We provide a semiparametric copula approach for estimating a classical sample selection model. We impose that the joint distribution function of unobservables can be characterized by a specifc copula, but the marginal distribution functions are estimated semiparametrically. In contrast to existing semiparametric estimators for sample selection models, our approach provides a measure of dependence between unobservables in main and selection equation which can be used to analyze the composition of, say, the female workforce. We apply our estimation procedure to a female labor supply data set and show that those women with the best skills participate in the labor market; moreover, we find evidence for the existence of an ability threshold which involves that women with high ability are to some extent advantaged and, therefore, have also obtained the best skills.
Sample selection model semiparametric estimation copula approach composition of the female workforce female labor force participation