短面板约束下固定效应空间动态面板模型的估计与模拟

Estimation and Simulation of Fixed-effects Spatial Dynamic Panel Data Model with Short Panels

空间动态面板数据(SDPD)模型中被解释变量初值极易带有内生性,采用一般拟极大似然(QML)方法容易造成参数估计偏误,特别是当样本结构为n大T小的时候。鉴于此,本文在一般QML基础上,通过重塑误差项的方差-协方差矩阵,修正拟似然函数表达式,得到修正QML,进而估计短面板下含空间、时间、误差三类关联项的固定效应SDPD模型,基于数值模拟和实例应用检验一般QML与修正QML的估计效果。数值模拟结果表明:修正QML比一般QML更精确、更稳健,均方误差修正率随样本短面板结构的增大而增大。实例应用不仅重新评估环境规制与技术创新之间的空间效应,回归结果也再次证实从数值模拟中得出的结论。

The initial value of the explanatory variable in the spatial dynamic panels data(SDPD)model is very prone to be endogenous.The general quasi maximum likelihood(QML)method is easy to cause parameter estimation bias,when sample structure is large n and small T.In view of this,based on the general QML method,the expression of the quasi likelihood function is modified to obtain the modified QML by reconstructing the variance-covariance matrix of the error term.It is applied to the estimation of fixed-effects SDPD model with short panels that incorporates three types of associations of space,time and error,and then the general QML and modified QML estimation results are tested based on numerical simulation and example application.The numerical simulation shows that the modified QML is more accurate and more robust than the general QML,and the mean square error correction rate increases with the increase of the sample structure.The example application not only re-evaluated the spatial effects between environmental regulation and technological innovation,but the regression results again confirmed the conclusions drawn from numerical simulations.