A Full Information Maximum Likelihood Approach to Estimating the Sample Selection Model with Endogenous Covariates

In this paper we establish a full information maximum likelihood approach to estimating the sample selection model with endogenous covariates. We also provide a test for exogeneity which indicates whether endogeneity is in fact a matter or not. In contrast to other methods proposed in the literature which deal with sample selection and endogeneity, our approach is computationally simple and provides exact asymptotic standard errors derived from common maximum likelihood theory. A Monte Carlo study and an empirical example are presented which indicate that not accounting for endogeneity in sample selection models may lead to severely biased parameter estimates.