SOURCES IN BIAS IN WOMENS WAGE EQUATIONS - RESULTS USING SIBLING DATA

We use data on sisters to jointly address heterogeneity bias and endog eneity bias in estimates of wage equations for women. This analysis yi elds evidence of biases in OLS estimates of wage equations for white a nd black women, some of which are detected only when the two sources o f bias are addressed simultaneously. For both white and black women th ere is evidence of upward bias in the estimated returns to schooling. Bias-corrected estimates of the effect of marriage on wages, for white women, suggest a positive marriage premium. We also use the sibling d ata to identify our models, and test a number of other commonly-used i dentifying assumptions as overidentifying restrictions.