面板数据的空间Hausman检验

在空间相关性存在的情形下,经典Hausman检验不再有效。以面板数据空间误差分量(Spatial ErrorComponents,SEC)模型为例,提出面板数据的空间Hausman检验,并通过Monte Carlo模拟实验比较空间Hausman检验与经典Hausman检验的有限样本性质。研究结果表明,空间Hausman检验具有较小的水平扭曲和良好的检验功效,空间权重矩阵的选取将影响检验的有限样本性质,增大样本量使得检验更为有效;经典Hausman检验虽然检验功效较大,但存在较大的水平扭曲,不适用于空间相关性存在的个体效应判定。

Spatial Hausman Test of Panel Data

Classic Hausman test won’t work effectively when spatial correlation exists.We propose Spatial Hausman test based on panel data spatial error components model,and through Monte Carlo simulation experiments,we research the finite sample performance of Spatial Hausman test and traditional Hausman test.The results indicate that,Spatial Hausman tests have smaller size distortion and excellent power performance,and the finite sample performance will be affected by spatial weight matrix,while larger sample will make the tests more effective.What’s more,though traditional Hausman tests have ideal power performance,they exist greater size distortions,which make them unsuitable for individual effect select.