Occupational Ghettos: The Worldwide Segregation of Women and Men (review)

Maria Charles and David B. Grusky (with substantial contributions from Kim Weeden, Mariko Chang, Joon Han and Jesper Sørensen) have provided the community of employment segregation scholars with a powerful review and extension of their sustained comparative and methodological work. This methodologically careful work has evolved through analyses of sex by aggregate occupation contingency tables, often comparing countries or time trends within countries and starting with the baseline question, "Is there a worldwide sex segregation regime?"

On a theoretical level the book proposes that in most countries there are two basic segregation processes at work. The first is a vertical mechanism, through which men tend to dominate the best jobs. The second is a horizontal mechanism, in which cultural notions of gender essentialism tend to match women to non-manual service work. They find in analyses of multiple countries, some over time, that both dimensions tend to be in play and that males tend to get the best jobs within the manual and non-manual sectors, but women tend to be overrepresented in the typically higher status non-manual sector. Thus the horizontal gender-essentialist mechanism tends to promote segregation but undermine inequality. Vertical gender segregation tends to be stronger in the manual sector, reflecting the stronger influence of egalitarian norms and politics on the non-manual sector. The mix of vertical and horizontal mechanisms varies across countries, and time and there is substantial national variability in detailed segregation patterns leading to the conclusion that there is no worldwide sex segregation regime.

One of the great puzzles in the sex segregation field has been that gender equalitarian countries often have high levels of sex segregation (e.g., Sweden). Charles and Grusky show that this is because Sweden has particularly high levels of horizontal segregation tendencies. Importantly, this insight demonstrates that gender essentialism and gender egalitarianism can operate simultaneously. Gender essentialism promotes a separate but equal segregation regime, while gender egalitarianism challenges the legitimacy of vertical segregation mechanisms such as sex discrimination.

Both methodologically and theoretically this line of research is linked to the classical analysis of mobility tables in sociology. A key strength of this approach to national sex segregation patterns is that it allows researchers to explicitly model the contributions of cross-country or cross-temporal occupational and sex distributions to observed segregation. This is an advantage relative to conventional summary indices of segregation (e.g., the Index of Dissimilarity) all of which are margin dependent in one way or another. That is, estimated segregation (depending on the index deployed) is partially a function of the national occupational distribution and/or the sex composition of labor supply. In classical mobility table analyses the key question was for a long time, "Net of changes in the marginal distribution (i.e., structural mobility), has intergenerational mobility increased with modernization?" Thus margin-free models were always implied by the theoretical question. It is not so clear that margin-free models are demanded by sex segregation research, perhaps because the theoretical questions have not been specific enough to provide clear guides. My sense is that this issue is always empirically important and that both the development of employment structures (the occupation margin) and labor supply (the sex margin) should be conceptualized and modeled.

Charles and Grusky argue against the use of margin dependent indices, proposing a new index based on odds ratios and so is margin free. They further propose that summary indices of all types hide as much as they reveal. Both proposals seem justified by their analyses. I am particularly convinced that disaggregated analysis that uses odds ratios to evaluate the degree of segregation is a useful approach. The use of odds ratios to study segregation is not limited to aggregate contingency table analyses. Thus, while most researchers in this field prefer workplace data when they can get it, we would do well to learn from the methodological contributions of this work.

One of the remarkable...