Testing for omitted variables in partially linear regression models

Abstract

This study conducts Monte Carlo simulation exercises to determine the finite sample properties of the Fan-Li test [1996] when the interest focuses on uncovering the impact of omitted variables on the semiparametric partially linear model (PLM). It essentially subjects the said test to various empirical situations such as the case of variable omissions in the linear or nonlinear components of the model or both. The study finds that the Fan-Li test achieves considerable power in rejecting the invalid null hypothesis. However, actual sizes still deviate from their respective nominal sizes.

Classification-JEL: C15, C14