This paper tests for racial discrimination by white and nonwhite male workers using the states as the unit of observation. It is shown theoretically that white (nonwhite) employee discrimination creates a component of white (nonwhite) earnings inequality within skill levels which is higher, the higher the fraction of nonwhites in the labor force. The "job-rationing" hypothesis predicts a negative (positive) partial effect of percentage nonwhite on white (nonwhite) earnings inequality. Other discrimination hypotheses offer no prediction. Empirically, the effect of percentage nonwhite on earnings inequality is positive and significant for whites, but not significant for whites.