基于参数空间的ST-ECM模型的协整检验问题研究

在非线性平滑转移误差修正模型（ST-ECM）的协整检验中，由于存在未识别参数而使协整检验统计量构造困难，同时由于目前文献普遍使用的泰勒展开近似法并不能精确替代原始非线性模型，从而导致协整检验统计量功效较低。本文首先在遍历未识别参数的参数空间的基础上构造了ST-ECM模型协整检验的supF统计量，推导了supF统计量的极限分布并说明了其收敛性质。接着，蒙特卡洛仿真模拟结果显示，supF统计量在ST-ECM模型协整检验中具有良好的检验水平和功效，且supF统计量的功效明显优于EG统计量、F*NEC统计量和inft统计量。最后，本文对亚洲六个国家的利率期限结构预期假说进行了验证，结果表明中国、新加坡和泰国三个国家的利率期限结构预期假说成立且存在非线性调整效应，supF统计量较其他统计量具有更高的检验功效。

Research on the Cointegration Test of the ST-ECM Model

Considering the difficulty of proposing cointegration test statistics caused by the unidentified parameter in nonlinear Smooth Transition Error Correction Model (ST-ECM) and the present existing cointegration test statistics'' lower power which is because that the Taylor approximation can''t replace the original nonlinear model accurately, this paper proposes the supF statistic based on grid search on the unidentified parameter space to test cointegration in ST-ECM.The Monte Carlo simulation results show that the supF statistic has good power and size, and it has better power than the other traditional statistics such as EG (Engle and Granger, 1987), F*NEC (Kapetanios et al., 2006) and inft (Ouyang, 2013).The last empirical results show new evidence that the term structure of interest rates holds true for three of the six Asian countries and their adjustment mechanisms are nonlinear, and the supF statistic has better power than the other traditional statistics.