Distribution-Free Statistical Inference for Inequality Dominance with Crossing Lorenz Curves

Distribution-free techniques of statistical inference are developed for the cumulative coefficients of variation of an income distribution, thus allowing one to test for inequality dominance when Lorenz curves cross. The full covariance structure of the cumulative sample means and variances is worked out. As an illustration, the procedures are applied to the 1984 and 1990 earnings distributions of male paid workers in the United States, and it is found that the 1990 distribution was significantly less unequal than the 1984 distribution.