非线性协整关系及其检验方法研究

RESEARCH ON NONLINEAR COINTEGRATION AND ITS TESTING METHOD

文献［４］所提出的协整概念描述了向量时间序列中的长期线性均衡关系，从而可以称为“线性协整”．然而，经济系统中的许多时间序列是长记忆的，这些序列本身以及序列之间往往存在非线性关系，所以对这些序列进行线性协整分析是不太合适的．针对时间序列的长记忆和非线性特点，本文提出了向量时间序列非线性协整的概念，并运用神经网络方法对非线性协整关系存在性进行检验．同时，还讨论了这种非线性协整关系检验的可行性，给出了运用神经网络估计非线性协整函数的算法．最后，通过模拟试验说明了检验方法的可行性与有效性．

The concept of cointegration presented by Engle and Granger(1987) describes the long run linear equilibrium in a given vector series.However,many economic time series are long memory and nonlinear,and the relations between them are also nonlinear.Based on the long memory property and nonlinearity of a given vector time series,this paper presents the concept of nonlinear cointegration,and applies the neural network technique to the testing of the existence of nonlinearly cointegrated relationships.Furthermore,we discuss the feasibility of the testing procedure and present its algorithm.In the end the paper illustrates the effectiveness of the testing methods with a simulation study.