分位数回归下的因果关系检验  

Test the Causal Relationship with Quantile Regression

作　　者：伍兴国[1,2] 

机构地区：[1]暨南大学经济学院,广东广州510632 [2]东莞职业技术学院财经系,广东东莞523808

出　　处：《统计与信息论坛》2018年第3期10-16,共7页Journal of Statistics and Information

摘　　要：Granger因果关系在计量经济学中占有重要地位。在实践中,采用Granger检验是在最小二乘回归下进行的,检验的是条件均值之间的因果关系,在此基础上扩充到检验条件分位数上的因果关系,即线性分位数回归下的Sup-Wald检验法;利用该检验法实证分析全国城镇居民人均可支配收入增量对人均消费性支出增量之间的关系,结果表明其之间存在单向的因果关系,并且在较低分位数时城镇居民人均可支配收入增量对人均消费支出增量的因果关系并不显著,只有在分位数较高时这种因果关系才显著存在。Granger causality plays an important role in the econometric system.Granger tests are carried out based on the least squares regression in practice,and the causal relationship between the conditional mean values of the variables is verified,On this basis,the causality relation of the test conditional quantile is extended,and the Sup-Wald test method under the linear quantile regression is proposed,and use the test method to empirically analyze the relationship between incremental per capita disposable income and per capita consumption expenditure in urban areas.The results show that there is a one-way causal relationship between them,and in the lower quantile,the per capita disposable income increase of urban residents per capita consumption expenditure causality is not significant;only in the higher quantile,this causal relationship is significant.

关 键 词：分位数回归 Sup-Wald检验 因果关系 可支配收入 消费性支出 

分 类 号：F062.9[经济管理—政治经济学] O212[理学—概率论与数理统计]