基于HAC估计视角的格兰杰伪因果关系检验

研究发现相互独立的弱平稳过程之间会产生伪因果关系, 经传统HAC法修正的Wald统计量甚至存在更高的伪因果关系概率. 文章认为数据过程的长期方差是发生伪因果关系的深层次原因, 通过改进传统HAC法的截断参数, 能获得格兰杰因果关系检验统计量(Wald)的不依赖于冗余参数的极限分布. 针对设定各种弱平稳过程并利用模拟技术, 研究发现新的Wald^*统计量大大减少了发生伪因果关系的概率, 并且对于数据过程持久性和样本容量是稳健的, 但是存在一定的检验水平扭曲.

Granger spurious causality test based on the perspective of HAC estimation

The study finds that spurious Granger causality occurs between independent weak stationary processes, and the Wald statistic modified by the traditional HAC method has higher probability of spurious causality than the primitive Wald statistic. At the same time, the paper shows that estimation of the long-run variance is the underlying reason of occurring spurious causality, the Granger causality test statistic constructed based on modified truncation parameter of traditional HAC method has a limiting distribution that does not depend on nuisance parameters. Through setting various weak stationary processes and using Monte-Carlo simulation technology, the paper finds that the new Wald^* statistic can greatly decrease the probability of occurring spurious causality, and it is robust to the persistence of data process and sample size, but has some test size distortions.