2010), “Bias reduction in dynamic panel data models by common recursive mean adjustment,” Oxford Bulletin of Economics & Statistics

LSDV estimation of the dominant root in dynamic panel regression is vulnerable to downward bias. This paper studies recursive mean adjustment (RMA) as a bias reduction strategy. We develop the RMA estimators for general AR(p) process under both cross sectional independence and dependence. We study its asymptotic properties as N;T!1 jointly and \u85nd that the pro-posed asymptotically normal estimator exhibits nearly negligible bias when log2 T (N=T) ! where is a non-zero constant. The proposed method is an e ¢ cient and e¤ective bias reduction strategy and is straightforward to implement. Our simulation experiments suggest that the RMA estimator performs quite well in terms of bias, variance and MSE reduction both when error terms are cross sectionally independent and correlated. It dominates comparable estimators particularly when T is small and/or the underlying process is highly persistent