中文摘要：

近年来运用空间计量经济模型进行实证分析的文献都普遍采用空间自回归（SAR）形式的设定,对参数的估计也多采用极大似然（MLE）的方法。在经典多元线性回归模型中,仅有被解释变量的测量误差并不会影响系数估计的一致性。本文证明对于SAR模型,即使仅当被解释变量存在测量误差时,且无论该测量误差是否与模型本身的扰动项相关,普遍采用的MLE都将是不一致的。为此,Hausman型的设定检验被推广到SAR模型中用以判别是否存在被解释变量的测量误差。当零假设被拒绝时,我们说明由Kelejian ＆ Prucha（1998）,Lee（2003）提出的二阶段最小二乘法仍然可以得到参数的一致估计。Monte Carlo模拟的结果与我们的理论预期一致。最后我们用一个估计地方环境支出外溢效应的实例说明如何运用本文所提的方法来检验应用空间自回归模型时可能存在的测量误差。

英文摘要：

Recently,spatial autoregressive models are widespread applicated in many empirical studies and most of which use ML method to estimate the spatial model. In this paper,we show that MLE is generally inconsistent even with measurement error only existing in the dependent variables,regardless of whether the measurement error is independent of the disturbance term or not. Hausman-type specification test is then generalized to the spatial context and we suggest using the 2SLS estimator proposed by KelejianPrucha （1998） and Lee （2003） when the null hypothesis is rejected. Monte Carlo simulation results are consistent with our theoretical analysis and an example on spillover effect of local environmental spending is used to illustrate the proposed procedures and methods.