An evaluation of an entropy based index of segregation

Abstract

This paper reviews the properties suggested in the methodological literature on the measurement of gender segregation by occupation. It is found that an index of segregation based on the entropy concept satisfies twelve basic axioms previously proposed in the single-dimensional case. This index can be expressed as the sum of a between-group and a within-group term in the two-dimensional case. In pair-wise comparisons, it can be meaningfully decomposed into three terms, one of which is independent of both the gender composition of the population and the population's distribution across occupations. Finally, it can be motivated as two different loglikelihood tests in nonparametric econometric models. Other existing measures of segregation either fail to satisfy one or more of the basic axioms, do not admit a between/within decomposition, have not been motivated from a statistical approach, or are based on more restricted econometric models.