Pairwise Difference Estimation of Nonlinear Models
James Powell, UC Berkeley
Bo Honoré, Princeton University

Abstract

This paper uses insights from the literature on estimation of nonlinear panel data models to construct estimators of a number of semiparametric models with a partially linear index, including the partially linear logit model, the partially linear censored regression model, and the censored regression model with selection.. We develop the relevant asymptotic theory for these estimators and we apply the theory to derive the asymptotic distribution of the estimator for the partially linear logit model. We evaluate the finite sample behavior of this estimator using a Monte Carlo study.