GMM and 2SLS estimation of panel data models with spatially lagg dependent variables and spatially correlated error components

摘要

In this article, the GMM based estimation of a typical family of spatial panel models with spatially lagged dependent variables and error components that are both spatially and time-wise correlated is addressed. We derive the best GMM (BGMM) estimator within certain class of optimal GMM estimators. We also discuss the asymptotic efficiency of BGMM estimator relative to the panel analogue of generalized spatial two stage least squares (GS2SLS) estimators and maximum likelihood (ML) estimators. We show that by including GS2SLS estimators as a special case, the BGMM estimator is generally more efficient than GS2SLS estimator and able to be as efficient as ML estimator under normality.