结构突变时间序列单位根的“伪检验”

Unit Root Spurious Test for Time Series with Structural Change

利用蒙特卡罗分析方法,本文对含一个结构变化点的经济变量单位根检验的有效性进行了探讨。分析的结果表明,当经济变量的数据生成过程存在一个结构性突变时,不考虑这种变化而进行常规的单位根检验只有在特定条件下才不会“失效”:只有当突变前后两期的样本数相差极大,或者选取的样本期总数很小时,单位根检验才不会“失效”。并且,随着结构变化程度的增大,不考虑结构变化而进行常规单位根检验得出“伪检验”的可能性也会增大。

Using Monte Carlo Methods, this paper examines the effectiveness of unit root test for time series with one structural change. The result shows that if there exists one structural change for the data generation process, the normal unit root test is effective under certain conditions, when the sample before and after the turning point varies greatly, or the total sample is relatively small, the normal unit root test holds. Furthermore, with the structural change increasing, the chance of reaching a spurious result with normal unit root test is also enlarged.