Efficient Semiparametric Estimation of Nonseparable Models

Abstract

This paper considers semiparametric estimation of quantile regression models in the presence of endogeneity. Imposing index restrictions in the triangular model framework, an efficient semiparametric estimator for nonseparable models is introduced. Endogeneity is addressed through a control variable approach. The estimator has four main characteristics: 1) It is based on conditional quantile restrictions; 2) it allows for dimension reduction; 3) it addresses endogeneity of one or more covariates; 4) it is efficient. Large-sample properties of the proposed estimator are derived and an empirical application on the effect of property rights on land profits is discussed.