【作者】 张洁；

【导师】 王少平；

【作者基本信息】 华中科技大学 ， 数量经济学， 2009， 博士

【摘要】 协整理论指出，如果某两个或多个同阶非平稳时间序列的线性组合可以得到一个平稳的误差序列，则这些非平稳时间序列存在长期的均衡关系，即存在协整。协整的基本概念刻画了变量之间的长期均衡，由此派生的误差修正模型(ECM)则进一步描述了协整变量短期偏离向长期均衡的调整行为。传统的线性协整和ECM在现代时间序列分析中得到广泛应用。但是，它们都暗含了两个基本假设：短期偏离向长期均衡的调整是连续的；调整的速度是对称的，即无论偏离大小，都以相同速度移向均衡水平。事实上，在现实经济中，由于调整成本等因素的存在，这两个假设未必总是得到满足。阈值协整正是针对传统协整理论的这一缺陷而提出，把非线性的阈值自回归引入协整模型，拓展了协整理论的适用范围，因而成为计量经济学研究的一个前沿领域。阈值协整模型把传统的线性协整理论扩展到非线性，对协整理论的发展做出了巨大贡献，有广泛的应用前景，在国外受到充分重视并发展迅速。但是在我国，相关的研究非常之少，还处于刚刚起步的阶段。本文对阈值协整模型的理论背景、模型设定、估计和检验方法等问题做了系统介绍，并在此基础上进行了深入的理论和应用研究，本文的主要贡献和意义体现在以下几个方面：(1)理论的创新。现有文献中，对阈值协整模型的协整系数分布特征鲜有提及。本文基于蒙特卡罗仿真试验，发现总体样本长期均衡系数的分布非常接近正态分布，而不同机制中，协整变量的均衡系数分布存在显著差异。这一研究在阈值协整系数分布特征这一空白领域做出了探索性的工作，具有显著的方法论意义。(2)研究方法和视角的创新。对我国CPI和PPI研究的文献，计量方法上大都基于传统的线性协整模型考察二者之间的关系，本文首次使用非线性的阈值向量误差修正模型(TVECM)进行实证研究，结论表明，在不同的通胀水平下，我国CPI和PPI的传导方向是不一致的，同时CPI向均衡调整的速度也显著不同。在对我国汽油价格和国际原油价格的分析中，本文建立三机制的阈值模型，从计量的角度证实了我国汽油价格确实存在人们所说的“跟涨不跟跌”现象。(3)本文的研究结论具有更丰富的经济学和政策含义。模型回归结果表明，我国通胀水平较高时，PPI向CPI的传导显著，CPI下调速度缓慢，通胀水平较低时，CPI的下滑带动PPI下降，CPI向均衡的调整较快。这也在一定程度上说明，我国的通胀中有很大的成本推动因素，而通缩多源于有效需求不足的冲击。这一结论隐含的政策意义为，对通胀的治理应注重供给面的改善，而应对通缩则要集中于刺激总需求。同时我国汽油价格随国际原油价格的调整也是上调迅速下调缓慢，即存在“跟涨不跟跌”的现象，暴露了现有成品油定价机制中的弊端。显然，传统的线性协整无法刻画这种非对称的调整关系，上述发现正是基于非线性的阈值协整模型所产生的结论，从这个意义上说，本文具有显著的学术和应用意义。更多还原

【Abstract】 Cointergration theory indicates that if a stationary error series could be conclude from linear combination of two or more same-order nonstationary time series, there is long-run equilibrium relationship in these nonstationary time series. The basic concept of cointergration characterizes the long-run equilibrium relationship between variables, while ECM deviding from cointergration characterizes the adjustment behavior of cointergration vector from short-run divergence to long-run equilibrium. Traditional linear cointegration theory and ECM can be used widely in time series analysis. But,they impose two basic restrictions :(1) It is continuous in the period of adjustment from short-run divergence to long-run equilibrium; (2)The speed of adjustment is symmetric,whichs means this adjustment shifts to equilibrium level at same speed regardless of divergence scale. However, these hypotheses were not always satisfied in the actual economy because of the adjustment costs and other factors. Then, the linear cointegration model may be incorrect. For this defect of the traditional theory, threshold cointegration theory, which introduced the non-linear threshold autoregressive(TAR) into cointegration model and expands the scope of application, has been a frontier study of econometrics.While extending traditional linear cointegration theory to non-linear theory, threshold cointegration model has made great contributions to the development of cointegration theory. Because of the wide applications, it receives adequate attentions and the rapid development in foreign countries. However, in our country, the relevant research is very little, even in its infancy stage. This paper introduces systematically the theoretical background, model specification, estimation and testing method of the threshold cointegration model, and do a intensive study for the theory and application. The followings are the main contributions of this paper: (1) Theoretical innovation. There are very little study for the distribution characteristics of cointegration coefficient of threshold cointegration model in available literature. Based on the monto carlo simulation ,we find that the overall samples distribution of long-run equilibrium coefficient is very close to normal distribution, but there are a significant difference between the distributions of cointegration variables’ equilibrium coefficient indifferent regimes. This study is an exploratory survey in distribution characteristics of threshold cointegration coefficient, and it has important methodology sigificance. (2) Innovation on research methods and perspectives. Most research on the relationship between CPI and PPI use traditional linear cointergration model, while in this paper we do empirical study with nonlinear Threshold Vector Error Corrector Model. The transmit direction of CPI and PPI is not the same in different inflation levels, and the speed which CPI adjust to equilibrium level varies significantly. This paper sets up a threshold cointergration model to analyze gasoline prices and international crude oil prices, and confirms the phenomenon that gasoline prices in China easily rise but not fall in the perspective of econometrics. (3) This paper has rich economic and policy sense. The estimated results shows that in higher inflation, the transmit from PPI to CPI is significant and CPI declines slowly, while in lower inflation, CPI declines following by PPI decline and adjusts to equilibrium level relatively fast. To some extent, this implies that cost-push is an important factor in China’s inflation, and deflation is mostly caused by the shortage of effective demand. Therefore,we should focus on improving the supply side to control inflation, and expanding aggregate demand to deal with deflation. Meanwhile, the adjustment of gasoline prices in China to international crude oil prices also rise rapidly and falls slowly, which exposes the defects of the existing pricing mechanism. Obviously, only from nonlinear threshold cointergration models could we get those findings. From this point, this dissertation has much significance of knowledge and application.更多还原