MODELS FOR DESCRIBING THE UNDERLYING STRUCTURE OF SEX SEGREGATION

This article introduces a structural approach to analyzing sex segrega tion data that rests on margin-free measures of the underlying associa tion in sex-by-occupation arrays. The starting point for the analyses is a log-multiplicative model that is formally consistent with the con ventional practice of summarizing cross-national variability in a sing le parameter pertaining to the overall strength of sex segregation. Un der this baseline specification, the segregation regime is forced to t ake on the same basic shape in each country, with the only form of per missible variability being a uniform compression or expansion of the p eaks and valleys characterizing the shared segregation profile. Althou gh the latter model does not account for the cross-national variabilit y in our illustrative data, it can be readily generalized in ways that both improve the fit and yield new insights into the structure and so urces of sex segregation. These elaborated models can be used to exami ne the hierarchical structure of segregation, to identify the dominant ''segregation profiles'' in industrial countries, and to parse out th e net residue of segregation at multiple levels of analysis.