An algorithm to reduce the occupational space in gender segregation studies

Abstract

This paper presents an algorithm based on the bootstrap to select an admissible aggregation level, that is, the minimum number of occupational categories which yield a gender segregation value which is not significantly smaller than that obtained from the large number of occupational categories usually available in any data set. The approach is illustrated using Labor Force Survey data for Spain for the comparison of gender segregation in 1977 and 1992, as well as 1994 and 2000. To measure gender segregation, an additively decomposable segregation index based in the entropy concept is used. Despite a substantial simplification in the size of the occupation's space, the decrease in the segregation index is very small and not significant, regardless of the year. Consequently, intertemporal changes in gender segregation can be studied using a greatly reduced classification of occupations that permits an easier interpretation of results.

Entropy-Based Segregation Indices

“…In empirical contexts where it is advisable to use decomposable segregation indices, such as the entropy‐based ones, a key question arises: Which index should be used? This is an important issue in a scenario in which, except for Frankel and Volij (forthcoming) in school segregation and Fuchs (1975), Mora and Ruiz‐Castillo (2003, 2004), and Herranz, Mora, and Ruiz‐Castillo (2005) in occupational segregation, the authors who have used an entropy‐based index have preferred the H index 2 . Taking as reference the school segregation problem in the multiracial case, this paper establishes the practical and conceptual advantages of the M index in multilevel studies of segregation and its trends for the following reasons.…”

Computing decomposable multigroup indices of segregation

“…The lesson here is that decomposable indices are useful for analyzing survey data that may give rise to small-cell problems. (See Herranz, Mora, and Ruiz-Castillo [2005] for an application of this approach to gender segregation in occupations and industries with the M index.) Estimation of random segregation.…”

“…By contrast, survey-based measurements of segregation are finite-sample estimates and, therefore, biased and subject to sample variability (Deutsch, Flückiger, and Silber 1994; Herranz, Mora, and Ruiz-Castillo 2005). Bootstrap methods can help estimate bias and basic bootstrap confidence intervals for segregation indices (Ransom 2000; Allen et al 2015).…”