Double kernel nonparametric estimation in semiparametric econometric models

In this paper we estimate the coefficients of a generated regressor in the context of a partially linear semiparametric regression model. The generated regressor is part of the linear part of the model and the estimator is obtained by double kernel estimation. It is established that the double kernel estimator is rootn-consistent and asymptotically normal. Monte Carlo results suggest that it has satisfactory small samples properties. The usefulness of the proposed method is illustrated in an application to a model of wage determination.