require a method for extrapolating the earnings function to work-experience levels not observed in the data. We find significant upward bias in the ordinary least squares (OLS) estimates, pointing to the importance of accounting for endogeneity of schooling.