Cointegration Analysis Based on Linear Error in Variable Model

Abstract

Linear regression coefficients in traditional cointegration analysis are estimated using ordinary least square criterion, when the response variables and input variables have ignorable observation error, using ordinary linear regression model will lead to cointegration regression model is not unique. In view of this situation, linear error in variable regression model instead of ordinary linear regression model is used to estimate cointegration regression model parameters. At last, computer simulation examples and the cointegration of the Shanghai Stock Index(SZZS) and Shenzhen Component Index(SZCZ) verified the correctness of this method.