R E G I 0 N A L A N D S E C T 0RA L S T U D I E S Women's Employment and Pay in Latin America Overview and Methodology FILE COPY Report No. :11360 Type: (PUB) Title: WOMEN'S EMPLOYMENT AND PAY IN I Author: PSAOHAROPOULDS, GEO Ext.: 0 Room: Dept.: NOVEMBER 1992 BOOKSTORE GEORGE PSACHAROPOULOS AND ZAFIRIS TZANNATOS -20 .;5 Z t.t.Y>X ,=t 10 1 ,i t' " Women's Employment and Pay in Latin America Overview and Methodology WORLD BANK REGIONAL AND SECTORAL STUDIES Women's Employment and Pay in Latin America Overview and Methodology GEORGE PSACHAROPOULOS AND ZAFIRIS TZANNATOS The World Bank Washington, D.C ( 1992 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, N.W., IWVashington, D.C. 20433 All rights reserved Manufactured in the United States of America First printing November 1992 The World Bank Regional and Sectoral Studies series provides an outlet for work that is relatively limited in its subject matter or geographical coverage but that contributes to the intellectual foundations of development operations and policy formulation. These studies have not necessarily been edited with the same rigor as Bank publications that carry the imprint of a university press. The findings, interpretations, and condusions expressed in this publication are those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to the members of its Board of Executive Directors or the countries they represent. The material in this publicationis copyrighted. Requests forpermission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for dassroom use is granted through the Copyright Clearance Center, 27 Congress Street, Salem, Massachusetts 01]970, U.S.A. The complete bacldist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'1ena, 75116 Paris, France. George Psacharopoulos is the senior human resources adviser to the World Bank's Latin America and Caribbean Technical Department. He previously taught at the London School of Economics. Zafiris Tzannatos is a labor economist with the Population and Humain Resources Department at the World Bank. He is an honorary research fellow at the Universities ofNottingham and St. Andrewsin the United Kingdom. Cover design by Sam Ferro Library of Congress Cataloging-in-Publication Data Psacharopoulos, George. Women's employment and pay in Latin America: overview and methodology / George Psacharopoulos and Zafiris Tzannatos. p. cm. Includes bibliographical references. ISBN 0-8213-2270-2 1. Women-Employment-Latin America. 2. Wages-Women-Latin America. I. Tzannatos, Zafiris, 1953- . II. Title. HD6100.5.P79 1992 331.4'098-dc2O 92-35611 CIP Contents Acknowledgments xiii Foreword xv 1 Introduction and Summary 1 2 Trends and Patterns in Female Labor Force Participation 1950-1985 37 3 The Industrial and Occupational Distribution of Female Employment 71 4 Potential Gains from the Elimination of Labor Market Differentials 135 5 Gender Differences in the Labor Market: Analytical Issues 151 6 Summary of Empirical Findings and Implications 183 Appendix A: Contents of Companion Volume 213 Appendix B: Authors of Companion Volume 217 References 219 Index 241 v vi Women's Employment and Pay in Latin America List of Tables 1.1 Female Participation Rate and Female-Relative-to-Male- Pay 5 1.2 Aggregate Statistics for Selected Latin American Countries 8 2.1 Total Labor Force in Latin America (selected years) 39 2.2 Relative (F/M) Labor Force in Latin America and the Caribbean 42 2.3 Participation Rates for Prime Age Groups 45 2.4 Female Participation Rate by Age Group and Country's per Capita Income Early 1980s 54 2.5 Female Labor Force Participation Rate and ReLigion Early 1980s 54 2.6 Ratio of Employees in the Female Labor Force by World Region 58 A2.1 Age-Specific Female Labor Participation by Region and by Country (early 1L980s) 63 3.1 Female Overrepresentation by Employment Status (1950s and 1980s) 80 3.2 Female Overrepresentation by Industry (1950s and 1980s) 83 3.3 Female Overrepresented Industrial Sectors in the 1950s and 1980s by Employment Status 84 3.4 Occupational Dissimilarity (Duncan index) 86 3.5 Structure and Sex Ratio Effects on Occupational Dissimilarity Over Time 92 3.6 Workers Who Would Have to Change Occupation to Reach Equality in the Employment Distribution of Women and Men, as a Percentage of the Total Labor Force 98 Contents vii 3.7 Dissimilarity Between Female and Male Occupational Employment in Selected Industrialized Countries (1970- 1982) 103 3.8 Dissimilarity Between Female and Male Industrial Employment 105 A3-la Occupational Distribution of the Labor Force by Employment Status - Argentina 112 A3-lb Occupational Distributionof the Labor Force by Employment Status - Bolivia 113 A3-1c Occupational Distribution of the Labor Force by Employment Status - Chile 114 A3-ld Occupational Distribution of the Labor Force by Employment Status - Colombia 115 A3-le Occupational Distribution of the Labor Force by Employment Status - Costa Rica 116 A3-lf Occupational Distribution of the Labor Force by Employment Status - Ecuador 117 A3-lg Occupational Distribution of the Labor Force by Employment Status - Guatemala 118 A3-lh Occupational Distribution of the Labor Force by Employment Status - Honduras 119 A3-li Occupational Distribution of the Labor Force by Employment Status - Jamaica 120 A3-lj Occupational Distribution of the Labor Force by Employment Status - Mexico 121 A3-lk Occupational Distribution of the Labor Force by Employment Status - Panama 122 A3-11 Occupational Distribution of the Labor Force by Employment Status - Peru 123 viii Women's Employment and Pay in Latin America A3-lm Occupational Distribution of the Labor Force by Employment Status - Uruguay 124 A3-ln Occupational Distributionof the Labor Force by Employment Status - Venezuela 125 A3-2a Industrial Distribution of the Labor Force by Employment Status - Argentina 126 A3-2b Industrial Distribution of the Labor Force by Employment Status - Bolivia 127 A3-2c Industrial Distribution of the Labor Force by Employment Status - Brazil 128 A3-2d Industrial Distribution of the Labor Force by Employment Status - Colombia 129 A3-2e Industrial Distribution of the Labor Force by Employment Status - Ecuador 130 A3-2f Industrial Distribution of the Labor Force by Employment Status - Jamaica 131 A3-2g Industrial Distribution of the Labor Force by Employment Status - Mexico 132 A3-2h Industrial Distribution of the Labor Force by Employment Status - Peru 133 A3-2i Industrial Distribution of the Labor Force by Employment Status - Venezuela 134 4.1 Results of the Within Industry Elimination of Occupational Differentials 142 4.2 Percentage Change in Female Wages 144 A4.1 Results of the Within Industry Elimination of Occupational Differentials 148 6.1 Female Participation by Selected Sample Characteristics 185 Contents ix 6.2 Decomposition of the Male Pay Advantage in the Region 185 6.3 Contribution (in log percentage points) of Selected Variables to the Male Pay Advantage in the Region 191 6.4 Female Wages (in local currency) and Female Relative to Male Wage in the Private and Public Sectors (selected countries) 201 A6. 1 Percentage of Male Pay Advantage Attributed to Differences in Endowments (E) and Rewards (R) 203 A6.2a Average Hours per Week and Coefficients on Log (hours) by Sex 204 A6.2b Contribution of Differences in Hours to the Male Pay Advantage 205 A6.3a Average Years of Schooling and Estimated Coefficients on Schooling by Sex 206 A6.3b Contribution of Differences in Schooling to the Male Pay Advantage 207 A6.4a Average Years of Potential Experience and Coefficients on Potential Experience by Sex 208 A6.4b Contribution of Differences in Potential Experience to the Male Pay Advantage 209 A6.5 Contribution of Differences in the Constant Terms to the Male Pay Advantage 210 A6.6 The Value and Significance of the Coefficient on the Sample Selection Variable (Lambda) in the Eamings Functions 211 x Women's Employment and Pay in Latin America List of Figures 1.1 Female Labor Force Participation Rate in Latin American and the Caribbean Countries 1950s and 1980s 16 1.2 Female Age-Participation Profile in Latin American and Caribbean Countries 1950s and 1980s (Stylized) 17 1.3 Female Participation Rate in Industrialized and Latin American and Caribbean Countries (Stylized) 19 1.4 Female Labor Force Participation Rate by Selected Characteristics, Costa Rica 1989 25 1.5 Decomposition of the Male-Female Wage Gap (Stylized) 26 1.6 Female Labor Force Participation Rate in Latin American and the Caribbean Countries 1950s and 1980s 30 1.7 Female Monthly Earnings by Educational Level, Costa Rica 1989 31 2.1 Female/Male Labor Force in Latin American and Caribbean Countries 43 2.2 Female Labor Participation Rate by Age 46 2.3 Female/Male Labor Force Participation Rate by Age 50 2.4 Female Participation Rate by World Region and Age Groups 52 3.1 Percentage of the Labor Force in Agriculture in Nine Latin American and Caribbean Countries 1980s 75 3.2 Percentage of thLe Labor Force in Industry in Nine Latin American and Caribbean Countries 1980s 77 3.3 Percentage of the Labor Force in Services in Nine Latin American and Caribbean Countries 1980s 78 Contents xi 3.4 Distribution of the Labor Force in Nine Latin American and Caribbean Countries 1980s 79 3.5 Percentage of Change in Labor Force in Latin American and Caribbean Countries by Broad Industrial Sector 1950-1980 80 4.1 Effects of Sex-Differentials in the Labor Market 140- 5.1 Decomposition of the Gender Wage Gap 157 5.2 Costs and Benefits of Investment in Education 160 5.3 A Truncated Distribution 171 Acknowledgments We have benefited from comments and encouragement from many people who read earlier versions of this volume and participated in seminars given at the World Bank, the University of St. Andrews, and conferences organized by the Comparative and International Education Society, the International Union for the Scientific Study of Population, and the European Society for Population Economics. In particular we would like to thank Ana-Marfa Arriagada, Alessandro Cigno, Deborah DeGraff, Barbara Herz, David Huggart, Emmanuel Jimenez, Philip Musgrove, and Michelle Riboud for their helpful comments, Professor William Greene for providing purpose built routines for LIMDEP which facilitated the estimation procedures used in the country studies; Diane Steele and Carolyn Winter for their reviews of the book; Hongyu Yang for preparing the graphics; and Donna Hannah for typing and preparing the earlier versions of this book and Marta Ospina for taking these tasks over and eventually putting the book into its present form. The completion of the present study would have not been possible without the generous support of the Norwegian Trust Fund. xiii I Foreword Women's role in economic development can be examined from many different angles, including feminist, anthropological, sociological, economic, and legislative perspectives. This study employs an economic perspective and focuses on how women behave and are treated in the work force in a number of Latin American economies. It specifically considers the determinants of women's labor force participation and male-to-female earnings differentials. Understanding the reasons for 'low' labor market participation rates among women, or 'high' wage discrimination against women, can lead to policies that will improve the efficiency and equity with which human resources are utilized in a particular country. The study is in two volumes. This volume presents aggregate data on the evolution of female labor force participation in Latin America over time, showing that in some countries twice as many women (of comparable age groups) work in the market relative to twenty years ago. The companion volume uses household survey data to analyze labor force participation rates and wages earned by men and women in similar positions, paying special attention to the role of education as a factor influencing women's decision to work. The results show that, overall, the more years of schooling a woman has, the more likely she is to participate in the labor force. In addition, more educated women earn significantly more than less educated women. The book also attempts analyses of the common factors which determine salaries paid to men and women in an effort to identify what part of the male/female earnings differential can be attributed to different human capital endowments between the sexes, and what part is due to unexplained factors such as discrimination. Differences in human capital endowments explain only a small proportion of the wage differential in most of the country studies. The remaining proportion thus represents the upper bound to discrimination. It is our hope that this work will be followed up by a more careful look at labor legislation and the role it plays in preventing women from reaching their full productive potential. S. Shahid Husain Vice President Latin America and the Caribbean Region xv 1 Introduction and Summary 1. Objective This is a fact finding, policy oriented study about working women in 15 Latin American and Caribbean countries.' The aims are: 1. To establish patterns and trends in women's characteristics in the labor market; 2. To identify factors affecting women's decision to work for pay and to quantify their impact on female participation; 3. To examine what part of the gap between women's and men's labor earnings cannot be explained by differences in their respective productive characteristics; and 4. In view of these findings, to explore policy options that can enhance the functioning of the labor market and contribute towards the alleviation of poverty. The evidence comes from country household surveys undertaken during the late- 1980s and published data (primarily population censuses) covering the period between 1950 and 1980. I The countries are: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, Guatemala, Honduras, Jamaica, Mexico, Panama, Peru, Uruguay, and Venezuela. These countries accounted for approximately 90 percent of the total labor force in the region both in the 1950s and the 1980s. (See Yearbook of Labour Statistics 1945-1989: Retrospective Edition on Population Censuses, Geneva: ILO, 1990). 2 Women's Emtployment and Pay in Latin America Though certain individual country peculiarities have been accommodated, this volume does not attempt to provide a complete analysis of specific country issues. Rather, a comparative methodology is applied in order to examine whether there are common factors operating in labor markets in the Latin American2 region which can be used to set a policy framework and an agenda for research. In this respect, the general directions for the design of policy and management of services are established with a view towards facilitating behavioral and economic change. "Behavioral" does not mean changing people's culture and resulting choices; rather, it means the removal of constraints imposed by market failure which prevent women from exercising their choices. There are two volumes to the present study. This volume provides an overview of women's characteristics in the labor market, summarizes the findings and outlines possible policy options. A companion volume deals with the situation in individual countries. 2. The Problem Research on women's role in development has been increasing rapidly in recent years.' Three factors have spurred this interest. First, developing countries have been moving from a traditional mode of economic activity towards systems followed by the industrialized countries over the past 40 years or so. Today, the functional roles of women and men in developing countries are characterized less by breadwinning men in the labor market and "bread-processing' women at home. Women's welfare depends now more than ever before on the labor market where women earn less than men and are more likely to be unemployed than men. This change has earned women the title of a "vulnerable' group.4 In many developing countries this vulnerability can, for practical purposes, be regarded as a synonym of, if not a euphemism for, poverty. 2 For brevity "Latin America" will be used to mean "Latin America and the Caribbean." I See the collection of papers in World Development, July 1989 (special issue on women in development) and references therein. 4 A consistent finding in recent developmental work has been that women and children bear disproportionately the burden of adjustment (Comia, Jolly and Stewart 1987; UNICEF 1987; ECLAC 1990, 1991). Introduction and Summary 3 Second, the belief which was widely held in the post-war era of high growth -- that moving the locus of employment from the traditional rural sectors to the modem urban sectors would more or less automatically solve the problem of regional disparities -- began to be seriously questioned in the 1970s and was largely repudiated following the adverse economic developments of the 1980s.5 Efficiency considerations have become more important than ever before. The omission of half a country's human capital from the development strategy clearly leads to inefficiency. Third, the policy implications of the changing national economic environment and social fabric are complex and worth examining in situ. Simply copying lessons from elsewhere is not likely to be effective. With respect to the labor market, women's employment and economic status are typically inferior to those of men. This is more so in developing economies than industrialized ones.6 Also, the economic and cultural characteristics of developing countries are more diverse than the characteristics of advanced economies and women's behavior is known to be more dependent upon these characteristics than men.7 Therefore, women's labor market status and welfare are greatly affected by the prevailing conditions in the macro-economy. The study of women's issues would be incomplete if country specific socio-economic constraints are ignored. Poverty, inefficiency and underdevelopment are interrelated issues which depend to a great extent on the functioning of the labor market. The greatest part of national income, and individual/family income, is generated from the buying and selling of labor in the market. Therefore, the way labor is priced and allocated in the market is of prime importance to economic development and social s See World Labour Report, ILO, 1987. 6 Anker, Buvinic and Youssef (1982); Buvinic, Lycette and McGreevey (1983); Bardhan (1984); Birdsall and Sabot (1991). 7 The view that women's reactions to changing socio-economic conditions are greater (more "elastic") than those of men is shared by all schools of thought and is also strongly confirmed by empirical evidence. For example, in the realm of conventional economic analysis women are categorized as "secondary" workers whose fate depends crucially on the stage of economic development in general and, within a particular developmental stage, on the cyclical fluctuation of the economy. At the other end of analysis, women are seen as a "reserve army" whose utilization depends on the productive forces and productive relations operating at a particular point in time. All theories assume or predict that most men will be permanently attached to the labor force, even if only as unemployed workers. 4 Women's Employment and Pay in Latin America welfare. This observation leads to the following line of reasoning in the case of women in development: If more women worked for pay and if women were paid more for their work, then the value of output would increase and poverty would be reduced. The evidence provided in. Table 1.1 suggests that indeed more women could work for pay and female pay could be higher than it is at present.8 The figures relate to female participation rates and relative (female to male) pay. Though the figures are subject to a number of qualifications (which are discussed in more detail in chapters that follow), they show that only one-third of women are currently in the labor market (column 2), and women's labor earnings are around 70 percent of the earnings of men (column 4). The former finding relates to extensive growth, that is, that more output can be produced if more persons are employed. The latter finding relates to intensive growth (that is, to a more efficient use of women workers) and, as a significant byproduct, to poverty and, especially, the feminization of poverty.9 It is clear, therefore, that the status of women in the labor market is a crucial determinant of economic development and social welfare. It is also a timely issue for Latin America where recent economic performance has been poor. This proposition does not necessarily imply that more women should work or that women should be paid more. ]n fact, it is possible that the observed low rates of female participation and low levels of female pay are efficient. Also, one has to take into account the labor demand side. However, it is also possible and, given the experience of industrialized countries, more probable that the utilization of female labor in the region is less than optimal. Our concern is about the removal of constraints which lead to "unjustified" gender differentials in the labor market. Some of these constraints are discussed below. 9 Tokman (1989). Introduction and Summary 5 Table 1.1 Female Participation Rate and Female-Relative-to-Male-Pay Female Relative Country (age group) Year Participation Year FIM pay Rate (percent) (percent) (1) (2) (3) (4) Argentina (20-60) 1980 33.1 1985 64.5 Bolivia (20-64) 1976 23.1 1989 62.3 Brazil (20-60) 1980 33.0 1980 61.2 Chile (20-60) 1982 28.9 1987 65.4 Colombia (25-60) 1985 39.4 1988 84.6 Costa Rica (20-60) 1984 26.4 1989 80.8 Ecuador (20-60) 1982 22.6 1987 63.7 Guatemala (20-60) 1981 14.7 1989 76.8 Honduras (20-60) 1974 18.0 1989 81.3 Jamaica (20-64) 1982 48.2 1989 57.7 Mexico (20-60) 1980 32.7 1984 85.6 Panama (20-60) 1980 35.7 1989 79.6 Peru (20-60) 1981 29.0 1990 65.7 Uruguay (20-60) 1985 46.0 1989 57.4 Venezuela (20-60) 1981 35.0 1989 70.6 Average 31.1 70.5 Note: Participation of prime-age women (aged 20 to 60 years). Weekly earnings in Venezuela, Mexico, Colombia, Jamaica, Honduras, Chile and Bolivia, and monthly earnings in all other countries. Source: Participation: constructed from ILO (1990), Table 1. Relative pay: based on information provided in the companion volume. 6 Women's Employment and Pay in Latin America 3. The Latin American Case In Latin America, the study of women's earnings in the labor market and labor force participation is characterized by significant analytical, practical and statistical problems. From an analytical point of view, one difficulty arises from the fact that the region is more diversified and complex than other developing areas. Latin America as a group is second only to industrialized countries in terms of development, being far ahead of most African and mainland Asian countries.' Hence, the region is quite diverse both in terms of composition of the final output (primary, industrial and service sectors) and also in terms of social structure. Table 1.2 presents some economic and social indicators which show this diversity. Column 1 shows the level of 1988 per capita income (in US$) in the 15 countries under consideration which contain the bulk of the region's population (almost 90 percent). In column 2, per capita income is expressed as a percentage of the corresponding figure for the United States in the same year. At the lower end is Bolivia and Honduras and at the higher end Argentina and Venezuela. Social indicators are equally diverse; only six countries have a life expectancy of more than 70 years. In Bolivia life expectancy is 53 years. In Guatemala and Peru the corresponding figure is 62 years (column 3). The across country diversity in women's family characteristics may be inferred from the variation in the total fertility rate (column 4): in some countries (such as Uruguay, Argentina, Chile and Jamaica) the total fertility rate is lower than three children per woman while in other countries (Guatemala, Honduras and Bolivia) it is more than five children per woman."1 The diversity of women's status within countries can be assessed from the adult female and male literacy rates, which can be taken as a measure of women's absolute and relative position in terms of socio-economic welfare (columns 4 and 5). Women's illiteracy rates are typically higher than men's and, in some cases, reach 35 percent -- in fact 53 percent in Guatemala. Finally, a commonly noted fact, '° For example, the weighted average for Latin America's per capita income is $1,840 (US$ in 1988) compared to $3,470 for high income countries, $320 for South Asia, $330 for Sub-Saharan Africa, $540 for East Asia, and $1,380 for the lower middle income economies taken as a group (World Bank, 1990: Table 1, p. 178). The figures for income are closely related to health, nutritional, educational and demographic data and to women's conditions (Iid., Tables 27, 28, 29 and 31). 11 Total fertility rate (TFR) is the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children at each age in accordance with prevailing age specific fertility rates. Introduction and Summary 7 confirmed in this study, is that low aggregate female labor force participation rates for the region mask women's high participation rates in urban areas. This contrasts with other developing regions such as South East Asia, which has high- urban/high-rural female labor force participation, and the Middle East countries with low-urban/low-rural participation.'2 Data for Chile, Colombia, Puerto Rico, Costa Rica, El Salvador, Peru and Venezuela suggest that the female labor force participation in rural areas is only about 50 percent of the female urban participation rate. 3 These remarks suggest that it may not be as easy to make broad generalizations about the region as elsewhere in the developing world. The study of Latin American women in the labor market today is beset with an additional practical problem: the data may reflect transitory characteristics of the labor force. More specifically, the region experienced considerable economic and social growth in the 1950s, 1960s and during some part of the 1970s. The main sources for economic growth have been, first, an expanding export sector dependent on primary commodities; and, second, an industrialization drive fuelled primarily by domestic demand and a sustained growth in investment (especially public investment). However, in the last 10 to 15 years the region experienced significant economic slowdown and macro-economic imbalances. The changing role of the public sector (in terms of capital formation and employment, the latter especially from the point of view of women) may have affected the present data. In addition, the importance of the growth and structure of the private/formal sector may be underplayed. The severity of the recession is shown in column 7 of Table 1.2. The annual percentage change of per capita GDP suggests that one-third of the countries under consideration (Venezuela, Argentina, Peru, Jamaica and Bolivia) have had negative growth in the last quarter century."4 Three countries (Chile, Guatemala and Honduras) have had growth rates lower than one percent per annum. In another three countries (Uruguay, Panama and Costa Rica) the growth rate was between 1.2 percent and 1.6 percent. In the remaining four countries (Brazil, Mexico, Colombia and Ecuador), the 2.3 percent to 3.5 percent annual growth rates during the 1965-1989 period mask the fact that recent growth has been slower than in the earlier period, as low as 0.7 percent per annum in the case of 12 Standing (1981. p. 15). 13 Ibid., Table 1, p. 17. 14 The annual rate of per capita GDP growth is based on World Bank's estimates calculated from constant price series using the least squares method. See World Bank, World Development Report 1991: The Challenge of Development, 1991 (chapter on technical notes). 8 Women's Employment and Pay in Latin America Table 1.2 Aggregate Statistics for Selected Latin American Countries Per Capita GDP Life Total Adult Annual rate (US$ % of Expectancy Fertility Illiteracy of per capita 1988) US (years) Rate Rate (percent) GDP Growth (children Female Male 1965-89 per woman) (percent) Country (1) (2) (3) (4) (5) (6) (7) Venezuela 3250 16.4 70 3.7 15 11 -1.0 Argentina 2520 12.7 71 2.9 5 5 -0.1 Uruguay 2470 12.4 72 2.4 4 6 1.2 Brazil 2160 10.9 65 3.4 24 20 3.5 Panama 2120 10.7 72 3.1 12 12 1.6 Mexico 1760 8.9 69 3.5 12 8 3.0 Costa Rica 1690 8.5 75 3.2 7 5 1.4 Chile 1510 7.6 72 2.7 - - 0.3 Peru 1300 6.6 62 4.0 22 8 -0.2 Colombia 1180 5.9 68 3.1 13 11 2.3 Ecuador 1120 5.6 66 4.2 20 6 3.0 Jamaica 1070 5.4 73 2.6 - - -1.3 Guatemala 900 4.5 62 5.7 53 37 0.9 Honduras 860 4.3 64 5.5 40 40 0.6 Bolivia 570 2.9 53 6.0 35 17 -0.8 - not available. Source: World Bank Development Report 1990, Tables 1, 2, 27, 29 and 32; and World Bank Development Report 1991, Table 1. Mexico in the 1980s.' Consequently, though some economic recovery has already taken place in the region, surveys undertaken in the late 1980s (like those utilized in the present study) may be still affected by the recession/adjustment that lias been under way for some time. Disentangling cyclical variation from longer term trends is thus a complicated task. Finally, an additional difficulty arises from the paucity of historical data on women's employment and pay in the region. Published data cover a few broad employment aggregates but do not pursue the distinction between women and men in more detailed and meaningful presentations. The case of pay is indeed is World Bank, World Development Report 1990, Table 1. Introduction and Summary 9 telling: nowhere in the most authoritative publication of world labor statistics can one find substantial information on women's wages in Latin America.16 The diversity in the characteristics and differences in the underlying trends in the Latin American economies suggest that the study of economic performance in general, and labor market performance and poverty in particular, are a challenging task for applied research in the region. These analytical, practical and statistical considerations have shaped the present study. The papers in this study make a comprehensive effort to identify, collate and analyze data that were available at the time of writing, and use these data to establish the characteristics of and trends in women's work and pay in the region and to examine whether women are treated differently than men in the labor market. 4. Methodology The present study does not attempt to advance our theoretical understanding of issues pertaining to women's time allocation between home and market work, the differentiation of gender roles within the family, household formation and dissolution, or other issues not directly related to women's employment and pay. Instead this study utilizes (with appropriate qualifications) existing analytical approaches to investigate women's status in the Latin American labor market in two particular ways. First, we do not (and, perhaps, we will never find out) what the 'appropriate' size of female labor supply or the "appropriate" level of female pay is. However, we know that women compared to men have lower rates of both labor force participation and pay. We also know (or, reasonably, assume) that there are no innate material differences between the sexes that necessarily justify the observed gender differentials in the context of modem production that is successively characterized by more capital intensive techniques. Hence, most of the analysis undertaken in this study is based on comparisons between female workers and male workers. Second, the country case studies adhere to similar techniques and utilize comparable specifications in order to facilitate comparisons of results. 16 See International Labour Office, Statistical Yearbook, any issue, old or recent. For many countries, wages are broken down by sex at the economy wide level and separately for the manufacturing (aggregate and by about 20 industries), agriculture, transport, and storage and communication industries. However, in Latin America, such information is available only in one country (El Salvador) while there are some sporadic aggregate estimates for female and male wages in the Netherlands Antilles and Chile. 10 Women's Employment and Pay in Latin America Obviously, the immediate objective was not to provide a set of indepth country studies but to examine whether there are some common patterns and factors at work in the Latin America region. Before discussing the findings, it is worth dwelling on some aspects of the methodological approach adopted in the present studies as well as their implications for the empirical specification of the models that were used. Basically a human capital framework is utilized, that is, it is assumed that education and acquired labor market experience are among the most important determinants of individual earnings. The merits and limitations of the human capital approach are well known.'7 Chapter 5 provides a detailed exposition of the methodological and practical problems in the study of discrimination and below we highlight two aspects of the discussion with additional reference to another important issue in the case of women, that is labor market selectivity. Type of education. There was no information in our data sets about the type of human capital held by women and men. The data on education (in effect, schooling) are reported simply in years (or highest grade completed) with no reference to the type of education which the individual has acquired. This lack of information necessitates the adoption of the uncomfortable assumption that there are no differences in the type of education acquired by women and men. In this way, our results may overstate the extent of sex discrimination in the labor market. However, we hasten to add that this may not be as serious a problem in Latin America as in the case of industrialized countries. The reason is that relatively few women in the region have attended school beyond the second education level. Many women workers have not even completed lower secondary education and it is at the end of lower secondary education when studies become specialized. In fact, even as late as in 1980, about 11 percent of all females in the region aged 15 to 24 were illiterate, 17 percent in the 25-34 age group and as many as 26 percent in the 35-44 age group."5 In conclusion, only a few observations in our samples are affected by the failure to standardize for the type of education women and men acquire. Actual versus potential experience. There was no information in the data sets about actual labor market experience. This statistical defect does not usually present problems in the case of men. Men are typically found in the labor force during most of their lives. Hence, potential experience (that is, the difference 17 For recent evaluations of the human capital specification in the study of labor earnings see Siebert (1985), Willis (1986) and Dougherty and Jimenez (1991). I UNESCO, 1990. Introduction and Summary 11 between age, and years of schooling and conventional school entry age) should be a fair approximation of men's actual experience. However, many women have interrupted work careers. Hence, potential experience usually overestimates the actual labor market experience of women. In the present context, the implication of using inappropriately measured experience understates the significance of this variable for women's earning power and overstates the extent of discrimination. There is no way out of this difficulty until more detailed data become available."9 In the meantime, it can be noted that studies that had access to more complete data sets have shown that a substantial part of the pay gap between women and men remains unexplained, even if data on actual experience for women are used.' This conclusion still holds when 'imputed" (that is, estimated from family characteristics) experience is used in an attempt to decrease the bias arising from the use of potential experience in the case of women.2" However, one can add that, as in the case of education mentioned in the previous paragraph, the use of potential experience in Latin American countries may not be as damaging as in the case of industrialized countries. The reason is that the average age of women workers in our samples was typically about 35 years and as low as 31-32 years in Bolivia, Mexico and Peru. Thus the average age of women in the region is lower than that in industrialized countries and the measurement error between actual and potential experience should be correspondingly lower. In addition, the typical female age participation profile in the region suggests that women do not usually reenter the labor market after an interruption in employment. As a result, it is possible that many of the working women in our samples may have been continuously in the '9 Ofecourse, to the extent that women's labor force participation decisions are affected by discrimination in the first instance, then even the use of actual experience in the earnings functions will produce biased results. This issue is explained in detail in Chapter 5. ' Wright and Ermisch (1991) report that in the case of Britain, the use of actual experience reduces the unexplained part of the pay difference between women and men by one-third compared to the results derived from potential experience. The reduction in the part of the sex wage gap attributed to discrimination is practically the same when uncorrected and selectivity corrected earnings functions are used. 21 Miller (1987), Wright and Ermisch (1991). In fact, the latter study attributes the .success of imputed experience" to the strong predictive power of childbearing patterns for women's actual work experience (Ibid. p. 519). Similarly, an earlier study on British women concluded that the use of actual experience versus potential experience increases the percentage of the sex pay gap attributed to differences in endowments by only 5 to 10 percentage points still leaving a substantial part (up to two-thirds) of the pay gap open to a number of alternative interpretations (Zabalza and Tzannatos, 1985, Chapter 1). 12 Women's Employment and Pay in Latin America labor market since they first started work. This presumption may be valid for another reason. In general, self-employment and family work are more prevalent in developing countries than in industrialized ones. These two types of work are more compatible with work at home than dependent employment and do not necessitate an interruption of employment when family formation starts. Therefore, a higher percentage of women in the region may have had continuous work experience since they started working compared with women in industrialized countries. Finally, it is possible that many women in the region who work in the formal sector have continuous work history as women are heavily employed in the public sector. These women have access to institutionalized maternity provisions which safeguard their return to work, if they wish to do so. Hence, family formation may not have severe adverse effects upon the building of women's labor market experience in the region and the bias arising from the use of potential experience may not be significant. Selectivity bias. The issue of selectivity refers to whether workers are a random or a "selected" sample of the population. If the former applies, statistical inferences from working women about all women should be valid (within a chosen margin of error). If, however, women workers are not a representative sample of all women, then there will be estimation bias. The bias arising from selectivity is not considered to be significant in the case of men as most men are usually in the labor force throughout most of their lives. In contrast, a relatively small number of women are in the labor force at any point in time. Are, then, those women working because they face high wages in the market or because they have low productivity at home? If either is correct, then working women are not representative of all women in the economy. Under these circumstances, one should correct for selectivity bias. However, one could equally argue that most women are working at some stage in their lives and not being observed as working at the time of the survey is a matter of chance, that is, it depends on what year the survey was conducted. In this case, the sample characteristics of working women can be taken to reflect the characteristics of all women. We cannot know ex ante which of these cases is relevant to a particular sample of female workers. Hence, the present studies attempted to identify selectivity and correct the estimated earnings functions for it. The correction amounts to an evaluation of wage offers for all, working and non-working, women in the sample (who can be taken to be representative of all women in the economy) rather than to concentrate on the actual wages of working women in the sample. Empirical estimates that have been corrected for selectivity bias provide a better insight for public policy: from a developmental point of view what is of interest is not simply what happens to female workers at present and whether they are treated in an efficient (that is, non-discriminatory) way compared to men. The more important issue is whether women overall are or can become as productive Introduction and Summary 13 in the labor market as men. This is an important extension in the recent literature on discrimination and all our country studies report results which have been corrected for selectivity. 5. Main Findings A. Cross-country comparisons In this volume the characteristics and trends of women in the labor market across countries are firmly established, a task long overdue.' The female labor force is examined with respect to its size as well as its distribution by age, industrial and occupational composition, and employment status (employee, self- employed, family worker). Then, women's characteristics are juxtaposed against the corresponding data for men in an attempt to standardize for possible cross-country differences in the treatment of labor by national statistical conventions and also to account for the fact that different countries are at different stages of development. The characteristics of female employment are thus put in context and then used to establish some general patterns to the extent possible by the region's diversity. Trend in participation. Considering the region as a whole, the labor force participation rate of women was initially low, averaging only 24 percent in the 1950s. However, by the 1980s it had risen to 33 percent -- an overall increase of more than one-third or about 1 percent per annum (Figure 1.1).3 The increase was as high as 20 percentage points in Colombia and 10 to 15 percentage points in Brazil, Panama, and Mexico. In the other countries under consideration, the increase was a high single figure (five to nine percentage points) with the exception of Venezuela (three percentage points) and Chile (a I Schultz (1989a; 1990) provides a world perspectiveof women's employment but little that relates specifically to women in Latin America's labor markets. 23 Unweighted average of the participation rate of prime age women (aged 20 to 60 years) calculated for 13 of the countries under consideration. Bolivia was excluded because there was no reliable information for the 1950s: the data suggest that the female participation rate was practically the highest in the region in 1950 (75 percent) and among the lowest in the 1980s (only 29 percent). Perhaps this counterintuitive and dramatic change is the result of a change in the national definition of what constitutes work. However, no conclusions could be made without more specific information. Honduras is also excluded because the labor force participation rate in the early period refers to all women and is artificially low as it includes children and persons aged 65 and over. For the actual figures and sources see Table 2.3. 14 Women's Employment and Pay in Latin America Figure 1.1 Female Labor Force Participation Rate in Latin American and the Caribbean Countries 1950s and 1980s 33% 24% 1950a 19808 gain of one percentage point). The only country where the female participation rate declined over time was Jamaica where it dropped from 53 percent to 48 percent. However, Jamaica still has the highest female participation rate in the region. It is difficult to attribute the rise in female participation in the region to any specific factor. For example, given the stagnant, and at times adverse, macro- economic conditions of the last 10 to 20 years, one could conclude that these changes for women were achieved not via growth, but by the removal of some of the inefficiencies that might have existed in the way women, as a factor of production, were treated in the 1950s and 1960s. If this were so, the economic crisis resulted in a more efficient use of female labor which has traditionally been underutilized in the region. However, this may not be the only explanation. It is possible that the increased participation of women in the labor Introduction and Summary 15 Figure 1.2 Female Age-Participation Profile in Latin American and Caribbean Countries 1950s and 1980s (Stylized) 1980s 0L 20 50 Age force has come primarily from the expanding employment opportunities for women in the public sector. It is a well established fact that the public sector is an increasingly important employer of female labor during development.' The expansion of the public sector could have caused, in turn, some of the stagnation in the macro-economic performance of these countries. If this has been the case, the distributional effects from the quantitatively greater and qualitatively better employment of women are suspect because the poor or poorest are less likely to end up with ajob in the public sector. With the data in hand we cannot establish whether the increase in female participation has been the result of 24 One may note that our data have not allowed us to establish the relative importance of public and private sectors for female employment growth. However, the service sector, where most of the operations of the government are included, now accounts for more than 50 percent of the region's total GDP (World Development Report 1990, The World Bank, Table 3). 16 Women's Employment and Pay in Latin America greater efficiency during the recession or greater role of the public sector. Perhaps the answer lies somewhere in between the "competitive" and 'public sector" explanations, though it is harder to say whether it is closer to the former or the latter. A complicating factor is that women's labor force participation rates in the (predominantly Catholic) Latin America region were initially the lowest in the world -- save for Middle-East countries.' Hence, if there were ever to be a change in Latin America it was bound to be in the direction of greater representation of women in the labor force. In fact, this study shows that there has been considerable "regression towards the mean" in the change of female participation rates over time: countries with the lowest female participation rates in the 1950s have shown greater increases compared to countries whose female participation rates were initially high. Tfhe age-profile of parhicipafion. The increase in female participation has come from an increase in the participation rates of women aged 20 to 50 years. Figure 1.2 presents a stylized profile of female participation in Latin America by age in the 1950s and 1980s (for detailed country profiles see Chapter 2). Naturally, younger and older women have lower participation rates than prime age women. This is a commonly observed pattern in many countries and there are straightforward explanations for this (for example, school enrollment for the younger groups and the existence of savings/pensions for the older groups). However, in the early period (1950s) female participation was relatively flat across all ages. By the 1980s, the two ends of the age distribution had dipped, but participation among prime age women had increased. Figure 1.3 attempts, again in a stylized presentation, a comparison between the age participation profiles of women in industrialized countries and in Latin America. First, female participation in industrialized countries is higher than in Latin America. Second, the age profile of female participation in industrialized countries is characterized by a double peak: the first peak occurs just before childbearing starts while the second peak is reached after the last child goes to school. In contrast, there are no visible signs, at least in the 25 The underutilization of female labor in Latin America could be even more severe than that suggested by a comparison with the Middle East region since the officially low rates of female participation in the latter may be significantly affected by a "cultural reluctance" to admit that a woman is working, even when she does so (Boserup, 1970; Standing, 1981; Kozel and Alderman, 1988). Introduction and Suwmary 17 Figure 1.3 Female Participation Rate in Industrialized and Latin American and Caribbean Countries (Stylized) 0 ~~~~~~~~Industrialzed E % CL a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 20 50 Age countries studied in this volume, that women reenter the labor market after an interruption in employment.' Comparison between female and male participation rates. When changes in the male participation rate over time are taken into account, the change in female participation is even more impressive. The male labor force participation rate was lower in all our countries in the 1980s than it was in the 1950s. For example, in relative terms female participation rose even in Jamaica where the ratio of female to male participation was 55 percent in 1962 and 62 percent in 1982. In some countries the increase was dramatic. In Colombia, Brazil and 26 The current age profile of female labor force participation in Latin America is similar to the profile observed in advanced countries some time ago when women's participation "reflected a straightforward career pattern: work for pay (if you work at all) before marriage. Then stop." (Levy, 1987, p. 142: on women's participation in the early postwar United States.) 18 Women's Employment and Pay in Latin America Mexico the female relative to male participation rate almost doubled. It is worth noting that the gains of women compared to men have been universal, across all countries and all age groups. 2 Is the increase in female participation temporary orpermanent? The increase in female participation in the region occurred somewhat contrary to expectations and textbook wisdom. In particular, the low, or even negative in some countries, rates of output growth have not prevented the composition of the labor force from moving toward greater representation of women. In this respect, Latin America has not followed the pattern of industrialized countries where women appear to have been 'pulled" into paid employment during periods of consistently high economic growth and labor shortages.' Are Latin American women's higher participation rates in the more recent period the result of the "added worker' effect which can be contrasted with the "discouraged worker" effect? The former effect suggests that more women enter the labor market during periods of economic recession in an attempt to preserve family income and level of consumption. The latter effect suggests that women drop out of the labor force during periods of recession because expected returns to search are not worth considering: wages are low and the probability of finding employment' is small. We do not think that either effect has much to do with the rise in female participation in the region, even more so for the dominance of the added worker effect over the discouraged worker effect. There are three reasons for this belief. First, both the added and discouraged worker effects are operating at the margin and relate to cyclical, not long term, variation. Our observation period is sufficiently long for the task in hand (from the 1950s to the 1980s) so that any cyclical effect should not be sufficient to distort the overall picture. Second, empirical studies have consistently found that in the case of women, the discouraged worker effect is the dominant one. One of the reasons for this is that during an economic recession women leave employment and become economically inactive in contrast to men who move from employment to unemployment.' Third, and finally, the drop in the participation rates of n See Chapter 2. 28 See Mincer (1962), Cain (1966), Oppenheimer (1974) for the American case; Nakamura, Nakamura and Cullen (1979a) for Canada; Joshi, Layard and Owen (1985) for Britain. For a survey of the experience of the industrialized countries see Killingsworth and Heckman (1986). 29 For example, Joshi (1981) found no evidence that women in Great Britain have a different degree of cyclical change in employment than men though specific groups of women and men (such as the younger and pensioners) have different trends of cyclical instability. Introduction and Summary 19 younger women is not compatible with a dominant added worker effect. We, therefore, conclude that the rise in female participation that our data suggest for the 1980s compared to the 1950s is due to an underlying trend and cannot be attributed to the recession which has hit the region so hard in the eighties. Employment dissimilarity between female workers and male workers. Another aspect of the labor force is the employment distribution of workers. We examined data by industries and occupations. The problems associated with the examination of qualitative characteristics of employment are well known.0 Despite these limitations, the present evidence supports the view that women's employment distribution across occupations and industries is substantially different than the employment distribution of men. However there are signs that occupational and industrial dissimilarity in the region is decreasing over time. Dissimilarity among all workers decreased in seven of the countries studied in this volume while it rose in six countries.3' In addition, the decrease in dissimilarity in the former countries was greater (in absolute terms) than the increase in dissimilarity in the latter countries. It is worth noting that when changes in dissimilarity are examined separately for the groups of self- employed workers, employees and family workers, the reductions in dissimilarity were greatest and more uniform among employees followed by self- employed workers. Therefore, workers in paid employment (rather than family workers) appear to have been the main beneficiaries from the increase in women's employment during the last 30 years. One should not, however, jump to the conclusion that dissimilarity is less of an issue today than it was in the 1950s. The reason is that, despite the reduction in employment dissimilarity between women and men over time, a higher percentage of the labor force today is in sex stereotyped employment because there are more female workers in the total labor force than ever before. This finding is important because it shows that the developmental effort of a country does not depend only on developments in the labor market but it also depends on the behavior of non-workers (this observation is of similar nature to the remarks made earlier with respect to selectivity bias). Two more observations can be made with reference to the decrease in dissimilarity over time. First, part of the reduction has been due to the decline in the agricultural sector and the concurrent decrease in the prevalence of unpaid family work. Obviously, this is a process that is bound to level off quickly -- in fact, it has done so in some of the more advanced 30 These problems are examined in detail in Chapter 3. 31 There was insufficient information for two countries, namely Bolivia and Brazil. 20 Women's Employment and Pay in Latin America countries in the region.32 Second, the reduction in the employment dissimilarity was greater among employees than among the self-employed and family workers. It is possible that the public sector "explanation" is relevant here as governments are often seen to be an equalizer of opportunities in employment matters.33 Formal and informal employment. The 'formal sector and in particular dependent employment ("employees") were found to be more important for women than men: the percentage of women workers in these sectors is higher than the corresponding percentage for men. In addition, the percentage of employees in the total female labor force has increased over time in all but six of the countries in the region.' These findings were somewhat unexpected: women are typically perceived to be working in the informal sector, most often within a family context, and dependent employment is usually affected more by adverse economic conditions than self-employment and family work. Again, it is possible that the increase in the share of employees in the female labor force and the greater reduction in dissimilarity among workers in the formal sector are due to the employment growth in the public sector. However, these findings may necessitate a reconsideration of the oft-quoted importance of the informal sector for women in developing countries. The experience of industrialized countries lends additional support to the importance of the formal sector in the long-run: historically the percentage of the labor force in dependent employment has increased in industrialized countries. 32 The contribution of agriculture to GDP decreased by almost 40 percent in Latin American and Caribbean countries between 1965 and 1988 and now accounts for only about 10 percent of all output. In other developing areas, such as East Asia and Sub- Saharan Africa, the reduction was less severe (about 20-30 percent), despite the fact that in both regions agriculture originally accounted for a much greater percentage of their GDP, around 45 percent (The World Bank, 1990: Table 3). 3 For example, among industrialized countries sex anti-discriminatory legislation was first introduced in the public sector and only later extended to the private sector (Gregory and Duncan, 1981; Tzannatos and Zabalza, 1984; Gunderson, 1989; Killingsworth, 1990). In fact, some empirical work in developing countries has shown that the unexplained wage difference among workers in the public sector are practically zero (see Birdsall and Fox, 1985, on the case of teachers' pay in Brazil). 34 The fact that women employees as a percentage of the labor force have increased over time in all but six countries in the region lends additional support to the view that the rise in female participation is not of cyclical nature. Introduction and Summary 21 Potential gains from gender equality in the labor market. The final issue examined in this volume is based on the observation that, if employment and pay differentials between women and men reflect some kind of inefficiency in the labor market (discrimination is only one such inefficiency), then the removal of this inefficiency should improve the competitive functioning of the labor market. The result will be, on the one hand, a reduction in labor market gender differentials and, on the other hand, an increase in the level of output and also an increase in real wages. In this context, we attempted to evaluate the following hypothetical question: what would happen to output and pay in the region, if pay and employment differentials were eliminated -- assuming that women and men are identical factors of production?' It is difficult to obtain a precise answer to this question. However, similar studies have been undertaken for advanced countries and we repeated this analysis for Latin America bearing in mind that the results are only 'upper bound" estimates. Subject to a number of qualifications, our simulations suggest that the potential gains in output and rise in female wages could be substantial. Output may increase by about five percent and female wages by about 50 percent. Of course, such gains may not be easy to achieve even within the time span of a whole generation of workers. In addition, achieving complete gender equality in terms of labor market outcomes may be neither feasible (for example, due to physical differences between women and men) nor desirable (many people may be content with the present division of labor in the market and at home). However, the magnitude of the results in our simulations suggests that, even if a relatively small part of existing gender differentials in the labor market is due to some form of discrimination, its elimination is a policy issue worth exploring further. B. Country studies The companion volume focuses on the position of women in the labor market within each country. In particular, two issues are singled out as the most important for analysis within countries. First, the decision to participate in the labor market, the sine qua non for paid work, and, second, female labor .prices' (wages), the prime signal for economic efficiency and an important 35 This question was answered by assuming that women change occupations within the industries in which they are currently employed until occupational wage differentials are eliminated. This simulation assumes that women and men are perfect substitutes in production (on the labor demand side) and women are willing to work in all types ofjobs currently undertaken by men and vice versa for men (on the labor supply side). In this respect, the results overestimate the gains that can be achieved but there are other factors which may mitigate this bias. These considerations are explored in more detail in Chapter 4. 22 Women's Employment and Pay in Latin America individual's material welfare. Representative participation functions and earnings fimctions are estimated, and an attempt is made to answer the question as to whether women's productive characteristics in the labor market are rewarded in the same way as men's. If not, it may be possible to improve efficiency and alleviate poverty by introducing measures to make the functioning of the labor market more competitive (sex blind). T7he determinants offemale participation. The majority of studies confirm that women's decisions to work for pay and enter the labor market is greater (1) as they enter adulthood and up to the age of 40 to 45 years (after controlling for fertility); (2) if they reside in urban areas;' (3) the higher their education qualifications; (4) the more general (rather than technical/vocational) their education;37 (5) the lower their family responsibilities (in terms of number of young children present in the household and whether they live in a male or female headed household); and (6) the lower other income and family wealth is (such as husband's earnings and house ownership). Figure 1.4 shows typical patterns of female participation in the region by some key factors with reference to Costa Rica. None of these findings is surprising but the effect of education on participation is particularly worth noting as education is potentially an important policy variable. Wage differentials. On the question of what accounts for the pay differences between female and male workers, most country studies suggest that differences in human capital endowments between women and men explain only a small I The exception to this is Kingston, Jamaica, where women have ceteris paribus a lower propensity to participate in the labor force than women in other areas of the country. 37 This finding derives from the few studies in this volume which were able to pursue this distinction. However in Argentina the possession of a commercial or technical secondary qualification appears to have a greater impact on the decision to participate in the labor market. This finding is subject to two qualifications. First, Argentina is among the most developed countries in Latin America. Hence, there may be more demand for specialized skills. Second, the sample is drawn only from Buenos Aires, the most developed area of the country, and it is unlikely to reflect accurately the characteristics of the whole country (for example, almost 30 percent of working women in Buenos Aires are employed in the public sector while more than one in three women are involved in part-time employment and almost one in ten working women are foreign born; none of these characteristics is likely to hold outside the capital city). In any case, other research on the issue Ihas typically found that returns to vocational education are not as high as those to general education (Grasso and Shea, 1979; Meyer and Wise, 1982; Psacharopoulos 1987). Introduction and Summary 23 Figure 1.4 Female Labor Force Participation Rate (per cent) by Selected Characteristics Costa Rica 1989 Married - O 18 Single - 0 40 Rural - 20 Urban - = 29 Non-head of Hshld - 23 Head of Hshld - 34 Education: Primary< - _ _ 17 Primary - 22 Secondary - 31 University - 38 proportion of the wage differential. The way the labor market rewards productive characteristics appears to be dependent on whether the holder of these characteristics is a woman or a man. Detailed estimates are shown in the individual country studies that follow, but as a summary statement one can say that on average only 20 percent (or five percentage points) of the sex wage differential can be explained by differences in the stock of human capital that women and men have acquired. The remaining part of the sex wage gap can be seen as the upper bound of discrimination in actual wages. However, after correcting for women's self-selection in the labor market, the estimates suggest that an additional 20 percent (another five percentage points) of the unexplained gap that was previously attributed to discrimination is due to women's earning power (wage offers) being lower than that of men. Hence, the unexplained part is reduced to about 60 percent of the sex wage gap (or 20 percentage points) (Figure 1.5). One can mention here that the value of our estimates rests on the fact that studies for other countries (mostly advanced ones) have estimated lower .upper bounds" than those presented in this volume. We are not aware of any study, or collection of studies, which have addressed the issue from the point of view of a region (rather than individual country). 24 Women's Employment and Pay in Latin America Figure 1.5 Decomposition of the Male-Female Wage Gap (Stylized) Wage Upper bound of discrimination SDue to selectivity Wf _t Due to differences in human capital 7 0 - - - -- - - - - - Women Men 6. Policy Issues T he direct and indirect effects of women's employment cannot be overstated. The relationship between female participation in the labor force, fertility, women's overall economic welfare and social status, and growth is well established in the literature. What is less certain are the causes for the observed differentials in employment and pay between women and men. The issue is a complex one but our findings suggest that policies which enhance women's productive characteristics (labor supply), eliminate differential treatment of workers at the workplace on the grounds of sex (labor demand) and which generally improve the competitive functioning of the labor market (institutional framework) can be both growth enhancing and self-financing. In addition, their distributional effects will be in the right direction as women are predominantly employed in low pay, low status jobs. Below, we highlight some key areas where (1) social policy can have rewarding returns; (2) more analytical work can increase our understanding of the functioning of the labor imarket; and (3) there is a need for better statistical information. Introduction and Sunmary 25 Women, childbearing, and the decision to work for pay. A consistent finding in all studies in this volume is that, after controlling for other factors, women's propensity to work for pay is high even during the childbearing age. In this respect, women's behavior appears to be ex ante similar to that of men. However, the actual age profile of female participation dips during the reproductive age and all country studies confirm the negative effect of children upon women's decision to work for pay. The conflict between productive and reproductive decisions is obvious. In fact, it is this asymmetry, in part biological and in part stemming from societal norms, which largely destines women to the observed employment and pay characteristics in the labor market. A number of options exist which can relieve women from some of the burden of family formation and increase women's contribution to production and women's welfare. One such measure is improving women's understanding, especially in rural or relatively poor areas, of how to avoid unwanted pregnancies. One may note here that education increases the level of contraceptive efficiency and lowers the expenditure on contraceptives necessary to reduce the risk of pregnancy at a given level.8 An additional increase in women's work effort can come through the encouragement of women's reentry into the labor market after an interruption in employment. This can be achieved by the provision of effective and cost- efficient pre-school and child-care facilities. Recall that the typical pattern for women in the region is to withdraw from the labor market upon childbearing with little tendency to reenter the labor market. The usual approach has been for governments either to provide such child-care facilities free heavily subsidized or not to provide them at all. Where free child-care facilities are offered, these have been largely urban based with a relatively limited number of places. As a consequence, the most needy groups have seldom been the beneficiaries of the subsidies. Offering pre-school care with selective cost recovery measures along social cost-benefit lines would enable more women to enter employment and, subsequently, to improve their income potential. It would also assist children from disadvantaged backgrounds by exposing them to organized pre-school education and by improving their socialization.39 In addition, day-care can provide a medium through which children can be reached with targeted immunization, nutrition and other programs. 38 Michael (1974); Rosenweig and Wolpin (1982). 39 "New research indicates that our fears about average day-care programs are baseless: it shows that typical, not just ideal, day-care seems to have no ill effects...' (Nakamura, Nakamura and Cullen, 1979a, p. 135). 26 Women's Employment and Pay in Latin America The family structure observed in industrialized countries is not that typical in the Latin American region. Internal and overseas migration ("women as urban domestic servants and men as industrial workers abroad") is quite significant while in some areas, especially in the Caribbean, visiting partnerships are not an uncommon form of arrangement. Also, given the longer life expectancy of women and the fact that in most marriages women are younger than their husbands, widowhood even during prime age is not uncommon."' In addition, societal norms may not encourage remarriage.41 Finally, and in more general terms, the growth in the number of divorced and separated mothers is considered to be one of the primary reason for the continuing increase in single parent families.' These complex socio-demographic effects throw women into a vicious cycle of inability to work and poverty. In countries examined in this volume, female headed households accounted for between 10 and 15 percent of the sample in Argentina and Venezuela, and for as much as one-third in Jamaica (and around 50 percent in the Kingston area alone). Consequently, policies which directly (via the elimination of provisions in family law and taxation regulations which induce asymmetry in the treatment of women with respect to family/employment decisions)4 or indirectly (via reducing the burden of child care) improve women's employment opportunities during the critical period of family formation are bound to have beneficial efficiency and distributional effects. The efficiency issue is self-obvious. In distributional terms, economic theory predictse and empirical evidence suggests' an inverse relationship between income/class position and marital instability. Whether such policies should be adopted does of course depend on costs. This is an area of research with potentially significant returns. Education and women's work and pay. Another systematic finding of the studies reported in the companion volume is the effect of education upon women's employment and pay. The participation functions show that, after 4 Mohan (1986). 41 Rosenhouse (1988). 4 Ermisch and Wright (1990). 43 A study of legal provisions which differentiate between women and men in the family and the labor market is already under way in the World Bank. 44 Becker, Landes and Michael (1977); Becker (1981). 45 Goode (1956, 1962); Bishop (1980); Kieman (1986); Peters (1986). Introduction and Swnmary 27 for other factors, the probability of participation is greater the higher the woman's educational qualification. Similarly, women's earnings increase as fozmally acquired educatioz increases. Although the issue of occupational choice has not been explicitly addressed in this study, the effect of education upon a woman's propensity to work and her level of pay is sufficiently clearcut to guide public policy. Increasing opportunities for female education would enhance efficiency and alleviate poverty.' When women stay longer in the education system their natural (maximum) fertility rate is reduced. In addition it has been shown that there is a strong negative effect between female education and family size through a price substitution effect as well as birth control knowledge and contraceptive efficiency.' Finally, women are exposed to influences which typically alter their preferences away from the traditional view of the family toward fewer children.4' These effects are implicit in Figure 1.6 which shows the relationship between education and fertility among women teenagers in four of the countries in the region. Apart from an effect via lower fertility, education increases women's propensity to work because the opportunity cost of staying at home (foregone income) also increases.4' Women's greater attachment to the labor market can subsequently augment family income (in a conventional family context) and can help reduce the incidence of poverty among prospective female headed households.' The increase in female human capital also assures a more effective use of half of the country's potential work force and induces men to work in a more competitive environment. Finally education enhances family production as broadly defined. Children's well-being and educational attainment has been found to be highly correlated to mother's education. In addition, educated women are in a better position to prepare '6 Blau, Behrman and Wolfe (1988); Psacharopoulos and Tzannatos (1989). 47 Heerand Turner (1965); Westoff (1967); Harman (1969); DaVanzo (1971); De Tray (1972, 1973); Cochrane (1979); Kelly and Da Silva (1980); Da Silva (1982); Mueller (1982). 4' Easterlin (1969); Tzannatos and Symons (1989). 49 Khandker (1987, 1988); Psacharopoulos and Tzannatos (1991). 50 Schultz (1969b). 28 Women's Employment and Pay in Latin America Figure 1.6 Female Labor Force Participation Rate in Latin American and the Caribbean Countries 1950s and 1980s Annual number of live births per 1,000 women aged 15-19 No education 278 m Primary Secondary 248 Higher 203 Oominican Rep. El Salvador Brazil Peru Source: The Al1an Guitmacher Institute, Today's Adolescents, Romorrow 's Parents: A Portrait of the Amerncas, 1991. meals in a more hygienic way and can look after ill members of the household in a better way.5' Female earnings also increase with schooling, and do so faster than male earnings. Thus, the distributional effects of more/better female education are warranted and desirable from a social cost-benefit point of view; the same marginal investment (one additional year of education) often yields higher returns for women than men. Figure 1.7 shows another typical profile in the region, this time with respect to education and earnings. 51 Chiswick (1974); ]_eibowitz (1974); Haveman and Wolfe (1984); Michael (1984). Introduction and Sunmary 29 Figure 1.7 Female Monthly Earnings by Educational Level Costa Rica 1989 C34,663 C18,162 C12,908 C10,389 No education Primary Secondary University Note: in current Colon Is a policy of expanding female education desirable given that the average length of schooling among female workers is already higher than that of men? The answer is wyes" because what is relevant is not the educational composition of the female labor force but that of the female population as a whole. The case even of the most advanced countries in this volume is telling indeed. In Venezuela, working women have, on average, 7.9 years of schooling while non- working women have only 5.5 years of schooling -- far behind the average attainment of men of 7.0 years. In Argentina, working women have 9.4 years of schooling compared to 8.8 years for men and only 7.8 years for non-working women (and the sample is drawn only from the capital city). The disparity between female and male length of schooling is even greater in the less advanced countries of the region and between urban and rural areas. Providing more education to women appears to be a sound policy direction. In terms of simple arithmetic, average female education will increase more, and in a more cost-effective way, if many illiterate women attend primary school than if a few secondary school graduates attend a four year university course. In this respect, the high rates of return to female university education reported in this 30 Women's Employment and Pay in Latin America volume need to be qualified accordingly.52 First, the most qualified workers, especially females, find employment in the public sector, and the present estimates may simply reflect this. Second, and more importantly, the earnings functions that are estimated in the conventional econometric form are based on the explicit assumption that the only cost of education is foregone earnings during the period of studies, which amounts to saying that education is a free good. This is clearly unrealistic and the difference between the returns to primary and university education is not necessarily so great as to justify the public provision of more tertiary education at the expense of lower levels of education.53 Third, and finally, the pro-rich distributional effects of the emphasis on tertiary education, rather than basic education, in developing countries have been widely documented.' T'he earnings gap. The final objective of this study was to determine what part of the difference in female and male labor earnings is accounted for by workers' individual characteristics and effort. Education, experience and weekly hours were singled out as appropriate variables. All the studies in this report found that these three variables accounted for only about one-third of the observed earnings differential. The rest of the difference is due to one of the "black boxes' of labor economics. One may argue that, as the theoretical foundation of earnings functions rests on a competitive market clearing condition, this approach to the study of discrimination does not allow differences in labor supply between the sexes to show up separately in the final decomposition of the pay gap. Alternatively, there may be genuine market imperfections, or government legislation, or societal conventions that are, at least in part, responsible for the unexplained difference. The usual approach has been to label the differences in earnings unexplained by human capital characteristics as the "extent of our ignorance" and assign it the interpretation "upper bound of discrimination." Though this may be correct in terms of semantics, it has little 52 This finding is quite common in developing countries (Haque, 1984; Khan and Irfan, 1985). 53 For estimates of the- cost-efficiency of investment on different levels and types of education see Adelman (1975); Colclough (1982); Mingat and Tan (1988); Psacharopoulos (1977, 1985); Lockheed and Hamishek (1988). 54 The unintentional distributional effects of public expenditure on education have been shown by among others Ribich (1968); Selowsky (1979); Stromquist (1986); Lockheed and Hanushek (1991). Introduction and Summary 31 practical significance. The debate this unexplained difference has spurred in advanced industrialized countries indicates the complexity of the issue.55 Despite these limitations, the present study makes some valuable points. First, the estimated unexplained difference in the countries included in this volume is larger than that found in advanced countries where up to two-thirds of the differential has been attributed to differences in the productive characteristics between workers of different sex.' In this respect, there seem to be greater differences in the treatment of women in Latin America than in industrialized countries. Part of the explanation for this difference can be the fact that markets in developing countries have not matured sufficiently to take over from custom as an allocative mechanism. Powerful norms of female seclusion, which may still apply, restrict women from achieving a status comparable to men outside the family.57 Second, it is likely that, in some countries, women's participation in certain types of employment and/or hours of work are still restricted by decree while women's pay may be determined (implicitly or explicitly) pro rata to male pay in the jobs in which they are employed. These overt restrictions continued to exist even in advanced countries (such as Australia, New Zealand, Britain and some other European countries) until they were repealed by equal pay/employment legislation enacted mostly in the 1970s.'s Despite the fact that the present authors did not examine institutional factors governing collective pay and employment determination in the countries studied, some evidence was identified supporting the view that such discriminatory arrangements exist in Latin America.59 There can be no economic justification for such employment or pay restrictions and, to the extent that such practices exist, concern for 5 For the detection of discrimination and applicability of policies designed to eliminate it see Zabalza and Tzannatos (1985); Dex and Sloane (1988); Gunderson (1989); Siebert (1990). s Cain (1986); Killingsworth (1990). 57 Cain, Khanam and Nahar (1979). " See, among others, Gregory and Duncan (1981); Tzannatos and Zabalza (1984); Tzannatos (1984, 1987b). 59 See chapters on Bolivia and Venezuela. 32 Women's Employment and Pay in Latin America overall economic (allocative and distributional) efficiency calls for their elimination.' Women's status and economic development. This is hardly measurable with the economists' tools. Most work in this area has come from other disciplines, especially sociology. However, if the general policy directions suggested in the present volume are followed and women's economic roles increase, it is clear that women's status will improve in social terms with subsequent beneficial effects on the macro-economy.6' For example, it has been argued that obtaining a job for wages outside the family enables women to control the returns to their own labor: the exercise of such a control has been found to augment women's relative power in the allocation of household resources.2 The improvement.in the economic status of women relative to men can, in turn, be associated with specific consumption patterns, investments in the health and nutrition of women and children, reduced child mortality, and eventual declines in fertility. This line of argument follows the observation that the reproductive function/power of women becomes less important for women the more they have other secure power (mostly economic) bases.' Future research. The present research has shed light on some female issues in Latin America's labor markets but a number of issues remain unresolved. An agenda for future analysis can include the following issues: 1. The existence of protective or other forms of legislation in the types of employment potentially accessible to women. If such provisions prohibit, 6D Something that should, perhaps, be stressed here is that, even if female and male productivity and wages rose at the same rate (that is, even if the gender wage gap persisted), this could be sufficient to increase women's participation in the labor market, reduce the onset of marriage and diminish lifetime fertility (Layard and Mincer 1985). However, the issue is not whether the economy will eventually get in an appropriate growth track in the long run (via the effect of real rather than relative wages upon female labor supply) but whether this process can be facilitated by a more competitive functioning of the market. 61 Deere et. al. (1982) note that there exists a complex interaction between various aspects/tiers of development (at international, national, regional, class and household levels) which result in a dynamic relationship between women's status and the macro- economy. 62 Boserup (1970). 63 Safilios-Rothschild (1982). See also Schultz (1969a). Introduction and Summary 33 in effect, women from certain activities, perpetuate sex stereotypes, and limit competition, their effects should be evaluated and corrected. 2. The nature of pay and employment determination -- especially public sector policies and collective labor market arrangements in the private sector. It is probable that wage setting is not geared to productivity and overall labor market conditions but it may reflect narrow interests of certain groups or societal norms no longer beneficial to modem production. 3. Labor market imperfections arising from different treatment of full-time vis-a-vis part-time, seasonal, and temporary work. The latter type of employment is undertaken primarily by women and can be discouraged by the existence of fringe benefit obligations on employers which are not fully prorationed compared to full-time employment. 4. The regional distribution of educational attainment and access to education, especially between metropolitan, other urban, and rural areas. The provision of public subsidies should be targeted to sectors where returns are highest. 5. The importance of the public sector in its capacity as an employer and as the main supplier of education. It is not uncommon for public policy to be dictated by short run political considerations rather than from the point of view of long term growth (dynamic efficiency). 6. The provisions included in public policy and their allocative and distributional effects, especially with respect to social policies associated with the labor market (such as unemployment benefits, income support measures, and pensions). 7. The distinction between, and relevance of, the formal and informal sectors and the implications for employment and overall developmental strategies adopted by the government. 8. The existence of statutes which provide for the differential treatment of women and men in the household (such as family law and taxation). The need for better statistical infornation. One should emphasize that the scope of the analyses undertaken in the present research has been limited by the quality and coverage of data raised by household surveys in the region. In particular, in some (and, sometimes, in all) countries: 34 Women's Employment and Pay in Latin America 1. The surveys did not always provide information about the industrial sector of the worker and/or the occupational sector or the number of weekly hours worked. 2. One of the most important variables for the analysis of women in the labor market, namely actual labor market experience, was not found in any of the databases. 3. There was no useful information about the relationship between individuals in the same household: it would have made a lot of difference in the precision of female labor supply estimates if the researchers knew whether an extra adult in the household survey was a mother-in-law, sibling of either spouse, temporary visitor, or a domestic laborer. 4. Another useful piece of information typically missing from our databases was the number of days or weeks an individual was absent from work and the reasons for absence (for example, ill health, family reasons and so on). 5. On the issue of ealnings, some surveys were undertaken in stages over a lengthy period and this has impaired the accuracy of earnings for estimates in countries with high inflation. Delays in data collection can also affect employment estimates, where employment is seasonal or cyclical. 6. Information on education/training needs to be more specific. Type of education needs to be included in the questionnaire (such as general academic, technical, vocational), as does quality, as proxied by the distinction between public/private school or location of school in rural/urban area. These questions should also be asked of training. 7. There is, finally, a need for improving the quality/stratification of the surveys. For example, in some databases the distinction between general and technical education was pursued but, since few women undertake the latter, the information that could be usefully utilized consisted of a handful of observations even at country level. Attempts to obtain more disaggregated information (for example, by region) proved of little value since in some cases there were hardly more than two or three observations left. Unless the quality and scope of data improve, it will not be possible to take the analysis beyond some aggregate quantitative issues. Given the regional diversity Introduction and Summary 35 that exists in developing countries with respect to their economies and societies, the need for broader and more accurate information is much greater than in industrialized countries. 2 Trends and Patterns in Female Labor Force Participation 1950-1985 1. Introduction This chapter examines the broad patterns and trends in the women's labor supply during the last 20 to 30 years. We look at the size of the female labor force and then at the female participation rate.' We also compare these two indicators of female labor supply with the respective figures for men. In this way we standardize for possible differences in the statistical treatment of labor at the national level, and also for the fact that countries are at different developmental stages. The female participation rate is also examined with respect to its age distribution and its distribution by employment status (dependent employment versus self-employment and family work). The main findings are, first, that female participation rates have increased as such and also in comparison to male participation rates. Second, the increase has come from prime age women. The participation rates of younger and older women has declined over time. Third, in all but six countries in the region there has been a shift in the locus of female employment from the informal to the formal sector. 2. The Size and Growth of the Labor Force Table 2.1 shows the size of the total labor force by sex in 35 Latin American countries sometime in the 1950s and 1980s. The figures refer to all those 1 The labor force participation rate is the ratio of the economically active population ("labor force') to the population as a whole ('population at risk"). For a discussion on the theoretical and statistical shortcomings of the labor force and participation rate see Bowers (1975); Standing (1981); Psacharopoulos and Tzannatos (1989). 38 Women's Employment and Pay in Latin America engaged in an economic activity broadly defined, that is, they include the self- employed/employers/own-account workers, the wage/salaried employees, and family workers as well as those classified as unemployed; and to all age groups. The data are drawn from national population censuses. Subject to the many caveats which apply to the theoretical classification of an activity as an economic one and to national differences in the statistical treatment of work and unemployment, two points can be made even at this level of aggregation.2 First, in the late 1950s the male total labor force was about 54 million while the female labor force was only 14 million. This implies that for every four male workers there was only one working woman. Therefore, the initial representation of women in the labor force was, by all standards, low. The only region with lower representation of women in the labor force was (and still is) the Middle East.3 Second, by the early 1980s the size of the male labor force had increased by about 33 million. The corresponding increase for the female labor force was 20 million. There were, therefore, 87 million male workers and 33 million female workers by the 1980s. This implies that the majority of new jobs during the last 25 years have been taken by men. However, the smaller initial size of the female labor force means that there has been an impressive rate of employment growth among women: it rose by 140 percent compared with half that figure for men (60 percent). As a result, the number of working women for every five working men has now increased to almost two. This compares more favorably with the rest of the world where the ratio of the male to female labor force is about 5:3.4 2 It has been shown that even within the same advanced country the long run measurement of female labor supply can vary considerably due to differences in either the census definition of work (Joshi and Owen, 1984, on "how elastic labor supply can be') or the particular year and stage of the economic cycle to which the census observations relate (Joshi and Owen, 1985, on whether "elastic retracts"). Part of the problem relates to women's flow, real or statistical, between work and inactivity rather than between work and unemployment (Lundberg, 1985). 3 One of the factors responsible for the low female participation rates in the Middle East (which is, however, less applicable to Latin America) is the cultural environment which does not encourage women to work in the open labor market (Boserup, 1970; Kozel and Alderman, 1988). 4 Sivard (1985). Female Participation: Trends and Patterns 39 Table 2.1 Total Labor Force in Latin America (selected years) Early Period Late Period County Year Male Female Year Male Female Bahamas 1953 31862 20086 1980 48275 38777 Barbados 1960 54478 36591 1980 57834 45199 Belize 1960 22123 4883 1980 36585 10742 Bermuda 1960 12700 6744 1980 17232 14204 Cayrnan Islands 1960 2229 930 1979 4711 3408 Cuba 1953 1706477 353182 1981 2434069 1106623 Dominica 1960 13328 10081 1981 16698 8635 Dominican Rep. 1960 732220 88490 1981 1361109 554279 El Salvador 1961 663273 143819 1971 914324 252155 Grenada 1960 16392 10922 1970 17482 11200 Guadaloupe 1961 70029 44238 1982 71220 52668 Guyana 1960 134828 39902 1980 180084 59247 Guyane Fr. 1961 8309 3672 1982 20786 11589 Haiti 1950 890756 856431 1982 1257416 872245 Martinique 1961 56952 35392 1982 72207 58293 Nicaragua 1963 379305 95655 1971 395003 110442 Paraguay 1962 453520 132895 1982 834308 204950 Puerto Rico 1960 449840 144260 1970 471369 212421 Suriname 1964 61196 19003 1980 58091 22730 Trinidad/Tobago 1960 203732 74415 1980 266592 108121 Total 5963549 2121591 8535395 3757928 Countries studied in this volume Argentina 1960 5879054 1645415 1980 7278034 2755764 Bolivia 1950 779691 581536 1976 1164619 336772 Chile 1952 1616152 539141 1982 2720822 959455 Brazil 1960 18597163 4054100 1980 31392986 11842726 Costa Rica 1963 330879 64394 1984 626633 177560 Colombia 1951 3054420 701189 1985 6419607 3138261 Ecuador 1962 1207235 235356 1982 1861652 484411 Guatemnala 1964 1196745 166924 1981 1449058 247406 Honduras 1961 494717 73271 1974 643056 119739 Jamaica 1960 401191 253391 1982 433312 275130 Mexico 1960 9296723 2035293 1980 15924806 6141278 Panama 1950 212248 52371 1980 394012 152840 Peru 1961 2445427 679152 1981 3978410 1335481 Uruguay 1963 759987 262280 1985 785944 390864 Venezuela 1961 1929421 421870 1981 3387892 1305876 Total 48201053 11765683 78460843 29663563 Grand Total 54164602 13887274 86996238 33421491 Source: National Population Censuses. See ILO (1990), Table 1. 40 Women's Employment and Pay in Latin America In conclusion, the size of the female labor force in the 1980s was about two and a half times bigger than in the 1950s. The annualized rate of growth of the female labor force (2.97 percent) was almost double the corresponding rate for men (1.59 percent) during the aforesaid period. Women's labor supply appears, therefore, to have increased considerably both in absolute and relative (to men) terms in a short period of time. 3. The Relative (Female to Male) Labor Force An examination of the female relative (to male) labor force can standardize in part for the across country differences in the statistical treatment of labor in a number of ways. It can also take into account the different conventions in the definitions of employment. For example, a country which appears to have a small female labor force may also have a small male labor force, if certain activities are not considered to be economic ones.5 In addition, expressing the female labor force in terms of the male labor force should eliminate some of the differences stemming from the countries' different socio-economic and demographic characteristiics. These differences may refer to variations across country in education enrollments (which affect the labor supply of younger cohorts); the existence and provision of pensions and other aspects of social policy (which affect older groups); and the different demographic profiles of the countries (such as rural/urban residence and average age/life expectancy). These factors are important because employment behavior typically varies with location and age. Table 2.2 ranks the countries in the region from those which had the highest ratio of female to male labor force to those with the lowest ratio in the 1950s and early 1960s, and examines the change in the ratio over time.6 Some countries have information from 1950 well into the 1980s while in others information exists only between 1960 and around 1980. Thus, it is better to concentrate on the annual percentage rates of growth rather than on the changes 5 The implicit assumption here is that the total population in a country is shared roughly equally between the two sexes and the age distribution of the sexes is similar. 6 Bolivia is excluded from Table 2.2 because it had one of the highest female to male ratios in the 1950s (75 percent) and one of the lowest in the 1980s (29 percent). This decrease corresponds to an annual rate of change of -3.6 percent. This dramatic decline and the initially high value of female relative labor force suggests some irregularity in the statistics for which no explanation can be offered. Female Participation: Trends and Patterns 41 themselves.7 There is some clear evidence of regression towards the mean: gains have been greater in countries where the ratio of female to male labor force was initially low. Among the four countries where the ratio decreased, three countries had the highest ratios in the earlier period (Haiti, Dominica, and Grenada) and only Paraguay was somewhere in the middle. Of the countries which experienced a high annual rate of increase, most are found toward the lower end in terms of the initial ratio of the female to male labor force. The effects of the differential growth in female labor force on the regional profile of the sex composition of the labor force are summarized in Figure 2.1. The countries are ranked in descending order of the female to male ratio around the 1950s/early 1960s (as they appear in the first column of Table 2.2).s Compared with the 1950s, when the female participation rate was below 25 percent in as many as 13 countries, there are now only two countries below the 20 percent mark (Honduras and Guatemala -- 19 and 17 percent respectively). In all other countries the female/male labor force ratio is now equal to, or higher than, 25 percent. Almost one-third of the countries have a ratio of 35 to 50 percent. In another one-third of the countries the ratio is more than 50 percent. In conclusion, the female gains in the labor force appear to have been significant and have been achieved over a period of only 20 to 30 years. The region has become more homogeneous with respect to the sex composition of the labor force. 4. The Labor Force Participation Rate: Broad Trends Having examined the behavior of the female labor force in both absolute and relative (to men) terms, we now turn to the participation rate (the ratio of the labor force to the total population). We concentrate on the countries studied in this volume, which represent about 90 percent of the total labor force in the region. Table 2.3 shows participation rates for prime age workers (20-60 years), by gender (columnns 1 and 2) and then in relative terms (female to male, 7 For present purposes, the annual rate of growth is calculated as the nth root of the ratio of the latest to the earliest figure minus 1, where n is the number years which elapsed between the earliest and latest observation. This is more appropriate than taking the difference between the later and earlier figures and simply dividing the result by the difference in the number of years, which amounts to a simple linear pattern of growth. 8 Three-country moving averages are used for the both periods in order to smooth the variation. 42 Women's Employment and Pay in Latin America Table 2.2 Relative (F/M) Labor Force in Latin America and the Caribbean (percent) Early Late Annual Countrys period period change (1950s/early 1960s) (1980s) (%) Haiti 96.15 69.37 -1.0 Dominica 75.64 51.71 -1.8 Barbados 67.17 78.15 8.0 Grenada 66.63 64.07 -4.0 Guadaloupe 63.17 73.95 0.8 Jamaica 63.16 63.49 0.0 Bahamas 63.04 66.66 0.2 Martinique 62.14 80.73 1.3 Bermuda 53.10 82.43 2.2 Guyane Fr. 44.19 55.75 1.1 Cayman Islands 41.72 72.34 2.9 Trin. & Tob. 36.53 40.56 0.5 Uruguay 34.51 49.73 1.7 Chile 33.36 35.26 0.2 Puerto Rico 32.07 45.06 3.5 Suriname 31.05 39.13 1.5 Guyana 29.59 32.90 0.5 Paraguay 29.30 24.57 -0.9 Argentina 27.99 37.86 1.5 Peru 27.77 33.57 1.0 Nicaragua 25.22 27.96 1.3 Panama 24.67 38.79 1.5 Colombia 22.96 48.89 2.3 Belize 22.07 29.36 1.4 Mexico 21.89 38.56 2.9 Venezuela 21.87 38.55 2.9 Brazil 21.80 37.72 2.8 El Salvador 21.68 27.58 2.4 Cuba 20.70 45.46 2.9 Ecuador 19.50 26.02 1.5 Costa Rica 19.46 28.34 1.8 Honduras 14.81 18.62 1.9 Guatemala 13.95 17.07 1.2 Dominican Rep. 12.09 40.72 6.0 a. Countries ranked in descending order of relative (F/M) labor force in early period. Source: Table 2.1 Femalk Participation: Trends and Patterns 43 Figure 2.1 F/M Labor Force in Latin American and Caribbean Countries 0.s 0.4 0.6 0.2 0.4 Countries Cin descending order, 19506) 0 19505 + 19SOs Source: Table 2.2 column 3). Column 4 shows the annualized rate of growth of the female relative participation rate. The data show that in the early period only two countries had a female participation rate greater than 30 percent (32 percent in Uruguay and 53 percent in Jamaica). In five other countries the female participation rate was between 20 and 30 percent, while in the remamning countries the rate was below 20 percent. In contrast, by the 1980s in only one country the -female participation rate was below 25 percent (15 percent in Guatemala).9 In the 1980s, the female labor force participation rate was 26 percent in Costa Rica, 29 percent in Peru and 30 percent in Chile. In most of the other countries the female participation rate had risen to between 30 and 35 percent. In Colombia, Uruguay, and Jamaica, however, it was 39, 46, and 48 percent respectively. ' Excluding Honduras where the data refer to all ages (not prine age only); hence, the participation rates for both sexes are low and not comparable to the other countries. 44 Women 's Employment and Pay in Latin America Thus, the recent figures represent significant gains for the ratio of workers in the female population. The unweighted average of female participation rates in the countries studied was 24 percent in the 1950s/1960s compared to 33 percent in the 1980s.10 The improvement in women's participation in the labor market is even more convincing when compared to male participation. While the male participation rate has declined in all countries without exception between the two periods, the female rate has risen."1 The decline in the male rate cannot be explained by the assumption that men remain in the educational system longer than women, since younger cohorts are excluded from the present comparison.2 Neither can the decline in male participation be explained by arguments pertaining to retirement as only groups of workers below the age of 60 are examined. Thus, the increase in female participation rates can be attributed to factors that are separate from those governing the growth of the labor force as such during the process of economic development. The relative improvement in female participation is further discussed (and confirmed) in the next section with respect to the age profile of female and male participation rates. 5. The Age Profile of Female Participation Figure 2.2 shows female participation rates by age group and by country in the 1950s and 1980s (in alphabetical order). To facilitate comparisons, the participation (vertical) axis has been standardized and is measured from 0 percent to 60 percent. Howvever, the age groups on the horizontal axis have been constructed ad hoc in order to enable within country comparisons between earlier and later periods. Consequently, some of the between country variations in age profiles may not necessarily reflect actual differences. 10 Excluding Honduras for both dates where the female participation rate refers to all ages, not prime age workers. 11 The only exception is Bolivia. ILO data suggest that the female labor force participation rate was 62 percent in 1950 - the highest in the region, even higher than in Jamaica. By 1976 the same data indicate that the rate had dropped to 23 percent - in effect the lowest in the region except Honduras where the participation rate for women aged 20 to 60 years was 18 percent. The levels of and magnitude of change in female participation rates in Bolivia suggest some irregularity. 12 In any case, the scholarity ratio of those aged 18 to 22 years is low and university enrollment accounts for a very small fraction of the total enrollment in education. Female Participation: Trends and Patterns 45 Table 2.3 Participation Rates for Prime Age Groups Annual Country Participation rate (%) percentage (Age Group) Year Male Female F/M change of F/M (1) (2) (3) (4) Argentina (20-60) 1960 92.8 24.4 26.3 1980 90.8 33.1 36.5 1.6 Bolivia (20-64) 1976 94.1 23.1 24.6 - Brazil (20-60) 1960 95.0 18.2 19.1 1980 92.4 33.0 35.7 3.2 Chile (20-60) 1952 94.5 28.6 30.6 1982 87.2 28.9 33.8 0.3 Colombia (25-60) 1951 97.4 19.0 19.5 1985 85.4 39.4 46.1 2.6 Costa Rica (20-60) 1963 97.0 18.6 19.1 1984 89.7 26.4 29.4 2.1 Ecuador (20-60) 1962 97.8 17.7 18.0 1982 87.7 22.6 25.3 1.7 Jamaica (20-64) 1962 95.9 52.7 55.0 1982 78.4 48.2 61.5 0.6 Guatemala (20-60) 1964 96.2 13.1 13.5 1981 91.3 14.7 16.5 1.2 Honduras (all ages) 1961 52.7 7.7 14.7 1974 48.8 8.9 18.3 1.7 Mexico (20-60) 1960 96.5 19.1 19.8 1980 92.4 32.7 35.4 2.9 Panama (20-60) 1950 97.0 24.9 25.2 1980 87.3 35.7 40.9 1.6 Peru (20-60) 1961 96.8 22.7 23.3 1981 91.3 29.0 31.3 1.5 Uruguay (20-60) 1963 93.0 32.0 34.4 1985 92.4 46.0 49.8 1.7 Venezuela (20-60) 1961 96.4 22.1 23.0 1981 89.0 35.0 39.3 2.7 - not available. Source: Constructed from ILO (1990), Table 1. 46 Women's Employment and Pay in Latin America Figure 2.2 Female Labor Participation Rate by Age Argentina Brazil 40 4~~~~~~~~~~~~~00 30 3~~~~~~~~~~~~~~00 10 1~~~~~~~~~~~~~~00 0 0. 1.-i 1S 0-24i 20-29 0- 50-4 0-55 00-64 e 0+ 015-1 20-U4 20-20 0-41 4 5 00- 5 6 -9 -64 05- Chile Colombia x ~~~~~~~~~~~~~~x 10 10 0~~~~~~~~~~~~~~~4 15-10 26-20 3-503 4i-49 56-59 05-69 IS-I D 0-2s 30 0-44 45-52 O0+ Costa Rica Ecuador 80~~~~~~~~~~~~~~~~~0 IM4~~~~~~~~~4 30 ~~~~~~~~~~~~30.80 20~~~~~~~~~~~~~~2 0 ~~~~~~~~~~~~~~~~~~0 15-15 20-20 30-38 U45-4 55-59 00+ 10-1S 20-24 20-20 30-49 WU-0 05-50 60-04 005 Female Participation: Trends and Patterns 47 Figure 2.2 (cont) Female Labor Force Participation Rate by Age Guatemala Mexico 2 00~~~~~~~~~~~~~~~~5 40 -4 30.0 20 Ias 20 10 16-19 25-29 36-S9 45-40 55-59 65+ 15-19 20-24 25-29 S0-49 00-054 SS-W0 0-64 6S+ Panama Peru 10-10 26-20 36-30 45-4 5 0-59 26-19 25-29 36-39 45-49 55-59 065 Uruguay Venezuela 50 5~~~~~~~~~~~~~~~0 40 ~~~~~~~~~~~~~~~~40 1698 30 ~~~~~~~~~~~~~~~30 20 ~~~~~~~~~~~~20 06 I to 15-19 25-29 45-46 55-59 85+ 16-19 25-29 35-39 45-49 55-59 65 + Source: ILO (1990), Table 1. 48 Women's Employment and Pay in Latin America The age participation profile for prime age women in the second period is clearly above that for the first period. However, in }most countries the participation rates of younger women (below the age of 20) are lower in the second period than the first period. The exceptions to this are Brazil, Colombia and Mexico. The decrease in the participation rate of younger women may reflect the increase in female school enrollment. The participation rates of older women (above the ages of 55 to 60 years) have also declined in all countries in the later period -- except in Colombia and Uruguay. An explanation for the decline in the participation rates of older women can be sought in the provision of pensions and the presence of other income effects which are associated with economic growth."3 A second observation based on the age profile of female labor force participants is that an increasing percentage of women enter employment until their mid- and late-twenties but then drop out of the labor force (Mexico is an exception but only in the early period). This pattern can be attributed to childbearing by women when they are relatively young. The data are cross-sectional, rather than longitudinal and evidence from longitudinal data in industrialized countries suggests that the cohort profiles do not dip as sharply as the corresponding profiles derived from cross-section data. The reason for the different pictures provided by cross-sectional and longitudinal data is that younger women are more work oriented than older women. Nevertheless, one can still argue that the cross-sectional variation of the age profile of female participation suggests that reentry of women into the labor force after family formation takes place is not typical in Latin America. In no country in our sample is there a "double peak" in the age participation profile. These observations hold for both the earlier and more recent periods under consideration. It should also be mentioned that the dipping of the age participation profile after the mid-twenties or -thirties appears to be more severe in the later than earlier period. This decline is not confined to older age groups (such as those above the age of 50) who may have been affected by the increasing provision of social welfare or availability of pensions in the more recent period. The decline in participation rates at the time of family formation has important welfare implications: past workers have already acquired human capital which is wasted if women do not reenter the labor market. The effects become even greater when one takes into account the fact that the number of women in the labor market increases over time (hence, the percentage of the total labor force which is potentially underutilized/wasted is greater). Finally, one should not 13 This argument is consistent with the experience of industrialized countries (see Mincer, 1962; Cain, 1966; Hornstein et aL, 1982). Female Participation: Trends and Patterns 49 forget that the empirical evidence conclusively suggests considerable downward occupational mobility for women rejoining the labor market after an absence for childbearing. 14 An examination of the relative (female to male) age specific participation rate can shed some additional light on women's inroads into the labor force in recent years. What Figure 2.3 suggests is that, in practically all countries, the recent relative (female to male) age participation profiles are above the older ones. This finding suggests that the improvement of female participation relative to male participation has come from all age groups. 6. Comparisons With Other World Regions In this section we focus on differences in the level and age distribution of the female participation rate between Latin America and other world regions. These differences can be assessed with reference to Figure 2.4. The data relate to the early-1980s.15 The rate of female labor force participation in Latin America is the second lowest among world regions. The region with the lowest female labor force participation rate is the Middle East. The regions with the highest female participation rates (in descending order) are the industrial market economies followed by the Southern European countries and the Sub-Saharan countries."6 14 Martin and Roberts (1984). Of course, part of the downward occupational mobility may be due to supply factors such as women's willingness to combine family life with "some kind' of work. This complication is examined further in Chapter 5. The prevalent opinion is that women are usually overqualified compared to men in the same job (Frank, 1978) and that wage discrimination often takes the form of occupational segregation (Lloyd and Niemi, 1979; Reskin, 1984). Is For a more complete description of the data see Psacharopoulos and Tzannatos (1991). 16 For the regional classification of countries see World Bank (1983) World Tables, Volume 11, Social Data, 3rd edition, Johns Hopkins University Press, Baltimore, pp. xix- xx. 50 Women's Employment and Pay in Latin America Figure 2.3 Female/Male Labor Force Participation Rate by Age Argentina Brazil 400 4X0 20 2~~~~~~~~~~~~~~00 I: 1~~~~~~~~~~~~~~0 o o~~~~~~~~~~~~~~~~~~~~~ts I ~~~~~~~~~~~~~~~0 0S-10 20-24 25-29 30- 50-64 65-59 e0-04 u 6+ 15-19 20-24 2-29 30-4a 50-u 659 o0-04 86-00 Chile Colombia 50 x0 40 ~~~~~~~~~~~~~~~40 30 ~~~~~~~~~~~~~~~~30 20 ao 1860 15-lS 25-2s 20- 45-41 55-59 05-e9 00-Z4 25-34 35-44 45-50 00+ Costa Rice Ecuador 00~~~~~~~~~~~~~~~~5 40 I440 20 20 15-19 25-20 35-39 45-40 55-50 05+ 15-19 20-24 25-29 30-49 50-54 55-59 00-04 05+ Female Participation: Trends and Patterns 51 Figure 2.3 (cont.) Female/Male Labor Force participation Rate by Age Guatemala Mexico 00~~~~~~~~~~~~~~~~5 40 40 30 3 20 20a 0 1I-19 25-29 25-39 45-49 5-59 A6+ 15-19 20-24 25-29 30-49 5-'54 55-59 50-54 95+ Panama Peru 15-19 25-2$ 35-S 45-49 55-59 15-19 2A-29 s5-39 45-49 525-9 65+ Uruguay Venezuela z % 50 ~~~~~~~~~~~~~~~50 40 de~~~~~~~~~~~~~~~5 20 ~~~~~~~~~~~~~~~24 I: o 05-12 25-29 45-49 58-29 50+ .t-19 25-29 S5-39 45-49 55-59 65 + Source: ILO (1990), Table 1. 52 Women's Employment and Pay in Latin America Figure 2.4 Female Participation Rate by World Region and Age Groups 70 60- ~40A 30- 20- 10 15-19 2D-24 25- 30-44 35-39 40-4 45-49 50-4 59 60-/ Ago GM0 3 SSA + MEO * AAP A LAT x ME V SEU Mean Female Labor Force Partidpation by Age and Region, 1980s. SSA = Sub-Saharan Africa MEO = MWddle East, North Africa and Oil Exporters AAP = Asia and Pacific LAT = Latin America and Caribbean IME = Industrial Market Economies SEU = Southern Europe Source: Appendix Table A2-1. Female Participation: Trends and Patterns 53 The age profile of female participation rate suggests that in general female participation in most world regions dips after the age of early- to mid- twenties.'7 However, the decline in female participation after the family cycle has started is not uniform across regions. Industrial market economies, and Asian and Pacific countries experience only a mild decline in female participation at older ages. The age participation profile exhibits a noticeable decline at older ages in the remaining three regions (Southern Europe, Latin America, and the Middle East). With respect to Latin America the decline in female participation at successively older age groups is rather dramatic: the rate drops from approximately 50 percent at the age of 20 to approximately 30 percent at the age of mid-to-upper 50s. The reasons for the regional differences in female labor force participation are many and complex. However, two explanations can be singled out. One explanation relates to changes in the sectoral pattern of production and employment during economic development. More specifically a subsistence economy makes heavy use of female labor as agricultural and related activities are mostly household based. During economic development the primary sector loses its importance and the structure of production moves at first toward manufacturing and then toward services. When industrialization is under way, it is usually the case that employment growth in the manufacturing sector is not fast enough to absorb the workers released from agriculture. The net result of the relatively fast contraction of agriculture and slow expansion of industry and services is a reduction in opportunities for female employment. In addition, the labor force status of a typical woman is either 'employed" or "inactive" in contrast to men who are in most cases classified as either 'employed" or "unemployed." This economic/statistical argument (called the U-hypothesis about the pattern of female participation during economic development) is consistent with the data presented in Table 2.4. The group of middle-income countries has lower female participation rates than the group of low-income countries and high-income countries. This observation is true for all age groups. As most Latin American countries can be classified as middle-income, the U-hypothesis may have some relevance to the observed patterns of female participation in the region. Second, female participation is more dependent on non-economic factors than male participation. Male participation rates are rather uniform across countries and over time. The male labor force participation rate is around 90 percent in Latin America (see Table 2.3). Similar rates of male participation in the labor 17 The exception is Sub-Saharan Africa where the female participation rate is quite constant across most part of the age distribution. 54 Women's Employment and Pay in Latin America Table 2.4 Female Participation Rate by Age Group and Country's per Capita Income Early 1980s (percent) Country Type 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 Low Income 40 48 50 50 51 50 49 47 44 38 Mfiddle Income 24 42 42 41 40 38 36 33 27 21 High Income 39 70 65 61 63 64 60 55 41 24 Source: Constructed from ILO Yearbookof Labor Statistics, 1985, Geneva, and World Bank World Tables: Volhne II, Social Data, 1983, Johns Hopkins University Press, Baltimore. Table 2.5 Female Labor Force Participation Rate and Religion Early 1980s (pereent) Country's Dominant Religion Mean Female Participation Islam 23 Roman Catholicism 33 Hinduism 42 Buddhism 48 Confucianism 48 Christianity (other than Roman Catholicism) 49 No major religion- 58 Note: A country is classified under a given religion if 30 percent or more of the population follows that religion. Source: Based on Psacharopoulos and Tzannatos 1987, Table A-2. Female Particpation: Trends and Patterns 55 force are observed in other world regions and countries.'" Thus one can argue that male participation is relatively independent of demographic, social, cultural and other factors. In contrast, female participation appears to bear a rather clear relationship to some non-economic factors. Reference has already been made to the low female participation rates in the Middle East where the cultural environment is not conducive to women's work in the open labor market. Table 2.5 uses religion as a discriminating variable in the sense that many social norms derive directly from the religious basis of the country concerned. The data suggest that the link between custom and female work may be more real than apparent. Muslim and Roman Catholic countries have the lowest female participation rates while countries with no major religion (mostly socialist countries at the time when the data were drawn) have the highest female participation rates. Though there exist religious minority groups in some Latin American countries, the region can be safely classified as predominantly Roman Catholic. This may be another explanation for the observed low female participation rates in the region. In conclusion, it seems that some common economic and non-economic explanations are relevant to the observed female labor force participation rates in Latin America. 7. The Employment Status of Working Women We finally examine the changes in the employment status of women (dependent employment, self-employment, and family work) and compare it with the experience in other world regions. We differentiate principally between dependent employment (i.e. wage and salaried workers) and self-employment/ family work. This distinction is more important for women than men for three reasons. First, as will be shown later, the share of employees in the labor force is greater for women than men. Second, wage/salaried employment becomes increasingly more important for women during the process of economic development. And third, dependent employment is more directly influenced by public policy than self-employment and work within a family context. In practically all world regions, the percentage of employees in the labor force is greater for women than men (though this statement is not necessarily correct for individual countries). The unweighted average of the share of employees in Latin America and the Caribbean was 57.9 percent for men and 68.5 percent for women in the 1980s. The share of employees in the male labor force exceeds Is Standing (1981). 56 Women's Employment and Pay in Latin America the share of employees in the female labor force in only three countries (El Salvador, Guyana and Peru) of the 21 countries for which data exist. An interesting question is whether women in wage/salaried employment increased as a fraction of the female labor force in Latin America. It is possible that the large increase in female participation rates during the past three decades reflects increases in the number of self-employed and/or family workers as the importance of the informal sector might have increased because of the recession. In addition, it is possible that the growth of women's employment in the formal sector in Latin America might have been restricted more by labor market regulations than in other developing regions (such as Africa or South-East Asia). For example, the observation has been made that, although the fraction of wage earners has increased on average across the world since the 1950s, it fell in a number of countries "many of which are in Latin America' possibly as a result of 'pervasive minimum wage legislation and general labor market regulations. "' 19 This is an important observation in terms of public policy formulation and can be examined against the available evidence. Table 2.6 presents the ratio of female wage earners in the labor force in 77 countries grouped in six world regions for two periods (typically sometime in the late 1950s and early 1980s). Among the regional groups examined, only the countries in Eastern Europe and East Asia experienced a consistent increase in the share of wage/salaried workers in the female labor force. In the industrialized world and African countries there has been a decline in only one or two cases. In contrast, Latin America and South and West Asia each have six countries with negative growth. The six Latin American countries which experienced a decline in the share of female employees in the labor force are El Salvador, Mexico, Chile, Ecuador, Paraguay and Peru. However, a number of qualifications apply. First, Latin America is represented by 21 countries and the South and West Asia group by only 13. Even so, the latter group is artificially large as it includes Cyprus and Israel.' As the Latin American sample is more numerous than the South and West Asia group, the number of countries showing negative growth in the share of female wage/salaried workers in the labor force may not be as important as it originally appeared. In any case, among the 14 Latin America and Caribbean countries where the share of employees in the female labor force rose over time, 19 Emphasis added; Schultz (1990). ' Israel has more in common with, and should be included in, the western group rather than with Nepal or Sri Lanka. Also, if Greece is included in the industrialized group, Cyprus should be. However, the original grouping is maintained to make the present analysis comparable with Schultz's (1990) study. Female Participation: Trends and Patterns 57 the increase exceeded five percentage points in more than half (Brazil, Bolivia, Panama, Nicaragua, Guatemala, Costa Rica, Puerto Rico and Haiti). As a result the share of employees in the female labor force has risen to 68.5 percent in the 1980s from 66.5 percent in the 1950s/1960s. Second, what matters is not only the number of countries which experienced a decline in the share of employees in the labor force, but also the magnitude of the decline. The annualized percentage rate of decline has been less than 1 percent in two of the six Latin American countries, namely Ecuador and Chile. The rate of decline in the other four countries (El Salvador, Mexico, Paraguay and Peru) did not exceed 1.4 percent per annum. In contrast, the decline in countries in other regions was substantially higher. For example, among the South and West Asia countries the rate of decline was more than 1 percent per annum in practically all cases and was sometimes as high as 3.5 percent per annum. Third, Latin America started from a "high" base compared to other countries which had initially low shares of female employees in the labor force. In the predominantly Muslim countries and South Korea the share was around 16 percent in the early 1950s while in Thailand it was as low as 2 percent. In terms of unweighted averages, the early figure for Latin America was almost 67 percent compared to 55 percent or less in the other developing regions (and as low as 38 percent in South and West Asia). The decline observed in a few Latin American countries has been, in quantitative terms, relatively unimportant and could easily be due to cyclical effects. The continuing importance of wage/salaried employment for women in the Latin America and Caribbean region is substantiated further by the evidence examined in the next chapter. In addition, some recent research does not support the view that Latin American labor markets are more inflexible than elsewhere. For example, some authors have argued that there is more flexibility in hours of employment and wages in the labor markets of Latin America than is the case for many other developing economies.2" In conclusion, from a female employment perspective there does not seem to have been a decrease in the importance of the wage/salaried sector in the region during the last three decades or so. The argument that labor market regulation has been responsible for the relative decline (or slow growth) of the share of employees in the female labor force needs to be further substantiated. 21 Behrman and Wolfe (1984, p. 264). 58 Women's Employment and Pay in Latin America Table 2.6 Ratio of Employees in the Female Labor Force by World Region Region/Country Year Ratio Year Ratio Eastern Europe Bulgaria 1965 0.98 1975 1.00 Czechoslovakia 1961 0.94 1970 0.99 Hungary 1963 0.81 1980 0.94 Poland 1960 0.41 1970 0.56 West Japan 1955 0.33 1980 0.64 Australia 1954 0.89 1981 0.88 New Zealand 1951 0.91 1981 0.92 Canada 1951 0.92 1981 0.96 United States 1960 0.92 1980 0.96 Denmark 1960 0.86 1981 0.91 Finland 1960 0.65 1980 0.87 Iceland 1950 0.73 1960 0.85 Ireland 1951 0.71 1981 0.92 Norway 1960 0.91 1980 0.92 Sweden 1960 0.91 1980 0.95 United Kingdom 1966 0.94 1971 0.96 Greece 1951 0.49 1981 0.56 Italy 1961 0.69 1981 0.80 Malta 1957 0.60 1981 0.88 Portugal 1960 0.87 1981 0.79 Spain 1966 0.56 1970 0.79 Austria 1951 0.54 1981 0.85 Belgium 1961 0.71 1980 0.83 France 1954 0.59 1975 0.83 West Germany 1959 0.70 1980 0.85 Luxembourg 1966 0.65 1970 0.79 Netherlands 1960 0.84 1981 0.88 Switzerland 1950 0.87 1980 0.96 Afrkca Mauritius 1962 0.91 1972 0.89 Reunion 1961 0.82 1982 0.93 Cameroon 1976 0.03 1982 0.03 Algeria 1966 0.74 1977 0.96 Egypt 1960 0.59 1976 0.83 Libya 1960 0.36 1973 0.58 Tunisia 1956 0.06 1975 0.43 Botswana 1964 0.04 1981 0.35 Continued - Female Particyation: Trends and Patterns 59 Table 2.6 (Cont.) Ratio of Employees in the Female Labor Force by World Region Region/Country Year Ratio Year Ratio Latin America Cuba 1953 0.88 1970 0.99 Dominican Rep. 1960 0.72 1981 0.76 Haiti 1950 0.10 1982 0.20 Martinique 1961 0.82 1967 0.84 Puerto Rico 1961 0.83 1980 0.93 Costa Rica 1963 0.88 1973 0.93 El Salvador 1961 0.72 1980 0.55 Guatemala 1973 0.67 1981 0.72 Honduras 1974 0.64 1977 0.65 Mexico 1960 0.80 1977 0.66 Nicaragua 1963 0.61 1971 0.69 Panama 1960 0.78 1980 0.89 Chile 1960 0.78 1982 0.66 Uruguay 1963 0.77 1975 0.78 Bolivia 1950 0.16 1976 0.41 Brazil 1960 0.51 1980 0.76 Ecuador 1950 0.85 1982 0.64 Guyana 1946 0.66 1965 0.67 Paraguay 1972 0.52 1982 0.46 Peru 1961 0.51 1981 0.41 Venezuela 1961 0.75 1981 0.78 East Asia Hong Kong 1958 0.62 1981 0.93 South Korea 1960 0.16 1980 0.37 Indonesia 1965 0.29 1978 0.36 Philippines 1960 0.35 1978 0.41 Singapore 1957 0.74 1980 0.90 Thailand 1954 0.02 1980 0.17 South and West Asia Bangladesh 1961 0.30 1974 0.19 India 1961 0.26 1971 0.53 Iran 1956 0.58 1976 0.47 Nepal 1961 0.10 1976 0.07 Pakistan 1951 0.15 1981 0.38 Sri Lanka 1963 0.84 1981 0.79 Bahrain' 1971 0.96 1981 0.99 Kuwait 1965 0.97 1980 1.00 Syrian Arab Republic 1960 0.53 1970 0.40 United Arab Emirates 1975 0.97 1980 0.99 Cyprus 1976 0.59 1982 0.61 Israel 1972 0.84 1976 0.81 Turkey 1975 0.09 1980 0.14 Source: Schultz (1990). 60 Women's Employment and Pay in Lain America 8. Discussion and Conclusions The increase in female labor supply has been well documented in this chapter with the use of a variety of indicators (size of the female labor force and its relation to the male labor force, the female participation rate and its comparison to the male rate). ODfficial statistics in Latin America suggest that the contribution of women to employment in the past was, in general terms, low: only one in four prime age women was found in the labor force thirty years ago. However, the latest censuses indicate that this is no longer the case. By the 1980s almost one in three prime age women were in the labor force. Countries with lower female participation rates in the 1950s have experienced higher rates of growth and the region has become more homogeneous with respect to the gender composition of the labor force. The improvement in female participation is even more convincing when examined in relative (to men) terms. The annual increase in the female relative (to male) participation rate has been approximately one to two percent with the exceptions of Jamaica, which had originally the highest female labor force participation rate, and Bolivia, for which a statistical irregularity is suspected. The increase in the relative size of female employment is impressive, given that most Latin American countries experienced economic difficulties during the period under consideration. In fact, in some of the countries, per capita income has been declining over the last 25 years. The rise in female participation appears to be at present a general trend determined more by the developmental stage of a country than the country's cyclical/transitory characteristics as such. Equally noticeable is the fact that employees, as a group, have increased their share in the region's total labor force, although this increase has been neither universal nor uniform across countries. The importance of the formal sector for women's employment was somewhat unexpected given the (real or apparent) high level of government intervention in the labor markets of these countries, and the adverse economic conditions that have prevailed in the region since the mid-1970s. Equally, one could argue that the rise in the importance of dependent employment for women has been the result of expansionary government policies. The government sector is typically more female dominated in terms of employment than the private sector - if only in the areas of education, health, and social services, where jobs are predominantly filled by women as teachers, nurses and junior non-manual workers. If it were clear that the absolute and relative rise in women's employment was the outcome of uninhibited market forces, rather than the result of the potentially distortionary employment growth in the public sector and government regulation, these changes should be welcome from an' efficiency point of view. However, what Female Participation: Trends and Pauerns 61 one observes can easily be the net effect of expansionary govermnent policies with contractionary effects upon the private sector. The historical data used in this study do not allow us to establish whether the govermnent sector has been responsible for the rise in female participation. One should also not forget that Latin America started with low levels of women's labor force participation in the late 1950s/early 1960s. Hence, future gains may be more difficult to achieve than in the past 20 to 30 years. Further country specific research needs be undertaken which can look, on the one hand, at the institutional framework within which labor is priced and allocated and, on the other hand, at the effects of employment growth in the government sector. The findings of such a study would definitely yield rewarding findings for guiding public policy formulation. Given the present state of information, future policies can move in two directions. First, an attempt can be made to create an environment which could be conducive to further increases in women's participation, should the present female participation rates reflect some kind of inefficiency (such as employment and/or pay discrimination against women). It should be remembered that, despite recent gains, women's participation continues to be low. There would be two kinds of benefits associated with an increase in female participation. First, as the incidence of poverty falls disproportionately upon women (and children), an increase in female employment will mean additional monetary income for certain types of households as well as a more continuous employment record and increased labor market experience for women. The latter can in turn increase female earnings in the future. In terms of efficiency, women as a factor of production will be utilized more fully than at present. Consequently, economic growth can be faster. Second, given the significant decline in female participation rates during childbearing and women's failure to reenter the labor market after this interruption, policies contributing to the current age participation profile can be reassessed. Candidate policies for abolition are those which place women on a different footing than men in either the family or the market. This includes overt or hidden arrangements governing the determination of female pay on the assumption that women are "supplementary' workers in a conventional family context (whereas female headed households are quite common in Latin America); and the unjustified exclusion or restriction of women from certain types of employment, in certain areas, or at certain times. Protective legislation in the labor market, ifjustified, should not revert to prohibiting norms. Neither should legislation induce an asymmetry in the treatment of individuals within the family, especially with respect to benefit and taxation matters. 62 Women's Employment and Pay in Latin America Statistical Appendix to Chapter 2 The tables in this Appendix present "age-specific' labor force participation rates and the 'all ages" labor force participation rates. Age-specific labor force participation rates are calculated as the ratio of the labor force in an age group to population in the same group. The all-ages labor force participation rates are calculated as the ratio of the total labor force to the total population. Appendix Table A2.1 Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Sub-Sahara Africa Year 15-19 20-24 25-29 30-34 35-39 40-44 4549 50-54 55-59 60-64 all ages Botswana 85 42.0 74.7 76.9 76.9 73.5 73.5 69.5 69.5 60.5 60.5 36.0 Chad 80 22.0 28.1 28.0 28.0 28.0 28.0 25.2 25.6 25.6 13.1 16.4 Comoros 80 21.8 26.2 28.2 29.4 30.3 31.5 30.7 31.2 28.9 27.2 15.6 Ethiopia 84 57.9 61.0 63.7 65.5 62.9 60.8 56.8 50.1 41.2 33.8 34.3 Gambia 83 61.9 68.3 71.1 73.9 74.6 77.2 75.3 76.2 72.4 71.5 43.6 Guinea-Bissau 79 2.2 4.0 3.2 2.7 2.3 2.3 2.3 1.8 2.2 1.7 2.6 Madagascar 75 58.2 69.3 78.1 78.4 79.0 79.4 82.5 82.9 76.7 76.7 44.8 Malawi 83 78.0 85.6 90.1 88.3 91.6 90.0 90.5 89.4 89.8 84.3 51.7 Mauritius 84 24.3 38.7 32.3 32.3 32.9 32.9 29.7 29.7 16.1 16.1 19.4 Mozambique 80 68.9 87.6 91.4 91.4 91.4 92.7 92.7 92.7 92.7 82.0 - Niger 77 6.3 6.5 6.5 7.1 8.0 9.1 10.2 10.1 10.4 8.0 4.4 Nigeria 83 13.6 24.6 37.2 40.3 47.8 49.3 55.4 57.1 78.3 43.6 20.6 Reunion 82 25.9 56.7 50.3 46.9 43.7 40.4 37.0 32.4 25.9 11.9 23.6 Rwanda 78 88.2 96.1 97.1 97.4 98.0 98.0 97.2 95.3 90.9 81.7 55.6 Sao Tome & Principe 81 23.2 46.4 48.9 48.9 48.6 48.7 44.0 42.7 30.8 24.1 20.4 Senegal 85 54.0 54.9 61.4 98.1 67.7 56.7 73.5 72.3 68.0 58.8 39.1 Seychelles 85 60.0 84.2 82.9 69.6 70.0 62.2 46.2 37.2 26.9 7.0 35.9 Sudan 73 16.1 18.5 20.3 21.4 24.1 27.0 28.1 28.8 27.4 26.4 11.9 Tanzania 78 53.4 86.0 90.3 93.1 94.2 94.9 94.9 93.0 90.8 84.0 45.2 N Zambia 84 44.9 29.8 24.0 24.1 24.1 24.1 34.4 42.3 37.6 37.6 17.4 Zimbabwe 82 46.8 50.9 48.5 50.2 51.3 52.6 52.4 50.6 50.7 31.7 25.4 Mean 41.4 52.3 53.8 55.4 54.5 53.9 53.7 52.9 49.7 42.0 26.9 Variance 559.9 757.6 823.6 885.7 797.8 774.4 777.4 767.1 831.5 801.1 250.7 Standard Deviation 23.7 27.5 28.7 29.8 28.2 27.8 27.9 27.7 28.8 28.3 15.8 - not available. Continued o Appendix Table A2.1 (cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Middle East & North Africa Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages Algeria 82 1.4 9.7 9.7 4.6 4.6 4.6 4.6 4.6 4.6 1.8 2.9 Babrain 85 7.3 35.0 30.0 27.0 16.0 8.0 5.0 4.0 3.0 2.0 10.0 Brunei 81 17.0 47.1 43.4 36.7 32.9 28.6 22.7 20.5 14.5 10.4 18.7 Egypt 83 12.6 21.9 20.6 19.9 19.9 13.7 13.7 13.7 13.7 5.9 12.5 Iran 82 8.8 19.6 18.8 14.8 10.9 7.5 4.9 4.0 3.5 3.0 7.2 Iraq 77 10.9 15.5 19.0 20.8 19.2 19.3 18.6 18.3 16.5 13.0 9.4 Jordan 79 3.4 15.7 13.5 8.7 5.2 3.3 2.4 2.0 1.8 1.1 3.3 Kuwait 80 5.0 21.4 29.9 27.4 23.9 24.1 21.8 18.9 13.2 7.6 10.9 Morroco 82 19.0 20.4 20.9 17.7 16.2 14.7 14.1 14.6 14.6 11.2 11.6 Syria 83 9.5 13.5 13.9 10.4 11.3 9.0 7.7 6.9 4.5 2.6 5.6 United Arab 80 4.5 14.6 21.9 24.7 19.4 18.3 13.2 9.5 5.9 3.4 8.8 Emirates Mean 9.0 21.3 22.0 19.3 16.3 13.7 11.7 10.6 8.7 5.6 9.2 Variance 28.1 105.7 81.4 80.9 61.7 60.7 47.9 42.2 29.5 16.5 18.4 Standard Deviation 5.3 10.3 9.0 9.0 7.9 7.8 6.9 6.5 5.4 4.1 4.3 Continued Appendix Table A2.1 (cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Asia and Pacific Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages American Samoa 80 8.6 40.1 46.0 48.8 50.0 50.0 40.4 40.4 23.7 20.2 21.0 Bangladesh 84 7.0 8.7 7.7 8.6 9.4 8.1 8.4 8.7 7.1 4.4 5.4 China 82 77.8 90.3 88.8 88.8 88.5 83.3 70.6 50.9 32,9 16.9 46.4 Fiji 82 22.8 22.8 22.9 22.9 22.9 22.9 17.0 17.0 17.0 17.0 1.3.3 HongKong 85 35.3 83.2 68.8 52.7 52.2 54.2 51.8 41.6 30.8 24.8 37.4 India 81 26.5 29.2 32.1 34.7. 36.4 36.0 36.0 29.8 29.8 14.0 19.8 Indonesia 80 31.1 33.2 36.1 38.5 42.3 45.1 46.2 44.7 40.1 32.0 23.2 Korea, Republic of 85 18.6 49.1 35.9 43.2 55.8 60.0 61.8 55.9 50.7 38.0 29.3 Malaysia 80 33.9 52.6 43.9 40.7 43.0 44.1 42.2 37.6 32.5 26.7 25.3 Maldives 77 52.2 62.1 64.7 64.8 70.7 71.9 73.3 68.1 61.5 52.3 - Nepal 81 51.3 47.6 44.9 43.3 44.1 44.7 44.9 44.7 43.3 39.9 - Pakistan 85 10.9 11.6 13.2 12.3 12.5 14.0 12.3 10.4 9.2 7.2 7.2 Philippines 85 31.4 47.6 53.4 53.4 60.0 60.0 58.9 58.9 49.1 49.1 - Singapore 85 33.8 78.9 66.5 48.8 44.7 39.6 36.3 25.9 18.4 11.9 34.3 Sri Lanka 81 19.0 36.8 36.5 33.9 32.1 28.7 25.6 19.8 13.3 7.6 -- Thailand 82 72.4 81.7 87.2 88.7 90.1 88.9 88.9 79.0 79.0 31.7 50.6 Vanuatu 79 64.9 80.1 79.8 81.3 84.3 83.3 85.7 85.8 86.1 82.4 42.5 Western Samoa 81 6.0 26.1 21.7 20.8 18.2 16.2 13.8 11.5 9.3 4.0 8.3 Mean 33.5 49.0 47.2 45.9 47.6 47.3 45.2 40.6 35.2 26.7 20.2 Variance 462.6 616.5 563.6 524.8 567.2 559.3 571.7 505.1 508.2 388.3 273.9 Standard Deviation 21.5 24.8 23.7 22.9 23.8 23.7 23.9 22.5 22.5 19.7 16.6 4 --not available. Continued Appendix Table A2.1(cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Latin America & 3 Caribbean Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages Argentina 85 24.6 46.7 41.3 37.9 35.7 34.1 30.9 26.1 18.0 9.9 19.9 Bahamas 80 34.5 72.3 75.0 72.3 72.4 68.2 63.5 59.5 49.4 37.8 36.0 Barbados 83 34.7 79.5 83.0 77.5 76.5 72.2 73.1 58.2 45.1 29.2 39.7 Belize 80 42.1 34.4 28.7 26.9 22.9 23.1 20.6 17.3 17.5 15.7 15.0 Bermuda 80 32.7 85.2 83.9 79.8 76.3 79.4 66.5 66.5 66.5 66.5 51.3 Bolivia 85 21.5 26.8 27.7 26.3 24.7 24.3 23.3 21.3 19.0 16.8 14.4 Brazil 80 31.2 39.1 35.9 34.2 34.2 .30.0 30.0 21.4 21.4 10.3 -- Chile 83 11.6 42.0 44.3 43.7 41.5 37.1 34.7 28.1 19.5 13.6 19.9 -. Costa Rica 84 20.8 33.7 33.7 35.4 35.4 27.2 27.2 16.3 16.3 8.7 17.5 Cuba 81 12.9 43.2 50.9 52.4 51.8 48.7 40.7 30.9 18.1 7.8 23.0 Dominica 81 33.2 54.9 52.6 48.5 45.2 45.3 41.4 35.9 33.5 25.0 23.3 Dominican Rep. 81 20.4 33.3 38.3 37.1 32.4 33.3 32.3 28.9 28.1 26.9 19.7 Ecuador 82 14.6 22.7 26.2 24.2 22.1 20.4 18.9 17.4 15.6 13.1 12.0 French Guiana 82 15.0 63.4 67.3 65.1 66.7 63.8 62.1 58.5 54.0 32.9 33.5 Guadeloupe 82 19.1 66.7 69.7 66.9 64.8 61.5 57.1 51.5 46.7 25.3 31.6 Guatemala 81 13.0 16.9 15.9 15.3 14.0 13.6 12.2 11.7 10.0 8.8 8.1 Guyana 80 21.3 36.3 33.2 30.2 27.4 27.0 26.9 25.4 21.8 15.1 15.5 Haiti 82 35.1 53.5 56.7 55.5 58.4 57.5 60.2 57.8 54.2 46.6 33.5 Honduras 84 16.1 23.5 22.5 21.0 19.5 17.1 15.7 14.3 13.2 10.9 - Jamaica 82 38.9 83.3 88.1 88.1 86.9 86.9 82.7 82.7 67.8 67.8 43.2 Martinique 82 17.6 70.0 77.0 74.2 71.9 66.7 60.8 52.1 44.8 25.3 34.6 Mexico 80 26.8 37.3 34.9 32.5 31.3 30.2 29.1 27.5 25.8 24.1 18.2 Neth. Antilles 83 22.5 73.4 70.4 61.3 53.9 44.7 37.7 29.7 22.0 12.7 32.1 Panama 80 16.8 38.3 41.6 40.2 38.1 35.3 30.8 22.6 15.0 7.2 18.2 - not available. Continued Appendix Table A2.1 (cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Latin America & Caribbean (cont.) Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages Paraguay 82 20.0 27.5 27.4 26.4 24.8 23.2 20.7 18.2 15.9 12.7 13.6 Peru 81 -18.3 29.2 31.2 31.0 29.4 27.9 26.7 25.8 23.4 22.9 15.7 Puerto Rico 85 9.1 31.6 43.5 43.5 46.9 46.9 34.1 34.1 14.1 14.1 - St Christopher & Nevis 80 43.2 76.5 71.9 66.7 61.9 54.1 49.9 44.5 39.1 29.0 31.3 Trinidad & Tobago 80 18.0 45.5 42.8 40.5 38.7 36.7 34.1 30.9 28.4 18.2 20.8 Uruguay 84 26.1 65.6 64.9 65.2 61.4 61.3 54.3 43.6 31.1 17.3 32.0 ! Venezuela 81 18.5 35.8 39.5 39.9 39.0 35.7 28.9 23.4 16.6 10.7 17.9 Mean 23.6 48.0 49.0 47.1 45.4 43.0 39.6 34.9 29.4 22.0 22.3 Variance 84.1 389.5 408.6 376.6 377.4 368.1 326.3 308.2 257.7 227.6 151.8 N Standard Deviation 9.2 19.7 20.2 19.4 19.4 19.2 18.1 17.6 16.1 15.1 12.3 - not available. Continued N4 0% Appendix Table A2.1 (cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Industrial Market Economies Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages Ai Australia 85 56.1 73.5 57.7 57.7 61.4 61.4 50.2 50.2 27.1 11.1 45.7 Austria 81 46.7 69.8 62.2 59.1 59.9 62.2 58.2 50.8 25.8 5.5 32.8 1 Canada 85 52.1 74.9 70.7 70.7 70.0 70.0 61.3 61.3 33.8 33.8 - f Denmark 85 61.0 82.1 86.9 88.8 86.4 84.7 80.3 71.1 57.3 25.6 48.6 Finland 85 37.2 71.4 83.2 85.5 89.7 90.8 89.2 83.3 66.5 38.9 49.3 France 84 13.7 66.0 72.8 67.8 67.3 64.9 61.0 54.1 41.4 18.0 34.8 Germany, Fed. Rep. 84 40.9 71.3 65.6 59.9 59.8 60.3 56.4 49.7 40.2 11.8 35.3 Ireland 84 33.7 75.2 52.2 30.9 24.2 25.8 25.3 25.8 21.1 15.3 21.9 Italy 85 25.9 59.6 58.6 56.9 51.6 46.0 40.4 32.7 20.8 10.2 28.2 Japan 85 16.6 71.9 54.1 50.6 60.0 67.9 68.1 61.0 51.0 38.5 38.6 Luxembourg 81 44.1 70.5 58.0 46.1 41.8 36.9 30.3 25.6 20.1 12.4 17.5 t4 Netherlands 86 25.5 71.7 62.9 48.0 47.8 48.5 42.4 31.1 20.7 8.9 - New Zealand 81 49.5 63.7 42.3 39.9 49.2 54.9 53.0 44.3 31.6 12.1 28.8 Norway 85 45.8 67.7 72.5 73.8 77.1 80.3 79.2 72.6 60.0 46.2 -- Spain 85 31.8 55.1 54.3 41.0 33.6 30.8 27.5 24.4 23.6 16.1 - N Sweden 85 48.3 81.3 87.3 88.4 89.2 92.1 90.5 85.6 74.4 46.4 - Switzerland 80 51.1 76.2 58.6 48.9 50.3 52.3 50.8 46.9 41.1 24.4 34.4 United Kingdom 81 45.0 69.4 55.5 53.4 62.3 68.4 68.1 63.1 51.9 22.4 35.8 United States 85 41.4 70.9 70.4 69.7 70.9 72.6 67.5 60.5 50.2 33.0 41.8 Mean 40.3 70.6 64.5 59.8 60.7 61.6 57.9 52.3 39.9 22.7 26.0 Variance 158.6 40.6 141.2 261.3 301.9 333.9 355.3 332.3 273.4 163.2 300.5 Standard Deviation 12.6 6.4 11.9 16.2 17.4 18.3 18.8 18.2 16.5 12.8 17.3 -- not available. Continued Appendix Table A2.1 (cont.) Age-Specific Female Labor Participation Rates by Region and by Country (early 1980s) Southern Europe Year 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 all ages Greece 84 22.9 45.9 48.9 49.3 47.5 44.6 44.6 36.1 28.1 22.7 26.6 Israel 83 31.4 31.4 58.6 58.6 56.2 56.2 47.2 47.2 29.9 29.9 -- Portugal 85 46.1 66.5 72.8 73.3 67.9 59.7 52.6 46.7 37.2 25.8 37.0 Turkey 80 51.8 49.5 44.7 44.6 47.2 49.4 50.5 49.5 47.0 43.7 31.4 Yugoslavia 81 10.1 58.8 68.4 67.4 61.2 54.2 47.9 37.7 27.8 22.1 32.9 Mean 32.5 50.4 58.7 58.6 56.0 52.8 48.6 43.4 34.0 28.8 25.6 Variance 230.5 142.4 - 117.0 115.2 63.7 28.0 7.6 29.7 53.9 62.9 174.6 Standard Deviation 15.2 11.9 10.8 10.7 8.0 5.3 2.8 5.4 7.3 7.9 13.2 - not available. Source: ILO Statistical Year Book (various issues). B. 0 N9 X0 3 The Industrial and Occupational Distribution of Female Employment 1. Introduction The previous chapter examined the size of the female labor force and found that women have improved their employment position both in absolute terms and relative to men. This chapter goes beyond the aggregate female participation rate and examines the employment distribution of women by status (self- employment, dependent employment and family work), industry and occupation. The first objective is to identify the important sectors for women's employment and changes -- if any -- in the importance of sectors over time. The second objective is to establish whether women are now employed in a more equitable way relative to men with respect to the jobs they undertake. 2. Methodology Of the three key aspects of labor supply -- participation, wages, and type of work - the last is the most difficult to measure. Various approaches to quantifying occupational attainment have been proposed in order to make the issue susceptible to empirical investigation. However many of these measures range from arbitrary to, at best, ordinal rankings.' The difficulty increases when one asks the tempting question "... and in what industry?" At this point the demand for labor creeps into the analysis. Are women found in some occupations because of their own lifetime optimization decisions? Or, are women absent from some industries because of overt exclusion or subtle discriminatory practices? If it is the former, then employment differentials are efficient in the sense that they reflect the choice of individual agents. If it is the See, for example, Goldthorpe and Hall (1974). 72 Women's Employment and Pay in Latin America latter, then one can argue that this is a prime example of a rather rare combination in economics: intervention and higher efficiency.2 The original difficulty has increased further: to what extent are employment differentials, whether occupational or industrial, justified? There is no easy answer to this.3 Below a methodology is employed which helps, first, to establish the dissimilarity in employment between women and men and, second, to estimate the percentage of the labor force who would, in theory, have to change jobs so that the employment distributions of women and men eventually look alike. This is a promising exercise as welfare losses of any type are conventionally (and more meaningfully) expressed in percentage terms of the target variable to which they refer -- in this case, the size of the total labor force. In this respect the analysis is carried to the limit: how far away from sex equality are labor markets in Latin America? In the absence of relevant country institutional information and specific knowledge of factors governing household formation and employment decisions, this is a legitimate question. Though equality may still sound hypothetical in some developing countries, experience elsewhere has shown the obvious: in many cases there are no real differences in what women and men can do in the labor market which justify the acute sex differentials still observed in many countries.4 Hence the present analysis and results stand if adjusted by a factor, the choice and magnitude of which are left to be decided by the reader. 2 In some countries women were (and still are) excluded from some types of employment by decree. Moreover, their wages were (are) set in an arbitrary way relative to male wages. The idea behind this preferential treatment of men is that men are usually expected to be the main breadwinners and supporters of the family while women's employment and pay is consideredto be "secondary" or "supplementary" within a family context. As a result, the introduction of relevant legislation can unleash competitive forces and enhance the working of allocative mechanisms in the labor market and the economy. 3 The difference between measurable variables, such as wages, can be established and analyzed more easily. In fiact, this is done in other sections of this volume. However, it is harder to measure the difference in the type of work between, say, a doctor and an economist, or an economist in the private sector and an economist in the public sector. 4 In some Scandinavian countries women's wages are almost equal to men's while sex differences between their respective employment levels and employment distributions have been drastically reduced over time. See the appropriate tables in the recent editions of the Statistical Yearbook, ILO. Distribution of Female Employment 73 Given the large number of countries to be studied, it would be tedious to proceed with a long list of tabulations5 to show the obvious, that women are overrepresented in a couple of sectors and that the importance of these sectors has increased, as expected, during the process of development.6 Instead, if one examined women relative to men, one could indirectly obtain this kind of information. What this study seeks to show is whether women, as a factor of production, exhibit different characteristics and are subject to different treatment than men. Consequently, this section examines, first, whether women, relative to men, are more likely to be found in the labor force as employers, self- employed and own-account workers, wage/salaried employees, or as family workers. Second, it examines which industrial sectors have higher proportions of women employees compared with the proportions of male employees found in the same sectors. After the scene is set, a dissirmilarity index is presented in order to capture in a quantitative way the difference in the employment distributions of women and men in the labor force.' The properties of the index are evaluated and applied to broad occupational and industrial data. Before detailed estimates are presented, it would be useful to get a bird's eye view of the employment distribution of women and men in the region. For brevity we concentrate on industries. Looking at broad industrial employment patterns could also provide a first feeling about the occupational and employment status of the labor force. For example, in terms of employment status "agriculture" means to a large extent family work for women while in terms of occupation "services' in many Latin American countries implies domestic work for women. Figures 3.1 to 3.3 illustrate the percentage of the labor force by country and sex during the early-1980s in the three main industrial groupings, namely agriculture, production industries, and services.8 Countries have been 5 There are 7 occupations, 8 industries and 15 countries which can be examined by three separate employment groups (self-employed, employees and family workers) at two points in time and separately for women and men - in all 10,080 information cells. 6 Women are typically found in non-manual occupations and service industries. It is precisely these two groups whose share in the labor force increases with time. This view was expressed early on by Clark (1940) and has subsequently been confirmed in a number of studies (Kuznets, 1955, 1957, 1971; Oberoi, 1978; Ahluwalia, Carter and Chenery, 1979). 7 The labor force includes the employed and unemployed. 8 Production industries in the present context are defined as mining, manufacturing and utilities. For the exact definition and dates see the Appendix to this chapter. 74 Women's Employment and Pay in Latin America ranked in descending order of the percentage of the male labor force who are engaged in agricultural activities. In this way one can observe patterns of employment from the least to the most 'modern" labor markets. In the top panel of Figure 3.1 two of the countries shown, Colombia and Bolivia, have more than 50 percent of the male labor force in agriculture. The corresponding percentage in five other countries, Peru, Mexico, Ecuador, Jamaica and Brazil, is between 37 and 43 percent. In two of the nine countries shown, Argentina and 'Venezuela, the percentage of the male labor force in agriculture is approximately 20 percent. With respect to women, the importance of agriculture seems to be declining when the male ranking is maintained -- though Colombia and Brazil appear to defy this observation (lower panel in Figure 3.1). However, the most important finding is that the share of agricultural employment in the female labor force is only a fraction of the corresponding share for men. For example, the female share is, first, approximately half that of the male share in Bolivia (27.2 and 54.2 percent respectively), Peru (24.6 percent to 43.4 percent) and Mexico (19.2 to 42.6 percent) and, second, approximately one-third to one-fifth in the other countries - and as low as one-tenth in Venezuela (1.7 percent to 17.5 percent). It is possible that this comparison is biased by some underreporting of women engaged in agricultural tasks given that the distinction between home production and work is not clear in agrarian environments. However, the census data used in the analysis (see Appendix to this chapter) include family workers and any statistical bias should not be sufficient to change the main conclusion that agriculture is not perhaps an important employer of women after some point in development.9 We explore this proposition in more detail in the next section after we examine the regional pattem of female and male employment in production industries and services. 9 Of course, one is tempted to ask what the wives and daughters of the 50 percent of male workers who are in agriculture in, for example, Colombia or Bolivia are doing. Our point here is that in terms of contribution to broadly defned market production and women's independent labor power agriculture is not as important as the other two main industrial sectors. This may be an important observation in a dynamic developmental context because historical trends suggest that women move further away from subsistence activities during periods of economic growth. Hence, though agriculture may still be an important sector in the short run, women are expected to be utilized in industry and services in the future even more than they are today. If polices which correctly anticipate the future are considered appropriate, then due emphasis should be given to the modern sectors in the economy. Distribution of Female Employment 75 Figure 3.1 Percentage of the Labor Force in Agriculture in Nine Latin American and Caribbean Countries 1980s 58 MALE 54 43 43 41 38 20 I.. Colombia Bolivia Peru Mexico Ecuador Jamaica Brazil Argent. Venezuela FEMALE 27 25 19 15 13 * ~~~~~~~~~~~~~~~02 Colombia Bolivia Peru Mexico Ecuador Jamaica Brazil Argent. Venezuela Source: Constructed from Appendix Tables A3.1 and A3.2. 76 Women 's Employment and Pay in Latin America The top panel in Figure 3.2 shows the share of industrial employment in the male labor force. In four countries about one-quarter of the male labor force is engaged in industry (Colombia, Bolivia, Peru and Ecuador) while it is about one-third in three other countries (Mexico, Jamaica and Brazil). In Argentina and Venezuela the share rises to 40-45 percent. The share of male industrial employment is increasing from left to right -- countries are presented in the same order as in Figure 3.1, from the most agrarian to the least agrarian economies. No clear increasing/decreasing pattem emerges in the cases of women (lower panel in Figure 3.2). The share of industrial employment among women workers is rather invariant across countries. If one argued that one in five female workers is occupied in industry this would be an accurate generalization. In addition the difference between the male and female shares is not as dramatic as in the case of agriculture. Similarly, Figure 3.3 shows the share of employment in services. Figure 3.4 takes another look at the change in sectoral employment of women and men. This time changes are expressed in terms of the original size of the respective labor force of women and men. (That is, changes refer to absolute numbers - not shares in the total labor force.) This is why the male labor force in agriculture shows an increase of 2.4 percent although the share of male employment in agriculture declined from 50.6 percent in the 1950s to 38.3 percent in the 1980s. What is interesting is that the size of the female labor force in agriculture actually decreased during this 30 year period. Two more observations are equally interesting. First, the increase in the number of women engaged in industry was greater than the increase for men (82 percent compared to 56 percent). And second, the increase in the labor force engaged in the services sector was practically the same for both sexes (90.2 percent for women and 89.5 percent for men). These findings are summarized in Figure 3.5 when a comparison with the situation in the 1950s is also attempted. The pie-charts show that in the 1950s more than half of the male labor force was in agriculture while more than half of the female labor force was in services. In contrast, services used to be the least favorable sector for men and industry for women. By the 1980s the representation of men in the industrial and services sectors had increased -- in fact, industry was not far behind agriculture (34 percent versus 38 percent). In terms of female work, the share of agriculture was reduced by half (from 26.1 percent in the 1950s down to 13.8 percent in the 1980s) while the share of industrial employment increased by about 10 percent (from 21.0 percent to 23.7 percent). The share of women workers in services increased almost 20 percent (from 52.9 percent to 62.5 percent) and, given the initial size of the services sector, this change was sizeable. Distribution of Female Employment 77 Figure 3.2 Percentage of the Labor Force in Industry in Nine Latin American and Caribbean Countries 1980s 45 MALE 41 35 35 30 28 ~~~~~27 24 ~~~25 Coloimbia Bolivia Peru Mexico Ecuador Janaica Brazil Argent. Venezuela FEMALE 30 22 2 19 19~~~la1 Coloxmbla Bolivia Peru Mexico Ecuador Jamaica Brazil ArgenL Venezuela Source: Constructed from Appendix Tables A3.1 and A3.2. 78 Women's Employment and Pay in Latin America Figure 3.3 Percentage of the Labor Force in Services in Nine Latin American and Caribbean Countries 1980s MALE 41 35 82 32 32 28 Colombia Bolivia Peru Mexico Ecuador Jamaica Brazil ArgenL Venezuela FEMALE 80 80 73 67 88 s 54 51 Colombia Bolivia Peru Mexico Ecuador Jamaica Brazil Argent. Venezuela Source: Constructed from AppendixTables A3.1 and A3.2. Distribution of Female Employment 79 Figure 3.4 Distribution of the Labor Force in Nine Latin American and Caribbean Countries 1980s MALE 1950s FEMALE 1950s Services Agriculture 26.1% 52.9% Agricultu e 50.8% Wll wIn2d9ustry I2l.7 29.5% MALE 1980s FEMALE 1980s Services Agriculture 27.8% ~~~~13.Z Agriculture 38.3%..... Industry Services 23.7% 62.5% 33.97% Note: Industry = Mining, Manufacturing, Construction, Transport, Utilities. Services = Commerce, Trade, Other Services. Source: Constructed from Appendix Tables A3.1 and A3.2. 80 Women 's Employment and Pay in Latin America Figure 3.5 Percentage of Change in Labor Force in Latin American and Caribbean Countries by Broad Industrial Sector 1950-1980 100.0 - 90.2 B9.5 50.0 0.0 ~~~2.4 I i 0.0 -14.6 a X b Agriculture Production Services Female M Male a. Mining, Manufacturing, Construction, Transport, Utilities. b. Commerce, Trade, Other Services. Source: Constructed from Appendix Tables A3 . 1. and A3 .2. Distribution of Female Employment 81 This introductory examination of the sectoral composition of the labor force runs counter to traditional thinking. One popular belief is that females are predominantly family workers and that more women than men are found in the agricultural sector. Neither of these two expectations appears to hold true in Latin America. In the next section we undertake a more rigorous analysis of these issues by distinguishing between industrial, occupational and employment status characteristics of the female iabor force and male labor force. A sector is defined as "female overrepresented' if the ratio of women workers in the sector relative to the female labor force exceeds the corresponding ratio for men."0 For the moment the analysis is restricted to eight countries for reasons of data availability." 3. The Pattern of Female Employment by Status and Industrial Sector Table 3.1 shows whether the percentage of the female labor force in the three main groupings of employment status (self-employment, dependent employment, family work) exceeds the corresponding ones for men. An easy generalization is that more women than men were not found among the group of employers, self-employed and own-account workers either in the early or the late period in any country. The picture is less clear for family workers. In only three countries were women initially overrepresented among family workers (Bolivia, Brazil and Peru) although more countries joined this group (Argentina and Mexico) in the second period. Thus, the relative representation of women among family workers has increased but this may be a short lived phenomenon. One explanation for this increase can be the adverse economic conditions which have prevailed in many countries in the region during the last decade: in the absence of rising opportunities in other employment sectors women may have found refuge in family work. Surprisingly, the highest representation of women workers is found in dependent employment. The exception is Bolivia where family work prevailed but only in the earlier period. In all countries, women are now overrepresented in wage/salaried employment. Consequently, the conclusion noted in the previous chapter, that wage/salaried employment has been, and still is, an important employment sector for women during the process of development, is maintained. 10 That is, a sector is over-represented if (F,/F)-(M,/M) > 0, where F and M refer to the number of female and male workers respectively and i is the sector. it The eight countries are Argentina, Bolivia, Brazil, Colombia, Ecuador, Mexico, Peru, and Venezuela. 82 Women's Emiployment and Pay in Latin Amenca Table 3.1 Female Overrepresentation by Employment Status (1950s and 1980s) Country Self-employed Employees Family Workers Argentina 0 xx x2 Bolivia 0 x2 xx Brazil 0 xx xx Colombia 0 xx 0 Ecuador 0 xx 0 Mexico 0 xx x2 Peru 0 xx xx Venezuela 0 xx 0 Notes: 0 No female overrepresentation in either period xx Overrepresentation in both the 1950s and the 1980s x2 Overrepresentation in the 1980s only Source: Calculated from ILO, 1990, Table 2. Table 3.2 gives a summary picture of female overrepresentation by industry. The construction and transport industries are not included because female over- representation did not occur in any of the countries in either of the two time periods. Again, a surprising finding is that agriculture was important in only one country (Bolivia) and only in the earlier period. This holds true for mining and utilities respectively for Colombia and Brazil. The service industry, however, has been overrepresented in all countries in both periods. Women are now also overrepresented in commerce in all countries but one (Argentina). Women's position in the manufacturing sector is mixed as there has been over- representation in four of the eight countries in both periods but one country has dropped out of the group (Venezuela) and another has joined (Bolivia). In short, women's 'favorite" place appears to be in the service and commercial sectors while the case for manufacturing is still to be proven. Women are less likely to be employed, in relative terms, in agriculture, mining and utilities. The transport and construction industries appear to have been, and still are, unimportant. Table 3.3 combines the information presented in the previous two tables and gives a more detailed view of industrial employment by status. The prominence of the service sector noticed earlier can now be examined further: in all countries, and for all three types of employment status, women are over- Distnbuton of Female Employment 83 Table 3.2 Female Overrepresentation by Industry (1950s and 1980s) Service Commerce Manufacturing Utilities Agriculture Mining Argentina xx 0 0 0 0 0 Bolivia xx x2 x2 0 xl 0 Brazil xx x2 0 xl 0 0 Colombia xx xx xx 0 0 xl Ecuador xx x2 xx 0 0 0 Mexico xx xx 0 0 0 0 Peru xx xx x 0 0 0 Venezuela xx x2 xl 0 0 0 0 No female overrepresentation in either period. xx Overrepresentation in both the 1950s and the 1980s. xl(x2) Overrepresentation in the 1950s (1980s) only. Note: No female overrepresentation in the construction and transport industries in either period. Source: Calculated from ILO, 1990, Table 2. represented in services. However, in four cases the female self-employed and family workers seem to have lost their lead over men in the service sector but, perhaps, this was due to women moving sideways and enforcing the ranks of employees. The commerce industry is almost as important for women as services (except for the self-employed) and its importance has increased over time (seven new cases were added in the later period). The third important sector is manufacturing, although women are overrepresented there primarily as self-employed or as family workers. The data, albeit aggregated, have somewhat shaken the conventional views on women's employment distribution. It is true that large numbers of women are engaged in family work (most often unpaid) and in agriculture. It is also true that the groups of self-employed and family workers are sizeable in absolute terms. However, when taking into account the structure of the male labor force, the evidence for manufacturing changes the view (at least in the countries studied) that the industrial worker is a man, and the evidence by employment status suggests that women's involvement in the open labor market as employees is, relative to men, more important than alternative forms of economic activity. 84 Women's Employment and Pay in Latin America Table 3.3 Female Overrepresented Industrial Sectors in the 1950s and 1980s by Employment Status Industrial sector Self-employed Employees Family workers Agriculture Ecuador(l) Bolivia Brazil Peru Mining Bolivia(l) Colombia Colombia Mexico(2) Mexico(2) Mexico(2) Manufacturing Argentina Argentina Bolivia Bolivia Brazil(l) Brazil Colombia Colombia Colombia Ecuador(2) Ecuador Ecuador Mexico Mexico(2) Mexico Peru Peru Venezuela Venezuela Utilities Bolivia(l) Brazil(l) Brazil(1) Brazil(l) Commerce Argentina Argentina Bolivia(2) Bolivia(2) Bolivia Brazil(2) Colombia Colombia Ecuador(2) Ecuador Ecuador(2) Mexico Mexico Mexico Peru Peru Peru Venezuela(2) Venezuela(2) Services Argentina Argentina Bolivia(l) Bolivia Bolivia(l) Brazil Brazil Brazil Colombia Colombia Colombia Ecuador(l) Ecuador Ecuador Mexico Mexico Mexico Peru(l) Peru Peru Venezuela Venezuela Venezuela a. Mining, manufacturing, construction industries combined (1)/(2) Overrepresentation in the first (second) period only Note: No female overrepresentation in the construction or transport industries in either period. Source: Calculated from ILO), 1990, Table 2. Distribution of Female Employment 85 With respect to the former finding, women are perhaps not often seen as industrial workers because women are rarely engaged in heavy industries, especially in manual tasks requiring physical strength, and they are outnumbered by men in absolute terms. However, "industry' in less developed countries need not be synonymous with the conventional use of the term in advanced countries: industrial activities in developing countries evolve predominantly around the production of food, drink, tobacco, clothing, and similar goods. The production of these goods is similar to tasks undertaken by women at home and, in fact, the majority of workers in these industries tend to be women. In addition, the presence of a sizeable cottage and handicraft industry in the earlier stages of development offers another explanation for the prominence of women in manufacturing. What is more difficult to explain is the second finding, the prevalence of employee status among women compared to men. A promising direction of research would be to examine the importance of the public sector as an employer of female labor compared to the private sector. Given this evidence, one is tempted to argue that the emphasis generally given to the informal sector as an employment generator and income guarantor for women is still to be proven.'2 However, one may suspect that the importance of wage employment for women, as evidenced in the previous analysis, is possibly the result of the way women's contribution to production, broadly defined, is statistically treated: the data are based on official statistics, and employment in the formal sector is more easily detected than other types of work. Hence, employees are disproportionately represented in the final estimates for women since about half of all women are reported to be economically inactive -- and this may not be a reliable estimate of the overall economic activity broadly defined."3 This does not apply to men as practically all prime age men are counted as workers. Consequently, male employees appear to be "2 Of course, this discussion does not deny that during periods of economic crisis women's involvement in the informal sector may increase at least in the short run. Therefore, measures which facilitate women's adjustment efforts during periods of transition can be wel justified. However, it should be clear that such measures may work against the longer term tendencies of the development process, if they persist after the economic crisis is over. 13 In many developing countries women are not reported to be economically active in the sense that men are: self-employment and family work by women may be easily classified as "housekeeping" (Standing, 1981; Kozel and Alderman, 1988). 86 Women's Employment and Pay in Latin America a less important subgroup of the total male labor force."4 This discussion can only advance if more detailed data become available but, given the findings of this section, it would be inadvisable to pay less attention to the importance of wage employment for women (compared with work in the informal sector) in the context of a dynamic development strategy.15 4. Measuring Employment Dissimilarities Between Women and Men The most commonly used summary statistic of dissimilarity in the employment distributions of any two groups of workers is the Duncan index.'6 In the case of sex differentials the index, D, takes the form 1N D = S -m, I where i (= 1,2,..., N) is the total number of sectors of interest (for example, industries or occupations), 1i and mn are the sectoral employment ratios of women and men to their respective labor force, and the summation refers to the absolute differences between these two ratios within each sector."7 The value of the index varies between 0, when women and men have identical employment distributions across sectors, and 1 when there is complete dissimilarity (no women and men work in the same sector). 14 While men's movements are usually confined to work and unemployment women's typical flows are between work and inactivity (Bowers, 1975; Jelin, 1982; Lundberg, 1985). 15 Even if salaried employment were not as important as work in the informal sectors, strategies which focus on the formal sector are bound to be more efficient in the longer run as this is the direction in which any developing economy is moving. In addition, public policies are usually considered to be responsible for the shift in the gains arising from technological changes toward male workers in the formal sector rather than (female) workers in the non-market sector (Blumberg, 1984; Ngwira, 1987). 16 Duncan and Duncan (1955). 17 Absolute differences are used because, by definition, the excess female and male labor in some sectors is exactly equal to their respective shortages in the other sectors. This also explains why the resulting sum is divided by two. Distribution of Female Employment 87 The popularity of the index is easily explained. On intuitive grounds, the sum of employment differences within sectors should give an idea of the extent of dissimilarity in the total labor force. On theoretical grounds, the index satisfies a number of criteria typically desired in similar exercises. Is And on practical grounds, a change in the value of the index can be decomposed into two parts: that due to changes in the sex ratio within sectors and that due to changes in the size of the sectors. 9 5. Interpreting the Index The micro-foundations of the Duncan index are obvious. The index simply reflects relative employment both within and across occupations. If women's employment doubled but remained distributed as before and this were the only change, then there would be no change in the value of the index. Equally, if the rate of employment growth were the same across sectors, then ceteris paribus the index would return the same value. However, the weakness of the index is that, while it identifies that a symptom is present, it makes no reference to its importance: the size of the labor force does not enter into these calculations, nor does the overall ratio of women to men. Despite the absence of any obvious macro-foundations, most authors have typically interpreted the value of the Duncan index as "the proportion of either women or men who would have to be transferred from one sector to another in order to obtain equal proportions across sectors" (emphasis added). This is clearly not the case as the index has a single value although there are usually fewer women than men in the labor force. The index does not even refer to the percentage of the total labor force which would have to change sectors to reach equality in the female and male employment distributions across sectors. In fact, it has been shown that the index is simply the "standardized" ratio of required reallocations to potential reallocations, had the employment distributions of 18 Such criteria include unidimensionality, boundedness, and an increase in the value of the index when segregation increases. See Hall and Tideman (1967). 19 See Fuchs (1974); Blau and Hendricks (1979); or Humphries (1988). 2D Quoted from Brown et al. (1980b, p. 515). In fact identical phrases are found in Blau and Hendricks (1979, p. 199); Joseph (1983, p. 147); Beller (1984, p. 12); and Beller and Han (1984, p. 91). 88 Women's Employment and Pay in Latin America women and men been as dissimilar "as possible."2' 'As possible' means that, even if the intention were to allocate women to certain sectors and men to the rest, it is almost certain that there would be some workers of either sex who cannot be accommodated in the sector of their respective sex and would have to work in the sector(s) originally reserved for the other sex. This could happen if the female labor force exceeds total employment in the sectors originally assigned to women. In conclusion, the index fails to say anything about how far from equality the employment distributions are, unless the overall sex mix and size of the labor force are explicitly taken into account in tandem with the size of the sectors.' 6. Results The Duncan index was estimated separately for seven occupational and eight industrial sectors in the 15 countries studied in this volume.' In general terms, the occupational data suggested a higher value of the index (an unweighted average of about 0.49 in both periods compared with about 0.40 for industrial data).' The s]light dominance of the occupational dissimilarity over the industrial one at aggregate levels is maintained for the group of employees, though the reverse is true in the case of self-employed workers. Family workers appear to be a mixed case and the one kind of dissimilarity dominates over the other in as many cases as the latter over the former. In other respects the results derived from the industrial data and occupational data were quite comparable. Part of the explanation for the similarity of the results between occupational data and industrial data is the fact that at the aggregate level of data examined in this 21 The index is standardized for (insensitive to) the size and sex mix of the total labor force. 22 See Tzannatos (1990b). 23 The international conventions for occupational and industrial classifications of employment changed in 1968. In order to enable comparisons between the data for the 1950s and the data for the 1980s the following groupings for occupations and industries were constructed. The seven occupational sectors refer to professional, administrative, clerical, sales, service, farming, and manual tasks. The eight industrial sectors are agriculture, mining, manufacturing, construction, utilities, commerce, transport and services. For more information see the Appendix to this chapter. 24 At the country level, only Colombia appears to have higher industrial dissimilarity in both periods and Peru in the first period only. Bolivia has a higher value for industrial dissimilarity in the later period but there are no data for occupations in the earlier period. Distribution of Female Employment 89 report the groupings were quite similar. For example occupation 'farmer' versus industry 'agriculture," or "professional-administrative-commerce" versus 'services," or "laborer" versus "manufacturing," or 'sales" versus "commerce" and so on. Given the broad uniformity of the two sets of results, we concentrate below on occupational data only. A summary picture of industrial dissimilarity is presented later in this chapter when estimates for Latin America are compared with estimates from other countries. The value of the Duncan index for occupational dissimilarity is presented in Table 3.4.Y Estimates are presented for the total labor force and are then broken down by employment status (self-employed/own-account workers and employers, employees and family workers). The table shows that dissimilarity is generally higher among employees than among self-employed and family workers. Family workers were much less differentiated in the earlier period than the self-employed with the exception of Costa Rica and Mexico, though in the latter country the difference between the two groups was only one percentage point. In the 1980s, the picture became somewhat mixed as dissimilarity among family workers, albeit still low, increased in three countries. In contrast, the groups of self-employed workers and employees have had a more uniform experience that suggests a decrease in dissimilarity over time. This finding may imply that during the process of development there are different forces at work between those who work for pay and those who offer their services within a family context.2' It is worth noting that the changes in dissimilarity among employees have been considerably more uniform than those among the self-employed. On the one hand, the rates of decline in occupational dissimilarity among the employees varied between 0.2 and 1.7 percent per year. On the other hand, the rate of decline for the self-employed varied between 0.3 and 2.9 percent per year while in Argentina the rate of decline was as high as 4.9 percent and in Chile, Honduras and Peru (practically) there has been no change. The discrepancy in the rates of decline of dissimilarity among workers of different employment status can be attributed to the fact that salaried employment is subject to more consistent forces than employment in the informal market (as proxied here by the self-employed). Self-employment may be less standardized across countries 25 The value of the Duncan index will in general be lower the more aggregated the data (that is, the fewer the sectors examined). This point is pursued in England (1981). 26 In fact, it has been found that in some developing economies there exists high substitutability between self-employment and salaried employment, but not between these two and family work (Hill, 1983). 90 Women's Employment and Pay in Latin Ameria Table 3.4 Occupational Dissimilarity (Duncan index) Year Index Year Index Annual change (%) All workers Argentina 1960 0.4270 1970 0.4486 0.5 Bolivia 1976 0.3670 Brazil 1970 0.4913 Chile 1952 0.4283 1982 0.5259 0.7 Colombia 1951 0.5543 1964 0.5177 -0.5 Costa Rica 1963 0.5894 1984 0.4981 -0.8 Ecuador 1962 0.4856 1982 0.4648 -0.2 Guatemala 1950 0.6220 1981 0.5732 -0.3 Honduras 1961 0.7515 1974 0.6584 -1.0 Jamaica 1960 0.4668 1982 0.5411 0.7 Mexico 1960 0.3891 1980 0.3182 -1.0 Panama 1950 0.5348 1980 0.5916 0.3 Peru 1961 0.3087 1981 0.3289 0.3 Uruguay 1963 0.4192 1985 0.4333 0.2 Venezuela 1961 0.5510 1981 0.4708 -0.8 Self-employed Argentina 1960 0.3584 1970 0.2174 -4.9 Bolivia 1976 0.4610 Brazil 1970. - Chile 1952 0.3416 1982 0.3452 0.0 Colombia 1951 0.5009 1964 0.4572 -0.7 Costa Rica 1963 0.6155 1984 0.4792 -1.2 Ecuador 1962 0.4997 1982 0.3230 -2.2 Guatemala 1950 0.7341 1981 0.6783 -0.3 Honduras 1961 0.7943 1974 0.7892 0.0 Jamaica 1960 0.5246 1982 - Mexico 1970 0.3950 1980 0.2948 -2.9 Panama 1950 0.5930 1980 - Peru 1961 0.3314 1981 0.3397 0.1 Uruguay 1963 - -1985 0.3168 Venezuela 1961 0.5057 1981 0.3953 -1.2 - Continued Distribution of Female Employment 91 Table 3.4 (Cont.) Occupational Dissimilarity (Duncan index) Year Index Year Index Annual change (M) Employees Argentina 1960 0.4675 1970 0.4880 0.4 Bolivia 1976 0.5882 Brazil 1970 - Chile 1952 0.5407 1982 0.4877 -0.3 Colombia 1951 0.6497 1964 0.6296 -0.2 Costa Rica 1963 0.6079 1984 0.4718 -1.2 Ecuador 1962 0.6164 1982 0.5251 -0.8 Guatemala 1950 0.6450 1981 0.6084 -0.2 Honduras 1961 0.7404 1974 0.6005 -1.6 Jamaica 1960 0.5723 1982 - Mexico 1970 0.4457 1980 0.3742 -1.7 Panama 1950 0.5590 1980 - Peru 1961 0.4968 1981 0.3991 -1.1 Uruguay 1963 - 1985 0.4640 Venezuela 1961 0.5946 1981 0.4467 -1.4 Family workers Argentina 1960 0.1877 1970 0.3534 6.5 Bolivia 1976 0.0969 Brazil 1970 - Chile 1982 0.5275 Colombia 1951 0.4898 1964 0.4766 -0.2 Costa Rica 1963 0.7370 1984 0.7342 0.0 Ecuador 1962 0.3797 1982 0.1320 -5.1 Guatemala 1950 0.4023 1981 0.4880 0.6 Honduras 1961 0.7262 1974 0.6924 -0.4 Jamaica 1960 0.1838 1982 - Mexico 1970 0.4093 1980 0.3155 -2.6 Panama 1950 0.1429 1980 - Peru 1961 0.0796 1981 0.0574 -1.6 Uruguay 1963 - 1985 0.2683 Venezuela 1961 0.2827 1981 0.5952 3.8 - not available Source: Calculated from ILO, 1990, Table 2. 92 Women's Employment and Pay in Latin America as it is affected more by local conditions and less by employment legislation, trade unions and competitive practices. The higher variance that characterizes the estimates for family workers, whose employment is even more informal than that of self-employed workers, is consistent with this explanation. With respect to all workers, dissimilarity decreased in seven countries while it increased in six countries. The average for the countries which experienced a decline comes to around -0.65 compared to an average of 0.45 for the countries which experienced an increase. One can, therefore, summarize the trend in dissimilarity as declining. The decline has come primarily from women working for pay, that is, on the one hand, employees and, on the other hand, the self- employed/own-account workers and employers. It was also found that the variance of the over time changes of the index within each subgroup varies indirectly with the level of the index of the group: the changes have been more uniform for employees who have the highest value of the index, while the changes have varied widely in the case of family workers who have the lowest value of the index. Consequently, it appears that there are more consistent forces operating in the formal sector during development than in the informal sector. 7. The Decomposition of Changes in Dissimilarity The Duncan index evaluates sex employment differentials within sectors and calculates their sum across these sectors. Thus the index is affected by changes in these two different aspects of employment, namely the prevailing sex ratios within occupations and the occupational structure of the total labor force. Since the index has generally declined during the period under examination, it would be interesting to examine the extent to which the change in dissimilarity was due to some equalization of the sex ratios within occupations as women have slowly made inroads into previously male dominated occupations (sex ratio effect). Alternatively, some of the decline in dissimilarity could be due to different growth rates in the size of occupations over time (structure effect). For example, if total employment in the female overrepresented sectors has increased faster than employment in the female underrepresented sectors, the Duncan index will also register an increase, even if the sectoral sex ratios have remained the same. To examine these two effects the change (A) in the value of the Duncan index can be decomposed in the following (stylized) way:' A(D) = S [A(sex rtio effect)i + A (total employment effect), + (cross effect),] 7 The full expression can be found in Humphries (1988). Distribution of Female Employment 93 Table 3.5 shows the sign of the structure and sex ratio effects on the change in the value of the Duncan index by employment status for the 13 countries for which this calculation was possible. The table reads as follows: a positive sign suggests that the value of the index increased over time. A positive sign with respect to a particular effect suggests that, had this been the only effect, dissimilarity would have increased over time. The interpretation of negative signs is the opposite. Bearing this in mind, the first column of the table shows that the structure effect is in the same direction as the change in the value of the Duncan index in all cases with the exception of Chile and Panama; the sex effect is also in-line with the structure effect in eight countries and also in line with the total change in 10 countries. Out of the five countries where structure effect and the sex ratio effect worked in different directions, the former dominated over the latter in three cases and the latter over the former in the other two cases.' A similar inspection of dissimilarity among the self-employed (column 2) shows that the structure and sex ratio effects worked in the same direction in all but three countries: in two of these countries the structure effect dominated (Honduras and Peru) while in the other country the sex ratio effect was the dominant one (Colombia). With respect to employees and family workers (columns 3 to 4), the structure and sex ratio effects moved together in almost half of the cases,' while in the other half the structure effect dominated the sex ratio effect in almost as many cases as the latter dominated the former. Given this evidence one could argue that there is a mild dominance of the structure effect at the aggregate level. One interpretation of this may be that the growth of occupations is a more important determinant of women's relative position in the labor market than the improvement of women's occupational distribution. This can be explained by the fact that where there exists an excess supply of labor, as is the case in Latin America (particularly among women who are characterized by low participation rates), women are utilized in the customary occupations unless (or, until) shortages arise which enable them to enter 'male' sectors. However, the prominence of the structure effect is not that strong and a closer examination of the effects by detailed employment status (columns 2 to 4) may dilute to some extent the significance of this view. Therefore, one could also argue that the reduction in dissimilarity has come 2 The interaction term has been omitted as it relates directly to the other two effects and its interpretation is not readily obvious for policy matters. 29 Ecuador, Peru and Venezuela for employees, and Argentina, Mexico, and Venezuela for family workers. 94 Women's Employment and Pay in Latin America Table 3.5 Structure and Sex Ratio Effects on Occupational Dissimilarity Over Time All groups Self-employed Employees Family Workers Argentina Changein the index + - + + Structure effect + - - + Sex ratio effect + - + + Chile Change in the index + + + Structure effect - + Sex ratio effect + + + Colombia Change in the index - Structure effect - + - + Sex ratio effect - - + Costa Rica Change in the index - - - Structure effect - - - + Sex ratio effect - - Ecuador Change in the index - - - Structure effect - - - + Sex ratio effect - - - Guatemala Change in the index - - - + Structure effect - + - + Sex ratio effect - + + + Honduras Change in the index - - - - Structure effect - - - + Sex ratio effect + + - Continued Distribution of Female Employment 95 Table 3.5 (Cont.) Structure and Sex Ratio Effects on Occupational Dissimilarity Over Time All groups Self-employed Employees Family workers Jamaica Change in the index + Structure effect + Sex ratio effect + Mexico Change in the index - - - - Structure effect Sex ratio effect + - + Panama Change in the index + Structure effect Sex ratio effect + Peru Change in the index + + Structure effect + + - + Sex ratio effect + - - + Uru2uav Change in the index + Structure effect + Sex ratio effect + Venezuela Change in the index - - - + Structure effect - - - + Sex ratio effect + - - + Note: n+" ('2') indicates an increase (decrease) in the value of the index. Source: Calculated from ILO, 1990, Table 2. 96 Women's Employment and Pay in Latin America from the simultaneous beneficial effect of the changing structure of the labor force and the improved distribution of female workers within occupations. It should be noted again that the change in the gender composition and distribution of paid emplo yment (self-employed and employees) has been more responsible for the improvement in the overall employment distribution of women workers than family work. This evidence lends additional support to the view that women's position in the labor market may improve more rapidly as more women enter employment in the formal sector of the labor market.0 8. Qualifications A number of observations have emerged with respect to the occupational dissimilarity between women and men workers in Latin America, the changes in employment dissimilarity over time, and the separate experiences of workers by status (employees, self-employed and family workers). First, dissimilarity varies directly with the "formality" of employment: the occupational distribution of family workers is least diverse, followed by the self-employed. The occupational distribution oif workers in these two groups is, in turn, considerably less differentiated than that of salaried/wage workers. Second, there has been a clear tendency, in aggregate, for dissimilarity to decrease over time. The decline has been more uniform among the salaried/wage workers compared to the self- employed. These findings may create some optimism since women appear to be employed successively more like men during the process of development. However, this has to be qualified in two respects. First, the nature and availability of data have allowed us to examine only the incidence of horizontal dissimilarity -- and this was done at a very aggregate level. It is common knowledge that within each occupation the extent of vertical dissimilarity is severe and, perhaps, more important than horizontal dissimilarity. An example of vertical dissimilarity within the non-manual group could be that the chairman of an organization is male while his secretary is female or, within a more narrow context, an occupational group labelled "medical" may include a doctor as well as a nurse. In the absence of more informative data, little more can be said in this respect other than to state that in practice dissimilarity must be greater than what the present estimates suggest and to caution about the reliability of the results, in 30 One must also take into account the effect of public sector growth and the implications for women's employment, something which could not be. examined in the present analysis. Distribution of Female Employment 97 that the change in dissimilarity over time might look different had changes in vertical dissimilarity been taken into account. Second, it should be remembered that the analysis was based on a comparison of the relative employment distributions of women and men. Though dissimilarity, if unjustified, is a sine qua non condition for misallocation, the extent of misallocation and resulting welfare loss cannot be estimated from such a comparison.3" Therefore the decline suggested by the index might not have been accompanied by a reduction in the percentage of the labor force who are employed in the "wrong' occupation. This "paradox" is worth pursuing further as it has important implications for efficiency and public policy. This is done in the following section. 9. From Dissimilarity to Misallocation While accepting that the value of knowing the extent of employment dissimilarity across occupations is important, it should be recognized that the inefficiency induced from the misallocation of the labor force cannot be assessed in relative terms (women to men). One should also identify the implications of dissimilarity for the whole economy (women and men). We propose a method which (1) enables the calculation of the number of workers who would have to change occupation for the employment distributions of women and men to become identical, and then (2) expresses it as a percentage of the total labor force. The Duncan index is explicitly incorporated. Hence, the proposed approach provides more insight to our understanding of employment dynamics as it encompasses both the type of information already provided by the Duncan index as well as the size and sex ratio of the total labor force. In the previous section the value of the Duncan index was derived, first, by aggregating the absolute differences in the sectoral employment differentials and then by dividing the resulting sum by 2. In fact, one could derive the same 31 The term "misallocation" is used in a narrow sense. It denotes the gross difference between the actual female (male) employment in a sector and "expected" female (male) employment in that sector, had women and men been identical factors of production and been distributed equally across the sectors. It is hard to say which part of the difference between the employment distributions of women and men indicate misallocation (inefficiency, in the economist's sense) in the absence of information about the characteristics of workers and their respective wages. Some of these aspects are tackled in other sections of this book. 98 Women's Employment and Pay in Latin America result by taking into account only the female overrepresented sectors.2 For example, consider an economy where women are overrepresented in some sectors. Let the level of employment in all such sectors, k, be Fk and Mk respectively, and the total female and male labor force F and M respectively. Thus the actual value of the Duncan index, D, when calculated only from the female overrepresented sectors, is: D =|Fk| | Mk| To reach equality a number, Rk, of female workers would have to move out of the k female overrepresented sectors while an equal number of male workers would have to move in, that is: (Fk- Rk) _ (Mk Rk) = o F M From the last two equations one can obtain R,, the number of fernale or male workers who should be reallocated R =DMf where f is the share of all female workers in the total labor force (F/(F +M)). This implies that reallocations depend on the extent of inequality within sectors (the Duncan index) but only in part as the size of the male labor force and the share of all women workers in the total labor force should be also be taken into account. To find out the percentage of the labor force which would have to change sector, one should multiply R by 2 and divide the resulting number by the size of the total labor force. Table 3.6 presents the estimates for the percentage of the labor force which would have to change occupation in order to reach equality in the employment distribution of the sexes. With respect to the total labor force, the estimates typically vary from 10 to 20 percent. This figure is high and implies that up to half of all female workers may be "misallocated" -- if, of course, the initial 32 The only difference in the second case, that is when one examines only the female overrepresented sectors, is that there is no need to divide the resulting sum of employment differentials within sectors by 2. Distribution of Femalk Employment 99 dissimilarity in the occupational distributions of women and men were due to discrimination.33 The required reallocations among family workers are estimated to be relatively low (typically between 5 percent and 15 percent). Family workers appear to be more uniform across countries than workers in the other two employment groups. However, in all but two countries (Ecuador and Peru) the percentage of family workers who would have to change occupation has increased or remained the same. Initially the self-employed had slightly higher values than family workers but have experienced a decline in all countries except in Honduras and Mexico, though the increase has been a trivial one in the latter country. Self- employed workers have also become more homogeneous in the recent period: with the exception of Argentina, reallocations in this sector vary typically between 8 percent and 12 percent. Employees have had an experience similar to that of the self-employed. Though negative changes are observed in six countries compared to positive changes in four countries, the positive changes are hardly greater than one percentage point. However, the percentage of reallocations among employees still remains about double the figure for the other two groups. Given the difference in the experience of the groups of self-employed, employees and family workers, the interesting question becomes whether overall reallocations, as a percentage of the labor force have been on the rise. The present estimates suggest that reallocations have increased over time in all but three countries which have, however, registered a negligible decline of less than half of one percentage point (Colombia, Guatemala, and Mexico). Comparing the results for reallocations to the values of the Duncan index reported earlier, one can easily notice that high levels of dissimilarity are associated with high percentages of misallocated workers. However, the changes in the two measures over time are only loosely related to each other, suggesting opposite movements more often than not. This is because the Duncan index is equal to the ratio of necessary to potential reallocations, while the percentage of the labor force who would have to change sector is the ratio of the necessary reallocations divided by the actual size of the labor force.' This observation 33 This is so because women typically comprise only about one-third of the total labor force. 34 Let R be the number of reallocations. The Duncan index divides R by 2FM/(F+M) while reallocations as a percentage of the labor force are simply derived by dividing R by (F+M), where F (M) is the size of the total female (male) labor force. 100 Women's Employment and Pay in Latin America Table 3.6 Workers Who Would Have to Change Occupation to Reach Equality in the Employment Distribution of Women and Men, as a Percentage of the Total Labor Force Percent Percent Country Year Reallocated Year Reallocated All workers Argentina 1960 14.6 1980 16.8 Bolivia 1976 12.9 Brazil 1970 16.3 Chile 1952 16.4 1982 20.3 Colombia 1951 17.2 1964 16.7 Costa Rica 1963 16.4 1984 17.2 Ecuador 1962 13.5 1982 15.3 Guatemala 1950 13.9 1981 13.8 Honduras 1961 16.7 1974 17.4 Jamaica 1960 21.7 1970 24.8 Mexico 1970 13.0 1980 12.8 Panama 1950 16.3 1980 23.5 Peru 1961 10.6 1981 11.6 Uruguay 1963 15.8 1985 19.6 Venezuela 1961 16.6 1981 18.6 Self-emploved Argentina 1960 8.4 1980 5.9 Bolivia 1976 14.3 Brazil 1970 - Chile 1952 13.8 1982 10.2 Colombia 1951 12.5 1964 10.6 Costa Rica 1963 8.0 1984 7.6 Ecuador 1962 11.8 1982 8.0 Guatemala 1950 13.2 1981 9.8 Honduras 1961 10.7 1974 16.9 Jamaica 1970 - Mexico 1970 10.6 1980 11.0 Panama 1950 8.8 Peru 1961 10.0 1981 9.4 Uruguay 1985 12.3 Venezuela 1961 10.9 1981 8.8 Continued Distribution of Female Employment 101 Table 3.6 (Cont). Workers Who Would Have to Change Occupation to Reach Equality in the Employment Distribution of Women and Men, as a Percentage of the Total Labor Force Percent Percent Country Year Reallocated Year Reallocated Employees Argentina 1960 17.4 1980 19.6 Bolivia 1976 21.2 Brazil 1970 - Chile 1952 20.1 1982 20.0 Colombia 1951 23.6 1964 18.1 Costa Rica 1963 20.8 1984 18.8 Ecuador 1962 19.1 1982 20.2 Guatemala 1950 18.9 1981 19.8 Honduras 1961 25.6 1974 20.9 Jamaica 1970 - Mexico 1970 14.5 1980 14.8 Panama 1950 24.0 Peru 1961 17.9 1981 15.0 Uruguay 1985 21.6 Venezuela 1961 21.2 1981 18.3 Family workers Argentina 1960 6.3 1980 13.4 Bolivia 1976 4.6 Brazil 1970 - Chile 1982 10.7 Colombia 1951 9.3 1964 10.9 Costa Rica 1963 6.2 1984 7.7 Ecuador 1962 9.6 1982 4.2 Guatemala 1950 6.1 1981 6.1 Honduras 1961 4.1 1974 6.2 Jamaica 1970 - Mexico 1970 13.5 1980 14.3 Panama 1950 3.7 Peru 1961 3.4 1981 2.9 Uruguay 1985 12.9 Venezuela 1961 2.5 1981 15.1 - not available. Source: Calculated from ILO, 1990, Table 2. 102 Women's Employment and Pay in Latin America and the evidence suggest that the decline in occupational dissimilarity between women and men has not been fast enough to compensate for the rising numbers of working women in the labor force." This finding has important policy implications to the extent that women's occupational status is not the result of genuine differences between the labor supply decisions of women and men, but are rather the direct or indirect result of possible discriminatory practices which result in an inefficient use of female labor in the economy. 10. Comparison with Other Countries Table 3.7 shows the dissinilarity between the occupational distribution of female and male workers in seven industrialized countries (excluding members of the European Commmunities, EEC, which are examined later).6 The data upon which the calculations are based refer to one digit ISCO37 level and refer to all workers. In this respect the data are comparable to those used in our estimates of dissimilarity for Latin America presented in the top panel of Table 3.4. As a general observation the index of occupational dissimilarity in the industrialized countries is below the 0.49 mark. The index varies between 0.379 and 0.486 in all countries except in Japan where the value of the index is 0.224. In contrast, our estimates JFor Latin America (Table 3.4) reveal that in as many as eight of the 15 Latin American countries studied in this volume the value of the index is greater than 0.49. In only three of the Latin America countries (Bolivia, Mexico, and Peru) was the value of the index below the 0.40 mark, but even in these countries the value of the index was above 0.30. Therefore, occupational dissimilarity appears to be more severe in Latin America than in the industrialized world. 3 This finding is consistent with the experience in some industrialized countries. For example, in Britain occupational segregation, as measured by the Duncan index, has decreased by more than 10 percent since 1900 while the percentage of the labor force who should change occupation has increased by about 3 percent (Tzannatos, 1990). Also, our own estimates for the United States suggest that the value of the index has decreased by more than 20 percent while the percentage of the labor force which should change occupation to restore equality in the employment distributions of the two sexes has increased by more than 25 percent in the 1900-1970 period. 36 The table is taken from OECD The Role of Women in the Economy: Report on Occupational Segregation hy Sex, Chapter II, Paris 1984. 37 International Standard Classification of Occupations. Distribution of Female Employment 103 Table 3.7 Dissimilarity Between Female and Male Occupational Employment in Selected Industrialized Countries (1970-1982) Country Value of the Annual % Change Structure Sex Ratio Index in the Index Effect Effect (1982) (1970-1982) (1) (2) (3) (4) Australia 0.479 0.1 + - Canada 0.379 -0.6 - - Japan 0.224 -0.9 - - Norway 0.478 -0.2 - + Sweden 0.422 0.5 + - United States 0.411 0.7 + + " +" ("-") indicates an increase (decrease) in the value of the index. Note: Norway and Sweden (1981) Source: Adapted from OECD (1984). This conclusion is reinforced by observing that the present comparison refers to all workers. Recall that the evidence for Latin America reveals that occupational dissimilarity is lower among the self-employed and family workers. These two groups employ a greater percentage of workers in developing countries than in industrialized countries. In industrialized countries around three-quarters of all workers are employees. If this difference is taken into account, then the dissimilarity among employees in Latin America should be even greater than among their counterparts in industrialized countries. In terms of over time changes in occupational dissimilarity among industrialized countries, the evidence is mixed in that dissimilarity increased in three countries while it decreased in another three countries. This finding is comparable to that for Latin America (last column in Table 3.4). However, the sex ratio effect in industrialized countries appears to be negative more often than in Latin America. Recall that Table 3.5 showed that among the 13 Latin America countries where the identification of the two effects was possible, the sex ratio effect was positive in nine countries. This means that, had changes in the relative employment of women and men within occupations been the only change in Latin America over time, the occupational dissimilarity index would have increased. The increase in the value of the index was, however, averted by opposite and stronger changes in the occupational structure. Hence, one may 104 Women's Employment and Pay in Latin America conclude that decreases in occupational dissimilarity during development come initially from structural changes and it is only after some critical point in time that "equal employment" effects become operative. We finally examine the case of the Member States of the European Communities (EEC) in the 1980s. There are many advantages in focusing upon EEC countries alone. The region can be said to be subjected to more homogeneous forces than the group of all advanced countries taken together as there are no trade barriers within EEC and there are no restrictions on migration. In addition EEC has fairly common attitudes toward the social sector and has enacted legislation at Community level which outlaws sex discrimination and promotes sex equality. As there is no information for occupational dissimilarity in Europe, we concentrate on industrial data for both Europe and Latin America. The results for Europe are presented in the top panel of Table 3.8 and the results for Latin America in the lower panel. What is striking is the sulbstantial uniformity of results across Europe. First, the value of the index is only between 0.30 and 0.40 across the Communities (column 1): in seven countries the index registers a value of between 0.30 and 0.35 and in the other five countries the value of the index is between 0.36 and 0.39. Some reasons for this uniformity were mentioned in the previous paragraph. Another reason may be that the results are based upon the harmonized labor statistics held at Statistical Office of the European Communities in Luxembourg (EUROSTAT). Hence, discrepancies due to differences in the statistical treatment of employment should be minimal. Second, the uniformity of the European results also applies to changes over time. All Community countries have experienced a decline in dissimilarity over time; a negative structure effect; and a negative sex ratio effect (except in Germany in which the positive sex ratio effect was not strong enough to counterbalance the negative structure effect). In contrast, the results for Latin America are mixed. Industrial dissimilarity in the region varies from about 0.30 in Mexico and Peru to over 0.40 in Bolivia, Brazil, Ecuador, Jamaica and Colombia.'8 The structure effect in Latin America was positive in seven cases and the sex ratio effect was positive in three cases. Only in three countries, namely Bolivia, Colombia, and Jamaica, were the two effects in the same 3 Note, however, that the estimate for Colombia refers to a relatively early year due to the unavailability of data in more recent periods. Distribution of Female Employment 105 Table 3.8 Dissimilarity Between Female and Male Industrial Employment A. European Community (1983-1989)" Country Value of the % change Structure Sex ratio Reallocations' index in the index effect effect (1989) (1989) (1983-1989) (1983-1989) (1983-1989) (1) (2) (3) (4) (5) Germany 0.3355 -0.3 - + 16.0 France 0.3059 -0.2 - - 15.0 Italy 0.3224 -0.2 - - 14.9 Netherlands 0.3650 -0.6 - - 17.0 Belgium 0.3658 -0.2 - - 17.1 Luxembourg 0.3958 -1.2 - - 18.1 UK 0.3448 -0.6 - - 17.1 Ireland 0.3326 -0.2 - - 15.8 Denmark 0.3336 -1.1 - - 16.6 Greece 0.3287 -1.0 - - 14.8 Spain' 0.3602 - - - 15.5 Portugald 0.3803 - - - 15.4 B. Latin America (Selected years)' Country Value of % change Structure Sex ratio Reallocations' the index in the indexe effect effect (latest years) (latest years) (1) (2) (3) (4) (5) Argentina 1970 0.3845 -0.5 + - 14.5 Bolivia 1976 0.4275 1.9 + + 15.0 Brazil 1980 0.4007 0.5 + - 15.9 Colombia 1964 0.5524 -0.6 - - 17.7 Ecuador 1982 0.4113 0.0 + - 13.4 Jamaica 1982 0.4726 0.5 + + 21.7 Mexico 1980 0.3062 -0.8 + - 11.0 Peru 1981 0.3011 -0.6 + - 10.6 Venezuela 1981 0.3803 -1.7 - + 15.2 - not available. a. "+' ('-') indicates an increase (decrease) in the value of the index. b. European results based on 11 industrial sectors uniformly defined across the Communities. c. Percentage of the total labor force who would have to change industrial sector to reach equality in the employment distributions of female and male workers. d. Spain and Portugal were not rnembers of the European Commnunities in 1983. e. The base years for Latin America countries are those indicated in Table 3.4 Sources: Based on EEC Labor Force Survey (1983 and 1989). Latin America: Calculated from Appendix Table A3.2. 106 Women's Employment and Pay in Latin America direction. In Ecuador the two changes completely offset each other. In three of the remaining countries the sex ratio effect was the dominant one (Argentina, Mexico, and Peru) while the structure effect was stronger in Brazil and Venezuela. The foregoing comparison with other regions implies a number of conclusions. First, employment dissimilarity in Latin America is still relatively high. Second, though there is some tendency for employment dissimilarity to decrease in Latin America, the tendency is not very strong. Third, the decrease in employment dissimilarity in Latin America is effected more by changes in the employment structure than by improvement in the sex ratios within given employment groups. Fourth, and finally, the assertion, that there are common factors operating during the process of development that lead toward some common patterns of women's employment, is not incompatible with the uniformity of results presented for European countries. Therefore it seems that women in Latin America are increasingly employed more "like men' but the process is slow and relies on structural (rather than "equal employment") effects. The issue becomes whether the "unbalanced" sex ratios reflect some kind of imperfection whose removal by appropriate policy measures can increase efficiency and accentuate, the process of growth. 11. Conclusion This chapter has examined some characteristics of the distribution of female employment in Latin America and Caribbean countries. Within a developmental framework (initial conditions and changes over time) it tried to establish the importance of work status, industries and occupations in terms of female/male employment. Some facts emerged but equally a number of qualifications apply. First, the service sector clearly stands out as an important employer for women. Manufacturing, rather than agriculture, is the next in the line. One relevant observation is that workers in agriculture may evade statistical enumeration more easily than workers in other sectors. Even so, the direction of employment change is known with fair a degree of certainty: during development the share of employment in the modern sector increases at the expense of agriculture/ traditional sector. Policies which facilitate this transition by making workers able to work in the modem sector may be preferred to corrective policies in the primary sector which are bound to be short lived and can even have adverse effects, if they are difficult to eliminate after the economic crisis is over. Second, the distribution of women in paid employment (versus family work) has become more equal to that of men. Within paid employment, female employees Distribution of Female Employment 107 (versus the self-employed) appear to have made most of the gains. This finding reinforces the significance of dependent employment for sex equality in the labor market. However, the role of the public sector upon these changes still remains unanswered and needs to be studied further, when appropriate data become available. Third, despite the decline in dissimilarity among employees, the dissimilarity for this group of workers is still almost twice as high as the dissimilarity among the self-employed and family workers. Given this difference and the fact that dependent employment becomes successively more important during development, an examination of occupational choice in the formal sector may be particularly rewarding. Finally, despite the gains achieved in paid and formal employment, the percentage of the total labor force who are potentially 'misallocated' has increased (subject to the use of the term adopted in this chapter). The validity of this conclusion does not necessarily depend on the assumption that dissimilarity between the employment distributions of women and men reflects to a great extent some economic imperfection (of which discrimination may be one). It would also be valid if the extent of (whatever small) imperfection has remained constant over time. Of course, both statements are tentative, given the quality of the available data. This remains an area where more analysis needs to be undertaken. 108 Women's Employment and Pay in Latin America Statistical Appendix to Chapter 3 The data reported in this Appendix exclude the members of the armed forces, the unemployed, those seeking employment for the first time and those not classified by either occupation/industry or employment status. These data were used in the sectoral analysis undertaken in this volume (Chapter 3). However, the aggregate analysis (Chapter 2 on the total size of the labor force) included these persons. The classification by employment status was based on the International Classification of Employment by Status (ICSE) as recommended by the Population Commission of the United Nations in 1948 and amended by the Statistical Commission of the United Nations in 1958 (ILO, 1990, p. XXXV). The basic classification includes six status groups, namely employers, own- account workers, employees, unpaid family workers, members of producers' cooperatives and persons not classifiable by status. In the analysis undertaken in this volume the first two groups were considered and presented together under the heading 'self-employed" workers while unpaid family workers were mentioned simply as family workers. The last two groups are numerically unimportant in most cases and were omitted from the analysis. The industrial breakdown was based on the 1958 International Standard Industrial Classification of all Economic Activities (ISIC-1958) for the early period and on ISIC-1968 for the late period. Similarly, the occupational breakdown was based on the 1958 International Standard Classification of Occupations (ISCO-1958) for the early period and on ISCO-1968 for the late period. The correspondence between the 1958 and 1968 industrial/occupational classifications is as follows: Distribution of Female Employment 109 Correspondence Between the 1958 and 1968 Industrial Classifications ISIC-1958 ISIC-1968 Industry Division Major Division Agriculture, forestry, hunting and fishing 0 1 Mining and quarrying 1 2 Manufacturing 2-3 3 Construction 4 5 Electricity, gas, and water 5 4 Commerce 6 Wholesale/retail trade, restaurants, hotels 6 Transport, storage, and communication 7 7 Services 8 Finance, insurance, real estate, and business 8 Community, social, and personal services 9 Activities not adequately described 9 0 Correspondence Between the 1958 and 1968 Occupational Classifications ISCO-1958 ISCO-1968 Occupation Major Major Group Group Professional, technical, and related workers 0 0/1 Administrative, executive, and managerial workers 1 Administrative and managerial workers 2 Clerical workers 2 Clerical and related 3 Sales workers 3 4 Farmers, fishermen, hunters, loggers, and related workers 4 Agriculture, animal husbandry, and forestry workers etc. 6 Miners, quarrymen, and related workers 5 Workers in transport and communication occupations 6 Craftsmen, production-process workers, and laborers 7-8 Production workers, transport operators, and laborers 7/8/9 Service, sport, and recreation workers 9 Service workers 5 Workers not classifiable by occupation X X Members of the armed forces 110 Women's Employment and Pay in Latin America The minimum age for classifying a person as active/inactive varies between countries and has also varied within countries at different time periods. The table below shows the minimum age at which national censuses drew the distinction between activity and inactivity in the early and late periods under consideration. Minimnum Age for Inclusion in the Labor Force Period Country Early Late Argentina 14 14 Bolivia 10 10 Brazil 10 12 Colombia 10 10 Costa Rica 12 12 Chile 12 15 Ecuador 12 12 Guatemala 7 10 Honduras 10 10 Jamaica 14 14 Mexico 12 12 Panama 10 10 Peru 6 6 Uruguay 10 12 Venezuela 10 12 Distribution of Female Employment 111 The industrial and occupational data were at times based on different minimum ages than those used in the calculation of the labor force participation action rates. The table below shows the minimum age for the industrial and occupational data. Minimum Age for Classification by Sector Period Country Early Late Argentina 15 14 Bolivia 10 10 Brazil 10 12 Colombia 10 10 Costa Rica 12 12 Chile 12 15 Ecuador 12 12 Guatemala 7 10 Honduras 10 10 Jamaica 14 14 Mexico 12 12 Panama 10 10 Peru 6 15 Uruguay 10 12 Venezuela 10 12 For more information see ILO Year Book of Labor Statistics: Retrospective Edition on Population Censuses 1945-1989, Geneva: International Labor Office, 1990. Appendix Table A3.la Occupational Distribution of the Labor Force by Employment Status ARGENTINA Total Labor Force Self-Employed Employees Family Workers 1960 (M) (F) (M) (F) (M) (F) (F) Professionrl 185565 263619 56909 21723 122170 237662 173 278 Administrative 169394 13477 92391 5978 72708 6755 428 166 Clerical 579980 235393 576838 234149 627 577 Sales 582058 118220 370727 44188 196660 68975 7963 4534 Service 259127 426300 35042 31294 220655 390243 972 2770 Farmers 1263114 68705 481295 16183 656852 26938 121228 25291 Manual 2255276 365335 410492 106072 1797447 238370 9169 3867 Total 5294514 1491049 1446856 225438 3643330 1203092 140560 37483 Total Labor Force Self-Employed Employees Family Workers 1970 (M) (F) M (F) M (F) M (F) Professional 306000 371500 109000 38150 190500 327750 1500 1000 Administrative 128350 9500 65150 3400 61750 5950 300 50 Clerical 660100 365300 3300 500 649600 360750 500 1150 Sales 816700 256100 458200 107100 331400 137950 18150 8950 Service 451300 685250 53150 43900 386800 603650 3550 19250 Farmers 1218150 77950 415750 25400 647700 27950 130500 23600 Manual 2743150 348200 510250 96000 2146550 236500 30150 9000 Total 6323750 2113800 1614800 314450 4414300 1700500 184650 63000 Appendix Table A3.lb Occupational Distribution of the Labor Force by Employment Status BOLIVIA Total Labor Force Self-Employed Employees Family Workers 1976 (M) (F) (M) (F) (M) (F) (M) (F) Professional 50183 35317 9607 1269 40081 33860 107 97 Administrative 7488 1604 3405 1128 4009 407 24 48 Clerical 41020 18589 928 225 39902 18241 31 49 Sales 41248 50137 33885 45837 6831 3259 377 913 Service 57153 71442 3671 6160 50061 64646 2812 296 Farmers 607950 89190 440334 41080 84791 3114 79387 44621 Manual 310073 60462 97783 44479 207355 10069 2122 5504 Total 1115115 326741 589613 140178 433030 133596 84860 51528 BRAZIL Total Labor Force Self-Employed Employees Family Workers 1970 (M) (F) (M) (F) (M) (F) (M) (F) Professional 575545 835201 Administrative 440076 57021 Clerical 1035150 526528 Sales 1873448 320213 Service 868018 2192291 Farmers 11782384 1256765 Manual 5560719 702852 Total 22135340 5890871 - not available. Appendix Table A3.1c Occupational Distribution of the Labor Force by Employment Status CHILE Total Labor Force Self-Employed Employees Family Workers 1952 (M) (F) (M) (F) (M) (F) (M) (F) Professional 52019 40176 12955 4602 38773 35341 Administrative 98400 39216 83313 32288 12709 2141 Clerical 116100 44521 3577 1107 112214 42971 Sales 40509 14292 21752 6197 18219 7227 Service 81277 218791 6872 26531 74066 191593 Farmers 570271 38119 155836 16648 392193 16203 Manual 540996 124350 74185 53359 462839 69333 Total 1499572 519465 358490 140732 1111013 364809 Total Labor Force Self-Employed Employees Family Workers 1982 (M) (E;) >1) (F) (M) (F) (F) (F) Professional 133370 148061 23952 8229 108773 139197 645 635 Administrative 73276 18850 49201 14314 23765 4312 310 224 Clerical 255778 163854 8366 3132 245885 158909 1527 1813 Sales 214653 101819 117824 46810 90121 50268 6708 4741 Service 127201 303457 9713 14536 116159 287636 1329 1285 Farmers 621661 18144 161854 4460 376908 10020 82899 3664 Manual 1049914 119177 170205 27479 865925 90156 13784 1542 Total 2475853 873362 541115 118960 1827536 740498 107202 13904 - not available. Appendix Table A3.ld Occupational Distribution of the Labor Force by Employment Status COLOMBIA Total Labor Force Self-Employed Employees Family Workers 1951 (M) (F) (M) (F) (M) (F) (M) (F) Professional 54660 32416 16924 1881 33504 28246 273 219 Administrative or 247846 57603 134408 21822 103076 31820 3829 2314 Admin./Clerical Sales 42727 19832 11586 5029 29288 13739 778 624 Service 83018 313680 13688 15034 54553 291479 430 776 Farmers 1907900 86617 777160 44138 823813 24444 262213 15576 Manual 583770 184194 133185 98226 409989 66808 9080 14410 Total 2919921 694342 1086951 186130 1454223 456536 276603 33919 Total Labor Force Self-Emoloyed Employees Family Workers 1964 (M) (F) (M) (F) A) 7 (M Professional 106174 95250 31530 6511 73711 87648 318 562 Administrative or 266922 104293 63256 11519 199550 90080 1755 1939 Admin./Clerical Sales 215157 73697 132716 33555 75927 34823 4917 4881 Service 146183 428297 20639 22988 122997 400351 1008 3242 Farmers 2320295 106404 974535 53106 994496 27728 343652 25354 Manual 897993 195028 240197 97444 626614 76048 13353 18223 Total 3952724 1002969 1462873 225123 2093295 716678 365003 54201 Appendix Table A3.1e Occupational Distribution of the Labor Force by Employment Status COSTA RICA Total Labor Force Self-Employed Employees Family Workers 1963 (M) (F) (M) (F) (i) (F) (F) Professional 9067 11577 1378 152 7572 11051 115 373 Administrative 4595 548 1810 343 2728 185 53 20 Clerical 15022 5645 172 19 14770 5571 80 55 Sales 24336 5585 11713 1381 11255 3733 1365 471 Service 11457 26244 898 356 10362 25603 164 283 Farmers 183529 3040 50234 262 97266 2449 36018 329 Manual 64848 10181 9716 3159 53809 6765 1312 257 Total 312854 62820 75921 5672 197762 55357 39107 1788 Total Labor Force Self-Employed Employees Family Workers 1984 (M) (F) (M) (F) (v) (F) (NO (F) Professional 44786 34251 6293 1320 38327 32811 166 120 Administrative 20017 3746 4325 732 15559 2965 133 49 Clerical 30093 28695 713 557 29266 28007 114 131 Sales 52486 15507 27805 5076 23816 9978 865 453 Service 41732 51638 2461 2448 38916 48410 355 780 Farmers 233670 5527 78682 581 123190 4595 31798 351 Manual 153606 24995 31186 3711 120717 21114 1703 170 Total 576390 164359 151465 14425 389791 147880 35134 2054 Appendix Table A3.lf Occupational Distribution of the Labor Force by Employment Status ECUADOR Total Labor Force Self-Employed Employees Family Workers 1962 (M) (F) (M) (F) (M) (F) (M) (F) Professional 24991 21982 5895 1011 17136 17837 Administrative 4462 329 2000 170 2430 142 Clerical 34661 13311 34488 13121 79 35 Sales 66466 19948 53685 13553 11105 4534 1654 1838 Service 32533 69642 7284 4841 24613 64206 221 189 Farmers 760763 39627 381419 18285 304909 13731 74256 7399 Manual 238950 66832 84561 47124 146997 14634 6906 4943 Total 1162826 231671 534844 84984 541678 128205 83116 14404 Total Labor Force Self-Employed Employees Family Workers 1981 (M) (F) (M) (F) (M) (F) (M) (F) Professional 105797 77782 19041 4475 84378 71760 203 109 Administrative 9394 1729 3140 413 6103 1267 20 12 Clerical 72850 59075 1176 1044 70464 57213 253 96 Sales 146681 62699 107896 44647 33828 14952 2717 2061 Service 71222 108222 11584 13460 56713 83255 517 1871 Farmers 726050 58717 387166 27668 216418 9412 88167 18190 Manual 531900 69474 197114 31801 298047 30830 10971 2743 Total 1663894 437698 727117 123508 765951 268689 102848 25082 Appendix Table A3.Ig Occupational Distribution of the Labor Force by Employment Status GUATEMALA Total Labor Force Self-Employed Employees Family Workers 1950 M) (F) (F) (F) (F) Professional 8910 6661 1976 326 6866 6299 68 36 Administrative 9147 5193 6440 4680 2524 289 183 224 Clerical 13441 3446 326 99 12979 3242 136 105 Sales 22419 9738 16702 6213 3801 2783 1916 742 Service 22214 44714 1565 2843 20483 41456 166 415 Farmers 635309 16493 292225 3014 189783 5597 153301 7882 Manual 129394 37204 41977 23047 80023 8891 7394 5266 Total 840834 123449 361211 40222 316459 68557 163164 14670 Total Labor Force Self-Employed Employees Family Workers 1981 (M) (F) (M) (i) (l (F) (M) (F) Professional 49490 31747 7980 1615 39687 28906 109 74 Administrative 16836 3281 3619 1194 12739 1807 31 32 Clerical 33121 23447 1065 562 31156 22118 72 71 Sales 66481 33035 42748 20127 20329 11192 1805 834 Service 38005 71843 2254 3883 34983 66154 112 354 Farmers 890628 20629 502366 5157 272805 11180 96260 3231 Manual 306635 43308 90821 22566 201866 16339 6146 2869 Total 1401196 227290 650853 55104 613565 157696 104535 7465 Appendix Table A3.lh OccuPational Distribution of the Labor Force by Employment Status HONDURAS Total Labor Force Self-Employed Employees Familv Workers 1961 (M) (F) (O (1F) (M) (E) (M) (F) Professional 6128 8172 895 139 5058 7966 8 11 Administrative 2975 367 519 139 2454 225 2 3 Clerical 8447 4186 87 13 8341 4100 18 73 Sales 14379 8029 9607 4956 4320 2501 449 570 Service 12392 32134 1071 1831 11224 30078 65 143 Farmers 371348 2874 183319 1545 98760 626 89262 703 Manual 52830 12431 10588 7597 41298 3605 927 1226 Total 468499 68193 206086 16220 171455 49101 90731 2729 Total Labor Force Self-Employed Employees Family Workers 1974 (M) (F) (M) (F) (M) (F) (M) (F) Professional 16398 14584 1780 284 14556 14267 15 15 Administrative 5420 1592 873 622 4540 968 1 Clerical 22479 9305 9 21 22451 9226 12 46 Sales 27715 16192 17591 10968 9093 4312 999 892 Service 13376 36298 964 2021 12363 34186 33 78 Farmers 447153 5960 222126 1840 123629 2682 101214 1433 Manual 97985 33423 21376 21132 74124 9527 2281 2734 Total 630526 117354 264719 36888 260756 75168 104554 5199 Appendix Table A3.li Occupational Distribution of the Labor Force by Employment Status JAMAICA Total Labor Force Self-Employed Employees Family Workers 1960 (M) (F) (F) (M) ()(MF Professional 6871 12309 633 472 5998 11465 236 366 Administrative 9824 2182 7043 1913 2781 269 0 0 Clerical, Sales 27893 41052 7528 19788 19811 19133 550 2103 Service 12321 76252 861 3378 11353 71881 99 863 Farmers 194504 38702 104709 15990 78510 16623 11264 6087 Manual 128208 48887 20006 24970 100556 18749 7519 5141 Total 379621 219384 140780 66511 219009 138120 19668 14560 Total Labor Force Self-Employed Employees Family Workers 1982 (M) (F) (MI) (F) (M) (F) (M) (F) Professional 20518 29750 Administrative 10904 8828 Clerical, Sales 25243 51212 Service 20747 45757 Farmers 114956 13702 Manual 117376 22108 Total 309744 171357 - - not available. Appendix Table A3.1j Occupational Distribution of the Labor Force by Employment Status MEXICO Total Labor Force Self-Employed Employees Family Workers 1970 (M) (F) (M (F) M (F) M (F) Professional 485268 247941 140529 39723 330134 201393 14605 6825 Administrative 267777 52051 128089 25378 139688 26673 Clerical 579347 397832 66001 38408 503061 350750 10285 8674 Sales 698258 269009 393747 126475 247498 107515 57013 35019 Farmers 4724803 227397 1925262 84082 2347287 113024 452254 30291 Laborers 2415701 353079 430887 97342 1919688 234660 65126 21077 Transport 876173 684441 183476 156991 671861 499278 20836 28172 Total 10488800 2466257 3407562 646506 6411327 1643495 669911 176256 Total Labor Force Self-Employed Employees Family Workers 1980 (M) (F) (F) (M) (F ) (F) Professional 976039 622967 187989 49739 603263 436217 13541 11231 Administrative 202877 38647 81032 15794 98909 17025 877 802 Clerical 1133961 883519 58415 32693 870557 684020 21660 17229 Sales 1094760 517562 523872 231999 340023 147940 59255 51069 Farmers 4854926 678320 2396363 291044 1096114 96817 339288 75438 Laborers 4767377 871115 797075 152804 2972986 490596 201039 61329 Transport 471875 1101776 78493 190437 284500 545390 19400 91034 Total 15924806 6141278 4474365 1484327 7125648 2640863 776715 416276 Appendix Table A3.lk Occupational Distribution of the Labor Force by Employment Status PANAMA Total Labor Force Self-Emoloyed Employees Family Workers 1950 (M) (F) (M) '(F) (F) (M) (F) Professional 4995 4896 1052 107 3931 4776 12 13 Administrative 6400 1109 4291 878 2072 188 37 43 Clerical 5648 4844 135 11 5501 4797 12 36 Sales 5979 3456 1724 521 3893 2557 362 378 Q Service 10729 17179 633 1403 10057 15726 39 50 Farmers 123057 7255 78267 2016 11692 226 33098 5013 Manual 38001 6246 7104 3233 30600 2540 297 473 Total 194809 44985 93206 8169 67746 30810 33857 6006 Total Labor Force Self-Employed Employees Family Workers 1980 (M) (F) () (F) () (F) () (F) Professional 25574 28111 Administrative 20162 4682 Clerical 18195 37176 Sales 22677 11738 Service 35275 43902 Farmers 135506 3838 Manual 114721 10363 Total 372110 139810 not available. Appendix Table A3.11 Occupational Distribution of the Labor Force by Employment Status PERU Total Labor Force Self-Employed Employees Family Workers 1961 (M) (F) (M) (F) (M) (F) (m) (F) Professional 55423 47289 12688 2491 41986 44129 66 86 Administrative 40338 4844 11106 2083 28739 2419 74 53 Clerical 91352 41993 1259 187 89479 41102 146 255 Sales 161116 65184 117088 49810 39679 10935 4123 4326 Service 111812 167450 8083 13311 103426 153654 303 485 Farmers 1321288 213488 720372 87838 413098 50088 187150 75417 Manual 535647 116947 152393 77394 376602 33329 4468 5698 Total 2316976 657195 1022989 233114 1093009 335656 196330 86320 N Total Labor Force Self-Employed Employees Family Workers 1981 () (F) (Mi)(F) (M) (F) Professional 254783 149690 38802 7786 211968 139329 268 277 Administrative 22031 1894 5357 423 16674 1471 0 0 Clerical 382307 184669 3492 1070 377857 182389 184 212 Sales 347166 180466 250657 141982 89460 28815 3115 6932 Service 184596 187135 17829 21652 164635 163609 387 1014 Farmers 1576719 263748 1152119 110519 291364 28489 106226 117114 Manual 903511 116897 290938 65110 591940 39446 3800 8320 Total 3671113 1084499 1759194 348542 1743898 583548 113980 133869 Appendix Table A3.lm Occupational Distribution of the Labor Force by Employment Status URUGUAY Total Labor Force Self-Employed Employees Family Workers 1963 () (F) (F) (F) () (F) Professional 24245 32739 Administrative 12404 709 Clerical 92071 34935 Sales 75401 19355 Service 52238 87222 Farmers 175581 3413 Manual 263911 55811 Total 695851 234184 Total Labor Force Self-Employed Employees Family Workers 1985 (M) (F) () (F) (M) (F) (F) Professional 43098 61969 11203 9173 30316 50935 505 536 Administrative 21808 5703 10498 2572 10855 2925 17 31 Clerical 75319 65326 1109 695 72582 63508 155 492 Sales 77334 40400 45832 19121 29505 18965 1241 1963 Service 54047 121634 4172 14253 49313 105809 199 565 Farmers 154874 10879 56494 4280 88767 2423 8514 4093 Manual 271945 60043 57598 16408 208121 41829 1411 397 Total 698425 365954 186906 66502 489459 286394 12042 8077 not available. Appendix Table A3.1n Occupational Distribution of the Labor Force by Employment Status VENEZUELA Total Labor Force Self-Employed Employees Familv Workers 1961 (W (F) (M) (F) M (F) (F) Professional 63745 63430 9220 2849 52181 58518 3 2 Administrative 29798 3628 17143 2908 12533 697 7 Clerical 102262 53120 1710 288 100165 52644 86 48 Sales 208740 20705 131092 10254 73411 9958 3718 433 Service 96925 167630 8510 24657 87823 141899 366 678 Farners 736127 25215 396034 14981 245368 6974 93461 3206 Manual 556794 73666 122737 40313 430247 32628 1940 521 Total 1794391 407394 686446 96250 1001728 303318 99581 4888 Total Labor Force Self-Employed Emplovees Family Workers 1981 (M) (E) (M) (E) (M) (F) Professional 262744 232291 22521 6019 228902 209781 158 177 Administrative 67523 8718 20761 1438 44568 6901 27 45 Clerical 236168 284316 6724 1992 217943 267053 229 495 Sales 381007 116789 183915 46311 185443 65907 1882 1544 Service 220800 332189 13465 24366 185905 144552 366 1655 Farmers 508373 15518 256640 6432 197674 6656 32484 1580 Manual 1637027 243982 284441 28702 978296 119940 4312 1431 Total 3313642 1233803 788467 115260 2038731 820790 39458 6927 Appendix Table A3.2a Industrial Distribution of the Labor Force by Employment Status ARGENTINA Total Labor Force Self-Employed Employees Family Workers 1960 (N) (F) (M) (F) (M) (F) (M) (F) , .1 Agriculture 1248624 75327 480487 16597 618260 30220 121140 25479 Mining 39341 1102 2399 92 36344 927 153 10 Manufacturing 1471339 384655 305631 108214 113539 255074 7297 4017 Construction 420888 5537 102741 875 308574 4522 1202 32 Utilities 79816 2902 1048 22 77716 2812 55 4 Commerce 744020 169460 374486 45433 339157 115997 8285 4662 Transport 496819 24746 76192 620 412919 23943 1202 49 Services 733188 794939 122340 57393 599656 726388 1517 3291 Total 5234035 1458668 1465324 229246 3528024 1159883 140851 37544 Total Labor Force Self-Employed Employees Family Workers 1970 (M) (F) (IM) (F) (M) (F) (M) (F) Agriculture 1243150 87950 416050 25650 671250 36400 132200 24300 Mining 42850 1750 950 41050 1750 100 Manufacturing 1357500 413750 184000 93500 1139900 306500 10150 7200 Construction 699050 12250 161000 1100 518700 10850 6500 100 Utilities 90750 5800 1100 88350 5650 50 1250 Commerce 1008500 316300 480450 110250 494550 192650 21750 10500 Transport 541450 51800 109500 4500 418700 45600 3250 1000 Services 1157500 1193900 215200 74700 919000 1077750 5850 19050 Total 6140750 2083500 1568250 309700 4291500 1677150 179850 62150 Appendix Table A3.2b Industrial Distribution of the Labor Force by Employment Status BOLIVIA Total Labor Force Self-Employed Employees Family Workers 1950 (M) (F) ( (F) (M) (F) (M) (F) Agriculture 274772 397358 65698 9674 59675 12917 147537 373722 Mining 38969 4132 1564 168 37153 3636 184 309 Manufacturing 64539 44981 20274 24899 42183 8827 1361 11191 Construction 25131 869 1325 307 23574 554 45 7 Utilities 30588 10050 3102 121 27337 9702 92 204 Commerce 32736 24310 23812 14850 7285 2413 1531 7009 Transport 20492 782 2733 97 17558 668 90 16 Services 17513 52005 1567 2446 15864 48806 48 680 Total 504740 534487 120075 52562 230629 87523 150888 393138 Total Labor Force Self-Employed Emplovees Family Workers 1976 (M) (F) (NOi) (F) (M) (F) (M) (F) Agriculture 604078 88971 440,206 40970 82310 3141 79331 44538 Mining 57194 3405 5,186 355 51749 3033 60 7 Manufacturing 88978 56426 35,019 42573 52244 8149 1047 5365 Construction 81918 529 23,033 66 58084 449 359 11 N Utilities 1987 156 42 2 1937 153 2 1 S Commerce 49650 57212 37,385 50259 11668 5746 475 1062 Transport 54250 1722 18,626 92 34586 1620 336 3 Services 176315 118537 30247 5965 140847 111540 3277 537 Total 1114370 326958 589744 140282 433425 133831 84887 51524 Appendix Table A3.2c Industrial Distribution of the Labor Force by Employment Status BRAZIL Total Labor Force Self-Employed Employees yajily Workers 1960 (M) (F) (M) (F) (M) (F) (M) (F) Agriculture 10523225 1174573 5301878 264819 2774344 206461 2445994 703216 Mining } Manufacturing } 2813576 550656 399234 133303 2354167 372689 60020 44664 Utilities } Construction 1579694 668547 171187 47932 1358777 585386 17283 28167 Commerce 1345301 174745 681660 35351 634501 132404 29063 6990 Transport 1044163 44635 216468 464 824193 44171 3109 393 Services 1291204 1440944 668275 443932 606744 982017 15939 14995 Total 18597163 4054100 7438702 925801 8552726 2323128 2571408 798032 Total Labor Force Self-Employed Employees Family Workers 1980 (M) (F) (M) (F) (M) (F) ( (F) Agriculture 11376454 1732961 5500588 477928 4466557 563595 1384747 688230 Mining } Manufacturing} 5790196 1733687 520300 109592 5234292 1607243 24571 12072 Utilities } Construction 3095756 55338 855041 818 2228296 53942 5584 95 Commerce 2941586 1169721 1205397 293665 1701831 844863 29189 27471 Transport 1670831 144710 493931 1482 1168140 142549 5337 48 Services 5430825 6515945 1179744 1030288 4207943 5385286 30218 57037 Total 30305648 11352362 9755001 1913773 19007059 8597478 1479646 784953 Appendix Table A3.2d Industrial Distribution of the Labor Force by Employment Status COLOMBIA Total Labor Force Self-Emploved Employees Family Workers 1951 X (F) (M) (F) (F) (F) Agriculture 1930229 93052 778317 44795 818726 28277 262513 15642 Mining 45256 15967 8738 8144 32762 2052 2111 5574 Manufacturing 303667 157240 88108 82389 195704 62744 5439 8547 Construction 131058 1864 16894 238 104683 1339 549 12 Utilities 9957 515 358 17 9176 474 1 Commerce 158398 45376 94441 20960 54127 21019 3811 2470 Transport 123974 6109 21309 370 95781 5358 133 11 Services 235874 362219 33488 23903 169865 330611 691 1197 Total 2938413 682342 1041653 180816 1480824 451874 275248 33453 Total Labor Force Self-Employed Employees Family Workers 197? (M) (F) (M) (:)() (F) (M) (F) Agriculture 2311058 116001 975517 53631 984892 36446 343552 25715 Mining 61150 20129 12538 11520 44787 3144 3323 5429 Manufacturing 476643 179318 123700 75920 340945 90514 6793 11748 Construction 217234 3471 42720 168061 3428 1765 Utilities 12248 1028 594 11550 1017 24 Commerce 331862 108658 190627 44374 132289 57435 6603 6251 Transport 179485 12332 43753 626 131944 11497 1125 113 Services 374192 551754 57022 36214 312734 508525 Total 3963872 992691 1446471 222285 2127202 712006 363185 49256 Appendix Table A3.2e Industrial Distribution of the Labor Force by Employment Status ECUADOR Total Labor Force Self-Employed Employees Family Workers N 1950 (M) (F) (M) (F) (M) (F) (F) Agriculture 552062 88517 279538 7764 193596 76481 78,556 4237 W Mining 4760 199 1000 18 3683 176 76 5 Manufacturing 124958 108334 56825 17288 61675 85529 6,378 5149 Construction 25578 1224 2528 35 22639 1047 206 9 Utilities 1282 32 12 1 1270 31 Commerce 49467 25658 33511 4559 14938 20789 974 302 Transport 25985 1388 5914 105 19853 1282 187 Services 65525 75693 9211 3584 55456 70469 832 1614 Total 849617 301045 388539 33354 373110 255804 87209 11316 Total Labor Force Self-Employed Employees Family Workers 1982 (M) (F) (F) (N (F) (M) (F) Agriculture 727880 59092 387573 27646 217721 9857 88144 18118 Mining 6912 494 1930 140 4697 315 101 23 Manufacturing 214063 72467 73665 30642 127451 36025 5091 2646 Constntction 154683 3326 46300 458 96800 2654 2580 56 Utilities 11946 1237 510 23 11205 1197 18 Commerce 185127 86787 118533 51004 60063 31762 3304 2511 Transport 96345 4976 45703 280 44539 4594 1681 13 Services 373492 225539 66004 16160 260138 182812 7582 3287 Total 1770448 453918 740218 126353 822614 269216 108501 26654 Appendix Table A3.2f Industrial Distribution of the Labor Force by Employment Status JAMAICA Total Labor Force Self-Employed Emtloyees Family Worker 1953 (M) (F) (M) (F) (M) () () (F) Agriculture 225885 74402 Mining 7572 220 Manufacturing 40397 35097 Construction 18698 1458 Utilities 2490 270 Commerce 19266 33107 Transport 13143 2441 Services 26141 79093 Total 353592 226088 Total Labor Force Self-Employed Employees Familv Workers 1982 (M) (F) (M4) (F) (M) (F) (M) (F) Agriculture 113306 14402 Mining 4855 534 Manufacturing 43649 13373 Construction 20517 610 Utilities 3757 605 Commerce 23458 34495 Transport 15534 5041 Services 68910 93050 Total 293986 162110 - not available. Appendix Table A3.2g Industrial Distribution of the Labor Force by Employment Status MEXICO Total Labor Force Self-Employed Employees Family Workers 1960 (M) (F) (M) (F) () (F) M (F) Agriculture 5479642 663898 2560714 126119 2776360 520105 92645 8183 Mining 131928 9602 7291 941 124184 8555 109 15 Manufacturing 1306435 249656 217148 62661 1082953 185319 2161 618 Construction 394118 14161 47489 1106 345372 12954 299 13 Utilities 37542 3901 3616 356 33753 3501 34 6 Commerce 784837 290337 503252 157043 275578 130145 3565 2327 Transport 338105 18834 41470 779 295802 17957 92 15 Services 760161 767068 106014 36407 650993 724850 321 537 Total 9232768 2017457 3486994 385412 5584995 1603386 99226 11714 Total Labor Force Self-Employed Employees Family Workers (MI) (F) (MI) (F) (MI) (F) (Mv) (F) Agriculture 4958146 742714 2415701 313078 1195008 120435 469923 91903 Mining 353847 159492 85259 39013 219717 92557 16629 12294 Manufacturing 1898457 681742 303939 111792 1256134 427871 99834 46852 Construction 1093593 214174 193389 27785 689416 136808 46506 11809 Utilities 92144 24053 11490 2302 69825 18690 2262 726 Commerce 1151019 599509 475249 215457 434229 224852 66683 58954 Transport 605614 78026 125614 12302 367685 47154 22117 2570 Services 1490555 1372849 255009 169015 926701 796809 63078 250560 Total 11643375 3872559 3865650 890744 5158715 1865176 787032 475668 Appendix Table A3.2h Industrial Distribution of the Labor Force by Employment Status PERU Total Labor Force Self-Employed Employees Family Workers 1961 (M) (F) (Ni) (F) () (F) (M) (F) Agriculture 1340483 215077 721299 87922 431479 51557 187223 75430 Mining 64614 1799 1529 51 62899 1690 46 12 Manufacturing 294983 115997 103028 77501 187471 32455 4002 5824 Construction 103712 984 22312 138 80924 812 207 14 Utilities 8187 397 8072 317 115 80 Commerce 202998 78849 120929 50426 77794 24151 4029 4140 Transport 89385 4586 29416 220 59515 4342 261 11 Services 242368 234346 23789 17500 217922 215767 657 1079 Total 2346730 652035 1022302 233758 1126076 331091 196425 86510 Total Labor Force Self-Employed Employees Farnily Workers 1981 (l4) (F) (O) (F) ) (F) (M) (F) Agriculture 1596816 267192 1152794 111201 310430 31029 106226 117301 Mining 91532 5160 4852 219 85370 4736 80 26 Manufacturing 420366 136064 108659 64206 303146 59500 2301 8156 Construction 193503 3708 56241 395 132519 3177 442 24 N Utilities 16123 1641 369 12 15585 1613 4 1 Commerce 412081 220425 260140 153503 143907 55722 3480 7822 Transport 195818 14189 74831 825 114477 13130 544 30 Services 750268 439092 103497 19094 641515 416449 939 677 Total 3676507 1087471 1761383 349455 1746949 585356 114016 134037 Appendix Table A3.2i Industrial Distribution of the Labor Force by Employment Status VENEZUELA Total labor Force Self-Employed Employees Family Workers 1961 M (F) (M) (F) (M) (F) (M) Agriculture 733320 26465 395265 15238 243185 7767 93479 3201 Mining 50660 2880 2187 92 48185 2768 55 1 Manufacturing 212988 74356 50236 39850 160600 33784 1355 517 Construction 129581 1420 34602 155 94173 1243 389 13 Utilities 22083 1243 274 12 21750 1225 7 2 Commerce 260181 37277 133181 10649 122660 26083 3761 440 Transport 112738 4910 42514 211 69621 4681 141 4 Services 298097 250201 26994 29931 269044 218284 429 718 Total 1819648 398752 685253 96138 1029218 295835 99616 4896 3. Total Labor Force Self-Employed Employees Family Workers 1981 ) (F) (M) (F) (W (F) (NO (F) Agriculture 515484 19101 257436 6654 201968 9532 32948 1762 Mining 49231 5533 2802 77 43430 5162 16 3 Manufacturing 493542 145488 49129 19116 413313 117323 610 386 Construction 378216 21587 96917 895 259340 19313 766 29 Utilities 45015 8085 1434 49 40404 7529 18 3 Commerce 523668 208780 194363 52815 305475 143565 2610 2339 Transport 244965 31522 77705 1885 152090 27495 294 31 Services 698328 686682 75754 25629 580969 476491 744 1579 Total 2948449 1126778 755540 107120 1996989 806410 38006 6132 4 Potential Gains from the Elimination of Labor Market Differentials 1. Introduction In the previous chapter, employment differentials were examined in a statistical way rather than an economic one, that is, the patterns and trends of the gender employment distributions were compared independently of pay. Pay, however, is the prime signal for the efficient functioning of the labor market. Consider, for example, two industries that are identical in all respects except in the ratio of female to male pay. If some part of the pay differential is due to non- competitive practices,' then the efficiency loss associated with this pay differential would be higher in the industry where relative female pay is lower. In order to estimate the welfare loss arising from the differential treatment of women and men in the labor market, one has to consider both the dissimilarity in employment patterns of female and male workers as well as any associated difference in their respective pay. As there is no information about which part of the sex differential is justified, one can estimate the potential gain in efficiency and the effects on pay and employment only on the assumption that women and men are identical factors of production.2 One can then adjust the I A non-competitive outcome may be the result of pay and/or employment discrimination (by the employer, fellow employees, customers), protective or prohibiting legislation and social customs. 2 In the words of an author sympathetic to women's reluctance to undertake some manual tasks "... women ... seem more likely to see real dangers in some blue collar work that men have accepted ... [M]en ... have lived with these conditions for years, but women may not be as willing to put up with them. Some of these very problems may be keeping other women from even trying these jobs" (O'Farrell, 1982, p. 155). Obviously, 136 Women's Employment and Pay in Latin America results to the extent that these differences arise from different considerations by the individual suppliers of labor. 2. Theoretical Framework The aim is to show how women's under payment in, or limited access to, some sectors (such as industries) would result in lower levels of production, even if women kept supplying the same amount of labor as if there were neither wage nor employment discrimination in the first instance. Obviously, to the extent that some women are discouraged from joining the labor market when rewards, in terms of either pay or employment prospects, are lower than what they should be, the loss of output will be even greater than that suggested by the present analysis. The cost (welfare loss) of an unjustified pay or employment differential between working women and men can be shown in a simplified way as follows. First, assume that there are two industries employing two identical factors of production (women and men). For simplicity, assume further that (1) the two industries are identical in all respects relevant to the present study;3 (2) the total factor supplies are fixed (perfectly inelastic) but can vary between the two industries; and (3) there are neither costs of adjustment nor any non-pecuniary differences associated with employment in the two industries.4 Under competitive conditions there will be a common wage in both industries, and an efficient production and consumption will prevail. Though one can proceed by imposing either a wage or an employment differential, let us assume the latter so that women are now excluded from the first industry and the displaced female workers seek employment in the second industry. This will lower wages in the second industry as employment has risen, while wages in the first industry will increase as employment has been reduced. The "male" industry has become more capital intensive while the 'female' "keeping from trying" is a supply decision quite different from "keeping out from a job," which appears as a demand determined condition in the labor market. Note, however, that both demand and supply can be affected by what is going on in the other side of the labor market. 3 This refers to the demand for the final product of the two industries as well as to available technology and the level of the capital stock. 4 Alternatively, if there are costs of adjustment and non-pecuniary aspects of employment, then they shouldl be the same in the two industries. Potential Gains 137 industry is now more labor intensive. The labor market is characterized by both wage differentials and (most likely, partial) segregation.5 Though this result is predicted from a hypothetical experiment, it conforms to common observations about women's and men's employment and wage patterns in practically all countries in the world.6 An interesting question arises as output in the second industry has increased while output in the first industry has dropped: what will be the net effect upon the total product of the two industries? The answer depends on whether the gain in output in the industry which provides refuge to women workers is greater than the loss in output in the discriminating industry. Assuming marginal productivity conditions, the answer is that the net effect will be negative. This is because the contribution of additional labor to production in the second industry will, at the margin, be less than the increase in output in the first industry. Though this analysis is based on partial equilibrium, its extension to a general equilibrium framework is feasible (though more cumbersome) and the basic results hold.7 So far, it has been established that there is bound to be an efficiency loss (lower level of total production, GDP) from a restriction in the labor market based on non-economic criteria (such as the sex of the worker). Unjustified differentials do, therefore, provide grounds for corrective policies from an economic (Pareto) point of view. However, the acceptability of any intervention depends also on its distributional effects, and one further question, therefore, needs to be answered -- that is, 'who benefits and who loses' if an established differential is disturbed? This amounts to establishing who suffers the consequences of the employment restriction. Three groups are affected by such a restriction, though in different ways: female workers, male workers and employers. Workers are 5 Whether there is complete segregation will depend on the initial total factor supplies and the new size of employment in the two industries but, most likely, employment in the second industry will be mixed. 6 Of course, the observed differentials can be due to supply side decisions, a possibility which has been ruled out in the present paradigm by the assumption of homogeneity. However, supply side differences do not invalidate the present conclusions unless all differentials are due to supply factors - and this is not likely to be the case. 7 See Tzannatos (1987a). 138 Women's Employment and Pay in Latin America interested in wages while employers are concerned with profits.8 In the presence of discrimination of the type analyzed in this section, the only conclusion that has been demonstrated consistently in the literature is that women are always the losers.9 Therefore, the candidate beneficiaries are either male workers or employers or both. Whether male employees and/or employers gain from discrimination has been a contentious issue and opinion is still divided.'" No attempt can be made to solve this issue here. However, the results on welfare gains presented below, as well as the estimated effects upon male wages, can give some insight as to how strongly male workers (or male dominated unions) may react to measures promoting sex equality in the labor market. In short, what is shown is that, if women competed with men on an equal footing, men would have a smaller share of the pie, but the size of the pie will be greater as the allocative mechanisms of the market will improve. As a result, men's real pay need not suffer any adverse consequence in the longer run. These considerations can be shown diagrammatically in the following simplified way. In Figure 4.1 the left panel represents an industry which is initially exclusively male in terms of employment and the right panel represents an industry which is exclusively female. The labor supply of men and the labor supply of women are assumed to be equal and perfectly inelastic. The labor supply curves are shown in the respective panels of Figure 4.1 as the vertical lines MSm and FS'. The negatively sloped curves, D, represent labor demand (value of marginal product) assuming that men and women are equally productive and perfect substitutes. The model assumes that there are neither adjustment costs nor any non-pecuniary differences between employment in the a There are also employment effects which, for simplicity, are ignored in the present analysis. 9 For a review see Tzannatos (1990a). 10 According to the taste theory of discrimination, if observed differentials between workers are taken to reflect employers' preferences, the "remarkable agreement in the proposition that [employers] ... are the major beneficiaries of prejudice and discrimination ..." is dismissed as a "non-sequitur" (Becker, 1971, pp. 21-22). The response to this assertion has been "if this deduction is correct ... do whites in South Africa ... have lower standards of living as a result of their discrimination?" (Thurow, 1969, p. 112). Though employers may willingly pay a price for treating homogeneous labor in different ways, there is still considerable ambiguity about who gains. Some empirical studies have shown that the elimination of sex differentials will also benefit at least some male workers (Pike, 1982) though other studies have shown losses for workers belonging to the "majority" group (Bergmann, 1971; Tzannatos, 1987b, 1988). Potential Gains 139 two industries. Under competitive conditions a common equilibrium wage, wD, will prevail in the two industries. Let us introduced discrimination in the diagram in the form of either an arbitrary wage differential in favour of men (say, wm-wo) or an employment restriction in the male industry (say, from M to M*). As a result displaced workers from the male industry will seek employment in the female industry thus increasing employment (to F + [M-M*]) and lowering wages (to wf) in the latter industry. The welfare implications can be shown with the reference to the areas A, B and C. The reduction in employment in the male industry has reduced output by A+B+ C. The gain in output from the additional employment in the female industry has increased output by only A+B. There has been a deadweight loss equal to the area indicated by the rectangle C in the male industry. Thus, the elimination of direct wage discrimination or indirect employment discrimination should lead to welfare gains, increases in female wages and decreases in male wages. It should, however, be noted that the welfare gain resulting from the elimination of gender differentials and productivity gains in the longer run may imply that the effect upon male wages need not be significant. The latter is in fact found to be the case in the simulations which we have undertaken and which are presented below. For empirical purposes, one would require information about the level of female and male pay and employment by industry as well as production and product demand conditions." If these were known, one could work out the consequences of eliminating unjustified differentials and then compare the outcome to the prevailing one. As this type of information is unavailable, researchers have usually adopted an inverse strategy. They first evaluate the total product under the current conditions and then estimate the gain in output if employment differentials were eliminated, assuming factor homogeneity.'2 As in the case of employment dissimilarity and the Duncan index, this exercise provides long run upper bound estimates of the new level of output as well as female and male wages. " More precisely, to estimate the effects of the elimination of sectoral labor market differentials one needs to know (1) the labor demand and supply curves of women and men; (2) the currently available capital stock; (3) the nature of the production function; (4) the demand schedule for the final product(s); and (5) the elasticity of substitution between women and men. 12 See, among others, Bergmann (1971), Dougherty and Selowsky (1973), Pike (1982) and Tzannatos (1987b, 1988). Figure 4.1 Effects of Sex-differentials in the Labor Market Wa-ge Sin glm Wage s~ f Sf Wm We~~~~~~~~~W C ~~~~~eAN B N',D wfB N1D M.* M Employment F F+(M-W*) Emnployrment Industry 1: Men Industry 2: Women Potential Gains 141 3. Data and Results The data on pay and employment are the same as those used in the country studies in the companion volume. They refer to monthly earnings and persons employed by industry.'3 It is visualized that workers in each industry are employed in two distinct occupations, namely a high-pay one and a low-pay one. The male workers are assumed to be employed in the former occupation while women workers are assumed to be employed in the latter occupation. The aim is to examine the effects of allowing women to enter the high-pay occupation until wages are equalized within each industry.'4 There is no information about the state of technology and the ease of substitution in production between female and male workers. In the absence of information about the nature of technology in the countries studied, the production function is assumed to be of a constant elasticity of substitution (CES) type. In theory, a high value (tending to infinity) of the elasticity of substitution would be appropriate for estimating the upper bound effects of the present simulation, as one wants to examine what would happen if women and men were equal in all respects.'5 Though estimates for this limiting case are presented, more modest values of the elasticity of substitution have been also used in order to accommodate the fact that there is some heterogeneity between female and male labor. Consequently, we also experimented with values of the elasticity of substitution (sigma) equal to 3, 6, 9 and 12.16 Finally, the assumptions " In the case of Guatemala and Uruguay hourly wages are used. In addition, the sectoral distribution of employment and wages in Uruguay refers to occupations, not industries. The data for Ecuador were taken from Finn and Jusenius (1975). 14 Across industry differentials are allowed to persist as long as there are no within industry differences between women and men. Obviously, such differences should, at the limit, be eliminated as well, something which suggests that the empirical estimates of the present analysis understate the upper bound of welfare gains. Is In practice, the complete exodus of women from the low-pay occupation was achieved (in the present data sets) when the elasticity of substitution reached a value between 40 and 80. 16 There is wide agreement in the literature that aU pairwise elasticities of substitution are substantially greater than unity, ranging between three and nine (Bowles, 1970; Psacharopoulos and Hinchliffe, 1972; Dougherty, 1972; Hamermesh, 1986) and most studies have utilized the present range of values of the elasticity of substitution (Bergmann, 1971; Dougherty and Selowsky, 1973; Pike, 1982; Tzannatos 1987b, 1988). 142 Women's Employment and Pay in Latin America Table 4.1 Results of the Within Industry Elimination of Occupational Differentials Percentage Change in Country Year Female Male Labor Wage Wage GDP Force, (1) (2) (3) (4) Argentina 1987 38.2 -8.9 4.0 25.3 Brazil 1980 96.6 -7.7 8.7 23.1 Chile 1987 40.5 -5.6 3.3 17.6 Colombia 1988 46.4 -7.6 4.6 20.2 Costa Rica 1989 34.8 -6.3 3.0 17.5 Ecuador 1966 58.6 -13.3 9.7 37.2 Guatemala 1989 25.2 -5.5 2.0 13.7 Jamaica 1989 60.7 -8.2 8.3 27.5 Uruguay 1989 29.6 -7.6 3.4 16.3 Venezuela 1987 23.6 -6.2 1.9 11.7 a. Percent of total labor force who would have to change occupation underlying the present calculations are those implied by the competitive labor market model.17 Table 4.1 presents the results for the countries under consideration when the elasticity of substitution is assumed to be equal to three (the complete set of results is presented in Appendix Table A4. 1). The first column indicates the average increase in the low-pay occupation (female earnings) as fewer workers will be employed in it. The second column provides the percentage reduction in the average wage in the high-paying occupation (male earnings) which would 17 These assumptions are: (1) that all workers, independently of their sex, are paid wages equal to the respective value of their marginal products before and after equality; (2) that adjustment is costless; (3) that the only reward to workers (cost to employers) associated with employment is the wage received (paid); and (4) that the capital stock (more importantly, utilization) is fixed. Though these assumptions are rather restrictive, especially if one wants to evaluate the labor market as a whole, they need not necessarily be so for comparative purposes. For example, even if wages are not equal to the value of their respective marginal products (one of the most argumentative issues in this area of research), this does not produce unacceptable results in the present analysis so long as the discrepancy between marginal products and wages is proportionately the same for women and men. Potential Gains 143 result from the increase in (female) employment in that occupation. The third column shows the welfare gain which would follow the influx of women into the high-pay sector in each industry. The fourth column shows the percentage of the labor force that needs to change occupation within industry for all these effects to be achieved. The findings can be summarized as follows. First, the new level of wages in each industry suggests that the decline in male wages would, in general terms, be small and the increase in female wages large. Second, the net change in output which would result from the above changes in (real) wages would be positive and sizeable (unweighted average about 5 percent) even for low values of the elasticity of substitution (equal to 3 in Table 4.1). Third, around one-fifth of the total labor force would have to change occupations within industries for all these effects to take place (unweighted average 22 percent). Finally, and taking into account the complete set of results included in Appendix Table A4. 1, the elasticity of substitution is a key parameter for the value of the results but is not so important as to render the simulation a wholly hypothetical exercise. Particularly striking is the fact that, even at the lowest levels of the elasticity of substitution, the effects are more than suggestive. The welfare gain will be higher, the reduction in male wages lower, the increase in female wages higher, and the percentage reallocation of the labor force higher, the easier it is to substitute women for men as was expected. What is also interesting is that these effects are not proportional to the change in the value of the elasticity of substitution. As a summary statement, one may say that raising the value of the elasticity from 3 to its maximum value increases the gain in output by only 3 times, reduces the decline of male wages from approximately 10 percent to around 1 percent, and increases the rise in female wages by no more than 50 percent. The associated increase in the percentage of the labor force who should change occupations to achieve equality in wages within each industry rises from approximately 20 to 40 percent. In fact, in most cases the changes suggested from the maximum value of the elasticity of substitution are only marginally higher than those suggested by an elasticity of substitution equal to 9 and are practically indistinguishable from those derived when the value of the elasticity of substitution was set to 12. The conclusion, therefore, is that even if most of the observed differentials were due to supply factors, and even if the easiness of substitution were not great, the gains in output and the rise in female wages could be sizeable at a trivial cost to male workers. One should also note that the number of reallocations required to achieve wage parity between women and men are greater than the 144 Women's Employment and Pay in Latin America Table 4.2 Percentage Change in Female Wages If Women Had Country Men's Wage Men's Employment Distribution Distribution Argentina 58.1 4.3 Brazil 129.4 -14.9 Chile 61.4 7.2 Colombia 63.3 4.7 Costa Rica 50.7 -18.0 Ecuador 103.0 9.0 Guatemala 39.3 -17.1 Jamaica 101.9 -3.9 Uruguay 50.5 -5.9 Venezuela 24.3 -15.4 Source: See text. reallocations required to reduce the value of the Duncan index to zero. A straightforward implication of this is that employment dissimilarity as such is less damaging, in allocative terms, than wage differences. Expressed in a different way, policies which enhance the pricing mechanism of the labor market and let individuals adjust accordingly may have a more direct effect on female pay and women's position in the labor market than policies aiming at eliminating employment differentials. One such policy could be the introduction of employment quotas. We can examine the last assertion in another simulation exercise.18 Let us assume that women are employed in the same sectors as at present but are paid male wages. Alternatively, let us assume that women are paid as at present but their employment is distributed across sectors as the employment of men. The results of these two simulations are presented in Table 4.2. Column 1 shows the percentage increase in female wages, if women were paid as men in their current employment. Column 2 shows the corresponding change in female wages, if women had the male employment distribution but were still paid women's wages. The difference between these two results is impressive but is We use the same data on pay and employment as before, that is, industrial employment and wages. Potential Gains 145 comparable to that reported for other countries, both industrialized and developing ones.19 In short, the employment distribution of women has a modest effect on female pay. Much of the difference between female and male pay should be sought in women's low relative pay in the sectors in which they are employed.' 4. Qualifications and Discussion The results of the previous analysis would be valid, if the underlying assumptions were true. This is not the case. However, the non-fulfillment of some of the assumptions biases the estimated effects upward while the non- fulfillment of others biases them downward. Therefore, the net effect of relaxing the non-realistic assumptions cannot be determined using available evidence. Only the direction of bias from selected factors is indicated below and precise estimates must await further research. The results were based on the assumption that women and men have identical labor supply functions. However, empirical studies have documented that this is not true. In particular, women's responses to changing wages have been found to be highly elastic2' while most studies imply that the male supply is totally or nearly perfectly inelastic.' Knowing that women's labor supply responds elastically to labor market conditions (so more women will be tempted to join the labor market when barriers are eliminated) and that the male labor supply 19 See Brown, Moon and Zoloth (1980a) for the United States, and Chiplin and Sloane (1976) and Miller (1987) for Britain. Similar evidence for developing countries is presented in the collection of studies in Birdsall and Sabot (1991). 20 Of course, our results may reflect to some extent the fact that broad employment categories have been used. 21 For a survey see Killingsworth and Heckman (1986) or the wide range of individual country estimates (which include the United States, Britain, France, Spain, Germany, Netherlands, Sweden, Italy, Australia, and Japan) provided in the Journal of Labor Economics, Vol. 3, No. 1, Part 2 (Special Issue), January. 2 For surveys see Pencavel (1986) or Fallon and Verry (1988). 146 Women's Employment and Pay in Latin America will not react adversely to a decrease in their wages, one is confident that the welfare gain should be even higher than that suggested by the present results.' Women's occupational choice, a key factor in the present analysis, is seen as more favorable to work which is compatible with family matters24 and where productive skills do not depreciate much because of labor market interruptions (atrophy).' In addition, biological or attitudinal differences may also be important for some jobs.2' In this respect, our estimates are upwardly biased. However, the following qualification may neutralize, or even reverse, this bias. One has in mind here the perennial chicken and egg question. That is the extent to which what appears to be a free occupational choice is not the result of a rational labor supply decision under labor demand constraints.' Consequently, the assumption of factor homogeneity may initially bias the estimates for efficiency gains upward but the (unavoidable) failure to account for the response of the female labor supply to the demand conditions" may well correct for this bias. There are many refinements that could be mentioned in this context although little can be done in the context of the present study. Given the lack of informative data, this section is concluded with one more observation. The 3 Of course, the rise in female labor supply will eventually result in lower levels of average wages but this effect has not been found to be important in practice (Pike, 1982). In any case, the equalization of the occupational distribution of the sexes would be a continuous process over a number of years and, given the increase in labor productivity due to improvements in technology, the level of male real wages need not suffer at all. 24 Easterlin (1968); Lehrer and Stokes (1985). 25 Mincer and Polachek (1974); Mincer and Ofek (1974); Polachek (1975). I For example, in Britain shortly after World War I it was estimated that women's productivity was about 10 to 15 percent less than men's in tasks involving heavy manual work. See Atkin Report (War Cabinet Committee on the Employment of Women in Industry), London: HMSO, 1919. 27 Gronau (1982). n The constraints attributed to the labor demand side can well be the result of legislation (Kanowitz, 1969), informal mechanisms (Bernard, 1971), ideological/cultural factors (Williams and Best, 1982), men's interest in maintaining their privileges (Goode, 1982), union practices (Rubery, 1988), as well as consumer preferences (Becker, 1971). Potential Gains 147 simulations assumed that women's and men's pay and employment characteris- tics are equalized within but not across industries. This eclecticism was dictated by the fact that it is even harder to incorporate movements in employment across industries.' Even so, one is quite sure about the direction of bias from this omission: it has already been established in the previous chapter that industrial dissimilarity is almost as significant as occupational dissimilarity and, had one allowed for across industry equalization of employment, the gains would have been higher. 5. Conclusions One always faces a risk when trying to incorporate all eventualities in a rather simple economic framework -- as the one employed in this chapter.2' Yet a number of conclusions may be relevant. In particular, even if only a small part of the observed sex differential in the labor market is due to demand factors or some (unspecified by the present analysis) imperfection, its elimination may result in sizeable gains in output (efficiency) which would in turn improve women's labor earnings (poverty). All this becomes possible from the more efficient use of any country's most abundant type of human capital. As women are disproportionately found in the low income segment of the population, the distributional effects are guaranteed. And, as real wages tend to rise with growth, the real wages of men need never suffer an actual reduction. The (albeit tentative) nature of the present findings suggests that a rewarding direction of research would be the exploration of factors which cause the observed labor market differentials between women and men in the region. In a broader social cost-benefit framework, imperfections arising from causes other than legislation or labor market regulation are difficult to identify, even more so to quantify their effects. Hence, a relatively easy policy option may be to revise statutes which differentiate between the sexes and force individuals to accept one stereotype or another, either in the family or in the labor market. 29 One requires a substantial number of additional assumptions for this case such as cross-industrial (as well as occupational) elasticities of substitution. 30 Mason (1984, p. 157) notes that economists have given more attention to the consequences of female employment aggregates and occupational sex segregation than to their causes. However, this practice is changing as the study of Gronau (1982) and references therein indicate. 148 Women's Employment and Pay in Latin America Statistical Appendix to Chapter 4 The table presents the simulation results of assuming that women are equally productive and have the same occupational distribution as men within each industry under different values of the elasticity of substitution. Appendix Table A4.1 Results of the Within Industry Elimination of Occupational Differentials Percentage change in Country/ Elas. of Female Male Labor Year Subst. wage wage GDP force (sigma) (1) (2) (3) (4) Argentina 3 38.2 -8.9 4.0 25.3 1987 6 45.0 -5.7 8.3 35.8 9 48.8 -4.0 10.8 39.3 12 51.1 -3.0 12.3 40.4 max 56.4 -0.7 16.2 40.9 Brazil 3 96.6 -7.7 8.7 23.1 1980 6 112.0 -3.7 15.2 27.8 9 118.0 -2.3 18.0 28.5 12 121.0 -1.7 19.5 28.7 max 126.6 -0.5 22.6 28.8 Chile 3 40.5 -5.6 3.3 17.6 1989 6 48.0 -3.3 6.7 24.1 9 52.0 -2.2 8.6 25.9 12 54.3 -1.7 9.7 26.5 max 60.0 -0.3 12.6 25.9 Colombia 3 46.4 -7.6 4.6 20.2 1978 6 51.8 -4.5 8.4 29.8 9 54.8 -3.1 10.4 34.5 12 56.7 -2.3 11.6 36.5 max 61.3 -0.7 14.7 38.2 (Continued) Potential Gains 149 Appendix Table A4.1 (Cont.) Results of the Within Industry Elimination of Occupational Differentials Percentage chan2e in Country/ Elas. of Female Male Labor Year Subst. Wage Wage GDP Forcea (sigma) (1) (2) (3) (4) Costa Rica 3 38.8 -6.3 3.0 17.5 1978 6 40.1. -4.0 6.0 24.2 9 43.0 -2.9 7.7 26.7 12 44.7 - 2.2 8.8 27.9 max 49.3 -0.5 11.8 29.1 Ecuador 3 58.6 -13.3 9.7 37.2 1966 6 76.4 -7.3 19.7 51.8 9 85.1 -4.7 25.1 54.3 12 89.7 -3.4 28.1 54.7 max 97.3 -1.4 33.4 54.9 Guatemala 3 25.2 -5.5 2.0 13.7 1989 6 29.0 -3.7 4.3 20.4 9 31.5 -2.6 5.9 22.8 12 33.0 -1.7 7.0 23.1 max 38.1 2.4 11.8 21.9 Jamaica 3 60.7 -8.2 8.3 27.5 1987 6 77.4 -4.4 15.7 35.3 9 84.7 -2.9 19.4 38.1 12 88.7 -2.2 21.4 39.5 max 94.0 -1.2 24.3 40.6 Uruguay 3 29.6 -7.6 3.4 16.3 1989 6 34.7 -4.9 6.9 24.2 9 37.8 -3.4 9.0 29.2 12 39.9 -2.6 10.4 33.3 max 48.1 -0.5 15.2 40.3 Venezuela 3 23.6 -6.2 1.9 11.7 1987 6 26.7 -4.4 4.1 21.5 9 29.0 -3.3 5.6 27.1 12 30.6 -2.6 6.6 30.1 max 36.4 -0.4 10.3 32.9 a. Percent of total labor force who would have to change occupation Source: See sources to Table 4.2. 5 Gender Differences in the Labor Market: Analytical Issues 1. Measuring and Interpreting Gender Wage Differentials Female labor is, on average, rewarded less than male labor. This is, and has been, true for all countries for which data exist, irrespective of the mechanism used in allocating economic resources (from market to plan) or the developmental stage of the country (from basic agrarian to mature industrialist). What is less common is agreement on the reasons giving rise to this. In previous studies a sizeable part of the pay gap remains unexplained even after accounting for differences between the employment characteristics of women and men (such as occupation and industry, whether they are employed in the public or private sector, or in rural, urban or metropolitan areas) and the personal characteristics of workers (such as education, labor market experience and unionization).' From an economist's point of view, any observed difference in pay can be seen as the interaction of labor demand and labor supply schedules that differ between women and men. In other words, if women and men are identical factors of production, women's inferior position in the labor market may be the result of 1 For a survey see Cain (1986). One author's summary statement of the empirical findings attributes up to two-thirds of the pay gap to differences in workers' characteristics (Killingsworth, 1990, p. 57) though this seems to be a rather high figure (Duncan, 1984, Chapter 1, Table 1). However, these studies relate primarily to advanced economies and it is probable that different considerations apply to countries characterized by large informal sectors and excess supply of labor. In fact, the present results are more in line with earlier studies in developed countries where differences in characteristics appeared to account for only a small part of the total wage gap between women and men (Oaxaca, 1987). 152 Women's Employment and Pay in Latin America some form of discrimination against them in the workplace (labor demand side).2 Alternatively, if the pricing and allocation of women's work in the labor market is gender blind, the observed differences may be the result of women's own choices and efforts to compromise their willingness to work with domestic activities (labor supply side).3 However, even if it were possible to determine that the initial reason for a particular differential was more on the supply than the demand side, or vice versa, the observed differential would most likely be a mixture of the two -- as individuals attempt to maximize their welfare and employers their profits subject to their own constraints and the constraints imposed by the behavior of other parties. For example, if women perceive it as unlikely that they will end up in a senior position, they will underinvest in education as rewards will be lower compared to men's. In this respect, it would be wrong to argue that women's low achievement in the labor market is the result of their free choice to underinvest in their human capital. Alternatively, if employers expect women to drop out of the labor market during family formation, they will underinvest in women's training as the expected benefits will be lower than in the case of men. In this respect, employers can be seen as rational agents (rather than discriminators) who balance their decisions between different current costs and expected returns.4 Given the difficulty in isolating which part of the female/male wage differential is due to demand reasons and which part is due to supply reasons, what one hopes to do is to 2 One should note that what appears as discriminatory practice in the labor market need not necessarily be, wholly or in part, the result of employers' practices. Fellow employees' attitudes ("men do not like to be led by women"), unions' practices ("most union members are men"), customers' preferences ("surveys have shown that air travellers prefer female hosts"), government regulations (protecting or prohibiting) and social norms ("a woman's place is at home") can all contribute to different treatment of women and men in the labor market. See Boserup (1970), Cain et al. (1979), Easterlin (1968), England (1982), Hofstede (1980) and Lehrer and Stokes (1985). 3 Of course, observed differentials in the labor market can be due to differences in productivity between women and men as well as difference in "fixed costs" associated with the decision to supply/demand labor in the market (Oi, 1962; Nickell, 1978; Cogan, 1980a). 4 This behavior of employers has been labelled "statistical discrimination" (Phelps, 1972; Thurow, 1975; Aigner and Cain, 1977). Analytical Issues 153 differentiate between the systematic relationship between earnings and workers' productive characteristics in general and the "gender effect' upon earnings.5 For practical purposes one can bypass the long list of theoretical arguments and address the issue in a way susceptible to empirical investigation. There are two different approaches. First, one can examine whether there is afixed premium/ disadvantage associated with the sex of the worker. Second, one can investigate whether individual characteristics of female workers are rewarded differently in the labor market than the corresponding characteristics of men. The former approach relates to a "shift" in the earnings function and the latter to a "difference in the slope coefficients" of the earnings function. The first approach consists of running a regression of earnings upon the characteristics of all (male and female) workers including a separate variable which indicates the sex of the worker.6 This can be shown as follows: In(W) = C + (X1)a + b(F;) + e, (5.1) where In(W;) is the logarithm of the ith worker's pay,7 C is a constant term, X is a vector denoting whatever measurable personal characteristics of relevance are utilized by the researcher, a is the vector of the estimated coefficients/effects of these characteristics upon pay, F is a (dummy) variable taking the value of 1 if the worker is female and 0 if the worker is male, and e refers to unobserved or unmeasurable characteristics.8 The interpretation of equation 5.1 is that individual earnings depend on the worker's observed characteristics (X's), the s Economists have conventionally referred to the sex effect as the "upper bound of discrimination," irrespective of origin, or "the extent of our ignorance" (Sloane, 1985; Siebert, 1990). 6 See Beller (1984), Fallon and Verry (1988, Chapter 5) or Killingsworth (1990, Chapter 3). For applications and extensions of this approach to measuring wage differentials in other areas of research see Smith (1977), Oswald (1985) and Ehrenberg and Schwarz (1986). ' The logarithm of earnings, rather than the level of earnings as such, is considered to be the appropriate regressand both on theoretical grounds (Mincer, 1974) and also on empirical grounds (Dougherty and Jimenez, 1991). 8 The error tern is assumed to be normally distributed with zero mean. 154 Women's Employment and Pay in Latin America worker's sex (F), and unobserved characteristics (the error term) assuming that e is not correlated to F at given X.9 The coefficient of interest is that on the variable representing the sex of the worker, which shows whether women receive on average lower pay than men (b <0) other things being equal (after adjusting for whatever the X's account for). This approach constrains, however, the values of the coefficients on the other explanatory variables, such as education and experience, to be the same for women and men. Given that sex specific earnings functions have produced coefficients on female characteristics that are significantly different than those for men,"0 a finding confirmed also by the present studies, this approach is bound to yield, in general, biased results. The second approach consists of running two regressions separately on women's earnings and men's earnings and comparing the outcome. This method requires the two regressions to have a strictly comparable specification, that is, the number and type of variables should be the same in both the female and male earnings functions. Thus the estimation can start with the following two regressions (omitting subscripts for notational simplicity): ln(1W=) = Cm~ + (X.)m + em, (5 .2) ln(W) = Cf + (Xf)f + ef (5.3) where C, (s= male or female) is the constant term, X, is the vector of male or female characteristics, m and f are the respective coefficients on these characteristics, and e, is the error term. Then, the 'adjusted' pay gap can be estimated in the following way: the difference in the average logarithms of male and female pay [ln(W,)-ln(Wf) - no subscripts] can be shown'" to be equal to the percentage difference of male to female average pay (W. and Wf): 9 If the error term is negatively correlated to F, then the coefficient on discrimination will be biased upwards as women will possess fewer unobservables than men with the same X's. This bias arises because the characteristics which are unobserved and affect women negatively will register an effect via the coefficient on the dummy variable measuring sex in addition to the pure effect of sex upon pay. 10 Psacharopoulos (1985); Tilak (1987); Sahn and Alderman (1988); Schultz (1989b); Bustillo (1989). "1 Oaxaca (1973). Anatytical Issues 155 fn(Wm) - ln(WY = Ifn(l +(Wmt-W)IWJ (5.4) = (Wm-WYWf Given the previous two equations and utilizing the regression property that the error term has a mean value of zero, one can rewrite the right hand side of equation 5.4 as: ln(W.) - ln(W) = (Cm - Cf) + [(Xm)m - (Cf)fl (5-5) where the first bracket refers to the respective constant terms in the male and female earnings functions, and Xm and Xf are the average values of the male and female characteristics in the sample. Adding to and subtracting from equation 5.5 the term (Xf)m or (X,)f and rearranging produces the following two 'decompositions' of the gross differential in average pay: In(W,) - In(Wf = [(Cm-CQ+(Xf)(m-oI + [(X.-Xf)m] or (5.6) = [(Cm Cf)+(X.)(m-f)I + [(Xm-Xf)fl (5.7) Thus, the percentage difference in pay can be seen to come from two different sources. First, the differential rewards to male and female characteristics (m-f) in the labor market including the difference between the constant terms and, second, the differences in the quantities of these characteristics held by men and women (X,-Xf). In this approach, the portion of the wage gap due to differences between the endowments of productive characteristics held by women and men cn be considered to be nondiscriminatory (or 'justified" discrimination). 12 On the other hand, the portion of the wage gap which is due to differences in the values of the coefficients, including the constant term, can be thought of as the upper bound of ("unjustified") discrimination. Obviously, this approach (equations 5.6 and 5.7), which utilizes two separate earnings functions, encompasses the previous one (equation S.1) which is based on a single regression and examines, in effect, only the difference in the constant terms. This explains the popularity of the decomposition based on separate earnings functions for women and men in applied research.'3 This is the approach followed here. 12 Blinder (1974). 13 It should be noted that, in practice, the two approaches (equation 5.1 and equations 5.6 or 5.7) may yield similar results as the constrained single equation estimation is, in effect, a matrix-weighted average of the results produced by the two equation method (see also Killingsworth, 1990, p. 96). 156 Women's Employment and Pay in Latin America One should note that equations 5.6 and 5.7 do not produce the same results. The former decomposition evaluates the justified and potentially discriminatory components of the pay gap if women were paid as men. The latter decomposition assumes that men are paid like women. This is a common problem with index numbers and is shown in Figure 5.1." The horizontal axis measures education (schooling in years) which can be considered a typical individual characteristic. The vertical axis measures wages. The lower line represents the earnings function for women, that is, it shows that female wages increase by f (the slope of the line) for an additional year of schooling. The upper line is the earnings function for men and m is the corresponding slope coefficient. Let Sf and Sm be the average level of schooling attained by women and men respectively. The way the diagram is drawn suggests that women have on average lower wages than men (Wf < Wm) because (1) they are less educated (S, YVU) -- given that both r and S are positive. In addition, this simple model provides a convenient (semi- logarithmic) relationship between annual earnings and length of schooling (in years) which is capable of being used in econometric work. In particular, one can run a regression of the log earnings of individual workers upon a constant term and their length of schooling. In this context, the constant term should be approximately equal to the logarithm of the wage for non-educated labor while the coefficient on schooling should be roughly equal to the rate of interest. In fact, and as the underlying assumptions necessitate, in a perfectly competitive environment the market rate of interest should be equal to the rate of time preference in the society and also to the rate of return to an additional year of schooling. The merit of this approach is that it is derived from an explicit economic foundation that dictates the functional form (log annual earnings on years of schooling) that econometric estimation should take. One disadvantage of the "earnings function" derived as above is that it ignores all other aspects of human capital formation and especially on-the-job training. In fact, in many jobs 162 Women's Employment and Pay in Latin America employer provided training or simply experience may be more significant forms of human capital and more important determinants of earnings than formally acquired education. The flat age earnings profile depicted in Figure 5.2 contradicts the typically observed age earnings profiles. In particular, age- earnings profiles increase at first, reaching a peak sometime in the working life of an individual until they eventually flatten out or even decrease (after the age of, say, 50 or 55 years). This pattern is suggestive of the fact that other types of human capital are formed during a person's working life. These observations can be accommodated by including into the model some proxy for post-education human capital formation on the right hand side of the last equation. The most popular version of earnings functions is by far the following one: ln(W") = Constant + rS + aE + bE2 where E (E2) is years of post-school work experience (and its square) and a (b) are coefficients. In theoretical terms the quadratic (in experience) formulation of earnings functions can be derived by assuming that investment in human capital declines linearly with time."8 In practical terms, the inclusion of experience and its square as an explanatory variable has been found to 'belong" to the earnings functions in the sense that regression analysis returns coefficients on experience which have the correct sign, are statistically significant and are 'intuitively' of reasonable size. The earnings functions thus specified have proven to be the most stable econometric relationship in the area of applied economics.'9 They have been estimated for practically all countries for which individual (cross-section) data exist. Their popularity rests partly on the very few variables that are required, namely education and experience. The former variable is usually available from many sources (labor force surveys or household surveys). The latter variable, namely labor market experience is relatively easily proxied, at least in the case of men, as follows: Experience = Age - Schooling - (school entry age) as most men are in the labor force after the completion of their formal education and throughout their prime age. Experience thus calculated is in effect potential experience (or, as referred to by some authors, Mincerian experience). is The proof can be found in many labor economics textbooks such as Fallon and Verry (1988), p.149-150. 19 Griliches (1977). Analytical Issues 163 A number of issues have been raised with respect to the use of Mincerian functions and the previously described decomposition in the attempt to identify the gender effect on pay. The arguments relate to whether the parsimonious formulation of earnings functions is sufficient and appropriate for the task in hand on both theoretical and empirical grounds. We do not dwell on the former as it is beyond the scope of this study and the relevant literature is rich.Yo With respect to the latter, applied research may encounter measurement and omitted variables errors as well as endogeneity, selectivity and specification problems. Such effects have a complex imapact on the decomposition results whose interpretation may not be immediately obvious. These aspects are examined below. 3. Errors of Measurement and Omitted Variables Education. Errors of measurement and omitted variables problems are both common and interrelated issues in the earnings functions approach. For example, the effect of formally acquired human capital is captured by the coefficient on the education variable, the latter being usually measured as years of schooling. The shortcomings associated with this approximation are, first, that it assumes an extra year of schooling augments earnings irrespective of whether it was acquired at an elite or deprived school, or whether it is added to 2 or 12 years of schooling, or whether education relates to studies in arts, social sciences or engineering.2" Second, there is also evidence that parents usually invest in higher quality education for boys than girls.2Y Hence, the nature of the schooling variable in the men's regression may be different from that in the women's regression. Third, formal schooling is not the only type of education 2 For the theoretical foundation and the common form of the earnings function see Mincer (1974) and Griliches (1977) and for surveys on the debate which has followed see Siebert (1985) and Wilis (1986). 21 Welch (1966); Behrman and BirdsaU (1983); Birdsal and Behrman (1983). 2 Becker and Tomes (1979) suggest that parents could distribute investment among their children with different characteristics differently and empirical evidence from developing countries confirms that, if such differential treatment is practiced (in areas such as health care, nutrition or expenditure on education), it favors boys (see Visaria, 1971, on India; Chen el aL, 1981, on Bangladesh; Aird, 1984, on China; Blau, 1984, on Nicaragua; Martorell et al., 1984, on Nepal; Amin and Pebley, 1987, and Schultz, 1982, on India; and Bardhan, 1984, for an overview). 164 Women's Employment and Pay in Latin America one can have; nonformal education is also important.? Consequently, education may not be measured precisely while its coefficient may be affected by the absence of variables (omitted/unobservable) which relate to the quality of education. The only way we could have taken into account some of the differences between women's and men's type and quality of education was by separating, on the one hand, general from vocational secondary education and, on the other hand, 'soft' and "hard' subjects at university level (such as arts and social sciences versus engineering and science). However, even in the seven countries for which such information existed, the number of identifiable cases was minute for women, especially in terms of vocational education.2' When a further breakdown was attempted (for example, by region or age) the number of useful cases dropped to single figures -- no more than two observations in some countries. The paucity of information in this respect is at the same time reassuring that the bias in our estimates may not be significant: too few women opt for vocational education and their enrollment in tertiary education is still low. In fact, one in ten of all females aged 15 to 24 years in the region were illiterate in the early 1980s and as many as one in four of those aged 35 to 44 years.' Given that specialized studies do not generally start before the beginning of upper secondary education, men and women workers in the region should have more homogeneous education than their counterparts in advanced countries. Experience for men. Similar considerations apply to the other human capital variable -- "experience" -- typically included in the earnings functions. Information about actual work history rarely exists in data sets. The usual strategy is to use potential experience (age minus schooling minus school entry age) and its square as right hand variables in the earnings functions in order to proxy the effect of the informal acquisition of human capital.26 The experience 2 Though the distinction between formal and nonformal education is far from clear (La Belle, 1986), nonformal education can be defined as the knowledge acquired outside the conventional primary/secondary/tertiary education system (Coombs, 1968). 24 The countries for which such information exists are Argentina, Colombia, Guatemala, Honduras, Panama, Uruguay, and Venezuela. 25 UNESCO (1990). 2 The inclusion of experience thus specified has been proposed on theoretical grounds (Mincer, 1974) and though this specification and its interpretation has been the subject of debate (Psacharopoulos and Layard, 1979; Griliches, 1977) it is still considered to be an appropriate simplification for estimating the returns to human capital Analytical Issues 165 variable so constructed has been routinely assumed to be a good proxy for the labor force experience of men. This might have been the case when the first earnings functions were tried using data sets in the 1960s in advanced countries following the post-war economic prosperity.' However, unemployment has increased considerably from the mid-1970s onward and the assumption of continuous work record is less accurate today than before.' In addition, neither the incidence nor the duration of unemployment is distributed equally among the various groups of workers. There are some workers who do not, in practice, experience unemployment (in tenured posts) while others may have an unemployment spell only rarely (the more educated and the more skilled). The bulk of those responsible for a 5 or 10 percent unemployment rate come from the lower end of the employment distribution, the less qualified workers.' As a consequence, the regression result will be inefficient and biased downward since potential experience in the sample will be greater than actual experience. Experience for women. In the case of women, the use of potential rather than actual experience is even more problematic. Women generally leave the labor market during their. family cycle and, when reliable data exist, their unemployment rates are usually higher than those of men. Hence, the measurement error is more serious and a decomposition of the pay gap based on potential experience would suggest a higher estimate for discrimination than the actual one. We cannot answer the question how the inclusion of potential rather than actual experience affects the estimates in this volume because there is no information on actual experience in the current data sets. However, the evidence from a few studies addressing this issue using cross-section or longitudinal data (Willis and Rosen, 1979; Heckman and Hotz, 1986). 7 Mincer's original application related to a sample of prime age men in 1959 - the unemployment problem was negligible in the United States during that period. 28 A recent study for Britain (Main and Elias, 1987) utilizing data from the National Training Survey 1976 shows even at that time, when the unemployment was less than half its level today, as many as 20 percent of the men in the sample did not have continuous work histories. 29 A comprehensive study of the incidence and duration of unemployment among male workers of different characteristics was performed by Nickell (1980) who confirms the broad patterns of the duration and incidence of unemployment described in the text. 166 Women's Employment and Pay in Latin America suggests that the estimated effect of potential experience upon female pay may be as low as 50 percent of the effect of actual experience.' Proxied and unobserved human capital characteristics. The standard human capital variables (schooling and experience) may serve only as proxies for other unobserved individual characteristics and education can be seen as a signal for these unobserved characteristics.31 In this context, if men have higher mean values of the proxy variables, the regression result for discrimination will be biased downward.32 In addition, education and experience are not the only attributes rewarded in the labor market. There are also individual differences stemming from innate ability, motivation -- perhaps related to different family circumstances -- and formal and informal training. These are unmeasured or unmeasurable aspects whose exclusion from the earnings function will again produce biased results.33 Endogeneity. Even if information on actual labor market experience for women were available, one could not use it as such in the earnings equation. The reason for this is endogeneity. Experience is nothing more than a measure of "accumulated" participation, and participation depends on pay. For men, this is not perhaps a very important issue as the inelasticity of male labor supply to wages can be interpreted as an indication of the fact that men participate and 30 Malkiel and Malkiel (1973), Mincer and Polachek (1974), Zabalza and Tzannatos (1985) and Miller (1987). These studies conclude that the earnings increment associated with actual labor market experience among females is comparable to that received by male workers. In fact one study (Levine and Moock, 1984) found that almost half of the wage gap between husbands and wives could be attributed to the fluctuations in the labor force attachment of wives, especially to depreciation of female skifls during interruptions in employment. However, their sample was drawn from a suburb of New York City in the late 1980s and may not be a representative one for other areas even within the United States. Wright and Ermisch (1991) estimated that the inclusion of actual experience in the earnings functions of women increases the part of the gender wage gap in Britain attributed to differences in endowments from 12 percent to only 17 percent - the latter being the estimate derived from using potential experience. Thus, in the present context one may argue that bias arising from proxying actual experience by potentia:l experience may not be that great. 31 Berg (1971); Arrow (1972). 32 Hashirnoto and Kochin (1980); Roberts (1980). 33 Unless, of course, the unobserved characteristics of relevance are distributed randomly among individuals - not a very realistic scenario (Polachek, 1975). Analytical Issues 167 accumulate experience independently of their earnings.3 However, for women, especially married women, the elasticity of labor supply to their prospective pay has generally been found to be positive and sizeable.35 Therefore, by focusing on one equation only, we ignore a more extensive (multi- equation) model where past and present female pay and work exhibit strong endogeneity.A This point has been forcefully made before and the endogeneity problem is clearly documented in the empirical literature.37 In fact, one study, which corrected for the endogeneity of female participation and used actual female labor market experience, estimated that the returns to experience in the case of women are significantly greater (by as much as 50 percent) than the corresponding returns for men.' A possible explanation for this is that more "experienced' women (women who are permanently attached to the labor force) are a relatively scarce factor of production compared to men. Another explanation, of course, is that those women with a long labor market history may be superior to the average man in the labor force. This issue relates in part to the selectivity problem addressed below. Selectivity. The use of earnings functions in the study of discrimination may have an additional complication, namely selectivity. Selectivity relates to the bias which results from the omission of unobserved variables from the analysis and can be highlighted with the use of the following example. Assume that all men work irrespective of their unobserved innate ability and their average pay is $100. This level of pay reflects the reward to the work of an average man. Assume that only half of women work and their average pay is $60. Finally 3 See among others Rosen (1969, 1976), Brown, Levin and Ulph (1976), Atkinson and Stem (1980), MaCurdy (1981), Blundell and Walker (1982). For a survey see Pencavel (1986) or Fallon and Verry (1988). 35 Though the labor supply elasticities for women have been estimated to be as high as 14 or more (Heckman and MaCurdy, 1980; Dooley, 1982), a value of around 1.5 to 3 could be considered quite typical. This range is also typical for a number of countries in Latin America (ECIEL, 1982). 3' In fact, the seriousness of failing to account for the endogeneity of female participation can be shown by reference to a study which reconciled the micro-evidence (cross-section data) with the observed patterns and trends in female participation (time- series data) only after correcting for (endogenizing) the experience of married women (Iglesias and Riboud, 1985). 37 Blinder (1973) and Mincer and Polachek (1974). 38 Zabalza and Arrufat (1985, p. 86, Table 5.4). 168 Women's Employment and Pay in Latin America assume that men and working women have the same observable characteristics of relevance to the earnings function. The decomposition described earlier will attribute all the pay difference to discrimination. This result will be correct if innate ability is shared equally between working and non-working women. However, if women workers are more able than non-workers, then discrimination would amount to more than $40, as the average female worker is more able than the average male worker (=average man). In the present context, in which the focus is on the economic aspects of women's work, wages are observed for working women but not for non- workers. Hence, a sample of working women can be taken as representative either of all (working and non-working) women, or only of those women who can attract a high offered wage, or only of women who have a low asking wage. Intuitively, it may be more reasonable to assume that women in the labor force are more likely to be a combination of women with high potential labor market rewards and low tastes for staying at home. Consequently, the sample is drawn from a self-selected group of women who are not likely to be representative of prospective female workers. The implication for the study of discrimination is that wage offers may constitute a more appropriate variable upon which the decomposition should apply, and not the actual wages that are, in effect, derived from the offered wage distribution that is acceptable to job seekers.39 A counterargument could be that women who are not currently in the labor force may be simply at a different stage of their lifecycle and they may not be different from women who do work: at another time, non-workers may enter the labor force when childrearing is over while workers may drop out of the labor force for reasons of family formation. If this is the case, there should be no concern for selectivity and restricting the basis of comparison to the sample of working women only should not overstate the value of the wage offers to women.4 Whether this observation is more relevant than the one mentioned in the previous paragraph is an empirical question which has been addressed by the so called 'second generation' models of labor supply whose rationale can be seen in the following analogy.4' 39 Gronau (1974). 40 Gronau (1973), Cogan (1980b). 41 The second generation models are invariably based upon the procedure suggested by Heckman (1974, 1979) and are surveyed/expanded in Lee (1978), Willis and Rosen (1979), Wales and Woodland (1980), Killingsworth (1983), Heckman and MaCurdy (1980, 1982) and Borjas (1987). Analytical Issues 169 Assume that a driver sets off for a journey between two points. He does not know the distance between the two points and the car mileometer is not working. He wants to find out the distance between the two points. Under normal circumstances this should present little problem. If the car travels at a constant speed of 50 miles per hour (mph) and gets to the destination in two hours, then the distance should be 100 miles. If traffic conditions are variable, then one can work out the implications. However, apart from the faulty mileometer there is another catch: the speedometer is sticky at 50 mph, that is, it stays at 50 mph whenever the car reaches a speed higher than this and up to the maximum of 100 mph the car is capable of doing. Fortunately, the driver is accompanied by a helpful passenger, who happens to be a statistician. They set themselves up to find out the distance in the following way: the driver will inform the passenger about the speed of the car and any changes in the speed while the passenger will keep a note of the speed and will monitor the time. At the end of the journey their information consists of the following. First, the journey took 1 continuous hour of driving. Second, they travelled at an average speed of 40 mph for 51 minutes, when the speedometer oscillated between 0 and 50 miles per hour. On this information, they know for certain that they covered 40 miles while the distance covered during the remaining 9 minutes of high speed (more than 50 mph) is to be guessed. The driver suggests the assumption that during the time the car was driven in excess of 50 mph the speed was 75 mph (half-way between 50 mph and 100 mph). This would bring the total distance to ([1160][51x40+9x75]) 45.25 miles. The statistician passenger considers this to be too high: he does not think they exceeded the legal limit that often (the driver is prepared to accept this). The passenger asserts that, since the observed average speed was 40 mph during the 51 minutes for which reliable information exists, the average speed should be around 42 mph and the total distance 42 miles. The driver is initially mystified but, after listening to the statistician's explanation - which is repeated in the next section, agrees. Let us get the basics for establishing this assertion. In Figure 5.3 a standardized normally distributed variable is depicted for which information exists up to point T (truncation) but not above. In line with the previous example, T corresponds to 50 mph. The left hand side of the distribution corresponds to the 51 minutes for which information exists. The shaded right hand side of the distribution corresponds to the observations that have been lost during the 9 minutes that the car speed exceeded 50 mph (due to the faulty speedometer). In other words, there exists information for 85 percent of the observations but not for the remaining 15 percent. 170 Women's Employment and Pay in Latin America Denoting = estimated mean from the sample (40 mph for 51 minutes) IA = true mean of the distribution (unknown) a = known standard deviation (say, 7 mph) X} = kth observation of the variable (speed in any minute) n = number of useful observations (51 minutes) 0(b) = height of the distribution at the standardized truncation point (50 mph) 1-I = percentage of observations which are known (51/60 minutes or 85 percent) one can establish the relationship between the calculated (sample) mean from the useful observations and the true mean as follows. The sample mean is: = EXk/n Obviously this is lower than the true mean as information on high speeds is lost. To correct for this (selectivity) bias something has to be added. The "correction" (or "adjustment") factor needs to take into account how thick the missing tail of the distribution is at the truncation point. The "thickness' depends on the standard deviation of the distribution (or, equivalently, its height) at the truncation point. The truncation point relates to how much is "lost" as one moves along the horizontal axis (note that the truncated normal distribution is summarized completely by its mean, standard deviation and the truncation point). Hence, to calculate the correction factor one needs information on the thickness of the distribution and the location of truncation. The derivation of the adjustment factor can be found in advanced econometric textbooks (see, for example, Greene, 1990, Chapter 21) but, for present purposes, what has to be added to correct for the selectivity bias is the product of the standard deviation times "lambda" (X), that is: correction factor = standard deviation x X - cv X [4(a)/(i-f)I and the true mean becomes: /= + 0{0(8)/(1-I)] Let us put some numbers into the last formula. The driver and the passenger estimated that the speed averaged 40 mph for 51 minutes. They know that 15 percent of time (observations) is lost, hence (1-I)=1-0.15=0.85. The standard deviation is assumed to be known and equal to 7 mph. The final thing to calculate is the height (+(a), called "abscissa") of the distribution at the Analytical Issues 171 Figure 5.3 A Truncated Distribution Observed Unobserved T (40) (?) (50) sample true truncation mean mean point truncation point: from the standard normal tables 0(6)=0.2323 when 1=0.15. Puttmg these numbers together implies that the (estimated) true average speed of the car was: = 40 + 7(0.2323/0.851 = 40 + 1.91 - 42 The faulty speedometer analogy in the context of selectivity bias shows two things. First, when one utilizes information which relates to only part of the sample, one obtains estimates which will in the general case be biased. Second, adjustment for sample selection is possible only when one is prepared make a 'distributional assumption." In our example, the driver assumed that the distribution of speed above 50 mph had zero variance (constant speed at 75 mph) while the passenger assumed a normal distribution. Though one might feel that the passenger's guess was more realistic, one has to bear in mind that selectivity correction is not 'something for nothing. " The benefit is that one can work even 172 Women's Employment and Pay in Latin America with truncated distributions. The cost is the price one pays in the distributional assumption. The similarity of the faulty speedometer to the study of female labor supply is obvious. One can get information on wages from existing workers, establish the known part of the wage distribution and make an informed guess about the unknown part, that is, the er ante wages of non-working women. We are now in a position to compare the econometric models which did not take into account selectivity in the estimation of earnings functions with those which explicitly accounted for it. The former models are conventionally called "first generation' labor supply models while the latter are called "second generation' models. The earlier models considered the market wages of working women only. In terms of regression analysis, these models postulate that the observed (market) wage of a woman relates to her characteristics in the following way (omitting individual subscripts for notational simplicity): WM = Xa + eM (5.8) where Wm is the market wage, X refers to characteristics relevant to the labor market, a is the vector of their respective coefficients, and em is a normally distributed error term with zero mean.42 The expected value of market wages thus specified is: E(WM) = E(Xa) + E(em) = Xa (5.9) under the assumption that the error term has zero mean. This specification is very much in accordance with the simplified earnings functions presented earlier in this chapter. The second generation models note that market wages are observed only for women who have already decided to work. For these women the market wage should be higher than their shadow wage at home, WH, otherwise women would stay at home where rewards are higher than in the market.43 As in the case of 42 For simplicity we do not specify explicitly the constant term which can be assumed to be contained in vector X. 43 Note that this does not imply that wage offers among working women are on average necessarily greater than wage offers among non-working women. Neither does this imply that productivity at home is necessarily greater for non-working women than working women. It is therefore possible that women who can fetch a high wage in the Analytical Issues 173 market wages, the value of the shadow wage can be thought as determined by women's personal characteristics, call them Z.' Hence, WH = Zb + e,1 (5.10) where b is the corresponding vector of coefficients and eH is another error term with the conventional properties. A woman's market wage is observed if: I=WM-WH= Xa-Zb + eM-en >0 (S.11) where I is the difference between the market wage and the value of work at home. I is a continuous variable and can be considered to be an index representing a woman's propensity to participate in the labor market. In the present formulation (equation 5. 1 1) when the value of the index for a particular woman is positive, the woman decides to work.45 Consequently, the expected value of a woman's market wage is not dependent only on her labor market characteristics, X, as was assumed in the first generation models (equation 5.9), but on: E(Wm) = Xa + E(eMII>O) (5.12) which depends also on her personal characteristics, Z, because equation 5.12 incorporates the arguments in the right-hand side of inequality 5.11. A comparison of the last equation with equation 5.8 reveals that not correcting for selectivity amounts to another omitted variables problem, that is, we omit the second term in equation 5.12 (the "sample selection rule") which indicates whether a woman would be in the labor market. The sample selection rule market do not work as they are also more productive at home than workers. In other words, the truncation point varies with the characteristics. This amounts to extending the car experiment to more than one car whose speedometers get sticky at different speeds: we do not know which car is faster than the others. 44 The vector of personalcharacteristics (Z) which determine women's productivity at home is assumed to include all the characteristics of relevance to the labor market (X) plus others but not vice versa. This assumption appears to be intuitively correct as there are many personal characteristics which do not affect productivity in the market but not vice versa. For example, being able to drive may augment home production but very few employers would care about it. 45 In fact, the critical value of I at which a woman decides to join or not the labor market does not have to be zero. However, without loss of generality, one can normalize the critical value to zero as we have done in the text for expository purposes. 174 Women's Employment and Pay in Latin America consists of two terms: it is the product between a scalar, a, and a variable (called lambda or the inverse Mill's ratio). The former term, o, is a function of the standard deviations of the error terms in the market wage equation (5.8) and the home wage equation (5.10), and their correlation."' The latter term, lambda, is the ratio of the ordinate of the standard normal density divided by the standard normal distribution both evaluated at (1). In other words, its value depends on where the critical point (truncation) lies and the height of the distribution at that particular point.' Lambda can be calculated from another regression of a woman's decision to participate in the labor market upon her personal characteristics (this regression is called the participation function).4 Then lambda can be included in equation 5.12 in order to solve, at least in theory, the problem of omitted variables as far as selectivity is concerned and it should attract a coefficient equal to v: E(Wm) = Xa + o(lambda) (5.13) Equation 5.13 is representative of the earnings functions used in the present studies. Two more observations need to be added. First, the theory as presented here does not unambiguously predict the sign of o, the coefficient on lambda (that is, the outcome of the interplay between the standard deviations of and the correlation between the errors terms).' For example, the coefficient "' The value of the scalar, a, is equal to uhw/h)-r.)>m)l where a(.) stands for the standard deviation of the appropriate error term and r for the correction between the two error terms. 4 Because the problem arises from the "truncation" of the distribution of a particular characteristic, economists have habitually referred to this situation as a "truncated" regression instead of the more accurate term "censored." In the present context, truncation would occur if there are observations only for working women. However, what we are presented with is censorship which occurs when there is information, though incomplete, about non-working women - though not, of course, with respect to their market wages. The correction consists in effect of predicting the missing information from whatever information is available in the censored data set. 4 The dependent variable in this case is a binary variable which takes the value of 1, if a woman is in the labor market, and the value of 0, if a woman is inactive. "9 In the early literature the expectation was that the coefficient on lambda would be positive. In fact, studies which reported negative coefficients were more or less dismissed as an "anomalous" result (Killingsworth and Heckman, 1986, commenting on the results for selectivity among Canadian women estimated by Nakamura at aL, 1979b). However, subsequent studies have often reported either insignificant or negative and Analytical Issues 175 on lambda will be positive (and significant), if the unobserved factors which induce women to work are also directly related to female pay. However, rewards to home activities may be more dispersed than those in the market and unobserved variables which boost productivity at home may relate positively to returns to market work. If these two conditions are met, then women who decide to join the labor market would be those who are least productive in terms of the unobservables5' and the coefficient on lambda will be negative (and significant). These remarks suggest that the inclusion of lambda in an earnings function may solve51 the econometric problem that arises when the error terms do not have the expected optimal properties but there is no 'correct sign' or unique interpretation of its coefficient.52 Second, in many studies the coefficient on lambda proved to be statistically insignificant. This insignificance can be interpreted as no evidence of self- selection.53 This can, in turn,, be taken to imply that women as a group are more homogeneous than initially perceived and they are, in general, expected to work as much as men do. In fact, all country studies in this volume confirm that, after controlling for other factors, women have a greater tendency to work as they enter prime age. Along these lines, the observed differences in participation are simply a reflection of household specialization during farmily significant coefficients on lambda. See Heckman and MaCurdy (1980) and Stelcner and Breslaw (1985). Behrman, Wolfe and Blau (1985) also find a negative coefficient on ambda for Nicaraguan women in rural areas. ' For a more detailed exposition of this argument see chapter on Venezuela by Cox and Psacharopoulos in the companion volume. 51 The simplified exposition of the sample selection effect and its correction adopted in our presentation masks the complexity of selection models and a number of difficult to resolve specification issues. More specifically, the conventional error distribution assumptions have been questioned by Lee (1982), Olsen (1980) and Duncan and Leigh (1985). In addition, problems of heteroscedasticity have not been fully resolved (Nelson, 1984). Finally, the assumed additivity in the effects of variables in the various equations in the model may not hold (Little, 1985). 52 Dolton and Makepeace (1987). 53 Cogan (1980b) shows that lack of adjustment for self-selection bias does not change the parameter estimates of the labor supply model for U.S. married women, who are the group for which the bias might be thought a priori to be more serious (for example, the corrected return to schooling increases to only 8.8 percent from 8.5 percent). 176 Women's Employment and Pay in Latin America formation. Thus, at a point in time some women are in and others out of the labor force. At another point in time those who were previously in are out and vice versa.' This relates to a "quantity" interpretation of women's labor supply and, in effect, challenges the "qualitative" view derived from the observation that not being in the labor force at a given time is highly correlated with not being in the labor force at any time.55 Though there is no ambition in the present volume to study whether "persistence" and 'habit formation"' is a more appropriate explanation for the pattem of female participation during the life cycle than the "intertemporal substitution" models of labor supply,57 one may feel tempted to accept that the forces determining women's roles within the family still dominate decisions about market employment in the developing countries. This may imply that selectivity may not be present in developing countries until the social groups are sufficiently differentiated -- when some critical level of per capita GDP is achieved. One should also add that results derived only from a sample of working women may well produce biased estimates even for working women. The reason that bias creeps into the estimation is that both asking and offered wages depend on unobserved variables (an educated woman may well have a lower taste for home chores). As a result, the error terms in the structural model (the decision to work, to work at what wage, and to work for how long) are correlated with other variables assumed to be exogenous to the model, and are also correlated between themselves across the different equations in the model.'8 Though in a different context, Nakamura, Nakamura and Cullen (1979a) find that the occupational and industrial distribution is the same for both married and unmarried women and this supports the view that the factors affecting women are common in the two groups. 55 Ben-Porath (1973); Heckman and Willis (1977); Keeley (1981). 5 Clark and Summers (1982); Blanchard and Summers (1988). According to these authors, the persistence in fcmale participation should be greater than in the working population as a whole. Cross and Allen (1988) echoed this theme at a macro-level and Behrman, Wolfe and Blau (1985) confirm that "there is a strong serial correlation in labor force participation because of differences across individuals in tastes, needs and returns from paid labor market participation" (p. 8). 57 Lucas and Rapping (1969); Altonji (1982). 5 Killingsworth and Heckman (1986). Analytical Issues 177 The implications of selectivity are usually visualized to be more important for women, as we usually observe only a fraction of women in the labor force while male participation is taken to be nearly universal. Although economists do not address the issue of male non-participation explicitly, especially for men in their prime age, one can be reasonably sure that some selectivity bias may be present even in the estimates for men. For example, during the process of development, life expectancy increases and older groups have lower labor force participation rates than prime age people. At the other end of the age distribution, the participation rates of younger men are also declining, though for other reasons such as rising school enrollment. Thus selectivity may also affect the estimates for men and its importance may be increasing over time.59 Having assessed, corrected, and interpreted the effect of selectivity bias, one can go back to the original question, that of discrimination. The correction in women's earnings functions allows us to estimate the wage offers of all women irrespective of current labor force status. Such knowledge can help us establish the productive potential of the country's womanpower, an important consideration in development strategies. However, one may argue that the kind of difference one wants to study is that arising in the labor market from demand factors. Under these circumstances, one needs to know what is paid in the labor market to those who work. Even if working women are a self-selected group with better than average characteristics than the whole group of women, these are the ones whose productive characteristics are evaluated in the labor market. Can or should the market pay non-working women with inferior attributes as much as women who are better qualified and actually working? In this respect, the appropriate decomposition of the pay gap should apply to the coefficients of the female wage equation uncorrected for selectivity and to the average value of characteristics held by working women only. This is, however, a procedure that practically all studies of discrimination have bypassed instead routinely using selectivity corrected wages for the sample characteristics of working women. If one wants to expand the notion of discrimination to include the pay potential of non-working women, one should use the selectivity corrected wage estimates and evaluate the adjusted pay gap at the value of the average characteristics of all 59 Nine of the country studies which follow report selectivity corrected results for men (Brazil - both studies, Chile, Colombia 1979, Guatemala, Honduras, Panama, and Peru - both studies). The findings suggest that the labor supply of men may be more affected by the presence of selectivity than the labor supply of women. This is not an unexpected finding: non-working men are less likely to be representative of all men. Therefore, we are somewhat reluctant to assign a great significance to selectivity for men in comparing wage offers between women and men because the typical non-working man may be not comparable to the typical non-working woman. 178 Women's Employment and Pay in Latin America women in the sample, both working and non-working. For reasons of completeness, and as a contribution toward this new direction of research, the country studies have evaluated the decomposition using both uncorrected wages cum working women means as well as corrected wages cum all women's means. Finally, one can add that the mechanics of selectivity correction is subject to the reservations raised in the case of earnings functions. (For example, is the participation function correctly specified? Are the explanatory variables good proxies of the theoretical variables they are assumed to represent?) In addition, the estimates would be affected not only by omitted variables and measurement errors and so on, but also by optimization errors, preference errors and budget constraint errors.' The net effect of these errors is as yet unknown to practitioners. The "chicken and egg" question. Even if the functional form of the earnings function is the appropriate one, and there are neither specification errors nor omitted or poorly measured variables, one cannot be certain that the decomposition results reflect accurately a properly standardized difference between female and male pay. This is because it is hard to distinguish to what extent endowments are the effect of past or expected discriminatory practices, something which can be said to constitute indirect discrimination.6' For example, women may suspect or know that they are less likely to enter a high- wage and/or senior position and as a result they are discouraged from acquiring human capital of the size and/or type they would have opted for, had they perceived equality of opportunity in the labor market.62 In this respect, the 0 Optimization errors refer to discrepancies in the measurement of optimal and actual values of the variables concerned; preference errors refer to unobservable differences in utility functions across individuals; and budget constraint errors refer to unobserved differences in the budget constraints across individuals. These issues are not pursued further in this volume since they are still at a theoretical stage (see, Killingsworth and Heckman, 1986). 61 In this case one has a different kind of endogeneity that is in the earnings function (earnings versus investment in human capital) compared with endogeneity in the participation function (employment versus wages). However, Griliches (1977) argues that accounting for the endogeneity of schooling typically does not alter significantly the coefficients in the estimated earnings functions. 6 England (1982); Weiss and Gronau (1981); Gronau (1982). Analytical Issues 179 measured wage discrimination would be underestimated more often than not by the present decomposition.' Wat is pay'? With respect to the dependent variable, the current approach assumes by necessity that the only reward to a worker from selling his/her labor, and the only cost from employing a worker, is what we observe as reported labor earnings at a point in time. This is deficient as, theoretically, one should use permanent earnings and, in practice, one should include fringe benefits and all other aspects of pay. Again, the implications are different for the male and the female results, but little can be done to accommodate them in practice. Another issue which relates to pay is what variables, other than education/schooling and potential experience/training, can be used in the earnings equation if the objective is to measure discrimination. As already argued, the semi-logarithmic formulation views the relationship between individual earnings and human capital characteristics as the result of an individual's willingness to maximize his/her lifetime earnings." This is a supply side story. However, here one is interested in identifying demand discrimination. The argument is, therefore, about rates of pay for comparable work and not earnings which are the product of wage rates and labor supply. This implies that either earnings on the left hand side of the regression equation should be adjusted by the amount of labor supply that generated them (such as weekly earnings divided by weekly hours of work), or one should include hours as an additional explanatory variable on the right hand side. Although the inclusion of hours in the right hand side of the earnings function equation can be labelled agnostic on theoretical grounds,' it has usually been justified as a useful device for the study of sex wage differentials as discrimination is more meaningfully analyzed in terms of rates of pay among homogeneous groups. 63 For the difficulty in distinguishing between the causes and effects associated with unequal endowments see Zabalza and Tzannatos (1985), Chapter 1. " Mincer (1974); Blinder and Weiss (1976). 65 Alternatively, the inclusion of variables other than human capital ones (that is, mainly education and experience) may be interpreted in many ways. For example, assume that the earnings equation relates to manual workers where the physique of the worker matters. Thus, the inclusion of hours as an additional regressor can somehow reduce the bias arising from omitted variables, as the two sexes are heterogenous in this respect, but this has little to do with the original justification offered for the conventional semi-log human capital specification of the earnings functions. 180 Women's Employment and Pay in Latin America Rhat is "comparable characteristicsn? The attempt to make the two groups homogenous before the gross wage differential is broken down into its constituent components can go beyond controlling for differences in the amount of labor supplied by women and men. For example, men could be paid more because they work in certain occupations and industries rather than because they are paid more than women who may also work in those sectors.' Consequently, one may be tempted to include explanatory variables relating to employment status in order to adjust for the effect of the different employment distributions of women and men on pay. To ask an employer to pay women in one industry or occupation as much as men in another industry or occupation may not sound immediately obvious.67 The same considerations apply to the effect upon pay of differences arising from the regional employment distribution of women and men. Therefore, the standardization of certain differences in the earnings equation before one attempts to establish the discriminatory part appears prima facie necessary. However, it makes a lot of difference if women choose to become nurses (instead of doctors) or employers do not promote women to managers and let them stay in junior administrative tasks. If restriction of entry is a determinant factor for the employment distributions of the sexes, the inclusion of employment variables will result in more standardization than needed for establishing the unjustified wage gap between women and men. This will be so because that part of the wage differential which is due to employment status would be attributed to differences in characteristics although it is really due to discrimination in the form of unequal opportunities in employment. On the other hand, if occupational choice is unconstrained and the occupational wage structure reflects compensating differentials, ignoring the occupational structure would lead to an exaggeration of the extent of discrimination. i The evidence suggests that women are predominantly found in low-pay sectors and are also paid less than men within these sectors. 67 There are countries where such cross-establishment comparisons are allowed "in the same trade or industry" in the context of sex equality legislation if due to segregation in the employment of women and men, the comparability principle cannot readily apply at firm level (this provision is made by the Dutch Equal Pay Act of March 20, 1975; see Hepple, 1984). Analytical Issues 181 In the absence of clear theoretical guidelines for the specification of the eanings equation in the study of discrimination, the contributors to the country studies have omitted the employment structure.' In effect, this approach amounts to the purest application of the human capital theory in the study of labor earnings.' There were also two practical considerations for this eclecticism. First, in some countries there was no information on either the occupational or the industrial status of those working. Second, the coding practices differ between countries. If industrial/occupational variables were included in some of the present studies but excluded in others, this would have limited to a significant extent the comparability of the findings. 4. Summary of the Methodology Adopted in the Present Study There are considerable unresolved theoretical and empirical problems in estimating which part of the gross difference between female and male earnings can be attributed to discrimination. Despite the analytical nicety of the distinction between demand and supply, in practice it is difficult either to define what constitutes a discriminatory practice or to disentangle it from what may be a rational individual choice. Further, one cannot be sure whether the mechanics of estimation result in the net over- or underestimation of discrimination. The existing empirical literature in developed countries has provided varying pictures of the situation depending on the time period examined, the specification of the model, the steps taken to correct for selectivity, the extent to which endogeneity was accounted for, and the type of data used.' Bearing these reservations in mind the authors of individual country studies were asked to: 68 There have been some recent studies which attempt to account for the wage effect of occupational segregation (Brown, Moon and Zoloth, 1980, for the United States, and Chiplin and Sloane, 1976 and Greenhalgh, 1980, for Great Britain). Brown et al. merge behavioral models of occupational attainment and of gender wage distributions and then derive a more accurate decomposition of the pay gap between women and men. However, it was deemed that the data requirements for such approaches was beyond the limits of the present study. This may not be a serious problem as Miller (1987) notices that eliminating differences by sex across occupational groups will have little impact on women's wages unless it is accompanied by changes in relative wage rates between occupations. 69 Mincer (1974). 7 For reviews see Lloyd and Niemi (1979), Treiman and Hartmann (1981), Sloane (1985), Cain (1986), Killingsworth (1990). 182 Women's Employment and Pay in Latin America 1. Give an aggregate picture of women's relative position in the labor market; 2. Estimate a female labor force participation function in order to show the impact of certain variables upon women's decisions to participate in the labor market and also to produce the selectivity correction variable for inclusion in the earnings functions; 3. Evaluate the determinants of pay using a relatively parsimonious, but also comparable across country studies, specification of the earnings functions; and 4. Decompose the pay gap, if women were paid as men and if men were paid as women both in respect to actual wages and also in terms of wage offers. The results are summarized in the following chapter. 6 Summary of Empirical Findings and Implications 1. Introduction This chapter gives representative results from the country studies included in the companion volume. Two warnings apply. First, the reader should note that some of the results may not be strictly comparable across countries. This point can be clarified with reference to the following case. There was no information in the Chilean and Mexican data sets about weekly hours of work. As a result, this particular variable does not appear in the earnings functions for these two countries and the effect of differences in hours worked by women and men is not reported in the corresponding summary table. The omission of hours may have also affected the coefficients of other variables included in the regression. Throughout the summary presented below we have, therefore, selected the most representative results that are susceptible to an analysis in a comparative context within the scope of the present study. The interested reader can always refer to the individual country studies for more detailed information. Second, our results as summarized below refer to all working women in the economy, that is, the distinction between employment in the formal and informal sector is not pursued in this section in great detail. We felt that any generalization about the informal sector could be misleading as (1) the informal sector is very much country specific (with respect to aspects such as racial composition of the population and regional pluralism) and not susceptible to easy categorizations; (2) even if the informal sector were comparable among countries which are at the same stage of economic development, it is not necessarily comparable across the countries examined in this volume. These remarks can be examined with reference to the few individual country studies which were able to tackle these differences. For example, Stelcner et al. conclude that Brazil's regional/economic/social diversification may preclude the consideration of the country 'as a whole' and note that the urban/rural residence is an 184 Women's Employment and Pay in Latin America important one for employment in the formal sector and the earnings of employees but not for the self-employed. They also argue that education not only enhances earnings but 'sorts" individuals among different types of labor force activity (dependent employment, self-employment and family work). Tenjo also notes that the decomposition results for Colombia are sensitive to the inclusion or not of domestic servants. If the latter are excluded from the analysis then the unexplained part of the gender wage gap is reduced from 77 percent to 10 percent. He is, however, careful to add that there are employment "ghettos" for women and there are more opportunities for occupational advancement for men than for women. In contrast, Gill's distinction between self-employment and wage/salaried employment in Peru produced insignificant results. These remarks suggest that as far as the formal/informal distinction is concerned, it may be better to refer to the country studies directly rather than attempt to produce a synthetic profile of the informal sector across the region.' We present the results in the following order. First, we look at some factors which affect the decision of a woman to participate in the labor force. Second, we present the aggregate results of the decomposition analysis, that is, the estimates for the percentage of the wage gap which can be attributed to differences in the average values of the characteristics between women and men ("differences in endowments" or "justified" part of the pay gap) and to differences in the rewards of these characteristics in the labor market ("upper bound of discrimination' or "unjustified" part of the pay gap). Finally, we attempt to identify the contribution of some variables included in the earnings functions to the observed pay gap the pay gap. 2. Participation Table 6.1 summarizes the effects of some key variables on women's decision to participate in the labor market. Education has a significant effect on participation. For example, in Argentina the observed average participation of all women is 36 percent. However, a woman with less than primary education has a ceteris paribus probability of participation of only 22 percent compared with a probability of 58 percent of a woman who is a university graduate. In Venezuela the probability of participation for the cofresponding education groups rises from about 30 percent to more than 85 percent. These estimates compare to an average female labor force participation rate in Venezuela of approximately 40 percent. I For some aspects of the diversity of women's labor market characteristics and treatment in the informal sector see Berger and Buvinic, eds. (1989). Table 6.1 Female Participation by Selected Sample Characteristics (percent) Argentina Chile Colombia Costa Rica Ecuador Guatemala Panama Peru Uruguay Venezuela Characteristic 1985 1987 1988 1989 1987 1989 1989 1990 1989 1987 Education Less than Primary 21.7 - 11.0 16.9 44.0 21.4 10.0 38.6 28.9 31.8 Primary 31.4 23.7 20.0 21.6 46.0 22.4 14.1 38.7 34.7 34.4 Secondary 32.6 32.5 34.0 30.9 47.0 40.7 33.6 40.0 46.9 62.8 University 57.6 60.7 53.0 38.1 49.0 47.2 47.7 63.2 54.1 87.4 Marital Status Single 55.9 40.8 40.4 33.0 47.3 56.2 Married 24.5 13.9 17.7 14.1 33.1 34.3 Number of Children None 37.2 28.6 25.0 25.2 49.0 22.3 26.7 42.6 39.8 One 33.6 23-0 )20.o 24.4 45.0 21.3 23.7 38.2 32.5 Two 30.2 18.0 ) 23.5 42.0 20.3 20.1 34.0 25.9 Household Head 37.8 47.0 34.1 65.0 30.3 57.2 57.9 65.8 64.7 Not Head - 21.0 22.7 43.0 19.6 20.4 - 34.2 41.7 Residence Rural 17.9 19.9 15.6 17.8 32.7 Urban - 28.9 29.7 29.6 45.6 - not available. Source: Based on case studies reported in Volume 11. 186 Women's Employment and Pay in Latin Amerca Women's family characteristics also exercise a strong effect on participation. Married women's probability of labor force participation is about half the probability for single women. For example, the probability for married women in Chile drops to 14 percent compared to 41 percent for single women. In Costa Rica the corresponding decrease is from 40 percent to 18 percent, in Venezuela from 56 percent to 34 percent and in Guatemala from 33 percent to 14 percent. Being head of household increases the probability of participation in all countries under consideration. In fact, this demographic aspect seems to be one of the strongest determinants of female labor force participation. For example, in Colombia the probability for a woman who is head of household jumps to 47 percent (from 21 percent for non-heads of household), in Panama to 57 percent (from 20 percent), in Uruguay to 66 percent (from 34 percent), in Venezuela to 65 percent (from 42 percent) and in Guatemala to 30 percent (from 20 percent). The effect of children is mixed depending on their age. As a general rule, results from countries which could not take into account the age of children suggest that the probability of participation drops by about three to five percentage points for each child. When the age of children could be taken into account, the results for young children (aged less than 6 years) suggest that the effect is even stronger.2 However, the presence of older children increases the probability of female participation in some countries.3 This can be explained by the fact that older children may be substitutes for women's services at home, as older children can both supervise their younger siblings and contribute toward other areas of family production.4 Individual country studies also report effects of variables which are country specific. For example, in Bolivia the probability of Spanish speaking women to be in the labor market is 42 percent while the corresponding probability for indigenous women is only 22 percent.5 Also, variables relating to the physical health of a woman have the predicted effect that is, ill-health affects adversely 2 See chapters on Chile, Panama and Uruguay in the companion volume. 3 See chapters on Jamaica and Peru in the companion volume. 4 Boserup (1970) and Standing (1981). S Indigenous women have a lower probability of participation in the labor force in Guatemala and in Brazil there are substantial differences in the participation probabilities of women belonging to different racial groups (white, black, Asian and Mulatto). Enpirical Findings 187 the probability of participation.6 In addition, the presence of adult non-workers in the household decreases the probability of female participation.7 This may suggest that there is higher demand for domestic services when older people are in the household and this translates into an increase in women's shadow wage at home.' Finally, residence is another significant factor for women's decision to participate. As a general rule women in urban areas have a much higher probability of participation than their counterparts in rural areas. In conclusion, the country studies confirm that women's decisions to participate in the labor market depend on education and on their demographic characteristics. As explained in the methodological section, these results are derived from the assumption that these characteristics are exogenous to the labor supply model estimated in our country studies, and this assumption is not necessarily appropriate. However, the magnitude and consistency of results are sufficiently clear for the limited generalizations presented in the introductory chapter to this volume with respect to policy implications. 3. Bird's Eye View of the Aggregate Pay Differential Table 6.2 presents the percentage of the pay gap which can be attributed to differences in the labor market endowments held by women and men and to differences in the labor market rewards associated with these characteristics.9 6 See chapter on Bolivia in the companion volume. S 5ee chapters on Ecuador and Jamaica in the companion volume. 8 These estimates may understate the effect of the presence of non-working adults in the household upon the probability of female participation as such adults may also assist in some household tasks performed by women. 9 The information presented in the Table 6.2 is derived from Appendix Table A6.1 where we report results which are standardized on female means (women paid as men: columns 2-3 and 6-7) and also on male means (men paid as women: columns 4-5 and 8- 9). We also report separately the results obtained from coefficients which were uncorrected (columns 2 to 5) or had been corrected for selectivity bias (columns 6 to 9). In some countries where selectivity did not appear to be statistically significant, the authors did not report separately results for uncorrected coefficients. In other countries, the authors reported coefficients uncorrected for selectivity for illustrative purposes. When the coefficient on the selectivity correction variable was statistically insignificant, the results for corrected and uncorrected coefficients were practically identical. 188 Women's Employment and Pay in Latin America Table 6.2 Decomposition of the Male Pay Advantage in the Region (percent) Evaluated at Selectivity Pay advantage Female means Male means Correction due to No Endowments 3.2 2.9 Rewards 96.8 97.1 Yes Endowments 11.7 13.7 Rewards 88.3 86.3 Totalr 100.0 100.0 a. The overall male pay advantage is 30 log percentage points. Source: Appendix Table A6.1 A number of clarifications need be made in order to correctly interpret the summary decomposition results. First, the pay gap is shown as log percentage points of the male pay advantage.10 This corresponds closely to the ratio of average male earnings to average female earnings in the sample." Second, differences in endowments refer to the difference in the average values of a particular characteristic in the sample between men and women."2 Third, 10 The reason for expressing the pay gap in terms of male advantage is that the decomposition is formulated in this way. Recall that our decomposition refers to the difference in the average logarithms of earnings between men and women, that is, log(Wj-log(Wf). See equation 5.6. " For example, in Argentina the pay gap is indicated as 43.2 percent and this should be taken to mean that men earn on average 43.2 percent more than women. In other words the pay differential is not expressed in one of the more conventional ways - - such as in terms of the relative female/male wage (which is 69.8 percent) or the underpayment of women (which is 30.2 percent). 12 Again with reference to Argentina working women have on average 9.4 years of schooling compared to 8.8 years of schooling for men (see Appendix Table A6.3). In this case the relevant difference is -0.6 years of schooling and the negative sign suggests that, had women and men been equally endowed in terms of schooling, the male pay advantage would have been even greater. Empirical Findings 189 differences in rewards refer to the difference in the corresponding coefficients as reported in the earnings functions. Again this difference is calculated between men and women. According to the data, the male pay advantage in the present data set varies between about 15-20 percent (in Colombia, Mexico and Peru) and 40-50 percent (in Argentina, Bolivia, Ecuador and Jamaica). This gives an unweighted average of male pay advantage in the region of about 30 log percentage points (or female/male pay of about 75 percent). The two most obvious conclusions that can be drawn from Table 6.2 are the following. First, the selectivity corrected estimates for the part of the pay gap attributable to differences in rewards (upper bound of discrimination) are lower than the results derived from the uncorrected estimates. Correcting for selectivity reduces the upper bound of discrimination from 97 percent to around 87 percent. The reason that selectivity correction reduces the unexplained part of the wage gap in the region is because in most country studies the coefficient on lambda was negative.'3 This implies that the difference in the wage offers of women and men is smaller than the difference in actual wages. Though it is not strictly appropriate to calculate averages from percentages, especially when these percentages refer to countries which differ in population size (and characteristics) as much as Jamaica differs from Brazil, the magnitude of the corrected and uncorrected results may be taken to suggest that only a small part of the difference in the actual wages between women and men reflects self- selection of women in the labor force. Our figures imply that, if the average female worker had the characteristics of the average woman in the population, then the observed pay gap would decrease by approximately 10 percent (or about 3 percentage points)."4 13 The coefficient on lambda was positive and significant only in the case of Uruguay. Insignificant positive coefficients are reported for Ecuador, Peru (1986) and Bolivia. Insignificant negative results are reported for Argentina, Colombia (1979), Costa Rica, Peru (1990) and Venezuela (1989). See Appendix Table A6.6. 14 This was derived by taking the difference between the uncorrected estimates for the part of the pay gap attributed to rewards (97 percent) and corresponding part suggested by the corrected coefficients (about 87 percent). Given that the male pay advantage is about 30 percent, this implies that (0. lx.0.3 =) 3 percentage points would be eliminated from the actual wage gap, if working women were representative of all women in the economy. 190 Women's Employment and Pay in Latin America A second conclusion is that, irrespective of whether the results are based on corrected or uncorrected coefficients, only a small part of the pay gap is attributable to differences in endowments. Even in the case of selectivity corrected results almost four-fifths of the pay gap is due to differences in rewards. This finding warrants further inspection in order to see which variables in the earnings functions give rise to such an impressive result. Contribution of specific variables to the decomposition. The decomposition was calculated from separate earnings functions for women and men. In particular, the earnings functions had the following simplified general form (omitting subscripts i for notational simplicity): ln(W,) = Cm + amSm + bmEm + c,ln(H) + (X,)m + em In(Wf) = Cf + afSf + bfEf + cfln(H,,) + (Xf)f + e, where W stands for weekly earnings, C is the constant term, S refers to education measured in years of schooling, E is potential experience also in years (age minus years of schooling minus school entry age), H is weekly hours worked and X is a vector representing whatever other variables might have been included in the earnings functions of individual country studies. Lower case letters attached to these variables stand for their respective coefficients, subscripts m and f refer to men and women, and e is an error term assumed to have the usual properties. Let us now dissect the contribution of each of the variables explicitly specified in the last two equations upon the observed pay gap. For brevity, we report results for coefficients derived from the sample of women workers only, that is, uncorrected for selectivity. As noted in the previous sections the difference between corrected and uncorrected coefficients was not that great. Also, uncorrected results are more comparable across countries than corrected results because the variables used in the participation functions (in order to estimate the lambda) varied between countries. In addition, recall that women workers are the appropriate reference group for explaining the observed (actual) gap in wages (rather than the gap in wage offers). Finally, the coefficient on lambda was found to be statistically insignificant in half of the countries studied.'5 As noted earlier, our regional averages were calculated from percentages and their sums do not match exactly the observed pay gap. In addition, there are effects from other variables which have not been taken explicitly into account. '5 See Appendix Table A6.6. Empirical Findings 191 Table 6.3 Contribution (in log percentage points) of Selected Variables to the Male Pay Advantage in the Region Variable Due to Difference in Total Explained by Endowments Rewards these Variables Hours 6.6 -32.1 -25.5 Education -11.1 -12.3 -23.4 Experience 9.9 10.2 20.1 Total exel. constant term 5.4 -34.2 -28.8 Constant term 52.2 52.2 Total incl. constant term 5.4 18.0 23.4 (22%) (78%) (100%) Source: Appendix Tables A6.1 to A6.5. Table 6.3 shows which part of the pay gap can be attributed to differences in endowments and which to differences in rewards with respect to each of the main variables in the earnings functions. In terms of endowments, women appear to be disadvantaged with respect to hours and also experience (though men have longer experience because of fewer years of schooling). These two variables taken together explain just over half of the pay gap (16.5 log percentage points). However, women's advantage in schooling reduces the pay gap and, in fact, schooling has the strongest effect of these three variables in terms of endowments. As a result, the part of the pay gap that can be attributed to difference in endowments is only 22 percent. In terms of rewards, women seem to benefit significantly from education and hours (though the latter may be taken as an unfavorable result because women work fewer hours than men).'6 These two factors would have more than 16 This is an appropriate point to raise a specific methodological issue in the present decomposition analysis. Can the greater female coefficients on hours be interpreted as discrimination against men? In fact, this is what a "mechanical" interpretation of the Oaxaca decomposition would suggest. However, we do not think this interpretation is correct. Another way of interpreting the female advantage in the rewards of weekly hours is the following: women are penalized in the labor market because working fewer hours than men reduces their pay proportionately more than men working shorter hours. In this respect, what the present decomposition assigns as a female advantage in terms of rewards is in effect a disadvantage because endowments are systematically lower for women than men. 192 Women 's Employment and Pay in Latin America eliminated the pay gap had it not been for the opposite effect from the difference in the coefficients on experience. However, the contribution of experience is not that important and the net effect of these three variables suggests a reversal of the pay gap (pay advantage for women). When the constant term is taken into consideration, the effect from the difference in rewards is inflated in the opposite direction to the point that differences in rewards account for 78 percent of the male pay advantage. These results may be subject to two different interpretations. First, ignoring the effect of differences between the constants terms,'7 one can argue that in many of the countries studied in this volume only a small fraction (if any) of the gross pay differential can be attributed to different wage structures (differences in rewards).8 However, this interpretation may be biased by the fact that formal sector employees are heavily represented in our samples. For example, in Guatemala the female participation rate is originally reported as 29 percent but the percentage of women with positive incomes and positive hours is only 24 percent. One suspects that the 20 percent of women missing in the second figure are women engaged in family work or other activities in the informal sector. Also, public sector employment accounts for a significant proportion of female workers in the formal sector and appears to be paying its female workers substantially more than women employees in the private sector. In fact, women's pay in the public sector is found to be higher than that in the private sector in many of our studies."9 To some extent this finding reflects the higher educational attainment of women workers in the public sector. However, the study on Ecuador confirms that women workers in the public sector have a wage premium of almost 25 percent (compared to the private sector) even after 17 The role of the constant term in measuring discrimination is questionable, if the constant term is taken to proxy the average effect on earnings of productivity characteristics omitted from the analysis. However, if the earnings functions are correctly specified, then the constant term should be included in the decomposition formula. is This is not an uncommon finding in the literature on discrimination for developing countries. Knight and Sabot (1982) report that in Tanzania in 1971 only 5-17 percent of the gross pay differential between women and men can be attributed to different wage structures when evaluated at male means while it is negative (-3 to -45 percent) if evaluated at female means. Similar results are reported in many of the studies included in Birdsall and Sabot (1991). 19 See chapters on Costa Rica, Guatemala, Honduras, Panama, and Uruguay in the companion volume and Table 6.4. Empirical Findings 193 controlling for the effects of human capital and other variables.' Similarly, the study on Uruguay suggests a corresponding premium of 15 percent again after controlling for other characteristics.2' The overrepresentation of public sector employees in our databases and the high wages paid to women workers in the governrment sector make us skeptical about the "no or little discrimination' results suggested by the decomposition analysis in some of our countries. In addition, as argued below, it might be true that the constant term cannot be considered separately from the other 'rewards' in the decomposition analysis but there is no practical objection that its place is among the "rewards' part of the gender wage gap. The second interpretation stems from the last observation, that is, that the part of the wage gap attributed to the difference in the constant terms falls well within the potential discrimination aspect of the results. The constant term represents the "reward to the sex of the worker" when all other characteristics are set equal to zero. In other words, the constant term can be interpreted as the earnings of an uneducated worker who is just about to enter the labor market. However, we are not prepared to say that the difference in constant terms represents actual discrimination because information is lacking. In particular, as noted earlier, the value of the constant term is affected by factors pertaining to both labor demand and labor supply. With respect to labor demand, one can mention genuine differences in productivity between the sexes or imperfect information while on the labor supply side there may be genuine differences in tastes between women and men. This is as pessimistic a conclusion to the present analysis as the light that was shed into the economists' "black box" (differences in rewards) which revealed another black box inside it: the constant term is another black box on its own. A more optimistic interpretation may, however, be relevant to our findings. The constant terms can be seen as a pure premium that is independent of a worker's other wage determining characteristics. Hence, if women are undervalued in the market when they have few characteristics (zero endowments) but recoup almost half of the lost ground because of the effects of schooling, hours of work and, possibly, experience, then one may have a policy prescription to the problem of growth and the feminization of poverty: if women's education increases and their labor force attachment and experience 20 See Table 17.4. The same study suggests that there is no corresponding premium for men working in the public sector. 21 See Table 25.5. It is interesting to note that the premium to male workers in the public sector is negative and significant (-7.6 percent). 194 Women's Employment and Pay in Latin America also increases, women's pay will increase proportionately more. In fact, we believe that this is the way the present findings should be interpreted. 4. Evaluation of Results The present volume added some information to the growing area of research on .women in development." We make no claim that our study has solved the problems associated with women's role in the economy. There are significant methodological issues and practical problems which remain unresolved. The reader had the opportunity to assess the merit and limitations of the preceding discussion and evidence. Below we attempt a synthesis, albeit aggregate and tentative, of we believe that can serve as a basis for some limited generalizations and potential policy implications. Most of the discussion derives directly from the previous chapters but is supplemented by additional information in some instances. Female participation/employment. Initially our analysis was based on one of the "grand ratios" in economics, namely the labor force participation rate. Conventionally defined, the participation rate refers to those who are engaged in an economic activity. Economic activity is taken to be one which results in an "economic = financial" reward. In advanced economies where production is very much monetized this interpretation of an economic activity may not be inappropriate. In these economies most workers are employees, thus, they are engaged in tasks that result in pay in one form or another. Also, if one were to focus on men, the value of men's household production can be thought to be low in the sense that few men, especially in developing countries, spend much of their time at home. As most "value" is deemed to derive from labor, even a relatively single minded approach to men's contribution to economic/social welfare may not lead to grossly misleading conclusions. This is not, however, the case for women, especially in developing countries. Women spend a substantial part of their time in and around the house. Analyses of women's issues that adopt the conventional dichotomy (between "work" and 'leisure=non-working time") are bound to miss some significant aspects of economic and social welfare. The reason is that non-working time includes activities such as meal preparation (from purchasing, if not growing, the basic inputs for meals to serving meals on a plate and cleaning the dishes); looking after and improving the mental well-being of persons (in the sense of providing informal education to younger members of the family - home economics has been the earliest subject taught much before the appearance of schools and its formal inclusion in the curriculum); and a wider range of activities, all of which are necessary for human existence and survival -- such activities extend from the Empirical Findings 195 purely "mechanical" activities (such as the "passive" supervision of minors -- if it could ever be passive) to the extremely "sophisticated" ones (such as the administration of medicines). In short a narrow interpretation of what constitutes an economic activity can easily lead to a practically meaningless measurement of women's contribution to societal output. To assume that women's work at home has zero value does not make sense. It must be obvious to the reader that we have not made such an assumption. The value of production at home, especially women's home value, is a prerequisite for all economic (monetized) activities. With this in mind we examined the female participation rate. We found that there has been an underlying upward trend in women's work in the market which, though we had no ambition to analyze it and offer explanations, suggests that where "other regions are already today" (the industrial economies), Latin America "will be there tomorrow." The question becomes whether Latin America is catching up with the rest of the world in an efficient and distributionally appropriate manner. "Efficient" means simply 'can a rearrangement of existing resources produce a pie of greater size?" 'Distributionally" means "does the resulting growth from a more efficient allocation of resources ensure that benefits accrue successively to a greater part of the population?" We found that female participation rates in Latin America are among the lowest in the world. If this is the result of women's or households' free choice and assuming that women are free to chose within households, then the issue of (in)efficiency does not arise. However, if female participation rates are low because of a market failure, then social policy may have something to offer. Let us clarify the relevant concepts and the conditions under which a market failure may arise. Efficiency implies that as much output as possible is created out of a given resource base at any given point in time (technical efficiency) and that output is also exchanged at the right relative prices (allocative efficiency). A third aspect of efficiency arises when intertemporal comparisons are made in which case dynamic efficiency refers to the maximum or optimal rate of growth. Distributional aspects have been often but erroneously considered to be independent of efficiency considerations and, at times, they are assumed to constitute a conflicting objective to wealth creation. However, it is more appropriate to think of "distributional efficiency," that unique initial allocation of resources among economic agents which will lead to the greatest feasible amount and most desirable product mix of output.2? In other words, in the X The first fundamental theorem of welfare economics postulates that a perfect competitive market economy wil of its own eventually reach a Pareto optimal equilibrium. The second fundamental theorem postulates that a given Pareto optimal equilibrium can be achieved only under a certain initial distribution of resources. In other words, that unique Pareto equilibrium which maximizes social welfare will not and cannot be achieved unless resources are distributed in a particular way. 196 Women's Employment and Pay in Latin America present framework it may be possible to achieve a superior outcome by changing the present distribution of resources among the sexes. It is in this context that we interpret our results by observing that initial conditions are different for women than for men (Table 1.2 showed that women's literacy rates were substantially lower than those for men); female labor force participation rates are also substantially lower than those for men (Table 2.2); and the decision of women to participate in the labor market is only determined to a small extent by what one can consider economic variables (Table 6.1). Do these findings suggest that there is a market failure somewhere? The answer is not an easy one in that what appears as a result may well be the cause. This is already covered ground (recall the relevant "chicken and egg' discussion in Chapter 5). However, we believe that on balance the biological asymmetry that has destined women and men to the 'traditional" roles is becoming successively less relevant for determining the functional roles of the sexes in the context of modem production. Today working for the market is characterized by more complex and capital intensive modes of production and by greater demand for coordination that calls for an expansion of the service sector. In this respect a market failure may consist of "less than perfect" information in the sense that roles and functions that are perceived today as appropriate by a girl and her family/parents may prove counterproductive later in life. The reason these expectations do not materialize is that until the steady state is reached present conditions are not representative of what would prevail in the future. The inability to see far enough in the future ("myopia") can easily divert the economy from its optimal path. In addition, the absence of perfect capital markets, even in industrialized economies, results in investment in human capital to be less than optimal: human bodies do not provide the kind of collateral for borrowing purposes that lenders are readily prepared to accept. We must next consider whether there exists anybody else other than individuals or households who is better equipped to reach more rational and non-myopic judgements.' The answer is that the general trend that economies follow 23 In technical terms the optimality properties of the perfect competitive market model breaks down if there is "irrationality" and "myopia." Irrationality means that the future is discounted at a higher rate than it should be discounted. For example, a sufferer from the severe results of smoking at the age of 50 would have in most cases decided against smoking earlier in life, had he "discounted properly" at the age of 20 how bad the quality of life would be 30 years later. This explanation of irrationality holds assuming that the smoker knew in advance the danger of smoking. If he did not know about the dangers of smoking but the information existed, then the market failure is due to imperfect information. In either case, there may be good grounds for policy interven- tion, if the benefits from reducing smoking outweigh the necessary costs of the policy. Empirical Findings 197 during development is better known to governments than to individuals. For example, individuals may assume that their employment conditions and family roles will continue to be in the future as they are at present. However, governments know that, even if conditions remained unchanged for some individuals, on average the employment conditions will move in certain directions (such as toward urban based activities in the services sector). Consequently, governments can undertake policies within a probabilistic context while, of course, individual actions will be decided by individual considerations. The relevance of these remarks to women's employment in the labor market is that it is possible that the currently observed low female participation rates in Latin America may well reflect a sub-optimal outcome in the sense that the developmental changes have not been correctly anticipated or have not been properly discounted by women or by their parents some time ago. As a result, it is probable that more women would have liked to be in the labor force today but find that they are not in possession of the right amount or mix of human capital. Failure of capital markets to ensure an optimal investment in human capital is also relevant in this respect. One does not have to engage in heavy theorizing in order to determine the role and effectiveness of social policy in market economies (such as those in Latin America) vis-a-vis actions undertaken by what is assumed to be 'well informed" individuals. Irrespective of persuasion, one is certain that Latin America is moving toward economic and family structures which are currently observed in industrialized countries. There are many and at times contradictory explanations about the "catching up" or "converging" process evident in many economies. The fact remains that in the longer historical perspective women become more like men in all aspects of life and the implications of the biological asymmetry of the sexes diminishes due to technical change. Those individuals and countries which can envisage these movements and adjust to them earlier rather than later are bound to find themselves in a better position to grasp the opportunities when changes occur. Of course, preparing for the future involves costs that have to be paid at present. For example, a campaign of information dispelling myths such as "the longer the hair, the smaller the brain"' has costs and, perhaps, 24 This phrase was obviously coined before the advent of long hair for men. It can be found in Weineger (1906) whose analysis led to the conclusion that "even the malest (sic) woman is scarcely worth more than 50 percent of men." Further back in history Aristotle wrote "the male is by nature (sic) superior and the female inferior: the one rules and the other is ruled" (Aristotle, Poliriks, bk. I, ch. 5) and the Book of Leviticus (27:1- 7) prescribed that the value of a woman shall be assessed at three-fifths the value of a man. 198 Women's Employment and Pay in Latin America not only economic ones. Also, the elimination of sex specific protective or prohibiting legislation would benefit some persons/groups but may result in costs for others. Those costs must be evaluated in comparison to the benefits before decisions are made that whatever involves costs is not desirable. In terms of female participation in Latin America, an optimist may argue that this could not have been any better in the past: the rate of increase in female participation has been remarkable by all accounts. A pessimist may want to point out that female participation is still lower than in other regions. Female pay. Pay is the price of labor. If pay accurately reflects competitive market conditions, then nothing more can or should be said or done. The concern arises when there are imperfections and less than competitive conditions either in the product market or in the labor market. If so, individuals and households do not supply the optimal amount of labor while employers do not utilize labor in the most efficient way. In this case total product is lower than it could be (inefficiency) while there are significant implications for the distribution of personal/household incomes (and poverty). Our study showed that women in the Latin American labor market are paid less than their male counterparts. This was not unexpected: so far we are unaware of any country study at any point in time that has come to an opposite conclusion when average female pay was compared with average male pay. As argued in the main text, the value of the present analysis rests on its comparison with other areas. In this respect it was established that a sizeable part of the gross wage differential between women and men in Latin America remains unaccounted for by some common economic characteristics. We repeatedly labelled the unexplained part as 'upper bound" of discrimination and we consistently stated that the true extent of discrimination should in all country cases be lower than its 'upper' bound. Still, the summary results presented in Table 6.2 and the detailed results presented in Appendix Table A6.1 are out of line with results from advanced countries. For example, the "discriminatory" part of the sex wage gap was found to be between half and three-quarters of the gross pay differential in Britain and the United States or even less.25 The country studies in the companion volume typically suggest that the discriminatory part in Latin America accounts for "three-quarters and up." It is possible that the present estimates of potential discrimination in Latin America are greater than those found for other countries because studies undertaken in other countries and world regions have used more accurate data 25 Wright and Ermisch (1991) for Britain and Killingsworth (1990) for the United States. Emnpirical Findings 199 or have accounted for more factors in deciding what determines earnings. This may be true. However, we expressed our skepticism about the practice to go on adding explanatory variables to the right hand side of the earnings equation in order to standardize for the difference between women's and men's pay. The inclusion of additional factors in the analysis (such as occupational or industrial types of employment) may bias the estimates of potential discrimination downward. Still, the uniformity of the results in Latin America with respect to the extent of potential discrimination is sufficiently clear to suggest that a greater part of the gross sex wage differential in Latin America (compared with results for other countries) is due to factors that are not immediately obvious. The identification of these factors (such as country specific legislation) requires a different and more in depth analysis than the regional study we have undertaken. One particular aspect of female pay in Latin America requires attention. This does not relate to the decomposition of the gender wage differential upon which our study has focused but on the pay differential per se. Sex wage differences in Latin America are by comparison to other regions small. The unweighed average of female relative pay comes to about 70-75 percent -- women workers even in many industrialized countries are still short of this figure in terms of their average remuneration in the labor market.' If this is taken at face value it may suggest that, though a good part of the pay gap in Latin America can be discriminatory in its origin, the labor market rewards women on average more like men than in other regions. We do not think so and the reasons are explained below. First, in our data sets there may be an overrepresentation of the earnings of workers engaged in activities in the formal sector. Such workers are more easily detected and included in databases than workers in the informal sector. In addition, the earnings of formal sector workers are less easily understated. The case of many informal sector workers reporting positive hours of work and zero labor incomes has already been discussed. As a result, our estimates may overstate the level of average female pay because women in the informal sector are not well represented (especially family workers). In contrast, in industrialized countries where most workers are employees such a bias either is smaller than in developing countries or does not arise at all. The implication is that, if women in the formal sector are overrepresented in our data sets and if reported pay is statistically greater in the formal sector than in the informal sector, then female average pay would be artificially greater than what a female worker would fetch on average in the formal and informal labor market. 26 Gunderson (1989). 200 Women's Employment and Pay in Latin America Second, and related to the previous point, women workers in the formal sector are primarily employed in the public sector. Consequently, not only are women in the formal sector overrepresented, but women workers in the formal sector tend to be dominated by public sector employees. The issue becomes whether Latin American women are paid more in the public sector than their counterparts in the private sector. An answer to this question is given in Table 6.4 (columns 1 and 2). To accommodate the fact that women employed in the public sector tend to be more educated than those in the private sector we broke down the information on pay by education level. The evidence suggests that women in the public sector are paid twice as much as women in the private sector - especially at lower levels of education. The difference tends to narrow at tertiary education level (save for Panama) but, as few women possess university qualifications, the average pay of all female workers is little affected by them. Third, we are not aware of any country in Latin America whose government or related organizations practice overt pay discrimination.' Consequently, one expects female pay relative to male pay in the public sector to be higher than that in the private sector. The last two columns in Table 6.4 show that women are paid more like men in the public sector than in the private sector -- indeed, women are paid more than men in the public sector in Guatemala. A more sophisticated way of establishing the role of the public sector in determining the average pay of women workers is through the use of earnings functions. As mentioned earlier in.this chapter women in the public sector enjoy a ceteris paribus pay premium (after controlling for other factors, such as human capital variables, hours worked and location). There is no statistical evidence that such a premium exists in the case of male workers. Consequently, the role of the public sector in Latin America may distort the overall estimates of female relative pay. In contrast, the difference between public and private sector pay is not that important in advanced market economies. If anything, the public sector in the latter group of countries is generally considered to be a low-pay employer because of the other non-pecuniary benefits it provides. Such benefits include job security, social benefits, longer holidays and better pensions. The foregoing provides some explanation why, somewhat unexpectedly, female relative pay in Latin America appears to be on the high side compared with women's relative pay elsewhere. This explanation has significant policy implications. In particular, it is probable that the public sector is paying 'distortionary' wages, that is, wages in excess of what a clearing labor market 27 It is, however, possible than indirect (employment) discrimination may be taking place in the sense that there is a prior expectation about the sex of workers in particular jobs and certain administrative ranks. Empirical Findings 201 Table 6.4 Female Wages (in local currency) and Female Relative to Male Wage (percent) in the Private and Public Sectors (selected countries) Educational Female Wages F/M Wage Country (pay) Level Private Public Private Public Guatemala (quetzalslhour) Primary 1.03 2.11 69 128 Secondary 2.03 4.07 79 127 Tertiary 3.90 4.26 72 96 Panama (balboas/hour) Primary 0.62 1.28 56 77 Secondary 1.52 2.13 85 77 Tertiary 1.03 1.83 58 71 Uruguay (pesos/hour) Primary 551 695 79 91 Secondary 646 853 71 94 Tertiary 1211 1268 56 89 Costa Rica (colon/month) All levels 10928 24954 66 91 Source: Country studies in the companion volume. would have established. The reasons for public sector pay being out of line with the competitive wage are well known. The public sector is not necessarily govemed to by immediate cost constraints but is relatively free to pay wages which conform to other considerations, even the satisfaction of group interests. Consequently, the public sector may not be a price taker. On the contrary, the public sector make act as a price maker given its size in terms of employment. In this respect, more detailed information and a deeper analysis is required before a concrete conclusion is reached. 5. Concluding Remarks Given the empirical results reported in this volume we believe that the labor market is quite well equipped to sort out a number of problems. We have noted the rise in female labor force participation which appears to have occurred during less than ideal macro-economic conditions. We also found that the employment distributions of women and men in the region have become more alike over time. In terms of wages women in Latin America appear to be paid relatively more than even in some advanced countries -- though a number of qualifications may apply. However, the percentage of the sex wage gap that is 202 Women's Employment and Pay in Latin America unaccounted by differences in the human capital characteristics held be women and men in Latin America is sizeable and greater than in other countries. We are not prepared to pinpoint the actual extent of wage (=price) discrimination against women in Latin America. But economics is not only about prices. Economics is also about constraints. Are the constraints that women workers and employers face "real" constraints? By "real"` we mean either genuine economic constraints (for example, limited consumer income or producers' resources) or "unavoidable' ones (such as the state of technology). In the case of real constraints nothing much can be done. However, if the constraints are "removable' and were indeed removed, then the economy could achieve a superior outcome. Such constraints may relate to labor market legislation or family law. They can also be the result of distortionary price setting by the public sector. Finally they may derive from limited rationality, myopic perceptions, capital market failure and externalities in the sphere of human capital. Many examples of these constraints have already been given in this volume. One feels certain that, if these constraints were removed, women would improve their position both at home and also in the labor market. Then inefficiency and poverty issues would become less acute in the region. Statistical Appendix to Chapter 6 The tables in this appendix show representative results from the country studies included in the companion volume. When means are reported, they are calculated as unweighted means of the countries for which information exits. Statistical Appendix 203 Appendix Table A6.1 Percentage of Male Pay Advantage Attributed to Differences in Endowments (E) and Rewards (R) Selectivity Uncorrected Selectivity Uncorrected' Evaluated at Evaluated at Male Pay Female Means Male Means Female Means Male Means Advantage . Country Year E R E R E R E R (1) (2) (3) (4) (5) (6) (7) (8) (9) Argentina 1985 43.2 22.0 78.0 32.0 68.0 26.0 74.0 38.0 62.0 Bolivia 1989 47.3 14.9 85.1 24.1 75.9 14.9 85.1 24.1 75.9 Brazil' 1989 35.7 - - - - 19.0 81.0 11.0 89.0 Chile 1987 33.8 - - - - -14.9 114.9 -13.7 113.7 Colombia 1988 16.7 12.3 87.7 22.1 77.9 8.0 92.0 14.8 85.2 Costa Rica 1989 21.3 -3.6 03.6 -3.2 103.9 5.5 94.5 6.7 93.3 Ecuador 1987 41.6 26.4 73.6 33.2 66.8 37.8 62.2 57.2 42.8 Guatemala 1989 26.4 -1.8 101.8 0.4 99.6 45.3 54.7 55.4 44.6 Honduras 1989 21.1 -69.2 169.2 -81.9 181.9 -50.6 150.6 -46.5 146.5 Jamaica 1989 55.1 - - - - -13.7 113.7 -19.1 119.1 Mexico 1984 15.7 - - - - 28.1 71.9 20.0 80.0 Panama 1989 22.1 -22.9 122.9 -40.6 140.6 13.9 86.1 14.7 85.3 Peru 1990 17.7 19.5 80.5 15.1 84.9 19.5 80.5 15.1 84.9 Uruguay 1989 29.5 24.0 76.0 26.0 74.0 23.0 77.0 23.0 77.0 Venezuela 1989 25.5 14.0 86.0 5.0 95.0 14.0 86.0 5.0 95.0 Averaged 30.2 3.2 96.8 2.9 97.1 11.7 88.3 13.7 86.3 - not available. a. Selectivity correction statistically insignificant in Argentina, Bolivia, Costa Rica, Peru and Venezuela. b. Measured in log-percentage points. It refers to hourly pay in Brazil, Ecuador and Peru and weekly/monthly pay in other countries. c. The figures for Brazil refer to married women working as employees. d. Unweighted average. Source: Based on results reported in the corresponding country cases in the companion volume. 204 Women's Employment and Pay in Latin America Appendix Table A6.2a Average Hours per Week and Coefficients on Log (hours) by Sex Male Male Average Hours Advantage Coefficient on Advantage in Country Year per Week in Hours Log (hours) Coefficients (percent) (percent) M F (1)/(2) M F (4)/(5) (1) (2) (3) (4) (5) (6) Argentina 1985 46.31 37.48 23.6 0.391 0.659 -40.7 Bolivia 1989 51.30 44.12 16.3 0.354 0.424 -16.5 Colombia 1988 49.90 46.10 8.2 0.426 0.458 -7.0 Costa Rica 1989 47.64 40.53 17.5 0.626 0.718 -12.8 Guatemala 1989 48.12 42.27 13.8 0.344 0.475 -27.6 Honduras 1989 45.73 43.75 4.5 0.301 0.438 -31.3 Panama 1989 42.76 40.14 6.5 0.660 0.600 10.0 Uruguay 1989 48.44 37.31 29.8 0.587 0.685 -14.3 Venezuela 1989 43.71 38.48 13.6 0.541 0.554 -2.3 Average 47.10 41.13 14.9 0.470 0.557 -15.8 Source: Based on results reported in the corresponding country cases in the companion volurne. Statistical Appendix 205 Appendix Table A6.2b Contribution of Differences in Hours to the Male Pay Advantage Evaluated at Total % of Male Female Means Male Means Effect Pay of Hours Advantage Country Year Effect Due to Effect Due to Upon Explained Differences in Differences in the Pay Gap by (7)+ (8) or Differences Endow. Coeff Endow. Coeff (9)+(10) in Hours (7) (8) (9) (10) (I11) (12) Argentina 1985 0.083 -0.971 0.139 -1.028 -0.888 -205.8 Bolivia 1989 0.053 -0.265 0.064 -0.276 -0.212 -44.8 Colombia 1988 0.034 -0.123 0.036 -0.125 -0.089 -53.2 Costa Rica 1989 0.101 -0.341 0.116 -0.355 -0.239 -112.2 Guatemala 1989 0.045 -0.490 0.062 -0.507 -0.446 -169.2 Honduras 1989 0.013 -0.518 0.019 -0.524 -0.504 -239.3 Panama 1989 0.042 0.222 0.038 0.225 0.263 119.3 Uruguay 1989 0.153 -0.355 0.179 -0.380 -0.201 -68.4 Venezuela 1989 0.069 -0.047 0.071 -0.049 0.021 8.4 Average 0.066 -0.321 0.080 -0.335 -0.255 -85.0 % of pay gap explained 23.4 -114.2 28.6 -119.4 -90.8 Source: Based on results reported in the corresponding country cases in the companion volume. 206 Women's Employment and Pay in Latn America Appendix Table A6.3a Average Years of Schooling and Estimated Coefficients on Schooling by Sex Male Male Average Years Advantage in Coefficient on Advantage in Country Year of Schooling Schooling Schooling Coefficients (percent) (xlOO) (percent) M F (1)/(2) M F (4)/(5) (1) (2) (3) (4) (5) (6) Argentina 1985 8.80 9.41 -6.5 9.1 10.7 -15.0 Bolivia 1989 9.510 8.97 5.9 7.1 6.3 12.7 Brazil 1980 4.86 6.96 -30.2 14.7 15.6 -5.8 Colombia 1988 7.60 8.70 -12.6 12.0 11.2 7.1 Costa Rica 1989 6.66 8.47 -21.4 10.1 13.1 -22.9 Ecuador 1987 9.70 9.05 7.2 9.7 9.0 7.8 Guatemala 1989 3.90 4.72 -17.4 14.3 16.4 -12.8 Honduras 1989 4.89 6.29 -22.3 15.4 17.8 -13.5 Jamaica 1989 7.37 7.84 -6.0 12.3 21.5 -42.8 Mexico 1984 6.26 7.56 -17.2 13.2 14.7 -10.2 Panama I989 9.21 10.45 -11.9 9.7 11.9 -18.5 Peru 1986 8.21 9.01 -8.9 11.5 12.4 -7.3 Uruguay 1989 8.34 9.06 -7.9 9.9 11.1 -10.8 Venezuela 1989 6.93 8.52 -18.7 9.1 11.1 -18.0 Average 7.30 8.22 -12.0 11.3 13.1 -10.7 Source: Based on results reported in the corresponding country cases in the companion volume. Statistical Appendix 207 Appendix Table A6.3b Contribution of Differences in Schooling to the Male Pay Advantage Evaluated at Total % of Male Female Means Male Means Effect Pay of Schooling Advantage Country Year Effect Due to Effect Due to Upon Explained Differences in Differences in the Pay Gap by (7) + (8) or Differences Endow. Coeff Endow. Coeff (9)+(10) in Schooling 7) (8) (9) (10) (11) (12) Argentina 1985 -0.141 -0.065 -0.151 -0.056 -0.206 -47.7 Bolivia 1989 0.076 0.033 0.072 0.038 0.109 23.2 Brazil 1980 -0.044 -0.328 -0.063 -0.309 -0.371 -130.2 Colombia 1988 0.061 -0.123 0.070 -0.132 -0.062 -37.4 Costa Rica 1989 -0.200 -0.237 -0.254 -0.183 -0.437 -204.7 Ecuador 1987 0.068 0.058 0.063 0.063 0.126 30.4 Guatemala 1989 -0.080 -0.134 -0.099 -0.117 -0.216 -82.1 Honduras 1989 -0.117 -0.249 -0.151 -0.216 -0.367 -173.9 Jamaica 1989 -0.678 -0.101 -0.721 -0.058 -0.779 -141.5 Mexico 1984 -0.094 -0.191 -0.113 -0.172 -0.285 -181.5 Panama 1989 -0.203 -0.148 -0.230 -0.120 -0.350 -158.6 Peru 1986 -0.074 -0.099 -0.081 -0.092 -0.173 -95.2 Uruguay 1989 -0.100 -0.080 -0.109 -0.071 -0.180 -61.1 Venezuela 1989 -0.139 -0.176 -0.170 -0.145 -0.315 -123.7 Average -0.111 -0.123 -0.129 -0.105 -0.234 -92.3 % of pay gap explained-40.4 -44.7 -47.0 -38.1 -85.1 Source: Based on results reported in the corresponding country cases in the companion volume. 208 Women's Employment and Pay in Latin America Appendix Table A6.4a Average Years of Potential Experience and Coefficients on Potential Experience by Sex Male Male Average Years Advantage in Coefficient on Advantage in Country Year of Potential Experience Experience Coefficients Experience (percent) (xlOO) (percent) M F (1)/(2) M F (4)/(5) (1) (2) (3) (4) (5) (6) Argentina 1985 24.19 21.30 13.6 4.9 3.8 28.9 Bolivia 1989 18.44 20.49 -10.0 5.0 2.8 78.6 Brazil 1980 26.31 21.01 25.2 4.2 3.9 7.7 Colombia 1979 7.04 5.56 26.7 2.5 2.2 13.6 Costa Rica 1989 22.45 19.10 17.5 3.5 3.1 12.9 Ecuador 1987 23.60 22.80 3.5 3.1 1.4 121.4 Guatemala 1989 24.90 22.16 12.4 4.5 4.1 9.8 Honduras 1989 23.81 21.33 11.6 5.2 5.0 4.0 Jamaica 1989 21.35 22.64 -5.7 7.7 8.2 -6.1 Mexico 1984 20.76 16.91 22.8 8.6 6.6 30.3 Panama 1989 20.36 18.36 10.9 7.9 10.3 -23.3 Peru 1986 19.22 15.86 21.2 5.5 7.6 -27.6 Uruguay 1989 24.47 22.48 8.9 5.8 4.2 38.1 Venezuela 1989 23.05 19.55 17.9 3.5 2.8 25.0 Average 21.4 19.3 12.6 5.1 4.7 22.4 Source: Based on results reported in the corresponding country cases in the companion volume. Statistical Appendix 209 Appendix Table A6.4b Contribution of Differences in Potential Experience to the Male Pay Advantage Evaluated at Total % of Male Femnale Means Male Means Effect Pay of Experience Advantage Country Year Effect Due to Effect Due to Upon Explained Differences in Differences in the Pay Gap by (7)+(8) or Differences Endow. Coeff Endow. Coeff (9)+(10) in Exper. (7) (8) (9) (10) (11) (12) Argentina 1985 0.266 0.110 0.234 0.142 0.376 87.1 Bolivia 1989 0.406 -0.057 0.451 -0.103 0.348 73.7 Brazil 1980 0.079 0.207 0.063 0.223 0.286 100.2 Colombia 1979 0.021 0.033 0.017 0.037 0.054 19.0 Costa Rica 1989 0.090 0.104 0.076 0.117 0.194 90.7 Ecuador 1987 0.401 0.011 0.388 0.025 0.412 99.2 Guatemala 1989 0.100 0.112 0.089 0.123 0.212 80.4 Honduras 1989 0.048 0.124 0.043 0.129 0.172 81.4 Jamaica 1989 -0.107 -0.106 -0.113 -0.099 -0.213 -38.6 Mexico 1984 0.415 0.254 0.338 0.331 0.669 426.3 Panama 1989 -0.489 0.206 -0.441 0.158 -0.283 -128.0 Peru 1986 -0.404 0.255 -0.333 0.185 -0.148 -81.4 Uruguay 1989 0.392 0.084 0.360 0.115 0.475 161.3 Venezuela 1989 0.161 0.098 0.137 0.123 0.259 101.8 Average 0.099 0.102 0.093 0.108 0.201 76.6 % of pay gap explained 32.6 33.9 30.9 35.5 66.4 Source: Based on results reported in the corresponding country-cases in the companion volume. 210 Women's Employment and Pay in Latin America Appendix Table A6.5 Contribution of Differences in the Constant Terms to the Male Pay Advantage % of Male Pay Male Pay Constant Term Difference in Advantage Country Year Advantage Male Female Constant Terms Explained (2)-(3) (4)/(1) (1) (2) (3) (4) (5) Argentina 1985 0.43 8.34 7.07 1.27 294.1 Bolivia 1989 0.47 1.58 1.35 0.23 48.7 Brazil 1980 0.29 2.40 1.75 0.65 227.9 Colombia 1988 0.17 5.66 5.66 0.00 0.0 Costa Rica 1989 0.21 4.53 3.69 0.84 393.6 Ecuador 1987 0.42 3.58 3.48 0.10 24.0 Guatemala 1989 0.26 2.01 0.97 1.04 394.7 Honduras 1989 0.21 1.25 0.33 0.92 436.5 Jamaica 1989 0.55 1.61 -0.44 2.05 372.3 Mexico 1984 0.16 6.66 6.58 0.08 51.0 Panama 1989 0.22 0.72 0.48 0.24 108.7 Peru 1990 0.18 2.10 1.78 0.32 180.5 Uruguay 1989 0.29 1.11 0.42 0.69 234.3 Venezuela 1989 0.25 3.92 3.52 0.40 157.1 Average 0.30 3.52 2.62 0.63 208.8 Source: Based on results reported in the corresponding country cases in the companion volume. Statistical Appendix 211 Appendix Table A6.6 The Value and Significance of the Coefficient on the Sample Selection Variable (Lambda) in the Earnings Functions Country Year Coefficient t-value Argentina 1985 -0.08 1.7 Bolivia 1989 0.07 1.2 Brazil 1980 -0.30 6.5 Chile 1987 -0.82 9.9 Colombia 1979 -0.09 1.3 Costa Rica 1989 -0.05 1.1 Ecuador 1987 0.03 0.5 Guatemala 1989 -0.29 7.5 Honduras 1989 -0.59 11.3 Jamaica 1989 -0.39 4.3 Mexico 1984 -1.45 6.7 Panama 1989 -0.39 12.3 Peru 1990 -0.05 1.0 Uruguay 1989 0.06 2.0 Venezuela 1989 -0.14 1.6 Source: Based on results reported in the corresponding country cases in the companion volume. Appendix A A companion to this Volume will be published and contains the following studies: 'Female Labor Force Participation and Gender Earnings Differentials in Argentina", by Y. C. Ng. 'Women in the Labor Force In Bolivia: Participation and Earnings", by K. Scott. "Labor Force Behavior and Earnings of Brazilian Women and Men, 1980", by M. Stelcner, J. B. Smith, J. A. Breslaw and G. Monette. 'Female Labor Force Participation and Wage Determination in Brazil, 1989", by J. Tiefenthaler. 'Is There Sex Discrimination in Chile? Evidence from the CASEN Survey", by I. Gill. "Labor Markets, the Wage Gap and Gender Discrimination: The Case of Colombia by J. Tenjo. "Female Labor Market Participation and Wages in Colombia", by T. Magnac. "'Women's Labor Force Participation and Earnings in Colombia", by E. Velez and C. Winter. "Female Labor Force Participation and Earnings Differentials in Costa Rica", by H. Yang. "Why Women Earn Less Than Men in Costa Rica", by T. H. Gindling. *The Effect of Education on Female Labor Force Participation and Earnings in Ecuador", by G. Jakubson and G. Psacharopoulos. 214 Women's Employment and Pay in Latin America 'Female Labor Participation and Earnings in Guatemala", by M. Arends. "Women's Labor Force Participation and Earnings in Honduras", by C. Winter and T. H. Gindling. "Female Labor Force Participation and Earnings: The Case of Jamaica", by K. Scott. "Women's Participation Decisions and Earnings in Mexico", by D. Steele. "Female Labor Force Participation and Wages: A Case Study of Panama", by M. Arends. "Women's Labor Market Participation and Male-Female Wage Differences in Peru", by S. Khandker. 'Is There Sex Discrimination in Peru? Evidence from the 1990 Lima Living Standards Survey", by I. Gill. "'Women's Labor Force Participation and Earnings: The Case of Uruguay", by M. Arends. "Female Participation and Earnings, Venezuela 1987", by D. Cox and G. Psacharopoulos. "Female Earnings, Labor Force Participation and Discrimination in Venezuela, 1989", by C. Winter. Appendix B The Authors of the Companion Volume Mary Arends is a Consultant for the World Bank's Latin America and Caribbean Technical Department, Human Resources Division. Jon A. Breslaw is Associate Professor in the Department of Economics, Concordia University, Montreal. Donald Cox is Associate Professor of Economics, Economics Department, Boston College at Chesnut Hill, Massachusetts. Indermit Gill is Assistant Professor in the School of Management at the State University of New York at Buffalo. T.H. Gindling is Assistant Professor in the Department of Economics, University of Maryland, Baltimore County. George Jakubson is Associate Professor in the School of Industrial Labor Relations at Cornell University. Shahidur Khandker is a Research Economist in the Women in Development Division, Population and Human Resources Department, The World Bank. Thierry Magnac is associated with INRA, ESR Paris, France and the Department of Economics, University College of London, United Kingdom. Georges Monette is Associate Professor in the Department of Mathematics, York University, Toronto. Ying Chu Ng is Assistant Professor at Hong Kong Baptist College. 216 Women's Employment and Pay in Latin America George Psacharopoulos is Senior Human Resources Advisor, Technical Department, Latin America and the Caribbean Region, The World Bank. Katherine Scott is a Consultant for the World Bank's Latin America and Caribbean Technical Department, Human Resources Division. J. Barry Smith is Associate Professor in the Department of Economics, York University, Toronto. Diane Steele is an Analyst at the American Institutes for Research, Washington D.C. Morton Stelcner is a Professor in the Department of Economics at Concordia University, Montreal. Jaime Tenjo is Assistant Professor in the Department of Management and Economics at the University of Toronto, Scarborough Campus. Jill Tiefenthaler is Assistant Professor in the Department of Economics at Colgate University. Zafiris Tzannatos is a Labor Economist in the Education and Employment Division, Population and Human Resources Department of the World Bank. Eduardo Velez is an Education Specialist in the Human Resources Division of the Latin American and Caribbean Region of the World Bank. Carolyn Winter is a Human Resources Specialist in the Women and Development Division, Population and Human Resources Department of the World Bank. Hongyu Yang is a Consultant in the Human Resources Division of the Latin American and Caribbean Region of the World Bank. References Adelman, I. "Development Economics: A Reassessment of Goals." American Economic Review, Vol. 65 Papers and Proceedings (1975). pp. 302-309. Ahluwalia, M.S., N. Carter, and H. Chenery. "Growth and Poverty in Developing Countries." Journal of Development Economics, Vol. 6, No.3 (1979). pp. 299-341. Aigner, D.J. and G.G. Cain. "Statistical Theories of Discrimination in the Labor Market". Industrial and Labor Relations Review, Vol. 30, No.2 (1977). pp. 175-187. Aird, J.S. "The Preliminary Results of China's 1982 Census." China Quarterly, Vol. 96 (1984). pp. 613-640. Amin, S. and A.R. Pebley. "The Impact of a Public Health Intervention on Excess Female Mortality in Punjab." Mimeograph. Princeton, New Jersey, 1987. Anker, R., M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. Altonji, J.G. "The Intertemporal Substitution Model of Labor Market Fluctuations: An Empirical Analysis." Review of Economic Studies, Vol. 49, No.5 (special issue) (1982). pp. 783-824. Arrow, K.J. "Higher Education as a Filter." Institute for Mathematical Studies in the Social Sciences. Technical Report No. 71. Palo Alto, California: Stanford University, 1972. 218 Women's Employment and Pay in Latin America Ashenfelter, 0. and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. Atkinson, A.B and N.H. Stern. "On the Switch from Direct to Indirect Taxation.' Journal of Public Economics, Vol. 14, No. 2 (1980). pp. 195- 224. Bardhan, P.K. Land, Labor and Rural Poverty: Essays in Development Economics. New York: Columbia University Press, 1984. Becker, G.S. The Economics of Discrimination. The University of Chicago Press, 1971 (2nd edition). -----. A Treatise on the Family. Cambridge, Mass.: Harvard University Press, 1981. Becker, G.S., E.M. Landes and R.T. Michael. "An Economic Analysis of Marital Instability." Journal of Political Economy, Vol. 85, No. 6 (1977). pp. 1141-1187. Becker, G.S. and N.Tomes. "An Equilibrium Theory of the Distribution of Income and Intergenerational Mortality." Journal of Political Economy, Vol. 87, No. 6 (1979). pp. 1141-1187. Behrman, J.R. and N. Birdsall. "The Quality of Schooling: Quantity Alone is Misleading." American Economic Review, Vol. 73, No. 5 (1983). pp. 928- 946. Behrman, J.R. and B.L. Wolfe. "Labor Force Participation and Earnings Determinants for Women in the Special Conditions of Developing Countries." Journal of Development Economics, Vol. 15 (1985). pp. 259-288. Behrman, J.R., B.L. Wolfe and D.M. Blau. "Human Capital and Earnings Distribution in a Developing Country: The Case of Prerevolutionary Nicaragua." Economic Development and Cultural Change. Vol. 34, No. 1 (1985). pp. 1-29. Beller, A.H. "Trends in Occupational Segregation by Sex and Race, 1960-1981" in B.F. Reskin (ed.). Sex Segregation in the Workplace: Trends, Explanations, Remedies. Washington DC: National Academy Press, 1984. pp. 11-26. References 219 Beller, A.H. and Han, K.K. "Occupational Sex Segregation" in B.F. Reskin (ed.). Sex Segregation in the Workplace: Trends, Explanations, Remedies. Washington DC: National Academy Press, 1984. pp. 91-114. Ben-Porath, Y. "Economic Analysis of Fertility in Israel: Point and Counterpoint." Journal of Political Economy, Vol. 81, No. 2 (Part 2) (1973). pp. S202-S233. Berg, I. Education and Jobs: The Great Training Robbery. Boston, Mass.: Beacon Press, 1971. Berger, M. and M. Buvinic (eds.). Women's Ventures: Assistance to the Informal Sector in Latin America. 1989. West Hartford, Conn.: Kumarian Press. Bergmann, B. "The Effect on White Incomes of Discrimination in Employment." Journal of Political Economy, Vol. 79, No. 2 (1971). pp. 294-313. Bernard, J. Women and the Public Interest. Chicago: Aldine-Atherton, 1971. Birdsall, N. and J.R. Behrman. "Does Geographical Aggregation Cause Overestimates of the Return to Schooling?" Washington DC: The World Bank, DEDPH, 1983. Birdsall, N. and M.L. Fox. "Why Males Earn More: Location and Training of Brazilian Schoolteachers". Economic Development and Cultural Change. Vol. 33, No. 3 (1985). pp. 533-556. Birdsall, N. and R. Sabot (eds.). Unfair Advantage: Labor Market Discrimination in Developing Countries. The World Bank, 1991. Bishop, J. "Jobs, Cash Transfers and Marital Instability: A Review and Synthesis of the Evidence." Journal of Human Resources. Vol. 15, No. 2 (1980). pp. 301-334. Blanchard, O.J. and L.H. Summers. "Hysteresis and the European Unemployment Problem" in R. B. Cross (ed.). Unemployment, Hysteresis, and the Natural Rate Hypothesis. Oxford: Basil Blackwell, 1988. pp. 306- 364. Blau, D.M., J.R. Behrman and B.L. Wolfe. "Schooling and Earnings Distributions with Endogenous Labour Force Participation, Marital Status and Family Size." Economica, Vol. 55, No.219, (1988). pp. 297-316. 220 Women's Employmenr and Pay in Latin America Blau, F.D. "Occupational Segregation and Labor Market Discrimination" in B.F. Reskin (ed.). Sex Segregation in the Workplace: Trends, Explanations, Remedies. Washington DC: National Academy Press, 1984. pp. 11-26. Blau, F.D. and W.E. Hendricks. "Occupational Segregation by Sex: Trends and Prospects." Journal of Human Resources, Vol. 14, No. 2 (1979). pp. 197- 210. Blinder, A.S. 'Wage Discrimination: Reduced Form and Structural Estimates". Journal of Human Resources, Vol. 8, No.4 (1973). pp. 436-455. -----. Toward an Economic Theory of Income Distribution. Cambridge Mass.: MIT Press, 1974. Blinder, A.S. and Y. Weiss. "Human Capital and Labor Supply: A Synthesis". Journal of Political Economy, Vol. 84, No. 3 (1976). pp. 449-472. Blumberg, R.L. "A General Theory of Gender Stratification" in R. Collins (ed.). Sociological Theory. San Francisco CA: Jossey Bass, 1984. pp. 23-101. Blundell, R. and I. Walker. "Modelling the Joint Determination of Household Labour Supplies and Commodity Demands." Economic Journal, Vol. 92, No. 366 (1982). pp. 351-364. Borjas, G.J. "Self-selection and the Earnings of Immigrants". American Economic Review. Vol. 77, No.4 (1987). pp. 531-553. Boserup, E. Woman's Role in Economic Development. New York: St. Martin's Press, 1970. Bowers, J.K. "British Activity Rates: A Survey of Research." Scottish Journal of Political Economy, Vol. 22, No.1 (1975). pp. 57-90. Bowles, S. "Aggregation of Labor Inputs in the Economics of Growth and Planning: Experiments with a Two-Level CES Function." Journal of Political Economy, Vol. 78 (1970). pp. 68-8 1. Brown, C.V., E. Levin, and D.T. Ulph. "Estimation of Labour Hours Supplied by Married Male Workers in Great Britain." Scottish Journal of Political Economy, Vol. 23, No. 3 (1976). pp. 261-277. References 221 Brown, R.S., M. Moon and B.S. Zoloth. 'Incorporating Occupational Attainment in Studies of Male-Female Earnings Differentials". Journal of Human Resources. Vol. 15, No. 1 (1980a). pp. 3-28. -----. "Occupational Attainment and Segregation by Sex." Industrial and Labor Relations Review, Vol. 33, No. 4 (1980b). pp. 506-517. Bustillo, I. "Female Educational Attainment in Latin America: A Survey." Washington DC: The World Bank (PHREE/89/16), 1989. Buvinic, M, M.A. Lycette and W.P. McGreevey (eds.). Women and Poverty in the Third World. Baltimore: Johns Hopkins University Press, 1983. Cain, G.G. Married Women in the Labor Force. Chicago: University of Chicago Press, 1966. 'The Economic Analysis of Labor Market Discrimination: A Survey" in 0. Ashenfelter and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. Cain, M., S.R. Khanam and S. Nahar. "Class, Patriarchy and Women's Work inBangladesh." Population andDevelopment Review, Vol. 5, No.3 (1979). pp. 405-438. Carline, C.A. Pissarides, W.S. Siebert and P.J. Sloane. Labour Economics. London: Longman, 1985. Chen, L., E. Hug and S. D'Souza. 'Sex Bias in the Family Allocation of Food and Health Care in Rural Bangladesh." Population and Development Review, Vol. 7, No. 1 (1981). pp. 55-70. Chiplin, B. and P.J. Sloane. Sex Discrimination in the Labour Market. 1976. London: Macmillan. Chiswick, B.R. Income Inequality: Regional Analyses within a Human Capital Framework New York: Colombia University Press, 1974. Clark, C. The Conditions of Economic Progress. London: Macmillan, 1940. Clark, K.B. and L.H. Summers. "Labor Force Participation: Timing and Persistence." Review of Economic Studies, Vol. 49, No. 5 (special issue) (1982). pp. 825-44. 222 Women's Employment and Pay in Latin America Cochrane, S.H. Fertility and Education: What Do We Really Know? Baltimore: Johns Hopkins University Press, 1979. Cogan, J. "Labor Supply with Costs of Labor Market Entry" in J.P. Smith (ed.). Female Labor Supply: Theory and Estimation. Princeton NJ: Princeton University Press, 1980a.pp. 327-364. "Married Women's Labor Supply: A Comparison of Alternative Estimates" in J.P. Smith (ed.). Female Labor Supply: Theory and Estimation. Princeton NJ: Princeton University Press, 1980b. pp. 90-118. Colclough, C. "The Impact of Primary Schooling on Economic Development: A Review of the Evidence." World Development, Vol 10, No. 3 (1982). pp. 167-185. Coombs, P. H. The World Educational Crisis: A System Analysis. New York: Oxford University Press, 1968. Comia, G.A., R. Jolly and F. Stewart. Adjustment with a Human Face. Oxford: Clarendon Press. 1987. Cross, R.B. and Allen in Cross (ed.). Unemployment, Hysteresis and the Natural Rate Hypothesis. Oxford: Basil Blackwell, 1988. Da Silva, L. "Greater Opportunities for Women Related to Population Growth." Working Paper No. 11, Belo Horizonte, Brazil: Federal University of Minas Gerais, Center for Development and Regional Planning, 1982. Da Vanzo, J. 'Family Formation in Chile." Santa Monica: Rand Corporation, R830 AID, 1971. De Tray, D.N. "Child Quality and Demand for Children." Journal of Political Economy, Vol. 81, no 2 (Part 2) (1973). pp. 70-95. ------. The Interaction Between Parents Investment in Children and Family Size: An Economic Analysis." Santa Monica: Rand Corporation, R1OOORE, 1972. Deere, C.D., J. Humphries and M. Leon de Leal. "Class and Historical Analysis for the Study of Women and Economic Change" in R. Anker, M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. pp. 87-115. References 223 Dex, S. and P. Sloane. "Detecting and Removing Discrimination at the Market Place." Journal of Economic Surveys, Vol. 2, No. 1 (1988). pp. 1-28. Dolton, O.J. and G.H. Makepeace. "Interpreting Sample Selection Effects". Economic Letters. Vol. 24, No. 4 (1987). pp. 373-379. Dooley, M. D. "Labor Supply and Fertility of Married Women: An Analysis with Grouped and Individual Data from the 1970 U.S. Census." Journal of Human Resources, Vol. 17, No. 4 (1982). pp. 499-532. Dougherty, C.R. S. "Estimates of Labor Aggregation Functions." Journal of Political Economy, Vol. 80, No. 6 (1972). pp. 1101-1119. Dougherty, C.R.S. and E. Jimenez. "The Specification of Earnings Functions: Tests and Implications". Economics of Education Review. Vol. 10, No. 2. (1991) pp.85-98. Dougherty, C. R.S. and M. Selowsky. "Measuring the Effects of the Misallocation of Labor." Review of Economics and Statistics, Vol. 55, No. 3 (1973). pp. 386-390. Duncan, G.J. Years of Poverty, Years of Plenty. Ann Arbor, Mich.: Institute for Social Research, 1984. Duncan, G.M. and D.E. Leigh. "The Endogeneity of Union Status: An Empirical Test". Journal of Labor Economics. Vol. 3, No. 3 (1985). pp. 385402. Duncan, 0. and B. Duncan. "A Methodological Analysis of Segregation Indices." American Sociological Review, Vol. 20, April (1955). pp. 210- 217. Easterlin, R.A. Population, Labor Force and Long Swings in Economic Growth: The American Experience. New York: National Bureau of Economic Growth, 1968. ------. Toward a Socio-Economic Theory of Fertility: A Survey of Recent Research on Economic Factors in American Fertility' in S.J. Behrman, L. Corsa Jr. and R.L. Freedman (eds.). Fertility and Family Planning: A World View. Ann Arbor: University of Michigan Press, 1969. pp. 127-156. 224 Women's Employment and Pay in Latin America ECIEL. Determinantes de la Oferta de Trabajo en America Latina, Programa de Estudios Conjunctos sobre Integracion Economica Latinoamerica. Rio de Janeiro, Brazil, 1982. ECLAC. (United Nations Economic Commission for Latin America and the Caribbean). Los grandes cambios y las crisis. Impacto sobre la mujer en America Latina y el Caribe. Santiago, Chile. December 1990 -----. Women in Latin America and the Caribbean: The Challenge of Changing Production Patterns with Social Equity. Fifth Regional Conference on the Integration of Women into the Economic and Social Development of Latin America and the Caribbean. Curacao, Netherlands Antilles. 16-19 September 1991. Ehrenberg, R.G. and J.L. Schwarz. "Public-Sector Labor Markets" in 0. Ashenfelter and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. pp. 1219-1259. England, P. "Assessing the Trends in Occupational Sex Segregation, 1900-1976" in I. Berg (ed.). Sociological Perspectives on Labor Markets. New York: Academic Press, 1981. pp. 273-295. -----. "The Failure of Human Capital Theory to Explain Occupational Segregation by Sex. " Journal of Human Resources, Vol. 17, No. 3 (1982). pp. 358-370. Ermisch, J.F. and R.E. Wright. "Entry to Lone Parenthood: Analysis of Marital Dissolution." Discussion Paper in Economics No. 9/90. University of London: Birkbeck College, 1990. Fallon, P. and D. Verry. The Economics of Labour Markets. Oxford: Philip Alan, 1988. Finn, M. and C.L. Jusenius. "The Position of Women in the Ecuadorian Labor Force". Ohio State University: Center for Human Resources Research, 1975. Frank, R.H. "Why Women Earn Less: The Theory and Estimation of Differential Overqualification." American Economic Review, Vol. 68, No. 3, (1978). pp. 360-373. References 225 Fuchs, V.R. "Recent Trends and Long-Run Prospects for Female Earnings". American Economic Review, Vol. 64, (Proceedings) (1974). pp. 236-42. Goldthorpe, J. and K. Hall. The Social Grading of Occupations. Oxford: Clarendon, 1974. Goode, W.J. After Divorce. Glencoe, Ill.: Free Press, 1956. -----. "Marital Satisfaction and Instability: A Cross-Cultural Class Analysis of Divorce Rates." International Social Sciences Journal, Vol. 14, No. 3 (1962). pp. 507-526. "Why Men Resist" in M. Yalom (ed.). Rethinking the Fanily: Some Feminist Questions. New York: Longman, 1982. pp. 131-150. Grasso, J.T. and J.R. Shea. "Effects of Vocational Education Programs: Research Findings and Issues" in Planning Papers for Vocational Study, Vocational Education Study Publication No. 1. Washington D.C.: USGPO: National Institute for Education, 1979. pp. 101-194. Greene, William. Econometric Analysis, McMillan, 1990. Greenhalgh, C. "Male-Female Wage Differentials in Great Britain: Is Marriage an Equal Opportunity?" Economic Journal, Vol. 90, No. 360 (1980). pp. 651-675. Gregory, R.G. and R.C. Duncan. "Segmented Labour Market Theories and the Australian Experience of Equal Pay for Women." Journal of Post-Keynesian Economics, Vol. 3, No. 3 (1981). pp. 403-428. Griliches, Z. "Estimating the Returns to Schooling: Some Econometric Problems." Econometrica, Vol. 45, No. 1 (1977). pp. 1-22. Gronau, R. "The Intrafamily Allocation of Time: The Value of the Housewives' Time." American Economic Review, Vol. 63, No.4 (1973). pp. 634-651. ------. Sex-Related Wage Differentials and Women's Interrupted Labor Careers: The Chicken and Egg Question." Cambridge, Mass: National Bureau of Economic Research, Working Paper No.1002, 1982. ------. Wage Comparisons - A Selectivity Bias." Journal of Political Economy, Vol. 82, No. 6 (1974). pp. 1119-43. 226 Women's Employment and Pay in Latin America Gunderson, M. 'Male-Female Wage Differentials and Policy Responses." Journal of Economic Literature, Vol. 27, No. 1 (1989). pp. 46-117. Hall, M. and N. Tideman. "Measures of Concentration'. Journal of American Statistical Association, Vol. 62 (1967). pp. 162-168. Hamermesh, D.S. "The Demand for Labor in the Long Run" in 0. Ashenfelter and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. pp. 429-471. Haque, N. "Work Status Choice and the Distribution of Family Earnings." Santa Monica: Rand Corporation, 1984. Harman, A. "Interrelationship Between Procreation and Other Family Decision- Making." Santa Monica: Rand Corporation, 1969. Hashimoto, M. and L. Kochin. "A Bias in the Statistical Estimation of the Effects of Discrimination." Economic Enquiry, Vol. 18 No. 3 (1980). pp. 478-486. Haveman, R. and B. Wolfe. "Education and Economic Well Being." in E. Dean (ed.) Education and Economic Productivity, Cambridge, Mass: Harper and Row, 1984. pp. 19-55. Heckman, J.J. "Sample Selection as a Specification Error". Econometrica, Vol 47 No.1 (1979). pp. 153-161. ------. Shadow Prices, Market Wages and Labor Supply." Econometrica, Vol. 42, No. 4 (1974). pp. 679-94. Heckman, J.J. and J. Hotz. "An Investigation of the Labor Market Earnings of Panamanian Males: Evaluating Sources of Inequality.' Journal of Human Resources, Vol. 21, No. 4 (1986), pp. 507-542. Heckman, J.J. and T. MaCurdy. "A Life Cycle Model of Female Labor Supply." Review of Economic Studies, Vol. 47, No. 1 (1980). pp. 47-74. -. "Corrigendum on a Life Cycle Model of Female Labor Supply.' Review of Economic Studies, Vol. 49, No. 4 (1982). pp. 659-660. References 227 Heckman, J.J. and R.J. Willis. "A Beta-Logistic Model for the Analysis of Sequential Labor Force Participation of Married Women." Journal of Political Economy, Vol. 85, No. 1 (1977). pp. 27-58. Heer, D.M. and E. Turner. "A Real Difference in Latin American Fertility." Population Studies, Vol. 18, No. 3 (1965). pp. 279-292. Hepple, B.A. Equal Pay and the Industrial Tribunals. Oxford: Martin Robinson, 1984. Hill, A.M. "Female Labor Force Participation in Developed and Developing Countries: Consideration of the Informal Sector." Review of Economics and Statistics, Vol. 65 (1983). pp. 459-468. Hofstede, G. Culture's Consequences: InternationalDifferences in Work-Related Values. London: Sage Publishers, 1980. Homey, M.J. and M.B. McElroy. "The Household Allocation Problem: Empirical Results from a Bargaining Model" in T. Paul Schultz (ed.). Research in Population Economics, Vol. 6. Greenwich, CT.: JAI Press, 1988. Hornstein, Z., J. Grice and A. Webb. (eds.). The Economics of the Labour Market. London: HMSO, 1982. Humphries, J. "Women's Employment in Restructuring America: The Changing Experience of Women in Three Recessions" in J. Rubery (ed.). Women and the Recession. London: Routledge, 1988. pp. 15-45. Iglesias, F.H. and M. Riboud. "Trends in Labor Force Participation of Spanish Women: An Interpretive Essay." Journal of Labor Economics, Vol. 3, No. 1 (Part 2) (1985). pp. S201-S217. ILO. World Labour Report. Geneva: International Labour Office, 1987. ILO. Yearbook of Labour Statistics: Retrospective Edition on Population Censuses 1945-1989. Geneva: International Labour Organization, 1990. Jelin, E. "Women and the Urban Labor Market" in R. Anker, M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. pp. 239-267. 228 Women's Employment and Pay in Latin America Jones, F.L. "On Decomposing the Wage Gap: A Critical Comment on Blinder's Method". Journal of Human Resources. Vol. XVIII, No. I (1983). pp. 126- 130. Joseph, G. Women at Work- The British Experience. Oxford: Philip Alan, 1983. Joshi, H. "Secondary Workers in the Employment Cycle: Great Britain, 1961- 1974". Economica, Vol. 48, No. 189 (1981). pp. 2944. Joshi, H., R. Layard and S. Owen. "Why Are More Women Working in Britain?" Journal of LaborEconomics, Special Issue (1985). pp. S147-S 176. Joshi, H. and S. Owen. "How Long is a Piece of Elastic? The Measurement of Female Activity Rates in British Censuses 1951-1981." London: Centre for Economic Policy Research, Discussion Paper No. 31., 1984. ----" -Does Elastic Retract? The Effect of Recession on Women's Labour Force Participation". London: Centre for Economic Policy Research, Discussion Paper No. 64., 1985. Kanowitz, L. Women and the Law: The Unfinished Revolution. Albuquerque: University of New Mexico, 1969. Keeley, M.C. Labor Supply and Public Policy: A Critical Review. New York: Academic Press, 1981. Kelly, A. and L. Da Silva. "The Choice of Family and the Compatibility of Female Workforce Participation in the Low-Income Setting." Review Economique, Vol. 31, No. 6 (1980). pp. 1081-1104. Khan, S. and M. Irfan. 'Rates of Returns to Education and the Determinants of Earnings in Pakistan." Pakistan Development Review, Vol. 24, No. 3/4 (1985). pp. 671-680. Khandker, S.R. "Determinants of Women's Time Allocation in Bangladesh." Economic Development and Cultural Change, Vol. 37, No. 1 (1988). pp. 111-126. -----. "Women's Role in Household Productive Activities and Fertility in Bangladesh." Journal of Economic Development, Vol. 12, No. 1 (1987). pp. 87-115. References 229 Kiernan, K. "Teenage Marriage and Marital Dissolution: A Longitudinal Study.' Population Studies, Vol. 40, No. 1 (1986). pp. 35-54. Killingsworth, M.R. Labor Supply. Cambridge: Cambridge University Press, 1983. -----. The Economics of Comparable Work. Kalamazoo Michigan: W.E Upjohn Institute for Employment Research, 1990. Killingsworth, M.R. and J.J. Heckman. "Female Labor Supply: A Survey" in 0. Ashenfelter and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. pp. 103-204. Knight, J.B. and R. Sabot. "Labor Market Discrimination in Tasmania". Journal of Development Studies. Vol. 19, No.1 (1982). pp. 67-87. Kozel, V. and H. Alderman. "Factors Determining Work Participation and Labor Supply Decisions in Pakistan's Urban Areas." Paper presented at the 5th Annual General Meeting of the Pakistan Society for Development Economists, March 1988. Kuznets, S. "Economic Growth and Income Inequality". American Economic Review, Vol. 45 (1955). pp. 1-28. ---. Economic Growth of Nations. Cambridge, Mass.: Harvard University Press, 1971. -----. 'Quantitative Aspects of the Economic Growth of Nations: Industrial Distributionof National Product and Labor Force." EconomicDevelopment and Cultural Change, Vol. 5, No. 1 (Supplement) (1956), pp. 1-94. La Belle, T.J. Nonformal Education in Latin America and the Caribbean. Stability, Reform or Revolution? New York: Praeger, 1986. Layard, R. and J. Mincer (eds.). "Trends in Women's Work, Education, and Family Building." Journal of Labor Economics, Vol. 3, No. 1 (Part 2) (1985). Lee, L.F. "Some Approaches to the Correction of Selectivity Bias". Review of Economic Studies. Vol. 49, No. 3 (1982). pp. 355-372. 230 Women's Employment and Pay in Latin America -----. "Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables". International Economic Review. Vol. 19, No. 2 (1978). pp. 415-433. Lehrer, E.L. and H. Stokes. "Determinants of the Female Occupational Distribution: A Log-Linear Probability Analysis.' Review of Economics and Statistics, (1985). pp. 395-404. Leibowitz, A. "Home Investment in Children." Journal of Political Economy, Vol. 82, No. 2 (Palt 2) (1974). pp. Sll1-S131. Levine, V. and P.R. Moock. "Labor Force Experience and Earnings: Women and Children". Economics of Education Review. Vol. 3, No. 3 (1984). pp. 183-194. Levy, F. Dollars and Dreams: The Changing American Income Distribution. 1987. New York: Sage Foundation for the National Committee for Research on the 1980 Census. Little, R.J.A. "A Note About Models for Selectivity Bias". Econometrica. Vol. 53, No. 6 (1985). pp. 1469-1474. Lloyd, C.B. and B.T. Niemi. The Economics of Sex Differentials. New York: Columbia University Press, 1979. Lockheed, M.E. and E. Hanushek. "Improving Educational Efficiency in Developing Countries." Compare, Vol. 18, No. 1 (1988) pp. 21-38. Lucas, R.E. and L. Rapping. "Real Wages, Employment and Inflation". Journal of Political Economy, Vol. 77, No. 5 (1969). pp. 751-754. Lundberg, S. "The Added Worker Effect". Journal of Labor Economics, Vol. 3, No. 1 (1985). pp. 11-37. MaCurdy, T. "An Empirical Model of Labor Supply in a Life-Cycle Setting." Journal of Political Economy, Vol. 89, No. 6 (1981). pp. 1059-1085. Main, B.G.M. and P. Elias. "Women Returning to Paid Employment." International Review of Applied Economics, Vol. 1, No. 1 (1987). pp. 86-108. References 231 Malkiel, B.G. and J. A. Malkiel. "Male-Female Pay Differentials in Professional Employment." American Economic Review, Vol. 63, No. 4 (1973). pp. 693-705. Martin, J. and C. Roberts. Women and Employment: A Lifetime Perspective. London: HMSO, 1984. Martorell, R., J. Leslie and P. Moock. "Characteristics and Determinants of Child Nutritional Status in Nepal." American Journal of Clinical Nutrition, Vol. 39, No. 1 (1984). pp. 74-86. Mason, K.O. "Strober's Theory of Occupational Sex Segregation" in B. F. Reskin (ed.). Sex Segregation in the Workplace: Trends, Explanations, Remedies. Washington DC: National Academy Press, 1984. pp.157-170. Meyer, R.H. and D.A. Wise. "High School Preparation and Early Labor Force Experience" in R.B. Freeman and D.A. Wise (eds.). The Youth Unemployment Problem: Its Nature, Causes and Consequences. Chicago: National Bureau of Economic Research, 1982. pp. 277-339. Michael, R.T. "Education and Derived Demand for Children." Journal of Political Economy in T.W. Schultz (ed.). The Economics of the Family. Chicago: University of Chicago Press, 1974. pp. 120-156. ------ Education in Nonmarket Production." Journal of Political Economy, Vol. 92, No. 2 (1984). pp. 306-327. Miller, P.W. "The Wage Effect of Occupational Segregation of Women in Britain". Economic Journal. Vol. 97 , No. 388 (1987). pp. 885-896. Mincer, J. "Labor Force Participation of Married Women: A Study of Labor Supply." Aspects of Labor Economics, National Bureau of Economic Research. Princeton NJ.: Princeton University Press, 1962. pp. 63-97. Mincer, J. Schooling, Experience and Earnings. National Bureau of Economic Research. New York: Columbia University Press, 1974. Mincer, J. and H. Ofek. "Family Investments in Human Capital: Earnings of Women." Journal of Political Economy, Vol. 82, No. 2 (Part 2) (1974). pp. S76-S108. 232 Women's Employment and Pay in Latin America Mincer, J. and S.W. Polachek. "Family Investments in Human Capital: Earnings of Married Women.' Journal of Political Economy, Vol. 82, No. 2 (Part 2) (1974). pp. S76-S110. Mingat, A. and J.P. Tan. Analytical Tools for Sector Work in Education. Baltimore: Johns Hopkins University Press, 1988. Mohan, R. Work, Wages and Welfare in a Developing Metropolis. New York: Oxford University Press, 1986. Mueller, E. "The Allocation of Women's Time and Its Relation to Fertility" in R. Anker, M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. pp. 87-115. Nakamura, A., N. Nakamura and D. Cullen. Employment and Earnings of Married Females. Ottawa: Ministry of Supplies and Services (1979a). -----. "Job Opportunities, the Offered Wage, and the Labor Supply of Married Women." American Economic Review, Vol. 69, No. 5 (1979b). pp. 787-805. Nelson, F. "Efficiency of the Two-Step Estimator for Models With Endogenous Sample Selection". Journal of Econometrics. Vol. 24, No. 1 (supplement) (1984). pp. 181-196. Ngwira, N. "Effectiveness of Agricultural Extension Service in Reaching Rural Women with Timely and Appropriate Agricultural Information." University of Malavwi: Center for Social Research, 1987. Nickell, S.J. "Fixed Costs, Employment and Labour Demand Over the Cycle". Economica. Vol 45, No. 180 (1978) pp. 329-346. ----" -A Picture of Male Unemployment in Britain." Economic Journal, Vol. 90, No. 360 (1980). pp. 776-795. Oaxaca, R. "Male-Female Wage Differentials in Urban Labor Markets." International Economic Review, Vol. 14, No. 1 (1973). pp. 693-709. ------. Sex Earnings Differentials" in G. Psacharopoulos (ed.). Economics of Education: Research and Studies. New York: Pergamon Press, 1987. References 233 Oberoi, A.S. Changes in the Structure of Employment with Economic Development. Geneva: ILO. 1978 O'Farrell, B. "Women and Nontraditional Blue Collar Jobs in the 1980s: An Overview" in P. A. Wallace (ed.). Women in the Workplace, (1982). pp. 135-165. Oi, W. "Labor as a Quasi-Fixed Factor". Journal of Political Economy. Vol. 70, No. 6 (1962). pp. 538-555 Olsen, R.J. "A Least Squares Correction for Selectivity Bias'. Econometrica. Vol. 48, No. 7 (1980). pp.1815-1820. Oppenheimer, V. The Female Labor Force in the United States: Demographic and Economic Factors Governing its Growth and Changing Composition. Berkeley, 1974. Oswald, A.J. "The Economic Theory of Trade Unions: An Introductory Survey." Scandinavian Journal of Economics, Vol. 87, No. 2 (1985). pp. 160-193. Pencavel, J. "Labor Supply of Men: A Survey" in 0. Ashenfelter and R. Layard (eds.). Handbook ofLabor Economics. Amsterdam: North Holland, 1986. pp. 3-101. Peters, H.E. "Marriage and Divorce: Informational Constraints and Private Contracting." American Economic Review, Vol. 76, No. 3 (1986). pp. 437- 454. Phelps, E.S. "The Statistical Theory of Racism and Sexism". American Economic Review. Vol. 62, No. 4 (1972). pp. 659-661. Pike, M. "Segregation by Sex, Earnings Differentials and Equal Pay: An Application of the Crowding Model." Applied Economics, Vol. 14, No. 5 (1982). pp. 503-514. Polachek, S.W. "Potential Biases in Measuring Male-Female Discrimination". Journal of Human Resources, Vol. 10, No. 2 (1975). pp. 205-229. Preston, S. "Health Programs and Population Growth." Population and Development Review, Vol. 20, No. 1 (1975). pp. 189-199. 234 Women's Employment and Pay in Latin America Psacharopoulos, G. (ed). Essays on Poverty, Equity and Growth. New York: Pergamon Press, 1991. 'The Perverse Effects of Public Subsidization of Education." Comparative Education Review, Vol. 21, No. 1 (1977). pp. 69-90. -----. "Returns to Education: A Further International Update and Implications." Journal of Human Resources, Vol. 14, No. 4 (1985). pp. 584-604. 'To Vocationalize or Not to Vocationalize? That is the Curriculum Question." International Review of Education, Vol. 33, No. 2 (1987). pp. 187-211. Psacharopoulos, G. and M. Hinchliffe. "Further Evidence on the Elasticity of Substitution Between Different Types of Educated Labor." Journal of Political Economy, Vol. 80, No. 4 (1972). pp. 786-792. Psacharopoulos, G. and R. Layard. "Human Capital and Earnings: British Evidence and a Critique." Review of Economic Studies, Vol. 46, No. 3 (1979). pp. 485-503. Psacharopoulos, G. and Z. Tzannatos. "Education and Female Labor Force Participation". Paper presented at the Sixth World Congress of Comparative Education, Rio de Janeiro, July 6-10. 1987. "Female Labor Force Participation: An International Perspective". The World Bank Research Observer. Vol. 4, No. 2 (1989). pp. 187-202. ----- "Female Labor Force Participation and Education" in G. Psacharopoulos (ed.) Essays on Poverty, Equity and Growth, Pergamon Press for the World Bank, (1991). Reskin, B. F. (ed.). Sex Segregation in the Workplace: Trends, Explanations, Remedies. Washington D.C.: National Academy Press, 1984. Ribich, T. Education and Poverty. Washington D.C.: Brookings Institute, 1968. Roberts, H.V. "Statistical Biases in the Measurement of Employment Discrimination" in E. R. Livernash (ed.). Comparable Worth: Issues and Alternatives. Washington DC.: Equal Employment Advisory Service, 1980. pp. 173-196. References 235 Rosen, H.S. "On Interindustry Wage and Hours Structure." Journal of Political Economy, Vol. 77, No. 2 (1969). pp. 249-73. 'Taxes in a Labor Supply Model with Joint Wage-Hours Determination." Econometrica, Vol. 44, No. 3 (1976). pp. 485-507. Rosenhouse, S. "Identifying the Poor: Is Headship a Useful Concept?" New York: Population Council, 1988. Rosenzweig, M.R. and K.I. Wolpin. "Govermnent Interventions and Household Behaviour in a Developing Country." Journal of Development Economics, Vol. 10, No. 2 (1982). pp. 209-225. Rubery, J. (ed.). Women and the Recession. London: Routledge, 1988. Safilios-Rothschild, C. "Female Power, Autonomy and Demographic Change in the Third World" in R. Anker, M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. pp. 239-267. Sahn, D. and H. Alderman. "The Effects of Human Capital on Wages and the Determinants of Labor Supply in a Developing Country." Journal of Development Economics, Vol. 28, No. 2 (1988). pp. 157-173. Sapsford, D. and Z. Tzannatos (eds.). Current Issues in Labour Economics. Houndmills: Macmillan, 1990. Schultz, T.P. "An Economic Model of Family Planning and Fertility." Journal of Political Economy, Vol. 77, No. 2 (1969a). pp. 153-180. "Retums to Women's Education." Washington DC: The World Bank (PHRWD/89/001), 1989b. -.-- "School Expenditures and Enrollments, 1960-1980" in D.G. Johnson and R. Lee (eds.). Population Growth and Economic Development. Madison, WI.: University of Wisconsin Press, 1987. "Secular Trends and Cyclical Behaviour of Iacome Distribution in the U.S." in L. Soltow (ed.). Six Papers on the Size Distribution of Wealth and Income. New York: Columbia University Press, 1969b. 236 Women's Employment and Pay in Latin America 'Women's Changing Participation in the Labor Force: A World Perspective". Economic Development and Cultural Change, Vol. 38, No. 3 (1990). pp. 457-488. 'Women and Development: Objectives, Frameworks, and Policy Interventions." Washington D.C.: The World Bank, 1989a. 'Women's Work and Their Status: Rural Indian Evidence of the Labor Market and Environment Effects on Sex Differences in Childhood Mortality" in R. Anker, M. Buvinic and N. Youssef (eds.). Women's Roles and Population Trends in the Third World. London: Croom Helm, 1982. pp. 202-238. Selowsky, M. Who Benefits From Government Expenditure: A Case Study of Colombia. Oxford: Oxford University Press, 1979. Shapiro, D.M. and M. Stelcner. "Public-Private Sector Earnings Differentials in Canada, 1970-1980'". Concordia University. Department of Economics: Working Paper No 1986-3. 1986, Montreal. Siebert, W.S. "Comparable Worth: An Evaluation." University of Birmingham: Institute of Economic and Business Studies. Mimeograph, 1989. "Developments in the Economics Gf Human Capital" in D. Carline, C. A. Pissarides, W. S. Siebert and P.J. Sloane (eds.). Labour Economics. London: Longman, 1985. pp. 5-77. Sivard, R.L. Women: A World Survey. Washington DC.: World Priorities, 1985. Sloane, P.J. "Discrimination in the Labour Market' in D. Carline, C. A. Pissarides, W. S. Siebert and P.J. Sloane Labour Economics. London: Longman, 1985. pp. 78-157. Smith, S.P. Equal Pay in the Public Sector: Fact or Fantasy? Princeton NJ: Princeton University Press, 1977. Standing, G. Labor Force Participation and Development. Geneva: International Labour Organization, 1981 (2nd edition). Stelcner, M., and J. Breslaw. "Income Taxes and Labor Supply of Married Women in Quebec." Southern Economic Journal, Vol. 51, No. 4 (1985). pp. 1053-1072. References 237 Stromquist, N.P. "Empowering Women Through Knowledge: Policies and Practices in International Cooperation in Basic Education." Stanford University: SIDEC School of Education, 1986. Thomas, D. "Intra Household Resource Allocation: An Informational Approach.' University of Yale Economic Growth Centre, 1989. Thurow, L.C. Generating Inequality. 1975. New York: Basic Books. Thurow, L. Poverty and Discrimination. Washington DC: Brookings Institution, 1969. Tilak, J.B.G. The Economics of Inequality in Education. New Delhi: Sage, 1987. Tokman, V. "Policies for the Heterogenous Informal Sector in Latin America:. World Development. Vol. 17, No. 7 (1989). pp. 1067-1076. Treiman, D.J. and Hartmann, H. I. (eds.). Women, Work and Wages: Equal Pay for Jobs of Equal Value. Washington DC.: National Academy Press, 1981. Tzannatos, Z. "The Economics of Discrimination: "Theory and British Evidence" in D. Sapsford and Z. Tzannatos (ed.). Current Issues in Labour Economics. Houndmills: Macmillan, 1990a. pp. 177-207. -----. "Employment Segregation: Can We Measure It and What Does the Measure Mean?" British Journal of Industrial Relations, Vol. 28, No. 1 (1990b). pp. 105-111. -----. "Equal Pay in Greece and Britain". Industrial Relations Journal, Vol. 18, No. 4 (1984). pp. 275-283. -----. "A General Equilibrium Model of Discrimination and Its Effects on Incomes." Scottish Journal of Political Economy, Vol. 34, No. 1 (1987a). pp. 19-36. -----. "The Greek Labour Market: Current Perspectives and Future Prospects." Greek Economic Review, Vol. 9, No. 2 (1987b). pp. 224-238. -----. "The Long Run Effects of the Sex Integration of the British Labour Market." Journal of Economic Studies, Vol. 15, No. 1 (1988). pp. 5-18. 238 Women 's Employment and Pay in Latin America -----. -Reverse Discrimination in Higher Education: Does It Reduce Inequality and at What Cost to the Poor?" International Journal of Educational Development, Vol. 11, No. 3 (1991). pp. 177-192. Tzannatos, Z. and J. Symons. "An Economic Approach to Fertility in Britain Since 1860." Journal of Population Economics, Vol. 2, No. 2 (1989). pp. 121-138. Tzannatos, Z. and A. Zabalza. "The Anatomy of the Rise of the British Female Relative Wages in the 1970s: Evidence from the New Earnings Survey." British Journal of Industrial Relations, Vol. 22, No. 2 (1984). pp. 177-194. UNESCO (United Nations Educational, Scientific and Cultural Organization). Compedio de estadisticas relativas al analfabetismo, 1990 Edition No. 31. Prepared for the International Education Conference. Geneva. September 3-8, 1990. UNICEF (United Nations Children's Fund). The Invisible Adjustment: Poor Women and the Economic Crisis. Bogota: The Americas and the Caribbean Office. April 1987. Visaria, P.M. The Sex Ratio of the Population of India, 1961. Census of India Vol. 1, Monograph No. 10. New Delhi: Office of the Registrar General of India, 1971. Wales, T. and A.D. Woodland. "Sample Selectivity and the Estimation of Labor Supply Functions." International Economic Review, Vol. 21, No. 2(1980). pp. 437-468. Weinger, 0. Sex and Character. (1906) London: Heinemann. Weiss, Y. and R. Gronau. "Expected Interruptions in Labor Force Participation and Sex-related Differences in Earnings Growth." Review of Economic Studies, Vol. 58 (1981). pp. 607-619. Welch, F. "Measurement of the Quality of Schooling". American Economic Review, Vol. 56, No. 2 (1966). pp. 311-327. Westoff, C.F. "Fertility Control in the U.S.A." in Proceedings of the World Population Conference, Vol. II, pp. 245-247. New York: United Nations, 1967. References 239 Williams, J.E. and D.L. Best. Measuring Sex Stereotypes: A Thirty-Nation Study. Beverly Hills, Calif.: Sage Publications, 1982. Willis, R.J. 'Wage Determinants: A Survey and Reinterpretation of Human Capital Earnings" in 0. Ashenfelter and R. Layard (eds.). Handbook of Labor Economics. Amsterdam: North Holland, 1986. pp. 525-601. Willis, R.J. and S. Rosen. "Education and Self-Selection." Journal of Political Economy, Vol. 87, No. 5 (Part 2) (1979). pp. S7-S36. World Bank. World Development Report, 1990. Washington D.C.: The World Bank, 1990. World Development Report 1991, Washington DC: The World Bank. 1991 World Tables. Johns Hopkins, Baltimore: University Press, 1983 (3rd edition). Wright, R.E. and J.F. Ermisch. "Gender Discrimination in the British Labour Market: A Reassessment". Economic Journal. Vol. 101, No. 406 (1991). pp. 508-522. Zabalza, A. and J.L. Arrufat. "The Extent of Sex Discrimination in Britain' in A. Zabalza and Z. Tzannatos. Women and Equal Pay: The Effects of Legislation on Female Employment and Wages in Britain. Cambridge: Cambridge University Press, 1985. pp. 70-96. Zabalza, A. and Z. Tzannatos. Women and Equal Pay: 7he Effects of Legislation on Female Employment and Wages in Britain. Cambridge: Cambridge University Press, 1985. I INDEX abscissa 170 added worker effect 18-19 Africa 6, 20, 49, 56 age and earnings 160, 162 and participation see participation agriculture 19, 20, 53, 74-75, 79, 83, 84 Argentina 6, 22fin, 26, 29, 81, 99, 106, 164, 188 Asia 6, 7, 20, 53, 56, 57 Atkin Report 146 Australia 31, 145 Austria 103 Bangladesh 163 Belgium 105 Bolivia 6, 11, 13, 19, 31, 44, 60, 81, 88, 102, 104 Brazil 13, 17, 19, 20, 48, 81, 106, 177, 183 Britain 11, 18, 31, 105, 102fiz, 145, 146, 166, 198 Buenos Aires 22fiS Canada 18, 103, 174 capital markets 196, 197 censorship, in data 174fr characteristics, comparable (in wage decomposition analysis) 180 childbearing (see also fertility) 6, 16, 22, 25-27, 49, 61 child care 25, 26 child mortality 32 children, and economic adjustment 2Jh Chile 6, 13, 43, 93, 177, 183 China 163 242 Women's Employment and Pay in Latin America Colombia 13, 17, 43, 48, 81, 88, 93, 99, 104, 164, 184 comparable characteristics (in wage decomposition analysis) 180 comparative findings decomposition of the wage differential by country 203 efficiency gains and employment differentials 142, 148, 149 employment dissimilarity by occupation 90-91 employment dissimilarity in EEC and Latin America 105 hours of work by country 204 lambda, sign, value and significance 211 potential experience by country 208 of earnings functions 204-210 of participation functions 185 of the constant term (and the male pay advantage) 210 simulated female wages 144 years of schooling by country 206 comparative statistics annual rates of output growth 8 employees in the labor force 58-59 employment in agriculture 75 employment in industry 77 employment in services 78 female overrepresentation by employment status 82, 84 female overrepresentation by industry 83, 84 female participation 5, 45, 54 female participation by income 54 female participation by religion 54 fertility 8 illiteracy 8 industrial classification 107-111 labor force 39 life expectancy 6, 8 male participation 45 minimum age and the labor force 107-111 occupational classification 107-111 participation and teenage fertility 28 participation by age by world region 62-71 per capita income 8 regional per capita incomes 6Jin relative (female to male) labor force 42 relative (female to male) participation rates 45 relative (female to male) wages 5 sectoral distribution of the labor force 79 Index 243 comparative statistics (continued) wages in the public and private sectors 201 compensating differentials 159 constant term and decomposition of wage differentials 191, 192fii, 193 in earnings functions 153, 154, 156-158, 161 and male pay advantage 210 consumer preferences, and discrimination 146, 152 contraception 25, 27 Costa Rica 22, 23, 29, 43, 89 Cyprus 56 decomposition, of employment changes 19, 92-96 decomposition, of wage differentials and discrimination against men 191 and education 191, 206-207 and experience 191, 208-209 and hours of work 191, 204-205 and the constant term 191, 210 aggregate findings 30-32, 187, 188, 189 effect of specific variables 153, 184, 190-191 as an index problem 156 shift and slope effects 153-157 Denmark 105 discouraged worker effect 18 discrimination and customer preferences 152 and efficiency gains 4.fz, 21, 135-149 and employers 152 and fixed costs 152 and legislation 20, 26, 31, 32-33, 104, 152 and training 152 as a slope effect in the earnings functions 153, 156 as a shift effect in the earnings functions 153, 156-157 demand factors 151, 152, 177, 178, 179 distributional effects 138 indirect 178 -justified- 155, 184 positive see quotas statistical 152 supply factors 151-152, 178 upper bound of 22-24, 187, 194, 198 244 Women 's Employment and Pay in Lain Amerca dissimilarity, by employment status 82, 84, 89, 90-91, 93-95, 99-102 by occupation 88-97, 102-104 decomposition of changes in 92-96 horizontal 96 in employment distributions 72-107 index see Duncan index industrial 83, 84, 103, 104 relation to misallocation 98 sex ratio and structure effects 92-93, 94-95, 103, 104, 105 vertical 96 distribution, assumption about (see also truncation) 171-172 domestic servants 26, 184, 187 Dominica 41 Duncan index 86-88, 144 Dutch Equal Pay Act, 1975 180 earnings functions and compensating differentials 159 and employment status 181 and hours of work 179, 183, 204 and on the job training 161 and potential experience 10-11, 162, 208 and selectivity 12-13, 167, 172, 211 evaluation of 10, 163-168 returns to education 29, 30, 161, 206 standardization before decomposition 180 theoretical basis 158-163 earnings permanent 179 see also wages Eastern Europe 56 Ecuador 81, 99, 106, 141, 192 education and decomposition of wage differentials 191, 206-207 and employment in the public sector 192 and fertility 27 and participation 23, 25-27, 184-187 and wages (see also eamings functions) 26, 28 distributional effects 29, 30 enrollment in 40 formal 163 Index 245 education (continued) nonformal 164 policy for women 29, 30 type of 10, 163-164 vocational 164 EEC see European Economic Communities efficiency gains 21, 61, 135-149 elasticity of female labor supply 145, 167 of male labor supply 166 of substitution 141fi, 148-149 El Salvador 56 employment dissimilarity see dissimilarity endogeneity 163, 166, 167, 178fin England see Britain error term in earnings functions 154, 155, 172 in structural models 176 errors and optimization 178 and preferences 178 from omitted variables 163, 173 in budget constraints 178 of measurement 163, 165 of specification 163 European Economic Communities 31, 102, 104 experience in eamings functions actual 165, 167 and decomposition of wage differentials 10-11, 164-165, 191 and industrialized countries 11 imputed II in earnings functions for men 164 in earnings functions for women 165 potential 10, 11, 165, 190 potential, definition of 162 family effects on participation 48, 175, 176 income/production 27 work 12, 19, 38, 55, 73, 92, 93Jfr, 103, 106, 184, 199 female overrepresentation in economic sectors) 81Jh fertility 6fn, 25, 32 246 Women 's Employment and Pay in Latin America fertility rate, total 6, 8 female headed households 26, 27, 61, 186 formal sector 60, 86, 96, 192, 199 France 105, 145 fridge benefits, of employment 33, 179 general equilibrium, and discrimination 137 Germany 105, 145 Greece 56, 105 Grenada 41 Guatemala 6, 41, 43, 99, 141, 164, 177, 192, 200 Guyana 56 habit formation, and participation 176 Haiti 41 health, and labor force participation 186, 187 heteroscedasticity 175ftu home production 27, 74, 194 Honduras 6, 13, 41, 43, 44, 93, 99, 164, 177 horizontal dissimilarity see dissimilarity hours of work 57, 183, 204, 179, 191 human capital informal 164 unobserved characteristics 166 see also education illiteracy 164 immunization 25 index of dissimilarity and misallocation 97 definition 86 estimates of 90-91, 94-95, 103, 105 interpretation 87 relation to employment reallocations 97-99 see also Duncan index India 163 industrialized countries 10, 20, 26, 31, 48, 49, 53, 102, 145, 165, 196, 199 intertemporal substitution (labor supply) 176 Ireland 105 Israel 56 Italy 105, 145 Index 247 Jamaica 6, 14, 17, 43, 44, 60, 10 Japan 102, 103, 145 Kingston, Jamaica 22fr, 26 labor force 38-55 by industrial sector 80 growth of 40 relative (female to male) 40-41, 42 size 37-38, 39 lambda definition of 170 empirical estimates 189Jh, 190, 211 sign of 174, 175, 211 significance of 175, 211 see also selectivity legislation 33, 56, 57, 60, 61, 92, 104, 198 life expectancy 40 Luxembourg 104, 105 marital stability 26 measurement errors see errors of measurement Mexico 11, 13, 18, 48, 81, 89, 99, 102, 106, 183 Middle East 7, 16, 38, 49, 53, 55 migration 104 Mill's ratio see lambda minimum wage 56 misallocation 97-102 and relation to dissimilarity 98 Nepal 56, 163 Netherlands 105, 145 Netherlands Antilles 9 New Zealand 31 Nicaragua 163, 175 nonworkers, effects on earnings functions see selectivity Norway 103 nutrition 6, 25, 32 250 Women's Employment and Pay in Latin America wages choice of, for decomposition analysis 179 if women had men's employment distribution 144 if women had men's wage distribution 144 in the public sector 192 permanent earnings 179 relative (female to male) 5 welfare gains see efficiency gains welfare implications of employment interruptions 48 younger workers 18fin, 40, 48, 53 k T H E W O R L D B A N K Other World Bank Regional and Sectoral Studies Nongovernmental Organizations and the World Bank: Cooperation for Development, edited by Samuel Paul and Arturo Israel Unfair Advantage: Labor Market Discrimination in Developing Countries, edited by Nancy Birdsall and Richard Sabot Education in Asia: A Comparative Study of Cost and Financing, Jee-Peng Tan and Alain Mingat Health Care in Asia: A Comparative Study of Cost and Financing, Charles C. Griffin Bolivia's Answer to Poverty, Economic Crisis, and Adjustment: The Emergency Social Fund, edited by Steen Jorgensen, Margaret Grosh, and Mark Schacter Crop-Livestock Interaction in Sub-Saharan Africa, John McIntire, Daniel Bourzat, and Prabhu Pingali Commodity Price Stabilization and Policy Reform: An Approach to the Evaluation of the Brazilian Price Band Proposals, Avishay Braverman and others The Transition from Socialism in Eastern Europe: Domestic Restructuring and Foreign Trade, edited by Arye L. Hillman and Branko Milanovic.