In this paper, we revisit the Mulligan and Rubinstein (2008: Selection, investment and women's relative wages over time. The Quarterly Journal of Economics, 123(3):1061-1110) analysis about the composition of the female workforce in the United States. Using a Heckman selection model, these authors found that the selection of women into the female workforce changed from negative to positive over time. However, the authors assumed the exogeneity of covariates, which is sometimes appropriate but not for a variable like education. We revisit the issue of the Mulligan and Rubinstein (2008) paper by developing and applying a Heckman selection model which also controls for the potential endogeneity of education. Applying this estimator to U.S. Census and American Community Survey data, we find that selection has become more positive over time (like in Mulligan and Rubinstein), but that selection has never been negative. We rather find an interesting puzzle concerning the correlation pattern of the unobservables in our model which requires further investigation.
Sample selection model endogenous covariates gender wage gap composition of the female workforce female labor force participation