BIAS CORRECTION ESTIMATOR FOR A DYNAMIC PANEL DATA MODEL WITH FIXED EFFECTS USING AN ITERATED BOOTSTRAP

Abstract

A bias correction estimator (BCE) for a dynamic panel data model with fixed effects is given, based on the alternating iterative maximum likelihood estimator (AIMLE). The new estimator is asymptotically unbiased and consistent. Monte Carlo studies are conducted to evaluate the finite sample properties of the MLE, AIMLE and BCE. It is shown that the BCE based on AIMLE appears to dominate the AIMLE approach both in terms of the median bias (Bias) and median absolute error (MAE) of the estimators.