中文摘要：

研究固定效应空间误差分量模型（Spatial Error Components,SEC）的广义矩估计（GMM）方法。为克服去均值转换方法导致的误差项方差奇异性,以及极大似然法在SEC模型中的运算困难,本文采用正交转换法去除固定效应,并基于正交转换模型,提出基于广义距估计的可行广义最小二乘法（GMM-FGLS）,证明了估计量的一致性。并通过Monte Carlo模拟实验,研究GMM-FGLS估计量的有限样本性质。结果表明,GMM-FGLS估计量在误差项正态分布下接近ML估计量,在非正态分布下远优于ML估计量,是实证研究中的理想估计量。

英文摘要：

In this paper,we study GMM estimation method of fixed effect spatial error components（SEC） model.In order to overcome singularity of error term variance by removing mean value methods,and the operation difficulties in SEC model by maximum likelihood method,we use orthogonal transform method to remove fixed effect,and propose feasible generalized least square method based on generalized moments of methods（GMM-FGLS）.Then we examine the consistency of estimators,and study finite sample performances using Monte Carlo simulation experiments.The result shows that,GMM-FGLS estimator is close to ML estimator when error term is normal distributed,while is far superior to ML estimator when error term is non-normal distributed,which is an ideal estimator in empirical research.