统计研究

• 论文 • 上一篇    下一篇

分位数单位根检验的拓展及其应用研究

司登奎等   

  • 出版日期:2017-05-15 发布日期:2017-05-16

The Extension of Quantile Unit Root Test with Its Application

Si Dengkuiet al.   

  • Online:2017-05-15 Published:2017-05-16

摘要: 在不同的分位下对时间序列数据是否含有结构性变化进行甄别,对于准确识别数据的动态变化及其分布特征具有重要的意义。文章首次在Koenker和Xiao(2004)研究的基础上提出傅立叶分位数单位根检验模型,并以此捕捉时间序列中所存在的结构突变点,进而刻画数据在不同分位下的动态变化特征。文章通过构建傅立叶QKS统计量并采用蒙特卡罗方法对傅立叶分位数模型的临界值、样本容量(Size)和检验“势”(Power)进行模拟,我们发现含有傅立叶级数的分位数单位根检验在刻画“尖峰厚尾”特征数据的非线性偏离动态调节特征具有更高的检验“势”。最后,我们利用拓展后的模型对中国通货膨胀的持久性(1990M01-2016M05)和失业的回滞效应(1978-2015)进行再检验,结果发现中国通货膨胀具有平稳的特征,而失业率却包含单位根过程。我们的研究为分位数单位根检验的拓展及其应用提供了一定的启示。

关键词: 分位数, 单位根, 结构突变, 傅立叶级数, 蒙特卡罗模拟

Abstract: Capturing structural changes in the time series under different quantiles is of great significance to accurately identify dynamic behavior and distributed characters of data. This paper for the first time incorporates the Fourier function into current Quantile unit root test suggested by Koenker and Xiao (2004) to detect smooth unknown breaks, and thus to describe the dynamic features of data in different quantiles. Through constructing Fourier QKS statistic and simulating the critical value by using Monte Carlo method, this paper tests different sample size and power. The findings show that the Fourier Quantile unit root test has higher power in depicting the features of nonlinear deviation and dynamic adjustment for heavy-tailed distributed data. Finally, we use the Fourier Quantile unit root test to investigate the persistence of inflation and unemployment of China. The results show that the inflation rate is stationary process, but the unemployment rate contains a unit root. Our study provides some implications for extension and application of Quantile unit root test.

Key words: Quantile, Unit Root, Structural Breaks, Fourier Function, Monte Carlo Simulation