HAC估计的性质与修正及其在伪回归中的应用研究  

A Study on the Properties and Correction of HAC Method and Its Application in the Spurious Regression

作　　者：刘汉中[1] 

机构地区：[1]湖南商学院经济与贸易学院

出　　处：《数量经济技术经济研究》2015年第11期148-160,F0003,共14页Journal of Quantitative & Technological Economics

基　　金：国家社科基金项目"弱平稳过程之间的伪回归研究"(11BJY012);湖南省高校创新平台开放基金"收入;城镇化和产业结构对环境污染的效应研究--基于稳健的HAC方法"(12K116);"湖南省高校科技创新团队支持计划"的资助

摘　　要：本文通过频谱分析揭示了预白化HAC和参数化VARHAC比非参数HAC具有明显优势,并指出传统预白化HAC法由于使用了存在有限样本偏差的OLS法来估计自回归参数,导致其存在有限样本偏差,由此构造的t统计量具有过度拒绝原假设倾向。为了减少预白化HAC的偏差,将OLS估计量的线性修正(LBC)和非线性修正(NBC)嵌入到预白化HAC中,研究表明该法能大大减少长期方差的估计偏差,并通过蒙特卡洛模拟证实对预白化HAC的修正估计能有效减少平稳过程之间的伪回归概率,从而提高了回归模型统计推断的可靠性。The paper reveals the prewhitening HAC and parametric VARHAC having an obvious advantage over the nonparametric HAC method through the spectrum analysis, and the traditional prewhitening HAC method having finite sample bias due to the OLS estima- tion existing finite sample bias in the process of estimating an AR model. Then, t statistic relying on the prewhitening HAC estimator has a tendency of over rejecting the null hypothesis. In order to reduce the bias of prewhitening HAC method, the linear correction method （LBC） and nonlinear correction method （NBC） is plugged into the prewhitening HAC method. The study shows that the methods can reduce the bias of long-run variance estimation, and through Monte Carlo simulations, the paper confirms that the modified estimation of prewhitening HAC can effectively decrease the probability of spurious regression between stationary processes, accordingly increases the reliability of statistical inference in the regression model.

关 键 词：HAC法线性偏差修正 非线性偏差修正 伪回归概率 

分 类 号：F224.0[经济管理—国民经济]