A1730
Title: Consistent generalized method of moments estimation of spatial autoregressive models based on non-Gaussian distributions
Authors: Fei Jin - School of Economics, Fudan University (China) [presenting]
Abstract: A generalized method of moments (GMM) estimation approach for the spatial autoregressive (SAR) model is proposed using generalized linear and quadratic moments based on non-Gaussian distributions. Both homoskedastic and heteroskedastic innovations are investigated. The proposed GMM estimator of parameters of interest is consistent and asymptotically normal under regularity conditions. It is computationally simple and can yield significant efficiency improvements over popular estimators of the SAR model. In the homoskedastic case, conditions for the existence of best moments that generate a GMM estimator with minimum asymptotic variance are provided. Simulation experiments demonstrate that the considered GMM estimator performs well in finite samples.
