AMH Copula ML Estimation for the Sample Selection Model

Abstract

In this paper, we propose a copula ML estimation method for the sample selection model using the Ali-Mikhail-Haq (AMH) copula. The proposed AMH copula ML estimation is compared with the well-known bivariate ML estimation and Heckman's two-step estimation. Monte Carlo experiments are conducted to compare their performance in terms of the mean squared error (MSE) depending on the following 2 conditions: (i) whether the imposed distributional assumption is correct, and (ii) whether some regressors of the participation and outcome equation are correlated. The results of the experiments show that the estimation results for the proposed method can be better than those of the two well-known methods, particularly when the imposed distributional assumption is incorrect and some regressors of the two equations are correlated. Hence, the proposed method can be a practically useful alternative for the sample selection model.