面板数据非线性阈值协整的检验方法

Panel Data Nonlinear Threshold Cointegration Test

非线性阈值协整是线性协整的后续发展。本文使用两机制TR模型对Westerlund和Edgerton(2005)的面板数据协整向量结构突变模型进行扩展,提出截距项具有阈值效应、截距项和斜率系数都具有阈值效应的面板数据非线性阈值协整模型。在此基础上,本文进而分别构造Zc、Ztc、Zr、Ztr统计量检验阈值协整,并对上述统计量的极限分布进行了数学推导,发现它们都收敛于随机泛函。仿真实验结果表明,有限样本下上述检验统计量具有较小的水平扭曲和较高的检验势。

Nonlinear threshold cointegration is the follow-up development of linear cointegration and it is also the central part of modern econometrics. This pa- per expands Westerlund, Edgerton＇s （2005） cointegration model of cointegration vector structural break of panel data, proposes panel data threshold cointegration model which contain the threshold effect of intercept, the threshold effect of inter- cept and slope coefficient. This paper constructsZФ、Zt^c、ZФ^R、Zt^rstatistics to test panel data threshold cointegration. We find the limit distributions of those statistics are converging to stochastic functional, and the simulation showes that those test statistics have small distortion size and high power.