Abstract: In the method of testing smooth transition cointegration, estimating parameters are uncertain and the problem of cointegration test is complex. This paper proposes a smooth transition regression model and conducts a Bayesian nonlinear cointegration analysis. Based on the selection of parameters prior of the model and the characteristics of the posterior conditional distributions of the parameters, Metropolis-Hasting within Gibbs sampling algorithm is designed to estimate the parameters and bayesian unit root test is utilized to test the stationarity of regression residual, addressing the uncertainty of parameters estimation and the complexity of cointegration test. At the same time, the research applies exchange rate of RMB against U.S. dollar and interest rate differential between China and U.S. to conduct an empirical analysis. The research outcome indicates that MH-Gibbs can effectively a estimate the parameters of the smooth transition model, and we find there is smooth transition cointegration relationship between exchange rate fluctuation and interest rate differential.

Key words: transition regression model, nonlinear cointegration, Bayesian analysis, MH-gibbs sampling, exchange rate fluctuation

摘要： 针对平滑转移模型参数估计不确定性导致的协整检验方法相对复杂问题,提出基于平滑转移模型的贝叶斯非线性协整分析。通过模型的统计结构分析,选择参数先验分布,结合参数的后验条件分布特征设计Metropolis-Hasting-Gibbs混合抽样方案,据此估计平滑转移模型的参数,并对回归残差进行贝叶斯单位根检验,解决参数估计过程中遇到的参数估计不确定性及协整检验复杂的问题;利用人民币对美元汇率与中美两国的利率数据进行实证分析。研究结果表明:MH-Gibbs抽样方案能够有效估计平滑转移模型的参数,中美汇率波动和利差之间存在平滑转移协整关系。

关键词: 平滑转移模型, 非线性协整, 贝叶斯分析, MH-Gibbs抽样, 汇率波动