Nearly Unbiased Estimation in Dynamic Panel Data Models with Exogenous Variables

This paper introduces a new estimator for the fixed effects dynamic panel data model with exogenous variables. This estimator does not share some of the drawbacks of recently developed IV and GMM estimators and has a good performance even in small samples. The nearly unbiased estimator is derived as a bias correction of the within estimator (least squares dummy variable estimator). The estimator is applied to a model of unemployment dynamics at the U.S. state level for the 1991-2000 period.