Growing Pains: Cross-National Variation in Sex Segregation in Sixteen Developing Countries

by Mariko Lin Chang
Growing Pains: Cross-National Variation in Sex Segregation in Sixteen Developing Countries
Mariko Lin Chang
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License

Growing Pains: Cross-National Variation in Sex Segregation in Sixteen Developing Countries

Mariko Lin Chang

Department of Sociology, Haward University

Using an occupational scheme with 43 categories (rather than the typical seven broad categories), this paper provides new information on cross-national patterns of occupational sex segregation in sixteen developing countries. Cross-national variation is investigated through the lens of a new theoretical framework that incorporates the importance of state policies that influence women's integration across the occupational structure. Multivariate analyses examine how three state policies (maternity leave, anti- discrimination, andprotective legislation) are related to levels andpatterns of segregation, while controlling for existing explanations for dzfferences in segregation in developing countries: economic structure (economic development, dependency, and size of the service sector) and human capital. State policies were the strongest determinant of cross-national dzfferences in levels of segregation, but two types of legislation had opposite effects: maternity leave is negatively related to levels of segregation while anti- discrimination legislation is associated with higher levels of segregation. Analyses of occupation-specific patterns of segregation reiterated the importance of state policies, demonstrating how different types ofpolicies have segregative and integrative effects throughout the occupational structure. Results also provided strong evidence of a culling effect, in which women 's representation in higher-status occupations declines as the

percentage of women in the labor force increases.

ross-national studies of sex segregation in industrialized countries have flourished, but few studies have examined levels of segre- gation in developing countries specifically (Anker 1998; Blackburn et al. 2000; Clark 199 1; Jacobs and Lim 1992; Rarnirez and Weiss 1979; Semyonov and Shenhav 1988; Semyonov and Jones 1999; Weiss, Ramirez, and Tracy 1976). Most studies conducted thus far were limited to

Direct all correspondence to: Mariko Chang, Department of Sociology, Harvard University, 33 Kirkland Street, Cambridge, MA 02138 ( This research was sup- ported by the Milton Fund, Harvard University. I wish to thank David Grusky, Barbara Reskin, and anonymous reviewers for graciously providing con- structive comments. JeeYoung Kim and Lin Tao pro- vided invaluable research assistance and Naomi Cassirer provided supplemental information regard- ing the ratification of ILO conventions.

highly aggregated occupational data, in which the distribution of men and women across occu- pations was measured using only a few broad categories: professional, managerial, clerical, sales, service, agriculture, and production (but see Anker 1998; Blackburn et al. 2000; Jacobs and Lim 1992; Semyonov and Jones 1999).'

Blackburn, Jarman, and Brooks (2000) employ detailed occupational categories in their analysis of the link between occupational sex segregation and gender inequality. They do not, however, address patterns of segregation (that is, how men and women are distributed across the occupational structure). Jacobs and Lim (1992) rely primarily upon aggre- gated occupational categories; however they con- duct analyses of trends in segregation using detailed occupational data (19 categories) for four develop- ing countries. The latest figures they have are for 1961 and, like Blackburn et al. (2000), they do not address patterns of segregation.

Much segregation is hidden at this level of aggregation. For example, while the profes- sional category as a whole may appear to be fair- ly well integrated by sex, the occupations that comprise the professional category may be quite segregated. At this level of aggregation, tradi- tionally female-dominated occupations such as nursing and teaching are in the same category as traditionally-male occupations such as doc- tors and physicists. Important cross-national differences are muted when such broad occu- pational categories are employed. Consequently, it is important to have a detailed occupational classification scheme and to examine which types of occupations are most segregated and by which sex.

This paper makes four contributions to the lit- erature on cross-national variation in occupa- tional sex segregation in developing countries. The first contribution is its use of more detailed occupational data, permitting the examination of segregation across 43 occupations rather than the usual six or seven. This high level of detail provides a more comprehensive portrait of seg- regation than has been possible to date. It also allows for a more rigorous examination of com- peting explanations for cross-national variation in segregation in developing countries.

The second contribution is theoretical. This paper is the first to examine explicitly the effects of state policies in developing countries that fall into two general categories: (1) policies that facilitate women's access to the labor market by helping women combine dual roles as workers and mothers, and (2) legislation that affects the integration of women across the occupational structure. The predictive ability of these types of state policies is tested in comparison to two general explanations for cross-national variation in segregation in developing countries: eco- nomic structure and human capital.

The third and fourth contributions to research on sex segregation in developing countries are methodological. The third contribution is that the analyses move beyond simple summary indices of overall levels of segregation and instead describe patterns of segregation (how men and women are distributed across the occu- pational structure). The fourth contribution is the use of log-linear models that allow for the pos- sibility that theoretical explanations for cross- national differences in sex segregation do not have uniform effects across the occupational structure. For instance, in her study of developed countries, Charles (1992) found that women's percentage of the labor force was positively associated with their access to clerical and man- agerial occupations, but negatively associated with their representation in professional occu- pations. We can learn more about the reasons for segregation if we recognize that different fac- tors may lead to male and female over-repre- sentation in different occupations.


This paper begins with a review of existing frameworks for understanding cross-national variation in sex segregation that fall within two general categories: economic structure (level of economic development, economic depend- ency, and size of the service sector) and human capitaL2 I then advocate greater attention to a variable that has to date received little attention by sex-segregation scholars: gender-related state policies.

The first set of explanatory factors focuses on three key characteristics of a country's eco- nomic structure: level of economic develop- ment, economic dependency, and size of the service sector3


Cross-national variation in levels of economic devel- opment is one of the most influential explanations for differences in the structure of the economy and economic stratification gen- erally. Modernization theories and other struc- tural-functional arguments posit that as

Other possible explanations, such as sex-role socialization, discrimination, and gender egalitari- anism will not be investigated here due to the lack of data for the countries in the analysis. Nevertheless, the analyses here cover the explanations that are most frequently found in the literature addressing occupational sex segregation in developing coun- tries and many of the findings speak to these alter- native explanations.

None of the explanations reviewed is intended to be complete explanations, but rather factors that influence cross-national variation in sex segregation in developing countries.

economic development increases, the alloca- tion of individuals across the social hierarchy shifts from ascription toward more universalis- tic, merit-based mechanisms (Kerr et al. [I9601 1994; Inkeles and Smith 1974; Parsons 197 1). The corresponding extension of citizenship rights to women and their greater access to edu- cational, economic, political, and social insti- tutions also facilitates women's access to more desired positions within society (Ramirez and Weiss 1979; Weiss et al. 1976). With modern- ization, fertility also generally declines, freeing women somewhat from extensive childrearing responsibilities and strengthening their ability to participate in the formal labor market. Hence, with economic development, gender-based inequalities in the labor market, including occu- pational segregation, should erode as more merit-based processes of attainment replace ascriptive processes.

Perhaps because the interrelationship between economic development and the status of women is incredibly complex (Boserup 1970; Nash and Fernandez-Kelly 1983; Tinker 1990; Ward 1984, 1990), the effect of level of economic develop- ment upon segregation is inconsistent across studies. For instance, while Clark (1991) found a positive relationship between level of eco- nomic development and women's access to high- er-status occupations (professional and managerial occupations), others found a nega- tive effect (Semyonov and Shenhav 1988), and still others found no statistically significant effect at all (Nuss and Majka 1983; Ramirez and Weiss 1979; Semyonov and Jones 1999; Weiss et al. 1976). In the few studies to examine the effects of economic development on levels of segregation beyond the high-status occupations, similar inconsistencies were found. For exam- ple Jacobs and Lim (1992) found that level of economic development was positively related to levels of segregation and Anker (1 998) found no statistical relationship.

ECONOMIC A second aspect of

DEPENDENCY. economic structure that may affect occupational segregation is economic dependency. Dependency theory argues that national devel- opment is affected by a country's position in the world system. More developed countries exploit the natural resources, cheaper labor, and often lax environmental regulations in developing countries. As a result, developing countries have become locations for the manufacturing of cap- ital-intensive products for export, shifting third- world residents from agriculture to jobs in factories and the tertiary employment sector, where they serve as cheap labor. Some typical features of dependency include the growth of export-based manufacturing, multinational cor- porate penetration, investment dependence, and export-processing zones. Although it has been argued that dependency may actually improve certain aspects of women's status (Lim 1990), most have found that the export of labor-inten- sive jobs to developing countries has resulted in the exploitation of women (particularly young unmarried women) to fulfill tedious capital- intensive jobs characterized by low wages, high turnover, and few benefits mash and Fernandez- Kelly 1983; Tinker 1990)~ Hence, dependen- cy increases occupational sex segregation by relegating women to the least desirable positions in the labor market (particularly the tertiary sector and labor-intensive production occupa- tions) and by restricting women's access to high- er status occupations.

Few studies have examined the effects of dependency on cross-national variation in occu- pational segregation; however there is evidence that dependency decreases women's access to professional and managerial occupations (Clark 1991; Semyonov and Shenhav 1988).5 While these studies shed light on women's access to managerial and professional occupations, they did not reveal how dependency may affect lev- els of segregation across the occupational stmc- ture. In the only study of the effects of dependency on segregation across the entire occupational structure, Semyonov and Jones (1999) found no evidence that dependency affectedlevels of segregation-but did find that it was a significantly negative predictor of women's access to higher-status occupations.

The demand for women to fill capital-intensive jobs may make it more difficult for some less-skilled men to gain access to the formal economy.

Nevertheless, each study utilized different meas- ures of dependency. Clark (1991) used a measure of multinational corporate penetration, Semyonov and Shenha\r (1988) employ a measurt: of in\.estment depcndencc, and Semyono\. and Jones ( 1999)incorporate dummy t.ariab1r.s to represent a nation's posi- tion in the world system-periphery. seml-per~phery. and core (reference group).

SERVICESECTOR.The size and growth of the service sector is linked to occupational sex seg- regation in numerous ways. First, segregation generally increases as the service sector expands, because responsibilities for services such as childcare, cleaning, and meal prepara- tion shift to the marketplace, where such occu- pations are filled primarily by women (Charles 1992; Chang 2000). Many service-sector jobs are particularly attractive to women with care- giving responsibilities because of the opportu- nities for part-time work and, in some cases, opportunities for self-employment (Lee and Hirata 200 1). Yet, the growth of the service sec- tor affects segregation in other ways as well. An expanding service sector may also pull women away from manufacturing because of the increased flexibility offered by many service jobs and because they are perceived as more consistent with women's domestic skills.

Empirical research on industrialized countries reveals that the size of the service sector is asso- ciated with higher levels of segregation overall6 (Charles 1992; Charles and Grusky, forthcom- ing; Nermo 2000) because it increases women's representation in the already female-dominat- ed service occupations and decreases women's representation in the already male-dominated production and managerial occupations.

Although researchers assume that the expan- sion of the service sector should have similar effects on segregation in developing countries7 (for example, Semyonov and Shenhav 1988; Marshall 1985), the only empirical study to test this proposition specifically, using a sample of 56 developed and developing countries, found no statistically significant effect of the size of the service sector on occupational sex segrega- tion (Semyonov and Jones 1999).8 The research presented here will be the first to employ

In her analysis of labor markets in the United States, however, McCall (2001) found that service- sector growth was associated with more occupa- tional integration.

'However, women working in the service sector in developing countries are more likely than service workers in developed countries to be in the informal sector, employed in occupations such as domestic ser- vants (Anker and Hein 1986).

The study by Semyonov and Jones (1999) employed data with only seven major occupational categories.

detailed occupational data to test empirically whether the size of the service sector is associ- ated with higher levels of segregation in devel- oping countries.

A second dominant explanation for inequalities in labor market outcomes broadly is differences in human capital. Human capital theory con- tends that individuals make rational decisions to invest in their human capital based on their expectations of future payoffs to their invest- ment. Because women's labor force participa- tion is assumed to be hindered by marriage and children, most women will either invest less in their job-related human capital or will choose the types of occupations for which they are able to retain favorable returns to their human cap- ital despite altering their labor force participa- tion to meet the dual demands of work and family. Occupations that provide flexible hours or lesser penalties for reentry after a period of absence are considered to be particularly appeal- ing to women because they allow women to maximize their returns to human capital with- in the constraints of childrearing responsibili- ties. For instance, teaching is described as being particularly appealing to women because the work day corresponds to the hours their children are likely to be in school. Teaching also permits relatively easy reentry should a woman take time out to take care of children before return- ing to work.

In a similar fashion, occupations that require more continuous labor force participation or full-time hours (such as many high-level man- agerial occupations) are likely to have few women--either because women do not chose these types of occupations or because employ- ers engage in statistical discrimination (by assuming that all women will alter the nature of their labor market participation if they marry or have children).

Increases in women's labor force participation are often considered indications of women's increasing labor force attachment and employ- ment continuity (which generally indicate high- er levels of human capital). However, women's odds of working in managerial and profession- al occupations are negatively related to rates of female labor force participation (Semyonov 1980; Semyonov and Shenhav 1988; but see Clark 199 1). And the few studies to examine segregation across the occupational structure more broadly have found that women's labor force participation is inversely related to levels of segregation (Anker 1998; Jacobs and Lim 1992; Semyonov and Jones 1999). In other words, as women's labor force participation increases, overall segregation decreases; but at the same time, women's odds of working in the higher-status professional and managerial occu- pations declines. To interpret these results, scholars have argued that when few women are in the labor force, these women are likely to be

(1) highly educated (relative to the population) and nontraditional, or (2) working due to eco- nomic necessity (Charles and Grusky 1995, forthcoming; Goldin 1990; Lewis-Eptsein and Semyonov 1992; Nermo 2000). This selective recruitment of women into the labor force includes a larger proportion of highly educated and career-oriented women who are employed in professional occupations. As more and more women enter the labor force, the composition of the female labor force becomes increasingly diverse (with respect to levels of education, skill, family responsibilities, and labor force attachment), and women become more equally distributed across the occupational structure, resulting in a decline in levels of segregation and a decline in women's odds of working in high- er-status occupations.

Another indicator of human capital is edu- cational attainment. Empirical findings demon- strate that women's participation in higher education increases women's access to profes- sional and managerial occupations (Clark 199 1 ; Ramirez and Weiss 1979; Weiss et al. 1976). Once again, however, studies of segregation across the occupational structure yielded incon- sistent relationships between levels of segrega- tion and women's educational attainment (Anker 1998; Jacobs and Lim 1992).

As the preceding discussion of empirical find- ings suggests, economic structure and human capital do not affect occupational sex segrega- tion in uniform ways across countries or across different types of occupations. The inconsis- tent findings, however, should be viewed in light of the limitations in data and method.

First, the majority of these studies used high- ly aggregated occupational categories, severe- ly limiting the researcher's ability to observe the

tremendous diversity and nuances of segregation across countries. Furthermore, many of these studies examined women's representation only in the more prestigious professional and managerial occupations. It is difficult to com- pare directly the results of studies focusing on women's access to professional and manageri- al occupations with those that focus on the entire occupational structure. It is entirely plau- sible that factors related to women's access to higher-prestige professional and managerial occupations may have different effects when used to predict segregation across the occupa- tional structure more broadly.Secon4 previous research was limited by reliance on summary indices, such as the Index of Dissimilarity, to represent the amount of segregation across countries. Summary indices provide only a lim- ited piece of information about the nature of sex segregation since they do not reveal which occu- pations are the most segregated and by which sex. Segregation can have different implica- tions for women's status, depending on where women and men are over-represented. Women's absence at the top of the occupational structure is considered more detrimental to gender equal- ity than women's absence at the bottom. For such reasons, sex segregation research should examine both levels of segregation and more descriptive indicators of the distribution of women and men across the occupational


This study overcomes these limitations because it utilizes new data with far more occu- pational categories than were available previ- ously and because it explores the effects of covariates on both levels and patterns of segre- gation. But this paper makes additional contri- butions to the literature on cross-national variation in segregation by providing a theoret- ical framework for exploring ways that state policies contribute to cross-national variation in patterns of segregation in developing countries.


The significance of state policies in explaining cross-national variation in sex segregation has been established only for developed countries (Chang 2000); however, state policies should be important for developing countries as well. In fact, there is every reason to believe that they may be particularly important because employ- ers in developing countries are more likely to practice overt discrimination against women, such as stating preference for male or female applicants in job advertisements and discour- aging women from seeking employment in cer- tain jobs (Anker and Hein 1986; Anker 1998; Brinton 2001; Nash and Fernandez-Kelley 1983). Legislation may be one of the only fac- tors countering such segregative pressures.

Studies of industrialized countries reveal two mechanisms through which state policies could impact occupational sex segregation (Chang 2000). The first mechanism consists of poli- cies that facilitate women's access to the labor market by providing services that help women combine employment and childrearing respon- sibilities. The second mechanism consists of legislation that promotes or inhibits women's unrestricted access to occupations. Both mech- anisms and their implications for sex segrega- tion in developing countries are reviewed in the following pages.

Maternity leave policies are the most wide- spread mechanism through which states facili- tate women's dual roles as workers and mothers. In more industrialized countries, states may also facilitate the combination of work and par- enthood by establishing state-sponsored child- care facilities. These types of policies are likely to affect occupational sex segregation in two ways, each of which fosters occupational inte- grati~n.~

First, maternity leave policies enable women to remain in the labor force more con- tinuously, allowing them to develop more years of work experience and greater labor force attachment. As women's labor force attachment and job-related human capital increases, they are more likely to be able to compete with men for jobs. Labor market attachment is critical in

However, some researchers (Anker 1998; Boserup 1970; Safa 1983)have warned that the exis- tence of maternity leave policies may render employ- ers more reluctant to hire female workers because

they are potentially more cost~y~espec~a~~y


employers bear the cost of maternity leave rather than the state.

lo Labor force attachment is also critical in many professional occupations, but less so in teaching and nursing, the most common professional occupations for women.

many manageriallo occupations and should therefore facilitate women's representation in these more elite male-dominated occupations. In addition, maternity leave policies may also weaken the tendency for certain jobs to be char- acterized by high turnover as young women continually exit to assume childrearing respon- sibilities, as occurs in some production occu- pations in developing countries.Second, maternity leave policies may encourage (or reflect) more gender egalitarian attitudes in the formal economy, as employers and employees come to expect women's labor force participa- tion to be decoupled from their fertility. As ideals of gender egalitarianism increase, women should seek entrance into occupations beyond those traditionally open to them and employers should be more willing to hire women workers for a wider range of occupations.

While policies such as maternity leave help fos- ter occupational integration by facilitating the combination of motherhood and paid employ- ment, a second type of state policy is important for reducing occupation-based sex discrimina- tion: legislation that provides women with unre- stricted access to all occupations. Two contradictory types of legislation contribute to women's occupational access: anti-discrimina- tion legislation and protective legislation. Anti- discrimination policies may help to open up opportunities to women directly, and they may also have a positive affect upon cultural beliefs about gender egalitarianism. On the opposite extreme, protective legislation should increase segregation by restricting women from certain types of work. Common examples of protective legislation include the prohibition of women from working underground at night, or with dangerous substances.


LEGISLATION. most basic level, anti-discrimination legislation is likely to open up opportunities for women throughout the structure by pro-

viding them with a legal justification for seek- ing entry into all occu~ations and by discouraging employers from discriminating against women." In addition, these policies


The effectiveness of anti-discrimination legis- lation depends on enforcement, which is sometimes

may contribute to (or reflect) a cultural shift in gender egalitarian norms, further legitimating women's position in the labor market and legit- imating their entrance into male-dominated occupations.

Although the effects of anti-discrimination legislation are likely to be broad, other researchers (Charles 1992; Chang 2000; Goldin 1990) have argued that in industrialized coun- tries, the cultural and legislative pressures brought to bear by these types of laws are most strongly felt in the more prestigious professional and managerial occupations, where jobs are typically allocated on the basis of educztional qual- ifications and less on ascriptive criteria.

Because the effectiveness of anti-discrimi- nation legislation is likely to be felt most strong- ly in the formal labor market, these policies may have uneven effects across occupational categories. The large, informal sector in devel- oping countries is most likely to include occu- pations in the service, sales, and production categories. Therefore, I expect that the effects of anti-discrimination legislation are likely to be weaker in sales, service, and production occu- pations where the informal labor market is larg- er. And conversely, anti-discrimination legislation should be more effective in the cler- ical, managerial, and professional occupations, where most jobs are in the formal labor market.

Anti-discrimination legislation, however, is likely to have additional effects that may actu- ally counteract the integrative aspects men- tioned earlier. Even if women are allowed by law to pursue the occupation of their choice, they are influenced by cultural assumptions about their role as wives and mothers that influence the types of jobs they are likely to pursue. Under these circumstances, married women in devel- oping countries will exercise their improved access to the labor market by seeking employ- ment that allows them to combine their pro- ductive and reproductive roles-jobs that provide flexible hours, part-time work, and short commutes. Consequently, married women will press for entrance into the types ofjobs that are most compatible with family obligations, such

weak in developing countries (Anker and Hein 1986; Brinton2001).Nevertheless, anti-discrimination leg- islation is an important (even if merely symbolic) first step in facilitating women with equal opportunities throughout the occupational structure.

as sales and service jobs because there are opportunities for flexible or part-time schedules and self-employment (Lee and Hirata 2001).

PROTECTIVE The second form of

LEGISLATION. legislation that affects women's access to occu- pations is protective legislation. Protective leg- islation restricts women's access to occupations with certain characteristics, such as working with dangerous substances, working under- ground, working at night, or carrying heavy loads. Although a wide variety of occupations are affected by protective legislation (particu- larly legislation prohibiting night work), I expect that the effects of protective legislation are most llkely to be evidenced in the production occupa- tions by virtue of the types of restrictions-par- ticularly the restriction on underground work and carrying heavy loads. In addition to baning women from certain occupations, protective legislation may Influence segregation more indirectly by rein- forcing the message that it is legal, and perhaps even preferable, for women to be banned fi-om cer- tain occupations, reinforcing the segregation of occupations by sex more generally.

But, there may be some counterintuitive effects. Whereas anti-discrimination policies may actually increase segregation in some female-dominated occupations by drawing mar- ried women into sales and service occupations, protective legislation may inhibit women's labor force participation,12 resulting in the more selec- tive recruitment of women into the labor mar- ket. Because the culling of women into the labor market typically results in a higher proportion of women workers who are highly educated and career-oriented, protective legislation may actu- ally increase the presence of women in profes- sional occupations that have traditionally attracted women, such as nursing and teaching.

To conclude, I have argued for the importance of state policies in shaping patterns of occupa- tional sex segregation and described a theoret-

l2 Women's labor force participation is likely to be inhibited because protective legislation restricts women's opportunities in specific occupations and also because protective legislation reinforces cultur- al norms that place women's primary societal roles outside the labor market, as wives and mothers.

ical framework for examining the likely seg- regative or integrative effects of different types of state policies throughout the occupational structure. Maternity leave should integrate women across the occupational structure because it facilitates women's continuous labor force participation and thus increases their human capital, rendering women in a better position to compete with men for jobs broadly. Anti-discrimination legislation should help to open up opportunities for women to enter the more prestigious white-collar professional, man- agerial, and clerical occupations because of the emphasis of educational criteria over ascriptive criteria and because these types ofjobs are more likely to be found in the formal labor market where anti-discrimination legislation is more effective. However, anti-discrimination policies may increase segregation in sales and service occupations because they help draw married women with family responsibilities into the labor market and into the types ofjobs that pro- vide flexible schedules and part-time opportu- nities (typical of many sales and service occupations). Protective legislation should increase segregation, particularly in production occupations. But, ironically, protective legisla- tion may contribute to the culling of highly edu- cated, career-oriented women into the labor market, resulting in women's greater represen- tation in professional occupations in compari- son to other occupational categories.


Data on the distribution of men and women across occupations for eleven countries were obtained from the "Segregat" database pub- lished by the International Labour Organization Bureau of Statistics. These data were compiled by the ILO and derived primarily from popula- tion censuses and labor force surveys. To sup- plement these data, I used published census data for seven countries that were either not included in the ILO database or for which I located more recent data. The total sample con- sists of 16 developing countries: Angola, the Bahamas, Bahrain, China, Costa Rica, Cyprus, Egypt, Ghana, Haiti, Hong Kong, India, North Korea, Kuwait, Mauritus, Singapore, and Taiwan.l3

l3 Whereas the countries in this sample are at dif- ferent levels of development, all (except Taiwan) are

Data on the distribution of men and women across occupations refer to 1990, or the closest available year. Eleven of the 16 countries had data pertaining to the 1989-1 992 time frame and the oldest data refer to 1980.14 A list of coun- tries, data source, and year are provided in Appendix A.

When publishing data on the distribution of men and women across occupational groups, some countries used their own classification scheme and others used the 1968 or 1988 International Standard Classification of Occupations, created by the ILO. In rendering the categories comparable across countries, it was most effective to base the common classi- fication scheme on the 1968 International Standard Classification of Occupations and then adjust the level of detail (by combining some of the categories when necessary) to ren- der one common classification scheme that worked for all 16 countries.15 The resulting classification scheme contains 43 occupations that fall within six standard broad occupation- al categories that correspond to conventional occupation-based representations of social class:

(1) professional and technical, (2) administra- tive and managerial, (3) clerical, (4) sales, (5) service, and (6) production, operators, and labor- ers.16 The complete list of occupational cate- gories is provided in Appendix B.

classified as "developing countries" by the United Nations (1995). The United Nations does not publish data on Taiwan; however, its GDP (an indicator of level of development) suggests it can be included with other developing countries. Nevertheless, results are unchanged when Taiwan is omitted from the analyses.

l4 Singapore is the only country with 1980 data. Although Singapore did conduct a population cen- sus in 1990, the published tables did not contain the distribution of men and women across detailed occu- pational categories

lS The logic underlying the creation of a common classification scheme is similar to the procedure used by Treiman (1 977), in which occupations were clas- sified into a series of "levels": major group, minor group, unit group, and occupation. The classification scheme used in this analysis differs because, given the different occupational classification schemes across the countries in this sample, certain occupa- tional categories had to be combined.

l6 The classification of agricultural workers is par- ticularly problematic because of the tremendous inconsistencies across countries in the classification


Two indices of occupational sex segregation are presented: the Index of Association, A, (Charles and Grusky 1995) and the Index of Dissimilarity,D. Each index provides a related but distinct view of segregation. The Index of Association represents the average factor by which women or men are over-represented in each occupation. If women are equally repre- sented across occupations, A equals 1.The more segregated men and women are from each other, the higher the value of A. The Index of Association weighs all occupations equally. The Index of Dissimilarity represents the proportion of men or women who would have to change occupations in order for the distribution of men and women to be equal across occupations. Unlike the Index of Association, occupations with fewer workers have less weight than occu- pations with more workers. I use the Index of Association to measure levels of segregation because it is not influenced by cross-national variability in the relative size of occupational categories, thus allowing a clearer distinction between cross-national variation in segregation and cross-national differences in the sizes of occupational categories. Nevertheless, I con- ducted analyses using both indices and found that the results are consistent across indices. The equations for both indices are presented below:

where J represents occupation, M, and F, rep- resent the number of men and women in the jth

and inclusion of agricultural workers, particularly unpaid family workers. Because of these problems, most researchers exclude agricultural occupations from their analyses (Anker 2000; Chang 2000; Charles 1992; Charles and Grusky 1995), and I do so here. Nevertheless, for purposes of comparison, I estimated models that included the agricultural cat- egory (comprised of three occupations), and the results were not substantively different.

occupation, and where M and F refer to the number of men and women in the labor force.

One of the main advantages of this analysis is that it moves beyond presenting an overall measure of segregation and examines patterns of segregation. To do this, I ran a saturated log- linear model for each country that fits a three- way association between occupation, sex, and country after purging the data of detailed-cate- gory compositional effects (see Charles and Grusky 1995, forthcoming). Purging the detailed-category effects ensures that the inter- action parameters at the major-category level are unaffected by the relative size of the detailed occupations comprising the major categories. This saturated model provides measures of seg- regation for each major occupational category and is represented by the following equation:

where i represents sex, j represents occupa- tion, k represents country, 2, is an indicator variable for gender (Z1= 0 and Z2 = I), qck refers to the scale values for the major occupa- tional categories (indexed by c), vjkrepresents the scale values for each of the 43 occupations that make up the major categories. To identify the model, the sex-by-country interaction effects have been constrained to sum to zero, with neg- ative values indicating male over-representa- tion, positive values representing female over-representation, and zero representing an equal distribution of men and women. The scale values for the major (or "broad") occupational categories incorporate the level of detail pre- sented by the 43 detailed occupational cate- gories. Each major category scale value can be though of, roughly, as the average of the scale values of the occupations constituting that broad category. (For example, the scale value for the professional category can be roughly thought of as the average scale value for the nine occupa- tions that make up the professional category.) But, the resulting major category scale values are adjusted so that they sum to zero across the six broad occupational categories.

The scale values provide a description of pat- terns of segregation and also a measure of the degree of female-representation across occu- pations that can be used in multivariate analy- ses. I use OLS regression to examine the strength of the three explanations for cross- national variation in both the level of segrega- tion, as measured by the Index of Association, and upon the patterns of segregation, as repre- sented by the occupation-specific scale values taken from the saturated model presented earlier.

The following independent variables are used:l7

ECONOMIC Three variables rep-

STRUCTURE. resent structural features of the economy: level of economic development, export dependency, and size of the service sector. Level of eco- nomic development is measured by the log of GDP per capita in U.S. dollars and export dependency measures exports as a percent of GDP, an indication of the extent to which trade in developing countries is focused on exports to developed countries (see Marshall 1985). Both measures were obtained from the World Bank (200 1) and from the 199 1 Statistical Yearbook of the Republic of China.

Dependency can be measured in many ways and previous studies of the effects of depend- ency on levels of segregation used different measures of dependency (see note 5). I considered using three different measures of depend- ency: export dependency, investment dependency, and multinational corporate pen- etration.18 The decision to use export depend- ency was driven by two factors: (1) data were available for all 16 countries in the analysis (which was not true of multinational corporate penetration), and (2) zero-order correlation coef- ficients were highest between export depend- ency and both investment dependence (r = .526) and multi-national corporate penetration (r = .565). Hence, the effects of export dependency on segregation were likely to be highly corre- lated with the effects of the two alternative measures of dependency. Nevertheless, I also conducted parallel analyses substituting export

l7 Covariates refer to the year closest to (but not greater than) the year the sex segregation data were collected. Further information on the year to which each covariate refers are available upon request.

l8 Investment dependency is measured by inward foreign direct investment stock as a percentage of GDP (obtained from the United Nations 2001) and the measure of multinational penetration was obtained from Bornschier and Chase-Dunn (1985), which was available for 10 of these 16 countries.

dependency with investment dependency and also with multinational corporate penetration and the results were identical except when noted.

Size of the service sector is measured as the percentage of the labor force employed in serv- ice indu~tries.'~ Data were obtained from the United Nations (1995) and the ILO (1999).


While human capital expla- nations focus on individual-level effects, the general arguments posited by human capital theory are often applied at the aggregate level.20 Two measures of women's human capital are commonly found in the literature on sex segre- gation and I include both here: women's share of tertiary education and women's percentage of the labor force.21 These two measures were obtained from the World Bank (2001), the United Nations (1995, 1996), and the 1991 Statistical Yearbook of the Republic of China.


Drawing on previous work (Chang 2000), I used three indicators of state policies, each designed to capture distinct dimensions of the state's commitment to occu- pational equality within the labor market. The first state policy dimension reflects the avail- ability of substantive benefits that facilitate the

l9 The ILO classifies the following industries with- in the service sector: wholesale and retail trade and restaurants and hotels; transport, storage, and com- munications; financing, insurance, real estate, and business services; community, social, and personal services.

Data limitations prevented the incorporation of individual-level factors. Results should be interpret- ed accordingly and any effects cannot be attributed to the human capital of individual women, but to human capital characteristics of the female labor force generally.

21 The zero-order correlation coefficient for these two variables is -. 170. Fertility is sometimes used as an indicator of women's human capital. It is exclud- ed, however, from the analyses here because (a) its high correlation with both the level of economic development (-,639) and the presence of maternity leave (-,642) and (b) because the effects of fertility upon the distribution of women across the occupa- tional structure are probably best realized indirectly through women's labor force participation and edu- cational attainment.

combination of motherhood and paid employ- ment. Here, the presence of maternity leave is used to measure substantive benefits.22 Data on maternity leave was obtained from the ILO's Conditions of Work Digest (1987) and the U.S. Department of Health and Human Services'

Social Security Programs Throughout the World

(1990, 1999).

The second mechanism consists of legislation that promotes or inhibits women's unrestricted access to all occupations. This mechanism is reflected in two indices that measure related but conceptually distinct types of legislation that affect women's access to all occupation^.^^ The first index refers to anti-discrimination leg- islation and is comprised of two measures: (1) the ratification of ILO Convention 1 1 1, which prohibits discrimination against women in employment and occupation, and (2) the ratifi- cation of the United Nations Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW).~~

The second index refers to protective legislation and is comprised of three measures: (1) ratification of ILO Convention 89, which prohibits women from night work, (2) ratification of ILO Convention 45 which prohibits women from working under- ground and (3) legislation restricting women from the manual transport of heavy loads. Each country's score on the two indices is presented in Table 1.Unfortunately, complete information was not available for Korea, Hong Kong, and

22 Chang's (2000) measure was based upon Wilensky's (1990,2002) family benefit index, which incorporates both maternity leave and the availabil- ity and accessibility of public daycare programs and government efforts to expand daycare. However, this index does not include developing countries. Because of the lack of data on public childcare in developing countries, I use only the maternity leave variable.

23 Chang (2000) combined these two dimensions into a single index and results were consistent when I also combined both dimensions. To retain the con- ceptual distinctions between anti-discrimination leg- islation and protective legislation, however, two indices are employed here.

24 Chang's (2000) index of equality of occupa- tional access included an additional measure: affir- mative action policies. But because none of these countries has affirmative action policies, the index would not differ if I had included this measure. Furthermore, Chang's (2000) index does not include equal pay legislation because it does not seek to inte- grate women across the occupational structure.

Taiwan because they were not members of the ILO (Korea became a member in December, 1991). To avoid deleting these countries from the analyses altogether, I substituted national leg- islation in place of the UN and ILO conventions for these three countries.25

Means, standard deviations, and zero-order correlation coefficients for the independent variables are presented in Appendix C. Size of the service sector is strongly correlated with women's tertiary enrollment (r = .777), log of GDP (r = .727), and export dependency (r = .606), and results should be interpreted accord- ingly. The substantive conclusions were similar, however, when the size of the service sector variable was excluded from the analyses.



Before proceeding with a discussion of the results, two issues should be addressed. First, the reader should bear in mind that this sample of 16 countries was driven by data availability and therefore results may not be generalizable to the population of developing countries. Nevertheless, the sample contains broad region- al representation except for South America. Comparing the average GDP26 in 1992 for the sample countries (8,737) with the average GDP for all developing countries (2,595) in that year suggests that the sample contains countries that, on average, are more industrialized than their counterparts. There is no reason, however, to expect the sample differs dramatically from other developing countries with respect to the patterns of segregation and explanations for segregation, but we cannot know for sure until data is available for more countries.

Second, due to the lack of time-varying detailed occupational data for most countries, this study incorporates data for only one time

25 TO determine whether the substitution of nation- al legislation for Korea, Hong Kong, and Taiwan biased the findings in any way, I ran identical mod- els without these three countries and the results were consistent.I prefer using the UN and ILO conventions as indicators rather than national legislation because they are consistent in scope and meaning. I also con- ducted analyses using national legislation, however, and the results were consistent with those presented here.

26 Data from the United Nations Human Development Report (1995).


Table 1. Indices of Anti-Discrimination Legislation and Protective Legislation for 16 Countries.

Anti-Discrimination Legislation   Protective Legislation
Equal UN   Night Underground Heavy
Country Opportunity' CEDAW2 Total Work1 Work1 Loads3 Total
Angola Yes Yes 2 Yes Yes No 2
Bahamas No No 0 No Yes No 1
Bahrain No No 0 Yes No No 1
China No No 0 No Yes Yes 2
Costa Rica Yes Yes 2 Yes No No 1
Cyprus Yes No 1 Yes Yes No 2
Egypt Yes No 1 Yes Yes No 2
Ghana Yes NO^ 1 Yes Yes No 2
Haiti Yes Yes 2 No Yes Yes 2
Hong Kong5 No No 0 Yes Yes Yes 3
India Yes No 1 Yes Yes Yes 3
Korea6 Yes No 1 Yes Yes Yes 3
Kuwait Yes No 1 Yes No Yes 2
Mauritus No No 0 No No No 0
Singapore No No 0 No Yes No 1
Taiwan7 No No 0 Yes Yes No8 2

Note: National legislation is substituted for UN and ILO conventions for Hong Kong and Taiwan. Korea did not accept the UN convention without reservations and national legislation is substituted for the ILO conventions for Korea.

Source: ILO (1992)
Source: Ladin (1994)
Source: ILO (1987)
Ghana did ratify the UN CEDAW in 1986, two years after the census data were collected.
According to the Women andYoung Persons (Industry) Regulations.
According to the Labor Standards Acts of 1982 and 1987 and the Employment Equality Act passed in 1987.
According to the Labor Standards Act of 1984 and the Labor Safety and Health Law of 1974.
A 199 1 revision of the 1974 Labor Safety and Health Law prohibits women from handling heavy loads;
however, this revision was passed one year after the census data were collected.

point. Consequently, it is impossible truly to and the Index of Dissimilarity (D), in ascend- test causal relationships or to observe how the ing order of fA.28 Overall levels of segregation covariates impact segregation trends. Despite vary dramatically across developing countries, these limitations, the richness of the data and the diverse sample of countries stand out as impres- sive features of the data, enabling a more to calculate the log-linear index (4% of the cells

detailed analysis of cross-national segregation were empty). I substituted a "1" in empty cells, but experimented with other substitutions as well, such

patterns than was possible with the highly aggre-

as advocated by Agresti (1990:250). As expect-

gated occupational categories that have char-

ed, the value ofA increased as the substitution value

acterized most segregation research thus far.

decreased, and vice versa. Since most data are sam- ple data and not population data, it seemed reason- able to replace empty cells with a value of 1,


representing the possibility that there could be one I begin by examining overall levels of segrega- person (of the opposite sex) employed in the partic- tion across countries. Table 2 displays levels of ular occupation who was not included in the sample.

Nevertheless, while the value of A shifted in accor-

segregation for the Index of Association (~)27

dance with the procedure for filling empty cells, the relative positions of countries with respect to their overall level of segregation was consistent.

27 Some occupations were completely segregated 28 The zero-order correlation for A and D is .573, and hence the issue of empty cells had to be addressed significant at the .05 level (two-tailed test).

one must bear in mind that they represent male MULTIVARIATE ANALYSES

or female over-representation in each occupa- tional category in comparison to the others. For example, in comparison to other occupational categories women are over-represented in the managerial occupations in four countries: Bahrain, Costa Rica, Egypt, and Kuwait. In these countries, the average percentage of women in the managerial categories is close to, or in some cases slightly exceeds, the percent- age of women in the labor force as a whole. This relative integration of women in these occupa- tions, coupled with women's almost complete exclusion from production occupations in these countries, makes their presence in professional and managerial occupations particularly pro- nounced. It is ironic that the extremely high lev- els of segregation in these countries, coupled by the selective recruitment of highly educated women into the labor market results in women's over-representation in professional and mana- gerial occupations relative to other occupation- al categories.

Figure 1 also reveals great variation in seg- regation patterns. For example, in most coun- tries, males are strongly over-represented in production occupations. However, in China, India, and Korea, the production category is close to being sex-integrated (and women are even slightly over-represented in production occupations in China). The demand for labor in these export-led economies has fueled women's integration in production occupations (Anker 1998).

The scale values provide a complementary piece of information when compared with the Index of Association, A. Countries with the highest overall levels of segregation (Bahrain, Costa Rica, Kuwait, Angola, and Haiti) are gen- erally those with the most extreme patterns of male or female over-representation. In a simi- lar fashion, countries with the lowest levels of segregation, such as India, China, Taiwan, and Hong Kong, tend to exhibit patterns of segre- gation across major categories that are much less pronounced. While there are indeed differences in the representation of men and women across these categories, differences are minimal in comparison to the much greater variation found in the more segregated nations.

I now turn to analyses that address how well existing explanations for cross-national varia- tion in segregation can account for these dif- ferences. Table 3 presents the results for the OLS regression analysis of the measures of the economic structure (economic development, export dependency, and size of the service sec- tor), women's human capital (women's per- centage of the labor force and women's share of tertiary education), and the three dimensions of state commitment to women's labor market equality (the presence of maternity leave, anti- discrimination legislation, and protective leg- islation) on levels of segregation, measured by A.32AS revealed in Model la, only two state policy variables are significant predictors of levels of segregation.

32 Results were similar when outliers were removed from the analysis.

Table 3. OLS Regression Results of Levels of

Segregation, measured by A, across 43

Occupational Categories for 16 Countries.

Model la Model lb


Economic Structure Log GDP

Export Dependency

Size of Service Sector

Human Capital Women's Share of Tertiary Education Women's % Labor Force State Policies Maternity Leave (1 = yes)

Protective Legislation

F Adjusted R2

Because of the small sample size, it is use- 11to limit the number of variables in the model. Therefore, I ran the same models utilizing back- ward selection.33 Model lb presents the results of the backward selection model with the high- est adjusted R2 value. The results of the back- ward selection model confirm the findings from the full model, indicating that two state policies significantly affect levels of segregation. However, the two policies have opposite effects on levels of segregation. The presence of mater- nity leave is associated with a decline in levels of segregation whereas the presence of anti- discrimination legislation has a positive effect on levels of segregation. These findings will be discussed in further detail as the analyses proceed.

Since covariates may not act in identical ways across all occupational categories, I have regressed the occupation-specific scale values (from the saturated model with 43 occupa- tions presented in Equation 3) for each major occupational category on the measures of eco- nomic structure, human capital, and state poli- cies. Because negative values indicate male over-representation and positive values indi- cate female over-representation, the effects of each covariate indicate their impact on female-representation in each occupational category. The reader should bear in mind that the effects on segregation depend upon whether the particular occupation was male or female-dominated initially. Increasing women's representation in female-dominated occupations will increase segregation; in con'trast, increasing women's representation in male-dominated occupations decreases seg- regation.

Table 4 presents the results from the occu- pation-specific analyses. Two models are pre- sented for each major occupational category. The first model for each occupational catego- ry (Model a) contains all explanatory variables. Because superfluous variables can have a con- siderable effect on analyses based on small sam- ple sizes, I have also presented results from backward selection models. The second model for each category (Model b) contains the coef- ficients for the variables in the backward selec-

33 Results were identical when stepwise selection was utilized.

tion model that yielded the highest adjusted R2 value.34

None of the three economic structure variables had a significant impact upon overall levels of segregation (Table 3), but export dependency decreases women's representation in profes- sional occupation^.^^ These results support pre- vious findings that while dependency had no effect on levels of segregation, it does decrease women's access to higher-status occupations (Clark 1991; Semyonov and Shenhav 1988; Semyonov and Jones 1999). In contrast, level of economic development and size of the service sector had no statistically significant effects on levels or patterns of segregation.

Women's tertiary education increases women's access to professional occupations, consistent with other studies (Clark 1991 ; Ramirez and Weiss 1979; Weiss et al. 1976). Professional occupations often require educational creden- tials, and as women increasingly attend college and obtain college degrees, it appears that they are able to use their credentials for entrance into professional occupations. Increases in women's tertiary education also draw women away from production occupations, presumably for occupations that require higher levels of skill or education.

34 I also conducted stepwise regression analyses and found that the results were similar, with two exceptions: (1) in the stepwise analyses, only women's labor force participation was included in the final model for professional occupations and (2) in the stepwise analyses, only maternity leave was includ- ed in the final model for production occupations. Nevertheless, the predictive power of the backward selection models was better (indicated by the adjust- ed R2values) than the stepwise selection models.

35 I ran two additional models that substituted investment dependency and multinational corporate penetration for export dependency. Multinational corporate penetration had a statistically significant (p < .05) negative effect on women's representation in professional occupations, consistent with export dependency. However, investment dependency was not statistically significant.

Table 4. OLS Regression Results for Covariates' Effects on Female Representation in Occupations (Maior Occupation Scale Values) for 16 Countries.
Professional Model la Model lb Managerial Model 2a Model 2b Clerical Model 3a Model 3b Sales Model 4a Model 4b Service Model 5a Model 5b Production Model 6a Model 6b
Economic Structure Log GDP  
Export Dependency Size of Service Sector  
Human Capital Women's Share of Tertiary Education % Women Labor Force  
State Policies Maternity Leave (1 = yes)  
Antidiscrimination Legislation  
Protective Legislation  
F Adjusted R2 Occupations (n) within each major category *p<.l; **p<.05; ***p<.Ol  

Of the two human capital variables, women's labor force participation had the most occu- pation-specific effects. Increases in women's labor force participation are associated with (a) increases in women's representation in sales, service, and production occupations, and (b) decreases in women's representation in pro- fessional and managerial occupations. These results are consistent with culling or selective recruitment arguments: when few women are in the labor force, they are more likely to be better educated than the general population and to have access to the higher-prestige pro- fessional and managerial occupation^^^ (Charles and Grusky, forthcoming; Lewin- Epstein and Semyonov 1992; Semyonov and Jones 1999). As more and more women enter the paid labor force, women's representation in these prestigious occupations declines, as women of all levels of skill, education, and commitment enter the labor force. As women's labor force participation increases, women are generally distributed (either by supply or demand forces) into sales, service, and pro- duction occupations.

In interpreting aggregate effects of women's human capital, one must also consider gender differences in human capital development before women enter the labor market. Because women in developing countries often receive less ele- mentary schooling than their male peers and have lower literacy (Nash and Fernandez-Kelly 1983; United Nations 1995; Ward 1984), many women are disadvantaged with respect to human capital before they even seek employment. Pre- market differences in human capital develop- ment are particularly relevant to understanding the effects of women's labor market participa- tion. If women in general have lower human cap- ital before entering the labor market, they are unable to compete with men for skilled or semi- skilled jobs. Consequently, increasing the per- centage of women in the labor market will result in the entrance of low-skilled women and the exacerbation of segregation in low-skilled occupations.

36 Bear in mind that the relative integration of women in these occupations is driven in part by women's almost complete exclusion from production occupations.

The availability of maternity leave decreases women's representation in sales occupations, but increases their representation in manageri- al, clerical, and production occupations. As expected, the presence of maternity leave poli- cies increases women's representation in man- agerial occupations, presumably because it facilitates women's continuous labor force par- ticipation, which is often a prerequisite for these more prestigious occupations. More con- tinuous labor force participation also increas- es women's representation in production occupations, typically held disproportionately by younger unmarried women in developing countries. These results clarify the earlier find- ing that maternity leave decreases levels of segregation overall (Table 3); although increas- ing the representation of women in the typically female-dominated clerical occupations has a segregative effect, increasing female represen- tation in the male-dominated production and managerial occupations has an integrative effect. (Sales occupations are not so clearly identified as male-dominated or female-dominated, so the net effects of decreasing women's representation in sales occupations are not as

easily generalized.)

I argued previously that anti-discrimination legislation is likely to have countervailing effects. On the one hand, women are more likely to press for entrance into professional and managerial (and to a lesser extent, cleri- cal) occupations because they are more pres- tigious and because they are more likely to be found in the formal sector where legislation is likely to be most easily enforced. On the other hand, it also facilitates married women's employment in sales and the female-dominat- ed service occupations. Results generally sup- port this second effect: anti-discrimination legislation increases women's representation in sales and service occupations, presumably because married women seek employment in jobs that are most easily accommodated to family responsibilities.

Contrary to expectations, protective legisla- tion had no effect upon women's representa- tion in production occupations. But as expected, protective legislation increases women's repre- sentation in professional occupations. If pro- tective legislation reflects less gender-egalitarian ideals generally, this finding may be a result of the culling of women into the labor market.


This paper makes several contributions to the sex segregation literature. First, it utilizes high- ly detailed occupational data that are better suited to analyzing segregation than aggregat- ed categories that mask the extent of segrega- tion (Anker 1998). Second it moves beyond summary indices of levels of segregation to analyze how men and women are distributed across the occupational structure. Third, it demonstrates the relevance of state policies in accounting for levels and patterns of segrega- tion in developing countries, while controlling for aspects of the economic structure and women's human capital.

State policies were the strongest predictors of cross-national variation in levels of segrega- tion. Maternity leave legislation is associated with lower levels of segregation in developing countries. Examination of occupation-specific effects reveals that maternity leave generally decreases levels of segregation throughout the occupational structure by increasing female representation in the male-dominated manage- rial and production occupations and decreasing female representation in the female-dominated clerical occupations. Maternity leave may be one of the most effective policies for lowering occu- pational sex segregation in developing countries and increasing women's presence throughout the occupational structure.

The most surprising finding was that legis- lation aimed to facilitate women's integration into the occupational structure was associated with an increase in overall levels of segregation. Examination of occupation-specific effects revealed the processes generating its overall segregative effect: equal access legislation increased women's representation in the female- dominated service occupations and decreased women's access to the male-dominated pro- duction occupations, contributing to higher seg- regation overall. In addition to the preceding explanation, the counterintuitive finding that anti-discrimination legislation increases segre- gation overall may be influenced by the fol- lowing factors.

First, countries with the highest score on the

anti-discrimination index (Table 1) also have the

highest overall levels of segregation (Table 2).

Without knowing more about why these coun- tries adopted these types of legislation, it is dif- ficult to interpret this finding. We don't know if the adoption of such legislation reflects a real national commitment to women's greater eco- nomic participation, whether countries with particularly high levels of gender inequality face greater international pressures to adopt more "women-friendly" legislation, or whether the adoption of legislation is driven by entirely different forces, resulting in a spurious rela- tionship with occupational sex segregation.

Another explanation for this finding is that the cross-sectional data may obscure the effect that anti-discrimination legislation has had in bringing about changes in segregation over time. If we had data on trends, we may find that anti- discrimination legislation helped to bring about agradual decline in segregation levels. If so, the positive relationship we see here might be a result of countries with high levels of segrega- tion enacting anti-discrimination legislation. However, because the data are cross-sectional, we can't determine whether this is indeed the case.

The lack of any statistically significant effect of anti-discrimination legislation upon women's representation in professional and managerial occupations is also particularly surprising given arguments that gender egalitarian pressures are most strongly felt in elite occupations where public pressure is strongest and where access is more closely tied to more meritocratic educa- tional qualifications.

But the most likely explanation is that any gains made by women in the professional and managerial occupations are overshadowed by the numbers of married women drawn into the already female-dominated service and sales occupations. It may be the case that as countries continue to develop, equal access legislation will have a larger impact upon women's ability to enter the more prestigious and higher-paying occupations found in the professional and man- agerial category.

One must also bear in mind that anti-dis-

crimination legislation is much less effective if

it is not enforced or if there are other institutional

barriers to occupational integration, such as

strong cultural restrictions that discourage

women from interacting with men in public,

and laws that require separate dining areas for

men and women in the workplace, making it

more difficult for employers to hire both male and female workers (Anker and Hein 1986). As stated by Whitehouse (1992:83), "the prospects for women in paid work are depend- ent on much more than the enactment of equal opportunity legislation. . . .This does not require the conclusion that legislation is unimportant, but rather that supporting institutions and poli- cies need to be seen as crucial to labor market outcomes."

Economic structure (as evidenced by GDP, export dependency, and size of the service sec- tor) was not a significant predictor of cross- national variation in levels of segregation. Export dependency, however, decreases women's representation in professional occu- pations, a finding that is consistent with earli- er studies (Clark 1991; Semyonov and Shenhav 1988). Yet export dependency had no other sta- tistically significant effects. This finding was somewhat surprising given arguments that dependency may affect women's participation in production and service occupations. It is pos- sible, however, that the effects of dependency are indirect. For instance, a growing body of liter- ature suggests that dependency impacts women's labor force participation and fertility (Clark et a1.1991; London 1988; Pampel and Tanaka 1986; Shen and Williamson 1999; Ward 1984). Further analyses should be done to adjudicate the direct and indirect effects of dependency upon levels and patterns of segregation.

Despite previous studies that suggest that growth of the service sector tends to increase levels of occupational sex segregation in devel- oped countries, no statistically significant effect was found in these analyses. Accordingly, some may conclude that size of the service sector is unrelated to levels and patterns of segregation in developing countries; however, I believe this conclusion to be premature. Given the high cor- relation between size of the service sector and both women's share of tertiary education and log of GDP (see Appendix C), any effects of the size of the service sector are likely to be captured by these other variables. The question of whether the service sector affects patterns of segregation in developing countries in ways similar to devel- oped countries is still an open question and worthy of further research.

The two human capital variables were poor predictors of cross-national variation in levels of segregation; however, women's labor force participation was one of the strongest predictors of female representation in five out of six of the major occupational categories. Increases in women's labor force participation decreased women's representation in professional and managerial occupations but increased their rep- resentation in sales, service, and production occupations. These results suggest that there is indeed a culling effect: in countries with high- er female labor force participation, women of all levels of education, skills, and commitment are found in the labor force. As a result, the pro- portion of women in the labor force who were selectively recruited because of their higher- than-average levels of education or commit- ment declines as the composition of the female labor force more closely resembles the general population. Women's tertiary education also increased women's representation in profes- sional occupations, consistent with expecta- tions that women's participation in higher education provides them with the necessary prerequisites for these occupations.

In addition to testing the strength of differ- ent explanations for cross-national variation in levels and patterns of segregation, the analyses demonstrate the importance of examining patterns of segregation in which one can observe gender differences in access to more and less desired positions in the occupational structure (Blackburn et al. 2000; Charles and Grusky 1995, forthcoming; Hakim 1996; Semyonov and Jones 1999). Because the processes shap- ing segregation are so complex, it is unlikely that any one factor can single-handedly change seg- regation. Nor should we expect any effects to be felt uniformly across all occupations--or even all women. Further analyses should address how individual-level characteristics such as age, marital status, and education interact with gen- der and state policies to shape patterns of segregation.

The results are also relevant to questions regarding the extent of cross-national variation in segregation worldwide. Recent cross-nation- a1 research has documented that most devel- oped countries exhibit similar patterns of segregation at the aggregated occupational level, with women generally concentrated in the "mid- dle" of the occupational structure in clerical, sales, and service occupations, whereas men are more likely to dominate both the top occu- pations (professional and managerial) and the bottom occupations (production) (Charles and Gmsky 1995; Nermo 2000; Roos 1985). In this regard, patterns of segregation are somewhat similar for developed and developing countries, although several of the developing countries in this analysis are characterized by female over- representation in professional and managerial occupations at the aggregate level, which is likely to due the selective recruitment of high- ly educated, career-oriented women into the labor market.

Despite these general similarities, the pat- terns of segregation in developing countries appear more diverse than their developed coun- terparts, both with respect to their gender-typ- ing and amount of variation. In addition, the absolute levels of segregation appear higher than in comparative studies of developed coun-

Given the tremendous diversity in levels and pat- terns of segregation in developing countries, the descriptive results suggest that developing countries do not follow a singular pattern. Future cross-national studies should focus on the mech- anisms generating cross-national similarities and differences among both developed and developing countries.

Some have argued that segregation is not always bad for women (Anker 1998;Blackburn et al. 2000). In some respects, this may be true since being in the middle of the occupational hierarchy is arguably better than being at the bot- tom. It is important to bear in mind, however,

tries that use detailed occupational categorie~.~~ that since the "separate but equal" ideal is

37 Charles and Grusky (forthcoming) found that the highest level of segregation in their sample of ten developed countries was 6.28 (in the United Kingdom), whereas in this study, Angola had a level of segregation of 61.6 and Haiti had a level of seg- regation of 69.6. While their occupational classifi- cation scheme is different from the one utilized here, theirs incorporates even greater detail; if anything, greater occupational detail reveals even more segre- gation (Anker 1998).

unlikely to materialize, we must remain skep- tical that gender equality can exist side-by-side with high levels of occupational sex segregation.

Mariko Lin Chang is Associate Professor of Sociology and of Social Studies at Harvard University. Her interests include social stratifica- tion, gendel; and economic sociology. She is cur- rently writing a book on the gender wealth gap and is engaged in research that addresses how race, sex, and social class impact financial decision-making and wealth-building opportunities.

Appendix A. List of Countries, the Year that Occupation by Sex Data were Collected, and the Data Source. Countrv Year Source

Angola Bahamas Bahrain China Costa Rica Cyprus

Egypt Ghana Haiti Hong Kong India Korea Kuwait Mauritius Singapore Taiwan Labor force survey*

Population census

Population census*

Population census

Labor force survey*

Establishment survey*

Population census*

Population census

Income and expenditure survey*

Population census*

Population census

Population census

Population census*

Population census*

Population census

Povulation census

* Indicates that data were part of the "Segregat" database, distributed by the ILO.

(Continued on next page)

Appendix B. Occupational Classification Scheme (Based on ISCO-68), with Average Minor-Category Scale Values in Parentheses.

A. Professional and technical workers

1.Physical scientists, Life scientists, and related technicians (-.27)

Architects, engineers and related technicians (-.84)

Aircraft and ships' officers (-3.12)

Jurists (-.12)

Teachers (2.02)

  1. Medical, dental, veterinary and related workers (2.19)
  2. Economists, accountants, statisticians, mathematicians, systems analysts and related technicians (.64)
  3. Sculptors, painters, photographers, composers, performing artists, athletes, sportsmen and related workers
  4. Workers in religion, authors, journalists, and professional and technical workers n.e.c. (1.17)

B. Administrative and managerial workers

10.Legislative officials and government administrators (-.53)

11. Managers (-.56)

C. Clerical and related workers

12. Clerical supervisors and government executive officials (.40)
13.Stenographers, typists and card- and tape-punching machine operators (3.62)
14.Computing machine operators, bookkeepers, cashiers and related workers (1 39)
15.Transport conductors and transport and communications supervisors (-1.03)
16.Mail distribution clerks, telephone and telegraph operators (.50)
17.Clerical and related workers n.e.c. (1.29)

D. Sales workers

18.Managers and working proprietors (wholesale and retail trade), sales supervisors, buyers (.49) 19.Technical salesmen, commercial travelers, manufacturers' agents, and insurance, real estate, securities and business services salesmen and auctioneers (.23)

20. Salesmen, shop assistants, other sales workers n.e.c. (1.40)

E. Service workers

2 1. Cooks, waiters, bartenders, maids and related housekeeping service workers, building caretakers, chanvork- ers, cleaners, and related workers (1.45)

Launderers, dry-cleaners and pressers (2.13)

Protective service workers (-2.18)

  1. Maids and related housekeeping service workers n.e.c. (3.68)
  2. Hairdressers, barbers, beauticians and related workers (2.22)
  3. Managers and working proprietors (catering and lodging services), housekeeping and related service supervi- sors, Service workers n.e.c. (.73)

F Production and related Workers, transport equipment operators and laborers

  1. Miners, quarrymen, well drillers and related workers (-1.05)
  2. Metal process&s, wood preparation workers and paper makers, chemical processers and related workers (-1.22)
  3. Spinners, weavers, knitters, dyers, tanners, fellmongers, pelt dressers and related workers (1.56)
  4. Food and beverage processers, tobacco preparers and tobacco product makers (.72)

3 1. Tailors, Dressmakers, Sewers, Upholsterers and related workers (2.11)

Shoemakers and leather goods makers (-.39)

Glass formers, potters and related workers (.20)

Painters (-2.89)

Transport equipment operators (-3.89)

  1. Cabinetmakers and related woodworkers, stone cutters and carvers, jewelry and precious metal workers (-1.81)
  2. Blacksmiths, toolmakers, machine-tool operators, plumbers, welders, sheet metal and structural metal prepar- ers and erectors (-2.77)
  3. Machinery fitters, machine assemblers and precision instrument makers (ex. electrical) (-3.03)
  4. Electrical fitters and related electrical and electronics workers, broadcasting station and sound equipment operators and cinema projectionists (-1.09)
  5. Rubber, plastics, paper, and paperboard product makers and printers and related workers (.06)
  6. Bricklayers, carpenters and other construction workers (-3.47)
  7. Stationary engine and related equipment operators, material-handling and related equipment operators, dock- ers, and freight handlers (-31)
  8. Production supervisors, Production and related workers n.e.c. and laborers n.e.c. (-.07)


Agresti, Alan. 1990. Categorical DataAnalysis. New York: John Wiley and Sons.

Anker, Richard. 1998. Gender and Jobs: Sex Segregation of Occupations in the World. Geneva, Switzerland International Labour Office.

Anker, Richard and Catherine Hein, eds.1986. Sex Inequalities in Urban Employment in the Third World. London, England: MacMillan Press.

Blackburn, Robert M., Jennifer Jarman, and Bardley Brooks. 2000. "The Puzzle of Gender Segregation and Inequality: A Cross-National Analysis." European Sociological Review 16: 1 19-35.

Bornschier, Volker and Christopher Chase-Dunn. 1985. Transnational Corporations and Underdevelopment. New York: Praeger.

Boserup, Esther. 1970. Womanj. Role in Economic Development. New York: St. Martin's Press.

Brinton, Mary C., ed. 2001. WomenS Working Lives in East Asia. Stanford, CA: Stanford University Press.

Chang, Mariko Lin. 2000. "The Evolution of Sex Segregation Regimes." American Journal of Sociology 105: 1658-701.

Charles, Maria. 1992. "Cross-National Variation in Occupational Sex Segregation." American Sociological Review 57:483-502.

Charles, Maria and David B. Grusky. Forthcoming.

Occupational Ghettos: The Worldwide Segregation of Women and Men. Stanford, CA: Stanford University Press.

. 1995. "Models for Describing the Underlying Structure of Sex Segregation." American Journal of Sociology 100:93 1-71.

Clark, Roger. 1991. "Contrasting Perspectives on Women's Access to Prestigious Occupations: A Cross-National Investigation." Social Science Quarterly 72:20-32.

Clark, Roger, Thomas W. Ramsbey, and Emily Stier Adler. 1991. "Culture, Gender, and Labor Force Participation: A Cross-National Study." Gender and Society 5:47-66.

Goldin, Claudia D. 1990. Understanding the Gender Gap: An Economic History ofAmerican Women.

New York: Oxford University Press. Hakim, Catherine. 1996. Kty Issues in Women j. Work. London, England: Athlone.

Inkeles, Alex and David H. Smith. 1974. Becoming Modern: Individual Change in Six Developing Countries. Cambridge, MA: Harvard University Press.

ILO (International Labour Office). 1999. Key Indicators of the Labor Market. (CD-Rom) Geneva, Switzerland: ILO.

. 1992. "Lists of Ratifications by Convention and by Country." Geneva, Switzerland: ILO.

. 1987. "Women Workers: Protection or Equality?" Conditions of Work Digest. Vol. 4. Geneva, Switzerland: ILO.

Jacobs, Jerry A. and Suet T. Lim. 1992. "Trends in Occupational and Industrial Sex Segregation in 56 Countries, 1960-1980." Work and Occupations 19:450-86.

Kerr, Clark, John Dunlop, Frederick Harbison, and Charles Myers. [I9601 1994. "Industrialism and Industrial Man." Pp. 659-69 in Social Stratification: Class, Race and Gender in Sociological Perspective, edited by David Grusky. Boulder, CO: Westview Press.

Ladin, Sharon. 1994. A Report on the Thirteenth Session of the Committee on the Elimination of Discrimination against Women. University of Minnesota, Humphrey Institute of Public Affairs, International Women's Rights Action Watch.

Lee, Yean-Ju and Shuichi Hirata. 2001. "Women, Work, and Marriage in Three East Asian Labor Markets: The Cases of Taiwan, Japan, and South Korea." Pp. 96-124 in Womenk Working Lives in East Asia, edited by Mary C. Brinton. Stanford, CA: Stanford University Press.

Lewin-Epstein, Noah and Moshe Semyonov. 1992. "Modernization and Subordination: Arab Women in the Israeli Labour-Force." European Sociological Review 8:39-5 1. Lim, Linda Y. C. 1990. "Women's Work in Export Factories: The Politics of a Cause." Chapter 7 in Persistent Inequalities: Women and World Development, edited by Irene Tinker. New York: Oxford University Press. London, Bruce. 1988. "Dependence, Distorted Development, and Fertility Trends in Noncore Nations: A Structural Analysis of Cross-National Data." American Sociological Review 53:606618. Marshall, Susan E. 1985. "Development, Dependence, and Gender Inequality in the Third World." International Studies Quarterly 29:21740. McCall, Leslie. 2001. Complex Inequality: Gendel; Race, and Class in the New Economy. New York, NY: Routledge.

Nash, June and Maria Patricia Fernandez-Kelly, eds. 1983. Women, Men and the International Division oflabor: Albany, NY. SUNY Press.

Nermo, Magnus. 2000. "Models of Cross-National Variation in Occupational Sex Segregation." European Societies 2:295-333.

Nuss, Shirley and Lorraine Majka. 1983. "The Economic Integration of Women: A Cross-National Investigation." Work and Occupations 10:2948. Pampel, Fred C. and Kazuko Tanaka. 1986. "Economic Development and Female Labor Force Participation: A Reconsideration." Social Forces 641599-619. Parsons, Talcott. 1971. The System of Modern Societies. Englewood Cliffs, NJ: Prentice Hall. Ramirez, Francisco 0. and Jane A. Weiss. 1979. "The Political Incorporation of Women." Pp. 238-49 in National Development and the World System: Educational, Economic and Political

Change 1950-1970, edited by John Meyer and Michael Hannan. Chicago, IL: University of Chicago Press.

Republic of China. 199 1. Statistical Yearbook ofthe Republic of China. Prepared by the Directorate- General of Budget, Accounting, and Statistics.

Roos, Patricia A. 1985. Gender and Work: A Comparative Analysis of Industrial Societies. Albany, NY: SUNY Press.

Safa, Helen I. 1983. "Women, Production, and Reproduction in Industrial Capitalism: A Comparison of Brazilian and U.S. Factory Workers." Pp. 95-1 16 in Women, Men and the International Division oflabor, edited by June Nash and Maria Patricia Femandez-Kelly. Albany, NY: SUNY Press. Semyonov, Moshe. 1980. "The Social Context of Women's Labor Force Participation: A Comparative Analysis." American Journal of Sociology 86:534-50. Semyonov, Moshe and Frank L. Jones. 1999. "Dimensions of Gender Occupational Differentiation in Segregation and Inequality: A Cross-National Analysis." Social Indicators Research 46:225-247. Semyonov, Moshe and Yehouda Shenhav. 1988. "Investment Dependence, Economic Development, and Female Employment Opportunities in Less Developed Countries." Social Science Quarterly 69:961-978. Shen, Ce and John B. Williamson. 1999. "Maternal Mortality, Women's Status, and Economic Dependency in Less Developed Countries: A Cross-National Analysis." Social Science and Medicine 49: 197-214.

Tinker, Irene, ed. 1990. Persistent Inequalities: Women and World Development. New York: Oxford University Press.

Treiman, Donald J. 1977. Occupational Prestige in Comparative Perspective. New York: Academic Press.

United Nations. 2001. World Investment Report 2001: Promoting Linkages. United Nations Conference on Trade and Development. . 1995. Human Development Report. New York: United Nations Development Programme. . 1996. UNESCO Statistical Yearbook.

U.S. Department of Health and Human Services. 1990 and 1999. Social Security Programs Throughout the World. Washington, D.C.: U.S. Department of Health and Human Services.

Ward, Kathryn B., ed. 1990. Women Workers and Global Restructuring. Ithaca, NY: ILR Press. . 1984. Women in the World-System: Its Impact on Status and Fertility. New York: Praeger.

Weiss, Jane A,, Francisco 0. Ramirez, and Terry Tracy. 1976. "Female Participation in the Occupational Structure: A Comparative Institutional Analysis." Social Problems 23:593-608.

Whitehouse,Gillian. 1992. "LegislationandLabour Performance. Berkeley, CA: University of Market Gender Inequality: An Analysis of OECD

California Press. . 1990. "Common Problems, Divergent

Countries." Work, Employment and Society

Policies: An 18-Nation Study of Family Policy."


Public Afairs Report 3 1 :1-3. Wilensk~, Harold L. 2002. Rich Democracies: World Bank. 2001. World Development Indicators, Political Economy, Public Policy, and CD-ROM.

  • Recommend Us