统计研究 ›› 2019, Vol. 36 ›› Issue (2): 101-111.doi: 10.19343/j.cnki.11-1302/c.2019.02.009

• • 上一篇    下一篇

基于带跳时变系数模型的PPI与CPI相关性研究

苍玉权等   

  • 出版日期:2019-02-25 发布日期:2019-03-07

A Study on Correlativity between PPI and CPI Based on Time-varying Coefficient Models with Jumps

Cang Yuquan et al.   

  • Online:2019-02-25 Published:2019-03-07

摘要: 2008年以来,我国PPI与CPI走势出现了多次背离与分化,从整体上看,两者相关性很弱。但从动态视角来看,由于相关关系可能会因时而变,整体相关性有可能被关系本身的方向和强弱变化所削弱甚至掩盖。为准确反映两者相关性的动态变化,本文放宽时变系数函数的光滑性假设,提出了带跳时变系数模型,并给出一种非参数三步估计方法:首先,估计系数函数中跳点的位置和个数;然后,基于估计的跳点和Bootstrap方法选择的窗宽给出系数函数的最终估计;最后,利用蒙特卡洛模拟评价本文提出的非参数估计和窗宽选择方法的有限样本性质。通过对2008年1月至2017年12月我国PPI和CPI月度同比数据的实证分析,我们发现该模型能较好地刻画PPI与CPI相关性的时变和带跳特征,进而也验证了该模型的应用价值。

关键词: 非参数估计, 跳点, 时变系数模型, PPI与CPI

Abstract: Since 2008, the trend of PPI and CPI in China has deviated and split up many times. On the whole, both correlate weakly. However, from a dynamic perspective, because the correlation may change over time, the overall correlation may be weakened or even concealed by the veer and extent of correlations itself. In order to reflect the dynamic changes of the PPI and CPI correlations, this paper propose a time-varying coe?cient models with jumps and offers a nonparametric three-step estimation procedure in a hypothesis of smoothness in the eased time-varying coefficient function. Firstly, the position and number of jumps are estimated in the coe?cient function; then the final estimation of the coe?cient function is given based on the estimated jumps and the bandwidth chosen by the bootstrap method; and finally, some Monte Carlo simulations are used to evaluate the ?nite sample performance of the proposed nonparametric estimation and bootstrap method. By empirically studying the PPI and CPI data in China from January 2008 to December 2017, it is found that the model can well depict the time-varying and band jumping characteristics of PPI and CPI, and furthermore justify the value in applying this model.

Key words: Nonparametric Estimation, Jumps, Time-varying Coe?cient Model, PPI and CPI