文档介绍：数理经济与计量经济学
协整向量误设对相关检验的影响
叶光张晓峒
(南开大学经济学院国际经济研究所,天津,300071)
【内容摘要】常用于检验既定协整关系的统计量有两种:和,但由于真实数据生成过程未知,估计模型中可能存在一定程度的协整向量误设,从而使统计量的分布特征受到影响。论文首先探讨检验的隐含系数约束,即短期弹性等于先验长期弹性;然后分析零假设下两种统计量的分布特征,以及先验设定对信号噪声比,进而对分布特征的影响;最后在局部备择假设下,给出两种统计量的渐近分布,并表明向量误设会降低协整检验的势,其程度与设定误差正相关。一定程度上,KED(1992)的很多结论可以视为本文分析结果的特例。
关键词:协整,向量误设,检验势
The Effect of Vector Misspecification on Cointegration Tests
【Abstract】When the cointegration vector is prespecified, people usually use and as test statistics. But because real DGP is unknown, it is inevitable to encounter the problem of cointegration vector misspecification in the estimated model, which will influence the characteristic of statistics distribution. Firstly, this paper discusses the coefficient restriction imposed by test that equals , that is, the short-run elasticity equals the prespecified long-run elasticity; then analyzes the distribution characteristics of these two statistics under the null hypothesis, and the effect of on “signal-to-noise” ratio and then the distribution of . In the last, this paper gives the asymptotic distribution of these two statistics under the local alternative, and proves that vector misspecification would result in the reduction of the power of cointegration tests that is positively correlated to the misspecification error. To some extends, many inferences in KED(1992)can be viewed as special cases of our conclusion.
Key words:Cointegration, Vector Misspecification, The Power of test
一、引言
协整关系反映的是非平稳时间序列之间存在的长期稳定关系,例如消费与可支配收入、工资与价格水平、长短期利率等经济变量序列,单个都是非平稳的,而这些序列之间却存在着一种稳定关系,以使彼此偏离不会太远。通常用来检验协整关系的方法有三种:一是EG两步法,也就是基于回归残差的ADF单位根检验,Engle和Granger(1987);二是通过对ADL模型误差修正项系数的显著性分析进行的沃尔德检验,Boswijk(1989),Boswijk(1991);三是在VAR模型基础上进行的似然比检验,Johnsen(1988),Johnsen和Juselius(1990)。似然比检验主要用来分析诸多变量组成的VAR系统,借助典型相关理论,进行协整检验的同时确定协整关系。如果需要对特定的协整关系进行检验,前两种方法就显得比较简单实用,但具体应用中发现,这两种方法有时对于同样的数据却可能得到完全相反的结论。Kremers,Ericsson和Dolado(KED,1992)在强外生性假设下对此进行了详细的分析,指出EG检验施加了共同因子约束,如果该约束无效,虽然检验依然是一致的,但相对于不施加共同因子约束的沃尔德检验,就存在检验势的损失,进而造成二者