基于VAR模型的中国居民消费水平贝叶斯单位根检验

Bayesian Unit Root Test for Chinese Residents' Consumption Using Vector Autoregressive Model

针对经济时序DF单位根检验方法在小样本条件下功效偏低问题,应用贝叶斯统计方法对中国居民消费水平进行单位根检验,提高单位根检验的功效水平,建立向量自回归居民消费水平模型,并通过马尔可夫链蒙特卡罗仿真结合贝叶斯因子分别对农村居民消费和城镇居民消费进行单位根检验,结果表明贝叶斯单位根检验方法解决了向量自回归模型超参数估计的难题,克服了经典单位根检验在经济时序小样本下功效偏低的缺陷,提高了模型预测精度。

As classical Dickey Fuller test could have very low power for small sample research, Bayesian method is implemented to deal with the unit root test for Chinese residents＇ consumption, improving the power of unit root test. Based on Vector Autoregressive model of residents＇ consumption, using Markov Chain Monte Carlo （MCMC） simulation combined with Bayes factors, unit root tests are conducted for rural residents＇ consumption and urban residents＇ consumption respectively. The results indicate that the problem of super parameters in Vector Autoregressive model and poor size property in classical unit root test methods could be solved by Bayesian unit root test, and the precision of forecasting could be improved.