一致有效的预测回归方法的构建及其在中国市场上的应用
Consistently Valid Tests for Predictive Regression and Empirical Study in Chinese Market
杨光艺 1赵琬迪2
作者信息
- 1. 南京审计大学金融学院,江苏南京 211815
- 2. 首都经济贸易大学统计学院,北京 100070
- 折叠
摘要
预测回归(Predictive Regression)是用于检验金融市场可预测性的重要计量方法,但现有检验方法,如传统的T检验以及Campbell和Yogo(2006)[1]提出的方法,均不能在|ρ|≤ 1范围内取得一致有效的检验效果.本文提出了 一种新的检验方法,该方法通过采用Bonferroni的方法,将在|ρ|≤ 1范围内一致有效的Hansen(1999)[2]格点自助法与文[1]提出的更具效用的Q检验结合.数值模拟显示该检验方法犯第一类错误的概率在|ρ|≤ 1范围内均靠近其5％的理论值,并且效用检验同样显示该检验方法优于目前学术界常用的文[1].采用本文提出的预测回归检验方法,本文发现在中国股票市场,距离52周最高价的接近程度、技术指标、PPI以及CPI等变量具有预测能力.
Abstract
Predictive regression is an important econometric model used to test the predictability of financial markets,but the existing testing methods,such as the traditional T test and the method proposed by Campbell and Yogo(2006)[1],are not consistently valid in the region of|ρ|≤ 1.In this paper,we propose a new test method,which combine Hansen(1999)[2]'s grid bootstrap method which is consistently valid over the region of|ρ|≤ 1 and Campbell and Yogo(2006)[1]'s Q test which is more powerful by Bonferroni method.Results of simulation analysis show that the probability of the Type I error of this test is close to the theoretical value 5％in the range of|ρ|≤ 1.Moreover,the power test also shows that this test outperforms the commonly used method from Campbell and Yogo(2006)[1].By using the method proposed in this paper,we find that nearness to 52-week high,technical indicator,PPI and CPI have the predictability in Chinese stock market.
关键词
一致有效/预测回归/可预测性Key words
consistently valid/predictive regression/predictability引用本文复制引用
基金项目
江苏省教育厅高校基础科学(自然科学)研究面上项目(22KJD630003)
江苏省高校哲学社会科学研究一般项目(2022SJYB0368)
国家社会科学基金青年项目(23CTJ021)
出版年
2024
NETL
NSTL
万方数据