Bias Corrected Instrumental Variables Estimation for Dynamic Panel Models with Fixed E¤ects

This paper proposes a new instrumental variables estimator for a dynamic panel model with .xed e¤ects with good bias and mean squared error properties even when identi.cation of the model becomes weak near the unit circle. We adopt a weak instrument asymptotic approximation to study the behavior of various estimators near the unit circle. We show that an estimator based on long di¤erencing the model is much less biased than conventional implementations of the GMM estimator for the dynamic panel model. We also show that under the weak instrument approximation such conventional estimators are dominated in terms of mean squared error by an estimator with far less moment conditions. The long di¤erence estimator mimics the infeasible optimal procedure through its reliance on a small set of moment conditions.