This article examines the sampling distributions of popular indexes of segregation: the dissimilarity index and the Gini index. Although applications of segregation indexes are common in the social sciences, researchers have usually failed to recognize their stochastic nature. This study addresses that failure by deriving the exact sampling distribution of two popular indexes of segregation and developing a convenient asymptotic test for hypotheses about changes in segregation. Monte Carlo simulations show that the proposed test has appropriate significance levels. An example applies the methods developed in this article to test whether segregation of men and women across college majors decreased during the 1980s. The analysis finds that gender segregation has decreased, and that the decrease is statistically significant.
