Semiparametric Varying Coefficient Models with Endogenous Covariates

The semiparametric varying coefficient model is used in a wide range of applications. However, the traditional specification does not account for endogenous covariates, which restricts its application. In this paper we consider the estimation of semiparametric varying coefficient models when the functional coefficients may contain (continuous) endogenous covariates thereby extending the reach of this flexible and powerful model. We provide theoretical underpinnings, assess finite-sample performance via simulations, and showcase its practical appeal via an empirical application that examines the degree to which returns to education factor into the documented growing disparity between more and less educated workers.