修正的KSS检验及其对中国通货膨胀率的应用

Modified KSS test with application to China's inflation rate

Kapetanios等^([1])提出在指数平滑转换自回归(ESTAR)模型框架下进行单位根检验.他们的检验是基于误差项为独立同分布的强假设下得到的,该假设在现实中很难成立.当误差项为平稳弱相依时,该检验统计量的极限分布包含冗余参数.通过构造修正的KSS检验,得到了不包含冗余参数的检验统计量.蒙特卡罗模拟结果表明该修正的统计量大大减少了序列相关性带来的水平扭曲(size distortion),且该检验统计量对于非线性平稳过程的检验功效高于PP检验.将该检验用于中国的通货膨胀率,发现它存在着一个单位根,是非平稳过程.

Kapetanios et al.[1] propose a unit root test in the ESTAR framework. The KSS test relies on the i.i.d, assumption for the error term, which is very restrictive in application. When the error term is stationary and weakly dependent, the limiting distribution of the KSS test statistic is not nuisance parameter free. The proposed modified KSS test is nuisance parameter free, and Monte Carlo simulation results suggest the modified statistic reduce the size distortion due to serial correlation in error term, and the power of the test statistic is higher than PP test. Applying the test on China＇s inflation rate suggests it＇s a unit root nonstationary process.