Statistical inferences for testing marginal rank and (generalized) lorenz dominances

This paper provides distribution-free inferences for testing marginal rank dominance and Lorenz, and generalized Lorenz dominances. Marginal dominances refer to ordinary dominance relationships holding between an income distribution and its dependent after-event distribution. Using the elegant Bahadur representation, I establish the asymptotic normal distributions of sample marginal changes and derive the variance-covariance structures. I also show that the inference procedures can be modified and applied to more general cases where samples are (partially) dependent. The approaches are illustrated by re-evaluating the marginal impacts of working wives on the U.S. family income distribution using 1990 census data.