固定效应面板数据模型中偏误更正的截面相关性检验研究

A Bias-corrected Cross-sectional Correlation Test in a Fixed Effects Panel Data Model

基于Schott(2004)检验统计量,本文在固定效应面板数据模型中提出了一个新的偏误更正的截面相关性检验。在大n和大T的框架下,本文首先研究了基于估计残差项构建的Schott统计量与标准正态分布之间偏误的极限性质;然后提出了一个偏误更正的新检验统计量并推导了其联合极限分布;最后采用蒙特卡洛模拟实验考察了本文提出的检验统计量的有限样本性质。新检验统计量联合收敛于标准正态分布,蒙特卡洛模拟结果显示该检验统计量有较为准确的水平尺度和较好的功效性质。本文完善和拓展了在固定效应面板数据模型中偏误更正的Schott检验统计量的极限分布理论,为相关的应用研究提供了一个新的非正态分布稳健的检验统计量。

Based on the test statistics of Schott(2004),this paper proposes a bias-corrected cross-sectional correlation test in a fixed effects panel data model.Under big n and big T framework,this paper first studies the asymptotic bias between the Schott statistics using the estimated within residuals and standard normal distribution;then it proposes a bias-corrected new test and derives its limiting distribution jointly;finally,it conducts Monte Carlo simulations to examine the finite properties of the proposed test.The new test converges to standard normal distribution jointly,and Monte Carlo simulation results show that the test has a correct size and relatively better power property.This paper improves and extends the asymptotic theory of the bias-corrected Schott test in the fixed effects panel data model,and provides a new non-normality robust test for application.