This paper argues that sex discrimination is an inefficient practice. We model sex discrimination as the complete exclusion of females from the labor market or as the exclusion of females from managerial positions. The former implies a reduction in GDP per capita; the latter distorts the allocation of talent and lowers economic growth. Both imply lower female-to-male schooling ratios. Our model predicts a convex relationship between nondiscrimination and growth. Although discrimination is difficult to measure, it will be reflected in schooling differentials. We present evidence based on cross-country regressions that is consistent with a convex relationship between schooling differentials and growth.