility opportunities, and internalized In Indonesia, the Kecamatan Develop- palities in 1997­2000 revealed that partici- beliefs regarding the legitimacy and ment Project (KDP), which operates in patory budgeting cities had significantly immutability of their circumstances. Break- 28,000 villages across the country, seeks to higher expenditures on sectors that affect ing out of such inequality traps and improving the "terms of recognition" and the poor most directly.4 improving the terms by which the poor are political agency of marginalized groups. 8 In Kerala, a large survey of key respon- "recognized" by others starts with building The project allocates grant money at the dents found that "disadvantaged" groups both a "capacity to aspire" and, equally subdistrict level, for which several groups of were the prime beneficiaries of targeted important, a "capacity to engage."6 This poor villagers (two of whom must be schemes. Case studies show that panchayats includes being able to envision and enact women) are invited to compete for funds have emphasized the need to bring all alternative futures, believing that it is desir- on the basis of the presentation of a formal households up to a certain basic level of able and possible to move out of poverty, subproject proposal. KDP's procedures, well-being, with a heavy emphasis on pro- and being able to more meaningfully par- institutions, and norms are largely decen- viding sanitation facilities, decent housing, ticipate in forums where decisions affecting tralized, they focus on joint public problem and safe water to needy families. their welfare are made. solving, they invite broad public participa- In both cases, there is also strong evidence Acquiring a "capacity to aspire" is largely tion and scrutiny, and they occur in a more that the incidence of rent-seeking has fallen a product of developing more broadly or less continuous and institutionalized sharply.5 The greater transparency of the accessible and equitable mechanisms for way.9 budgeting process alone has raised the trans- interaction between the poor and elites, Recent work assessing the impact of action costs of predation and patronage. mechanisms that are reciprocally linked to KDP on local decision making finds that attaining greater voice in associational KDP helps marginalized groups cultivate interactions. It thrives in and through access to more constructive spaces and pro- Empowering the most group organizing and public dialogue, and cedures for addressing project and non- marginal groups the opportunities these afford for practice, project conflicts.10 The beginning stage of The Kerala and Porto Alegre cases illustrate repetition, exploration, conjecture, and such a transformation--in which unequal the value of improving the accessibility, refutation. groups build the capacity to peacefully transparency, and accountability of local An association of sex workers in a Cal- engage one another--is a humble but non- government. However, even when such ini- cutta slum, for example, gave its individual trivial outcome for a development project. tiatives have been undertaken, and where members a voice, a public presence, and a Why does equity matter? II IN THE FIRST PART OF THIS REPORT, WE SUMMARIZED some of the evi- dence on inequalities in several dimensions. In addition to affecting well-being directly, such dimensions as health, education, income, P A R T voice, and access to services shape the opportunities people face for future progress and achievement. We emphasized the interconnections between these various dimensions. Not only is there inequality in the distributions of income, health status, and educational attainment, but--even more important--these indicators tend to be correlated. The rich tend to be both healthier and better educated than others. The poorest of the poor tend to have the lowest attainment in years of schooling and some of the worst health indicators. These correlations generally also extend to public services, with the poor gaining access to infrastructure, electricity, water, sanitation, and garbage disposal much later than others, if at all. Because education and wealth help a person gain influence in soci- ety, voice and political power are also generally thought to be correlated with economic well-being. The interaction between these mutually reinforcing economic, social, and political inequalities perpetuates them across generations. Chapter 2 discussed evidence indicating that a 10 percent difference in economic status between two families in one generation tends to imply, on average, a 4 percent to 7 percent differ- ence in the next generation, depending on the country and measure- ment details. Opportunities clearly are not independent from social and family background, or from group identity. Do such disparities matter? Are people concerned with the large observed differences in access to education and health, and in eco- nomic opportunities, or merely with the fact that some people have low absolute levels of income, years of schooling, and access to serv- ices? Should policymakers worry about the unequal opportunities 74 WORLD DEVELOPMENT REPORT 2006 that arise from discrimination, unequal political voice to most or all of society. One access to justice or other unfair processes? set of reasons for this arises from failures in Should an institution like the World Bank, capital, land, and labor markets. Those fail- whose primary objective is to assist its client ures imply that productive opportunities countries in eradicating extreme poverty, are not necessarily seized by those with the care about inequalities--in opportunities, highest potential returns on their talents or outcomes, and processes--at all? ideas, but instead by those with greater Opinions on these questions are wide- wealth, better connections, or larger land ranging. Support for equal opportunities parcels. This would not happen if markets has long been a theme in domestic policy worked perfectly, as resources would flow to in the United States, for instance. Franklin those with the most productive investment D. Roosevelt once said that "We know that projects. But given that markets are not per- equality of individual ability has never fect, scope arises for efficient redistribution existed and never will, but we do insist that schemes. equality of opportunity still must be Chapter 5 documents cases in which sought." 1 Some participants in the consul- aggregate efficiency could be improved by tations for this report were even offended redistributing wealth or power toward that the question "Does inequality mat- poorer or marginal groups. Sometimes, the ter?" was asked at all, because they consid- evidence of inefficiency is seen in differ- ered its answer to be "Obviously, yes." One ences in marginal products of capital across participant felt that the very question indi- firms. We know that smaller entrepreneurs cated that "we are suffering [from] a terri- pay interest rates much higher than the ble tolerance to horror."2 marginal product of capital accruing to The next three chapters in this report other firms. We know that some farmers address the following question: should allocate effort between plots in a way that is good development policy be concerned not socially efficient, because they own one with equity? Equity, as discussed in chapter plot and sharecrop in another. We have 1, is understood here as the pursuit of equal experimental evidence suggesting that opportunities and the avoidance of severe groups discriminated against perform deprivation. Equity is not the same as equal- below their own capacity, either because ity in incomes, or in health status, or in any they internalize the stereotype or because other specific outcome. It is the quest for a they expect to be treated unfairly. Each of situation in which personal effort, prefer- these pieces of carefully researched empiri- ences, and initiative--rather than family cal evidence, and others discussed in chap- background, caste, race, or gender-- ter 5, provide reasons why more equitable account for the differences among people's economies would, in most cases, also be economic achievements. A situation in more efficient.3 which all institutions are color-blind and Chapter 6 complements this picture by nonmarket institutions are equally respon- looking at historical evidence, suggesting sive to the rich and the poor. In which per- that large inequalities in political rights and sonal and property rights are enforced power give rise to exclusionary institutions equally for all. And in which all have access that generally impair development processes. to the public services and the infrastructure Greater political equality, by contrast, estab- to leverage their productivity and their lishes limits on predation by the most pow- chances of success in the markets. erful in each society. This tends to lead to The evidence we review here has been institutions that level the playing field and assembled in disciplines ranging from eco- provide opportunities for advancement and nomics and history to sociology and mobility to those from underprivileged anthropology. On balance, this evidence backgrounds. suggests that the pursuit of sustainable, Such institutions seem to be associated long-term prosperity is inseparable from a with more sustained growth. One example broadening of economic opportunities and comes from contrasting the exploitative Why does equity matter? 75 labor practices of the Spanish conquista- Equity and fairness matter not only dores in the mining centers of their Ameri- because they are complementary to long- can colonies from the sixteenth to the eigh- term prosperity. It is evident that many teenth century, with the greater freedom people--if not most--care about equity and opportunity afforded to early settlers in for its own sake. Some see equal opportu- North America. Another example of nities and fair processes as matters of inequitable treatment of citizens by the social justice and thus as an intrinsic part state, which was also enormously costly for of the objective of development. In chapter efficiency, was the very high taxation of 4, we briefly review arguments and evi- poor African farmers by state-owned or dence suggesting that most societies parastatal agricultural marketing boards in exhibit a pervasive and long-standing con- Ghana, Nigeria, and Zambia, which pre- cern for equity. vailed a few decades ago. Equity and well-being 4 People from many cultures seem to share a come from religion. Several major world reli- concern for equity that is reflected in reli- gions endorse the notions of social justice gious and philosophical traditions, as well and a duty toward the poor. Buddhists see a c h a p t e r as in legal institutions, both national and duty to care for the poor. Christians are to international. Religions from Islam to Bud- "love their neighbor as themselves." The dhism and secular philosophical traditions Hebrew word for "charity" is the same as the from Plato to Sen have shown both a con- one for "justice." One of Islam's five pillars of cern for equity and an aversion to absolute faith is zakat, providing for the poor and deprivation. In modern legal institutions, needy. The World Faiths Development Dia- equity remains a fundamental tenet of the- logue (1999) states that "all religions would ory and practice. see the extreme material poverty in the world That a concern with equity is so pervasive today as a moral indictment to contemporary across cultures, religions, and philosophical humanity and a breach of trust within the traditions suggests that a fundamental pref- human family."And religious views on equity erence for fairness is deeply rooted in human are not restricted to poverty. Despite varying beings. We review experimental evidence interpretations, and a wealth of differences in showing that many people place a monetary perspective, a belief in the fundamental dig- value on "fairness" and are prepared to give nity of human beings is a theological tenet in up real money if they feel that a process they most major religions. While there are impor- are involved in is unfair. Complementing this tant differences in how this belief manifests evidence are data from opinion surveys, and itself across faiths, and even among different surveys on subjective well-being, suggesting groups within the main religions, some ana- that higher inequality in incomes is, on aver- lysts see a growing emphasis on this principle age, associated with lower aggregate levels of of equality within various faiths.1 subjective well-being. Equity is also a key theme in secular An empirical link between income philosophical traditions. Western political inequality and poverty reduction reinforces and ethical philosophy, for instance, has the conceptual link between the aversion to long been concerned with distribution. In inequality and the quest to avoid absolute ancient Greece, Plato argued that "if a state deprivation. We highlight the obvious fact is to avoid . . . civil disintegration . . . extreme that, if inequality falls during a growth poverty and wealth must not be allowed to spell, poverty generally falls by more than if rise in any section of the citizen-body, inequality had not changed. We also docu- because both lead to disasters."2 Roman law, ment the less obvious fact that higher- while discriminating against slaves, as in all income inequality reduces the effectiveness ancient empires, also laid the foundations of future economic growth in reducing for some of the principles of equality that absolute income poverty. underlie modern legal principles in many countries. Those principles applied only to Roman citizens who were free, but in mod- Ethical and philosophical ern nations they have become all-inclusive. approaches to equity In the modern era, Western thinking Perhaps the oldest manifestations of concern about social justice was greatly influenced with equity and the avoidance of deprivation by utilitarianism--the idea, originally from 76 Equity and well-being 77 Bentham (1789), that the social goal should tunities of the least privileged group. (The be to achieve "the greatest happiness for the Difference Principle is also known as Rawls's greatest number." Although utilitarians "maximin"principle). were essentially unconcerned with the dis- Sen (1985) thought that different peo- tribution of happiness, enjoining societies ple might have different "conversion fac- simply to maximize the sum of utilities tors" from resources to actions and wel- across all individuals, the approach has fare. He argued that all goods, including earned the somewhat misbegotten reputa- Rawls' "primary goods," are inputs to a tion (at least among economists) of having person's functionings--the set of actions a egalitarian implications.3 person performs and of states the person Modern theories of distributive justice values or enjoys. For Sen, the concept to be have largely moved beyond utilitarianism, in equalized across people is the set of possi- part because of its fundamental lack of con- ble functionings from which a person cern with the distribution of welfare. Since the might be able to choose (which he called a early 1970s, a number of influential thinkers, "capability set"). including John Rawls, Amartya Sen, Ronald Dworkin (1981b) and Dworkin (1981a) Dworkin, and John Roemer, have made sepa- argued that justice required that individuals rate and important contributions to the way should be compensated for aspects of their we think about equity. Although the theories circumstances over which they had no con- of justice and social choice proposed by each trol, or for which they could not be held of them are different in important respects, responsible. He argued for a distribution of they share much in common. resources that compensated people for All four reject final welfare (or utility) as innate differences that they could not have the appropriate space in which to judge the helped, including differences in talent. fairness of a given allocation or system. All Roemer (1998) argued that equity acknowledge the importance of individual demanded an"equal opportunity policy."He responsibility in moving from resources to acknowledged that individuals bear some final outcomes, including welfare. All prefer responsibility for their own welfare, but also to see some combination of the set of liber- that circumstances over which they have no ties and resources available to individuals as control affect both how much effort they the right space to form a social judgment. invest and the level of welfare they eventu- All seem to appeal, at some stage, to the "veil ally attain. He argued that public action of ignorance" argument, from Harsanyi should therefore aim to equalize "advan- (1955), that a fair allocation of resources tages" among people from groups with dif- should be one that all "prospective members ferent circumstances at every point along of society"would agree on, before they knew the distribution of efforts within the group. which position they would occupy. They Despite important (but subtle) differ- used this thought experiment to argue that ences, all four thinkers have contributed to justice implies equality in the allocation to shifting the focus of social justice from out- all people of some fundamental concept, comes to opportunities. We also take a leaf such as primary goods. from Nozick (1974), who is usually regarded What Rawls, Sen, Dworkin, and Roemer as an anti-egalitarian. He argued that theo- disagree on is what exactly this concept should ries of justice generally placed an excessive be. Rawls (1971) argued that social justice emphasis on outcomes, such as welfare, required that two basic principles should utility, or even capabilities. Nozick reminded hold. The first "demands the most extensive us of the (obvious) fact that outcomes are liberty for each, consistent with similar liberty the result of processes and argued that the for others."4 The second requires that oppor- correct focus for a theory of justice should tunities--which he related to the concept of be on the fairness of processes. If a particu- "primary goods"--should be open to all lar allocation departs from a fair initial members of society. Under the Difference state, and is arrived at through a fair Principle, he proposes that the chosen alloca- process, it should be judged to be fair-- tion should be one that maximizes the oppor- even if it is unequal. 78 WORLD DEVELOPMENT REPORT 2006 The concept of equity we adopt in this B O X 4 . 1 A simple representation of different concepts report draws on the contribution of these of equity four thinkers by focusing on opportunities, rather than on welfare, utilities, or some The figure below (adapted from figure 11-1 in · If, instead, it wanted to maximize the other corresponding individual outcome. Atkinson and Stiglitz 1980),* can help sum- opportunity set of the"poorest"group, it We do not dwell on the fine distinctions marize what this report hopes to achieve-- should aim for point R. and what it does not.Suppose that a society · If it insisted on absolute equality of between Sen's capabilities and Roemer's consists of only two groups of people (1 and opportunities, it must lie along the 45- opportunities. As in both frameworks, we 2) and let the axes of the diagram depict the degree ray through the origin, and would acknowledge the central role of individual opportunity levels of each group.Opportu- aim for point E. nity sets are obviously multidimensional,but responsibility and effort in determining out- What this report will not attempt to do imagine for simplicity that the various dimen- comes. We focus on eliminating disadvan- is to advise countries on which of these cri- sions can be meaningfully conflated onto an tage from circumstances that lie largely teria of social justice, or indeed any other, an "opportunity index,"O1 for type 1s,and O2 for individual society should aim for. Each of beyond the control of the individual but that type 2s.Now let the curved frontier AC repre- these three points can be defended by logi- sent the"opportunity possibility frontier"for powerfully shape both the outcomes and the cal arguments, under different degrees of this society.** It reflects the maximum oppor- actions in pursuit of those outcomes. aversion to inequalities in opportunity. tunity indexes that types 1 and 2 can obtain, These different perspectives on what a What the report will try to do is as follows: given the available resources and technology. social optimum should be can shed light on The fact that it does not monotonically · Describe the inequalities in opportunity decline from A to C incorporates the fact that actually observed in society (at a point an important point for this report--one when type 1 individuals have very limited such as X). previously mentioned in chapter 1. It is not opportunities,type 2 people can also benefit · Investigate whether some of those dis- for us to advise countries on what exactly from an expansion in type 1 opportunity sets, parities (which in this diagram greatly constitutes an equitable distribution in and vice versa.Over some ranges,improve- favor type 2) might actually be prevent- ments in the opportunity sets of the"poor- their societies. Instead, our role is to point ing the society from enjoying higher est"types can be Pareto-improving--that is, aggregate opportunities (and welfare, on out the inequities we can observe and to can benefit everyone.Put another way,there another space). note that reducing them may be perfectly is scope for efficient,growth-promoting redis- · Suggest possible policy and institutional consistent with--perhaps even necessary tribution. approaches that might help move from Eventually, however, tradeoffs set in. for--greater efficiency and prosperity in the points such as X to whichever point soci- Between points P and R, if society is on the long run. Box 4.1 draws on a classic public ety considers equitable, on the opportu- opportunity possibility frontier, any nity possibility frontier. economics discussion of these philosophi- improvement in type 1 sets must imply a reduction in type 2 sets, and vice-versa. cal perspectives to illustrate this point. Points B, R, and E are translations to this *Atkinson and Stiglitz (1980) refer, in turn, to a "opportunity space"of the welfare concepts Equity and legal institutions figure 1 in Buchanan (1976). associated with Bentham, Rawls, and full **In Atkinson and Stiglitz (1980), utilities are The concerns with equity that feature in egalitarianism, respectively. used instead of opportunities.While that dis- tinction is fundamental in almost every moral, religious, and ethical debates around · If this society wished to maximize the respect, it is not for the point being made here, the world are reflected in real-world institu- sum of total opportunity indexes, it namely that different social justice criteria should aim for point B. imply different optimal allocations. tions, through which people have sought to An illustration of choices between the opportunities of two types of people promote justice throughout history. Chief among them are legal institutions, where "equity" has a distinct--and specific--inter- P O2 B pretation as a set of principles intended to B: Maximize sum of total opportunity guide and correct the application of the law. X (O1 + O2) R According to Kritzer (2002), how these prin- R: Maximize opportunities for group 1 (O1) ciples merge with the written, codified law E: Absolute equality of opportunities varies across legal traditions, but the over- between group 1 (O1) and group 2 (O1) arching concept of "fairness" is a cross- A cultural reality. And in practice, definitions E often refer directly to shared values within a particular community5 as well as to the belief that people should not suffer before the law Opportunity possibility frontier as a result of having unequal bargaining C O1 power.6 In Western philosophy, Aristotle is regarded as the first author to distinguish between justice and equity.7 He found that Equity and well-being 79 courts enact justice according to law--that is, not a purely Western concept--it can be by applying general rules that give an equi- found in legal systems around the world, table solution in the majority of cases.In some including those that do not share European cases, however, the results are inequitable. origins.11 For example, the distinction Equity then rectifies law in so far as the law is between justice and equity is also found in defective on account of its generality.8 The Islamic law, in which the former is referred to Romans operationalized this concept of as adala, the latter as insaf, and in Jewish law, equity by distinguishing between ius strictum with the distinction between din and tsedek.12 (strict law) and ius aequum (equity), with the In today's more integrated world, legal latter used to interpret the law and to comple- understanding of equity has also influenced ment it. Equity prevailed in instances of con- international law--serving as the basis for flict between the two. individualized justice, creating specific In modern legal traditions, equity re- principles of fairness and reasonableness, or mains a fundamental tenet of legal theory being identified with international equi- and practice. In common law systems, equity table standards for sharing resources and was historically a separate branch of law redistributing wealth. Perhaps the foremost administered by Chancery Courts.9 The example of the development of interna- Judicature Act of 1873 in the United King- tional principles of equity is the interna- dom "fused" the courts of law and equity, tional human rights regime. International doing away with a bifurcated system of human rights law is rooted in a commit- courts, while establishing the supremacy of ment to protect the "equal and inalienable equity in cases of conflict between equity and rights of all members of the human family," common law. Equitable principles, based on which itself is considered to be the "founda- conscience and fairness, have continued to tion of freedom, justice and peace in the develop and be applied in common law juris- world."13 dictions around the world to mitigate harsh The U.N. Charter laid the foundation for and unfair results produced by the applica- contemporary international human rights tion of formal legal rules in specific cases.10 law. The preamble to the Charter states that In general, the use of equity as a source the U.N. community "reaffirms faith in fun- of law in the civil law traditions of the Euro- damental human rights, in the dignity and pean continent is more limited than in the worth of the human person, in the equal common law tradition. Civil legal codes, rights of men and women and of nations which have their origins in the Enlighten- large and small."14 The Universal Declara- ment era, aim at integrating equity into tion of Human Rights, adopted by the Gen- formal law--that is, by designing laws eral Assembly of the United Nations on aimed at producing equitable results. December 10, 1948, is viewed as the "source Equity is seen as part of law and, therefore, of inspiration and . . . the basis for the U.N. should be achieved by applying the formal in making advances in standard setting as rules. Provisions in the codes that refer contained in the existing human rights explicitly to equity, however, are used to instruments."15 It has become a highly visi- correct inequitable results of the applica- ble and widely recognized statement of tion of other formal provisions, in a way moral, ethical, and political standards at the which is similar to common law systems. international level.16 Both the common law and the codified The contemporary international human systems from the continental European tra- rights regime comprises a broad array of dition have spread to countries around the legal instruments,17 many operating under globe, and equity is now a global legal con- the aegis of the United Nations. There also cept. The legal systems of Latin American exist regional human rights regimes in countries, such as Argentina, Brazil, and Europe (European Convention on Human Mexico, have approaches to equity similar Rights and Fundamental Freedoms), the to those in continental Europe, while Americas (Inter-American Convention on Bangladesh, India, and Nigeria follow the Human Rights), and Africa (African Char- common law tradition. Importantly, equity is ter on Human and Peoples' Rights). In 80 WORLD DEVELOPMENT REPORT 2006 addition, the laws of some other interna- First, some people behave in ways clearly tional entities, such as the European Union, inconsistent with the rational self-interest incorporate human rights norms (Treaty of hypothesis. According to Fehr and Fisch- Nice, Charter of Fundamental Rights of the bacher (2003), such people regularly display European Union). Together, these different a willingness to engage in two specific forms legal instruments are aimed at protecting of behavior: "altruistic rewarding, a propen- people against a variety of harms, including sity to reward others for cooperative, norm- potential harm by their governments, and abiding behavior, and altruistic punishment, at committing to the fundamental princi- a propensity to impose sanctions on others ples of equality and nondiscrimination. for norm violations" (785). These behaviors are observed in contexts in which it is possi- People prefer fairness ble to rule out individual motivation by a Different cultures and religions around the desire for reciprocity or a concern with rep- world may differ in important respects, but utation. While reciprocity and reputation they all share a concern with equity and fair- are important additional determinants of ness. This suggests something quite funda- cooperation in many settings, the experi- mental about the value human beings place mental evidence suggests that they are not on them. A fairly recent body of literature in the only factors that influence cooperative economics sheds some light on these shared behavior. human preferences. It has amassed convinc- A classic example is the Ultimatum ing evidence on individuals' preferences for Game, in which a player (the Proposer) is fairness, based on controlled laboratory asked to suggest a one-time division of a experiments. In these experiments, individ- certain sum of money (say, $100) between uals interact through behavioral games and himself or herself and another player. The play with real money under tightly con- second player (the Responder) has the trolled conditions. Results from such experi- power to simply accept or reject the offer. ments over the last 10 or so years reject the Acceptance leads to the implementation of hypothesis in standard economic models the offer, whereas rejection leads to a zero that all individuals are exclusively concerned payoff for both players. Monetary stakes are with their material self-interest. This new for keeps, and neither player knows the real body of literature is large and rich, but its identity of the other player. Both players are main findings can be summarized under told that they will never play with each three main points. other again. In such circumstances, standard game Figure 4.1 The distribution of observed offers in theory predicts a unique equilibrium: the ultimatum games Proposer should offer the smallest possible Offers and rejections in high- and low-stakes ultimatum games amount, and the Responder should accept (since a penny is higher than zero). But time Frequency and again, across hundreds of experiments 0.4 in highly heterogeneous cultural circum- Low monetary stake ($10 pie) stances and with amounts ranging from one 0.3 High monetary stake hour's to one week's local wages, observed ($100 pie) offers are substantially higher and, even so, rejections are often observed.18 In many 0.2 experiments, the modal (most frequent) Accept offer is actually at 50 percent. Figure 4.1 Reject Accept depicts the actual distribution of observed 0.1 Reject offers in two sets of Ultimatum Games, one with lower monetary stakes (bar on the left) 0.0 and one with higher (bar on the right). 0 1­10 11­20 21­30 31­40 41­60 51­60 61­70 Proportion of pie offered (%) Second, people are heterogeneous. A sizable Source: Based on data from Hoffman, McCabe, and Smith (1996). fraction of people in most experiments (20 Equity and well-being 81 to 50 percent) engage in altruistic giving or toward us that attract reward or punish- altruistic punishment--expending real ment, while others think it is their outcomes resources in a way that is unambiguously or opportunities.19 These studies do not costly to them, without any hope of eliciting usually distinguish explicitly between out- personal gain if the other person is ration- comes or opportunities. But it is possible to ally self-interested. But the behavior of oth- speculate that the aversion to very unequal ers (a majority) is consistent with rational payoff distributions in the Ultimatum Game self-interest. This is brought out quite arises from the arbitrary and unequal nature starkly in the Dictator Game, a variant of of the endowments (or power) implicit in the Ultimatum Game in which the respon- the initial allocation of the roles of Proposer der is purely passive. The second player is and Responder. simply a Receiver, with no right to reject the The experiments also show that people's offer. Positive offers in the Dictator Game views of fairness are complex and do not are observed, but they are both rarer and depend entirely on outcomes. Some players smaller on average than under the Ultima- are prepared to punish noncooperators until tum Game, in which the Responder may-- they receive less than other people, because always at a cost--punish the Proposer. of what they perceive to be the unfairness of These results point to the importance of their actions in the process of the game. This investigating more precisely the conditions is consistent with our emphasis that observed under which people exhibit self-interested distributions of certain outcomes--such as and other-regarding behavior. incomes--are the product of complex pro- cesses, and that the primary interest for Third, fair-minded people can behave self- those concerned with equity is not the out- ishly, and self-interested people can behave come, but the fairness of the processes they fairly. Behaviors depend on the rules of the participate in over their lifetimes. An game. In games where competitive pressures income distribution in which some people are introduced, mimicking a competitive are much richer than others because, given market, players tend to quickly converge similar chances, they have worked much toward actions consistent with self-inter- harder, may be regarded as fair. But the same ested behavior. An example is an Ultimatum income distribution may be regarded as Game with Multiple Proposers. If the unfair if it was generated by the richer group Responder can choose among various offers having access to much better schools or jobs, from different Proposers, with all noncho- solely because of the wealth or connections sen Proposers receiving zero, observed of their parents. behavior quickly tends to the Nash equilib- A separate but related point is made by rium. In the Nash equilibrium, all Proposers the social identity literature in social psy- offer the full amount--or very close to it-- chology (see Haslam 2001) and epidemiol- despite this giving rise to a very unequal dis- ogy (see Marmot 2004), which suggests that tribution, in which the Responder captures individual behavior and performance are the entire surplus, and all Proposers get zero. heavily conditioned by group identity (for In other settings, however--such as the example, caste, gender, occupation); by Repeated Public Good Game with Punish- whether those groups are seen as subordi- ment in Fehr and Gachter (2000)--even a nate to others (for example, doctors and small number of altruistic players can sus- patients, the status accorded minority eth- tain a cooperative equilibrium. nic communities); and by whether the These findings have been interpreted to boundaries among groups are regarded as suggest that a sizable fraction of human permeable (for example, the rules shaping beings in most societies care not only about whether and how employees get promo- their own individual opportunities and out- tions, immigrants become citizens, and so comes but also about"fairness."There is also on). Civil servants with low status and few broad agreement that fairness consists of a upward mobility prospects suffer from concern for others, although some authors higher mortality.20 Employees of low-status suggest that it is other people's intentions firms undergoing a merger more readily 82 WORLD DEVELOPMENT REPORT 2006 laboratory, it is plausible that large inequal- B O X 4 . 2 Capuchin monkeys don't like inequity either . . . ities in real life also reduce their well-being (particularly if the inequalities are not seen Research is under way on the roots of the equality test, but this rate rose to more to reflect only differences in effort or human altruism and the aversion to than 50 percent under the inequality test. inequity, whether cultural or genetic. But The refusal rate rose even further (to merit). This provides support for the statis- there is some evidence that aversion to more than 80 percent) under an alternative tical association that the subjective well- unfairness is not just human. In a recent arti- treatment, known as"effort control."In this being literature finds between income cle in Nature,"Monkeys Reject Unequal Pay," treatment, the first monkey received a inequality and self-reported happiness--a Brosnan and De Waal (2003) report the grape with no effort--with no need to pick results of exchange experiments with up a token and exchange it for food. subject we now turn to. brown capuchin monkeys (Cebus apella). Although few monkeys were used, these The animals were given a token that they differences were all statistically significant. Income inequality could immediately redeem for food by In both treatments,refusal rates and subjective well-being returning it to the experimenter.They were increased over time,as the experiment was placed in adjoining compartments with repeated (never more than once a day) many To what extent do the concerns with equity visual and vocal contact. times.Interestingly,only female monkeys shown in tightly controlled lab experiments In the baseline treatment (the equality completed these tests,as earlier experiments test), both specimens received a quarter of suggested that male capuchins are much also manifest themselves in the attitudes, a cucumber slice for each token exchanged. less sensitive to the distribution of rewards. feelings, and opinions of "regular people?" A In one treatment of interest (the The authors concluded that"tolerant recent study of labor strife in the United "inequality test"), the first monkey received species with well-developed food sharing States suggests that worker perceptions of a grape, while the second was given the and cooperation, such as capuchins [. . .] usual slice of cucumber.That capuchin may hold emotionally charged expectations whether they have been treated fairly or monkeys prefer grapes to cucumber slices about reward distribution and social unfairly can affect their efforts, and thus had apparently been amply established by exchange that lead them to dislike product quality, in important ways (box 4.3). previous research.The results were striking. inequity"(Brosnan and De Waal 2003, 299). Other studies investigate the associations Monkeys failed to exchange their tokens for food around 5 percent of the time under Source: Brosnan and De Waal (2003). between the narrower concept of income inequality and measures of subjective well- being. One recent study of European nations and the United States relies on indi- accept the new organizational structure vidual answers to the following question: because it benefits them individually, while "Taken all together, how would you say members of the higher-status firm are more things are these days--would you say that likely to resist change and act collectively in you are very happy, pretty happy, or not too terms of their premerger identity.21 happy?"22 Based on the variation in answers The experimental and subjective well- to this question, across European countries being literature in economics and social and U.S. states, and on objective income- psychology remind us that there is some- inequality measures, Alesina, Di Tella, and thing deep and fundamental about our taste MacCulloch (2004) find that "individuals for fairness and equity. Such "human altru- have a lower tendency to report themselves ism," argue Fehr and Fischbacher (2003) in happy when inequality is high, even after Nature, may be what accounts for the much controlling for individual income, a large greater complexity of cooperative patterns set of personal characteristics, and year and in human societies compared with those of country [. . .] dummies" (2009).23 other animals (box 4.2). Equity, it seems, One reason that inequality might make matters intrinsically and fundamentally for people less happy, even when controlling human beings. for absolute income levels, is that it violates Whatever the exact form of the true their sense of fairness. It may be that (at motives of individuals, the main implica- least some) people feel that a very unequal tion for this report from this large body of income distribution reflects unfair processes experimental evidence is that a good pro- and unequal distributions of opportunity. A portion of people in most societies appear 2001 study of Latin American countries by to dislike unfair outcomes and behaviors, so Latinobarometro, a reputable Chile-based much so that they are prepared to pay to opinion survey company, asked respondents punish those responsible for them. If people the following question: "Do you think that are prepared to pay real money to reduce the income distribution in your country is inequalities that appear unfair to them in a very fair, fair, unfair, or very unfair?" On Equity and well-being 83 B O X 4 . 3 Worker perceptions of unfairness, product quality, and consumer safety Do people change their behavior because they were related to 271 deaths and more than 800 time concessions were demanded and the old feel they are being treated unfairly, when they injuries.The most common source of failure of contract expired,which occurred before replace- are outside laboratories? And if so, is this likely the recalled tires was tread separation, a sudden ment workers were hired.Through early 1995, to have any serious consequence? A study of detachment of the rubber tread from the steel when many replacement workers were making industrial relations in Illinois, United States, sug- belts that caused the tire to blow out. tires,there were no excess claims in the Decatur gests that the answer to both questions is yes. Krueger and Mas (2004) compare the num- plant.It was not until the end of 1995,when many Since the 1940s,Firestone tire plants had ber of claims of defective tires in the Decatur returning strikers were working side by side with adhered to the industrywide bargain with labor plant with those from the two other North replacement workers,that the excess number of unions.But when negotiations for a new contract American plants where Firestone ATX tires were claims became high.On the basis of this and a began in 1994,Bridgestone/Firestone proposed produced: Joliette, Quebec, and Wilson, North much broader analysis,it appears likely that the deviating from the industrywide bargain in a way Carolina. Neither of these two plants chemistry between the recalled strikers and the that worsened the terms for labor at a time when experienced labor strife in the relevant period. replacement workers,or the cumulative impact of company profits were increasing.The company Tires produced during the labor dispute labor strife in general,created the conditions that proposed moving from an 8- to a 12-hour shift (1994­96) at Decatur had a much higher failure led to the production of many defective tires. that would rotate between days and nights.It rate than those produced at Joliette or Wilson, The authors of the study"recommend that also proposed cutting pay for new hires by 30 although before and after the dispute period the reader exercise caution in extending our percent.The union at the plant in Decatur,Illinois, the rate of claims was similar for tires manufac- results to other settings; our paper provides a called a strike and,shortly after the strike,the tured in Decatur and in the other plants.The detailed case study of only one firm in one company hired replacement workers. pattern suggests that changes in technology unique period of its history"(Krueger and Mas Labor strife at the Decatur plant closely cannot explain the rise in complaints against 2004, 257). At least in that instance, it appears coincided with lower product quality and defec- Decatur tires, because no such rise occurred in that perceptions of unfair treatment influenced tive tires. In August 2000, Firestone announced Joliette or Wilson. worker attitudes--and product quality and the the recall of 14 million ATX and AT tires, most of Nor does it appear that the lack of experience safety of consumers. them on the Ford Explorer.The U.S. government of replacement workers is to blame.There was a reported that Firestone tires under investigation spike in claims in the first half of 1994,around the Source: Krueger and Mas (2004). average, 89 percent of respondents re- based not exclusively on actual incomes, but garded the distribution in their countries also on processes, and that some differences to be either unfair or very unfair. In 17 of in outcomes may be considered fair (for the 18 countries surveyed, fewer than 20 example, because of differences in effort), percent of respondents answered fair or while others are not (for example, because of very fair.24 differences in opportunities). People are Such results may be particularly strong in clearly aware that income differentials can Latin America, which is one of most provide incentives for work and investment, unequal regions in the world, but they are including in education, if they are coupled not exclusive to that region.A recent analysis with opportunities for reward to those of several OECD countries (which are less actions. This comes across very clearly from unequal than most developing countries) the answers to one question in the latest wave was based on data from the International of the World Value Survey, which split Social Survey Program. To construct a proxy respondents around the world more or less measure of cross-national attitudes toward evenly into those who felt that income income inequality, Osberg and Smeeding inequality was too high and those who felt it (2004) ask what a number of different pro- was too low. fessions25 "should earn" and what they "do earn." They find that citizens of most high- Income inequality and incentives: income countries26 appear on average to What do people say? have similar attitudes toward inequality, The World Value Survey is a multicountry generally thinking that less well-paid profes- survey of individuals designed and spon- sions should be paid more and that better- sored by the Inter-university Consortium paid professions should be paid less. for Political and Social Research, based at Osberg and Smeeding (2004) findings the University of Michigan. The survey reinforce the view that the normative prefer- aims to "enable a cross-national compari- ences people have over distributions are son of values and norms on a wide variety 84 WORLD DEVELOPMENT REPORT 2006 Figure 4.2 Views on inequality from the World Values Survey want more of them (although this ten- dency was less pronounced in countries Income: more equal or more unequal? Views on inequality vary by income Mean score (lower values indicate with either very low or very high levels of Frequency (percent) preference for more equality) inequality). 20 7.0 The balance of the survey evidence sug- gests that, although inequality in incomes 15 seems to be associated with lower aggregate 6.5 levels of subjective well-being, there is con- 10 siderable heterogeneity in opinions about 6.0 whether it should be reduced or not. 5 Poorer people, and people in countries at 0 5.5 very high or very low inequality levels, 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 seem likelier to favor a reduction in in- Preference for inequality Income deciles equality. People recognize that some inequality is important to generate incen- Source: Inglehart and others (2004). Note: Author's calculations are based on data for the years 1999­2000. Preference for inequality ranges from agree- tives for investment and effort; however, ment with 1, "Income should be more equal," to agreement with 10, "We need larger income differences as incentives when asked about relative pay scales across for individual effort." professions, they would on average prefer smaller differentials. While in Latin Amer- of topics." Four main waves have been ica, for instance, a majority judges the fielded since the early 1980s. In the latest income distribution to be unfair, there is no wave, Inglehart and others 2004) asked rep- worldwide agreement that income dispari- resentative samples of people in 69 coun- ties should be reduced everywhere. This is tries to place their views on a scale from 1 to generally consistent with a view that what 10, where 1 implied agreement with the matters for ethical judgment is not income, statement that "Incomes should be made but fair processes and opportunities. more equal," and 10 implied agreement with the statement that "We need larger Income inequality and poverty income differences as incentives for indi- reduction vidual effort." Figure 4.2a suggests considerable polar- To the philosophical and legal arguments ization on views about inequality. The for equity, and to the survey-based and median answer is 6, suggesting no strong experimental evidence that fairness matters agreement with the two polar statements. intrinsically to people, we add a final argu- Yet almost 20 percent of all respondents ment: high levels of inequality make it more were in strong agreement with each of the difficult to reduce poverty. First, we high- two extreme views, represented by scores of light the fact that if inequality falls during a 1 and 10. Figure 4.2b shows a positive corre- growth spell, poverty generally falls by more lation between the score (which is nega- than it would have if growth had been dis- tively correlated with inequality aversion) tribution-neutral. Second, we document and a respondent's own income. This is the finding that the effectiveness of future consistent with the evidence on the impor- economic growth in reducing absolute tance of relative incomes for welfare: if you income poverty declines with initial income are richer, you are less inclined to favor a inequality. reduction in income inequalities than if you are poorer. If inequality falls, poverty falls more The World Values Survey results caution during spells of growth against any preconceived notion that By raising the incomes and consumption of income inequality is viewed everywhere as people across the distribution of income, inherently undesirable. When asked about economic growth is the main driver of income differences explicitly "as incentives poverty reduction in the developing world. for individual effort," (many) people seem The negative association between the aver- quite happy to have them and, indeed, to age annual rate of change in poverty and Equity and well-being 85 the average annual rate of growth in mean Figure 4.3 Growth is the key to poverty reduction . . . incomes is immediately clear from figure Change in log poverty headcount index (x100) 4.3, suggesting that countries experiencing 500 higher rates of economic growth can be BUL expected to reduce poverty much faster 400 LVA than those that grow more slowly.27 The 300 PER slope of the simple regression line, ­2.4, is MDA ARG the average total elasticity of poverty with 200 HUN MLI respect to economic growth. It implies that, POL CIV LAO 100 PAR YEM VEN SVN without controlling for any characteristics TTOCOL MNG ZWE BGD ROM SLV LSO BDI of the country, 1 percentage point growth in BWA 0 GEO NERHRVRUS 0ZAF VNMNICGHA ZMBNGA UGA UZB MDG ETH EGYLKA BRA BOL INDBFAPHLHON MEX a country's mean income can be expected to KEN TURPAN DZA LTU SEN MAU ECU GMB CMR ­100 URY ALB GUY CHN GTM reduce the incidence of poverty in that AZE CRI MAR EST IDN PAK country by about 2.4 percentage points. ­200 y = ­2.3841x + 2.3517 IRN TUN THA This powerful association between eco- t-stat = 9.30 JAM R2 = 0.5225 LYS nomic growth and poverty reduction is ­300 KAZ N = 73 one of the central stylized facts of develop- ­400 ment economics. Its qualitative nature has ­150 ­100 ­50 0 50 100 150 long been understood, and it has recently Change in log mean consumption or income between surveys (x100) been quantified by Ravallion and Chen Source: Authors' calculations. (1997), Dollar and Kraay (2002), and others. Indeed, the growth-poverty relationship is Figure 4.4 . . . and, on average, growth is distribution-neutral probably more powerful than surprising: it Change in log Gini index (x100) merely reflects the fact that, on average, 50 the growth in the incomes of the poor is 40 PAR similar to the growth of mean incomes LVA NER MLI CRI (figure 4.4). Put differently: aggregate eco- 30 NGA BUL POL ZWE BDI SVN EST LAO BWA nomic growth is, on average, distribution- 20 CHN HRV MDA ARGHUN BGD GHA neutral.28 LSO PHL MEX PER ECU COL NIC 10 CIV SLV EGY GMB There is, however, considerable variation MDG VNM AZEBRAIDN HON GTM GEO URYMAR PAN 0 IND LYS around those averages. About half the total ROM TTO RUS LTUALBUGACMR PAK THA VEN ZAF IRNTUR TUN BFA variation in poverty reduction is accounted ­10 MNG DZA ZMB JAM GUY YEM KAZ MAU for by economic growth (see the explana- BOL ­20 KEN y = ­0.0288x + 3.3717 SEN tory power of the underlying regression for ETH ­30 t-stat = 0.515 figure 4.3).29 The other half must reflect R2 = 0.004 changes in the underlying distribution of ­40 N = 73 relative incomes. This happens because the UZB ­50 LKA incidence of economic growth (its distribu- ­60 tional pattern) can vary dramatically ­150 ­100 ­50 0 50 100 150 across countries. Two countries with simi- Change in log mean consumption or income between surveys (x100) lar rates of growth in mean incomes can Source: Authors' calculations. have very different growth profiles across the population. As one would expect, reductions in inequality at a given growth mean incomes from the household survey rate add a "redistribution component" to was close to 2.5 percent. In Tunisia, where the "growth component," leading to faster the distribution of this growth was rela- overall poverty reduction. tively more beneficial to the poor, the head- The contribution of inequality reduc- count index of poverty fell by 67 percent tion alongside growth is illustrated by a (from 30 percent to 10 percent). This cor- comparison of the growth incidence curves responds to an annual rate of decline in (GIC) for Tunisia (1980­2000) and Senegal poverty of 5.4 percent. In Senegal, where (1994­2001) (figure 4.5). In both coun- growth was less pronounced for the bottom tries, the average annual growth rate in the half than for the upper half of the distribu- 86 WORLD DEVELOPMENT REPORT 2006 Figure 4.5 The national growth incidence curves for Tunisia 1980­1995 and Senegal 1994­2001 reduction in two countries with the same distribution-neutral growth rate may well Tunisia Senegal Annual growth rate, % Annual growth rate, % be different. Perhaps the most flexible way to capture 4 4 Growth incidence curve the variation in growth elasticity with 3 3 inequality across the sample of countries available for these exercises is simply to 2 Growth rate in mean 2 compute the total and the partial growth Growth rate in mean elasticity of poverty reduction for each sin- 1 1 Growth incidence curve gle country (in a single spell per country) 0 0 and to plot it against the initial Gini coeffi- cient (figure 4.6).30 A positive relationship is ­1 ­1 apparent for both partial and total elasticity 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 concepts, for all four poverty line/poverty Population percentile Population percentile measure combinations.31 The absolute value Sources: Ayadi and others (2004) for Tunisia and Azam and others (2005) for Senegal. These are two of 14 country case studies from the World Bank's "Operationalizing Pro-Poor Growth" Study. of the growth elasticity of poverty reduction falls as countries become more unequal, both for the total and for the partial con- tion, poverty fell by only 15 percent (from cepts. The slope of the line fitted through 68 percent to 57 percent), corresponding to panel (a) suggests that a 10 percentage an annual rate of poverty reduction of 2.3 point increase in the Gini coefficient is, on percent. Although some of this difference is average, associated with a decline of 1.4 in due to the fact that the actual growth rate the (absolute value) of the elasticity. Given was marginally higher in Tunisia (2.7 per- that the average elasticity is 2.53, this is not cent versus 2.3 percent in Senegal), much of a small effect. it is clearly due to the different patterns in The fact that very unequal countries the incidence of growth, which is evident in (with a Gini coefficient near 0.6) have a figure 4.5. total elasticity near zero in this sample This contribution of declines in should not be overemphasized. It is caused, inequality to poverty reduction holds more in part, by increases in inequality in some generally. According to Datt and Ravallion of these countries during the recorded (1992), a decomposition of changes in growth spells. This is evident from the fact poverty into growth and inequality compo- that the partial elasticity (which controls nents has been widely applied. Redistribu- for changes in distribution) does not reach tion components are usually smaller than zero for the same sample. Growth still con- growth components and, because inequal- tributes to poverty reduction, even in high- ity often rises, they often have the "wrong" inequality countries. The robust finding sign. But when inequality falls, this helps relates to the sign of the slope of the line, reduce poverty. not its exact intercepts: higher initial A second and separate point is that the inequality means that growth reduces power of growth to reduce poverty declines poverty by a lesser amount. with higher initial income inequality. A It has been argued that this is a mechan- reduction in inequality today therefore ical result in that, given a fixed functional also tends to have a future impact on the form for the income distribution, greater effectiveness of (even distribution-neutral) inequality results in slower poverty reduc- growth in reducing poverty. This occurs tion even if each individual's income grows because the shape of most income distri- at the same rate. Indeed, as indicated here, butions means that the growth elasticity distributional change is on average uncor- of poverty reduction tends to be smaller related with mean growth rates so that, on in more unequal countries. Put another average, the poor see their incomes grow at way, because the initial distributions of the same rate as other people's. That does income are different, the rate of poverty not, however, follow from any law of nature. Equity and well-being 87 Figure 4.6 Greater inequality reduces the power of growth to reduce poverty a. Total elasticity vs. Gini (headcount ratio, $1) b. Partial elasticity vs. Gini (headcount ratio, $1) Total elasticity Partial elasticity 10 10 DZA 5 PAR KEN 5 UKR ETH KGZ NGA NIC LSO RUS 0 UGAUZB RUS BWA HON GTM 0 CHNU CRI NGA ZWE BFA BWAHON GTM MAU UGAECUGMB PAN CHNU PAKCRIINDU PHL ECU MAU INDU THA MEX BFA GUY PAN BRA PAK PHL GUYSEN ROM IDN NERGEO TUR GMB INDR IDN BDIAZENER THA MDGMEXNIC BOL EST BOL PAR JAMTUN LSOKEN BRA MNG VNM AZE ZWE MDG CMR SEN ZAF BGD VNM ETH CMR MYS ALB TUN EST ROM GEO TUR UZB COL YEM LYS DZA IRN KGZ ZAF ­5 BDI JAM VEN MAR URY INDR MLIMAR ­5 ALB MNG KAZ EGY CIV COL ARG VEN MLI KAZ IRN BUL EGY CIV BGD LAO UKR YEM PER ­10 ­10 y = 13.758x ­ 8.2332 y = 8.2813x ­ 6.0475 t-stat = 2.45 t-stat = 2.97 ­15 ­15 R2 = 0.0911 R2 = 0.1231 LAO URY N = 62 N = 65 ­20 ­20 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 Gini Index, initial period Gini Index, initial period c. Total elasticity vs. Gini (squared poverty gap, $2) d. Partial elasticity vs. Gini (squared poverty gap, $2) Total elasticity Partial elasticity 11 11 8 8 RUS KEN 5 PAR 5 DZA ECU NIC LSO 2 NGA ETH KGZ 2 BDI VNM DZA ECU UKR UZB BGD KEN EGY PAKCRI UGA UGA MEXRUS BWA GTM CHNU CRI ETH HON ZAF GUY ZAF MAU MDG PAN ­1 CHNU MAU LAO PAK MEX GTM INDU PHL GMB NERNGAZWEPHLTHA GMBBFA COLBWA THA ­1 HON INDR TUN EST IDN MDGCMRMYSBFA BUL IDN INDU GUY SEN AZEMAR TUN BOL PAN BRA EST PAR JAM VEN LSO ALB MNG VNM YEM TUR AZE MLI CIV UZBCMRMYS BRA ­4 MNG ARG BGD COL SEN BOL INDR NIC NER ­4 UKR CIV ALB TUR KGZ ZWE KAZ YEM BDI JAM IRN VEN IRN GEO MAR PER ­7 LAO MLI ROM KAZ ­7 y = 15.039x ­ 8.7066 URY y = 6.5129x ­ 4.8068 GEO ­10 ­10 EGY ROM t-stat = 3.24 t-stat = 1.97 URY R2 = 0.1468 R2 = 0.0585 ­13 ­13 MDA N = 63 N = 65 ­16 ­16 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 Gini Index, initial period Gini Index, initial period Source: Authors' calculations. Note: The figure shows the scatter plots of country-level elasticities against initial-year Gini coefficients. Panel (a) shows the total elasticity for the headcount measure of poverty incidence, with a $1 per day poverty line. Panel (b) shows the partial elasticity for the same measure and the same line. Panels (c) and (d) also show the total and partial elasticities respectively, but now for the squared poverty gap index FGT(2) and with respect to a $2 per day line.32 Income distributions in individual coun- reduces the effectiveness of economic tries can and do change during spells of growth in reducing poverty. This means growth (see figure 4.5).33 There is no that, if all else remains the same, a reduction mechanical rule that states that the incomes in income inequality today has a double of the poor must grow at the same rate as dividend: it is likely to contribute to a con- the rest of the population.34 If on average temporaneous reduction in poverty, and it they do and if, given the shape of the empir- is likely to make future growth reduce ical income distributions, the poverty elas- poverty faster. ticities are lower in countries with higher Evidently, the caveat "if all else remains initial inequality, this is an empirical fact. the same" is of crucial importance. The dis- The balance of the evidence does not, tribution of incomes is a reflection of the therefore, allow much room for doubt that general equilibrium of an economy, based growth elasticities of poverty reduction are on the social, political, and institutional stronger in more equal societies. Inequality structures that condition its behavior. 88 WORLD DEVELOPMENT REPORT 2006 Simple-minded attempts to change the way cies are likely to lead to faster poverty incomes are distributed, without taking reduction in the future, for any amount of into account the effects of policies on the growth that the economy generates. incentives of all agents in the economy, are It turns out, however, that some inequali- bound to fail. We return to the issue of ties--not necessarily those of incomes--are appropriate policy design in part III of this also detrimental to economic growth itself. report. All that can be said about the results Such inequalities in power, assets, and access here is that, if policies exist that can lead to a to markets and services are most likely to be less unequal distribution of resources with- the ones on which policy can productively out major costs to the (static and dynamic) focus. The next two chapters turn to a dis- efficiency of resource allocation, such poli- cussion of these "inefficient inequalities." Inequality and investment 5 In a world in which markets worked per- One of the great advances in development fectly, investment decisions would have lit- economics in the past 15 years is the accre- c h a p t e r tle to do with the income, wealth, or social tion of a substantial body of evidence on status of the decision maker. They would be documenting how well (or badly) asset and determined by the returns an investment financial markets work in developing coun- promises and by the market price of capital, tries. The fact that these markets rarely meas- adjusted for the extra risk it entails. If peo- ure up to their ideal creates the possibility ple had good investment opportunities, it that wealth and social status, defined as one's really would not matter whether they had position in society both in ascriptive identity the money--they could always borrow and in connections, will have an important what they needed, and if the risk bothered influence on investment decisions. It seems them, they could always sell shares in their natural to start with this evidence. business and buy safer assets with the money from the sale. Markets, wealth, status, However, for various reasons--mainly and investment behavior economic but also political--markets are not perfect. If borrowers can willfully default The market for credit on their loans, lenders prefer to make loans In a perfect credit market, there is a single to borrowers who can provide collateral interest rate and everyone can borrow or assets. Private returns for politically con- lend as much as they want at that rate. That nected firms can be higher than for those individuals can borrow as much as they want without such connections, and so these firms at the current rate explains the presumption may attract more capital, even though social of a separation between the wealth or status returns may not be any greater.1 Members of of the investors and the amount they invest. groups subject to discrimination may ration- Whether they are rich or poor, well-con- ally invest less in their human capital than nected or just off the streets, an extra dollar they would in the absence of such explicit or of investment will be profitable for them subtle stereotypes. only if the return they get from it is more After we give up the idea that markets than the interest rate. If the interest rate is work anywhere close to perfectly, the scope higher, they would be better off lending that for a direct link between investment and money if it was their own, or borrowing less the distribution of wealth or power widens if it were someone else's. So, two people with substantially, in many instances leading to the same return on investment would end up underinvestment by those who have good investing the same amount.4 growth opportunities.2 Correcting the How close are real markets to this ideal market failures directly is often not feasi- market? Chambhar is a market town in ble, and in these cases certain redistribu- Sindh (Pakistan), on the east bank of the tions of wealth, power, and resources can Indus. In 1980­81 farmers from the area serve as second-best alternatives.3 In other around Chambhar got most of their credit words, interventions to enhance equity can from about 60 professional moneylenders. improve efficiency. Based on detailed data from 14 of these 89 90 WORLD DEVELOPMENT REPORT 2006 lenders and 60 of their clients, Aleem The same pattern--high and variable bor- (1990) calculated the average borrowing rowing rates, much lower deposit rates, and interest rate charged as 78.5 percent. But if low default rates--shows up in the "Sum- these farmers wanted to lend their money, mary Report on Informal Credit Markets in the banking system would pay them only India."5 This report summarizes results from about 10 percent. It is possible, however, case studies commissioned by the Asian that they may not have been depositing in Development Bank and carried out under the banks. An alternative measure of the the National Institute of Public Finance and deposit rate that is relevant for these farm- Policy. ers is the opportunity cost of capital to the For the urban sector, the data are based moneylenders, 32.5 percent. In either case, on various case surveys of specific classes of it suggests a gap of at least 45 percentage informal lenders. For the broad class of points between the borrowing and lending nonbank financial intermediaries called rates. finance corporations, the maximum deposit The borrowing rate also varied enor- rate for loans of less than one year is 12 per- mously across borrowers. The standard devi- cent. These corporations offer advances for ation of the interest rate was 38.1 percent, one year or less at rates that vary from 48 compared with an average lending rate of percent per year to the utterly astronomical 78.5 percent. In other words, an interest rate 5 percent per day. The rates on loans of of 2 percent and an interest rate of 150 per- more than one year varied between 24 per- cent are both within two standard deviations cent and 48 percent. Default, once again, is of the mean. One possibility is that these dif- only a small part of the story: default costs ferences in interest rates reflect differences in explain only 4 percent of total interest costs. the default rate: perhaps the expected repay- For hire-purchase companies in Delhi, the ment was the same for everybody, because deposit rate was 14 percent and the lending those who paid higher rates were more likely rate was at least 28 percent and could be as to default. Also the expected repayment high as 41 percent. Default costs were 3 per- could have been equal to the actual interest cent of total interest costs. rate paid to the depositors, if the default rate For the rural sector, interest rates are was high enough. But default was rare: for high, but they are also variable (figure individual lenders, the median default rate 5.1). This finding is based on surveys of was between 1.5 percent and 2 percent, with six villages in Kerala and Tamil Nadu, car- a maximum of 10 percent. ried out by the Centre for Development Studies, Trivandrum. The rich (with Rs 100,000 or more in assets) get most of the Figure 5.1 In rural Kerala and Tamil Nadu, the rich access most of the credit and pay relatively low credit (nearly 60 percent) and pay a rela- rates tively low rate (33 percent), while those with assets between Rs 20,000 and Rs 125 Average interest rate (%) [left scale] 5000 30,000 pay rates of 104 percent and get Average loan size (Rs) [right scale] only 8 percent of the credit. The average 100 Cumulative 4000 interest rate charged by professional mon- proportion of credit eylenders (who provide 45.6 percent of 75 [left scale] 3000 the credit) is about 52 percent. While the average deposit rate is not 55 2000 reported, the maximum from all the case studies is 24 percent, and in four of them it is no more than 14 percent. In the category of 25 1000 professional moneylenders, about half the loans were at 60 percent or more, but 0 0 0­ 5,000­ 10,000­ 15,000­ 20,000­ 30,000­ 50,000­ 100,000 another 40 percent or so had rates below 36 5,000 10,000 15,000 20,000 30,000 50,000 100,000 and above percent. Default rates were higher than in the Asset group (Rs) urban sector, but they still cannot explain Source: Dasgupta, Nayar, and Associates (1989) more than 23 percent of the interest costs. Inequality and investment 91 The fact that credit access depends on relative to what would happen if markets social status is also shown by Fafchamps' functioned properly. The capital released (2000) study of informal trade credit in because they underinvest is absorbed by the Kenya and Zimbabwe. It reports an average non-poor, who may actually end up overin- monthly interest rate of just over 2.5 per- vesting relative to how they would invest in cent (corresponding to an annualized rate perfect markets. This is the reason: because of 34 percent), but it also notes that the rate the poor cannot borrow, the non-poor can- for the dominant trading group (Indians in not lend as much as they would like to (this Kenya, whites in Zimbabwe) is 2.5 percent a is why deposit rates in developing countries month, while the blacks pay 5 percent a are often very low). And because the non- month in both countries.6 Chapter 9 also poor cannot lend, it makes sense for them to provides evidence that in many countries keep investing in their own firms, even when "insiders" effectively lobby to limit access to the returns are low. financial institutions and that lending is Because the poor underinvest, and be- skewed toward the rich, consistent with the cause the opportunity cost of capital to the evidence in figure 5.1. non-poor is thus lower than it would other- None of these facts is surprising. Contract wise be, the composition of the investors enforcement in developing countries is often also changes. In particular, firms that would difficult, and it is not easy to get courts to not be viable if markets functioned per- punish recalcitrant borrowers.7 As a result, fectly (for example, because the interest rate lenders often spend at lot to make sure that would be too high) can survive and even their loans get repaid: it is plausible that expand because markets are the way they these are the resources that drive a wedge are. In other words, the "wrong" firms end between the borrowing rate and the lending up investing. rate. Indeed, Aleem (1990) shows that the resources spent by lenders to monitor bor- The market for insurance rowers explain the nearly 50 percentage The ideal insurance market is one in which point gap between the lending and borrow- people bear no avoidable risks. In a setting in ing rates in his data. It is easy to imagine that which a single village constitutes a separate borrowers who are easier to monitor will insurance market closed to the rest of the enjoy better rates, which would explain why world (so that only people in the village can lending rates vary so much. insure other people in the village, in some These imperfections in credit markets kind of mutual insurance arrangement), have immediate implications for the rela- individual consumption should respond tionship between wealth and investment. only to aggregate (village-level) income fluc- First, with the rate of interest on deposits tuations and not to fluctuations in the much lower than that on loans, the oppor- income of specific individuals. Put in blunter tunity cost of capital for those who just terms, as long as aggregate consumption is want to invest their own money is much unchanged, individual income fluctuations lower than the opportunity cost for those should not translate into fluctuations in who have to borrow. This means that the individual consumption. When insurance wealthy will end up investing much more markets work well, risk considerations than the indigent, even if they face exactly should not have a significant impact on the the same returns on their investment. Sec- choices people make, irrespective of their ond, the lower interest rates charged to rich wealth, given that what an individual does people reinforce this conclusion, because has little impact on aggregate uncertainty. the rich then face a lower opportunity cost While a perfect insurance market is more when they too are borrowing. Third, in complex than a perfect credit market, and some cases, those who are unable to provide thus harder to detect, there have been collateral will have no access to credit at any attempts to test the prediction about the interest rate. irrelevance of fluctuations in one's own We would thus expect the poor to under- income. The Côte d'Ivoire Living Standards invest, certainly relative to the rich, but also Measurement Surveys from 1985 to 1987 92 WORLD DEVELOPMENT REPORT 2006 provide panel data on the income and con- be much more effective than others in pro- sumption of nearly 800 households, with viding insurance to their residents. each household tracked for two consecutive Townsend describes in detail how insurance years (1985 and 1986 or 1986 and 1987). In arrangements differ across villages. While in table 5.1, the relationship between changes one village there is a web of well-function- in consumption and changes in incomes is ing, risk-sharing institutions, the situations reported separately for the three main in other villages are different. In one village, regions and separately for 1985­86 and the institutions exist but are dysfunctional; 1986­87. The first row of the first block for in another, they are nonexistent; in a third, each year reports the basic correlation close to the roads, there seems to be no risk- between income and consumption: a fall in sharing whatsoever, even within families.10 income always hurts consumption, although As for credit, the failure of insurance the coefficient varies between a low of 0.15 could have something to do with informa- (a $1 reduction in income means that con- tional asymmetries. It is not easy to insure sumption goes down by $0.15) to a high of someone against a shock that he alone 0.46. The next row does the same thing, but observes, because he has every incentive to now there is a village dummy intended to always claim that things had gone badly. But pick up any village-level changes in con- as Duflo and Udry (2004) demonstrate, sumption. Remarkably, the coefficients on spouses in Côte d'Ivoire do not seem to be own income, which under perfect insurance willing to insure each other fully against should have fallen to zero after controlling rainfall shocks that affect them differen- for village-level changes, barely budge.8 tially. Because rainfall obviously is observ- Not all the evidence is quite so pes- able, at least part of the problem has to be simistic. Townsend (1994) used detailed elsewhere. One possibility is limited com- household-level data from four villages, mitment. People may be happy to claim which were intensively studied by the Inter- what was promised to them when it is their national Crop Research Institute in the turn to be paid, and then default when the Semi-Arid Tropics (ICRISAT) in India, to time comes for them to pay. This may be see whether the full insurance hypothesis is particularly easy in a setting in which the consistent with the data. He found that social relations between the sets of people while the data did reject the exact predic- who are insuring each other are not partic- tion, it did not miss by very much. In other ularly close, perhaps explaining why words, his evidence suggested that villagers Townsend found no insurance in the village do insure each other to a considerable closest to the road. extent: movements in individual consump- Lack of insurance should have an effect tion in his data seem largely uncorrelated on the pattern of investment. That many with movements in income. insurable risks are uninsured means that Later work by Townsend, based on data one cannot invest without personally bear- he collected in Thailand, turned out to be ing a significant part of the concomitant less encouraging.9 Some villages seemed to risk. Indeed, big corporations able to sell Table 5.1 The effect of income shocks on consumption, Côte d'Ivoire West Forest East Forest Savannah All Rural OLS 1985­6 No dummies 0.290 (6.2) 0.153 (3.2) 0.368 (5.8) 0.259 (8.8) Village dummies 0.265 (5.7) 0.155 (3.5) 0.373 (5.7) 0.223 (7.7) OLS 1986­7 No dummies 0.458 (8.8) 0.162 (5.3) 0.168 (4.0) 0.239 (10.4) Village dummies 0.424 (8.1) 0.173 (5.6) 0.164 (3.8) 0.235 (10.1) Source: Adapted from Deaton (1997), table 6.5, 381. Note: Absolute value of t-statistics are shown in brackets. The first row of each panel shows the coefficient on income change of a regres- sion of consumption changes on income changes. The second row reports the same result when village dummies are included in the regression. OLS = Ordinary Least Squares. Inequality and investment 93 their equity in organized equity markets lates maximum contract terms of 3 years may be the only players who can really hope for traditional technologies and 15 years for to diversify away a large part of the risk of a modern technologies. particular project. Given this fact and the It is often unclear who has the right to reasonable assumption that the poor are sell a particular plot of land, when no single more risk-averse than the rich, we are likely person or family has a clear, undisputed, to be in a perverse situation in which the legal title to the land. This ambiguity reflects poor may also find it hardest to reduce their encroachments and land grabs in the evolu- exposure to risk. Thus, they are likely to shy tion of land rights, as well as the importance away from riskier and higher-return invest- of custom in governing land relations, espe- ments, reinforcing the prediction that the cially in Africa. The recent popularity of poor invest too little. land titling as a social intervention is a direct consequence. The market for land Where lease contracts exist, they are not In a perfect land market, individuals can always of the fixed-rent type, at least when buy or lease as much land as they want for the land is used for cultivation. Many coun- as long as they want at a price that depends tries, including the United States, have a only on the quality of the land (and the long tradition of an alternative contractual length of the lease). The lease should be at a form: sharecropping. Under sharecropping, fixed rent, so that the lessor is the residual the farmer gets only a fraction of the pro- claimant on the produce of the land. That duce, but he does not need to pay a fixed land can be freely bought and sold ensures rent. As Alfred Marshall pointed out more that there is no particular advantage or dis- than one hundred years ago, this weakens advantage to owning land compared with incentives and reduces the productivity of any other asset of similar value. That the the land, but the near universality of share- lessor is a residual claimant means that the cropping suggests that it is a response to a land is put to optimal use. Not so, however, real need. There is some disagreement in practice. among economists about the exact nature Many developing (and some developed) of that need.11 It is plausible, however, that countries have regulations about who can the need is related to the fact that farmers buy land and how much or how little they are often poor, and making them pay the can buy. Binswanger, Deininger, and Feder full rent when their crop does poorly is dif- (1995) argue that almost every developing ficult and probably not desirable. country today has gone through a phase Leaseholds in developing countries tend when it had regulations intended to con- to be short-lived. The norm is either a year centrate landownership. By contrast, Besley or a season. Longer leases are not unknown, and Burgess (2000) provide a list of regula- but they are rare. This might reflect the fact tions from different states in India, each of that custom, rather than law, secures most of which is an attempt to limit the concentra- these leases: perhaps it is too much to rely on tion of ownership in land. custom to enforce leases of arbitrary length. Governments also directly limit trans- The imperfect salability of land can, of actions in land, with the ostensible aim of course, hurt anyone who owns it. But the preventing the accumulation of land in the rural poor probably have more of their hands of a few people. In Ethiopia in the wealth in land than most people, so making late 1990s, Deininger and others (2003) land nonsalable might be particularly harsh note that selling and mortgaging land were on them. against the law. While rentals were offi- What tends to discourage investment in cially allowed (after being disallowed for the land is the lack of an explicit title, or the two decades), local leaders and govern- insecurity of tenure more generally (caused, ments were free to restrict even these for example, by the short duration of leases rental transactions in land. For example, and the possibility that the landlord might the Oromia region allowed farmers to rent threaten to take the land away at the end of only 50 percent of their holding and stipu- the lease). It clearly helps if land is owned 94 WORLD DEVELOPMENT REPORT 2006 by the person contemplating the invest- may be more a product of culture and tradi- ment. That most who work in agriculture tion than of the cold calculation of benefits. tend to be too poor to buy out the land they Benefits are relevant, but the responsiveness are cultivating is thus a potential source of to them may not be as large as one might underinvestment. have expected. In the market for human capital, the The market for human capital reward should be based entirely on the One thing makes the market for human human capital supplied, not on other attrib- capital different from all the other asset utes of the person supplying the skills. Dis- markets: many decisions about investing in crimination based on gender, caste, religion, human capital are made by parents (or or race obviously violates this, but so does a other family members) for their children. In system of job allocation based on contacts. other words, those making the decisions are Until very recently, job discrimination based different from those who receive the human on gender was the norm all over the world, capital. It is not hard to imagine why this and the number of countries where such separation might introduce important dis- discrimination is still either legally or tortions to the functioning of this market. socially accepted is dwindling but signifi- Gary Becker's classic formulation avoids cant. Even where such discrimination is this issue by assuming that the family can explicitly frowned on, there is some evi- borrow against the child's future income, dence of continuing discrimination. The turning the problem into a conventional same is true of race, caste, and religion. Most investment decision. Under that assump- discrimination--unless legally mandated tion, the amount invested will not depend through affirmative action in favor of a his- on the family's means. torically disadvantaged group, such as low In practice, however, although human castes in India and African Americans in the capital is an asset, it cannot be legally United States--flies in the face of explicit pledged or mortgaged, for the simple rea- laws against it. son that pledging your human capital One reason discrimination is so hard to would be tantamount to selling yourself eliminate comes from its sheer insidious- into slavery.12 This obviously constrains ness. Beliefs about differences are embedded people's ability to borrow money to finance in everyday attitudes and practices in a way investments in their education. that neither the discriminator nor the dis- When parents cannot borrow against criminated against may be conscious of, their children's future income--true most of even though these beliefs transform how the time in most developing countries-- they both behave. This is what underlies the they may still hope that those children will power of the stereotype. In a telling exam- take care of them in their old age. The hope ple, Stone, Perry, and Darley (1997) asked all might be that the children do grow up to participants in a recent experiment (Ameri- reap the benefits of their parent's investment can Caucasians, hereafter referred to as and that they will pay their parents back. But whites) to listen to the same running children know that they have no legal obli- account of an athlete's basketball perform- gation to do so. If they do repay their par- ance on the radio. Half the participants were ents, it is because they love their parents or led to believe that the target player was because society expects them to do so. white, half that he was African American. Investments in human capital may thus The results indicated that information was be driven as much by parents' sense of what less likely to be absorbed if it was discordant is the right thing to do, as by any calculation with the prevailing U.S. stereotypes that of costs and benefits. Once we accept this, it whites are more academically talented than becomes clear that children's human capital African Americans, and that African Ameri- may not be very different from any other cans are more athletically gifted. The white consumption good--so richer families will target player was perceived as exhibiting less tend to invest more in their children's health natural athletic ability but more "court and education. And human capital decisions smarts." The African-American target player Inequality and investment 95 was perceived as exhibiting fewer court erate self-reinforcing behavior. If members smarts but more natural athletic ability. of the discriminated group invest less in Such biases have also been documented their own education, or in searching for in real-world settings. A recent study of the employment, others might use this under- effect of stereotyping on judgment finds investment to confirm their prejudice that prison inmates with more Afrocentric against that group. features receive harsher sentences than Stereotypes can be self-fulfilling not only those with less Afrocentric features, con- because they influence perceptions of the trolling for race and criminal history.13 target of the stereotype, but also because Bertrand and Mullanaithan (2003) show they influence the behavior of the individu- evidence from a field experiment proving als who are stereotyped. Stone and others beyond reasonable doubt that there is a (1999) asked college undergraduate volun- high degree of African-American discrimi- teers to play a miniature golf course. Perfor- nation in the United States. They sent the mance was measured by how many strokes same resumes to a large number of compa- were needed to put the ball in the hole: nies under either a stereotypically white fewer strokes meant better performance. name or a stereotypical African-American The variable that the experimenters manip- name, and found a 50 percent higher call- ulated was the description of the task. In back rate when the name was white. The one treatment, the task was described as a data say that having a white name is worth "standardized test of natural athletic abil- as much as eight additional years of job ity," in the other as a "standardized test of experience. Moreover, the discrimination sports intelligence." When the task was tended to be greater when the resume cor- described as a test of natural athletic ability, responded to someone who was better edu- the African-American participants per- cated, suggesting that investment in human formed better than the whites: they aver- capital among African Americans probably aged 23.1 strokes to complete the 10-hole is significantly underrewarded. golf course, compared with 27.8 for the A very different form of discrimination whites. But when the task was described as a comes from the allocation of jobs based on test of sports intelligence, the race gap was contacts. Munshi (2003) presents persua- reversed: African Americans averaged 27.2 sive evidence that contacts are very impor- strokes, whites 23.3. tant in the allocation of jobs for migrant One way to interpret this behavior is that labor in the United States. The employment social ideas--stereotypes about the talents prospects for Mexican migrants there, it of different social groups--impose bounds turns out, are much better when they are from within. Under the rational, self-inter- from areas where there was an earlier out- est hypothesis, individuals change their flow of migrants. Quite remarkably, it helps behavior only when their preferences or if migrants are from an area where there external constraints change. But the behav- was a drought several years ago, which ior of real individuals depends as well on pushed out a cohort of migrants to the belief systems that society impresses on United States. These migrants then help the them. Negative stereotypes create anxiety later generations of migrants from that area that may interfere with performance: that is to find jobs. This is the clincher: it does not why the psychologist Claude Steele termed help to be from an area where there was a this kind of behavior "stereotype threat."14 recent drought. The beliefs underlying the stereotypes, if The perception of discrimination, con- deeply internalized, can affect early deci- scious or not, can affect investments in sions about prospective careers, and atti- human capital. Those who expect to be dis- tudes toward society, by changing what criminated against in a particular labor Appadurai (2004) calls a person's "capacity market--rightly or wrongly, consciously or to aspire." The reader may recall the exam- otherwise--will tend to invest less in ple (from chapter 2) of the Batwa girl who acquiring the type of human capital that the wanted to be a cleaner upon completing market rewards. This could, perversely, gen- school. Positive stereotypes, by contrast, can 96 WORLD DEVELOPMENT REPORT 2006 boost self-confidence and lead individuals formance of the low caste, regardless of to expend greater effort. whether the payment scheme was piece rate Stereotypes influence behavior twice-- (that is, participants were paid 1 rupee per through their impact on individuals' self- maze solved) or tournament (that is, the confidence, and through their impact on the participant who solved the most mazes was way individuals expect to be treated. To paid 6 rupees per maze solved, while the examine the effect of stereotypes on the abil- other participants received nothing). When ity of individuals to respond to economic caste was announced, the low-caste children incentives, Hoff and Pandey (2004) under- solved 25 percent fewer mazes on average in took experiments with low- and high-caste the piece-rate treatments, compared with children in rural north India. The caste sys- the performance of subjects when caste was tem in India can be described as a highly not announced. When caste was announced stratified social hierarchy in which groups of and groups were composed of six children individuals are invested with different social drawn from only the low caste (a pattern of status and social meaning. segregation that for the low caste implicitly In the first experiment, groups composed evokes their traditional outcast status), the of three low-caste ("untouchable") and decline in low-caste performance was even three high-caste junior high school students greater. While we cannot be sure from these were asked to solve mazes and were paid data what the children were thinking, some based on the number of mazes they solved. combination of loss of self-confidence and In one condition, no personal information expectation of prejudicial treatment likely about the participants was announced. In a explains the result. second condition, caste was announced with The expectation by the low-caste sub- each participant's name and village. In a jects of prejudicial treatment may be third condition, participants were segre- rational given the discrimination in their gated by caste and then each participant's villages. But the discrimination itself may name, village, and caste were announced in not be fully rational. Cognitive limitations the six-person group. may prevent others from judging stigma- When caste was not announced, there tized individuals fairly. That people are was no caste gap in performance (figure bounded in their ability to process informa- 5.2). But increasing the salience of caste led tion creates broad scope for belief sys- to a significant decline in the average per- tems--in which some social groups are viewed as innately inferior to others--to Figure 5.2 Children's performance differs when their influence economic behavior. If such beliefs caste is made public persist, it will generally be rational for those Average number of mazes solved, by caste, discriminated against to underinvest (with in five experimental treatments respect to others) in the accumulation of 8 skills for which the return is likely to be Piece rate Tournament lower for them. This rational calculation is 6 additional to any reduction in their "capac- ity to aspire," arising from the internaliza- tion of those beliefs. 4 The evidence on 2 High caste underinvestment Low caste Highly imperfect markets suggest consider- 0 able scope for underinvestment. Caste Caste Caste Caste Caste not announced not announced announced announced announced and Industry and trade segregated Direct estimates of marginal products show Source: Hoff and Pandey (2004). that there are many unexploited investment Note: A vertical line in the figure indicates that the caste gaps are statistically significant. opportunities. For small Mexican firms Inequality and investment 97 with less than $200 invested, the rate of Figure 5.3 Returns to capital vary with firm size: return reaches 15 percent per month, well evidence from small Mexican firms above the informal interest rates available in Monthly returns (%) pawn shops or through microcredit pro- 30 grams (on the order of 3 percent a month) (figure 5.3).15 Estimated rates of return 20 decline with investment, but the rates remain high--7 percent to 10 percent a month for firms with investments between 10 $200 and $500, and 5 percent for firms with investments between $500 and $1,000. All 0 these firms are thus too small and could reap large gains from increased investment. 0 100 200 300 400 500 600 700 800 900 1000 Trade credit is an important form of Nonrented capital (US$) credit everywhere, perhaps especially where Source: McKenzie and Woodruff (2003). the formal institutions of the credit market are underdeveloped. Fisman (2001a) looked at the relation between access to trade credit includes small industry. In January 1998, and capacity utilization for 545 firms in the limit on total investment in plants and Côte d'Ivoire, Kenya, Tanzania, Zambia, and machinery for a firm to be eligible for inclu- Zimbabwe. He finds that firms that receive sion in the small industry category was trade credit from three main suppliers (on raised from Rs 6.5 million to Rs 30 million. average, about one of the three suppliers The researchers first show that, after the provides trade credit) have 10 percent better reforms, newly eligible firms (those with capacity utilization than firms that receive investment between Rs 6.5 million and Rs no trade credit. The relation is much 30 million) received, on average, larger stronger in industries in which it is impor- increments in their working capital limit tant to carry large inventories. than smaller firms. They then show that the Such studies present serious method- sales and profits increased faster for these ological issues, however. The basic problem firms during the same period. Putting these comes from the fact that investment levels two facts together, researchers can estimate are likely to be correlated with omitted vari- the impact of the increased access to work- ables. For example, in a world without credit ing capital on the growth in profits. Allow- constraints, investment will be positively ing for the possibility that the firms in the correlated with the expected returns to priority sector were paying less than the investment, generating a positive "ability true cost of capital for the extra money bias."16 McKenzie and Woodruff (2003) from the bank, they estimate that the attempt to control for managerial ability by returns to capital in these firms must be at including the firm owner's wage in previous least 94 percent. employment. This goes only part of the way, A different kind of evidence for underin- however, if individuals choose to enter self- vestment comes from the fact that many employment precisely because their expected people pay the high interest rates reported productivity in self-employment is much earlier. Given that this money typically goes higher than their productivity in an into financing trade and industry, the pre- employed job. Conversely, if capital is allo- sumption is that the people borrowing at cated to firms to avoid their failure, there these rates of often 50 percent or more must could be a negative ability bias. have a marginal product of capital that is Banerjee and Duflo (2004a) take advan- even higher. But the average marginal prod- tage of a change in the definition of the uct in developing countries seems to be "priority sector" in India to circumvent nowhere near 50 percent. One way to get at these difficulties. All banks in India are the average of the marginal products is to required to lend at least 40 percent of their look at the incremental capital-output ratio net credit to the "priority sector," which (ICOR) for the country as a whole.17 For 98 WORLD DEVELOPMENT REPORT 2006 Figure 5.4 Inefficient allocation of resources; the example of the Gounders vs. the outsiders twice as much capital as the outsiders on average. Gounder firms of all ages own Gounder firms of all ages own more capital Gounders lose early lead in sales around fifth year more capital, although there is a strong ten- log capital stock (100,000 Rs) log sales value (100,000 Rs) dency toward convergence as the firms age 3.5 7.0 (figure 5.4a). The Gounders, despite own- 6.5 ing more capital, lose their early lead in sales by about the fifth year, and end up sell- 3.0 6.0 Gounders ing less (figure 5.4b). In other words, out- 5.5 siders invest less and produce more. They Gounders are clearly more able than the Gounders,19 2.5 5.0 but because they are less cash-rich and do 4.5 not have the right connections, they end up Outsiders Outsiders working with less capital. 2.0 4.0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Agriculture Years of experience Years of experience There is also direct evidence of high rates of Source: Banerjee and Munshi (2004). returns on productive investment in agricul- Figure 5.5 Average returns for the late 1990s the International Monetary ture. In the forest-savannah in Southern switching to pineapples as an Fund (IMF) estimates the ICOR to be more Ghana, cocoa cultivation, receding for many intercrop can exceed 1,200 percent than 4.5 for India and 3.7 for Uganda. The years because of the swollen shoot disease, Per hectare profits (1,000 Cedis) implied upper bound on the average mar- has been replaced by a cassava-maize inter- 20,000 ginal product is 22 percent for India and 27 crop. Recently, pineapple cultivation for Pineapple profits 15,000 percent in Uganda. export to Europe offered a new opportunity 10,000 That many firms in India have a mar- for farmers in this area. In 1997 and 1998 Non-pineapple profits ginal product of 50 percent or more, while more than 200 households cultivating 1,070 5,000 the average marginal product is only 22 per- plots in four clusters in this area were sur- 0 cent or so, is strong prima facie evidence for veyed every six weeks for about two years. ­5,000 the misallocation of capital. The firms with Pineapple production dominates the tradi- 0 20 40 60 80 100 the marginal product of 50 percent and tional intercrop (figure 5.5),20 and the aver- Cumulative percent of plots more are clearly too small, while other firms age returns associated with switching from Source: Goldstein and Udry (1999). (the ones who bring the average down to 22 the traditional maize and cassava intercrops percent) must, in some sense, be too large. to pineapple is estimated to be in excess of A specific example of this kind of misal- 1,200 percent! Yet only 190 out of 1,070 location of capital comes from a study of plots were used for pineapple. When the the knitted garment industry in the south- authors asked farmers why they were not ern Indian town of Tirupur.18 Two groups farming pineapple, the virtually unanimous of people operate in Tirupur: Gounders response was, "I don't have the money,"21 and outsiders. The Gounders, who issue although some heterogeneity in ability from a small, wealthy, agricultural commu- between those who have switched to pine- nity from the area around Tirupur, moved apple and those who have not, cannot be into the readymade garment industry entirely ruled out. because there were not many investment Evidence from experimental farms sug- opportunities in agriculture. Outsiders gests that, in Africa, the rate of returns to from various regions and communities using chemical fertilizer (for maize) would started joining the city in the 1990s. also be high. But the evidence may not be The Gounders, unsurprisingly, have realistic if the ideal conditions of an experi- much stronger ties in the local community, mental farm cannot be reproduced on and thus better access to local finance. But actual farms. Foster and Rosenzweig (1995) they may be expected to have less natural show, for example, that the returns to ability for garment manufacturing than the switching to high-yielding varieties were outsiders, who came to Tirupur precisely actually low in the early years of the green because of its reputation as a center for gar- revolution in India, and the returns were ment export. The Gounders own about even negative for farmers without an educa- Inequality and investment 99 tion. This, despite the fact that these vari- Table 5.2 Farm size productivity differences, selected countries eties had been selected precisely for having Farm size Northeast Brazil Punjab, Pakistan Muda, Malaysia high yields, in proper conditions. But they Small farm 563 274 148 required complementary inputs in the cor- (hectares) (10.0­49.9) (5.1­10.1) (0.7­1.0) rect quantities and timing. If farmers were Largest farm 100 100 100 not able or did not know how to supply (hectares) (500+) (20+) (5.7­11.3) them, the rates of returns were actually low. Source: Berry and Cline (1979). Chemical fertilizer, however, is not a new Note: 100 = land productivity in the largest farm size. technology, and the proper way to use it is well understood. To estimate the rates of Figure 5.6 Profit-wealth ratios are highest for the returns to using fertilizer on farms in smallest farms Kenya, Duflo, Kremer, and Robinson Profit/wealth ratio (2004), in collaboration with a small non- 0.4 governmental organization (NGO), set up 20th percentile small randomized trials on people's farms. 0.3 Each farmer in the trial delimited two small plots. On one randomly selected plot, a field 40th percentile officer from the NGO helped the farmer 0.2 apply fertilizer. Other than that, the farmers continued to farm as usual. The rates of 0.1 60th percentile 80th percentile return from using a small amount of fertil- izer varied from 169 percent to 500 percent, 0 depending on the year, although marginal 10 12 14 16 18 20 22 24 returns declined quickly with the quantity Monsoon onset, standard deviation (weeks) of fertilizer used on a plot of a given size. Source: Rosenzweig and Binswansger (1993). Evidence for a different type of underin- Note: The standard deviation of the date of monsoon onset is a measure of underlying risk. The onset date of the monsoon was the vestment in agriculture is the negative size- single most powerful of eight rainfall characteristics to explain productivity relationship, the idea that the gross farm output. The data come from the Indian ICRISAT villages. smallest farms tend to be the most produc- tive (table 5.2). The gap in the productivity of small and large farms within a country reasonable measure of land quality, which, can be enormous: a factor of 5.6 in Brazil however, is not entirely clear. There are also and a factor of 2.75 in Pakistan.22 It is residual doubts about whether the returns smaller in Malaysia (1.5), but a large farm in are well measured--it is possible that the Malaysia is not very large. This is strong land of the smaller farms is degrading faster, prima facie evidence that markets are some- but the degradation is not being counted how not allocating the right amount of land while calculating the returns. to those who currently farm the smaller For these same firms, when risk goes up, plots. the average return goes down. In part this The problem with this kind of evidence is may be inevitable, but it may also reflect the that it ignores the many reasons why the big- fact that the lack of insurance encourages ger farm may be inherently less productive, people to avoid risky (but remunerative) for example, lower soil quality. Even so, sim- choices.23 This is consistent with the fact that ilar (but somewhat less dramatic) results profitability falls faster for the poorer farm- show up even after controlling for differ- ers (less able to self-insure) as the risk goes ences in land quality. The profit-wealth ratio up. Specifically, a one-standard-deviation in- in Indian ICRISAT villages is the highest for crease in the coefficient of variation of rain- the smallest farms, and when risk is compar- fall leads to a 35 percent reduction in the atively low, the gap is more than 3:1 (figure profit of poor farmers, a 15 percent reduc- 5.6). Because wealth includes the value of tion in the profit of median farmers, and no the land, the measure implicitly takes into reduction in the profit of rich farmers. The account differences in the quality of the study also finds that input choices are land. It does so as long as land prices are a affected by variability in rainfall, and in par- 100 WORLD DEVELOPMENT REPORT 2006 ticular, poor farmers make less efficient input and Rosenzweig (1986) and Shaban (1987) choices in a risky environment. both show that productivity is 30 percent In related work, Morduch (1993) specifi- lower in sharecropped plots, controlling for cally investigated how the anticipation of a farmers' fixed effects (that is, comparing the credit constraint affects the decision to productivity of owner-cultivated and farmed invest in high-yielding variety seeds. Using land for farmers who cultivate both their a methodology inspired by Zeldes (1989), own land and that of others) and for land he splits the sample into two groups--one characteristics. Shaban (1987) shows that group of landholders expected to have the all the inputs are lower on sharecropped ability to smooth their consumption, and land, including short-term investments one group that owns little land, expected to (fertilizer and seeds). He also finds system- be constrained. He finds that the more con- atic differences in land quality (owner- strained group devotes a considerably cultivated land has a higher price per smaller fraction of land to high-yielding hectare), which could in part reflect long- variety seeds for rice and castor. term investment. Another consequence of the lack of On the impact of security of property, insurance is that it may lead households to Do and Iyer (2003) find that a land reform use productive assets as buffer stocks and that gave farmers the right to sell, transfer, or consumption smoothing devices, which inherit their land-use rights also increased would be a cause for inefficient investment. agricultural investment, particularly the Rosenzweig and Wolpin (1993) argue that planting of multiyear crops (such as coffee). bullocks (an essential productive asset in Laffont and Matoussi (1995) use data from agriculture) serve this purpose in rural Tunisia to show that a shift from sharecrop- India. They show, using ICRISAT data cov- ping to owner cultivation raised output by ering three villages in semiarid areas in 33 percent, and moving from a short-term India, that bullocks, which constitute a large tenancy contract to a longer-term contract part of households' liquid wealth (50 per- increased output by 27.5 percent.25 cent for the poorest farmers), are bought Security of property rights is often and sold quite frequently (86 percent of linked to the local power structure. The households had either bought or sold a bul- connection between inequalities in power lock in the previous year). Moreover, they and underinvestment is nicely exemplified buy when they are flush with money and by the Goldstein and Udry (2002) study of sell when they are broke. investment in land in a setting where land is Since people are not simultaneously sell- allocated by custom (rural Ghana). They ing and buying land, they are not selling show that individuals are less likely to leave these animals because they no longer need their land fallow (an investment in long- them for production. Indeed, from the view run productivity of the land) if they do not point of production, most of these farmers hold a position of power within either the should own two bullocks and never sell hierarchy of the village or the hierarchy of them. If they are selling, the reason is that the lineage. The problem is that the land they need the money for consumption. The gets taken away from them when it is lying data suggest that, for poor or midsize farm- fallow. Because women rarely hold these ers, there is considerable underinvestment positions, women's land is not left fallow in bullocks, presumably because of the bor- enough and is much less productive than rowing constraints and the inability to bor- men's. row and accumulate financial assets to smooth consumption: almost half the Human capital households in any given year hold no bul- According to the report of the Commission locks (most of the others own exactly on Macroeconomics and Health (2001), two).24 returns to investing in health are on the There is also compelling evidence that order of 500 percent. But these numbers, sharecroppers lack incentives. Binswanger arrived at through cross-country growth Inequality and investment 101 regressions, are not as easy to interpret as Measured returns to private investment in what would actually happen if someone education tend not to be quite so high. were to invest an extra dollar in health. That Banerjee and Duflo (2004b) survey cross- said, there clearly are examples of specific country evidence, and conclude that-- health interventions that have enormous Using the preferred data, the Mincerian rates private and social returns. There is substan- of returns seem to vary little across countries: tial experimental evidence that supplemen- the mean rate of returns is 8.96, with a stan- tation in iron and vitamin A increases pro- dard deviation of 2.2. The maximum rate of ductivity at relatively low cost. returns to education (Pakistan) is 15.4 per- · Basta, Soekirman, and Scrimshaw (1979) cent, and the minimum is 2.7 percent (Italy).29 study iron supplementation among rub- ber tree tappers in Indonesia. Baseline But most of the educational benefits of health measures indicated that 45 per- deworming mentioned above would be cent of the study population was anemic. captured by a child whose parents are The intervention combined an iron sup- willing to spend $0.50 on the deworming medicine. This clearly offers a return plement and an incentive (given to both much higher than the measured Mincer- treatment and control groups) to take ian returns at affordable absolute cost, the pill on time. Work productivity although they are not strictly compara- among those who got the treatment ble. Deworming does not require the increased by 20 percent (or $132 a year), child to spend more years in school, but at a cost per worker-year of $0.50. Even it does help the child get more out of the taking into account the cost of the incen- years he or she is already spending in tive ($11 a year), the intervention sug- school. However, when the deworming gests extremely high rates of returns. medicine was offered free to the chil- dren, the take-up was only 57 percent. In · Thomas and others (2005) obtain lower this sense, it is clear that at least some but still high estimates in a larger experi- causes of underinvestment have to be ment, also in Indonesia. They found that found in the way the family makes deci- iron supplementation experiments in sions, rather than in the lack of Indonesia reduced anemia, increased the resources. probability of participating in the labor market, and increased earnings of self- The fact that a lack of connections alters employed workers. They estimate that, the nature of human capital investment is for self-employed males, the benefits of nicely demonstrated in a recent paper by iron supplementation amount to $40 per Munshi and Rosenzweig (forthcoming). year, at a cost of $6 per year.26 They show that, in India, trade liberaliza- · The cost-benefit analysis of a deworming tion increased returns to knowing the Eng- program27 in Kenya reports estimates of lish language in families with connections a similar order of magnitude. Taking in the blue-collar sector compared with into account externalities (because of the families with no connections. However, contagious nature of worms), the pro- there is a much bigger gap between girls and gram led to an average increase in boys in the increase in enrollment in Eng- schooling of 0.14 years. Using a reason- lish-medium schools. This is attributed to able figure for the returns to a year of the fact that girls never really expected to education, this additional schooling will get these blue-collar jobs, while for their lead to a benefit of $30 over the life of the brothers, it depended on whether they had child, at a cost of $0.49 per child per year. the right contacts. Not all interventions have the same rates Inequalities and investment of return, however. A study of Chinese cotton mill workers28 led to a significant Four important points follow from this increase in fitness, but no corresponding body of evidence: first, markets in develop- increase in productivity. ing countries are highly imperfect, and 102 WORLD DEVELOPMENT REPORT 2006 those who do not have enough wealth or might depend on the set of economic social status tend to underinvest. The opportunities available. resources underused because of this under- The microcredit community, in particu- investment end up being used for some less lar, has long debated this last issue in trying productive purpose, reducing overall pro- to decide whether microcredit is best instru- ductivity. In the example from the knitted ment for helping the poorest of the poor. garment industry in Tirupur, the Gounders This clearly turns partly on whether the were overinvesting in their own relatively poorest are the ones who have the projects unproductive firms, while the much more with the highest returns, which could be the productive firms of the outsiders were case if the poor and the less poor have the starved of capital. The land owned by same kinds of production functions, and if Ghanaian women was getting degraded, there are diminishing returns to scale. If, because they did not have the social status instead, the most productive technology in needed to hold on to the land during the this area had a fixed cost of production but fallowing period. This, once again, is a pure (say) diminishing returns otherwise, giving loss for society. The fact that other people the poorest access to more capital may not who do have status and can fallow their be very productive: even with all the capital land as needed is not, in any way, compen- they can get, they may not be able to cover sating for the loss of productivity on the the fixed cost. It may be more effective to lands of the powerless. This creates a strong help people who are slightly richer, because presumption that certain specific types of with some help they may actually be able to redistribution, by empowering certain peo- start a business. ple or increasing their access to resources How good or bad is the assumption of or contacts, can promote efficiency and decreasing returns in the production func- equity. tion of an individual firm? As mentioned Second, this hypothesis would imply a bias above, McKenzie and Woodruff (2003) esti- in favor of those kinds of redistribution that mate a production function for small Mexi- target the specific lack of access to resources or can firms, suggesting strong diminishing influence causing the inefficiency. In some sit- returns. Mesnard and Ravallion (2004) find uations this will mean redistributing assets, weak diminishing returns using Tunisian but it also might mean redistributing access to data. But estimating a production function capital, perhaps by promoting microcredit, that exhibits local increasing returns is strengthening women's land rights or access inherently difficult. A firm is likely to grow to jobs and welfare programs, designing (or shrink) quickly when it is in the region affirmative action programs to break down of increasing returns. So we will observe few stereotyping, and improving access to justice firms in this region, and be likely to reject systems. too often the assumption of local increasing Third, because investments build wealth returns. Certainly the natural interpretation and wealth makes it easier to invest in a of the results in Banerjee and Duflo (2004a), world where markets do not function very showing close to 100 percent returns in well, a little help can go a long way. Starting medium firms in India, is that there are the right business might be the biggest chal- increasing returns over some range. lenge: once started, the business might pro- A corollary of this discussion is that the pel itself forward without any further help. redistribution that maximizes productivity Fourth, it is not clear that the beneficiar- growth is not necessarily the one that has ies from this kind of efficiency-promoting the strongest immediate effect on poverty. redistribution have to be the poorest of the Nor is it the one that does most to reduce poor. Because the ideal is to promote pro- inequality. Indeed, except under very spe- ductive investments, the target should be cial circumstances, this discussion tells us those most likely to make these invest- nothing about the relation between some ments. Whether the poorest are the right global measure of inequality and the effi- people from this point of view is an empiri- ciency of resource use or investment. Con- cal question, and one for which the answer sider the case, discussed above, in which the Inequality and investment 103 production function has a fixed cost but inequality and changes in growth, including also diminishing returns. If all firms are several studies that do the analysis at the sub- equal and the maximum they can each national level within the same country, find a invest is less than the fixed cost, no one will positive effect. be able to start a firm. Increasing inequality Both Banerjee and Duflo (2003) and will raise the productivity of capital by Voitchovsky (2004) conclude that there is no making it possible for some firms to pay the reason to give one of these sets of results pri- fixed cost. Because there are also diminish- ority over the other. Indeed, both could be ing returns, however, there will be a point at right. For example, in the short run, policies which any further increase in overall that allow large cuts in real wages might inequality would be counterproductive. encourage investment, but in the long run, More generally, the effect of inequality the consequent increase in poverty might will depend on the shape of the production make it harder for the population to maintain function, and the size of the investment its human capital. Or both could be wrong. potential of the average person relative to Most important among the many reasons for the fixed cost. Obviously, the issue gets even both the cross-sectional and the time series more complicated if different firms have dif- evidence to be misleading are the following: ferent production functions and if produc- the possibility of a nonlinear relationship tivity is correlated with the owner's wealth between inequality and growth, problems (as it might be if the owner's education is an with comparability of cross-country data, important input into production and richer and the difficult question of identifying the people tend to be more educated). direction of causality when both variables are Several authors have tried to look for a likely to influence one another. systematic relation in cross-country data This lack of clear-cut results is perhaps dis- between inequality and growth (presumably appointing, but it is worth emphasizing that what investment is meant to achieve). A our focus here has been on redressing spe- lengthy body of literature30 estimated a long- cific inequalities in productive opportunities run equation, with growth between 1990 and rather than some overall measure of inequal- 1960, for example, regressed on income in ity. Despite the great attention devoted to the 1960, a set of control variables, and inequal- question of a systematic relationship between ity in 1960. Estimating these equations overall inequality and growth at the country tended to generate negative coefficients for level, the body of evidence remains uncon- inequality. But there are obvious concerns vincing. But there clearly are situations in about whether such a relation could be which there is a strong presumption that driven entirely by omitted variables. To reducing a specific inequality would promote address this problem, Li and Zou (1998), better investment. Forbes (2000), and others used the time One such example comes from Opera- series dimension of the Deininger and Squire tion Barga, a tenancy reform in the Indian data set to look (effectively) at the effect of state of West Bengal in the late 1970s and short-run changes in inequality on changes 1980s. It has been known, at least since the in growth.31 The results change rather dra- work of the great Victorian economist matically: the coefficient of inequality in this Alfred Marshall, that sharecropping pro- specification is positive and significant. vides poor incentives and discourages A recent review paper by Voitchovsky effort. In such an environment, a govern- (2004) concludes that both these effects are ment intervention that forces the landlords quite robust. Most studies that look at the to give their sharecroppers a higher share of cross-sectional relationship between inequal- the output than the market would give ity and subsequent growth over a relatively them should increase effort and productiv- long period in cross-country data, and espe- ity. This is exactly what happened in West cially those that use measures of asset Bengal, India, when a Left Front govern- inequality, find a negative relationship, often ment came to power in 1977. The tenant's significant.32 By contrast, most studies that share of output was set at a minimum of 75 look at the relationship between changes in percent as long as the tenant provided all 104 WORLD DEVELOPMENT REPORT 2006 inputs. In addition, the tenant was guaran- money goes to the grandfather. (Boys are teed a large measure of security of tenure, essentially unaffected.) The estimates sug- which may have encouraged him or her to gest that receipt of the pension (which was undertake more long-term investments on about twice the per capita income among the land. Survey evidence shows a substan- blacks) was enough to help girls bridge half tial increase in both the security of tenure the gap in height-for-age between South and the share of output going to the share- African and American children. cropper. The fact that the implementation These examples show that it is possible to of this reform was bureaucratically driven, enhance both equity and efficiency simulta- and proceeded at different speeds in differ- neously. Judicious redistribution--of income ent areas, suggests the possibility of using to grandmothers, of power to poor women variation in the implementation of the farmers, of credit to entrepreneurs in small reform to evaluate its impact. The evidence firms--can increase the productivity of suggests that there was a 62 percent increase resources, such as land, human capital, and in the productivity of the land.33 physical capital. If markets fail, resources do A different program, also promoting not always flow to where their return is great- equity and efficiency, had to do with redress- est, particularly if that happens to be in proj- ing the effects of intrafamily inequality. A ects run by people with limited wealth or long line of research claims that income and influence. Careful microeconomic case study expenditures are often controlled by the evidence, some of which was summarized in male members of the family and that this this chapter, suggests that certain forms of leads to underinvestment, especially in the redistribution can reduce waste and con- health and education of girls. One fallout of tribute to a better use of resources, while also dismantling the apartheid regime in South reducing inequality of opportunity. In fact, it Africa was the expansion of the South enhances efficiency precisely because it African social pension program to the black reduces inequality of opportunity. population. Pension entitlements would This is not to say that one cannot easily accrue to elderly males and females, and imagine certain types of redistribution that many older women living alone were entitled hurt efficiency. But given the near universal- to receive the benefit. In many cases, children ity of market failures and underinvestment of very poor parents were sent to live with in poor countries, it should be possible, with grandparents who began to receive these a combination of good research and careful pensions. Duflo (2003) compared the impact thinking, to identify opportunities for redi- of these new transfers on the nutrition of recting resources to poorer people who are children living with their grandparents, sepa- in a position to make good use of them. rately for households in which the pension In making the case for improvements was given to the grandmother and those in in equity that are also efficiency-enhanc- which it was assigned to a grandfather. ing, this chapter used mainly micro- For children born before the expansion, economic evidence on markets, wealth, in 1990 and 1991, height-for-age was slightly and agency of individuals. The next chap- lower in families in which the grandmother ter uses a different set of historical, macro- would eventually get the pension. For chil- economic, and institutional evidence to dren born after the expansion, in 1992 and argue that complex historical processes, 1993, the children are significantly taller combined with inequalities in influence and (except for the newborns) in those families. power, may lead to bad political and eco- There is no difference between noneligible nomic institutions, which severely impair families and families in which pension the development of poor countries. f o c u s 3 o n Spain Equity and development in the Spanish transition to democracy In the last half century, Spain has gone from authoritarianism and underdevelopment to democracy and wealth. Spain's history illustrates how the distribution of political agency and economic assets greatly influences the policy choices avail- able to a society. The fundamentals of economic and political structure influence and constrain the choices. But the process is not deterministic: political agency and policies can shift the underlying fundamentals (as happened in Spain in the 1960s and 1970s) and open the space for new choices. Before the civil war: new elections were held, with a victory for contracts. Emphasis on permanent jobs and social and economic polarization the left. The threat of more radical policies cheap housing was seen as a substitute for Until the second half of the twentieth cen- prompted a military uprising, supported by the lack of direct social policies, an attempt tury, Spanish contemporary history was a the landed classes, much of the bourgeoisie, by the regime to win legitimacy. tale of political and economic failure. After and the church. Spain became engulfed in a In the late 1950s, Spain eventually a period of territorial expansion and Euro- three-year civil war. The social polarization moved to break with this interventionist pean hegemony in the early modern ages, locked Spain into a zero-sum gain over the system. An acute political crisis--associated Spain lapsed into economic decline and distribution of wealth. There was little with a wave of strikes, an economic re- cultural stagnation in the following cen- political space for compromise or reformist cession, and severe balance-of-payment turies. During most of the nineteenth cen- solutions. The "haves" opposed all attempts crisis--led the government to adopt a stabi- tury, its industrial takeoff was blocked by at even minimalist reform. The "have-nots" lization plan in March 1959. In addition to political instability, inefficient legal institu- wanted radical change, not gradual reform. fiscal and monetary restraint, the plan tions, substantial inequalities, and a poorly included wide-ranging measures to liberal- educated population. In 1929, per capita The Franco regime: ize the economy. It was an outright success. income was $3,000 (in 1990 dollars)-- from autarky to growth From 1960 to the outbreak of the first oil two-fifths that of Britain and less than two- With the defeat of the Republican govern- crisis, output expanded at an average annual thirds that of France. ment by the Nationalist army in 1939, Gen- rate of more than 7 percent with very little Spain was polarized by entrenched social eralissimo Francisco Franco established an interyear volatility. Per capita income and economic inequalities. In a country still authoritarian regime that lasted until his almost tripled from about $3,000 (in 1990 eminently agrarian, the distribution of land death in 1975. The destruction caused by the dollars) to $8,500 in 15 years. Productivity was very unequal. About 1 percent of the civil war depressed the Spanish economy. Per growth averaged 6 percent. holdings occupied 50 percent of the land. capita income fell to its 1900 level and did The transformation of the Spanish econ- Educational attainment remained low, not reach its 1918 level until 1950. The pro- omy led to significant structural changes in strictly linked to circumstances of birth. portion of the active population in industry Spanish society. The combination of eco- Social mobility was almost nonexistent. declined to 22 percent in 1940 (the level in nomic growth, industrial expansion and Except for Catalonia and the Basque coun- 1920) and the share of employed in agricul- internal migration produced a substantial try, which industrialized in the nineteenth ture rose above 50 percent. Growth averaged decline in the levels of interregional century, Spain lacked a large middle class. only 1.2 percent a year in the 1940s. inequality (from a standard deviation in per Against this backdrop of relative stagna- Spain's economic recovery was ham- capita income of 0.37 in 1955 to 0.27 in tion and high inequality, democratic insti- pered, above all, by the autarkic and statist 1973). Interhousehold inequality also tutions were introduced in 1931--only the policies of the Franco regime. Inspired by declined considerably: the Gini coefficient second time in Spanish history. They did the corporatist ideologies of Italian Fascism for wages and salaries of employees (agrar- not last long. The brief democratic period and German Nazism, Franco's regime gen- ian and industrial) declined from 0.29 in (1931­36) was characterized by huge politi- eralized a system of price controls and 1964 to 0.23 in 1973; the Gini coefficient for cal instability and social agitation. The first rationing and regulated foreign trade household income fell from 0.39 in 1964 to Republican government pushed ahead with through quantitative controls. This inter- 0.36 in 1974. The income share of the three a strong reformist program: separation of ventionist strategy extended to the labor central deciles went up from about 51 per- church and state; a single system of state and housing market. To quell one of the cent to 59 percent in that decade. schools and a goal to universalize educa- main forces that opposed the military insur- Still, significant social and economic tion; a process of land reform; a law to rection, Franco outlawed any independent inequities remained. Although the illiteracy decentralize political power to Catalonia; labor unions. Instead, workers and employ- rate had fallen to 10 percent by 1970, only 6 and stepped up efforts to reform the army. ers had to affiliate in a national trade union percent of the population had completed These reforms elicited a strong reaction organization. This repressive stance was secondary studies. Wages remained damp- from the right, which came to power in "compensated" by strict labor legislation ened by repressive labor institutions. Taxa- 1933 and quickly moved to halt them. Two that made it hard for employers to dismiss tion and public spending were low, and and a half years later, in the spring of 1936, workers or to hire them through temporary redistributive social programs nonexistent. 106 WORLD DEVELOPMENT REPORT 2006 Transitioning to democracy the 1960s and 1970s. Rapid industrializa- secondary education--10 times more and building the welfare state tion and urbanization deflated past con- than in the mid-1970s. An ambitious pub- Following Franco's death in 1975, King flicts around the distribution of land. The lic investment program tripled the public Juan Carlos became the Spanish head of expansion of literacy and the increase in highway network, revamped and expanded state. He immediately launched a process of productivity and incomes generated a large metropolitan transportation, and modern- political change. Employing the legal mech- middle class. Sustained growth defused ized the railroad system. anisms put in place by the very technocratic social conflict with the credible promise of Spain's transition to democracy and the generation that had reformed the economy higher incomes and more social mobility. resulting expansion of its welfare state in the early 1960s, as well as pointing to In short, Spain had overcome the zero-sum shows how a mutually reinforcing package wide popular support for democracy, he game it had been locked in for the past cen- of policy and institutional choices leading secured the consent of the old Francoist tury and a half. to greater equity helped underpin the Cortes to establish a truly democratic par- Economic growth resulted in a different development and modernization of the liament elected through direct, competitive economic structure and better distribu- Spanish economy and its integration into elections. tional outcomes, supporting a swift and the European Union. It illustrates how The political reform was ratified with successful transition to democracy. In turn, political and economic structures shape the overwhelming popular support in a ref- the transition to democracy changed the possibilities for policy choice, a theme of erendum in December 1976. Although role and size of the public sector. chapter 6. But it also illustrates that specific conducted in a climate of uncertainty, par- Democratization reinforced social de- policy choices matter--across social sec- ticularly over the reaction of the army and mands for progressive and redistributive tors, infrastructure, the workings of mar- the extent to which terrorist violence or policies--especially for public infrastruc- kets, and international integration--and labor mobilization could disrupt the negoti- ture, and education, health, and social that there can be important complementar- ations, democratic elections were held in programs. In 1979 more than 70 percent ities for both equity and dynamic growth, June 1977. After protracted negotiations, a of Spaniards agreed with the statement notably between greater social provisioning new constitution was approved in 1978 with that "the distribution of wealth in this and greater reliance on markets. This takes the support of all parliamentary groups. To country is totally unjust." In 10 years, us to the issue of practical policy design, the reinforce the political pact in parliament, social expenditure almost doubled to central theme of part III of this report. the government also struck a wide eco- reach 80 percent of the European average. nomic and social deal with employers and Public expenditure in education steadily trade unions that same year. increased from 2 percent of GDP in 1975 Spain's democratization was rooted in to 4.5 percent in 1995. By 2001 almost 50 Sources: Synthesized from Boix (2005), with references to Gunther, Montero, and Botella (2004); North and Thomas the new economic and social conditions of percent of the population had completed (1973); and Revenga (1991). Equity, institutions, and the development process 6 Product, land, labor, and capital markets are such as Haiti or Barbados in the eighteenth crucial for the allocation of resources and century. Even though property rights in land c h a p t e r development. Market institutions, however, and people were well defined and even well exist and function in the context of a whole enforced (although subject to potential slave set of nonmarket and political institutions. rebellions), most people had no property The nature of these other institutions--and rights and were thus subject to expropriation the way they function--are influenced by by others, particularly their masters. For 95 inequalities in the political and social realm. percent of society, there were no incentives to The most obvious of these other institu- engage in socially desirable activities. A simi- tions are those that define and enforce prop- lar, although somewhat less extreme, example erty rights and contracts. People will not of inequitable institutions is South Africa invest if property rights are not well defined under apartheid. Institutions there were good and enforced, or if they believe that the con- for the whites but left 80 percent of the popu- tracts they write will not be honored or that lation without incentives or opportunities to courts of law will not be fair. The state must engage in economically productive activities. also provide a whole set of other inputs apart from social order and fair contract enforce- The distribution of power ment. These include various types of public services and regulations. Lying behind well- and institutional quality: functioning markets are legal systems, judges, circles vicious and virtuous policemen, and, ultimately, social groups and How do societies develop equitable non- politicians. market institutions? First, there must be This chapter considers the circumstances sufficient political equality--equality in and processes for creating institutions that access to the political system and in the dis- promote prosperity. These circumstances are tribution of political power, political rights, closely related to the concerns of this report. and influence. In essence, societies that create institutions to Poor institutions will emerge and persist generate sustained prosperity are equitable in in societies when power is concentrated in important ways. Because talent and ideas are the hands of a narrow group or an elite. Such widely distributed in the population, it is cru- an elite may grant property rights to itself,but cial that the property of all people is secure the property rights of most citizens will be and that there is equality before the law for unstable. There may be equality before the all, not just for some. Predetermined circum- law for a particular elite group, but not for the stances should not constrain anyone's inno- majority of people. Government policies may vation or investment opportunities. This favor such an elite, granting them rents and implies that a good institutional environment monopolies, but most people will be will not block entry into new lines of business excluded from entering profitable lines of and that the political system will provide business. The education system may invest access to services and public goods for all. heavily in the children of such elites, but most Institutions must be equitable. will be excluded. To take an extreme example, institutions Many things determine the distribution of were severely inequitable in slave societies, political power in society--the constitution, 107 108 WORLD DEVELOPMENT REPORT 2006 the nature of checks and balances, and the ous today are so because they have developed ability of different groups to solve collective more egalitarian distributions of political action problems. But economic inequality power, while poor societies often suffer from often underpins political inequality. In a soci- unbalanced distributions. We also consider ety with large inequalities of assets and how some societies made the transition from incomes, the rich will tend to have more one equilibrium to the other. influence and an advantage in adapting and Because institutions have distributional distorting institutions to their benefit. effects, conflict arises naturally. One set of Because the distribution of power, institutions will benefit some people, while through its impact on institutions, helps to another will benefit different people. Thus, determine the distribution of income, the there will be incentives for people to control possibility of vicious and virtuous circles is power to create or keep the institutions that clear. A society with greater equality of con- benefit them and to avoid or weaken the trol over assets and incomes will tend to institutions that disadvantage them. If the have a more equal distribution of political groups in conflict are defined along ascriptive power. It will therefore tend to have institu- lines, such as ethnicity, then this may induce a tions that generate equality of opportunity more severe form of conflict than when for the broad mass of citizens. This will tend groups are defined along other lines, or when to spread rewards and incomes widely, there are cross-cutting cleavages. More polar- thereby reinforcing the initial distribution of ized conflict seems to be an independent incomes. In contrast, a society with greater force leading to bad institutions that can help inequality of assets and incomes will tend to to explain the relatively weak performance in have a less egalitarian distribution of power some societies (discussed below in a compar- and worse institutions, which tend to repro- ison between Guyana and Mauritius). duce the initial conditions. Political equality also matters for the qual- The evidence in this chapter suggests that ity of public policy. The basic role of the state the first type of society will tend to be more is to provide public services. But politicians prosperous. We argue that societies prosper- have the correct incentives to provide public services only when they have to appeal to the broad mass of citizens to attain power. If they can win power with a small number of Figure 6.1 Countries with more secure property rights have higher average incomes key supporters, or with few votes, they will GDP per capita, PPP in 1995 (log) tend to be clientelistic and more inclined to 11 buy votes or make individual exchanges of LUX patronage for support without providing the USA KWT JPN DNK CAN CHE AUS 10 QAT SGPSWEBELNOR ITA FRA ISLNLD AUT FIN GBR goods and services critical to raising the mass OMN BHS IRL ESP NZL ISR MLT SAUGRCBHR KORCZE PR T of people out of poverty. ARG URY TTO GAB MYS HUN 9 ZAF RUS Some simple patterns in the cross-country VEN MEXPOLCHL PRY TURTHA BWABRA BGR IRN PAN TUN CRI COL SLV DZA data show that more egalitarian distributions PER ROM JAM GTM SYR DOM ECU EGY PHL JOR MAR IDN 8 BOL GUYLKA ZWE GIN PNG CHN HND of political power and income are associated COG PAK CMR CIV IND NICAGO SEN HTI VNM STP with sustained and enduring prosperity. Fig- SDN MNG BGD KEN GHA 7 GMB BFA SLE YEM ZMB TGO ure 6.1 indicated that more secure property MLI MDG NER UGA NGA MOZ MWI GNB ETH ETH TZA rights are associated with higher incomes. 6 ZAR Crucially, however, better institutions and 5 secure property rights are associated with 3 4 5 6 7 8 9 10 greater political equality. Average protection against risk of Although there is no perfect way of expropriation, 1985­95 measuring political equality, protection Sources: Political Risk Services, International Country Risk Guide (ICRG) and World Bank database. against expropriation risk is highly corre- Note: The figure shows the relationship between GDP per capita in 1995 and a measure of the security of property rights, "protection against expropriation risk," averaged over the period 1985 to 1995. The data on institutions come lated with measures of democracy and from Political Risk Services, a private company that assesses the risk that investments will be expropriated in different countries. These data, first used by Knack and Keefer (1995) and subsequently by Hall and Jones (1999) and Acemoglu, measures of "constraints on the executive" Johnson, and Robinson (2001, 2002a, 2004), are imperfect as a measure of the relevant institutions because they pertain from the Polity IV database. This second to investments by foreigners only. Even so, they seem in practice to capture how stable property rights are in general. The findings are robust to using other available measures of related institutions. variable is designed to capture the extent to Equity, institutions, and the development process 109 which those who control political power are constrained or checked by others. The types B O X 6 . 1 Banking in the nineteenth century, of checks and balances and separation of Mexico and the United States powers written into the U.S. Constitution are classic examples of such constraints. Much recent work on growth and develop- had done so.The broadening of the suffrage, how- ment has focused on financial and capital ever, served to undermine the political coalitions There is a negative correlation between con- that supported restrictions on the number of bank markets. A central issue is to understand charters.That is, it created a second source of polit- straints on the executive and the Gini coeffi- why financial systems differ. For example, ical competition--competition within states over cient of income distribution. studies of the development of banking in who would hold office and the policies they would enact (10). The simple correlations suggest comple- the United States in the nineteenth century demonstrate a rapid expansion of financial mentarities between a relatively egalitarian The situation was very different in Mex- intermediation, which most scholars see as distribution of political power, good institu- ico. After 50 years of endemic political insta- a crucial facilitator of the economy's rapid bility, the country became unified under the tions, and prosperity, and a relatively egali- growth and industrialization. Haber (2001) highly centralized 40-year dictatorship of investigated the development of banks in tarian distribution of economic resources. Porfirio Díaz until the revolution in 1910. the nineteenth century in Mexico and the The correlations are consistent with many In Haber's argument, political United States. He shows that"Mexico had a institutions in the United States allocated different causal stories, but recent research series of segmented monopolies that were political power to people who wanted suggests that one can tell a causal story awarded to a group of insiders"(24). In 1910 access to credit and loans. As a result, they "the United States had roughly 25,000 about this data along exactly the lines we are forced state governments to allow free banks and a highly competitive market suggesting, which the rest of this chapter competitive entry into banking. In Mexico, structure; Mexico had 42 banks, two of political institutions were very different. discusses. The different evolutions of bank- which controlled 60 percent of total bank- There were no competing federal states, ing systems in Mexico and the United States ing assets, and virtually none of which actu- and suffrage was highly restrictive. As a ally competed with another bank." in the nineteenth century provide a good result, the central government granted Why this huge difference? The relevant example of the sort of historic argument we monopoly rights to banks, which restricted technology was certainly widely available, credit to maximize profits.The granting of rely on (box 6.1). and it is difficult to see why the various monopolies turned out to be a rational way types of moral hazard or adverse selection for the government to raise revenue and connected with financial intermediation redistribute rents to political supporters Institutions and political should have limited the expansion of banks (North 1981). in Mexico but not the United States. Indeed, inequality matter for Haber (2001) documents that market Haber shows when the U.S. Constitution regulation was not aimed at solving market development: historical evidence was put into effect in 1789, the structure of failures, and it is precisely during this period U.S. banking looked remarkably like that Figure 6.1 showed the relationship between that the huge economic gap between the arising later in Mexico. State governments, security of property and prosperity for the United States and Mexico opened (on stripped of revenues by the Constitution, which see Coatsworth 1993, Engerman and whole world, but to interpret this causally we started banks as a way to generate tax rev- Sokoloff 1997). Haber and Maurer 2004 enues and restricted entry to generate need to find a source of variation in institu- examined in detail how the structure of rents.Yet this system did not last because tions. Doing this is not easy, but Acemoglu, banking influenced the Mexican textile states began competing among themselves industry between 1880 and 1913.They Johnson, and Robinson (2001) provide a for investment and migrants. As Haber show that only firms with personal contacts partial answer. They show that the same (2001) puts it, with banks were able to get loans and that basic pattern holds for a smaller sample of The pressure to hold population and business in such firms were less efficient. Even though the state was reinforced by a second, related, fac- countries--those colonized by Europeans economic efficiency was hurt by tor: the broadening of the suffrage. By the 1840s, regulations, those with political power were after 1492. Indeed, colonization of much of most states had dropped all property and literacy requirements, and by 1850 virtually all states . . . able to sustain them. the world by Europeans provides something of a large natural experiment. Beginning in the early fifteenth century and massively intensifying after 1492, Euro- tions in development. It also provides fairly peans conquered many other nations. Colo- clear-cut evidence to support our conjectures nization transformed the institutions in about the joint evolution of prosperity and many diverse lands conquered or controlled political and economic equality. by Europeans. Most important, Europeans created very different sets of institutions in Colonial origins of contemporary different parts of their global empire, as institutions exemplified most sharply by the contrast Acemoglu, Johnson, and Robinson, building between the institutions in the northeast of on the research of Engerman and Sokoloff America and those in the plantation societies (1997), explain that Europeans created good of the Caribbean. This experience persua- institutions in some colonies, particularly the sively establishes the central role of institu- United States, Canada, and Australasia (what 110 WORLD DEVELOPMENT REPORT 2006 Crosby (1986) calls the neo-Europes),and bad ment, and the factor endowments that influ- ones in others (particularly in Latin America enced economic organization.1 There is a and Sub-Saharan Africa). These institutions strong inverse relationship between popula- had a strong tendency to persist and thus, tion density in 1500 and current protection today, generate the results seen in figure 6.1. against expropriation risk for former Euro- Why did different institutions develop in pean colonies (figure 6.2).And colonies with different European colonies? The simplest disease environments that were worse for answer is that Europeans shaped the institu- European settlers also have worse institu- tions in various colonies to benefit themselves. tions today (figure 6.3). And because conditions and endowments dif- Other aspects of factor endowments are fered among colonies, Europeans con- more difficult to measure directly, but Enger- sciously created different institutions. There man and Sokoloff (1997) point out that are several important empirical regularities where the climate and soils were suitable for connecting initial conditions to current out- crops such as sugarcane--which could be comes. Of particular importance are initial grown on large plantations with slave labor, population density, the disease environ- such as northeastern Brazil--much worse institutions and more skewed distributions of political power evolved than in climates Figure 6.2 Low population density in 1500 is associated with a lower risk of expropriation today where wheat or other nonplantation crops Average protection against risk of expropriation, 1985­95 could be grown. 10 USA Why did Europeans introduce better CAN NZL AUS SGP institutions in previously relatively unset- 9 tled and healthy areas than in previously GMB IND 8 BRA MYS densely settled and unhealthy areas? How GAB BWA CHL BHS MEX IDN did factor endowments influence institu- COL TTO VEN PNG JAM MLT 7 URA MAR PR Y tions? Europeans were more likely to intro- CRI CIV ZAF MWI TZA TGO EGY ZMB MOZ ECU GIN DZA ARG CMR TUN DOM GHA VNM duce or maintain bad institutions where 6 ZWE KEN SEN LKA PPAK GUY PAN SLE MMR BOL PER AGO PHL NGA there were a lot of resources and rents to NIC HND LBY 5 GTM BGD NER SLV extract--gold, silver, and, most important, COG GNB MDG UGA BFA people to provide the labor. In places with a 4 MLI SDN HTI large indigenous population, Europeans ZAR 3 SOM could exploit the population through taxes, ­4 ­3 ­2 ­1 0 1 2 3 4 5 tributes, or employment as forced labor in Population density in 1500 (log) mines or plantations. And where plantation Source: Political Risk Services, International Country Risk Guide (ICRG) and Acemoglu, Johnson, and Robinson (2002). crops could be profitably grown, slave- based societies emerged. These types of Figure 6.3 Worse environments for European settlers are associated with worse institutions today colonization were incompatible with insti- tutions providing economic or civil rights or Average protection against risk of expropriation, 1985­95 equality of opportunity to the majority of 10 USA the population. So, a more developed civi- NZL CAN AUS SGP 9 lization with a denser population structure, IND GMB and particular climatic and agricultural con- 8 MYS BRA CHL GAB MEXBHS ditions, made it more profitable for the IDN Europeans to introduce bad institutions. MLT COL VENTTO PNG MAR JAM CIV 7 URY CRI ZAF EGY PRY TGO ECU TZA GIN DZA VNM In contrast, in places with little to extract, TUN CMR ARG DOM GHA PAK LKA SEN 6 GUY PER KEN where plantation agriculture was not prof- PAN SLE MMR BOL NGA itable, and in sparsely settled places where HND AGO GTM BGD NIC 5 SLV NER COG GNB BFA the Europeans became the majority of the MDG UGA 4 SDN population, it was in their interests to intro- HTI MLI ZAR duce much better institutions. In addition, 3 2 3 4 5 6 7 8 the disease environments differed markedly Settler mortality (log) among the colonies, with obvious conse- Source: Political Risk Services, International Country Risk Guide (ICRG) and Acemoglu, Johnson, and Robinson (2002). quences for the attractiveness of European Equity, institutions, and the development process 111 settlement. When Europeans settled, they Figure 6.4 A worse environment for settlers is associated with fewer constraints on the established institutions under which they executive at independence themselves had to live. Constraint on the executive at independence This research suggests that most of the 1.0 AUS USA CAN MUS IND BOLURY PNG NZL ECU CRITTO ZAF CHL JAM gap in per capita income between rich and COL PAK PRY poor countries today is due to differences in FJI NIC BRA MDG 0.8 VEN PAN institutions. More precisely, if one takes two ARG DOM HND typical countries--in the sense that they LKA SLV BEN BGD MLI 0.6 HTI MYS both lie on the regression line--with high MEX NER GUY GTM and low expropriation risk, such as Nigeria CAF PER and Chile, almost the entire difference in 0.4 GNB COG TUN VNM MRT GHA SGP MMR EGY KENSEN AGO GMB income per capita between them can be MAR LAO TZA IDN CIV DZA UGA GIN explained by the differences in the histori- 0.2 DJI CMR SLE BDI BFA TGO cally shaped measure of the security of TCD RWA NGA AFG property rights.2 The research also pre- 0.0 SDN ZAR sented regression evidence showing that 2 3 4 5 6 7 8 Settler mortality (log) once the effect of institutions on GDP per Source: Acemoglu, Johnson, and Robinson (2002a). The analysis indicates that that the same factors that gave rise to capita is properly controlled for, geographic good institutions gave rise to a more egalitarian distribution of power. Without some measure of voice, it is impossible variables--such as latitude, whether or not for a person's property rights to be guaranteed or for them to have real access to the legal system to make sure that contracts are honored. A more egalitarian distribution of political power is also associated with a more egalitarian dis- a country is landlocked, the current disease tribution of economic resources. To get a better understanding of the mechanisms, we need to look further into histori- environment--have no explanatory power cal analysis. for current prosperity. Different types of societies thus devel- ish conquistadors the right to Amerindian oped in different colonies with radically labor),3 the mita (a system of forced labor different implications for subsequent devel- used in the mines), and the repartimiento (the opment. Crucially, the societies that emerged forced sale of goods to Indians, typically at in the neo-Europes had distributions of eco- highly inflated prices). Pizzaro created 480 nomic resources and political power that encomenderos, under whose care the entire were much broader. And they placed con- Indian population was placed. In other straints on the exercise of political power colonies the situation was similar. For and the ability of elites to adopt policies instance, in the territory of modern Colom- favorable to themselves but deleterious for bia, there were about 900 encomenderos.4 society (figure 6.4). The encomienda did not last for long in all parts of the empire because the Spanish Development and inequality Crown attempted to curtail it by the end of in the Americas: A case study the sixteenth century. But the mita (from in colonial origins the Quechua word mit'a, meaning "turn") The colonization of Latin America began became a central institution until inde- with the discovery of the"Indies"by Colum- pendence, and forced labor lasted far bus in 1492, the assault on Mexico by Cortés beyond this in most of Latin America (until after 1519, and the conquest of Peru by Piz- 1945 in Guatemala). The effects of the zaro after 1532. From the beginning, the encomienda also persisted because the con- Spanish were interested in the extraction of centration of political power that it was gold and silver, and later in taking tribute associated with led to the emergence of and raising taxes. The colonial societies that large landed estates.5 The feasibility and emerged were authoritarian, based on the attraction of this type of economic system political power of a small Spanish elite who was determined by the higher population created a set of institutions to extract wealth densities of indigenous people in many from the indigenous population. parts of the Spanish empire and the extent After Pizzaro conquered Peru, he imposed to which such societies had already devel- institutions to extract rents from the newly oped into "complex societies."6 conquered Indians. The main such institu- Other institutions were designed to rein- tions were the encomienda (which gave Span- force this system. For instance, indigenous 112 WORLD DEVELOPMENT REPORT 2006 people were not allowed to give testimony Historical accounts show that initial con- in some cases, and in others the testimony ditions had a large impact on the institutions of 10 indigenous people was equal to that of that the settlers built. Because there was low 1 Spaniard.7 Although indigenous people population density and no way to extract did use the legal system to challenge aspects resources from indigenous peoples, early of colonial rule, they could not alter the commercial developments had to import main parameters of the system. In addition, British labor. And, relative to much of the the Spanish Crown created a complex web colonial world, the disease environment was of mercantilistic policies and monopolies benign, stimulating settlement. Indeed, the from salt to gunpowder, from tobacco to Pilgrim fathers decided to migrate to the alcohol and playing cards, to raise revenues United States rather than Guyana because of for the state. the high mortality rates in Guyana.10 But Spanish colonies that had small popula- these same conditions made it impossible to tions of Amerindians, such as Costa Rica, profitably exploit labor, whose bargaining Argentina, or Uruguay, seem to have fol- power forced elites to extend political rights lowed different paths of institutional devel- and create equal access to land and the law. opment. The sharp contrasts along many These forces were reinforced by the fact that institutional dimensions between Costa plantation agriculture and slavery were not Rica and Guatemala (where population profitable, at least in the northern United density was greater) have been much stud- States and Canada. ied. Although the formal political institu- These colonies ultimately provided ac- tions of the Spanish empire were the same cess to land to a broad cross-section of soci- everywhere, the way they functioned ety and the legal system became fairly depended on the local conditions.8 impartial, ensuring secure property rights The institutions that emerged in the for smallholders and potential investors. main Spanish colonies greatly benefited the The new institutions made investment pos- Spanish crown and the Spanish settler elite, sible through financial development and but they did not promote prosperity in secure contracting and business relation- Latin America. Most of the population had ships. Underpinning these institutions were no property rights, nor incentives to enter fairly representative political institutions socially desirable occupations or to invest. and a fairly egalitarian distribution of Europeans developed coercive regimes resources. As in Latin America, there was a monopolizing military and political power synergy between economic and political and respecting few constraints on their institutions, but this time it was virtuous, power (unless imposed by the mother not vicious. Institutions giving and protect- country in Europe).9 ing property rights for the mass of people In North America, the initial attempts at and institutions of democratic politics colonization were also based on economic complemented each other, ensuring an motives. British colonies were founded by environment conducive to investment and such entities as the Virginia Company and economic progress. the Providence Island Company with the aim Representative political institutions in of profits. The model was not so different Virginia were a direct result of the authori- from that of the Spanish or Portuguese (a sys- ties realizing that, because of the different tem that other British colonizing entities, conditions, the colonization strategy that such as the East India Company, used to great worked in Peru would not work in the effect). Yet these companies made no money. United States. Virginia had many competing Indeed, both the Virginia Company and the and fragmented tribes, not a large central Providence Island Company went bankrupt. tribal empire. It had no gold or silver, and Because of the absence of a large indigenous the Indians, not used to paying tribute or population and complex societies, a colonial engaging in forced labor, would not work. model involving the exploitation of indige- So, the settlers of Jamestown starved.11 In nous labor and tribute systems was simply response to these early failures, the Virginia not feasible in these places. Company tried various incentive schemes, Equity, institutions, and the development process 113 B O X 6 . 2 Growth with poor institutions does not last The elite had good investment opportunities in ate institutions to take advantage of the huge suffrage"(21). From this point onward political Argentina in the golden age from the 1870s to economic opportunities emerging on world conflicts intensified, with a stream of coups and the 1920s, in Czarist Russia in the decades lead- markets, but the structure of the political rules, redemocratizations that lasted until 1983. ing up to World War I, in Colombia in the half such as their overrepresentation in national Though among the richest countries in the century after 1900, and in the Côte d'Ivoire for political institutions, guaranteed the interior world in the 1920s, Argentina gradually slid back the first two decades after independence (Wid- provinces a large slice of the benefits (Samuels to being a developing country. ner 1993). Such situations are rarely sustainable, and Snyder 2001). Argentina shows that, even with poor insti- for three reasons. First, the possibilities for sus- Although the majority was excluded from tutions for political inclusion and conflict man- tained growth are, by definition, limited because the political system, the economy boomed with agement, growth is possible if elites have good institutions exclude the majority of the popula- the property rights of the Pampean elite guar- investment opportunities and can manage to tion from effectively investing. Second, in the anteed. But the huge rents created by this sys- forge compromises. But the booms eventually rare situations in which elites manage to create tem began to cause conflict. In the 1890s, the unravel. Even when elites, such as the agricul- arrangements so that they can benefit directly Radical Party emerged under Hipólito Yrigoyen, turalists of the Argentine Pampas, face very from growth without the need to create good and after a series of revolts it was incorporated good investment opportunities, growth cannot institutions more generally, such arrangements into the political system by the democratizing be sustained forever by agricultural export tend to be fragile, vulnerable to shocks or crises. impact of the Sáenz Peña Law in 1912. booms. Moreover, the rents created by bad Third, bad institutions create power struggles Although Yrigoyen was elected president in institutions create conflict without fundamen- that undermine growth, because they generate 1916, the traditional interests were confident tal balances of power in society.This meant that large rents for those who control power. that they could keep control of the polity and democracy in Argentina after 1912 was unsta- Consider the growth of Argentina in the half the economy.They were mistaken. Significant ble.The unchecked power of President century before 1930. After its independence changes in the social structure had occurred, Yrigoyen in the 1920s induced a coup in 1930, from Spain in 1816, Argentina plunged into 50 with rapid immigration from Europe, induced by as did that of Perón in the 1940s and in 1955 years of civil wars and conflicts over control of economic success, and the associated urbaniza- and again in 1976 after his return from exile. the country, mainly clashes between those in tion.The vote share of the Conservatives Although temporary political solutions can control of Buenos Aires and the littoral and declined rapidly and the prospect of a Radical sometimes ease conflict for a while, as they did those in the interior.These conflicts abated after Party majority was a key factor behind the coup in Argentina after 1853, in the absence of the 1853 constitution and the presidency of Bar- of 1930. Smith (1978) notes"this situation con- broader institutional inclusion, conflict tolomé Mitre with a compromise between the trasts sharply with that in Sweden and Great ultimately reemerges, undermining the incen- Pampas and the interior. Pampean mercantile Britain . . . where traditional elites continued to tives to investment. and agrarian interests would be allowed to cre- dominate systems after the extension of including a highly punitive, almost penal, Institutions and political effort to make money. Such efforts quickly inequality matter collapsed, however, and by 1619 the Com- for development: pany had created an unusually representa- tive set of institutions for that era: a general contemporary evidence assembly with adult male suffrage. Our review of comparative history supports The early history of the United States two conclusions. First, institutions, espe- shows a possible path to good institutions. cially those that underpin property rights Early attempts to create an oligarchic soci- for all and broad-based investment, have a ety with close control of labor quickly causative influence on long-run develop- collapsed. What emerged instead was a ment processes. And second, greater polit- relatively egalitarian society, with represen- ical equality can lay the basis for better tative institutions giving even the poorest economic institutions. By greater political colonists access to the law and some politi- equality, we mean, in particular, checks on cal representation. This laid the basis for the predatory behavior of political and eco- economic and social institutions that nomic elites, and the political need for the underpinned the takeoff of the United state to be responsive to middle and poorer States in the nineteenth century and its population groups. The basis for greater divergence from the fortunes of much of political equalities is often associated with Latin America. Some countries with weak underlying economic structures, although and unequal institutions have experienced causation can run both ways. periods of rapid growth, but these have How does this perspective relate to the proved to be unsustainable over the long variety of contemporary development expe- term (box 6.2). riences? It is consistent with the perspective 114 WORLD DEVELOPMENT REPORT 2006 that institutions and governance are central anything more than a transitory solution, as to a wide variety of development perform- may have been the case in Indonesia under ance, from growth to service delivery.12 the New Order. While debate continues, an important thrust The rapid economic development of the of this research has been to support the view Republic of Korea after the mid-1960s was that causation runs, at least in part, from bet- not due to a set of institutions put in place ter institutions to higher incomes, rather through a domestic balance of political than the other way.13 What is additional to power. Instead, as in Indonesia under the this (ongoing) debate is the second part of New Order regime, a precarious geopolitical the argument--that the nature and manage- situation, particularly after the rundown of ment of inequalities in power shapes the for- U.S. aid in the early 1960s, induced the Park mation of institutions. Some cross-country regime to create a pro-growth environ- analysis is suggestive: Rodrik (1999a) argues ment.14 This at least led to a contingent com- that the capacity of societies to manage mitment to good institutions, as it did under adverse shocks--itself a crucial determinant an authoritarian regime in Taiwan, China, of growth--depends on the depth of latent where a fairly egalitarian distribution of social conflict and the strength of conflict assets and incomes, perhaps eased the transi- management mechanisms. tion in the 1990s toward democracy, a To illustrate the argument, we continue greater equality of political influence, and to draw on comparative development expe- good institutions. As in much of East Asia, riences. We first look at East Asia, and then there was a political necessity to deliver look at agricultural pricing polices in income growth and services to the peasantry. Africa. We then examine in greater depth In Indonesia, Suharto's New Order gov- the comparative experience of Mauritius ernment also recognized that economic and Guyana, countries that started with growth was necessary to keep the regime in similar initial conditions, but then followed power and that, to achieve this, good eco- radically different development paths. This nomic policies had to be in place. This is also related to different experiences in induced Suharto to delegate macroeconomic managing polarization, which can be con- policy to technocrats and to respond to the tributory factors for violent social conflict. oil booms wisely. It also led him to intervene to attempt to control corruption and excesses Shared growth in East Asia: that would put in jeopardy the underpin- the Republic of Korea, Taiwan nings of the regime.15 (China), and Indonesia Yet this constraint, real though it was, at Elites may be forced by threats of social disor- least in the 1960s and 1970s, is only part of der to promote the prosperity of most citi- the story about Indonesian growth. Suharto zens. Indeed, societies that have a political managed to create a system that, while not necessity to appeal to or appease middle and introducing good institutions, induced lower groups (initially the peasantry) can investments and growth from which the grow substantially in the short run. Long- regime could benefit. One of the secrets run prosperity, however, requires institution- behind this appears to have been the role of alized, rather than contingent checks and Sino-Indonesian businessmen, the cukong balances on elite power and capacities to entrepreneurs. Many firms and businesses adjust to changing circumstances. The were controlled by Indonesians of Chinese response of elites to social disturbances origin who were very marginal politically. sometimes leads to solutions that perma- Suharto granted such businessmen mono- nently change the political equilibrium in a poly rights and placed members of the mil- beneficial way, as may have happened with itary and his supporters on their boards of the agrarian reforms in the Republic of directors.16 Rock (2003) argues, "There is Korea and Taiwan, China, in the late 1940s little doubt that the . . . distortions in New and early 1950s. More often, however, the Order microeconomic policies thwarted transitory ability of citizens to act collectively competition, rewarded cronies, and en- dissipates without elites having to propose couraged substantial investment in uneco- Equity, institutions, and the development process 115 nomic projects" (10). Yet they also gener- through co-optation or repression, poses a ated wealth, economic growth, and rents serious challenge to their interests . . . Their for the regime. It was precisely the political response has been to try to appease urban marginality of the cukong entrepreneurs interests not by offering higher money wages that made them an attractive business part- but by advocating policies aimed at reducing the cost of living, and in particular the cost of ner for the regime. food. Agricultural policy thus becomes a by- The economic success of Indonesia after product of political relations between gov- 1966 elevated it into the class of an Asian ernments and urban constituents (33). "miracle economy."17 The East Asia financial crisis in 1997, however, exposed and exacer- In contrast to the situation in Ghana, bated Indonesia's institutional weaknesses, Nigeria, and Zambia, Bates (1981), Bates plummeting the country into an economic (1989) showed that agricultural policy in and political crisis from which it is only now Kenya over this period was much more pro- beginning to recover, doing so on the basis of farmer. The difference was due to who con- a new foundation of decentralization and trolled the marketing board. In Kenya, farm- democracy, which have progressively institu- ers were not smallholders, as they were in tionalized greater relationships of accounta- Ghana, Nigeria, and Zambia, and concen- bility between citizens and state. (See focus 4 trated landownership made it much easier on Indonesia for a further discussion of the to act collectively. Moreover, farming was relationship between social and political con- important in the Kikuyu areas, an ethnic text and policy choices.) group closely related to the ruling political party, the Kenya African National Union Agricultural pricing policies (KANU), under Jomo Kenyatta.19 Farmers in Africa in Kenya therefore formed a powerful lobby and were able to guarantee themselves high Another important example illustrating the prices. Even though the government of connections between institutions, the dis- Kenya engaged in land reform after inde- tribution of political power and growth pendence, Bates (1981) argued that-- comes from the seminal studies of price regulation prices in agricultural markets in 80 percent of the former white highlands were Africa by Robert Bates.18 Bates (1981) left intact and . . . the government took elabo- demonstrated that poor agricultural per- rate measures to preserve the integrity of the formance in Ghana, Nigeria, and Zambia large-scale farms . . . [which] readily combine in defense of their interests. One of the most was due to government-controlled market- important collective efforts is the Kenya ing boards systematically paying farmers National Farmer's Union (KNFU) . . . The prices much below world levels. The mar- organization . . . is dominated by the large- keting board surpluses were given to the scale farmers . . . [but] it can be argued that government as a form of taxation. As a the KNFU helps to create a framework of result of this pernicious taxation, reaching public policies that provides an economic up to 70 percent of the value of the crop in environment favorable to all farmers (93­4). Ghana in the 1970s, investment in agricul- Bates concluded that in Kenya "large farm- ture collapsed, as did the output of cocoa ers . . . have secured public policies that are and other crops. In poor countries with a highly favorable by comparison to those in comparative advantage in agriculture, this other nations" (95). meant negative rates of economic growth. Bates demonstrated why economic poli- Why were resources extracted in this way? cies were better in Kenya than Ghana in the Although part of the motivation was to pro- 1960s and 1970s, but this advantage did not mote industrialization, the main one was to survive the coming to power of Daniel arap generate resources that could be either expro- Moi in Kenya.20 The change in the ethnic priated or redistributed to maintain power. basis of the regime, from Kikuyu to Kalenjin, As Bates (1981) put it, undermined the coalition that had sup- governments face a dilemma: urban unrest, ported good agricultural policies, because which they cannot successfully eradicate the export farmers were not only large, but 116 WORLD DEVELOPMENT REPORT 2006 Figure 6.5 Constraints on the executive also predominantly Kikuyu. As a result, eco- tion in which parties led by Indo-Mauritians are greater in Mauritius than in Guyana nomic performance declined precipitously and Indo-Guyanese faced a coalition of par- Constraint on the executive in the 1980s and 1990s. The balance of ties supported by the non-Indian population, 1.0 power that sustained good policies in the led by Gaetan Duval in Mauritius and Forbes Mauritius 1970s did not endure. Burnham in Guyana. Yet, at independence, 0.8 politics and economics diverged. 0.6 The contrasting experience The Mauritian Labour Party won power Guyana of Mauritius and Guyana 0.4 initially and quickly abandoned its radical Mauritius and Guyana, in the 1960s, were policies--by the early 1970s, investment in 0.2 both poor societies dominated by the pro- the export processing zone had begun. The 0.0 duction and export of sugarcane. They had political hegemony of the Labour Party was 1970 1980 1990 2000 similar histories, factor endowments, social quickly contested by a strong socialist party, Source: Polity IV data set, downloaded from Inter- and political cleavages, and institutions. If the MMM (Mouvement Militant Mauricien) University Consortium for Political and Social Research. Variable described in Gurr (1997). anything, Guyana, although slightly poorer, led by Paul Berenger and Dev Virahsawmy. In had better prospects, because of its proxim- response, the Labour Party entered a coali- ity to the large U.S. market. Yet Mauritius tion with Duval and his PMSD (Parti Figure 6.6 GDP per capita is rising in Mauritius, not in Guyana has become one of the most dynamic and Mauricien Social Democrate) and the previ- successful (and equal) developing countries, ous opposition groups. The Labour Party GDP per capita (log) industrializing and maintaining competitive drew back from repressing the new political 9.5 democratic politics. Guyana slumped into forces, allowed the MMM to contest the 1976 Mauritius 9.0 dictatorship and poverty. election, and instead adopted social policies, The divergence between Mauritius and such as the provision of universal secondary 8.5 Guyana since independence is a fascinating education, to improve its popularity. It also Guyana example of economic and political diver- quickly dropped populist macroeconomic 8.0 gence in apparently similar societies (fig- policies and, in the late 1970s, implemented a 7.5 ures 6.5 and 6.6). serious stabilization program under the IMF. 1970 1980 1990 2000 What can explain this? Both countries The final test of Mauritian institutions was Source: World Bank (2005g). have similar histories. Mauritius was taken the election of an MMM government for the from the French and Guyana from the Dutch first time in 1982. Once in power, the MMM during the Napoleonic wars.21 In the nine- abandoned its more radical policies, and teenth century both developed sugarcane when the broad political consensus for good economies and, after the abolition of slavery institutions became clear, the export process- in the British Empire in 1834, imported large ing zone boomed. numbers of indentured laborers from India. The contrast with Guyana is stark. The Both have a similar population structure, first election on the eve of independence was with Indo-Guyanese and Indo-Mauritians won by Burnham and his People's National forming the majority of the population with Party in a coalition against Jagan's People's significant minorities of people of African, Progressive Party. Burnham maintained European, and Chinese descent. power by increasingly fraudulent means, After World War II, both colonies were finally changing the constitution in 1980 to moved by the British toward independence make himself executive president. He assassi- with early elections for democratic legislative nated opponents, most famously the radical assemblies dominated by pro-independence economist and political activist Walter Rod- political parties led by Seewoosagur Ram- ney in 1980. The economic policies of Burn- goolam in Mauritius and Cheddi Jagan in ham's regime were a disaster. He expropriated Guyana. Both groups used extensive socialist the sugar plantations, creating highly ineffi- rhetoric and proposed land reforms and cient state industries, and he aggressively fairly radical policies. Many of the political promoted his party members through struggles with British administrators over patronage, particularly in the civil service. The postindependence institutions, such as the implied or actual threat to property and per- form of the electoral system, were fought over son led to a huge diaspora of Indo-Guyanese similar issues. As independence arrived how- from the country, including most of the pro- ever, political forces re-formed into a situa- fessional and middle-class people. Only in the Equity, institutions, and the development process 117 1990s did a democratized Guyana begin to tee British companies. The authoritarian slowly recover from this legacy. But the ethnic tendencies of the colonial state were rein- divide endures, and the country continues to forced by British military intervention, pro- suffer from weak governance, a lack of politi- moted in 1953 by the United States, to cal transparency, and ethnic tensions that remove Jagan from power because of his hamper economic and social development. socialist tendencies. Guyanese politicians, What can explain such divergent outcomes unlike those in Mauritius, had far less abil- in such apparently similar circumstances? In ity to get what they wanted from the colo- Guyana, there were fewer constraints on the nial state. This meant that there were fewer use of power, and political conflict was indigenous checks on the exercise of power, more polarized, defined solely along ethnic and unfettered use of political power was lines. And although both countries started the norm. The best example here is the elec- independence as democracies, what the toral system. Britain imposed a propor- majority could do (or wanted to do) to the tional representation system on Guyana minority was limited in Mauritius, but not because it was afraid that the overrepresen- in Guyana. tation of large parties inherent in majoritar- In Mauritius, the British colonial state ian systems would allow Jagan to win an faced a powerful and homogeneous French absolute majority in the 1964 election (the planter class that did not leave the island People's Progressive Party won 42.6 percent after Mauritius was annexed to Britain in of the vote in the 1961 election). This sys- 1812. In the 1870s, when Britain was tem facilitated Burnham's rise to power. reducing the autonomy of colonial admin- Although the British tried to do the same istrations, it was forced to create a legisla- thing in Mauritius, political elites there held tive assembly. Although this was initially out and forced a compromise: a system with dominated by the planters, by the turn of relatively large electoral districts with the the twentieth century the first Indo-Mau- three politicians who got the most votes ritians were elected. This was a clear sign being elected and with the eight best"losers" that the greater political autonomy of the from the entire country being elected to par- island was allowing for a more open soci- liament. This system maintained elements ety with greater upward mobility of for- of the majoritarian institutions that Maurit- mer indentured laborers. The power of the ian leaders believed were essential to main- colonial state was checked, evident in the taining the country's governability. Politics fact that Mauritian independence leaders in Guyana became completely defined along were able in the 1960s to negotiate postin- ethnic lines. This occurred because the pre- dependence institutions closer to the ones vious evolution of the economy, and the they wanted. dominant power of colonial interests, left This juxtaposition of different local little room for the varied interests that interests and the weakening of the legacy of emerged in Mauritius. While Guyana has the colonial state gave rise to a more bal- not suffered outright social conflict, high anced distribution of political power in levels of polarization and weak conflict Mauritius. And from this situation more management institutions can be contribu- fluid interests emerged. Though ethnic tory factors to civil wars (box 6.3). identities were certainly important in poli- tics, so were different cleavages, as is clear Implications from the development of the MMM into a In Mauritius, property rights are secure and powerful political force and the coalition of the country has experienced open demo- Ramgoolam and Duval in the 1970s. Poli- cratic politics. There has been intensive tics became much less polarized than they investment in education and free access into might have been. profitable investment opportunities, illus- In Guyana, there was no indigenous trated most clearly by the export processing planter class to check the power of the colo- zone. In Guyana, the opposite was true in the nial state. After the departure of the Dutch, 1970s and 1980s. The puzzle is why institu- the plantations came to be owned by absen- tions have been so good in one case and so 118 WORLD DEVELOPMENT REPORT 2006 rural social disorder. The spillover from the B O X 6 . 3 Polarization, conflict, and growth conflicts of 1965 and 1966 was a redistribu- tion of power toward the rural sector, with Researchers have long recognized that sity and conflict, and through this channel, sustained, inclusive growth necessary for deep social divisions make it harder to economic growth (Esteban and Ray 1994). implement policies that benefit all. Getting By this measure, a country with three the political survival of the regime. a more precise measure of the nature and groups that comprise, respectively, 49 per- Yet the redistribution of power in extent of such divisions, however, has cent, 49 percent and 2 percent of the popu- Indonesia was not institutionalized, unlike proved problematic. For much of the lation will be more polarized than a coun- what occurred in the Republic of Korea, for 1990s, scholars used a measure known as terpart country where those same groups "ethno-linguistic fractionalization"--first comprise 33 percent, 33 percent, and 34 example. Moreover, it did not force the New compiled by Russian social scientists in the percent of the population.The polarization Order Regime to improve institutions out- 1960s--to show that economic growth measure is a far more robust predictor of side the rural and education sectors, was slower, controlling for other factors, in civil conflict than either measures of the societies where there was a low probability inequality of individual incomes or although the connection between promot- that two citizens drawn randomly from a fragmentation.This statistical association is ing economic development and social order population group were of the same ethnic illustrated by the fact that, by this measure, may well have helped the government to group. Africa's "growth tragedy"was, in 9 of the 10 most polarized societies in the sustain its relationship with the cukong part, blamed on its high level of "fractional- world have experienced major civil conflict ization"(Easterly and Levine 1997). in the past few decades, including Eritrea, entrepreneurs. As the constraints on eco- More recent work has sought to refine Guatemala, Nigeria, Sierra Leone, and Bosnia nomic policies of the New Order Regime measures of social diversity by focusing and Herzegovina (García-Montalvo and relaxed in the 1990s, it appears to have been instead on polarization, or the extent to Reynal-Querol forthcoming).This is only one more difficult to avoid a massive and debili- which a small number of influential groups influence on conflict, of course, and other dominate a society, thereby providing a work has emphasized the role of resource tating upsurge in corruption and rent-seek- more theoretically informed basis for dependence and state capacities (see World ing. Moreover, the collusive agreement that explaining the relationship between diver- Bank 2003h). the state forged with the Sino-Indonesian entrepreneurs appears to have been very fragile. It rested on shared expectations bad in the other, given such apparently sim- about the longevity of the relationship, ilar histories and circumstances. expectations that clearly deteriorated with But the two cases make sense in more Suharto's failing health and could not sur- detail. The colonial history of Mauritius vive the financial crisis in 1997.22 diverged from Guyana's in significant ways that allowed the development of a stronger Transitions to more equitable domestic political society. Mauritius resis- ted the colonial state more effectively and, institutions ultimately, generated a more egalitarian So far we have examined cases illustrating distribution of political power and a less the mechanisms that create good institu- polarized structure of political conflict. In tions and sustain prosperity. They involve Guyana, however, there was no powerful institutions that allow for greater equality of domestic interest group that had a vested opportunity, and behind such a set of insti- interest in opposing the colonial state or that tutions lies a relative balance of economic was able to block the state from expropriat- resources and political power. Such institu- ing land and other assets after independ- tions have emerged in some societies but not ence. The use of power was unconstrained, others. Although systems of institutions and politics were highly polarized along often tend to reinforce one another and per- ethnic lines. sist for long periods, they also change. Indonesia shows that growth is possible Countries with unequal distributions of even with underlying bad institutions when resources and political power become more elites can credibly make a contingent com- egalitarian and democratic, and previously mitment to improve institutions and when powerless people gain power and influence. they manage to forge mechanisms that indi- Although institutions are sometimes created rectly benefit from encouraging the invest- by colonialism or military conquest, they ment opportunities of others. The accelera- can often evolve through good decisions, tion of growth after 1966, and particularly virtuous paths, and the intrinsic dynamics the pro-poor aspect of growth, was clearly of the development process, as in Mauritius. driven by the threat of communism and It is also possible that even transitory condi- Equity, institutions, and the development process 119 tional solutions lead to permanent change, distribution of resources and power, and because growth unleashes transformations subsequent changes in institutions. These that induce beneficial changes in institu- changes included the collapse of feudalism tions. This message from modernization and serfdom and the move to a free labor theory23 is precisely what may have hap- market, the changes in land distribution, the pened in the Republic of Korea. commercialization of agriculture and the The biggest challenge is to understand development of interoceanic commerce.25 processes of change and to distill from them Yet, even after 1688, the political system lessons about how poorer societies can was at root oligarchic. Further changes undergo beneficial institutional transitions. were needed in the distribution of power This does not appear to have happened in toward greater political equality to sustain Argentina (box 6.2) or Guyana, but it did Britain's development path and eventually happen in Britain in the seventeenth, eigh- deliver a more egalitarian society. Even teenth, and nineteenth centuries and in Fin- though Britain was a constitutional regime, land, Sweden, Spain, and the Republic of it was a very limited democracy in 1800. Korea in the twentieth century. It also hap- Before the first reform act of 1832 set in pened in Mauritius. Here we briefly review motion political liberalizations that culmi- three such transitions: early modern Britain, nated in full democracy in 1918, fewer than Finland and Sweden in the early twentieth 10 percent of adult males could vote. The century, and China in the last 20 years. The reason for these changes seems to have been transitions and policy choices in Spain are the effect of early industrialization and discussed in focus 3 on Spain. urbanization on the ability of the disenfran- chised to contest the power of political Early modern Britain elites.26 British democratization in the nine- Around 1500 most European countries teenth century was the outcome of a series were highly hierarchical feudal societies of strategic concessions by political elites to ruled by absolute monarchs whose powers avoid social disorder.27 were endowed by God. The most prosper- While the political system of the eigh- ous places, such as the Italian city states of teenth century was consistent with individ- Venice, Genoa and Florence, had escaped ual initiative, invention, and the start of the feudalism and were ruled by republican industrial revolution in Britain, sustained governments strongly representing mercan- long-run growth called for broad invest- tile interests. The Netherlands also escaped ment, particularly in human capital. Such intense feudalism and was relatively pros- institutions had to wait for mass democ- perous, but it was part of the autocratic racy to begin to arrive after 1867.28 How- Habsburg Empire. Nevertheless, the differ- ever, the longer history of the Poor Laws ences in income between the most and the provide an example of how provisioning least prosperous places were relatively for adverse risks was also supportive of small. After 1500, this picture began to greater dynamism (box 6.4)--a theme we change rapidly. First the Netherlands and return to in chapter 7. then Britain became much more prosper- The types of political reforms in nineteenth- ous than the rest of Europe, and the century Britain led to economic institutions Mediterranean world went into decline. that clearly influenced the distribution of As North and Thomas (1973) argued, the income, most obviously the promotion of most plausible explanation for these changes education after 1867. But the same period is the emergence of constitutional govern- also saw extensive labor market reforms that ment in the Netherlands and Britain: diverg- strengthened the bargaining power of labor ing prosperity within the early modern period and led to the rise of the Labor Party. After was tied to the evolution of political institu- 1906, the Liberal government of Herbert tions.24 Institutions improved because of a Asquith also began to introduce the basics of change in the distribution of resources and a welfare state, further extended by the Labor political power. Indeed, there was a virtuous government after 1945. As Britain began to circle of changes in institutions, the broader adopt institutions that promoted prosperity, 120 WORLD DEVELOPMENT REPORT 2006 1867, which was the first reform that really B O X 6 . 4 Aiding equitable growth in early modern expanded voting rights to working people. Britain: the role of the Poor Laws When democracy enfranchises the relatively poor, they usually can use democracy to tilt Far from being a consequence of successful parish registers,which was instituted in economic institutions and the distribution of economic growth, recent historical research 1538.This placed the English population on income in society in their favor.30 on seventeenth- and eighteenth-century an entirely different basis, in terms of social Britain has found that widespread but security,from that of the rest of Europe. unique institutions of social security were in The comprehensive social security system Twentieth-century Finland existence for several centuries before the provided by the Poor Laws had a number of and Sweden31 industrial revolution. Indeed, scholars highly significant economic consequences.In increasingly argue that a previously under- combination with laws (dating from the thir- Finland and Sweden are popularly identi- estimated influence on Britain's industrial teenth century) granting complete alienability fied as prosperous countries with generous revolution, in fact, lies in its prior agricultural of land,it encouraged labor mobility and welfare states that, in some measure, are revolution.The principal comparator here is reduced the attachment to land holding as products of a small and ethnically homoge- with the immensely advanced Dutch rural the only form of security for peasants.Individ- and trading economy of the sixteenth and uals had a relative certainty of being provided nous population. But, a closer reading of seventeenth centuries. Many of the most for,wherever they moved to work in the econ- their economic histories shows that their important technical innovations in British omy,no matter what their property-owner- contemporary "virtuous circles"--with agriculture during this period, such as land ship status.Landlords and farmers could reap growth and equity mutually reinforcing-- drainage engineering, new crop types, and the economic gains to be had from increased rotations, were directly borrowed from the farm sizes,from enclosure,and from laying off are the outcome of a long and difficult Dutch.Yet it was the British agricultural and workers or changing their labor contracts to political struggle to establish institutions service economy that was increasingly out- more efficient weekly or day labor,without and enact policies that provide broad eco- pacing the Dutch as the seventeenth and provoking the same degree of peasant nomic opportunities and respond to the eighteenth centuries progressed.Why? protest as occurred on the continent.But Attention has recently been given to equally,employers in England had a strong inherently wrenching social transitions of one major institutional difference between incentive only to do this if it made economic positive (economic growth, structural the two countries--the nationwide system sense because,through the Poor Law,they change) and negative shocks (macroeco- of social security created in England by the would also have to reckon with their liability nomic crises, civil war). Poor Laws, which gradually evolved during to pay for the families of the laid-off workers. the course of the sixteenth century,culmi- What the Poor Law created in England Finland was part of Sweden in the Middle nating in the famous Elizabethan statutes of was a public system of acknowledgment of Ages, but following a war between Russia and 1598 and 1601.This was a Christian human- collective responsibility for the basic subsis- Sweden in 1808­09, it became part of the ist response,imbued with a new optimism tence of all, including for a strikingly non- about what government could and should moralistic approach to the support of single Russian empire. It experienced one of the last be able to achieve in the face of perceptions mothers and their illegitimate children.The European famines in 1867­68, an event that of increased poverty amid plenty in a time comparative evidence suggests a relative ushered in major demographic and eco- of population growth.The Poor Law was lack of correspondence in England--alone nomic changes as entire regions were devas- mandated by the central state but--most in all of Europe--between fluctuations in important for its practical effectiveness--its the price of food and the death rate, and tated. The Russian revolution of 1917 led to a implementation was entirely locally England--but not Ireland--was the first collapse of imperial authority in Finland, and devolved: it was funded by a local tax on nation in the world to cease to experience the country soon declared its independence. property in every parish, administered by famine-related mortality. But this immediately gave birth to a bloody local officials but also rigorously enforced by local magistrates.It went side by side with a civil war between "white guards" (bourgeois Sources: Szreter (2005) drawing on Slack (1990), relatively efficient nationwide population Wrigley (1998), Solar (1997), Solar (1995), King nationalists) and "red guards" (socialists loyal registration system, the Church of England's (1997), King (2000), Lees (1998). to Russia). More than 30,000 troops alone lost their lives. In the aftermath, however, many progres- sive reforms laid the foundation for the mod- it was still a highly unequal society, and ern Finnish economy and society. Land inequality almost certainly increased until the reform--a major cause of the civil war--was early or mid-nineteenth century (figure 6.7). enacted almost immediately. A law passed in Although precise measures of inequality dif- 1918 allowed sharecroppers to buy their land, fer depending on the sources, inequality and amendments in 1922 facilitated the sub- appears to have risen until the early and per- sidized expansion of small farms. Progressive haps mid-nineteenth century.29 After about income and wealth taxation were in place by 1870, there is wide consensus that inequality 1920, soon followed by expansions of fell substantially for the next century. women's rights (although universal suffrage The fall in inequality after 1870 is closely in parliamentary elections had been in place correlated with the Second Reform Act of since 1906) and commitments by the central Equity, institutions, and the development process 121 government (not just local municipalities) to Figure 6.7 Inequality in Britain began to fall around 1870 primary education. Gini coefficient From the late 1940s until the early 1990s 0.8 the economy expanded steadily, with per Lindert and Williamson (1982, 1983) capita incomes catching up with Great 0.6 Britain in the 1980s and Sweden in the 1990s (from roughly half a century earlier). 0.4 Williamson (1985) Bourguignon and Morrison (2002) This success was a product of Asian-style 0.2 "governed markets": collaboration between the state and private sector was harnessed to 0.0 rapidly industrialize an economy that, as 1759 1801 1823 1867 1871 1881 1890 1901 1910 1929 1950 1960 1970 1980 1992 late as the 1950s, generated 40 percent of its Sources: Lindert and Williamson (1982), Lindert and Williamson (1983), Williamson (1985), output from agriculture.32 A crucial coun- and Bourguignon and Morrisson (2002). terpart to Finland's activist industrial policy (based on high rates of capital accumula- table institutional arrangements be in tion and public saving, low interest rates on place. credit, and major investments in manufac- In Sweden, these prior arrangements turing infrastructure), however, was the were unusually favorable to upward mobil- construction of a welfare state to cushion ity by subordinate groups: a long history of citizens of all ages against the unsettling peasant autonomy, a correspondingly weak social changes wrought by such a rapid eco- aristocracy, and an emerging nation-state nomic transformation. able to secure support from farmers while Strong and credible political leadership also repudiating aristocratic claims on its was central to making this possible. In the powers. Sweden was also the first country to aftermath of World War II, President Urho have a central bank (in 1668) and among the Kekkoken famously asked his nation,"Do we first to grant basic property rights. As such, have the patience to prosper?" Thereafter, he "inclusion of the peasantry in the trans- set about negotiating the arrangements formation of the agrarian economy and ("social corporatism") among industrialists, institutional arrangements that sustained trade unions, and citizen groups that would egalitarianism were to become fundamental enable all to act as complements. The Finnish elements in the rise of the Swedish industrial model has its problems (high unemploy- market economy."33 This was an economy ment), but it shows how state, market, and increasingly grounded in broad political society can jointly generate the institutions, rights and social opportunities. policies, and spaces needed to generate equi- But history is not destiny. Equitable table development outcomes. development is as much a function of key Sweden is perhaps most closely associ- choices and decisions at pivotal historic ated with the welfare state today. Less well junctures. The Middle Ages, the industrial known is the timing and sequencing of revolution, and the tumultuous twentieth events putting it in place. Importantly, the century unleashed sweeping forces on Swedish welfare state was the product of, Swedish society. Some were leveling (rising not a precursor to, the country's transition agricultural productivity), others wrench- to modern economic growth. Indeed, it was ing (mass unemployment). Each attempt to designed in response to the very problems respond to these forces established the (old-age security, unemployment) gener- political contours for subsequent attempts. ated by such growth. But to make such Drawing on and extending the equitable growth possible, and to have in place institutional foundations during these piv- sociopolitical conditions that would enable otal historic junctures have been the unify- the articulation of and sustained support ing elements of Sweden's development for something like the welfare state (when strategy. Its achievements to date have been such a system existed only in rather embry- remarkable, even as twenty-first-century onic forms elsewhere in the capitalist realities present distinctive challenges to its world), it was vital that a prior set of equi- welfare state. 122 WORLD DEVELOPMENT REPORT 2006 The main implications of the Finnish and check on the discretionary use of power by Swedish cases for today's developing coun- the central government. China's experience tries are that economic growth and sociopo- also demonstrates that what is important for litical equity can be powerfully reinforcing, equitable development are credible checks and can be underpinned by institutional on the arbitrary use of power, assurance of transitions. These cases should not be seen as property rights and fair treatment for a blueprints for others to follow. Instead, they broad segment of society. The particular should be read as examples of how commit- form that institutions take to deliver these ments to equity in a given context help lay the functions can vary, especially during periods foundations for short- and long-term pros- of transition. perity by consolidating virtuous circles link- The key to China's equitable development ing institutions and incentives. was the combination of initial conditions and the economic reforms launched in 1978 China in the late twentieth century that unleashed entrepreneurial initiative and Economic development in China since 1978 legitimized the profit motive. China's eco- has been nothing short of spectacular. With nomic policies following the 1949 revolution the quadrupling of GDP per capita over the proved seriously flawed: they stifled incen- last 25 years, China has transformed itself tives for investment and innovation. But the from a poor centrally planned economy to social policies of the Mao Tse-Tung period a lower-middle-income emerging market leveled the distribution of assets in important economy. As a result, the number of people and durable ways. As a result, both land and living in poverty (under $1 per day) fell from human capital were equitably distributed on 634 million in 1981 to 212 million in 2001.34 the eve of reforms. With the adoption of the From the perspective of this chapter, what rural household responsibility system, peas- is interesting is that the world's largest coun- ants became the immediate beneficiaries of try has undergone profound economic reform. This helped to reinforce equity, while transformation without substantially chang- unleashing entrepreneurial initiative and ing the political institutional structure, that boosting productivity. remains dominated by the Chinese Commu- The economic reforms launched in nist Party. Yet institutional improvement did 1978 aimed at decentralizing economic take place in China along with economic decisions--to individual farm households, reform. And the large increase in nonstate enterprise managers, local governments-- investment and free entry into profitable so as to generate incentives for investment economic opportunities suggest that prop- and innovation. Importantly, the form erty rights are secure, despite the absence of a these policies took and the transitional Western-style judicial system. institutions that were created were While the particular institutional form is designed to preserve the political support different from other cases reviewed here, the for reforms, by compensating potential experience in China is broadly consistent losers. with the thesis of this chapter. The earlier The aftermath of the cultural revolution, discussion of equitable transitions in Britain and the recognition that China's economy and Scandinavian countries illustrated the had fallen behind--not least in relation to the argument that a successful economic system East Asian Tigers--led to a growing consen- depends on the political system to assign sus on the need for and urgency of change, and enforce property rights and contracts, and paved the way for the economic reforms and to protect the market from political initiated under Deng Xiaoping's leadership. encroachment. China's recent history sug- These reforms were inspired by the wide- gests that the starting point for reforms spread recognition of the failure of central does not necessarily have to be in political planning as an instrument for economic institutions. Changes in economic institu- organization,and reflected the need to deliver tions and in economic relations among lev- on economic growth for the legitimacy of the els of government can also establish credible new leadership. The political need for growth commitment to a reform path and act as a implied a new focus on liberating markets Equity, institutions, and the development process 123 and incentives. The sequencing of reforms which generated incentives that focused on and the transitional institutional arrange- collection to provide local public goods that ments that accompanied the economic attracted local investments. These changes decentralization, on the other hand, reflected provided for significant autonomy from the the premium the leadership placed on social central government and considerable inde- and political stability. pendent authority over their economies. The impetus for economic decentraliza- tion on the one hand, and the need for an The modern Chinese system includes a divi- sion of authority between the central and local integrated national market on the other, governments. The latter have primary control helped to shape a dynamic relationship over economic matters within their jurisdic- between the central government and local tions. Critically, there is an important degree governments that held them mutually of political durability built into the system.35 accountable and limited discretion on both sides. Over time, the result of these policies China's reforms are also replete with inno- was to create a stake in new economic institu- vative mechanisms for protecting potential tions for all the main actors, including the losers during transitional periods. This often local governments which served as a credible involved designing reforms that sustained check on the powers of the central govern- sources of income for incumbents, by ment in the economic domain. The reforms keeping important elements of pre-existing also fueled the emergence of strong economic pricing and payment mechanisms, while centers, such as Guangdong province and the providing incentives at the margin. "The Shanghai municipality. These centers now transitional institutions [were] not created wield considerable influence and bargaining solely for increasing the size of [the] pie, [but power relative to the central government and also] to reflect the distributional concerns of can serve as important countervailing forces. how the enlarged pie is divided and the polit- How did economic decentralization rein- ical concerns of how the interests of those in force private incentives? According to Walder power are served."36 and Oi (1999), "For almost 20 years, reform Dual pricing at the start of reforms is a in China has proceeded through the gradual prime example. The system obliged farmers reassignment of specific property rights from and enterprises to sell specified quantities to higher government agencies to lower govern- the state at "plan" prices, while allowing them ment agencies, or from government agencies to obtain market prices for any above-quota to enterprises, managers, families, or individ- production. This maintained the planning uals" (7). All of these reforms enhanced the system for those who benefited from it, while power of economic agents to make decisions creating incentives for efficient production. over economic activities in their respective Equally important, it allowed time for market domains, and boosted productivity through institutions to emerge, avoiding the institu- better incentives.Farmers retained their earn- tional vacuum that plagued many transition ings and therefore worked harder and economies when state institutions were dis- invested more. Township and village govern- mantled. Fiscal contracting guaranteed the ments had rights to the profits made by central government a certain level of rev- township and village enterprises (TVEs) and enues,37 but it generated incentives for local therefore adopted policies that promoted governments to collect more because the business. But because they had no revenue marginal retention rate was much higher. authority, they did not have the ability to bail Similarly, labor contracting allowed state out poorly performing TVEs, which made for workers to retain the guarantee of lifetime hard budget constraints and higher efficiency. employment while introducing greater flexi- Higher levels of local governments (coun- bility in labor policies for new contractual try and province) acquired control over local workers. These arrangements made reforms a enterprises and therefore also had a stake in win-win game, ensuring social stability and their performance. They were allowed to the support of those in power. retain more local revenues through fiscal con- But there is a danger in such a strategy of tracting and to have extrabudgetary funds, incrementalism: getting stuck in incomplete 124 WORLD DEVELOPMENT REPORT 2006 reforms if local governments and incum- Qian (2003) notes the following: bents acquire too much power and are able There is apparently a larger room than we to block further progress. The prevalence of thought for institutional innovation to interprovincial barriers to trade in the simultaneously address both the economic 1990s, with each province vying to boost and political concerns, that is, to make a profits for the enterprises it owned, is an reform efficiency improving and interest example. But there are some checks and bal- compatible for those in power (305). ances in the system that help maintain the But there are many challenges ahead, some direction and momentum of reforms. These of which will not be amenable to win-win include competition among local govern- solutions and therefore are likely to be polit- ments, hard budget constraints for local ically and socially more costly. Continued governments, the central government's reforms in the state enterprise and financial insistence on enforcing a unitary market, sectors, managing rural-urban migration, and a growing economy that reduces the and addressing increasing regional disparities economic influence of incumbents. (see focus 6 on regional inequality) are some The struggle for the right balance of these challenges. Macroeconomic policy between economic centralization and decen- and structural reforms will need to be under- tralization is constantly evident in many of pinned by further institutional improvement China's domains of intergovernmental rela- to ensure broader participation and account- tions. The 1994 tax reforms recentralized fis- ability so that the interests and desire of the cal revenues, in part to ensure greater people are better reflected in decision mak- regional equity in spending, and the central ing, and to further strengthen the govern- government continues to apply strict con- ment's capacity to lead market-oriented trols on deficit financing by local govern- reform while maintaining economic and ments. Constraints on labor mobility have social equity. eased considerably over time, helping to cre- ate a more unified labor market, despite con- cerns by some provincial governments that Conclusion this might aggravate problems of unemploy- A few simple principles go a long way ment for established urban residents. toward unifying different development There are also some more recent and experiences in the historic and the contem- more permanent institutional changes that porary worlds. There is little disagreement reconfirm the government's commitment to among scholars that basic institutions, such market-oriented reform. These include as security of property rights and equality mechanisms that strengthen accountability before the law, are keys to prosperity. These at the local level and empower local popula- institutions lie behind the capital, financial, tions. Local elections are the most important land, and labor markets that we saw in of these mechanisms but others include, for action in chapter 5. Because talent and ideas example, recent regulations to eliminate nui- are widely distributed in the population, a sance taxes on the rural population. China prosperous modern society requires the has also successfully used the external com- mass of people to have incentives--and a mitment device of WTO accession to signal state that can and will provide key com- its resolve to move ahead with market plementary inputs and public goods. It reforms and impose discipline on incum- therefore requires an underlying set of bents. For example, it is no longer possible institutions that generate the equality of for every province to have its own inefficient opportunity for individuals and assure the automobile factory erected behind trade bar- accountability of politicians to all. riers designed to provide local employment Why do some societies have such insti- and local taxes. More broadly, China's desire tutions and not others? A relatively egali- to carve for itself an important place in the tarian distribution of political power global order and to be recognized as a underpins the institutions that promote responsible global power places constraints prosperity. Institutions clearly have distrib- on the shape of its future policies. utional effects, and bad institutions often Equity, institutions, and the development process 125 arise because they benefit some group or onto a dynamic path toward a virtuous cir- elite. Good institutions arise when checks cle of equity and prosperity? The organiza- are placed on the power of elites and when tion of society is highly persistent, but we the balance of political power becomes more have seen many cases of transitions to bet- equal in society. Often, equality of political ter institutions. Sometimes, as in early power is supported by economic equality, modern Britain, economic changes lead to and this connection gives rise to the possi- changes in the distribution of power, which bility of both virtuous and vicious circles. promotes a more equitable society and bet- Growth certainly can occur in societies ter institutions. Contemporary China fol- in which these conditions do not apply. But lows a similar pattern albeit with a different the preponderance of evidence suggests that configuration of institutions. In other such growth is unsustainable. This perspec- times, as in the Republic of Korea and tive is consistent with historical narratives, Indonesia, regimes are forced, by external basic patterns in cross-country data, and or internal threats, to change the trajectory more careful causal empirical work on the of their society in ways that become institu- sources of prosperity. tionalized. In still other times, such as Mau- The crucial question for the promotion ritius and Botswana, leaders make good of development is this: how can poor soci- decisions that lead to reinforcing paths of eties improve their institutions and move better institutions and development. f o c u s 4 o n Indonesia Growth, equity, and poverty reduction in an East Asian giant I ndonesia presents an illuminating "politics in command" in 1959 and pro- poverty line until the 1990s and all still sub- example of the long-term interactions duced a ruinous inflation that brought sisting on less than $2 a day, grew at the of the three basic themes of this report much of the population to near starvation in same rate (or possibly slightly faster). The on equity and development: the mid-1960s. It was with just cause that distribution of household expenditures had Gunnar Myrdal pronounced in Asian been remarkably stable, with the overall · The importance of market-driven Drama, 1967, that "no economist holds out Gini coefficient staying within a narrow processes in determining the distribu- any hope for Indonesia." range between 0.31 and 0.36.3 Rural tion of opportunities and incomes. Indonesia's rapid, pro-poor growth for inequality had actually declined signifi- · The role of political processes, and the the 30 years after the fall of Sukarno aston- cantly since the 1970s, when access to land engagement of the poor in these ished the development profession and, allowed substantial benefits to be reaped processes, in determining the policy along with other countries in East and from the green revolution. By the mid- framework for market and asset accu- Southeast Asia, Indonesia became the object 1980s, the labor market had become the mulation. of intense analysis.1 In Indonesia, the weak primary determinant of income in rural · The overriding dominance of institu- starting conditions significantly influenced areas. tions in determining the long-run con- how the economic planners approached the But when the Asian financial crisis hit in ditions of governance for markets and task of linking growth to the poor. They 1997 and President Suharto was forced to politics to operate. designed a three-tiered strategy for pro- resign in the face of widespread rioting in poor growth, which connected sound 1998, the country was entirely unprepared in These complex interactions require long macroeconomic policy to market activities political or institutional terms to cope with periods of developmental evolution to that were facilitated by progressively lower the rapid changes needed in corporate and observe and identify. transaction costs. Those policies were linked public governance. The crisis sharply low- Indonesia has substantial variance to household decisions about labor supply, ered inequality, as urban real estate and across all three of these themes. There is agricultural production, and investment in financial markets collapsed. But the dra- enough independence in the variance for the nontradable economy. matic reduction in GDP--over 13 percent in each factor to sort out, if only roughly, what The extent to which the poor benefited 1998 alone--caused poverty rates to triple. is driving what. In chapter 6, the political from growth depended on the array of Only after 2002 did poverty rates return to dimension of the economic performance of assets they controlled: their labor, human the previous lows observed in 1996. By 2004 the Suharto regime was discussed. Here, we capital, social capital, and other forms of they still had not returned to the trend rate discuss the connections with policy capital, including access to credit.2 Appro- of decline disrupted in 1998. choices. priate government policies also influence Explaining these trends in per capita Because Indonesia has been so impor- those dimensions, especially in health and incomes and their distribution requires an tant to the development profession, it has education. The "road to pro-poor growth" understanding of how markets, politics, been studied for a long time. The Dutch started from desperately poor economic and institutions jointly shaped the rapid, exploited the Netherlands East Indies from conditions, weak institutions, and a decade pro-poor growth strategy, its subsequent the seventeenth century to early in the of political instability. It seemed that every- collapse, and current efforts to revive it. twentieth century. Then, under political thing needed to be done at once. The key Any such explanation is bound to be con- pressure at home, the Dutch experimented was to focus on restarting and then sustain- troversial, and there is no formal model with an "Ethical Policy" for the colony, and ing rapid economic growth, empowering behind the story about to be told.4 But the the poor benefited significantly. During the poor households to enter the market econ- story is plausible and anchored in the his- Great Depression, World War II, and the omy, and reducing the costs and risks of torical record. fight for Independence, the Indonesian doing so by investments to lower transac- The story begins with two concerns of economy deteriorated rapidly, and the poor tion costs. the emerging Suharto government in the suffered disproportionately. Java was the The strategy worked for three decades: late 1960s. The first was the misery and dis- original home of the "dual economy" ana- between 1967 and 1996, income per capita content of the rural masses, who had sup- lyzed by Boeke (1946) and formalized by increased by 5 percent a year. The incomes ported Sukarno's communist leanings and Lewis (1954). After declaring independence of the bottom quintile of the income distri- populist rhetoric. After a decade of active in 1945, President Sukarno eventually put bution, all individuals below the national discrimination against their livelihoods, Focus on Indonesia 127 rural households were near starvation and by putting down the riots and imprisoning national budget be balanced quarter by thus an obvious source of opposition unless the student leaders. Then it mounted a seri- quarter--a law Suharto basically imposed the new government could incorporate ous effort to make the economy more equi- on himself, but then touted to all con- them in its development plans. Second, the table. The result, also stimulated by the stituents as a rule the government had to hyperinflation of the mid-1960s, the total world food crisis in 1973­4, was a major shift live under. To build confidence among the disintegration of the market economy, and in priorities toward rural development and a Chinese business community, the govern- the political chaos meant the entire popu- specific push toward increasing domestic ment opened the capital account in 1970 lation was ready for a more stable life. A rice production. Behind this push were the when it unified the exchange rate. The flow strategy that promised stability and rural objectives of stabilization and equity. To of foreign exchange to and from Singapore recovery would win wide support (as it lose control of the rice economy was to lose and Hong Kong was a sensitive barometer would throughout densely settled East and control of what mattered to Indonesian of the investment climate. Southeast Asia). society. Thus the two constraints on the presi- This is the message that Suharto deliv- The restructuring of Indonesia's develop- dency, which Suharto felt personally and ered to his technocrats. This economic ment approach after 1974, especially the pre- used as motivation for his bureaucracy and team had engaged Suharto and other senior emptive devaluation of the rupiah in 1978, government (not the same thing in Indone- military officials in economic training exer- signaled the government's determination to sia), were the need for rural areas to partici- cises at the Military College. The tech- include the poor in the development process. pate in growth, and the need to keep the nocrats were handed the macroeconomic The stability of the Gini coefficient seen from investment climate highly favorable for portfolio and told to deliver on what the late 1960s to 2004 should not be taken as Suharto's business partners. The response to became known in Indonesia as the develop- the result of market-driven forces in the face both constraints was an economic package-- ment trilogy--growth, equity, and stability. of given technology, but as a conscious gov- low inflation, food price stability, an open To many in the political and military arena, ernment effort, led from the macroeconomic economy, and massive investments in rural stability meant repressive measures to stifle arena by the technocrats, to stimulate pro- infrastructure--that generated rapid pro- dissent, but to the technocrats it meant poor growth.6 This effort succeeded in spec- poor growth. But another part of the restraining inflation (which they did in tacular fashion until the mid-1990s, when investment climate, a part only for those spectacular fashion in just three years) and cronyism and the growing influence of favored business partners, involved special stabilizing the rice economy, which was still Suharto's children on economic decision licenses, trade protection, and lucrative a quarter of GDP and providing half the making caused the approach to unravel. access to domestic markets. This part average Indonesian's daily calories. The Part of the problem of post-Suharto unraveled the "open economy" part of the institutions built to provide this stability, in governments has been their need to dis- growth package. both macro terms and in the food econ- tance themselves from this record of repres- The Suharto legacy, despite the deep omy, became essential to the Suharto sion and cronyism, despite three decades of commitment to pro-poor growth, did not regime's success.5 pro-poor growth. This tension brought the build the groundwork for a political and Thirty years of rapid economic growth, failure of political and institutional devel- institutional framework that would ulti- with equally rapid rates of poverty reduc- opment during the Suharto era to the fore. mately support it. A deep tension developed tion, was politically popular (the elasticity Questions about causality remain, particu- between the institutional framework to of reduction of the headcount poverty larly whether rapid, pro-poor growth can keep the open economy functioning effi- index with respect to growth in per capita be implemented by authoritarian regimes. ciently and the political controls to keep the incomes was about 1.3 during the Suharto Indonesia's record, along with that of most cronies' businesses profitable. Without polit- era). Every five years, the polling results for of East and Southeast Asia, indicates that ical feedback about these very same political parliament were gleaned for signs of disap- they can. But is such growth sustainable? controls, the regime was blindsided by the pointment with the development program. And which is more important for managing ferocity of the opposition to its manage- Despite the heavy hand of Golkar, the pres- long-run, pro-poor growth: good econom- ment of the Asian financial crisis. The ident's party, real information was flowing ics or good institutions? depth of the crisis, both economic and from villages up to the center through these In Indonesia, there was no "chicken or political, reflected the vacuum of institu- elections. egg" problem. Something had to be done at tions in place to cope with an alternative Almost despite the intentions of the once in view of widespread destitution and political system. Suharto regime, political institutions were political chaos, and the sequencing was The climb out of the chaos of 1998 mir- taking root (people expected to vote) and clear. Rapid, pro-poor economic growth rors that from the 1965 era, but this time these institutions provided feedback to the was imposed by an authoritarian regime without order imposed from above. The policy approach of the government. There concerned about its survival. But this same eagerness and skill with which the Indone- were other feedback mechanisms as well, regime also imposed on itself commitment sian population has participated in the and the ones that threatened stability were mechanisms to make the growth process democratic process suggests that social and taken very seriously. After the 1974 riots in market friendly to rural households and to political order will now be far more sustain- Jakarta in reaction to the visibly widening Chinese capitalists--that is, both ends of able. The challenge now is to translate the income distribution, especially in urban the economic system. Inflation was brought same democratic process into rapid and areas, the government responded brutally under control by a law requiring the sustainable pro-poor economic growth. Leveling the economic and political playing fields III WHAT CAN BE DONE TO INCREASE EQUITY IN THE WORLD? Can this be done in ways that also spur long-term prosperity? We read in part I that there are large inequalities of opportunity between people within coun- P A R T tries and--even more--between people in different countries. These inequalities are perpetuated through interlocking economic, political, and sociocultural mechanisms, creating inequality traps. Individuals from different groups and countries face a highly uneven playing field, both in their capacities to acquire endowments and aspire to a better life, and in their opportunities to reap returns from those endowments through market and nonmarket processes. Because differences between countries often exceed within-country differences, it is of particular importance that national policies support, or are at least consistent with, the narrowing of international differences, notably through the growth process. We argued in part II that many inequalities not only violate peo- ple's concern for fairness, but actually have costs for the development process. The effects on development depend on specific forms of inequality and their interactions with market imperfections and insti- tutions. Unequal opportunities are associated with inefficiencies and wasted economic potential. Pronounced inequalities in the distribu- tion of power are often associated with weak economic institutions, undermining the investment and innovation that is central to long- run growth. Greater equity is thus not only intrinsically desirable but also is complementary to long-run growth and prosperity. For poorer and excluded groups, a focus on equity can bring a double benefit--a bigger pie and a greater share. But the scope for such a complementary relationship between equity and aggregate development is often not exploited. When exam- ining this, we suggest there are two kinds of pathology in policy 130 WORLD DEVELOPMENT REPORT 2006 design. First, there is the pathology associ- institutions (even though action to support ated with oligarchic dominance--institu- empowerment of the poor is now empha- tions and policies that further the interest of sized in the design of specific policies--see elites but not those of the whole society. Narayan 2002). In part III, we focus on a set This may take the form of extreme preda- of areas that do lie squarely in the arena of tion and high-level corruption, as in development analysis and practice--in poli- Mobutu's Zaire or Haiti under the Duva- cies affecting the sectors, markets, and in the liers. Or it may take the form of enmeshed global arena. This recognizes the influence alliances between economic and political of the political and sociocultural context, elites that favor rent-seeking, as in the but focuses rather on what an equity prism, Philippines under Marcos, in much of Latin based on the analysis of parts I and II, has to America in past decades, and in more subtle say about the policy design to break inequal- forms in many countries of the world. ity traps and support aggregate growth The Second, there is a more complex pathol- lesson from part II is that this implies paying ogy of policies pursued with the intent, or in attention to specific inequalities and their the name, of equity that have high efficiency interactions with markets, social structure, costs or perverse effects. Communist eco- and power. This involves both issues of tech- nomic policy was disastrous for efficiency, nical design and mechanisms that provide even while many communist societies did the political underpinnings for change, much in social provisioning. Directed notably through broader accountability, credit--in India, for example--was intended coalitions for change, or compensation of for the poor (and reached some of the poor), losers. And while an overarching message is but proved a high-cost strategy. Populist of the potential complementarity between macropolicy is always bad for growth, and greater equity and long-run prosperity, almost always bad for equity sooner or there will often be tradeoffs in specific areas later--witness Argentina during much of the and context. One cross-cutting area con- second half of the twentieth century. Per- cerns the need to raise taxes to finance desir- verse or growth-sapping effects of policies able public spending. The design of tax under this pathology can be caused by instruments is of great importance to mini- adverse consequences for incentives, unaf- mize adverse efficiency effects, while also fordable fiscal burdens, or the capture of the promoting equity where feasible. benefits, often by middle groups, which We organize the discussion of domestic "hoard opportunities" at a cost for other action into three areas. First is building and groups and the overall growth process. protecting people's human capacities-- What can be done? At a fundamental from the start of people's lives and through level, the analysis underscores the centrality adulthood and old age. Here we focus on of shifting to a state that is more account- equalizing from the bottom up--equalizing able, has checks on predatory behavior of up the opportunities of the least advantaged political and economic elites, is responsive in terms of skills, health, and risk manage- to all citizens--especially from middle and ment. There are certainly issues of equity poorer groups--and has effective conflict among more advantaged groups, but we management mechanisms. In part II, we give priority to the disadvantaged (in part sketched cases of transitions in this direc- for reasons of space). As seen in part II, tion from history and contemporary expe- there are major market imperfections in riences, and at the local level. The emphasis human capital formation and insurance in the development community on issues of that affect poor or lower-status groups governance and empowerment is entirely most, yet political action has also often been consistent with this perspective. biased against these groups. While such overall shifts are central to Second is ensuring equitable access to jus- development, the World Bank has neither tice and complementary assets. A fair and the mandate nor the comparative advantage accessible justice system is crucial for con- to discuss specifics of the design of political straining the power of the political and eco- Leveling the economic and political playing fields 131 nomic elite, avoiding discrimination, and nation-states is uneven--and has uneven protecting property rights and personal safety effects on different groups within countries. for all--with important implications for the There is substantial scope for making the willingness to invest and innovate.Inequitable playing field more even. But as in the domes- access to land and infrastructure--by wealth, tic arena, policy design involves both techni- location, or social group--is typical of devel- cal questions (such as the details of migra- oping societies and often enmeshed with tion arrangements and the application and political structures. Policy design can help design of patent legislation) and the political shifts to more equitable and often more effi- underpinnings of rules and institutions for cient patterns (chapter 8). global governance. We examine the potential Third is the domain of markets--finan- for change both in the key global markets-- cial, labor, and product--that have a power- for labor, products, ideas, and capital--and ful influence on the returns to people's in the potential scope for designing aid in endowments. As chapters 5 and 6 argued, ways that support (rather than undercut) markets are typically far from ideal, work- domestic development, and through more ing in noncompetitive and discriminatory effective and equitable management of the ways, whether because of intrinsic market global commons (chapter 10). imperfections, or because power structures The epilogue links the report's perspective have shaped them to serve the purposes of on equity to the thinking and agreements that those in power. In these areas, and notably have evolved in the development community in the case of finance, a primary concern is in the past decade--captured, for example, in equalizing down, by reducing protecting the Millennium Declaration (2000) and the privileges of incumbents. Closely related is Monterrey Consensus (2002)--as well as the the conduct of macroeconomic policy World Bank's own strategic pillars of an (chapter 9). enabling investment climate and promoting In the global arena, concern remains with empowerment. We argue that an approach to individuals--and the enormous, unjustified development that is deeply informed by differences in opportunity that people face equity is fundamental to the full integration of through the morally irrelevant fact of coun- these frameworks into an effective develop- try of birth. The global playing field between ment strategy. Human capacities 7 Expanding people's capabilities to lead ence of the powerful and the wealthy to be fuller lives, the aim of all development, able to shape public policies to benefit the rest c h a p t e r means investing in their education and of society. As we have seen, successful transi- health and in their ability to manage risks. tions are far more likely where the power of But as chapters 5 and 6 discussed, failures in the excluded to influence public action has markets and governments conspire to gen- been enhanced. erate large inequalities in people's opportu- There are strong complementarities among nities to build their capabilities. Children the different investments in people. Better from poorer families start out life with nourished children have higher cognitive greater disadvantages than their wealthier abilities. Well-educated parents, especially peers, attend lower-quality schools, have mothers, invest more in their children's less access to health services, and are not as education and health. More educated indi- protected from economic downturns and viduals are likely to be more resilient to family crises. By the time they are adults, shocks. Instruments to smooth consump- they are far less equipped to be productive tion will spur people to take on not only members of society. Economic, political, higher risk but also potentially higher and sociocultural inequalities fuel such dif- return activities and prevent them from dis- ferences in life chances, perpetuating them investing in themselves (lowering food across generations. intake, forgoing treatment) or in their chil- Public action can level the playing field dren (pulling them out of school) in times and broaden opportunities by addressing of shocks. And people with more human inequalities in access to quality education, capital and better risk management capabil- health care, and risk management. Well- ities can reduce the variability and increase designed policies will result in more equi- the level of their incomes. tably distributed opportunities to acquire The policies we consider in this chapter endowments and boost overall productiv- are particularly important in arresting the ity. As potentially talented and productive intergenerational transmission of inequali- individuals gain access to the services from ties. We begin with a review of the rationale which they may have been excluded for rea- and potential for early childhood develop- sons that have nothing to do with their ment programs. We next consider broader potential, societies make gains through education and health policies for expanding greater efficiency and greater social cohe- access to quality education and care, and sion in the long run. finally discuss transfer policies that help Still,there are challenges.Programs require manage risks and provide for efficient and resources, administrative capacity, and politi- equity-enhancing redistribution. cal support. This means paying attention to the design of tax systems, tailoring program Early childhood development: intervention to context,and above all building constituencies for change. We focus on level- a better start in life ing the playing field mainly through aug- By the time poorer children in many countries menting the capacities of those with the reach school age, they are at a significant dis- fewest opportunities, but we recognize that it advantage in cognitive and social ability. The may be necessary to attack the undue influ- Ecuadorian study cited in chapter 2 docu- 132 Human capacities 133 mented substantial differences at six years of children, regularly monitoring their growth, Figure 7.1 Children from better-off age, related to socioeconomic status and stimulating the development of their cogni- households have a big edge in cognitive abilities by age three parental education. Differences in childhood tive and social skills through more frequent cognitive abilities are indeed apparent as early and structured interactions with a caring Early vocabulary growth as 22 months of age. One study in the United adult, and improving the parenting skills of Cumulative vocabulary States shows that by age three the gaps in caretakers. The evidence suggests that these 1200 learning, measured by vocabulary, are already programs can be highly effective in address- High large among children from different social ing problems experienced later in schooling SES groups (figure 7.1).1 Cognitive learning is and adulthood. Middle affected by a child's socioeconomic status A recent study in the United States shows SES 800 through health (malnutrition, iron and that investments in the early years of life, micronutrient deficiency, parasite infections) before children reach the formal school sys- Low and the quality of the home environment, tem, give greater returns than later invest- SES including care-giving and cognitive stimula- ments (figure 7.2).7 Well-designed longi- tion.2 Scientific evidence on brain develop- tudinal studies--mainly from developed 0 12 16 20 24 28 32 36 ment supports this. Recent research findings countries--indicate that programs typically Age, months revamp earlier thinking that assumed that the register improvements for children in Source: Hart and Risley (1995). structure of the brain was genetically deter- health, cognitive ability, academic perform- Note: SES refers to socioeconomic status. mined at birth and point to the determining ance, and tenure within the school system influence of early experiences--from concep- and, later in life, higher incomes, higher tion to age six, and particularly the first three incidence of home ownership, lower propen- years--on the architecture of the brain and sity to be on welfare, and lower rates of capacities in adulthood.3 incarceration and arrest.8 This suggests a As a child ages, environmental effects strong productivity case for investing in appear to accumulate. Poor cognitive and early childhood development; the argu- social abilities are associated with weaker ments for public subsidies to disadvantaged future academic performance and lower families are compelling on both productiv- adult economic and social outcomes, ity and equity grounds. As Heckman argues, including poor health, antisocial behavior, early interventions in children from disadvan- and violence.4 These underachieving adults taged environments raise no efficiency-equity influence the cognitive abilities of the next trade-offs; they raise the productivity of indi- generation of children, creating an inter- viduals, the workforce and society at large, and generational cycle of poverty and unequal reduce lifetime inequality by helping to elimi- opportunities.5 Studies using internation- nate the factor of accident of birth.9 ally comparable student achievement tests Studies of ECD programs in developing confirm that socioeconomic background is countries also document strong benefits the overwhelming determinant of learning outcomes, with schools accounting for no more than 20 percent of the variation in test Figure 7.2 Early childhood interventions are good performance.6 investments Return per $ invested 8 Benefits of early interventions Preschool programs Early interventions can substantially en- 6 hance a child's life chances and loosen the Schooling intergenerational grip of poverty and in- Opportunity equality. In recent years, interest has ex- 4 cost of funds r panded in early childhood development Job training (ECD) in low- and middle-income devel- 2 oping countries, paralleling greater atten- tion in developed countries. Preschool School Post-school 0 Early childhood development programs 0 6 18 comprise a range of interventions that Age include providing nutritional supplements to Source: Carneiro and Heckman (2003). 134 WORLD DEVELOPMENT REPORT 2006 Figure 7.3 Catching up through early intervention ing school by 5.6 percentage points and led to higher completed schooling attainment Development quotient 110 and higher adult cognitive achievement test Children of normal Stimulation and nutritional supplement scores. height 105 Evidence is also mounting that interven- Stimulation tions in early childhood particularly benefit 100 Nutritional supplement poor and disadvantaged children and fami- lies.10 In Jamaica, nutritional supplementa- 95 tion and stimulation administered to under- Control group nourished children between the ages of 9 to 90 24 months--who are most likely to come from disadvantaged families--improved 85 Baseline 6 12 18 24 their mental development.11 Malnourished Months children who received a milk supplement Source: Grantham-McGregor and others (1991). developed more than those who did not (fig- Note: The Development Quotient is an index composed of ratings in four behavioral and cognitive development indica- tors: locomotor (large-muscle activities, running, and jumping), hand-eye coordination, hearing and speech, and per- ure 7.3). Children who benefited from stim- formance (shape recognition, block construction, and block patterns). "Months" refers to time after entry into the pro- ulation did even better, while the benefits of gram, generally at around 9 months old. supplements and stimulation were additive and allowed disadvantaged children exposed B O X 7 . 1 ECD programs are an essential ingredient for the to them to almost catch up with the develop- attainment of education for all ment trajectory of "normal" children over a period of 18 months. The results suggest that There is sufficient evidence from studies significantly for children from the higher ECD programs may be one of the most cost- throughout the world to make a case for caste, but fell a dramatic 46 percent for the effective avenues for reaching the Millen- placing early childhood education among lower caste and an astonishing 80 percent nium Development Goals for universal edu- the key interventions to achieve the educa- for the middle-caste students. In Colombia tion Millennium Development Goals the third-grade enrollment rates for cation and an important contributor to the (MDGs). children who participated in the PROMESA attainment of gender parity in primary com- Higher school enrollment. Colombia's program increased by 100 percent, reflect- pletion (box 7.1). They also help mothers ECD program PROMESA reports ing their lower dropout and repetition rates. participate in the labor force--poor women significantly higher enrollment rates in pri- In addition, 60 percent of the children who mary school for children participating in the participated in the ECD program attained who had access to free child care in the fave- program, compared with children not par- the fourth grade, compared with only 30 las of Rio de Janeiro increased their income ticipating. ECD programs in India (Haryana) percent of the comparison group. by as much as 20 percent--and improve the and Guatemala resulted in a significant Higher intelligence. ECD programs decline in enrollment age for girls. encourage young children to explore and academic performance of older children, as Less grade repetition. In Colombia's facilitate the social interaction that documented for a community nursery pro- PROMESA program, the Alagoas and Fort- promotes cognitive development. Children gram in rural Colombia.12 aleza PROAPE study of Northeast Brazil and who participated in Jamaica's First Home- the Argentina ECD study, children who par- Visiting Program, Colombia's Cali Project, ticipated in the programs repeated fewer Peru's Programa No Formal de Educación Designing ECD programs grades and progressed better through Inicial (PRONOEI),and the Turkish Early Interventions to improve young children's school than did nonparticipants in similar Enrichment Project in low-income areas of circumstances. Istanbul averaged higher scores on intellec- capacity to develop and learn can focus on Fewer dropouts. In India's Integrated tual aptitude tests than did nonparticipants. improving parents' teaching and child care Child Development Services program in Evidence from other studies, however, sug- skills, delivering services directly to chil- Dalmau, attendance of children ages six to gests that these effects dissipate over time. dren, or improving child care services in a eight in primary school increased by 16 per- cent for children who had participated in community. Programs may be established Sources: Chaturvedi and others (1987), Myers the program; dropout rates did not change (1995),Young (2002). in homes, day-care centers, or communi- ties. The evidence suggests that three design features are important for the full realiza- for all children, with cost-benefit analyses tion of benefits from ECD programs: start- showing returns of $2­5 for every $1 in- ing early, having strong parental involve- vested. For example, preliminary results ment, and focusing on child health for experimental nutritional interventions (especially nutrition) and cognitive and at 6 to 24 months of age in rural Guatemala social stimulation. The focus on health show that consuming a nutritional supple- leads to a virtuous cycle, because improved ment increased the probability of attend- health also helps increase cognitive and Human capacities 135 social abilities.13 Overly formal programs The second approach would target disad- can be too expensive for poor families, cul- vantaged families. This may be more cost- turally irrelevant, and insensitive to fami- effective in view of the evidence presented lies' needs.14 They thus run the risk of being earlier on larger gains from interventions for abandoned even when they demonstrate disadvantaged children. To bolster partici- high returns. pation, the program could be supplemented What then are the impediments to the by a cash-transfer scheme, with transfers widespread implementation of ECD pro- conditioned on various desirable behaviors, grams given that they are such good invest- including changes in the homecare environ- ments? Political economy constraints arise ment, as well as regular health center visits from the difficulty of making a case for for growth monitoring, immunizations, and spending resources on a program with the nutrition interventions.17 This would con- promise of (uncertain) benefits to come only centrate even more resources on the poor, years in the future. Such a case is often made but the political economy implications are by the immediate beneficiaries (parents of less clear. While targeted programs have a school-age children) or intermediate benefi- smaller constituency and thus would not ciaries (teachers), who organize themselves benefit from a broad coalition of support, a into powerful political forces. But the institu- national program, with transparent criteria tional setup for ECD program delivery-- for eligibility and good monitoring of "con- with funds in many instances channeled ditionality," could mobilize support not only to myriad small NGOs, community cen- from the direct beneficiaries but also from ters, or home-based caregivers--and the other stakeholders in society. absence of strong central responsibility It is possible to combine a universal pre- inhibit organized political pressure. The school approach with a conditional cash same institutional setup generates prob- transfer (CCT) program. This would yield lems of integration with other government the highest benefits in the participation of programs and of coordination across sev- the poor and the productivity gains for all, eral government departments.15 but it would also be more costly. The Thinking about the politics and the approach adopted in any country setting design of ECD from the start is thus impor- will have to emerge from considerations of tant. Getting information to parents, com- costs, benefits, and fiscal capacity--and munity leaders, and policymakers about the reflect the political economy. objectives and efficacy of ECD services can Basic education: expanding build public awareness and strengthen demand. Monitoring systems build support opportunities to learn by providing timely feedback on a range of Prominent in the Millennium Development intermediate outputs to policymakers and Goals, education is a great equalizer of program managers, while proper evaluations opportunities between rich and poor and provide more convincing evidence of impact between men and women. But the equaliz- and broader lessons from interventions. ing promise of education can be realized Integrating ECD programs into the broader only if children from different backgrounds development frameworks and involving par- have equal opportunities to benefit from ents, families, and community members quality education. In the previous section, enhance the sustainability of programs.16 we argued that children's ability to benefit There are two possible approaches to from school is strongly influenced by the scaling up ECD interventions. The first is to cognitive and social skills they acquire in expand publicly funded preschool pro- their early years. Evidence suggests that the grams to all children by making it a statu- gains from early interventions can dissipate tory right, as in several European countries. if disadvantaged children go on to low- This would have significant funding impli- quality primary schools.18 cations, but the benefit is potentially wide- Chapter 2 documented the large inequali- spread support from middle-class and poor ties in educational attainment within coun- families with children. tries by income, region, gender, and ethnicity. 136 WORLD DEVELOPMENT REPORT 2006 Chapter 5 presented the economic reasons the probability of their attendance. Even why credit-constrained households underin- uneducated parents will pull their children vest in education, making the efficiency case out of school if they perceive low quality.19 for subsidizing education for the poor. There There may be tradeoffs, however, if resources are other reasons for parents to choose a level devoted to upgrading quality benefit prima- of education for their children that may be rily the privileged who are already in school lower than what is optimal for the child and at the expense of reaching excluded groups for society. Educational attainment has vari- or areas--or if the rapid expansion of access ous societal benefits that are not fully cap- reduces the quality of instruction. While the tured by the individual. For example, it is long-run objective for school systems generally associated with enhanced democ- around the world is clear, priorities will vary racy and lower crime, while girls'schooling in by country, region, or group. particular has been shown to reduce fertility, empower women, and thereby contribute to Expanding access, particularly the welfare of children in the family. In addi- for excluded groups tion, education has intrinsic value, enabling Expanding access for all. More than 100 people to lead fuller lives as informed and million children of primary school age are active participants in society. out of school, either because they never The case for moving to equalizing access entered the system or because they dropped to education is therefore strong on both out before finishing.20 As a result, some 52 equity and efficiency grounds, especially for countries risk not reaching the goal of uni- basic education. Beyond basic education, versal primary completion.21 In most coun- there is an important efficiency rationale for tries, improving opportunities in education ensuring that the most talented and produc- means ensuring affordable access, especially tive people in society have access to higher for poor rural children and disadvantaged education. In today's globalized world, with groups. competition largely on the basis of skills and Higher public spending on the supply of ideas, countries need to cultivate latent talent, schools is one way to expand access. Analy- wherever it may reside. Motivated and tal- sis of the determinants of school enroll- ented children from poorer households ment in various countries suggests that deserve the opportunity to excel as much as proximity to schools is a major factor.22 A their wealthier peers. While we acknowledge careful evaluation of Indonesia's school the important equity dimension of policies construction program in the 1970s, the for tertiary education, the discussion here is largest such program on record, finds evi- devoted primarily to policies that expand dence of significant increases in both edu- access to and quality of basic education. cation and earnings.23 The program yields We argue that there is a case for public large positive returns, but it takes more than action to enhance equity in learning so that 30 years to do so because upfront construc- outcomes reflect not merely circumstances tion costs are high (more than 2 percent of of luck--parental endowments, sociocultural Indonesia's GDP in 1973), while the bene- environment, birth place, one dedicated fits are spread over a generation's lifetime. teacher--but genuine differences in prefer- But for every success story there are many ences, effort, and talent consistent with the others in which higher spending has not notion of equal opportunities. This requires translated into better access to infrastructure, expanding affordable access and upgrading inputs, and instruction for children. In many quality, with a particular focus on excluded cases, the resources are not used effectively-- groups, through various interventions that too much is spent on teacher salaries or increase both the demand for schooling and reducing class size and not enough is spent the capacity and incentives of the school sys- on instructional materials. tem to respond. Incidence studies suggest that the poor There are clear complementarities in this stand to benefit more from expansion when approach: quality improvements help only if mean levels of access to services are already children are in school, but they also influence reasonably high, now the case for primary Human capacities 137 schooling, even in many low-income coun- tries. But spending alone is clearly not B O X 7 . 2 School fees--an instrument of exclusion enough to get the children in school (and or accountability? even less effective in ensuring that they learn). In many countries, the main prob- There are two schools of thought about quality services from the schools,and point school fees.Some claim that school fees to studies that show even poor households' lem is not facilities but children dropping deter poor families from sending their kids willingness to pay for good quality services. out or not attending available schools.24 to school.Even nominally small amounts can Sympathetic to the arguments in favor Recent efforts to boost access thus focus on be a large share of poor households'income, of greater accountability, we argue for elimi- demand-side interventions: reducing the and these come on top of the forgone bene- nating user fees when the fiscal impact of fits of children contributing to family busi- forgone revenues can be managed without cost of schooling or providing incentives, ness or household chores.Schooling costs large efficiency costs or harmful spending even paying for attendance. often figure in parents'responses about con- cuts.The desirable voice and accountability In many countries, parents have to pay a straint to enrollment,and eliminating school aspects of school fees can be harnessed fees appears to have spurred a large equally or better through contributing lot, either for school fees or for other inputs, increase in enrollments in a number of labor for school improvements or working such as uniforms and textbooks. Eliminat- countries,including Kenya,Tanzania, on parent-teacher advisory committees. ing these costs can boost participation. Free Uganda,and Vietnam.Others see user fees Such in-kind fees are cheaper to the parent uniforms and textbooks provided by an as an important accountability tool,a mech- and engage the parent more fully in school anism for empowering parents to demand decision making. NGO program in Kenya (along with better classrooms) reduced dropout rates consid- erably: after five years, students in the pro- gram completed about 15 percent more tunity cost of schooling for poor families schooling. In addition many students from and represent one approach to addressing nearby schools transferred to program failures in credit markets and the imperfect schools to take advantage of the benefits. agency of parents. Many of the cash-for- The result was a 50 percent increase in class school-attendance programs are large, rep- size--an increase that does not appear to resenting significant commitments of public have deterred parents nor has it led to a resources. The biggest are Oportunidades measurable negative impact on test scores. (previously PROGRESA) in Mexico, the This is at least suggestive that a reallocation Bolsa Escola in Brazil, and the Food for Edu- of the education budget--larger class size cation Program in Bangladesh.26 with the savings used to pay for the inputs The budgets allocated to these programs under the program--could raise school are between just under 1 percent of total participation at no cost to quality.25 government current expenditure in Brazil Eliminating user fees for basic schools and more than 5 percent in Bangladesh. has also been shown to boost student These significant, but not prohibitive, sums enrollment, but quality may be compro- could be generated from savings on other mised if reliable alternative sources of expenditures, such as regressive subsidies financing are not available to schools (box for public services, including tertiary edu- 7.2). In both Tanzania and Uganda, elimi- cation. A question remains about how cost- nating school fees became an important effective the programs are in expanding political issue when the population could education: the answer depends on how suc- voice its discontent, helped by the demo- cessful they are in reaching households that cratic process, an active civil society, and (in would not have participated in the school Tanzania) the Poverty Reduction Strategy system without the transfers. Paper process. A careful evaluation of PROGRESA In some cases, there may be a need to go found an average increase in enrollment of beyond removing the direct financial costs 3.4 percent for all students in grades one of schooling to induce poor parents to enroll through eight, with the largest increase their children. This could be accomplished (14.8 percent) for girls who had completed by providing CCTs and free meals. CCT pro- grade six.27 Morley and Coady (2003) esti- grams make payments to poor families, typ- mate an internal rate of return (taking into ically mothers, on the condition that chil- account the cost of grants) for the program dren attend school regularly. The programs of 8 percent a year and report that the can be seen as compensating for the oppor- transfers are 10 times more cost-effective 138 WORLD DEVELOPMENT REPORT 2006 than building schools. But De Janvry and put forward proposals for reaching disabled Sadoulet (2004) find that most program children. In the two years since the fund was benefits are received by those who would set up, 6 percent of all schools in Uruguay have gone to school anyway. They suggest have been awarded grants to cover expenses calibrating transfers to increase program to adapt school materials, equipment, and efficiency--for example, through larger infrastructure and to train teachers in transfers to the eldest child, to children with appropriate pedagogical approaches. an indigenous father, or to children, espe- Improving gender equity in access to cially girls, living in villages without a sec- schooling often requires making special ondary school. provisions for girls, especially older girls.28 Specific grants for girls have been effective Reaching excluded groups. Schools with in Bangladesh and Mexico. Private latrines adequate supplies and well-trained and for girls are essential. Other structural motivated teachers, who are accountable for improvements including boundary walls, the learning they produce, are good for flexible or double sessions when sharing a everyone. But additional support may be facility with boys, and perhaps even gender- necessary to improve access for excluded specific schools may allay parents' concerns groups, such as disabled children, girls, and about girls' privacy and safety. It is impor- indigenous groups. tant for schools to undermine, not under- Including disabled children is possible at score, stereotypes and unequal treatment of relatively modest costs. In Uruguay, grants women--and to be wary of giving boys of up to $3,000 are awarded for schools that more resources, leadership, and attention. Female teachers are good role models for boys and girls, and even young women can be effective teachers with training, support, B O X 7 . 3 Desegregating Roma schools in Bulgaria: and a programmed curriculum. Govern- the Vidin model ments might consider setting national goals In Vidin, the Open Society Institute and the At the end of the first semester, atten- for hiring women and being flexible with Roma NGO known by the initials DROM dance was 100 percent, and first-term final- age and education requirements for female have been collaborating to integrate Roma grade averages were identical to those of teachers (while still providing adequate in- students into the mainstream school non-Roma pupils. Parents and teachers service training).29 system.Vidin is a town of 85,000 in north- were satisfied, especially with the absence west Bulgaria, and 6 percent of its popula- of reported incidents of anti-Roma To expand access for ethnic groups, teach- tion was identified as Roma in the 1992 cen- prejudice. Education authorities were ers or teacher aides from the target ethnic sus. In the 2000­01 school year, 460 Roma encouraged to scale up in other cities. In group are particularly helpful in their ability students, or half the school-age students, addition, 35 Roma parents of the bused chil- to connect with the students as powerful role were enrolled in the mainstream school sys- dren returned to school in adult education tem. Students are bused from the programs, and three teenagers who had models. Bilingual schools have also been settlement to school and back. And Roma dropped out in the third grade asked to join effective. In Mali, bilingual programs were monitors interact with parents and the the program, prompting teachers to work associated with large declines in dropout and school to encourage attendance. Low- extra hours with them. On the negative side, repetition, and rural students outscored income students also receive shoes and 24 pupils received failing grades in one or school lunches, with lunches given on the more subjects, and three left the program. urban children. In Mexico, geographic target- bus to reduce the stigma of receiving it at The success is attributable to three ing under PROGRESA (now Oportunidades) school. major factors. First, parents feel that their led to the relatively high participation of While preparing the program, DROM children are protected from prejudice went door-to-door in the Roma settlement because they are bused and monitored indigenous people (but not those in the most and sought the support of the schools, the throughout the day by adult Roma. Second, remote areas without schools).30 An innova- mayor, and the media.The project eventu- Roma monitors in the schools assure that tive approach to encourage the attendance of ally gained the support of all the stakehold- the children are not mistreated, encourage Roma children in Vidin, Bulgaria, appears to ers except the mayor, who nevertheless parental engagement and student partici- agreed not to block it. DROM invited the six pation in extracurricular activities, and help have paid off (box 7.3). mainstream schools in Vidin to present the the teachers ease cultural differences.Third, program, philosophy, and teachers on tele- the children are happy to be in schools Upgrading quality vision. Roma parents then selected a school where real learning takes place. Better quality for all. Expanding access to for their children.This marked the first time that their views had been solicited by the basic education is necessary but not enough; authorities. Source: Ringold, Orenstein, and Wilkens (2005). the quality of education matters for op- portunities. But even children in middle- Human capacities 139 income countries do a lot worse than the Figure 7.4 Boosting enrollments is not enough to average OECD kid on international tests of overcome the learning gap learning achievement, suggesting that much As a fraction of cohort of the learning in schools does not prepare Completed grade 9 but did poorly on tests children to be productive adults, let alone Dropped out before completing grade 9 for the rigors of competition they will face Dropped out before completing grade 5 in the global labor market.31 The quality Never enrolled 100 deficit is undoubtedly greater for children 90 from poorer families, because the better-off children can go to better public schools or 80 leave the public system and opt for private 70 schooling. 60 Based on the results of a standardized 50 international achievement test--the Third 40 International Mathematics and Science Study 30 (TIMSS)--Pritchett (2004a) estimates that 20 the overwhelming majority of children ages 10 15 to 19 lacks education (not completing 0 grade nine or performing poorly in the Colombia Indonesia Morocco Philippines Turkey TIMSS) in five middle-income countries Source: Reproduced from Pritchett (2004a). with data (figure 7.4).32 But the enrollment a. Based on TIMSS-R scores for eighth graders on mathematics in 1999. To calculate the fraction of students with scores below 400-- problem remains large only in Morocco. one standard deviation (100 points) below the OECD median of Indonesia and Turkey have difficulty retain- 500)--Pritchett uses the country mean and standard deviation and assumes a normal distribution. This assumes that scores are ing kids in secondary school; in Colombia, roughly constant over time so the 1999 test represents the cohort ages 15­19 in the survey year, and that eighth- and ninth-grade Morocco, and the Philippines, three of four competencies would be roughly similar. children who have completed grade nine have failed to learn enough. How can countries improve basic learning outcomes for all? We know broadly from a indicate that increased availability of text- large number of studies that have tried to books helps improve test scores, but only account for the "production" of schooling among the better-performing students, and outcomes that higher public spending does that performance-based prizes for teachers not always translate into better student learn- increased test scores initially, but the gains ing.33 A recent study analyzing the determi- dissipated later. What did work in raising nants of student performance on the test scores were merit scholarships for 13- TIMSS--using data for more than 260,000 to 15-year-old girls--with positive effects students from 6,000 schools in 39 countries-- also on learning for boys, who were ineligi- finds that education spending (spending per ble, and girls with low pretest scores, who student, class size, student-teacher ratio) at were unlikely to win the scholarships. The either the school or country level has no posi- scholarships were the most cost-effective of tive impact on student performance. Among all the interventions tested, achieving the factors at the school level, the only ones that same learning results at less than 20 percent have a significant impact on student perform- of the cost of textbook provision.35 ance are instructional material and teachers The results underscore the importance of with an adequate formal education.34 combining additional spending (of the right These results are confirmed by several kind) with interventions that strengthen careful microlevel studies. Since 1996, a incentives to teach and to learn.As the teacher group of researchers working with a Dutch incentive program shows, project design--in NGO, International Chirstelijk Steunfonds this case, the behavior rewarded--matters.36 Africa, has been involved in the design and evaluation of a series of randomized exper- Better quality for the most disadvantaged. iments to improve learning outcomes in the Many of the programs just discussed focus Busia district in rural Kenya. The results on improving performance at the school 140 WORLD DEVELOPMENT REPORT 2006 about half the cost of private secondary B O X 7 . 4 Remedying education: schools; vouchers could be renewed as long as the Balsakhi program in India students maintained satisfactory academic performance. An evaluation of this random- The Balsakhi program is a large remedial program has highly significant positive ized natural experiment (vouchers were education program now implemented in 20 results on student learning. On average, the awarded by lottery) found lower repetition Indian cities by an NGO--Pratham--in col- program increased learning by 0.15 laboration with the government. Pratham standard deviations in the first year and rates and higher test results among voucher hires young women from the community to 0.25 in the second year.The gains were winners.39 But targeted voucher schemes may teach basic literacy and numeracy skills to largest for children at the bottom of the dis- be politically difficult to implement--and the children who reach the third or fourth tribution, with those in the bottom third grade without having mastered them. Stu- gaining 0.20 deviations in the first year, and Colombia program was discontinued. dents are pulled out of regular classes for 0.32 in the second year (0.51 for math two hours of the school day for the reme- alone).The results were similar in the two Strengthening accountability dial education.The program is inexpensive: grade levels and in the two cities. At the Dismal learning outcomes in many coun- $5 per child per year. Easily replicable, it has margin, extending this program would be been scaled up rapidly since its inception in 12 to 16 times more effective than hiring tries are due to the combination of inade- Mumbai in 1994, now reaching tens of new teachers. quate resources and the low responsiveness thousands of students in the 20 cities. and accountability of school systems. A recent two-year randomized evalua- Efforts to improve school performance will tion in Mumbai and Vadodara finds that the Source: Banerjee and others (2004). therefore need to focus on strengthening accountability processes: achieving societal consensus on expanding education, dealing level. What about improving learning out- with capture by vested interests, and tackling comes for disadvantaged or poorly per- the weak incentives for service providers to forming students? The merit scholarship raise the quality of learning. program for 13- to 15-year-old girls in Achieving societal consensus on expand- rural Kenya mentioned earlier is one such ing education helps tackle the pathology of example. The Balsakhi program in India-- elite capture, whereby the wealthy oppose a large remedial education program--rep- increased spending in public education. His- resents another highly successful and cost- torically, expansion of voice in a country has effective approach to giving poorly led to wider access and quality improve- performing students a leg up (box 7.4). ments in basic education, notably in Europe Because children with the lowest ability and North America.43 Democratic transi- registered the largest gains in test scores, tions have spurred recent expansions in basic the program had an equalizing effect on education also in Brazil, Guatemala, and student achievement. Uganda.44 But these are long-term processes, Many countries group students together and it is essential to make progress now by similar abilities on efficiency grounds. toward meeting the urgent needs of millions However, recent findings in 18 to 26 coun- of children around the world. tries show that such tracking increases edu- Some progress can be made by counter- cation inequality, possibly by reinforcing ing the stranglehold of interest groups on the effects of family background, but it equity-enhancing reforms, such as when does not contribute to higher mean levels teachers unions block reforms that would of performance.37 strengthen the link between performance Another option to improve learning out- and accountability.45 Significant payoffs comes for disadvantaged children is to pro- can come from systemic reforms that vide school vouchers. There is significant strengthen accountability from clients controversy around the equity and efficiency directly to frontline providers.46 The most impacts of generalized voucher schemes (box crucial steps in any such reform are to 7.5). Targeted means-tested voucher pro- increase the schools' accountability for per- grams may be more promising.38 Results formance and to ensure the availability of from one such scheme in Colombia are relevant information to monitor their per- encouraging. The PACES program provided formance. Accountability for performance more than 125,000 students from poor also requires autonomy to manage results. neighborhoods with vouchers that covered This means delegating responsibility and Human capacities 141 B O X 7 . 5 School vouchers: efficient and equitable? School voucher programs increase the power of choice imply that weak public schools will lose equity effects of universal voucher programs. parents to choose schools for their children. Par- students and could be forced to close. Success- They could lead to increased racial and socioe- ents are given a voucher by the government, ful schools would have to be enlarged, or new-- conomic stratification of schools as parents seek which (at least in theory) can be applied to the and presumably more effective--schools would to improve the quality of their children's peers school of their choice, public or private.The have to be built. Such institutional change pres- (such as middle-class flight in Chile). Such strati- expectation is that competition among schools ents significant political, technical, and adminis- fication could occur if all parents were given and the availability of public resources to access trative hurdles.The hurdles are particularly vouchers but low-income families were in a less private schools would improve the overall effi- acute under a universal voucher program that favorable position to exercise choice because of ciency of the school system and student enables large student migration. lack of information, prohibitive transportation achievement. But research into the impacts of Solid evidence on productivity differences costs, or extra fees. Disadvantaged students vouchers has not produced definitive or gener- between public and private schools is also lack- would simply be more concentrated in low- alizable results--in large part because of ing. Again consider Chile, whose voucher pro- quality schools. Echoing similar concerns, a methodological challenges and the differences gram generated a large number of new secular recent study concludes that, in the United in the specific design and institutional context private schools that operated alongside more States,"a large-scale universal voucher program of various reforms. Design can vary according to established Catholic schools. An analysis of would not generate substantial gains in overall the size of the voucher, the pool of eligible stu- Chilean fourth-grade achievement data showed student achievement and ... it could well be dents, whether schools can charge more than that Catholic schools had higher achievement detrimental to many disadvantaged students" the value of the voucher, and regulations gov- than public schools in math and Spanish, while (Ladd 2002, 4). erning school choice (such as whether or not secular private schools had lower achievement. There are ways to make voucher programs religious schools are eligible). Institutional man- Another study found that unrestricted nation- more beneficial for disadvantaged students, but agement, bureaucratic control, governance of wide school choice in Chile resulted in middle- these may reduce political support for such pro- public schools, and oversight of eligible private class flight into private schools, but without grams. For instance, vouchers and school choice schools also vary and influence the results of achievement gains.42 can be limited to low-income families. Program programs.40 Evidence on peer effects that could influ- design can also be enhanced by providing Chile has more than 20 years of experience ence student achievement is equally inconclu- transportation to school, requiring that schools with large voucher programs.Yet detailed analy- sive. It is not clear whether peer effects are lin- do not charge extra tuition or fees on top of the sis on the effects of competition on school qual- ear, meaning that gains for students who move voucher, and requiring oversubscribed schools ity in Chile has not led to a consensus on to a higher-quality peer group are offset by to select students randomly. Irrespective of impact. In the United States, one study found losses for either their new or old classmates-- design specifics, a voucher program needs to be that competition improved achievement in the or nonlinear, meaning, for instance, that posi- embedded in a larger strategy of education city of Milwaukee, while another found no tive peer effects can disproportionately benefit reform that improves the overall institutional impact. Similar variance is found in the related students with low socioeconomic status. incentive environment for schools and gives literature on the impact of school choice.41 While impacts on efficiency are ambiguous, underperforming schools the instruments and Competition between schools and school there are reasons to be cautious about the resources to improve. power for decision making to the lowest local communities. As a result of greater feasible level consistent with incentives.47 parental involvement, Educo schools had Once the responsibilities of the school rapid enrollment increases without giving system are well defined, the resources and up quality, reduced absenteeism among decision-making powers of providers are teachers and students, and increased math consistent with their responsibilities, and and language scores. information is available to track perform- ance, various mechanisms become avail- able to pressure schools to deliver better Toward better health for all performance. School autonomy, commu- The large inequalities in health care use nity control, nongovernment providers, and health outcomes in many developing voucher programs, and public sector countries do not just reflect different pref- reforms can strengthen the ability of citi- erences or needs--they arise from con- zens, communities, and public organiza- straints on the ability of individuals to tions to hold schools accountable for deliv- achieve good health (chapter 2). Income is ering results.48 El Salvador's experience one important constraint, especially given with rebuilding much of its education sys- incomplete financial markets. Low-income tem following the destructive civil war of people around the world have worse health the 1980s is a good example of what can be and use fewer health services (chapter 2). accomplished through partnership with Ethnicity, race, and location also influence 142 WORLD DEVELOPMENT REPORT 2006 outcomes. Infant mortality rates among affordable care, and enhancing the account- blacks in South Africa are 5.5 times higher ability of providers. than those among whites; life expectancy among the rural Chinese is almost 6 years Expanding knowledge lower than among urban dwellers, while the Underinvestments in health by patients life expectancy gap between China's richest may reflect a lack of knowledge and agency and poorest provinces (Beijing and and incentive issues within the household, Guizhou) is 10 years.49 as well as a lack of these resources. Lack of These stark differences in outcomes and knowledge can keep people from seeking use reflect large group-based inequalities in care when they need it, even when price is access to information, facilities with rea- not an issue. As chapter 5 showed, when sonable standards of care and financial deworming medicine was offered free to protection from health risk. A lack of children in Kenya, the take-up rate was only knowledge about hygiene, nutrition, avail- 57 percent. Similarly, in Bolivia, many poor able services, and treatment options, par- babies are not delivered by a trained atten- ticularly among the uneducated, lowers dant even though mothers are eligible for demand for health services. Within the free care. In India, 60 percent of children household, some family members have less have not been fully immunized, although voice (women and children) and this can immunization is free; mothers cited igno- affect the level of resources used in their rance of the benefits of vaccination and not interest. Health clinics, especially in poor knowing the clinic locations as the major and remote areas, are often inaccessible, reasons for why their children had not been have high rates of absenteeism and low qual- immunized. ity and responsiveness to clients. Finally, ill- Lack of knowledge can also lead people ness is certainly a burden on poor people, to pay for inappropriate care. Unqualified or but catastrophic health shocks can also unethical providers can overprescribe treat- have disastrous consequences for the not so ments for patients who do not know what is poor, mainly through loss of income but in their best interest. For instance, instead of also through high out-of-pocket payments effective and inexpensive oral rehydration for health care. therapy, a poor child in Indonesia gets more These large group disparities in health than four (often useless) drugs per diarrhea outcomes are inequitable, because they attack.50 imply vastly different opportunities to lead Education is a natural way to address the productive lives. And because they often lack of patient knowledge. Elo and Preston arise from failures in markets and agency, (1996) estimate that one year of extra edu- reducing these disparities would have large cation nationally reduces mortality rates by payoffs in efficiency and productivity. We about 8 percent--half directly and half focus here on ways to level the playing field through the effects of additional earnings. for attaining good health by boosting peo- Female education is particularly powerful. ple's knowledge about basic health practices Better-educated mothers are associated with and services, expanding their access to better child-health practices, including hand washing, proper disposal of feces, antenatal care, delivery assistance by trained person- B O X 7 . 6 Working with mothers to treat malaria nel, immunization, and well-baby clinics. Community health agents also provide Malaria kills nearly 1 million children in chloroquine and information on how to cost-effective instruction in disease preven- Africa each year. Empowering mothers to administer the drug at a cost of $0.08 per tion and healthy behavior. By employing take actions to treat their children in the child treatment dose. By educating mothers, home can be highly effective in reducing Tigray provided rapid and effective these nonspecialized personnel, many coun- mortality.The Tigray region of Ethiopia treatment without forcing the child to relo- tries have increased knowledge among the trained"mother coordinators,"who were cate, which reduced under-five mortality by general public at low cost, as with Brazil's selected from among the community to 40 percent and alleviated the burden of Family Health Program and Ethiopia's educate other mothers on the symptoms of severe malaria cases on hospitals. fever and malaria. Mothers were provided Source: World Bank (2004k). "mother coordinators," supporting home- based malaria treatment (box 7.6). Com- Human capacities 143 munity health workers have also helped greater burden on rural dwellers through increase coverage of poor populations cost- additional travel time and hospice costs. effectively. City dwellers are within easier reach of Public information campaigns can im- health centers. In Burundi, 98 percent of the prove health knowledge by working through urban population was within one hour of a existing health clinics or by directly targeting health center, but only 65 percent of the the community. It is also possible to collabo- rural population was. Even within rural rate with the private sector in marketing areas, there is large variation. Only half of socially valuable products, such as insecticide- the poorest rural Nigerians were within an treated mosquito nets, water purification hour of a clinic, but 84 percent of the rich- methods, foods rich in vitamin A, and est were. soap--as with the Central American Hand- Even when health facilities are accessible, Washing Initiative in Costa Rica, El Sal- they vary hugely in quality. Some have med- vador, and Guatemala.51 Media campaigns icines and drugs in stock, are run by well- can also be effective. For instance, frequent trained and motivated staff, and are well broadcasts of AIDS messages in Thailand, maintained. But many are not. They are Uganda, and Brazil were a key element in the often dilapidated, rarely have medicines in campaign to reduce the spread of the dis- stock, and are run by poorly trained and ease. The Thai media campaign is credited rude medical staff, who frequently fail to with reducing the incidence of AIDS to a come to work. It is often precisely the peo- point at which the country is now able to ple who are materially disadvantaged who consider a fiscally viable treatment program also have to struggle with poor quality and for AIDS patients.52 But neither information nor free services may be enough to boost use among the less B O X 7 . 7 Poor people and ethnic minorities empowered or those without voice. Mater- receive lower-quality care nal and child health is often viewed as merit- ing additional intervention. Through condi- New studies from India,Indonesia,Mexico, Mexico,indigenous women receive prenatal tional transfers, Mexico's PROGRESA (now and Tanzania demonstrate that the poor sys- care from doctors who rank only in the tematically receive lower-quality care from twenty-fifth percentile in quality,while Oportunidades) program was designed to private and public providers.54 The situation is equally poor nonindigenous women receive encourage women to attend pre- and post- often worse for ethnic minorities.Evidence care ranking in the fortieth percentile.The natal clinical visits and bring their children from Mexico suggests that,even in poor rural wealthiest fifth fare much better,but even villages,there is a difference in the quality of among the wealthy,the indigenous receive for immunization and growth monitoring. care between wealthy and poor and between worse care than the nonindigenous,suggest- The program saw an 8 percent increase in indigenous and nonindigenous groups. ing that discriminatory practices or cultural clinic visits by pregnant women in their first Among the poorest fifth of the population in barriers may be at play (see figure below). trimester, which led to a 25 percent drop in Indigenous Mexicans receive lower-quality care, regardless of income the incidence of illness in newborns and a 16 percent increase in the annual growth rate of Doctor quality (percentile) children between one and three. An impor- 70 tant design feature of the program is trans- Nonindigenous 60 ferring funds to women. Although the pro- gram puts more demands on mothers' time, 50 Indigenous participants felt that the benefits were worth 40 it. Women also reported feeling more self- confident and having more control over 30 household resources and their time and 20 travel. Similar schemes are delivering mater- 10 nal and child care services in Brazil, Colom- bia, and Nicaragua.53 0 Poorest Second Average Fourth Wealthiest quintile quintile quintile quintile Expanding access Access to quality health facilities remains a Source: Barber, Bertozzi, and Gertler (2005). problem in many areas, often imposing a 144 WORLD DEVELOPMENT REPORT 2006 inaccessible health facilities. Ethnic minori- in poor areas may not be affordable for many ties often fare even worse in terms of the poor countries. But there are other quality of health care received (box 7.7). approaches to reducing the indirect costs An important obstacle to the provision of (for transport and time spent in transit) and equitable health services everywhere is the the medical isolation of poor communities. difficulty of enticing urban-educated doctors Roving extension clinics visit sparsely popu- to work in poor areas. Chile, Mexico, and lated areas in Afghanistan, Somalia, and Thailand have used financial and other Tunisia to provide care locally and offer incentives to encourage qualified staff to transportation to better-equipped facilities work in rural areas.55 In Indonesia, doctors when required. Bangladesh, Cuba, Gambia, had to complete compulsory service in health India, and Madagascar have trained commu- centers before they could obtain a lucrative nity health workers to scale up service deliv- civil service post. Compulsory service was for ery for a wide range of services, including five years, with shorter periods allowed for malaria prevention, immunizations, family work in remote provinces. This system planning, treatment of TB patients, home increased the number of doctors in health visits, and neonatal care. The results have centers by an average of 97 percent from 1985 often yielded a substantial increase in cover- to 1994, with gains of more than 200 percent age and measurable improvements in out- for the most remote provinces.56 comes at much lower unit costs. Through a Expanding rural health infrastructure village health worker program that moni- and providing incentives to doctors to work tored infants' weight and health for the first month of life, rural infant mortality in Maharasthra, India, was cut in half from 75.5 B O X 7 . 8 Better maternal health in Malaysia to 38.8 per 1,000 live births between 1995 and Sri Lanka and 1998.57 Other outreach programs focus on Despite huge improvements in health,sur- (in Malaysia) and transportation subsidies maternal health and safe delivery. By mak- vival,and fertility around the world in recent (in Sri Lanka) were provided for emergency decades,global maternal mortality has not visits to the hospital. In Malaysia, health pro- ing professional midwives and supervisory declined significantly.Two exceptions are Sri grams were part of integrated rural devel- nurse-midwives widely available in rural Lanka and Malaysia.In Sri Lanka the mater- opment efforts that included investment in areas, Malaysia and Sri Lanka dramatically nal mortality ratio--the number of maternal clinics, rural roads, and rural schools. Simi- reduced maternal mortality rates (box 7.8). deaths per 100,000 live births--dropped larly, in Sri Lanka, the government invested from 2,136 in 1930 to 24 in 1996.In Malaysia in free primary and secondary education, In Bolivia, expectant mothers with high it dropped from 1,085 in 1933 to just 19 in free health care, and food subsidies for all obstetric risk are transported to larger clin- 1997.What can account for this impressive districts.The concept was that basic health ics a few days before their due date; in Sri decline? Improving access for rural and dis- care acts in synergy with education and Lanka, they are picked up by radio- advantaged communities was an important other types of infrastructure. For example, part of the strategy in both countries. better roads make it easier to get to rural dispatched four-wheel-drive vehicles. Sri Lanka and Malaysia made health facilities and facilitate transportation competent, professional midwives and of obstetric emergencies. By addressing the Financing affordable care supervisory nurse-midwives widely multidimensionality of equity, these coun- For consumers, health care finance systems available in rural areas. Midwives assisted tries made significant health gains. deliveries in homes and small rural hospitals Dramatic improvements in maternal have two goals: affordable access to a basic and performed initial treatment in the mortality are thus possible. Just as impor- package, and financial protection in the event of complications.They were given a tant, the experiences of Malaysia and Sri event of catastrophic illness costs. The clas- steady supply of appropriate drugs and Lanka show that these can be attained with equipment and supported by improved only modest expenditures. Since the 1950s, sic case for government intervention (pub- communication, transportation, and back- public expenditures on health services have lic subsidies) is when the full benefit of a up services. Besides reducing financial and hovered between 1.4 and 1.8 percent of "treatment" accrues not just to the individ- geographic barriers, they also helped over- GDP in Malaysia and averaged 1.8 percent ual but also spills over to the community come cultural obstacles and allegiances to in Sri Lanka, with spending on maternal and traditional practices. Because midwives child health (MCH) services amounting to more broadly. Interventions to avoid the were available locally and were well less than 0.4 percent of GDP in both coun- spread of malaria fall into this category. A respected, they developed links with com- tries. Countries with similar income levels bed-net distribution program--involving munities and partnerships with traditional have significantly higher health the Red Cross and national ministries of birth attendants. expenditures and similar, if not higher, Malaysia and Sri Lanka pursued other maternal mortality ratios. health--increased use among the poorest complementary strategies.Transportation Source: Pathmanathan and others (2003). quintile from 3 percent to nearly 60 percent in a northern district of Ghana and from 18 Human capacities 145 percent to 82 percent in five rural Zambian events.61 If so, poor people may be better districts.58 Immunizations, vector control, served by having protection against these and interventions for tuberculosis, HIV/AIDS, low-risk, high-cost events through some and other communicable diseases are simi- type of pooling mechanism. It is no easy larly deserving. task, however, to cover catastrophic health But the case for government interven- risks in ways that reach the poor. tion goes beyond these well-accepted public Reducing out-of-pocket costs involves a health reasons: inequality in access to finan- combination of pooling health risks and pre- cial protection from health risk based on payment--through contributory insurance wealth, ethnicity, and location provides schemes, national health services that are another important rationale. Out-of-pocket funded out of general revenues, or a mix of payments are the dominant form of health the two. In all instances, reaching the poor care finance in lower-income countries. But requires some means of subsidizing their liquidity constraints and imperfect credit health care costs, so fiscal room and political markets often make out-of-pocket pay- commitment are crucial. In very low-income ments more difficult for the poor, reducing countries, community insurance schemes, their use rates, and health and productivity. sometimes supplemented through NGO or In Vietnam in 1998, before the establish- donor funding, can provide some protection ment of health insurance, 30 percent of to some people, but generally these services poor households' nonfood budget went to do not reach the poorest. medical costs, while only 15 percent of Contributory schemes--private or social-- spending for the richest fifth of the popula- operate best where the share of the formal tion was health related. In Cambodia, a sin- labor market is high and administrative gle hospital stay absorbed 88 percent of an capacity is strong. And because premiums average household's nonfood consumption and copayments can be unaffordably high, in 1997 and, for the poorest among them, purely contributory schemes generally bypass the cost was higher than the entire nonfood the poor. Private insurance is a significant budget. In the transition economies of part of health finance systems in Brazil, Europe and Central Asia, with the collapse Chile, Namibia, South Africa, the United of prepayment in the 1990s, out-of-pocket States, Uruguay, and Zimbabwe. But in all spending skyrocketed, accounting for as seven countries, private insurance is used by much as 80 percent of health resources in formal sector workers, leaving the ministries Georgia and Azerbaijan. In Armenia, 91 of health to provide public funds for pro- percent of patients reported having to make grams for the poor and underserved.62 some payment for service received.59 While Social insurance is characterized by health care use has plummeted in the compulsory coverage financed by employ- region, the collapse of prepayment espe- ment taxes. Benefits are often limited to con- cially hurts poor people. tributors, and providers are often from the The regressive nature of out-of-pocket public sector even when private providers payments is well understood, but there are are eligible. Social insurance has the appeal no easy answers, especially in low-income of generating a large risk-sharing pool and countries. Given the small formal sector can, in principle, reach the poor through and limited administrative capacity, these cross-subsidization. But, when the formal low-income countries have limited capacity sector is small this potential is limited, to mobilize resources to pay for essential because of the difficulties of enrolling a large health services and to establish large enough share of the population. This can enough risk pools. So, developing countries turn the system into a ticket to privileged face a difficult tradeoff between providing a access to health services for some, while basic package of health services and extend- leaving the bulk of the population under- ing financial protection.60 Some evidence served. For example, in Mexico, social secu- suggests that the poor are better able to rity health spending per person is five times cover low-cost, high-frequency health higher than what the Ministry of Health shocks than low-frequency, high-cost spends per person.63 And, the payroll tax 146 WORLD DEVELOPMENT REPORT 2006 required for social insurance introduces 1,000 births to 15) among the insured. But labor market distortions, especially in set- there are questions about the program's tings characterized by dual labor markets. sustainability in the face of mounting fiscal The challenges from both an equity and cost--reflecting the difficulty of systemic efficiency perspective are enormous, but a reforms that threaten the privileges of handful of mainly middle-income countries established interests, in this case, public have made important attempts to make hospitals and the prereform social security social insurance systems work. Colombia, for institution.64 example, has a cross-subsidization scheme Ministries of health in many developing for the poor, topped up by general revenues. countries operate essentially as national The scheme has delivered considerable ben- health services, with nationally owned health efits: higher coverage among the poor (48 sector inputs and funding from general tax percent, up from 9 percent, in 10 years); revenues. The systems they manage are often lower out-of-pocket costs for ambulatory inefficient and inequitable, reflecting severe care; large increases in physician-assisted resource and institutional capacity con- delivery (by 66 percent) and prenatal care straints but also a bias in favor of the wealthy among rural women (by 48 percent); and and influential. Services are meant to cover lower child mortality rates (from 44 per everyone, but high out-of-pocket payments keep many poor people from participating. Countries have tried various approaches to improve equity in access to health care pro- B O X 7 . 9 Mobilizing support for universal coverage vided by the national health system, such as in Thailand eliminating user fees for all, waiving fees, or giving vouchers to poor people. With the introduction of the universal cover- Years of corruption and political In 2001, Uganda abolished user fees for age scheme in 2002,almost the entire Thai inequity in the early 1990s, together with all. The result was a significant increase in population now has health coverage (box the intervening period of military rule fuel- figure 1).This was possible largely because ing social discontent, prepared the ground health care use, lower probability of sickness, the democratic transition of 1997 ushered in for democratic reforms and a liberal consti- and better anthropometric measures, partic- a period of increased voice and openness tution in 1997.Two provisions of the new ularly for the poor.65 But the elimination of and raised the political profile of poor peo- constitution were important for the ensuing ple's concerns.Technical preparation--with health sector reforms: the principle of user fees, if effective, can reduce the resources design details that had been under consid- equity in health care access; and the scope for the health sector, and thus its quality, eration and subject to experimentation for for civil society to propose national legisla- unless budgetary funding is topped up to some time--also helped to garner support tion affecting citizen rights and the role of make up for the shortfall. Uganda appears to for the reform, while prior investments in the state, if the measure had 50,000 signa- health care infrastructure, establishing a tures. In March 2000, nearly at the same have avoided a fall in quality, thanks to a health center in nearly every rural subdis- time that a feasibility study for the scheme large increase in the health budget that more trict, provided assurance of implementation was completed, a network of 11 NGOs sub- than compensated for the loss in revenues success. mitted to parliament a draft bill calling for from eliminating user charges. universal health care coverage. Between Thailand's increasing coverage 1999 and 2001, the press also captured the Introduced in 2002, Thailand's "30 baht" general public's interest and kept the issue or universal coverage scheme aims to guaran- Uninsured on the political agenda by highlighting the Private health insurance tee health care to every Thai citizen. It com- shortcomings of the current health system. Universal coverage scheme bines previously existing schemes targeted to This political foment attracted the Voluntary health card scheme the poor and uninsured, and allocates budg- Medical welfare scheme attention of the opposition Thai Rak Thai Social Security (TRT) party.TRT adopted the policy because etary resources to providers on a capitation Government/state enterprise it was broadly supported, administratively basis, with only a small copayment per visit and technically feasible, and consistent with 100% (30 baht). The Ministry of Public Health the party's ideology.TRT effectively turned the 1997 financial crisis into an opportunity remains a strategic manager and central fin- 75% by highlighting the health issues ancier, but the district offices make the deci- precipitated by the crisis. Following its land- 50% sions on choice of providers. The scheme has slide victory in 2001,TRT introduced its own universal coverage legislation, which passed markedly increased use and coverage, with 25% in November 2002. roughly three-quarters of the country bene- 0% fiting from the scheme and 95 percent of the 1999 2002 2003 Source: Pitayarangsarit (2004). population insured overall, all at a limited additional budgetary cost (box 7.9). Human capacities 147 But there is broad agreement that exist- burden on populations in poorer or ethnic ing capitation rates are too low and that the minority regions.68 system is underfunded. This provides lim- Many developing countries have a mixed ited incentives for private providers to par- system, with ministries of health, private ticipate and could lead to a financial insurance, social insurance, and targeted squeeze on public providers with adverse schemes coexisting to serve different seg- impacts on quality.66 Higher use has also ments of the population. These multitrack put a strain on human resources, with systems tend to fragment, increasing admin- increased workloads (and low pay) acceler- istrative costs, limiting pool sizes, and under- ating the number of physicians leaving the mining both equity and efficiency objectives. public system. Clearly supply-side measures Chile's two-track universal coverage system need to be considered in tandem with has caused severe segmentation, with the health finance reforms to expand access. healthy and wealthy in the private scheme, Still, the achievements have been consider- leaving the public scheme overburdened able, and the scheme has broad popular with the poor and ill. Chile is trying to over- support. Thailand implemented the reform come this by creating a "virtual pool," man- in large part thanks to the popular support dating a common basic benefits package, that democratic reforms made possible; instituting catastrophic insurance, allowing previous investments in sound design and portability of benefits between schemes, and health infrastructure also helped. initiating minimum quality and maximum Other countries reduce costs for the wait-time standards. poor through targeted programs funded from general revenues. Armenia's targeted Community-based health insurance (CBHI) fee-waiver system curbed plummeting use schemes have developed in some poor com- among the poor. But in many instances, munities outside the reach of national simple legislation of free or reduced-price health systems. Communities pool health services can be counterproductive without risks through voluntary contributions to a good funding and targeting. Targeted pay- local fund used when any member incurs a ments, through government vouchers and health shock. The schemes are reported to civil society partnerships with hospitals, can reduce out-of-pocket spending and in- help. Vouchers issued to poor patients, as by crease use by their members, but they gen- the MCH poverty alleviation fund in Yun- erally do not reach the poorest and nan Province in China, give providers a socially excluded groups or offer members greater assurance of payment. In Cambo- enough protection from financial risk. dia, a promising partnership has emerged Many are limited by their small risk pools, between government hospitals, Médecins exposing them to low-frequency but high- Sans Frontières (MSF) and a small local cost catastrophic events that can outstrip NGO, covering the hospital fees of those the community fund. Some communities considered indigent by the local NGO's address large health risks by increasing the social workers. Because the hospital is fully maximum benefit, as in Cameroon. But compensated, poor patients receive the they do so by limiting the number of fam- same care as those who can pay.67 ily's claims to one a year and by requiring Vietnam has introduced health cards for high premiums (which prohibit the poor the poor. More than 11 million of 14.3 mil- from participating).69 lion eligible people benefited from Viet- Insurance alone is not enough for equi- nam's program in its first year of implemen- table use. Inadequate knowledge of the tation in 2002. The program has already scheme's benefits and processes and even the significantly increased the flow of govern- paperwork for submitting claims to commu- ment health funding to the poor and to pre- nity insurance schemes can be a deterrent. dominantly poor areas of the country. The Hospitals often require payment on or funding per beneficiary, however, is consid- before discharge, but insurance claims are ered inadequate and the cost-sharing not settled until later, requiring patients to arrangements are likely to impose too large a pay up front. India's Self-Employed Women's 148 WORLD DEVELOPMENT REPORT 2006 Association (SEWA) has been seeking to network. NGOs are particularly helpful in remedy similar difficulties encountered in its serving remote areas and hard-to-reach pop- large and well-established CBHI scheme. ulations: the Bangladesh Rural Advancement SEWA is testing door-to-door visits for Committee (BRAC) trains community work- member education, reimbursement assur- ers to seek out the extremely poor in need of ance with selected hospitals, and reimburse- urgent medical care. In Jordan, half or more ment to members while still in hospital.70 of outpatient visits are to private providers.72 Many private providers offer excellent serv- Enhancing provider incentives ices. But some do not--and misdiagnose, misprescribe, or overprescribe treatment. In Addressing knowledge, access and afford- Mexico, even wealthy women receive worse ability constraints are important, but they care from private providers than from public may not be enough to raise health use and providers (Barber, Bertossi, and Gertler outcomes. Hours of operation, waiting time, 2005). Without unduly discouraging benefi- staff disposition, competence and integrity, cial private enterprise in health, governments and the cultural appropriateness of services need to ensure accreditation and appropriate are all important. Complaints of unprofes- regulation for nonpublic providers. sional treatment, abuse, and corruption abound worldwide. Public medical staff Social protection: managing risk who take authorized or unauthorized leave from public clinics undermine the credibil- and providing social assistance ity of the public health sector, drive up costs Social protection policies typically have for poor families, and induce the poor to use been thought of as a form of redistribution. private providers, including traditional heal- This certainly is important. But more recent ers. In Bangladesh, such absenteeism rates theoretical and empirical work also high- amount to 40 percent for physicians in lights a crucial opportunity-enhancing role larger clinics and 74 percent in smaller sub- for social protection.73 As chapter 5 showed, centers with a single physician. More gener- pervasive financial market failures in devel- ally, poor service delivery has to do with oping countries lead to widespread unin- weak management and incentives within the sured risks and credit constraints. Unequal public health system--ineffective technical capacity to manage risk means unequal and structural backup, lack of professional opportunities to engage in risky but high- career structures, and inadequate financial return activities. Families may deal with incentives all contribute. But the weak crises in ways that narrow future opportu- demand for service provider accountability nities, such as distress sales and forgoing and quality is also a problem.71 health care, schooling, or food intake. By If they are organized, poor citizens and helping poor people manage risks, social communities can have more voice and protection programs expand their opportu- greater power to influence health providers. nities and enhance overall efficiency. Governments can help support organization Even purely redistributive programs can by communities and enhance provider have important opportunity-enhancing im- accountability. It helps to have well-defined pacts. Take the example of social pension objectives for health service delivery with schemes in Brazil and South Africa. These transparent metrics for monitoring progress. schemes are pure transfers targeted to the This allows for community oversight of elderly, geared strictly to avoiding destitu- health workers and facilities, and when cou- tion, but they have important welfare pled with sufficient management autonomy impacts beyond that. They improve the for providers to reach the established objec- recipients' access to credit, thanks to the tives, can lead to improved provider incen- regularity of pension payments, and lead to tives and accountability. higher investments in the household's phys- There is also a need for governments to ical capital and in the human capital of its engage with nonpublic health care providers: children and elderly.74 in many countries, NGO and private But social protection systems do more providers make up a large part of the health than help individual households avoid des- Human capacities 149 titution and expand their opportunities-- Figure 7.5 Almost all countries spend more on social insurance than on social they can also help societies embark on assistance (percent of GDP) reforms that would have insurmountable % of GDP equity and political costs without them. 14 Reforms desirable for their beneficial im- Social insurance 12 pacts on efficiency and the government's spending Social assistance fiscal position--such as increasing utility 10 spending prices, eliminating general food subsidies, 8 introducing a defined contribution pension system, liberalizing trade--may not be 6 politically feasible unless policies are in 4 place to compensate losers. Importantly, permanent social protection can help 2 reduce the need for special compensatory 0 programs for each and every reform75--all Sub- Middle East Europe East Asia South Asia Latin OECD Saharan and and and the America the more important because such programs Africa* North Africa Central Pacific and the are difficult to start and stop and are not Asia Caribbean always very efficient. Source: Data on 74 countries taken from World Bank Public Expenditure reviews or other similar work. All of this confirms that there is a OECD data are from the OECD Social Expenditure database. dynamic efficiency rationale for social pro- Note:* The average for Africa is based on data for only two countries. OECD excludes those OECD mem- bers (such as Poland and Mexico) that are already accounted for in the regional averages. tection. But there are also important effi- ciency arguments against transfer policies. Design issues are of particular concern, poor. These include a variety of cash or because poorly designed programs can have in-kind programs targeted at the poor. large negative consequences on efficiency. These are complemented by labor mar- Taxes or contributions have distortionary ket regulations (for example, on hiring and costs, especially when they are not directly firing of workers) that are discussed in linked to benefits (see focus 5 for a discus- chapter 9. There is large variation in the sion of tax policies), while transfers can share of GDP spent on social protection, dampen work incentives, reduce private sav- with more-developed regions devoting con- ings, and weaken informal insurance mech- siderable sums (figure 7.5). Almost all anisms. Europe's experience in the second countries spend more on social insurance half of the 1900s suggests that well-designed than social assistance programs. social (and tax) policies can indeed be con- There is no consensus on the appropriate sistent with strong growth thanks to careful balance of interventions--even in countries attention to productivity impacts.76 that have sufficient resources and capacity to implement any combination desired. Some Program choices vary by country observers argue for the universality of social Social protection generally encompasses two insurance programs over the targeted nature classes of interventions: of social assistance programs that are based on political economy considerations. They · Contributory schemes (social insurance) argue that targeted programs are exclusion- in which the primary focus is on manag- ary, by definition, and divisive as a result.77 ing risks through smoothing an individ- But a significant group of OECD countries ual's income over time and in the face of (notably New Zealand, Australia, the United difficulties These programs often pool States, and the United Kingdom) have opted risks across large numbers of individuals for systems with heavier components of tar- and include old-age and disability pen- geted transfers and less generous or less uni- sions and health and unemployment versal programs.78 insurance. Many developing countries face con- · General tax funded transfers (social straints on the choice of systems because of assistance) in which the focus is on limited fiscal and administrative capacity. redistribution from the better off to the Many poor and even middle-income coun- 150 WORLD DEVELOPMENT REPORT 2006 Table 7.1 Examples of social protection programs Complementary Group served Social insurance Safety net Labor Market Regulation Working poor or unemployed Unemployment insurance Transfers Minimum wage laws Public works programs Job security regulations General subsidies to food, utilities, or housing Severance pay Nonworking young Universal child allowances Means-tested child allowances Child labor laws Maternity benefits Transfers linked to MCH programs School feeding Conditional cash transfers Nonworking elderly Contributory pensions Transfers Retirement age Social pensions Special groups Disability insurance for disabled Transfers Affirmative action for minorities tries lack the administrative sophistication Programs for the working poor and levels of urbanization and formal Most people, especially poor people, rely on employment needed to administer a social labor earnings for their livelihoods, many in insurance scheme, and high social security the informal sector, through subsistence taxes have segmented the labor market and farming, or as agricultural laborers for others. encouraged informality. Labor market risk can be reduced signifi- What then are the options for the many cantly by improving the functioning of labor developing countries that are far from able markets and by pursuing sound macroeco- to achieve universal social insurance sys- nomic policies (chapter 9). But even a well- tems? There is a large range of social assis- functioning labor market will not fully elimi- tance programs, each different in groups nate the risk of unemployment. Moreover, in served, administrative requirements, com- years with bad crops or low prices, earnings plementary benefits, incentive effects, and may not be enough to stave off poverty. political factors (table 7.1). A judicious A range of instruments can help address blending of these programs can usually the risk of inadequate incomes--for exam- result in a social protection system that cov- ple, unemployment insurance, needs-based ers the appropriate groups with feasible social assistance, or public works. Even instruments. The mix of programs selected food, utility, and housing price subsidies are and their specific characteristics will depend geared in part to solving the problems of on context--that is, the risks faced, the level inadequate labor incomes, although with of urbanization, the age structure, the size of notably poor targeting and sometimes large the formal sector, the administrative capac- distortionary costs. ity, and the complementary social policies Unemployment insurance, the obvious and sociocultural or political factors. instrument for mitigating the risk of job Next, we discuss programs for four key loss in the formal sector, will not work well groups: in countries with large informal sectors. Even so, schemes may be able to cover a use- · The working poor ful share of workers and take some burden · The nonworking young off programs more tailored to those in the · The nonworking elderly informal sector (chapter 9). For example, in · Special vulnerable groups 1998 in response to the East Asian financial crisis, Korea expanded its young unemploy- In many cases, the second two groups are part ment insurance program to smaller firms as of households that could benefit from pro- well as to temporary and daily workers.79 grams that target the working poor. So the Needs-based cash transfers, the classic more comprehensive the programs for the social assistance instrument, are common first group, the less the need for programs for in high-income countries. Such programs the latter two, and the smaller and more are potentially very efficient. Nontransfer focused they might be. costs can be low, usually 5 to 10 percent of Human capacities 151 total program costs. The programs need not pare customized contracts with the house- impose significant forgone earnings on par- hold to address the most important of these ticipants. And they give cash and, thus full barriers over a period of two years. consumer sovereignty, to the recipient. Bangladesh's program of income generation But these programs face two challenges. for the development of vulnerable groups First, they require a targeting mechanism. (IGVGD run by BRAC) gives in-kind assis- The classic mechanism in high-income coun- tance to destitute rural women for a period tries has been a verified means test. Because of 18 months. During this time, they are income in these settings is mostly formal, it is required to save some money and participate possible and not too costly to collect accurate in business training. At the end of the cycle, information on income and assets. Eastern women have the opportunity of "graduat- Europe also has successful experience with ing" into the regular microfinance program. means-tested programs, although verifying A few programs, as in Romania and Bulgaria, incomes and assets is more difficult and less add a public service requirement (thus blur- accurate there than in high-income coun- ring the line between the means-tested social tries. Latin America's proxy means tests assistance and public works programs).82 So, (relying on easily observable indicators of whether through traditional or more inno- income) have been shown to be fairly accu- vative mechanisms, the disincentive problem rate and low cost. Low-income countries with can also be mitigated. large shares of income from the informal sec- Public works programs that support the tor experience greater difficulties in setting working or unemployed poor have been used up targeting mechanism. Although the evi- in many countries (box 7.10). By offering dence is less clear-cut, community-based sys- employment for low wages, these programs tems have been shown to work well in some self-select the able-bodied poor, avoiding countries around the world, especially in both the means-testing and work disincen- fairly homogenous rural communities tives. In good programs, the work is in high- (Albania, Bangladesh, Ethiopia, Indonesia, return activities that create assets and serv- Uganda, and Uzbekistan) where elite capture ices. The self-targeting aspect is useful is not a major concern. So the targeting issue because informality is widespread in develop- could be surmounted.80 ing countries and incomes are hard to assess. The second challenge is perhaps bigger-- It is doubly useful as part of a countercyclical both for technical design and political sup- measure in fighting poverty during periods of port. Needs-based transfers inherently pres- crisis--workers leave the program when their ent a disincentive to work for those of regular source of livelihood picks up again working age because entry into the pro- after the crisis. Public works programs for gram (or the benefit level) depends on infrastructure are especially welcome in low- income. Traditional mechanisms to partly income countries, postconflict settings, and mitigate the work disincentives include sometimes post­natural disaster settings. keeping benefits substantially lower than Public works programs have some disad- minimum wage, as in Bulgaria or Romania, vantages too. The administrative capacity to or lower than the earning of low-skilled select and run the programs is significant. agricultural laborers, as in Kyrgyz Republic, Indeed the often-cited good programs are a or using a sliding withdrawal of benefits as minority of all the public works programs incomes rise, as in much of the industrial implemented around the world. Many have world, or an earned-income tax credit as in failed, often over the inability to line up and the United Kingdom and United States.81 deliver useful public works, to provide suffi- A newer wave of efforts takes a more active cient nonlabor inputs, or to set the wage right. approach to encouraging independence or Even when programs are well run, the net "graduation" from the need for assistance benefits transferred to participants are often than under the traditional mechanisms just a small share of total program costs. First, mentioned. Chile's Puente program uses management, materials, equipment, and extensive social work to diagnose each house- skilled labor requirements can run up to 40 hold's barriers to independence and to pre- to 60 percent of program costs. Second, 152 WORLD DEVELOPMENT REPORT 2006 B O X 7 . 1 0 Public works programs: key issues Public works programs have been tional options as community-based selection of same setting.There is a tradeoff between imme- demonstrated to work in some middle-income the neediest or a lottery may be implemented. diate income gains through employment of the countries (Chile, Argentina, and South Africa) Least desirable is rationing with entry poor, and gains to the poor from the quality and and low-income countries (Senegal, Kenya, determined by foremen or political figures. durability of the assets created. In a crisis situa- India, and Bangladesh)--and not to work in Women's participation can be enhanced through tion, in which current transfers to the poor have many others.This international experience nondiscriminatory wages,the provision of onsite high weight, a high labor intensity is desirable. offers several lessons in the design and imple- child care,and adequately private latrines. Illustrative average labor intensities range from mentation of public works programs. Employment guarantee. A workfare program 0.5 to 0.65 percent in low-income countries and Wage rate. The key to self-targeting is setting that guarantees employment can reduce the are somewhat lower (0.4) in middle-income the wage rate low enough--no higher than the longer-term risk the poor face.While highly countries, although labor intensity often varies market wage for unskilled manual labor in agri- desirable, guarantees have not been a feature of significantly by subprojects. culture or the informal sector during a normal most public workfare schemes. One exception is Administration and implementation. Admin- year.While determining the precise level of the the Employment Guarantee Scheme in Maha- istering and implementing an effective scheme wage rate may not be easy,it is better to start rasthra, India, which guarantees unskilled man- is hard--requiring the selection and manage- with a wage rate that is too low--if there is no ual work within the district within 15 days of ment of a plethora of small projects over a wide demand at the offered wage rate,it can be registering for employment with the scheme. geographic area and many administrative enti- raised.Setting a low wage rate level does more While this does not necessarily mean locally ties. Ideally, public works schemes require a than ensure that the workfare scheme will be accessible employment, the scheme comes clos- menu of works well-integrated into the local well self-targeted.It also maintains the incentive est to offering a guarantee of any kind. India planning process yet elastic in size and timing. to take up regular work when it becomes avail- recently announced its intention to extend the This can be difficult in low-capacity settings able,and it helps ensure that the program can guarantee by providing 100 days of because of the forward planning and reach as many of those in need as possible. employment on rural public works projects at a interagency coordination needed. In high- Conditions of eligibility. Rationing should be minimum wage.The scheme is not far enough capacity settings, fitting many small labor- avoided; ideally the only requirement should be into implementation to draw lessons. Murgai intensive projects into the sophisticated and the willingness to work at the offered wage.If and Ravallion (2005) simulate some possible often capital-intensive infrastructure plans of rationing is unavoidable (for example,if the outcomes for a range of design parameters: the large- and middle-income cities can be difficult. demand for employment at the wage set targeting could be good and impacts of poverty Moreover, ensuring that the workfare program is exceeds the available budget),explicit secondary large, but the costs could also be substantial--1 poverty focused is not easy because of conflict- criteria should be used--the program may tar- to 2 percent of GDP for the 100-day scheme. ing pressures from alternative target groups, get poor areas,work may be offered only in sea- Labor intensity. The labor intensity--that is, such as the skilled unemployed. sons of greatest need,the length of employment the share of the wage bill in total costs--should Sources: Subbarao (2003) and Murgai and Ravallion of any individual may be limited,or such addi- be higher than normal for similar projects in the (2005). workers have to forgo some income to par- Programs aimed at children can be uni- ticipate: in the absence of the program, they versal or means-tested, free-standing or usually would have been able to pick up linked to the use of health and education some part-time work or engage in low-return services. In both Eastern and Western self-employment activities. In Argentina's Europe, the traditional approach to income Trabajar, forgone earnings were half of gross support for children is through child earnings; in Maharashtra's Employment allowances, independent from but comple- Guarantee scheme, 53 percent; and in mented by extensive public education and Bolivia's ESF, 60 percent.83 Theoretically, for- health care. Most Western European pro- gone earnings can be minimized by allowing grams are universal, although a few are flexible working hours or part-time work, means-tested (Italy and Spain). A higher but this can complicate the supervision of fraction of Eastern European programs and the public works and the workers. those in other middle-income countries are means-tested (Bulgaria, Belarus, the Czech Programs for the nonworking young Republic, the Kyrgyz Republic, Poland, Earlier, we focused on interventions geared to Romania, the Russian Federation, Serbia overcoming the disadvantaged family circum- and Montenegro, the Slovak Republic, stances of young children through a variety of Argentina, Chile). In Africa, Latin America, services that improve their cognitive and social and Asia, there is a long and extensive his- skills and prepare them for learning in school. tory of school feeding programs and mater- Here we focus on ways of augmenting the nal and child health programs that distrib- family's income to alleviate poverty among ute food (or occasionally food stamps). children and improve their chances in life. Many of these programs rely on existing Human capacities 153 service delivery mechanisms to make their Table 7.2 Targeting performance of conditional transfer schemesa administration feasible or cheaper. Quintile PRAF RPS PROGRESA SUF FFE A new wave of CCT recognizes that (Honduras) (Nicaragua) (now Oportunidades) (Chile) (Bangladesh) (Mexico) imperfect markets can lead to underinvest- ment in human capital (chapter 5) and 1 43 55 40 67 -- explicitly seeks to enhance the opportunity- 2 80 81 62 89 48 generating potential of income support 3 94 94 81 97 -- through links to the use of services. CCT 4 98 99 93 100 -- programs are now being implemented in 5 100 100 100 100 100 about two dozen countries, mainly in Latin Source: Morley and Coady (2003), table 5-3. America--but they are being discussed by Note: PRAF = Programa de Asignación Familiar; RPS = Red de Protección Social; SUF = Subsido Unitario Familiar; many other countries and in all regions. FFE = Food For Education; -- = not available. a. Cumulative share (percent) of benefit captured, by income quintile. These programs transfer income in cash or in kind to poor households with children. They grant benefits only if children comply In settings with low access to health and with standards for attendance in school or education services, this tension means that participation in a health care program. In the conditional transfer programs may not be CCT programs with good data, the targeting appropriate vehicles for social assistance. outcomes have been quite good at generally The conditions would keep the program reasonable administrative costs. All five pro- from serving the poorest. The opposite may grams reviewed by Morley and Coady (2003) be true as well: when the use of services is distribute far more than a proportional share already satisfactory, it may not be worth to the bottom quintiles (table 7.2). On aver- using administrative resources to verify age, the share of program benefits going to compliance with service use conditions. the bottom 40 percent of the population is an impressive 81 percent. The evidence on Programs for the nonworking elderly poverty impact is more limited, but PRO- Most countries have public pensions pro- GRESA (now Oportunidades) had a power- grams for the elderly. Two arguments pro- ful effect: program communities experienced vide a rationale for governments to mandate declines of 17.4 percent in the incidence of a pension system to provide for old-age poverty compared with the control group.84 security: imperfect financial markets limit The conditioning of benefits on use of the scope for redistribution over one's life, health and education services serves the and human "failures" to see far enough into dual objectives of avoiding severe depriva- the future may lead to undersaving for old tion and enhancing opportunities for age. The need for old-age security will grow. human development. But there is a tension The population of 60 year olds, about 10 between these goals. A simulation of the percent of the world population today, is results that might be expected from the fed- projected to reach about 21 percent by 2050. eral Bolsa Escola program in Brazil shows Within this group, the fraction of people only a small reduction (1 percentage point) over the age of 80, about 12 percent today, is in the poverty index because of the (simu- expected to reach 19 percent by 2050.86 lated) loss of labor income of children who Contributory pension programs have drop out of the labor force to attend school. not solved the problem of old-age security. Mexico's PROGRESA (now Oportunidades) Coverage is low--only 20 percent of the had impressive poverty impacts but global workforce. Even in pension systems increased primary enrollment rates by only with extensive coverage, the lifetime poor about 1 percentage point because they were cannot contribute enough to have a pension already above 90 percent. Cambodia's pro- at old age that would keep them out of gram, which focuses on grades seven poverty. Elderly women who have not through nine, may well help with the transi- worked outside the home are particularly tion to secondary school, but it misses some vulnerable. Moreover, in some countries, of the poorest households because so many such as in Kenya, Uganda, Sri Lanka, and have dropped out by then.85 Zambia, poorly governed schemes gave 154 WORLD DEVELOPMENT REPORT 2006 lower-income workers returns less than stantial. Kakwani and Subbarao (2005) sim- bank deposits and the alternatives of invest- ulate various options for 15 African coun- ing in land, tools, or a vehicle.87 tries, and conclude that the best--taking into Options for assisting the elderly poor account poverty impact, fiscal cost, and include the following: broadening pension incentive effects--is to keep the benefit low systems to include more people, adding a (about one-third of the poverty threshold), redistributive element as part of an existing the eligible age limit at 65 or older, and to contributory pension scheme, or covering target only the elderly poor, thus sacrificing them through a separate "social pension" the administrative simplicity and political financed by general revenue. A fourth advantage of universalism. There is enough option is a general needs-based social assis- variation across countries to warrant country- tance program. specific efforts to determine benefit and eli- Broadening the coverage of contributory gibility levels and targeting methods rather pensions has been attempted, generally with than relying on rules of thumb. little success. In the Republic of Korea, man- How should we think about the balance dating the expansion of coverage to farmers, between social pensions for the elderly and fishermen, and self-employed was met with other programs, such as those targeting massive protests; in the end, the government families with children? Are the elderly poor had to subsidize in full or in part the contribu- more deserving than other poor? Brazil tions of almost two-thirds of the target popu- spends 1 percent of GDP to transfer $70 a lation. Adding a redistributive element is month to 5.3 million elderly poor and only common, but as pension reforms strengthen 0.15 percent of GDP to transfer $6 to $19 the link between contributions and benefits per month to 5 million families to support for efficiency reasons, that redistributive ele- school attendance through the Bolsa Escola ment is becoming smaller. program.90 When considering whether this Social pensions provide transfers to the the right balance, one can argue generally elderly without requiring prior contributions that young families with children, who have or withdrawing from the labor force.88 They their entire lives ahead of them, should have can be universal, as in Botswana, Mauritius, higher priority. Indeed some argue for shift- Namibia, or Bolivia. Or they can be means- ing public spending away from pensions tested, as in South Africa, Senegal, India, and toward families with children in Brazil Bangladesh, a number of Latin American and others suggest that a focus on unem- countries, Australia, Italy, and New Zealand. ployment may be more appropriate to Many of them complement contributory sys- reduce poverty in South Africa.91 tems that cover higher-income groups. When There may be important political econ- the transfers are means-tested, the programs omy reasons why programs for the elderly are really a special case of needs-based cash garner such political support. There is direct transfers limited to the elderly. The targeting evidence from attitude surveys, across soci- challenges discussed earlier and the potential eties and age groups, that concerns about old solutions are similar. Labor disincentives are age poverty are strong and widely shared-- lessened, however, because societies expect perhaps because most people expect to be old lower work efforts from the elderly.89 one day (but not necessarily unemployed, or Evidence from various countries imple- a single parent, or disabled) and also because menting large social pension schemes indi- old age is more easily verifiable and less sub- cates that the costs are 1 to 2 percent of GDP, ject to moral hazard, for example, when com- not negligible for low-income developing pared with unemployment insurance.92 countries. Schwartz (2003) simulates the costs in six African countries of providing Programs for special social pensions, limiting the benefit to 40 vulnerable groups percent of GDP per capita and eligibility to Some groups are vulnerable regardless of those age 75 and above. The costs would age--the disabled, HIV infected, ethnic range from 0.2 percent of GDP in Kenya to minorities, certain castes, internally dis- 0.7 percent of GDP in Ghana, still not insub- placed households, refugees, and orphans Human capacities 155 (box 7.11).93 One of the key issues in provid- ing transfers for these groups is whether to B O X 7 . 1 1 Africa's orphans and public action set up specific programs or to include them Conflict and the HIV/AIDS pandemic are interventions should first try to strengthen in a more general program. There is no uni- generating a major humanitarian crisis for grassroots responses to orphan care,and turn versal answer to this, and a complex set of families in Sub-Saharan Africa.There are as to supplementary interventions only when issues must be diagnosed in each case. One is many as 43 million orphaned children in the the extended family is no longer sufficient or region today, 10 percent of whom have lost capable.When no other living arrangement is targeting. Not all orphans, widows, or dis- both parents. Orphans make up more than possible,experience and research show that abled are poor, so universal programs will 15 percent of all children in 11 countries, orphanages must be the"last resort."Recog- include some non-poor. A second issue and the numbers are rising . nizing the scope for exploitation of vulnera- relates to the special needs of the groups.Will The death of an earning family member ble children under all arrangements,appro- is most likely to drive a family into penury priate checks and balances must be in place, income support alone be sufficient, and if because of the costs of funeral, the loss of including oversight by NGOs or community- not (as is usually the case) does it make sense regular income, and the risk of losing one's based organizations. to link the income support to other pro- property. Erosion of human capital is When access to basic education and grams for the group? For example, when another major risk: microstudies and analy- health services is generally low, waiving sis of household surveys suggest that, rela- school fees and uniform obligations would large groups of internally displaced people or tive to other children in the household, fos- help increase enrollment rates of all refugees emerge suddenly, their needs for tered children are underenrolled in schools, children including orphans, as in Uganda. housing, food, and health care may strain work longer hours doing household chores, When average access to services is high, but and have lower immunization coverage-- the difference in access between the poor local availability. In such cases in-kind provi- and the disadvantage is stronger for and the non-poor, and between orphans sion to the group is usually the first response. fostered girls than for fostered boys. Psycho- and non-orphan children, is large, cash Only after the groups become long-standing logical risks are also high because the death transfers conditional on children attending or somewhat smaller does the question of of a parent often leaves the child in a state school seem appropriate. Innovative of trauma, lacking nurturance and programs along these lines are just begin- whether to switch to a needs-based cash guidance, and impeding socialization. ning (as in Swaziland). assistance or workfare program arise. Special The main coping strategy in Africa is fos- Sources: Subbarao and Coury (2004); USAID, programs for groups viewed positively or as tering by the extended family.When possible, UNAIDS, and UNICEF (2004). deserving, such as veterans or the disabled, may have adequate political support, but if the group is excluded, as ethnic minorities or tion through impacts on innovation, pro- the HIV positive are in many cases, such pro- ductivity, and social cohesion. But there are grams may not garner sufficient support. big challenges to equitable provisioning-- While standard transfers may protect getting the relevant issues on the policy these groups, a broader set of policies can agenda, fighting political capture of institu- help expand their opportunities and facili- tions so that they do not only serve the tate their integration into society. Some powerful and the influential, and managing countries have used affirmative action (see efficiency-equity tradeoffs, especially in the chapter 8). Others have used regulations short term. There are also good prospects and awareness campaigns that sanction for incremental change through advocacy harmful local practices to help prevent dis- to point out long-run benefits even when crimination. Policy responses include creat- there are short-run costs, through sound ing a framework to hear advocacy groups program and tax design to minimize effi- and mainstreaming such concerns into gov- ciency costs and build accountability struc- ernment practice, often building on infor- tures, and through political coalitions that mal or private arrangements, such as those can thwart elite holdups. of faith-based organizations. The power of greater equity in human capacities to unhinge inequality traps is Summary tremendous--through directly contribut- Equity in the acquisition of human capaci- ing to leveling the economic, political, and ties--through early childhood develop- sociocultural playing fields. But achieving ment, formal education, health services, greater equity in human capacity is not and social protection--is at the core of a enough to break the inequality trap. It strategy to equalize the opportunities for needs to be complemented with fairness in people to lead productive, fulfilling lives. the returns to those capacities and in the Broad provisioning of these services is also access to complementary assets, topics dis- good for development and poverty reduc- cussed in the next chapter. Justice, land, and infrastructure 8 It takes more than building human capacities inaccessible, because they are incompatible to broaden people's opportunities. People with local norms and customs and they are c h a p t e r also need complementary assets, access to the physically or economically inaccessible, or marketplace, and security of person and because people lack the knowledge or capac- property. This chapter starts with a descrip- ity to navigate the system. Third, elite tion of justice systems, showing how critical capture and the inaccessibility of the legal they are in ensuring a level playing field and system may mean that policies relating to fair returns. It then turns to policies for crime and personal security are inequitable expanding access to the complementary assets and perpetuate crime-related inequality traps. of land and infrastructure. Promoting fair- ness in markets is addressed in chapter 9. Combating elite capture and discrimination Building equitable justice systems Political and economic elite interests often Society's rules, and the institutions that coincide at the expense of a disempowered establish, maintain, and transform them, majority. When power is in the hands of a govern market and nonmarket interactions. narrow elite, the rights of most citizens are They determine people's endowments, their unstable. A century of banking in Mexico, rights and obligations, and their ability to outlined in chapter 6, illustrates how deals generate fair returns. Reflecting and pro- between the political and economic elite led ducing the distribution of power among to the establishment of banking monopo- groups, good institutions (so necessary for lies and laws that maintained a system of prosperity) emerge only when the distribu- rent-sharing between banks and govern- tion of political power and enforceable ments.1 Another striking example of elite rights is equitable. capture comes from the transition economies Legal institutions play a key role in the dis- and the rise of oligarchs who manipulated tribution of power and rights. They also politicians and shaped institutions to get underpin the forms and functions of other rich.2 Legal systems that cater to narrow institutions that deliver public services and interests also tend to discriminate against regulate market practices. Justice systems can other groups through inequitable laws and provide a vehicle to mediate conflict, resolve practices. disputes, and sustain social order. But Ensuring equality before the law and inequitable justice systems may perpetuate securing both personal and property rights inequality traps by maintaining or reproduc- for a broad section of the community give ing elite interests and discriminatory prac- individuals the incentive and the opportu- tices. Equitable justice systems are thus cru- nity to take part in economic and political cial to sustained equitable development. life. This requires an independent and Building more equitable justice systems accountable judiciary and laws and prac- runs into three main challenges--often tices that protect citizen's rights in a nondis- interrelated and reinforcing. First, legal criminatory way. institutions may be open to capture by elite interests or may discriminate against certain Enhancing judicial independence and groups. Second, these institutions are often accountability. In many countries, a rule of 156 Justice, land, and infrastructure 157 law system--administered by multiple arms accountable. In Colombia public informa- of government--constrains political power.3 tion centers in major courts disseminate In this system, an independent judiciary acts information and help people use the court. as a safeguard against abuses of state and In Venezuela information is provided to the nonstate power. Because judges are also open public through an Internet-based judicial to elite capture and corruption, accountabil- portal for the Supreme Tribunal.11 ity mechanisms are a key aspect of legitimate Strengthening the relationship among judicial independence.4 civil society, the media, and the courts has In many developing countries, shifts also improved public awareness and scrutiny toward an accountable and independent of the judicial system. Bad judges have judiciary require a change in culture and resigned because of high-pressure media institutional practice. Ethiopia established an campaigns, such as the recent media independent judiciary for the first time in scrutiny in the Philippines.12 The media can 1995.5 In Vietnam "telephone justice" was also disseminate information, such as the common, with party elites habitually contact- "My Rights" television show in Armenia ing judges to direct decisions.6 Changing (box 8.1). Similar shows have been devel- ingrained institutional practices in both oped in other parts of Eastern Europe. In countries has been a slow process. Poor con- Georgia, an NGO disseminating informa- ditions of judicial service in many countries tion about the courts increased public satis- can increase corruption.7 For example, low faction with the courts. remuneration for magistrates in Kenya made The existence of an independent and them open to alternative funding for their accountable judiciary is not enough to pro- services; Kenya removed almost one-third of tect citizens against abuses of state power. judicial staff for corruption in 2004.8 Adequate laws and institutional mecha- Promoting judicial independence with- nisms are also needed. In Thailand, for out establishing accountability mecha- example, separate administrative courts nisms can further entrench elite interests. were established for the first time in 2001 to Institutional safeguards, transparency, and protect citizens against arbitrary uses of the existence of a civic constituency are key state power. The courts aim to ensure that to both accountability mechanisms and state authorities act in accordance with state judicial independence. Institutional safe- laws and regulations. They also aim to guards include providing for security of enhance citizen participation in public pol- tenure and improving conditions of serv- icy formulation and oversight. In the first ice for judges; rigorous and transparent three years, the courts processed almost appointment and disciplinary processes; transparent mechanisms of case allocation and case management; transparent and B O X 8 . 1 Increasing legal literacy and public awareness: open hearings; appeal rights and the publi- "My Rights" on Armenian public television cation of judicial decisions; and public information about the courts.9 Many Many people in Armenia have no understand- The show airs once a week on Armenia's countries have enshrined judicial inde- ing of the legal system or the rights afforded state television channel. After only five or six pendence in the constitution or state them under the law.And distrust of the courts shows,"My Rights"became the number one laws.10 Bolivia has established open com- is widespread.In a recent public awareness show in Armenia.There have been numer- campaign,the government funded a televi- ous reports of viewers requesting legal doc- petitions for judgeships and ethical stan- sion show to provide citizens with examples, uments and decisions from notaries, judges, dards for judges. Courts in the Philippines advice,and information on their legal rights. and other legal officials based on what they have a performance management system "My Rights"uses mock trials to depict learned from the show. And when the real-life disputes in Armenian courts.The tel- power went out in one village a few for judicial and nonjudicial personnel. evision judge is a deputy minister of justice, minutes before"My Rights"was going on Public information campaigns can en- and the parties are often those in the real the air, the people in the town marched on hance the independence and accountability dispute.The topics--such as rental and prop- the mayor's office and accused the local erty disputes,customs issues,and family law officials of intentionally cutting the power of the courts, increasing public confidence in matters--are timely and of broad interest.A so that people could not watch the show! and commitment to the system, and enhanc- live studio audience of judges,lawyers,legal ing people's capacity to demand better gov- officials,and others discusses the trials on air. Source: Decker and others (2005). ernance and hold those in positions of power 158 WORLD DEVELOPMENT REPORT 2006 17,000 cases, most concerning corruption tage may mean, however, that legal equality is or other unlawful acts by public officials. not enough. Some countries have passed laws Many of the cases made front-page news in that discriminate in favor of certain groups, Thailand because of their social impact. creating affirmative action programs on the basis of race, ethnicity, and gender or for peo- Combating discriminatory norms and prac- ple with disabilities. An assessment of two of tices. Laws that reinforce exclusionary prac- the most widely implemented affirmative tices in norm-based institutions perpetuate action programs, in India and the United unequal power relations. Some laws may States, suggests mixed impacts (box 8.2). discriminate against particular groups, such The mere existence of "equitable laws" as laws affecting indigenous people or the for affirmative action does not guarantee laws in apartheid South Africa. The absence their equitable implementation or enforce- of laws can also reinforce unequal power ment. For example, in Peru and Honduras, relations as for domestic violence, often rele- gender discrimination in judicial decisions gated to the nonlegal private realm. and treatment by police and judges discour- In many countries, antidiscrimination age women from using the system to resolve and equal opportunity laws have reduced dis- disputes.13 Such disadvantaged groups are criminatory practices. Historical disadvan- more likely to experience the law-and-order side of the law than the protection of their rights (as discussed below under crime and B O X 8 . 2 Affirmative action in India and the United States personal security). The affirmative action program in India is minorities in education and employment Making justice accessible based primarily on caste and gender and (Deshpande 2005). that in the United States primarily on race. The programs in both countries have People's legal rights remain theoretical if the Before independence in India, the British become centerpieces of political battles institutions charged with enforcing them are government introduced affirmative action over race and caste.Critics argue that they inaccessible. Accessibility depends on how to address discrimination against tend to benefit the upper echelon of minor- compatible laws are with the norms and "untouchable" castes (now known as Dal- ity groups, and they are difficult to end.In its) and "tribals" (now known as Adivasis). India, the programs are said to apply to sub- understandings that shape people's lives. After independence in 1947 the policy of castes that have not traditionally faced dis- Legal institutions need to be physically and reserving 22.5 percent of seats in educa- crimination (Sowell 2004).They may also economically accessible and people need to tion institutions, government jobs and reinforce negative stereotypes by placing have the knowledge and capacity to claim electoral seats was written into the consti- minorities in positions they are not qualified tution. Since 1991, a further 27 percent for (Coate and Loury 1993).Despite these their rights. quota has been introduced for other low weaknesses, India's program has provided castes (called Other Backward Castes), but formal sector employment and higher edu- Addressing the compatibility of state and with no constitutional guarantee. And cation for many Dalit and Adivasi families, customary justice systems. Forms of cus- since 1993, 33 percent of the seats in local freeing them from subservient roles.With governments have been reserved for the reservations in local government, tomary or nonstate law operate in a women, Dalits, and Adivasis (Deshpande elected women leaders make decisions in majority of countries.14 Yet they are often 2005). line with women's needs (Chattopadhyay neglected in justice sector reform policies. In the United States, slavery was perva- and Duflo 2004).Low-caste representatives Engaging with customary systems is an sive for more than two centuries, and not in state assemblies increase the allocation of until 1866 were blacks granted citizenship quota-based jobs to low-caste constituents important part of equitable reform strate- rights.The system that replaced slavery was (Pande 2003).And Dalit representatives in gies for two main reasons. First, customary only marginally better, with several features village government improve the targeting of law is often a fundamental part of a com- similar to the Indian caste system: segrega- benefits to Dalits (Besley and others 2004). munity's identity and belief system; thus, a tion, denial of education, restrictions to low- In the United States, disparities between paid, menial jobs, social and economic dis- blacks and whites continue to be significant lack of recognition can be intrinsically dis- crimination, negative stereotyping, and on all economic indicators, and there is evi- criminatory and serve to exclude communi- violence.The Civil Rights Act of 1964 and dence of discriminatory gaps in earnings. ties from the wider state system. Second, a subsequent legislation, Supreme Court rul- But affirmative action in jobs has increased ings, and executive orders in the 1970s intro- black employment and enrollment in failure to engage with customary systems duced affirmative action into the political, higher education (Holzer and Neumark may leave inequitable and inefficient prac- judicial, administrative, and economic 2000, Bowen and Bok 1998). But the U.S. tices at the local level unchecked. spheres of American society.Starting with program's quasi-voluntary element means Where state and nonstate systems have the label"equal opportunity,"selection pro- that litigation can dilute the program, and cedures incorporate compensatory correc- black representation in government bodies developed in tandem, they often comple- tion to ensure adequate representation of continues to be very low. ment each other and reinforce socially accepted codes and rules. But in communi- Justice, land, and infrastructure 159 ties where the state systems lack legitimacy lack capacity, be inaccessible, or dramati- and political reach, customary systems cally increase transaction costs.20 In rural often act independently from the state legal Tanzania, a perception that state institutions system, which may be rejected, ignored, or can not supply law and order has led to the not understood. Real difficulties arise when emergence of "new" forms of organized vil- local customary systems are at odds with lage defense groups called sungusungu. the rights and responsibilities articulated in While technically illegal, the sungusungu are state law. often informally supported by the state, In many developing countries, custom- given their success in reducing crime.21 ary systems are the dominant form of regu- Ignoring or trying to stamp out custom- lation and dispute resolution. In Sierra ary practices can also have serious negative Leone, about 85 percent of the population implications. Top-down reform can under- fell under customary law as of 2003.15 Cus- mine informal institutions without provid- tomary tenure, discussed below, affects 90 ing viable alternatives, and the vacuum can percent of land transactions in Mozam- lead to power grabbing, lawlessness, or even bique and Ghana.16 Customary justice violent conflict. When neither formal nor depends on local traditions, as well as the informal mechanisms are functioning, political history of a country or region. human rights abuses and serious conflict Ethiopia officially recognizes more than 100 are more likely. For example, a study in distinct "nations or peoples" and more than rural Columbia found the incidence of vigi- 75 languages. lantism,"mob justice," or lynching to be five Customary systems can be incompatible and a half times greater in communities in with economic, social, and civil rights. which informal mechanisms are no longer Many forms of customary law are seen to functioning effectively and the state pres- discriminate against marginal groups. In ence remains limited.22 much of Sub-Saharan Africa, for example, A failure to engage with customary customary systems systematically deny systems may mean that discriminatory women's rights to land, assets, or opportu- prac