Bias-Corrected Instrumental Variable Estimation in Linear Dynamic Panel Data Models

This paper introduces a new estimation method for linear dynamic panel data models with endogenous explanatory variables. The proposed approach adapts the estimation methods based on bias corrections of the least-squares dummy-variable or maximum-likelihood estimators to a common situation, where some explanatory variables are endogenous. The estimation approach relies on combining several simple instrumental variable estimators and correcting their biases using the analytically-derived bias expressions. We prove the consistency and asymptotic normality of the proposed bias-corrected instrumental-variable estimator under weak assumptions. The finite sample performance is compared with existing estimators by means of Monte Carlo simulations, which demonstrate good performance with the simplest choice of instrumental variables.