Sample Selection Models with Common Endogeneity in the Selection and Outcome: Revisiting the Family Gap

We develop a fully parametric estimation procedure for unbalanced panel data models with unobserved effects that allow for a common binary endogenous variable in both the selection equation and the outcome equation. We test the finite sample properties of the estimator using Monte Carlo simulations and find that our estimator performs better in the presence of high endogeneity and high sample selection compared to the estimators that ignore either or both of these issues. In addition, our estimator is also robust to distributional misspecification. We apply our econometric methodology to estimate the effect of fertility decisions on wages for white women using the National Longitudinal Survey of Youth 1979 (NLSY79) data from 1982 to 2006