Nonparametric Estimation in a One-Way Error Component Model: A Monte Carlo Analysis

Abstract

This paper considers the problem of nonparametric kernel estimation in panel data models. We examine the finite sample performance of several estimators for estimating a one-way error component model. Monte Carlo experiments show that the pooled estimator that ignores the dependence structure in the model performs well in each trial, but it is not the most efficient estimator since it is generally outperformed in the mean squared sense by both the two-step estimator and the nonparametric feasible generalized least squares estimator. Although the asymptotic bias and variance of most of the estimators converge at the same rate, the two-step estimator, which has smaller asymptotic variance and can have smaller asymptotic bias, generally outperforms most of the other estimators incorporating dependence when the technology is nonlinear and the variance of the error component is large relative to the variance of the random disturbance. Finally, we use an empirical example regarding the public capital productivity puzzle to showcase the estimators in a real data setting.