Statistical inference for testing inequality indices with dependent samples

This paper develops asymptotically distribution-free inference for testing inequality indices with dependent samples. It considers the interpolated Gini coefficient and the generalized entropy class, which includes several commonly used inequality indices. We first establish inference tests for changes in inequality indices with completely dependent samples (i.e., matched pairs) and then generalize the inference procedures to cases with partially dependent samples. The effects of sample dependency on standard errors of inequality changes are examined through simulation studies as well as through applications to the CPS and PSID data.