/3522 71 WORKING WOMEN IN LATIN AMERICA: PARTICIPATION, PAY AND PUBLIC POLICY By Carolyn Winter -~~~~~ ~ ~ ~ 77 4 4 The World Bank Latin America and the Caribbean Region Technical Department WORKING WOMEN IN LATIN AMERICA: PARTICIPATION, PAY, AND PUBLIC POLICY by Carolyn Winter World Bank Latin America and the Caribbean Region Technical Department 1818 H St., N. W. Washington, D.C. U.S.A. 1994 TABLE OF CONTENTS Introduction ...................................................... 1 Women's Labor Market Participation ............................................ 3 Trends in Women's Labor Force Participation Rates ...............5............. Participation and Childbearing .................................... 6 Women's Sectoral and Occupational Distribution ................................. 7 The Gender Wage Gap ........................................... 10 The Diminishing Wage Gap .................... ....................... 1 1 Wage Trends by Occupational Category ......................................... 12 The Wage Gap and Recession ......................... .................. 12 Gaps in Participation and Pay: The Role of Discrimination ................... 13 "Discrimination" ........................................... 15 Wage versus Employment "Discrimination" ..................................... 15 Public Policy: Current Interventions and New Directions ...................... 18 Existing Laws and Policy Interventions ........................... 19 (i) Protective Laws .......................... 19 (ii) Maternity Protection Laws .......................... 21 (iii) Laws on Child Care .......................... 24 (iv) Equal Pay Provisions ........................... 24 Conclusions and Recommendations .............................. 25 Implications for Policy ............................. 26 Selected References ............................. 29 Appendix .............................. 31 BOXES, FIGURES AND TABLES Boxes Box 1: Educational Attainment in Latin America ..............................2 Box 2: Explaining the Gender Wage Gap ..................................... 16 Figures Figure 1: Labor Force Participation Rates by Gender ........................4 Figure 2: Trends in Female Participation Rates ................................5 Figure 3: Age Participation Profiles of Women Workers in Three Countries ...........................................8 Figure 4: Ratios of Female/Male Hourly Wages ............................. 11 Tables Table 1: Changes in the Ratio of Female/Male Workers inthe 1980s ............................................. 6 Table 2: Venezuela: Occupational Distribution of Workers by Gender ........................................... 10 Table 3: Explaining the Wage Gap: Decomposition Results from Six Country Analyses ........................................... 17 Table 4: Protective Laws Regulating Women's Employment in Six Latin American Countries ......................... .......... 20 Table 5: Maternity Protections and Benefits Provided in Six Latin American Countries ............................................ 22. PREFACE This report has been prepared by Carolyn Winter of the Education and Social Policy Department of the World Bank. Jessica Youniss contributed substantially to the preparation and editing of the report. Evelyn de Castro and Benjamin Crow also contributed to the production of the report. The findings, interpretations, and conclusions expressed in this publication are those of the author and should not be attributed in any manner to the World Bank, to its affiliated organizations, or the members of its Board of Executive Directors or the countries they represent. This report summarizes the findings of a more extensive, analytical study, "Gender Discrimination in the Labor Market and the Role of the Law: Experiences in Six Latin American Countries", prepared by Carolyn Winter for the Latin American and Caribbean Region Technical Department. Background studies for this study were prepared by Ricardo Paes de Barros, Alejandra Cox Edwards, T.H. Gindling, Francisco J. Tapia Guerrero, Gustavo Marquez, Ivon Peres, Maria Isabel Plata, Alba Alonzo de Quesada, Lauro Ramos, Judith Roberts, Eleonora Santos, A. di Silvestro, Jaime Tenjo, Victor Perez Vargas and Carolyn Winter. i I Executive Summary In most Latin American countries the economic problems of the 1980s have been supplanted by a period of vigorous and robust economic growth. The region is now well-placed to increase its competitiveness in the world market. To ensure that growth is sustained, countries must continue on the path of economic restructuring. But, it is also critically important that they continue to develop their human resources and ensure that these resources are deployed and utilized as effectively as possible. Unnecessary restrictions or barriers to workers' opportunities will depress individual incentives to pursue education and training, limit employment opportunities, lower the economic returns to public investments in human resource development, and ultimately, reduce economic output. Women's labor force participation rates and pay rates are surprisingly low in Latin American countries. This is interesting given that women, on average, now have as much, and sometimes more, formal education than men. In the region at large, only around 33 percent of women work for pay. The reason for this is not clear, but increasingly attention has focused on gender labor market discrimination. It is important to ascertain whether such discrimination is indeed widespread; the underutilization and inefficient use of female labor would represent a tremendous waste of human resources. This paper summarizes the findings of a larger analytical study "Gender Discrimination in the Labor Market and the Role of the Law: Experiences in Six Latin American Countries". This study is a first attempt to ascertain whether gender discrimination is indeed widespread in the labor markets of six countries: Chile, Honduras, Colombia, Venezuela, Costa Rica and Brazil. Using household survey data covering the 1980s, the study examines changes in women's participation rates and relative pay over the period. It uses econometric analysis to examine whether the observed differences in participation rates and pay may be attributed to discrimination. The study also reviews labor laws which govern or regulate women's working conditions and pay to see whether they reduce opportunities for gender labor market discrimination. The study shows that, although women's participation rates continue to be low in the six countries, they have risen dramatically over the past decade. Indeed, one of the most important structural changes in the labor markets of these countries over the past decade has been the change in the sex mix of the work force. Women now comprise a greater proportion of the workforce. This increase in participation is evident among women of all ages but, interestingly, it is most evident among women of peak childbearing age. The participation data, if plotted as age/participation profiles, also suggests, contrary to what has long been assumed, that there is not a widespread tendency among women to withdraw permanently, or for extended periods of time, from the labor market once they have children. Examining data on women's participation rates over the decade also yields another interesting finding: although women continue to be concentrated in the non-formal employment sector, the proportion of women in this sector has declined sharply over the decade. This decline has been matched by a sizable increase in the proportion of women employed in the highest-paying employment sectors. The study also shows that the gender wage gap has diminished sharply over the past decade. This is undoubtedly explained largely by women's increased educational attainment; as their formal education level has risen relative to men's they have been better positioned to compete for job positions and higher earnings. The data also indicate that the economic recessions of the 1980s did not impose particular hardships on women who regularly participated in the labor market. Regular female employees do not appear to have been significantly more likely than men to have suffered wage cuts or retrenchment during the recessions. The study can yield only tentative findings about the extent and nature of gender "discrimination" in the labor market. Yet, the validity of these findings is strengthened by the fact that the econometric analyses yield remarkably consistent findings across each of the countries. The findings strongly suggest that "discrimination" is indeed widespread. This discrimination appears to be practiced mainly as wage discrimination whereby women receive lower returns to their human capital than men. A review of labor laws governing women's employment and pay in the six countries shows that these laws affect women's working opportunities in various ways. Certain laws, and particularly the older protective laws, are actually discriminatory towards women. In many cases they impose unnecessary restrictions on women's employment opportunities. Few would dispute the need for laws to protect and extend special benefits to women during pregnancy and childbearing. Yet, a review of these laws shows that they often work to encourage employers to discriminate against female workers. This happens when the laws require employers to finance much, or all, of the maternity benefits. This raises the cost of female labor relative to male labor for employers, so encouraging employers to hire the cheaper male labor whenever possible. Laws which require employers to provide and fully subsidize childcare facilities when they employ more than a certain number of female workers have similar effects. The evidence gathered for the study also suggests that equal pay and equal opportunity laws have not been particularly effective in reducing gender discrimination in the labor market. This is largely because the agencies responsible for enforcing these laws are poorly funded and ill-equipped to enforce these provisions. Based on the empirical analyses and the review of the labor laws, the study recommends several courses of action which may be pursued to reduce gender discrimination in the labor market. It recommends that countries wanting to reduce such discrimination: * Assess how laws affect women's economic opportunities: Countries should review labor laws to determine whether they work to reduce women's employment opportunities and if they increase the cost of female labor relative to males for employers. This information should help governments determine which laws adversely affect women's employment opportunities and pay and indicate which laws should be modified or revoked. * Take steps to blur gender distinctions in the law: It is important, to the extent possible, that laws are phrased in gender neutral terms. Laws which require employers to provide special benefits for women raise the cost of female labor relative to males' for employers. This encourages employers to discriminate against the more expensive female labor. * Strengthen existing enforcement mechanisms: Effective enforcement of carefully conceived laws is important if they are to have their intended effect. In most Latin American countries this will require that the institutional capacity of the Labor Inspectorate systems be strengthened. * Identifying necessary judicial reforms: The labor courts and tribunal systems are often overburdened and very slow. Highly formalistic procedures followed by the courts are also costly and difficult for the average worker to comprehend. A careful review of the operation of the labor courts would help identify what steps could be taken to make them more accessible to their clients. * Improving workers' knowledge of their rights: Laws will have little effect if their target population is not educated about their rights. Governments could reduce discrimination by promoting women workers' knowledge of the law. I l WORKING WOMEN IN LATIN AMERICA: PARTICIPATION, PAY AND PUBLIC POLICY INTRODUCTION The concern over the need to ensure human resources are deployed as Optimism over Latin America's effectively and efficiently as possible is economic future is currently wide- causing attention to focus more on spread. The economic problems experi- women workers. Significant, and enced by many Latin American coun- largely inexplicable, differences exist tries in the 1980s have been largely re- between men and women in terms of placed by vigorous and robust economic their labor force participation and pay. growth characterized by rapid expan- Women's participation rates are surpris- sions in inter-regional and foreign trade ingly low, with only around 35 percent and increases in foreign investment. of women in the region working for The continuing growth and increased pay. This is below levels recorded in competitiveness of Latin America in the the developing countries in Asia with world market, however, will depend which Latin America must compete in upon many factors, key amongst which the international market. A large gender will be the continuing development and wage gap also exists in most Latin effective utilization of its human re- American countries; on average, work- sources. Latin American countries have ing women earn only 66 percent of been, and are continuing to, invest men's wages. The gender differences in heavily in the education and training of participation rates and pay are of par- their populations. Most countries have ticular concern given that most coun- already achieved quite remarkable in- tries have invested heavily in female creases in school enrollments with the education. result that the average educational at- tainment of men and women has risen Why are women's participation sharply over the past two decades (see rates so much lower than men's? And Box 1). what explains the large gender wage gap? Circumstantial and anecdotal evi- To maximize the returns to these dence suggests that women may be investments in human resource devel- subject to different hiring and employ- opment, however, it is essential that ment practices, many of which are dis- countries ensure that educated man- criminatory. If gender discrimination in power can participate effectively in an the labor market is indeed widespread it open and competitive labor market. Ar- raises two concerns. First, it can impose tificial or unnecessary restrictions on serious economic costs on society. workers' opportunities will depress in- Studies in industrialized countries have dividual incentives to pursue education shown that such discrimination reduces and training, lower the returns to in- national income. And, second, such dis- vestments in human resource develop- crimination raises serious equity con- ment, and, ultimately, reduce economic cerns. Not only is a sizable proportion output. of the population subject to unjust 2 Box 1: Educational Attainment in Latin America A substantial body of empirical research now exists attesting to the significant economic benefits derived from investments in formal education and training. Higher worker productivity and increased national output are directly linked to investments in education. Growth accounting estimutes for Latin American countries report that investments in education accounted for over 4 percent of annual growth in Chile, Colombia, and Ecuador, and as much as 16 percent in Venezuela in the 1970s. This education-economic growth nexus has spurred governments in most Latin American countries to invest heavily in expanding educational access over the past few decades. Particular attention has been paid to increasing school access to underserved population groups. As a result, girls' educational access and attainment has increased very substantially. In many countries girls' participation rates now equal, or exceed, boys' rates at the primary and secondary levels. Even at teriary levels, women make up almost 50 percent of enrollments in many countries. Expanded educational access for girls over the past few decades has translated into a more educated pool of female labor. Indeed, working women in many countries now average more years of formal education than men as shown in the Table: Average Years of Formal Schooling by Gender (Late 1980s/Early 1990s) Woring Men Working Women All Women Colombia 7.6 8.7 7.1 Costa Rica 6.6 8.5 6.2 Venezuela 6.9 8.5 6.3 Honduras 4.9 6.3 4.6 The expected outcome of women's increased human capital would be a growth in women's labor market participation and a decrease in the gender wage gap as women become more competitive in the labor market. treatment, but fewer employment oppor- the Role of the Law: Experiences in Six tunities and lower wages translate into Latin American Countries". Drawing on poorer standards of living for this employment and earnings data available group. The consequences are likely to in country household survey data cover- be particularly severe for female headed ing the 1980s, the study reviews trends households. in women's participation and relative pay. The study also makes a first at- This paper summarizes the tempt, using econometric analysis, to findings of a recent study conducted assess the extent and nature of gender under World Bank auspices, "Gender labor market discrimination. Building Discrimination in the Labor Market and on this information, the study then re- 3 views and, to the extent possible, logical techniques used, and the results evaluates public policies and labor laws obtained in the six separate country which govern or regulate women's analyses, are referred to the larger working conditions and pay. In sum, study. this information provides a fairly com- prehensive overview of women's This paper begins by summariz- working conditions. It provides a basis ing information from the larger study on on which to formulate policy measures women's labor force participation rates. to ensure that women's productive po- Existing gaps, trends and changing pat- tential is realized more effectively in the terns in women's participation rates are labor market. discussed. The paper then presents sununary information on the gender The six countries included in the wage gap, describing how it has dimin- study are Chile, Honduras, Colombia, ished over time and discussing how the Venezuela, Costa Rica and Brazil. They size of the gap has fluctuated during are broadly representative of economic periods of economic recession. In the and social conditions in the region at next section of the paper the discussion large. The countries are variously ex- turns to the possible explanations for the periencing rapid and slower rates of observed gender differences in partici- economic growth. Some are highly ur- pation rates and wages. The focus is on banized, while in others a significant gender discrimination in the labor mar- proportion of the population continues ket. The results of an empirical analysis to be engaged in rural production. Rates which estimates the extent and nature of of un- and under-employment are high this discrimination are presented. Policy in certain of the countries and very low measures to combat gender discrimina- in others. Despite these differences, the tion in the labor market, common to the separate country analyses yield re- six countries, are discussed in the sub- markably consistent findings which sequent section. The effectiveness and suggests that the conclusions drawn outcomes of these policies are reviewed have broad applicability to the region at and discussed. The final section of the large. paper outlines recommendations and policy actions which might be consid- This paper provides a very con- ered by countries interested in promot- cise and much simplified discussion of ing women's participation in the labor the findings of this more comprehensive market. study. It aims to inform readers of the major findings of the larger study and stimulate thought about appropriate and WOMEN'S LABOR MARKET effective policy interventions. For this PARTICIPATION reason, explanations of the various methodological approaches used in the Until recently, little attention has larger study are not included here and been given to women's labor market are reviewed only briefly in the Ap- participation rates in Latin America. pendix. Readers interested in obtaining The rate and changes observed in more information about the methodo- women's labor force participation how- 4 Figure 1: Labor Force Participation Rates by Gender (Late 1980s/Early 1990s) 95 85 75 65 CWomen ever, can provide insight into the ro- donesia and Thailand, women's partici- bustness and competitiveness of the pation rates had climbed above 50 per- economy and into changing societal atti- cent by the late 1980s. tudes towards women. In a vigorous and open economy, women's participa- The low participation rate for tion rates will increase as the demand the region at large does, however, mask for labor increases and as the opportu- considerable differences in women's nity cost of women's non-market work participation rates across Latin Amenr- rises. The increasing participation of can countries. In some, women's par- women is also a reflection of changing ticipation rates are comparable to those attitudes about women (particularly in Asian countries. In Colombia, for married women) working outside the instance, close to 50 percent of women home, of perceptions about what consti- worked in the labor market in 1989. A rutes "suitablea work for women, and of fairly high proportion of women women's right to plan their families so (around 43 percent) also worked for pay they might pursue careers mn the labor in Honduras and Brazil by the end of market. the 1980s. In the latter two countries, however, the high rates are probably Given the growth of Latin more a consequence of the continuing American economies and the industrial- importance of agricultural production in ized nature of this growth, one would the economy than of women's absorp expect women's labor market participa- tion into manufacturing and industrial tion rates to be fairly high in the region. activities. Interestingly, women's par- These rates however, are surprisingly ticipation rates were relatively low in low. In the region at large, only around Venezuela (40 percent in 1989) and 35 percent of women worked in the la- Chile (37 percent in 1987) despite their bor market. By comparison, in fast- higher levels of industrialization. Ex- growing Asian economies, such as In- ceptionally low participation rates of 5 around 30 percent were recorded in past decade (see Figure 2). Some of the Costa Rica at the end of the decade. most impressive gains are recorded in the faster growing economies and in Women's participation rates countries where women's participation compare unfavorably with men's in all rates were initially very low. Such is the six case study countries. The proportion case with Chile, for example, where the of women working for pay is well be- rate surged from 31 percent in 1980 to low that for men (see Figure 1). This 37 percent in 1987, a 17 percent in- disparity raises questions about why so crease in women's participation rates few women choose to work for pay. over a seven year span. A substantial Gender differences in educational at- increase is also recorded in Costa Rica tainment clearly do not provide a satis- where the proportion of women work- factory explanation. There must, then, ing for pay was initially very low; be either economic factors or cultural or women's participation rates jumped legal barriers which discourage women from around 24 percent in 1980 to al- from entering or remaining in the labor most 30 percent in 1989. Less dramatic, market. but still very substantial increases also occurred in Venezuela and Brazil. Rates increased rather more slowly in Co- Trends in Women's Labor Force lombia where the proportion of women Participation Rates in the paid labor market was already fairly high at the beginning of the dec- Although women's participation ade; a seven percent increase in partici- rates continue to be low in much of pation rates was recorded between 1980 Latin America, there has been a quite and 1989. dramatic increase in these rates over the Figure 2: Trends in Female Participation Rates (Major Urban Areas Only) 55 50 45 40 __ {, 35 == 30- . _ 25 20 I l i 1980 1985 1987 1989 1990 Year I -Chile -Venezuela * Brazil o Costa Rica - Honduras + Colombia| 6 Table 1: Changes in the Ratio of Female/Male Workers in the 1980s Country Base/End Ratio of Female/Male Point Change over Years Workers Period Base Year End Year Chile 1980/1987 0.47 0.55 0.08 Honduras 1986/1990 0.55 0.58 0.03 Colombia 1980/1989 0.57 0.62 0.05 Venezuela 1981/1990 0.39 0.55 0.16 Costa Rica 1980/1989 0.32 0.39 0.07 Brazil 1981/1990 0.43 0.51 0.08 The rapid increase in women's influence policy have thus often focused labor force participation rates, coupled on the need to expand access to child with steady or slightly declining rates of care and improve the maternity benefits participation among men, has meant accorded to working women. that the sex mix of the labor force changed significantly in all six countries The assumption that childbear- over the 1980s. Women now comprise a ing responsibilities limit women's mar- substantially larger proportion of the ket participation has been fueled by paid work force than previously, as is findings in many other countries, both evident from the observed changes in developing and industrialized, that the female/male sex ratio over time. As women's participation rates, when Table 1 shows, the change in the gender graphed by age group, often show a composition of the labor force has been single- or double-peaked pattern. In the particularly striking in Venezuela, but it single-peaked pattern, women's partici- has also been marked in Chile, Brazil pation rates rise steadily after the and Costa Rica. school-leaving age and then drop sharply around the prime childbearing age, never to resume an upward trend. Participation and Childbearing This pattern indicates that women tend to withdraw permanently from the labor If it can be determined why market following childbearing. The women's participation rates are low, double-peaked pattern, evident in sev- effective policy measures can be intro- eral fast-growing Asian countries (as in duced to remove impediments to their Korea and Malaysia) and in many indus- participation. Women's childbearing trialized countries, shows that women and childrearing responsibilities have tend to withdraw from the market at long been assumed to be an important childbearing and re-enter the labor force factor limiting their participation in again when their childbearing respon- Latin American countries. Efforts to sibilities have diminished. Clearly, 7 though, both patterns of participation tional attainment and the probability of are linked to reduced labor force par- labor force participation in all six coun- ticipation rates among women. tries. In Venezuela, for instance, the probability that a woman will work for Interestingly, age/participation pay was 29 percent among those with profiles of women in the six Latin primary education but rose to 50 per- American countries exhibit neither of cent for women with secondary educa- these expected patterns. Instead, the tion. profiles have a "plateau" shaped curve which rises after school-leaving age, A second explanation may lie in peaks at around 30 to 35 years, holds women's declining fertility rates. Over steady until about 40 or 45 years, and the past decade, tremendous strides then declines (see Figure 3). No "dip" have been made in Latin America in in participation is evident during improving women's access to family women's prime childbearing years. This planning. As women's capacity to delay age/participation profile closely resem- and space their children has increased, bles that of men and indicates that they have been more able to participate women tend to remain attached to the in higher education and training and labor market once they have entered it. commence work with a definite career There is no evidence that women typi- path. Women in this situation will be cally withdraw from the market because less likely to withdraw from the labor of childbearing. It is also notable that market after childbearing because the there has been a general upward shift in opportunity costs of their time in the this age/participation profile for women home will be so much higher. Some over the past decade with particularly circumstantial evidence exists to support large increases being observed among this explanation; a strong negative rela- women of prime childbearing age (20 to tionship is found in all the countries 35 years). between the number of children a woman has and the probability that she Why do women's age/parti- will work for pay. Women with fewer cipation curves resemble men's in these children are much more likely to be in countries? And, why is the expected dip the labor force. in women's participation not observed? It is not possible to definitively answer these questions, but there are several Women's Sectoral and Occupational possible contributing explanations. One Distribution explanation centers on the fact that women's educational attainment is gen- It is well-known that women erally high in these countries and is of- workers in Latin America tend to be ten equal to men's. More educated concentrated in lower-paying, less sta- women may enter the labor market and ble employment sectors and occupa- choose to remain there in order to re- tions. This means that women work coup their investments in their educa- predominantly in the non-formal sector, tion. Indeed, a very strong, positive, rather than in the public sector or pri- relationship is found between educa- vate sector. As employees in the non- 8 Figure 3: Age/Participation Profiles of Women Workers in Three Countries Chile Greater Santiago Area so. 30 20- 10I 15-19 20-24 25-34 35-44 45-54 55+ Age Groups --1974 -1980 - 1987 Honduras Major Urban Areas 40*X so.- 40- 0' 30- AL. 10 -4 15-19 20-24 25-34 35-44 45-54 55+ Age Groups - -1986 -1989 _1991 Colombia Bogota Only 2 w so 40- 0. 30 20 - 15-19 20-24 25-34 35-44 45-54 55+ Age Groups --1980 -1987 -.1989 9 formal sector, they typically are not non-formal sector fell from 40 percent covered by minimum wage regulations in 1981 to around 32 percent in 1990. and do not benefit from labor provisions Women's private sector participation which establish minimally acceptable rose considerably over the same period. working conditions. Of the women who These changes in women's employment work in the formal sector, a large pro- patterns are likely to have had far- portion are generally concentrated in reaching consequences; women in the lower-paying occupations such as the formal sector will have profited from service, clerical and sales occupations. more stable employment, higher aver- age salaries, and coverage under the It is not well-understood why various labor laws providing benefits women are concentrated in the lower- such as social security coverage and paying employment sectors and occupa- maternity leave. tions. Gender differences in educational attainment clearly do not explain this. It This trend is not, however, is possible that these employment dif- common to all countries. In Honduras, ferences may be partly explained by for example, the proportion of women women's tendency to pursue training working in the non-formal sector has and careers in "female" lines of work remained relatively constant over the six (where wages and employment oppor- years for which data are available. tunities are typically more limited). This Likewise, little change in the distribu- does not seem to provide a completely tion of women across employment sec- satisfactory explanation, however. tors is observed in Costa Rica during Other factors must also be important the 1980s. and increasingly attention is turning to the role discrimination might play in Limited data make it more diffi- explaining these differences. cult to ascertain whether similar changes have been occurring in the dis- Although working women are tribution of women across occupational predominantly concentrated in the non- categories. Available data for Venezuela formal employment sector, employment suggests at first glance that more and earnings data show that the propor- women have moved into higher-paying tion of working women in the non- occupations, including the professional formal sector has declined sharply over and technical and managerial occupa- the decade in several of the countries. tions. However, when a simple measure In Colombia, for example, around 38 (the Duncan Index) is used to assess the percent of working women were in the extent of dissimilarity that exists in the non-formal sector in 1980, but by 1989 employment distributions of women and less than 32 percent worked in this sec- men, the picture becomes less clear. As tor. This decline was matched by a siz- Table 2 shows, a higher value of the able increase in the proportion of index (55.5) is obtained in 1990 than in women employed in the highest-paying 1981 (50.6), indicating that the dissimi- sector, the private sector. Similar larity in the gender composition of the changes are observed in Venezuela work force across occupational catego- where the percentage of women in the ries has increased rather than de- lo Table 2: Venezuela: Occupational Distribution of Workers by Gender, 1981 and 1990 1981 1990 Occupational Grouping Women Men Duncan Women Men Duncan (%) (%) Index (%) (%) Index Professional & Technical 23.9 5.5 9.2 26.6 6.6 10.0 Managerial 1.3 3.7 1.2 2.6 5.9 1.7 Clerical & Sales 34.1 17.6 8.2 39.4 17.1 11.2 Agricultural Workers 2.4 14.0 5.8 1.1 10.1 4.5 Miners 0.0 0.3 0.2 0.0 1.1 0.6 Transport 0.6 14.2 6.8 0.5 12.3 5.9 Operatives 13.2 35.7 11.3 7.1 37.1 15.0 Service Workers 21.8 8.9 6.5 19.3 9.7 4.8 Domestic Workers 2.8 0.1 1.4 3.5 0.0 1.8 Duncan Index Across all 50.6 55.5 Occupational Groupings creased. It appears, then, that women considerably across countries. It is es- have become more, not less, concen- pecially large, for instance, in Brazil trated in certain occupations than they where women earned only 55 percent of were in 1981. Estimates using data from men's hourly wages in 1990. It is ex- Brazil suggest a similar pattern; women ceedingly small in Venezuela and Costa continue to be concentrated in the low- Rica where women's hourly wages were est-paying occupation (Services). They 93 percent and 97 percent of men's are, however, more heavily represented wages in 1989, respectively. in the two highest-paying occupational groupings (Managerial and Technical) Increasingly, questions are being than men. raised about why women's wages are below men's. The fact that working women's educational attainment equals THE GENDER WAGE GAP men's means that it is no longer possi- ble to attribute the wage gap to gender Women workers, on average, differences in educational attaimnent. earn less than men workers in all Latin And, preliminary information shows American countries. This remains true that women are not necessarily more even after controlling for differences in unreliable workers and are not prone to the hours worked by women and men. higher absentee or quit rates than men On average, women workers in Latin as has often been posited. Attention, America earn only 66 percent of men's then, is turning to the role which gender wages. This average figure, however, discrimination might play in explaining masks the fact that the wage gap differs the wage gap. But, although anecdotal 11 Figure 4: Ratios of Female/Male Hourly Wages, 1980-1991 100 865 75 o , ,> 70 , y/ / \ 60- 55 -~~~~~~~~~~ 50~ ~ 45- 40- 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 991 Years 1-Chile aVenezuela -CostalRica +-Brazil eColombia eHonduras Cowwry Base/End Ratio of Female/Male Point Cange Year Hourly Wages over Period Base Year End Year Chile (a) 1980/1987 0.68 0.71 0.03 Honduras (a) 1986/1990 0.65 0.68 0.03 Colombia (b) 1980/1989 0.63 0.86 0.23 Venezuela (a) 1981/1990 0.87 0.93 0.06 Costa Rica (c) 1980/1989 0.90 0.97 0.07 Brazil (a) 1981/1990 0.55 0.55 0.00 Notes: (a) Data for larger metropolitan areas only. (b) Data for Bogota only. (c) Data for entire country. evidence of wage discrimination is ished sharply over the past decade or widespread, empirical evidence attesting so. This is undoubtedly largely ex- to its existence has been exceedingly plained by women's increased educa- limited to date. Before discussing the tional attainment; as their formal educa- findings of the empirical analyses of tion level has risen relative to men's wage discrimination, it is useful to pre- they have been better positioned to sent some basic data on the gender wage compete for job positions and higher gap and examine trends over time. earnings. Figure 4 charts changes in the The Diminishing Wage Gap female/male ratio of hourly wages over time. In five of the six countries In most Latin American coun- women's wages have increased substan- tries, the gender wage gap has dimin- tially relative to men's over the dec- 12 ade. The largest change is recorded in occupational category, Managerial oc- Colombia where the female/male hourly cupations. Here, the female/male ratio wage ratio was 0.63 in 1980 and 0.86 in of hourly wages increased remarkably 1989, a 0.23 point change. Despite this from 0.51 in 1980 to 0.81 in 1989 (a 58 very substantial change, however, percent increase in the ratio). The in- women still, on average, earned only 86 crease in the ratio was also substantial percent of men's hourly wages in 1989. in the Services occupations where the The decline in the wage gap was more bulk of women work. Similar changes modest, but still significant, in Costa are observed in Venezuela where the Rica and Venezuela. In Costa Rica the largest decrease in the ratio is recorded ratio of female/male wages increased in the Professional/Technical occupa- from 0.90 in 1980 to 0.97 in 1989. tions, followed by Clerical Occupations Women's average hourly wages also and Service Occupations. increased relative to men's in Chile and Honduras, but rather less substantially. Only in Brazil did the ratio of fe- The Wage Gap and Recession male/male hourly wages remain more or less unchanged over the decade. There has been much interest and speculation about the effects eco- nomic recessions have had on women's Wage Trends by Occupational Cate- working conditions. It is often asserted gory that women have fared particularly badly during economic downturns. It is interesting to examine Charting female/male hourly wage ra- changes in women's wages relative to tios over time provides a more accurate men's by occupational category. Have picture of how women have fared dur- the increases in women's relative wages ing these periods. As Figure 4 shows, been uniform across occupational cate- the decline in the gender wage gap did gories or have they been greater in cer- not occur steadily, but was interrupted tain occupations? by sharp, but temporary, increases in the wage gap. These fluctuations in the The available country data sug- wage gap coincide with periods of re- gest that the increases in women's rela- cession in the respective countries, tive wages have been very significant in providing support for theories that the higher-paying occupational catego- women's working conditions deterio- ries (Managerial and Technical) where rated during recession. In Costa Rica, women's participation rates have in- for instance, the wage gap increased creased sharply, and the lowest-paying sharply during the recession of 1982/83; categories where the majority of the female/male hourly wage ratio de- women workers are typically concen- creased from 0.91 in 1981 to 0.83 in trated. This is well-illustrated by data 1983, but then resumed its upward from Colombia. Although female/male climb in 1984 following the recession. wage ratios decreased in most occupa- A similar pattern is observed in Chile tional categories in Colombia, larger where the wage gap increased during decreases occurred in the highest-paying 13 the recession of 1982 but contracted GAPS IN PARTICIPATION AND sharply after the recession in 1984. PAY: THE ROLE OF DISCRIMI- NATION On the basis of these graphs it is tempting to conclude that women work- The reason why women's par- ers are indeed more likely than men to ticipation rates and pay are low relative suffer wage cuts and retrenchment dur- to men's is not clear. Evidently, human ing recession. Regression estimates run capital differences between men and on the time-series data available in the women do not provide a satisfactory six country cases, however, suggests explanation; women's average educa- that such conclusions might well be in- tional attainment now equals, or ex- correct. While the regression estimates ceeds, men's in most countries. confirm that the wage gap increased during recessions, ,.hey also show that Other explanations why such a the average human capital endowments small proportion of women work for of women workers declined during pay can be proposed, most of which are these periods. Once the economy picked rooted in societies' attitudes and per- up, women's human capital endow- ceptions towards women. Two explana- ments again increased relative to men's. tions appear to be of particular import. The most feasible explanation for this One is that women suffer wage dis- finding is that less educated women crimination, meaning that they receive temporarily entered the market during lower wages than men even when they recession, possibly in an attempt to perform the same work and have the "shore up" falling family incomes. same human capital endowments. In Their low human capital probably other words, women may be receiving meant that it does not pay them to work lower wages than men simply because in more prosperous times, so they with- of their sex. To the extent this occurs, drew from the market when economic women may be discouraged from enter- conditions improved. If this hypothesis ing the labor market. Women, far more is correct, and an increasing number of so than men, face very clear opportunity empirical studies suggest it is, the wid- costs to their time in the market; they ening of the wage gap during recession are primarily responsible for childrear- is not necessarily a consequence of ing and family maintenance tasks in the worsening employment and pay condi- home that compete with the time they tions for the pool of women workers can allocate to market work. Rationally, who regularly work in the labor market. then, women will compare the potential Instead, the increased wage gap reflects returns to labor market work with the a temporary change in the composition competing demands for their time in the of the female work force and a decline home. Lower wages in the market will in women's average human capital. raise the opportunity cost of their time in the home, possibly tipping their de- cision in favor of remaining at home. 14 The second possible explanation exists. The Brown, Moon and Zoloth is that women are not given equal op- model provides a more accurate assess- portunities and equal access to job posi- ment of discrimination and indicates tions in the market -- that they are sub- whether it is practiced principally as ject to employment discriniination. If wage discrimination or employment dis- this is the case, women may either fail crimination. The methodological un- to gain employment or may be unable to derpinnings of each of these models are obtain employment in positions com- briefly described in the Appendix. mensurate with their education or skill Readers interested in more a more in- level. In the latter case, they may be depth discussion of the methodologies unwilling to compromise and accept and applications of the models are re- employment positions below their skill ferred to the larger study. level and choose to withdraw from the labor market. Both econometric models are rooted in economic theories which hold Discrimination, be it wage dis- that rational employers hire workers and crimination or employment discrimina- establish wage rates on the basis of tion, presents a feasible, if as yet em- workers' potential productivity, which pirically untested, explanation as to why is most easily assessed from their hu- women's participation rates and pay are man capital endowments (level of for- low in Latin American countries. The mal education, labor market experience, larger study on which this report draws skills training, etc.). A gender wage makes a first attempt to assess the extent gap, then, should simply reflect differ- and nature of gender labor market dis- ences in the human capital endowments crimination in Latin America. The ob- of women and men workers. Yet, ex- jective of the study is to ascertain perience in many industrialized coun- whether wage and employment dis- tries has shown that this is not neces- crimination are important factors con- sarily the case; a gender wage gap may tributing to the observed gender wage exist even though women and men have and participation gap. By establishing equal human capital endowments. It ap. the extent and nature of discrimination, pears, then, that employers are not en- the study should provide a basis for de- tirely economically rational in their re- riving policy measures which could be sponse to labor; their decisions to hire implemented to limit such discrimina- and reward labor may be influenced, tion. among other factors, by personal prejudices and propensities to discrimi- The study utilizes two different nate. The two econometric models at- econometric models, the Oaxaca De- tempt to provide estimates of the extent composition and the Brown, Moon and to which gender discrimination is prac- Zoloth Decomposition, to arrive at an ticed. An important caveat regarding estimate of gender labor market dis- this estimate of discrimination should be crimination in each of the six countries. borne in mind, however. The models The Oaxaca model yields only a broad attribute the part of the gender wage assessment of whether discrimination gap not explained by gender differences in human capital (here, measured as is years of formal education) entirely to terms of a male advantage in human discrimination. It is quite possible, capital endowments. The wage gap is though, that other factors also contrib- therefore entirely attributed to ute to the wage gap. Because these "discrimination." other factors remain unidentified or unmeasured, their affects will be incor- Results from the separate coun- rectly attributed to discrimination. try analyses thus strongly indicate that Thus, the measure of "discrimination" "discrimination" plays a significant role obtained using the econometric models in explaining the gender wage gap. The is very likely overestimated and should question remains, though, as to whether not be taken as a precise measure of this discrimination is practiced as wage discrimination. Instead, it is only "discrimination" or employment "discri- broadly representative of the extent of mination." discrimination. Forthwith, this measure is always referred to as "discrimination" to make it clear that this is an estimated, Wage versus Employment "Discri- not an actual, measure of discrimina- mination" tion. The determination as to whether "discrimination" is practiced principally "Discrimination" as wage "discrimination" or employ- ment "discrimination" was undertaken The econometric analyses con- using the Brown, Moon and Zoloth De- ducted in each of the six Latin Ameri- composition method. This technique can countries yield remarkably consis- recognizes that women workers are tent findings. In each country, generally concentrated in certain em- "discrimination" largely accounts for ployment sectors and occupational cate- the gender wage gap. Gender differ- gories and acknowledges that this may ences in human capital endowments be a consequence of differences in explain only a very small part of the women's and men's work opportunities gender wage gap (see Box 2). In Chile, and unequal access to employment op- for example, only 21 percent of the portunities. The Brown, Moon and wage gap is explained by such differ- Zoloth Decomposition technique thus ences over the decade. The remaining seeks the causes of the gender wage gap 79 percent of the wage gap not ex- by examining gender differences in hu- plained by human capital endowments is man capital endowments in each em- therefore ascribed to "discrimination." ployment sector (or occupational cate- A similarly large portion of the wage gory) while taking the different em- gap in Colombia is also attributed to ployment patterns of women and men "discrimination. " Interestingly, in three into account.' of the six countries--Venezuela, Costa Rica, and Brazil--women workers' av- The technique actually requires that average erage human capital endowments are wages of women and men be weighted according to each sex's representation in the respective actually higher than men's. Hence, no employment sectors (or occupational categories). part of the wage gap can be explained in 16 Box 2: Explainmg the Gender Wage Gap It is well established that human capital endowments, and specifically formal education level, are the best predictors of worker productivity. Rational employers typically hire workers and set saaes based on workers' human capital endowments. Hence, two equally qualified workers performing the same, or very similar, work should receive much the same wages.; But, it is argued that this is not the case in many Latin American countries and that employers choose to discriminate against women workers. The fact that a gender wage gap is found in all Latin American countries suggests this may be true. Does such discrimination exist? Do women tend to receive lower returns to their human capital? If the-gender wage gap is not largely explained by gender differences in human capital endowinkt)' it woild suggest that employers are influenced by other, less rational, factors in setting wages. Econometric analysis of wage and employment data in six Latin American counaies shows very clearly that differences in humnn capital endowments between the sexes do not explain the gender wage gap. In Colombia and Chile such differences explain 25% and 21% of the wage gap, respectively, and they explain only 4% in Honduras. In Venezuela, Costa Rica, and Brazil, women's average human capital endowments are actually higher than men's, so such differences do not help explain the wage gap at all. Percent of Wage Gap Cowuiy, Total explained by not Expkined by Female/Male Human Capitul Human Capital WageGap Diferences Differences Chile (1975-1987) .259 21 79 Hondur (1986-1991) .431 4 96 Colombia (1980-1989) .417 25 75 Venezuela (1981-1990) .173 0 122(a) Costa Rica (19801989) .216 0 120(a) Brazil (1981-1990) .544 0 113(a) Note:: (a) Percentages grater than 100 indicate that women's average human capital endowments exceed men's. Human capital differences between women and men therefore do not help explain the wage gap. Women do, then, on average, appear to receive lower returns to their human capital than men in the six Latin American countries. This is indicative of discrimination. For ease of interpretation, Table representative year, 1987.2 This par- 3 presents the results of the Brown, ticular analysis examines gender differ- Moon and Zoloth Decomposition for each of the six countries for just one 2 Results covering the years 1980 through 1990 are presented separately for each of the six countries in the larger study. 17 ences in human capital endowments by tral question, though, is how much of employment sector (private, public and the wage difference is explained by non-formal) rather than by occupational gender differences in human capital category. The separate country analyses (WE) and how much is attributed to yield very consistent results and show wage "discrimination" (WU). The De- that the total female/male wage gap is composition results clearly indicate that largely a consequence of wage human capital differences did not ex- "discrimination". plain the wage gap. Of the gender wage gap within employment sectors (W = The Decomposition results are 0.140), exceedingly little (WE = best understood by referring to one 0.001) was explained by gender differ- country example. In the case of Chile, ences in human capital. Wage for instance, the total female/male wage "discrimination" thus accounted for al- gap was 0.181 in 1987. The decompo- most all of the wage gap (WU = sition results show that wage differences 0.139). These findings are replicated in within employment sectors accounted the other country analyses. One factor for much of this wage gap (W = should be emphasized, however. In 0.140). Gender differences in access to three countries (Venezuela, Costa Rica, employment sectors explained very little and Brazil) women, on average, have (J = 0.041) of the wage gap. The cen- higher human capital endowments than Table 3: Explaining the Wage Gap: Decomposition Results from Six Country Analyses, 1987 Total Gender w J WE WU JE JU Wage Gap Chile .181 .140 .041 .001 .139 .009 .032 Honduras .467 .440 .027 .009 .431 .001 .026 Colombia .297 .281 .015 .045 .236 .006 -.036* Venezuela .174 .212 -.038* -.017* .229 .002 -.040* Costa Rica .199 .230 -.030* -.039* .268 .006 -.036* Brazil .530 .460 .070 -.090* .550 NA NA Notes: W- Portion of the wage gap explained by differences in wages between employment sectors J - Portion of the wage gap explained by gender differences in access to employment sectors WE- Differences in earnings explained by gender differences in human capital WU- Differences in earnings unexplained by gender differences in human capital JE- Differences in access to employment sectors explained by gender differences in human capital JU- Differences in access to employment sectors unexplained by gender differences in human capital * The negative signs indicate that women actually have an advantage over men in terms of human capital endowments. 18 men. This advantage means that all of more analytical research still needs to the wage gap within employment sec- be undertaken before a clear and accu- tors is attributed to wage rate understanding of gender labor mar- "discrimination" in these countries. ket discrimination is obtained. It is notable that in all countries, although gender differences in access to PUBLIC POLICY: CURRENT IN- employment sectors (J) contributed only TERVENTIONS AND NEW DI- minimally to explaining the total gender RECTIONS wage gap, the different employment patterns of women and men workers is Both widespread anecdotal evi- largely explained by employment dence and the results of the preliminary "discrimination." Gender differences in empirical analyses indicate that gender human capital endowments do not ex- discrimination may well be common- plain the differences in the employment place in Latin American labor markets. patterns of men and women. This is rather puzzling given that all countries in the region have imple- The results from each of the six mented policies and enacted labor laws country analyses indicate that whose intent is to reduce opportunities "discrimination", practiced principally for discrimination and guarantee work- as wage "discrimination," accounts for ers the right to equal opportunities and much of the gender wage gap. How equal pay. The existence of such dis- much faith should be put on these re- crimination suggests that the laws and sults? Are they sufficiently robust to policies are not being fully effective. form the basis for policy formulation? Clearly, if women's potential Several factors increase confi- contribution to economic development dence in the decomposition results. is to be fully realized, gender discrimi- First, very consistent findings are ob- nation in the labor market must be tained from each of the different coun- minimized. This will require that loop-- try analyses. Second, decompositions holes and shortcomings in current laws run using different population samples and policies be identified and rectified. (one which excluded domestic servants To date, however, almost no efforts from the analysis and one which in- have been made to document the laws cluded only single/unmarried workers) governing and regulating women work- also yielded very similar findings. And, ers or to assess their effectiveness. As a third, decompositions run substituting first step in this direction the larger more finely disaggregated occupational study on which this summary is based categories for the broader employment documented and reviewed labor laws sectors also provided the same general which specifically cover women work- findings. Thus, it would appear that the ers in the six selected countries. On the results are indeed robust and may form basis of this information, the study pro- a basis upon which to formulate policy. poses some tentative recommendations The findings should still, however, be as to how existing policies might be re- considered preliminary. Considerably 19 formed to better facilitate women's la- cally restrict women's employment in bor market participation. certain occupations and settings, includ- ing in night and shift work, mining oc- cupations, work around "dangerous" Existing Laws and Policy Interven- machines and chemicals, (lead, for ex- tions ample), "morally hazardous" occupa- tions, and in work requiring that heavy Although the intent of existing weights be lifted or pushed. The laws laws and policy interventions is to im- also sometimes mandate a shorter legal prove women's working conditions, work week for women, grant them many actually appear to have an adverse longer work breaks during the day, and effect on women's economic opportuni- set earlier retirement ages for women. ties. Rather than decreasing the gaps in women's participation rates and pay, Most industrialized countries many of the laws work to maintain which had adopted these laws moved to these gender gaps. Clearly, these laws revoke them in the early 1970s as per- have very different outcomes than those ceptions about what was "appropriate" which were originally intended. There work for women changed. Many of the is a need, then, to fully evaluate their protective provisions were considered to effect so that they might be revised and unnecessarily restrict women workers' reworked to better promote women's economic opportunities. The laws con- economic opportunities. tinue to be in effect, however, in most Latin American countries as Table 4 Laws regulating and governing shows. Among the six case study coun- women's employment may be grouped tries, only Brazil has revoked these laws into four broad categories, each of on the grounds that they are discrimina- which have somewhat different effects tory. and outcomes on women's employment opportunities and pay. A careful examination of these laws shows that they actually do work (i) Protective Laws to discriminate against women in sev- eral ways. In some cases, they impose Protective laws, designed to unnecessary restrictions on the em- safeguard women from employment ployment of women in certain occupa- conditions considered hazardous to their tions or industries. Such is the case, for health or mental well-being, are en- instance, with laws which place a blan- trenched in the Labor Codes of most ket prohibition on women's employment Latin American countries. Generally in industries using chemicals which are adopted in the mid-1950s, these laws potentially hazardous to the unborn were intended to protect women from child. While the rational for prohibiting exploitation and the unduly heavy and women of reproductive age from such dangerous industrial work associated industries is clear, there is no reason to with the mechanization of production bar women who are sterilized or past occurring at that time. The laws typi- childbearing age. Yet, employers are prohibited from hiring these women Table 4: Protective Laws Regulating Women's Employment in Six Latin American Countries PrMtecnve Prnisra'u Qaie Honduras Colomia Veszuela Coa Rica Braul General Employment Physical Capciy: LAbor Ph ysical Capuiy: Labor P LUp CAjIY: 2by5a1[CApril: LAbor Code, Art. 114: Women prohibited lW vRa&&an: Low Prohibitions /Liuitations Code: Art. 15: Prohibits Code, Art. 127: Work of Labor Code: Art. in: k -dng/uloadig of ships; loadiglunloadnig with Labor Code: Art. 87: maxunm dose women from work which women to be adapted to their 242: Prohibits cranes/witches: work in quafies; work underwater; work as Employment of levels set for exceeds their strength or age, state of health. women from jobs trimmers or stokers, or in foundries; mtaitainin moving women prohubited in women of child- constitutes a danger tn their reqiing et machinery; work using circular sws or other dangerous work which is bering age- physical condition because strength, from machinery; work in marfacture of metal or glass; work in physically unhealthy, of their sex. dngerous jobs, distillaon of alcohol; work with explosives, infmmable or hard or dagenrous. Lntl: Women prohibited and where workig caustic materials. from work involving use of conditions might be L1A: Women prohibited from work involving use of lead paint. white kad or lead sAlfate. unhealthy for Radiatian: Higher maximm dose kvls of expomre set for Fadiation: Low dose levels women women. of radiation exposmre set for Ld: Women Weighn: Weight limits set for women. women of childbearing age. prohubited frm work involving use of lead. Prohibitions on Labor Code, Art 15: Labor Code, Art. 84: LAbor Code, Art. Labor Code, Art. 112: Women prohibited from work in mines. Laws 2512 & N-3877: Employmnt in the Prohibits women from Women's employment 242: Prohibits prohibit women's Mining Industy underground miung. prohibited in mines. women from employment in Exceptions for women in mimng. work. undergrund mines. kadership positions, health service workers, or where it is required as part of an educational program. Prohibitions on Night Labor Code, Art. 147: Labor Code. Ail. Labor Code, Arts. 11l & 208: General prohibition between Labor Code, Art 88: Work Women may not work +5 242: Prohibits 19:00 and 06:00 except women in responsibk managemerit Prohibits employment hours at night. Ruling does night work except positions not involving mantdal work. Exemptions set for somne of women except in not apply in public sector or in family industries, fam y undertakings. agricultural undertakings. undertakings from headth and domestic 18:00 to 06:00. work from 19:00 to 05:00. Partdal exemptions can be obtained from Dept of Labor Inspection. The law has been suspended in times of high unemployment. Prohibitions on Labor Code, Art. 127: Work Labor Code, Art. 88: employment unsuitable of women to be appropriate to Prohibits employment to women's moral their mental and moral of women in nmorally developrment development. inppropriate' work. Special requirements Work Tine: Pregnant Work Time: Labor Code, Work Tini: Dady rest of at least 11 consecutive hours for Facilites: Special women nay not work Arts. 130/140: Women and women working shifts in dilustry. A reduction to 10 hours for a ventiation overtine. imnors graned one 2 hour maximun of 60 days per year may be authorized by the Labor requiremnents for work break per day. Other Department. A minimum of 9 hours in exceptional female workers. rest periods granted if a circumnstances for women in hotels, restaurants and domestic medical certificate indicates a service. need. 20 21 too. A similar argument can be made of very possible that these laws contribute laws which bar all women from occu- to the observed gender gaps in pay and pations requiring hard physical labor participation in Latin America. The (underground mining, for example) modification and reform of protective without giving due regard to the indi- laws is evidently a necessary first step vidual's capacity to perform such work. in removing obstacles to women's par- ticipation and in increasing women's Other protective laws impose relative pay. broader restrictions on women's em- ployment. Laws prohibiting women's (ii) Maternity Protecton Laws employment in night or shift work, for example, prevent women from pursuing Following the guidelines estab- flexible work arrangements which may lished by several International Labor better suit their dual role as mother and Office Conventions, all Latin American worker. They also preclude them from countries have enacted laws granting earning the higher wages and premiums pregnant and nursing women workers often associated with such work. special protections and benefits. The maternity protection laws currently in Protective laws may also, in force in the six case study countries are some instances, actually induce employ- summarized in Table 5. Usually, the ers to discriminate against female labor. benefits and protections provided for The laws do this by raising the cost of under Latin American law are similar to female labor relative to male labor. those extended to women in the indus- Laws requiring that women workers be trialized countries. Typically, women granted longer work breaks than men, qualify for a period of maternity leave for example, will cause women's aver- of around 18 weeks, are awarded a sti- age daily output in the workplace to be pend equal to their pre-leave salary below men's. Employers, then, will during maternity leave, and are pro- consider women to be less productive tected from dismissal during pregnancy employees than men and their rational and for some time following childbirth. response will be to discriminate against The manner in which these benefits and female labor by either not hiring women protections are financed in Latin Ameri- or by offering them a lower wage. A can countries, however, differs from sinlilar situation will occur under pro- that in industrialized countries. visions which establish shorter maxi- mum work weeks for women than men. The method of financing the benefits often works to increase the cost Thus, protective laws, while of female labor relative to male labor. designed to safeguard women, often This situation arises largely because of work against women's economic inter- the way the laws are implemented; the ests. They tend to limit women's em- burden of financing the maternity ployment opportunities and sometimes benefits is often passed on to the em- actually induce employers to discrimi- ployer rather than being covered by nate against female labor. Although the public sector programs. The employer, link cannot be clearly established, it is Table 5: Maternity Protections and Benefits Provided in Six Latin American Countries Country Qualifying Conditions Length of Maternity Leave Benefits Payments Prohibition of Disnissal Nursing Breaks Chile Social insurance coverage arnd I1 18 weeks (6 pre- and 12 post-natal) of non- Benefits are calculated as the Disnmissal prohibited during Two 30 minute paid breaks daily consecutive months of renounceable leave. May be extended if need is workers taxable salary for 6 mwnths pregnancy and for one year after until child is 2 years. Traveling contributions before commencing established through medical certification. preceding the 5 months before the completion of maternity leave except time may be added if infant is in maternity leave. beginning of maternity leave divided when just cause is approved by a child care facilities outside the by 10. Funded by the State - neither judge. In such cases, the worker workplace. The employer must employers or the social insurance continues to receive all benefits until also pay travel costs. system contribute. the end of maternity leave. Honduras Social insurance coverage (in - 18 weeks (6 pre- and 12 postsatal) of non- Benefits equal the worker's average Dismissal prohibited during Two 30 minute paid breaks daily regions where there is coverage) renounceable leave for married and cohabiting salary in the 180 days prior to pregnancy and for the 3 months while infant under 6 months. and attendance at work for 75 days women in the private sector. maternity leave. In regions covered following birth unless just cause is in the 10 months preceding pre- - 18 weeks (6 prc- and 12 post-natal) of non- by social insurance, it pays 66% of established and approved by the natal leave. Must have provided renounceable leave for married women in the public the salary and employers 44%. In Ministry of Labor. medical certificate confirming sector. regions not covered by social pregnancy to the employer in the insurance, employers pay the full first months of pregnancy. salary. Colombia Social insurance coverage and 12 12 weeks which mnay include one week of paternity Benefits equal the worker's regular Dismtissal prohibited during One 30 minute paid break daily weeks of contributions, 4 of which leave. If paternity leave is taken, women qualify for salary. Funded by the social pregnancy, maternity leave or until infant is 6 months. must have been during the 9 11 weeks. 24 weeks for spontaneous abortion. insurance system through a 13.5% breast-feeding unless just cause is months preceding request for payroll tax to which employers established and approved by a Labor maternity leave. contribute 9% and workers 4.5%. Inspector. Venezuela Social insurance coverage. 18 weeks (6 pre- and 12 post-natal). Allowances for Benefits equal the worker's regular Dismissal prohibited during Two 30 minute paid breaks if Workers must provide employers pre-natal leave may not be trnnsferred and used as salary. Social insuance fund 66% pregnancy and for one year after nursing facilities available in the with medical certification of post-natal leave. 10 weeks provided for adoption. and employers 34%. confunement urness just cause is firm. Two 60 minute breaks if pregnancy to qualify for maternity established and approved by the nursing facilities not available in benefits. Labor Inspector. tie firm. Costa Rica Social insurance coverage and - 16 weeks (4 pre- and 12 post-natal) in the public Benefits equal the worker's regular Dismissal prohibited during Either two 30 minute paid breaks provision of medical certification and private sector. salary. Funded by the social pregnancy and breast-feeding unless daily or 15 minutes every 3 to employer and social insurnice - 16 weeks (8 pre- and 8 post-natal) for teachers. insurance system (50%) and the just cause is established and hours. Firms employing more agency early in pregnancy. - 12 weeks for adoption. employer (50%). approved by the National Central than 30 female workers must - May be extended if need established through Benefits for dailyhourly workers Administration and General provide special facility for medical certification. estimated from average pay in 180 Inspector of Labor. If just cause is feeding. Length of benefits not days prior to conftnement. If the established tie worker loses all established. Employer required worker has worked less than 180 maternity benefits. to provide mother with additional days, payment estirnated on average paid rest time if requested. pay during that period. Brazil Social insurance coverage 17 weeks. Leave may be extended on presentation of Benefits equal the worker's regidar Dismissal probibited during Two 30 minute paid breaks daily a medical certificate. 5 days of paid paternity leave salary. Funded by social insurance pregnancy and confinement and for until infant is 6 months. provided. through a 0.3% monthly payroll levy 5 months after childbirth. on all employers. Employers pay the benefits and deduct the amount from their monthly social security contributions. Domestic servants receive benefits directly from social insurance offices. 22 23 then, is faced with higher costs when countries such as Chile, many employ- employing women. This is best illus- ers hire women, especially lower-skilled trated in the case of the maternity sti- women, only in a temporary capacity pends paid to women during their ma- since such workers do not qualify for ternity leave. In three of the six case maternity benefits. study countries (Costa Rica, Honduras and Venezuela), the law requires em- The question, then, is how to ployers to contribute directly, and safeguard working women during child- sometimes very substantially, to the bearing without prejudicing their eco- worker's maternity benefits. The finan- nomic opportunities. Clearly, maternity cial burden on employers is greatest in laws which impose direct costs on em- Honduras. Here, employers are re- ployers create incentives for discrimi- quired to pay 44 percent of the absent nation and non-compliance. What ap- worker's wage in regions covered by proaches have other countries used to the social security system. In regions minimlize distortions in labor pricing not covered by the social security sys- caused by maternity protection laws? tem they must pay the full salary of the The most straightforward approach is to absent worker. In Costa Rica, employ- ensure maternity benefits are funded ers are required to pay 50 percent of the through social security systems or spe- maternity stipend. cial public funding sources rather than by direct contributions from employers. Other maternity protection laws This is not always feasible, though, also work to increase the relative cost of where social security systems are poorly female labor. Requirements that em- resourced or where coverage is limited. ployers subsidize breaks (generally one Another approach used is to modify hour per day) during which mothers can maternity leave laws so they no longer nurse their infants, for example, have require women to remain off work for this effect. the full period of maternity leave. This directly reduces the cost of the leave to Given the extra costs which the employer. It does, however, raise maternity protection laws impose on the risk that women will be pressured to employers, employers have every in- return prematurely by employers. centive to either avoid hiring women or seek ways to avoid compliance with the A more innovative approach laws. The rather scanty information used in several European countries is to available with which to assess the ef- broaden laws so either parent can qual- fects of these laws suggests that they do ify for maternity leave and benefits. indeed encourage employers to dis- This, to some extent, helps blur the dis- criminate against women. In all six tinction between male and female em- countries, employers reportedly fre- ployee benefits, so reducing (but obvi- quently (and illegally) require female ously not eliminating) employers' pro- job seekers to provide medical certifi- pensity to associate the costs of mater- cation that they are not pregnant. In nity benefits directly with women work- Brazil, certificates of sterilization are ers. This approach has been adopted, reportedly commonly required. And, in for example, in Sweden. Here, the law 24 requires the mother to take 29 days of sessments of the cost of female labor. maternity leave following the birth of Most employers, therefore, expressed the child. The father is also accorded a an unwillingness to hire women if it short period of paternity leave. After meant they then became liable for child the woman's 29 days of leave either care provision. parent may elect to take the remaining 151 days of maternity/paternity leave. The fact that child care laws of- ten seem to have adverse effects on (iii) Laws on Child Care women's employment opportunities does not, however, mean that efforts to Believing that access to adequate provide such assistance should be aban- and convenient child care will facilitate doned. Rather, it means that there is a women's labor market participation, need to carefully assess the economic many Latin American countries have consequences which alternative policy passed laws requiring employers to measures may have on women. Child provide and fully subsidize child care care provision in other countries, such facilities. Preliminary assessments of as Colombia and Costa Rica, financed these laws, however, suggest that they, either through general payroll taxes or too, sometimes work against women's through a parent fee structure, have economic interests. This is the case generally been more successful and do where the laws require employers to not adversely affected women's eco- bear the costs of providing these serv- nomic opportunities. ices and where mandatory provision of services is linked to the number of (iv) Equal Pay Provisions women workers employed in the firm. All six Latin American case Such is the case with laws in study countries have enacted equal pay Chile and Brazil. In Chile, the law re- policies, often following ratification of quires employers of 20 or more female international Conventions promoting workers to provide child care facilities. such policies (including the United Na- In Brazil, employers with more than 30 tion's Convention on the Elimination of women workers must subsidize child All Forms of Discrimination Against care. Clearly, these laws create strong Women). However, widespread anecdo- incentives for employers not to hire tal evidence, now supported by the re- more than a certain number of women sults of the empirical studies undertaken (19 in Chile and 30 in Brazil). In other in this study, indicates that wage dis- words, they encourage employers to crimination persists despite the exis- discriminate against women. Interest- tence of these laws. What accounts for ingly, interviews with employers re- this? vealed that even though enforcement of these laws is often weak and the Preliminary assessments in the chances of being fined for non- six countries indicate that two factors compliance small, employers factored partly explain why equal pay laws have the costs of a fine into their overall as- had such limited effectiveness. The 25 first, and perhaps most important of CONCLUSIONS AND RECOM- these factors, is the structure of the MENDATIONS laws. The laws are oftimes too narrow in scope to have any real impact. Al- Remarkably few women partici- though this is true in many Latin pate in the labor market in Latin Amer- American countries, equal pay laws in ica. Towards the end of the 1980s, only Honduras provide a useful illustration around 35 percent of women in the re- of this. Here, the law allows for a very gion at large worked for pay. Time se- narrow interpretation as to when equal ries data from six selected countries, pay for equal work policies may be however, shows that very considerable applied; employees must work in pre: changes have taken place over the dec- cisely the same occupation, must work ade. The proportion of women working the same number of hours per week, for pay rose sharply throughout the and provide proof they are equally effi- 1980s. And, as more women have en- cient. Clearly, these provisions make it tered the labor force, they have become very difficult for workers to press better represented in the higher-paying, claims for unequal pay. In other coun- more stable job positions in the formal tries, broader provisions establishing sector. It also appears that women, once workers' rights to equal pay for similar in the labor market, choose to remain work have generally been used more there; they do not seem to withdraw effectively by workers claiming wage from the labor market because of child- discrimination. bearing. The second factor which has The increases in women's par- clearly limited the effectiveness of equal ticipation rates have been matched, in pay laws has been the institutional five of the six countries, by very con- weakness of enforcement agencies in siderable increases in women's wages most Latin American countries. These relative to men's. This undoubtedly re- agencies, typically Labor Inspection Bu- sults largely from the fact that women reaus, are generally severely under- workers now generally have the same, resourced, poorly equipped, and have or more, formal education than men. not received any training on issues re- But, the fact that women are increas- lated to women workers. Their weak- ingly moving into higher-paying job ness means that little pressure is brought positions must also partly explain the to bear on employers to comply with declining wage gap. these laws. Moreover, the next level at which workers can seek redress --the Despite these changes, women's labor courts-- also suffers inefficiencies labor force participation rates and rela- and problems. Not only is it generally tive pay remain inexplicably low in extremely costly for a worker to bring a most Latin American countries. These case to the courts, but it is also a very countries have entered a phase of rapid slow process; in many countries four economic growth in which they must years may pass before a case is heard. increase their competitiveness on the world market. This means that they must use their available resources, par- 26 ticularly human resources, as effectively All Latin American countries as possible. Presently, women are not have already instituted policies and laws fulfilling their real economic potential. designed to limit gender labor market They have high levels of formal educa- discrimination. A preliminary assess- tion but, for reasons that are unclear, do ment of these strategies, however, indi- not participate actively in the labor cates that they are not achieving their market. intended effects. In fact, in many cases it appears they are actually adversely Preliminary, and as yet tenta- affecting women's employment oppor- tive, studies in the six selected countries tunities and pay. They are often poorly suggest that women's low relative pay conceived, are not designed to be part and participation rates may be largely a of a coherent strategy to combat dis- consequence of "discrimination". crimination, are implemented without Analysis of earnings and employment sufficient consideration being given to data indicates that women are very the economic incentives they create, and probably subject to wage are sometimes only weakly and errat- "discrimination", meaning that they are ically implemented. The following paid lower wages than men even though problems and shortcoming in countries' they have the same human capital as policies and strategies are identified in men. Women also seem to experience the larger study: some limited "discrimination" in gain- ing access to higher-paying employment * Many of the labor laws govern- sectors and occupations. ing women's employment work counter to women's economic interests. Protective laws, for in- Implications for Policy stance, actually discriminate against women by limiting their If gender discrimination in the employment opportunities. Other labor market is indeed commonplace, laws, particularly the maternity there are several reasons why Latin protection laws and some child American countries should take steps to care provision laws, work to eliminate or reduce it. First, discrimi- raise the cost of female labor nation reduces economic efficiency by relative to male labor. The laws arbitrarily restricting women from high therefore establish incentives for productivity, high-paying jobs. This employers to discriminate against clearly prevents women from making female labor; employers can re- their maximum contribution to national duce operating costs by hiring output. Second, the returns to public male, rather than female, work- investments in women's education, ers. which have been very significant over the past few decades, will be lowered if * Laws and policies designed to women are blocked from making their ensure women have equal oppor- full economic contribution. And, third, tunities and rights in the market the countries' broader equity objectives place are largely ineffective. will be frustrated. Their wording and application is 27 generally too broad and diffuse to ties and whether they work to in- effect any real change. crease the female cost of labor relative to male's. This informa- * Laws which have potential to re- tion should help governments de- duce unfair labor practices and termine which laws adversely af- discrimination against women are fect women's employment oppor- not routinely enforced because of tunities and pay and provide the institutional weaknesses in the basis for further actions to modify Labor Inspection Bureaus. or revoke such laws. These factors suggest that there is . Ensuring laws governing a real need for governments to review women's employment are rooted policies and laws currently in place to in a consistent premise: In most reduce gender labor market discrimina- Latin American countries, differ- tion. A careful and deliberate assess- ent laws simultaneously discrimi- ment of the effects of these laws should nate against women, encourage help guide the revision and modification the differential treatment of of these policies so they work to en- women on the grounds that they hance women's employment opportuni- have different physical capabili- ties. Strong, carefully-conceived, and ties, and require the equal treat- well-implemented policies and laws ment of women. A disparate con- have been shown to be effective in re- glomeration of laws works to un- ducing gender labor market discrimina- dermine commitment to ending tion in many industrialized countries. discrimination by sending very Public action in Latin American coun- different and often contradictory tries can be similarly effective if it is messages about what is to be cate- well-conceived and implemented effec- gorized as discrimination. A clear tively. The findings of the present study and consistent approach to gender suggest that public policy reform would discrimination needs to be formu- need to be based on the following ac- lated in each country and laws and tions: policies modified accordingly. . Assessing how laws affect . Taking steps to blur gender dis- women's economic opportuni- tinctions in the law: In their pre- ties: The most immediate need is sent form, several laws, and par- for countries to systematically re- ticularly the maternity laws, work view the labor laws which govern to raise the cost of female labor women workers. This needs to be relative to male labor, thereby done with an eye to determining creating incentives for employers what effect they have on women's to discriminate against female la- economic opportunities. Special bor. It is important that, in grant- attention should be given to as- ing women specific benefits, the sessing whether the laws reduce law does not require employers to women's employment opportuni- bear the financial costs of 28 providing these benefits. It is also * Identifying necessary judicial re- important that laws be reformu- forms: In many countries the la- lated so that the benefits accorded bor courts and tribunals are over- to the workers are, as far as pos- burdened and very slow. In addi- sible, gender neutral. Laws which tion, highly formalistic procedures require employers to subsidize followed by the courts are costly benefits if they hire more than a and difficult for workers with less specific number of women, for education to comprehend. These example, are particularly injurious problems discourage workers to women and should be reworked from seeking reparation if their to be gender neutral. labor rights are violated. A careful review of the operation of the la- Strengthening existing enforce- bor courts would help identify ment mechanisms: Effective en- what steps could be taken to make forcement of carefully conceived them more accessible to their cli- laws is important if they are to ents. have their intended effect. Im- proved enforcement in Latin * Improving workers' knowledge American countries will require of their rights: Laws will have governments to strengthen the little effect if their target popula- institutional capacity of the Labor tion has limited knowledge of Inspectorate systems. As a first them. In many Latin American step, the operational budgets of countries women suffer discrimi- the Inspectorates will need to be nation and job loss because they increased. This should pave the are not informed about their way for better staffing composi- rights. Governments could reduce tions, improved training pro- discrimination by promoting grams, and greater responsiveness worker education about the labor to reported violations. Efforts law. Various groups could assist should be made simultaneously to with this process--unions, inter- increase the accountability of In- ested non-government organiza- spectorates through improved re- tions, and women's ministries, for porting procedures. example. 29 SELECTED REFERENCES Brown, R.S., M. Moon, and B.S. Zoloth. 1980. "Incorporating Occupational Attain- ment in Studies of Male-Female Earnings Differentials," Joural of Human Re- sources, 15(l):3-28. International Labor Office. 1993. Regulaci6n del Trabajo de la Mujer in America ILatina. ILO. International Labor Office. 1987. Women Workers: Protection or Equality. Conditions of Work Digest, Vol. 6, No. 2. Geneva: ILO. Oaxaca, R. 1973. "Male-Female Wage Differentials in Urban Labor Markets." Intrna- tional labor Review, 14(3):693-709. Psacharopoulos, G.P. and Z. Tzannatos (Eds.). 1992. Women's Employment and Pay inLatin America. Washington D.C.: The World Bank. Schuler, M. (Ed.). 1990. Women, Law and Development -- Action for Change. Wash- ington D.C.: OEF International. Winter, C. "Gender Discrimination in the Labor Market and the Role of the Law: Ex- periences in Six Latin American Countries." Unpublished Mimeograph. Latin American and Caribbean Technical Department. Washington, D.C.: The World Bank. Winter, C. and T.H. Gindling. 1992. "Women's Labor Force Participation and Earn- ings in Honduras," in Psacharopoulos, G.P. and Z. Tzannatos (Eds.). Women's Employnt and Pay in Latin America. Washington D.C.: The World Bank. 31 APPENDIX 1: ANALYTICAL METHODOLOGIES The Oaxaca Decomposition Oaxaca's (1973) technique decomposes the male/female earnings differential into two parts: a part which can be explained by gender differences in human capital endowments (E) (years of education, work experience, etc.) and a part that cannot be explained by differences in these productive characteristics (U). This second part is commonly taken to represent the "upper bound" or upper extent of "discrimination."' The decomposition is performed as follows: Let the mean log wages (W) of men (M) and women (F) be represented by: Wi = j,'Tx i=M,F (1) Where: = a vector of human capital characteristics of men and women = a vector of coefficients to be estimated The Oaxaca technique then decomposes the wage differential as follows:2 'FM WFFPM (XM XF)+Q(M F) XF (2) (E) + (U) Where: XM = a vector of average endowments for men One caveat about the Oaxaca decomposition should be noted here. The vector of mean huran capital characteristics (K ) in the equations may not capture all skill components that influence wage levels. Any omitted explanatory variables will cause the part of the wage differential not explained by human capital differences (U) to be larger. Hence, it must be recognized that the unexplained pay gap represents an upper bound to "discrimination." The difficulties in interpreting U are well documented in Cain (1986). 2 Note that the Oaxaca decomposition could be estimated in two ways: BmXm -Bf X = (Bm-B )Ym + Bf (X.,-Xf) = (B. - Bf )Yf + Bm (Xm - X ) where Bi i = m,f are the estimated coefficients of the earnings functions and Xi i = m,f are the averages of the explanatory variables in the earnings functions. There is no 'best solution" to this index number problem which is frequently experienced in applied economics. The country studies in this study estimate the Oaxaca using the male means. 32 XF = a vector of average endowments for women E = the portion of the wage differential explained by differences in male/female human capital endowments U = the portion of the wage differential that is not explained by differences in human capital endowments. U is actually the difference between women's present earnings and the earnings they would receive if they were paid the same wages as men for their human capital. Hence, the higher the value of U, the greater the level of labor market "discrimination." The Brown, Moon, and Zoloth Decomposition The principal methodology used in the six country studies in this study is the Brown, Moon, and Zoloth Decomposition (hereafter referred to as BMZ). The important contribution of this methodology is the recognition that the different occupational distributions of male and female workers could be a significant factor explaining male/female wage differentials. It recognizes that the distribution of male and female workers across occupations/sectors might be the result of labor market "discrimination;" women may be denied access to higher paying positions because of employer prejudice or because protective labor legislation prohibits their employment in certain sectors. The BMZ technique goes one step beyond the Oaxaca decomposition by treating average wages as weighted averages of wages for each occupation, where the weights are the employment shares of those occupations/sectors. Thus, while the Oaxaca technique decomposes aggregate wage differentials into two elements (coefficients and endowments), the BMZ does this and also identifies the impact of occupational distribution. The BMZ methodology requires a three stage analysis. The first stage estimates the parameters of multinomial logit equations for the choice of occupation/sector by gender. The estimated coefficients from the males' sample are then used to predict the occupational/sectoral distribution that would be observed for females if females had the same occupational attainment as males, given their observed characteristics. In the second stage, wage equations are estimated by gender for each occupation/sector. In the third stage, the estimated wages and occupational/sectoral distributions are used to decompose the gender wage differentials. It is assumed that women's observed occupational/sector attainment is somehow constrained, either by emnployer wage "discrimination," or unequal access to job opportunities. The BMZ decomposition method thus allows us to estimate how average wages would change if these restrictions were removed. The procedure also assumes that, in the absence of these restrictions, women's occupational/sectoral attainment distribution would be similar to men's. Hence, BMZ calculates the expected occupational attainment of women using the multinomial logit equation estimated for men and uses the difference between the actual and expected occupational/sectoral distribution as the basis for the estimate of the portion of the wage differential that is attributed to occupational segregation.3 3 As with the Oaxaca method, there is an index number choice to be made. The choice made will have a small effect on the results. 33 The BMZ methodology is developed as follows: First, let the proportion of men or women (i) working in a given occupation/sector (j) be represented by Pij. Then, let the mean natural logarithm of wages of men (M) and women (F) be represented by: Wii = i I j'Xj i(1) Where: Wii = the mean log wages of men or women in a given occupation/sector X = a vector of individual characteristics B = a vector of coefficients The BMZ technique treats male and female average wages as a weighted average of wages in various occupations/sectors. Hence: WM -WF = EJ P 'j M Mj XA4 - j (P 'XM4 [PM p FJ 'XFJ (2) Here, J is the portion of the wage differential due to the different occupationallsectoral distribution of male and female workers. W is the portion of the wage differential attributed to male/female wage differences within occupations/sectors (the "within sector wage differential"). It is possible to further decompose J and W into "explained" and "unexplained" portions along the lines of Oaxaca. By adding and subtracting Ij P M) XMPFJ and X. Z Fj PW QXF&, expression (2) becomes: WM -WF Ej (PF [j XFj XjXFj D+Z ( AIJXl PFJ F D Yi~(J'413;xAl, -P XFJj +Ji MX4P P)(3) Ej (PFj [P Mj xM pM FJD (j Mj [PMJ P^ D(3 = (WU) + (ju) +(WE) + (JE) JE is the portion of the wage gap that is explained by the different human capital endowments of male and female workers through its effect over sector allocation. It corresponds to the "across sectors explained wage differential." JU is the portion of the wage gap that occurs because male and female workers with the same measured human capital endowments are employed in different occupations/sectors. JU corresponds to the increase in females' wages that would occur if women had the same occupational/sectoral distribution as men. The change is calculated over males' actual wages. JU represents the "unexplained difference in occupational/sectoral allocation" or "discriminaton in occupational/sectoral allocation." 34 WE is the portion of the wage differential within occupations/sectors that is explained by differences in the human capital endowments of male and female workers. It yields the increase in female wages that would be observed if females had the same human capital endowments as males and if they were paid the samne wages as males. It is calculated as the difference in average wages that would result if we used the wage equation for males but assumed that the occupational/sectoral distribution is the sane as that of females workers. This is referred to as the "within sector explained wage differential." WU is the portion of the wage differential within occupations/sectors that is not explained by differences in the human capital endowments of male and female workers. It is a weighted average of the difference between what women earn now in each occupation/sector and what women would earn in each occupation/sector if they were paid according to the same wage structure as men. WU is thus referred to as the "within sector unexplained wage differential" or "within sector wage discrimination." Of course, to calculate WE, WU, JE, and JU, estimates of P.,, PFj, f3, and 1 FJ are needed. That is, equations for the probability of employment in each occupation/sector and wage equations estimated within each occupation/sector must be estimated. The probability of employment in each occupation/sector is estimated using a multinomial logit model where the probability that the average male or female works in sector j is: i-i l+ k- epSx" (4) Where: Pij = the proportion of men or women in each occupation or sector ,ij = a vector of sex- and sector-specific parameters j = represents the occupational/sectoral categories Xi = a matrix of exogenous individual characteristics The wage equations in each sector are estimated using a simple Mincer-type equation where log wages is the dependent variable. Selectivity Issues Much consideration was given to the potential effect sample-selection bias might have on the results of the six country studies. Sample-selection bias could be caused if the analysis were to be undertaken using non-random samples. A case in point would be an analysis of the differential returns to education for men and women which used a sample of working men and women. The sample of working women very probably is not representative of all women (working and non-working); working women are likely to be those with higher potential returns (i.e., higher education levels) and lower tastes for remaining at home. They are likely to be a self-selected sample. Similarly, it is also very possible that workers in a specific occupation/sector are a self-selected sample. 35 Various approaches are available to correct for selectivity bias, the most common of which is the Heckman-type correction used when estimating a multinomial logit model.4 The use of selectivity correction in the Brown, Moon, and Zoloth decomposition is, however, problematic because of the three-stage analysis required. To correct for selectivity bias in the different stages, a Heckman-type correction would need to be undertaken in the first stage and a set of Lee's Aj be used in the second stage to correct for selectivity in each of the occupation/sector wage equations.5 To use Lee's selectivity correction the multinomial logit estimation in the first stage must pass the test for Independence of Irrelevant Alternatives (IIA), which is based on the assumption that workers decide simultaneously to enter the labor force and to work in a particular sector.6 This assumption may not hold true for our model, however. Instead, it mnay well be the case that the real selectivity effect occurs at the stage when women select into the labor force. Tests were run in three of the country studies (Brazil, Costa Rica, and Venezuela) to determine whether selectivity corrections should be undertaken. The outcomes were clear and consistent in all three countries; selectivity corrections should not be undertaken. In all the countries, for most years, most sectors, and for both sexes, the selectivity correction term in the wage equation was not statistically significantly different from zero.7 Performing the HA test on the multinomial logit equations for Costa Rica and Venezuela also pointed to the inappropriateness of selectivity corrections. The test indicated that the hypothesis of IIA should be rejected. This would mean that if the studies corrected for selection bias in the female sample they would possibly be using inconsistent estimates of the parameters of the selection equation. Moreover, when estimating the second stage equation by OLS as ln(wage, )= y 'Zj + (pp)" j +rT the coefficients for x are not statistically different from 0 (using a simple t test). Hence, the decision was made not to correct for selectivity; an uncorrected OLS is used in all six country studies to estimate the wage equations. 4The methodology for this is developed in Heckman (1979) and Lee (1983). 5 Using Lee's method, a selectivity correction term would be added in estimating the wage equations. 6 Hausman and McFadden (1984). 7This conclusion was drawn from the Ordinary Least Squares (OLS) standard errors. In the presence of selectivity, the OLS standard errors are heteroskedastic and not consistently estimated. Under the null hypothesis that selectivity does not exist OLS standard errors consistently estimate the true standard errors.