（1.西南财经大学经济数学学院；2.西南财经大学统计学院）

【摘要】考虑累积误差和有限样本问题，本文构造了一种固定效应模型截面相关性的新型检验统计量 。研究表明，累积误差情形下，Schott(2004)的检验统计量存在与 和 相关的渐进偏差，引起检验统计量水平的扭曲和势的下降。据此，本文对存在的渐进偏差进行了修正，并采用回归的办法得到具体的修正项的系数。本文提出的固定效应模型截面相关性检验统计量 ，在有限样本条件下，具有渐近正态分布以及良好的统计检验水平和势，并能用于异方差情形下的截面相关性检验。

中图分类号  F224.0    文献标识码  A

A New Method for Testing Cross-sectional Dependence in the FE Panel Data Model

Abstract：Considering accumulative error and finite sample problem, we propose a new testing statistic for cross-sectional dependence in the FE panel data model. The results show that using Schott’s statistics for testing cross-sectional dependence in a FE model, accumulative error will appear and produce asymptotic bias of the testing statistic in finite sample. The bias is found to be constants related to and . In view of this, this article suggests a bias-corrected Schott’s test for cross-sectional dependence in FE model, and the bias-corrected coefficients are obtained by the linear regression. In the finite sample, the bias-corrected Schott’s statistics (named ), converges to standard normal distribution asymptotically and performs well with respect to the size and the power, even under the heteroscedasticity.

Key words：Cross-sectional Dependence；Accumulative Error；Fixed Effects；Finite Sample