Cointegration, Error Correction, and Vector Autoregression

A regression model with unit root variables may provide spurious results. The random error of a regression model should be white noise for robust results. White noise error implies disequilibrium is a temporary phenomenon. Although the distributional assumptions of random error imply its white noise behaviour, the time series variables used in a regression model which are themselves random may distort the white noise behaviour of the error term. The concept of cointegration put forward by Clive Granger resolves the problem of getting robust result in the presence of unit root in the variables used in a regression model. To understand the nature of dynamics, all variables used in a regression model may be considered as endogenous that generates a system of equations in a regression model in vector form. Vector autoregression is appropriate when the theory fails to provide the direction of causality uniquely. This chapter deals with these problems of a regression model with time series data.