面板数据模型中随机效应存在性检验的理论研究及其实证分析

Theory Study and Empirical Analysis on Testing for the Existence of Random Effects in Panel Data Models

在反映异质性的个体效应和时间效应的设定上,经常存在人为的主观性和随意性,因此容易导致错误指定事件的发生。本文提出了一个稳健的方法分别检验面板数据模型中随机个体效应和随机时间效应的存在性。具体而言,通过对残差进行正交化变换消去可能存在的时间效应,并建立人工自回归模型,然后基于该模型自回归系数的最小二乘估计构造检验统计量检验个体效应。构造的检验是单边的,零假设下渐近服从标准正态分布。在检验时间效应时,可类似得到统计量及其渐近性质。功效研究表明这些检验敏感性较强,能检测到以参数速度(最快的速度)收敛到零假设的备择假设。通过模拟试验研究了检验统计量的小样本性质,并进行了实际数据分析。

Due to the excellent properties of showing heteroscedasticity,Panel Data models have been greatly used in economics.However,there are many subjectivity and randomness on the specification of individual and time effects,which easily leads to misspecification and wrong conclusions.This paper proposes a robust method to respectively test for the existence of individual effects and time effects in panel data models.Specifically,in testing for individual effects,we can eliminate the potential time effects by an orthogonal transformation of residuals and consider an artificial autoregressive model of the transformated residuals,and then construct the test statistics based on the OLS parameter estimates of the artificial autoregressive model.Similarly,the tests can be constructed for the existence of time effects.The resultant tests are one-sided,and are asymptotically normally distributed under the null.Power study shows that our tests can detect local alternatives distinct at the parametric rate from the null.The small sample properties of the tests are investigated by Monte Carlo simulation experiments and real data is analyzed for illustration.