Measuring Trends in Segregation

Abstract

Social scientists have tended to rely solely on the index of dissimilarity as a measure of segregation, without questioning what precise aspect of segregation is measured by this index. A recent paper by Lieberson and Carter has argued the merits of a less commonly used class of measures of pa interaction probabilities to tap an aspect of segregation not reflected by the index of dissimilarity. Standardizing these interaction probabilities by the proportion of minority/majority group members in the population yields yet a third class of measures, of which the best known is the correlation ratio or eta². We argue that these three kinds of measures reflect conceptually distinct aspects of racial residential segregation. We analyze the relationships among the three and demonstrate with both simulated and real data that the results of empirical studies will depend on the measure chosen.