Two-Stage Estimation of Unobserved Effects Panel Stochastic Frontier Models with Endogenous Regressors

This paper proposes an alternative estimation procedure for estimating the unobserved effects panel stochastic frontier models with endogenous regressors. Specifically, two-stage estimation method is used, where in the first stage, the frontier parameters are estimated based on GMM procedure, which is robust to the distributional assumptions of the composed error; whilst in the second stage, the remaining parameters are estimated by using MLE technique. Monte Carlo simulations indicate that the proposed estimation approach behaves well in finite samples