Garch effects on a test of cointegration

Abstract

This article discusses the effects of GARCH type error processes on the use of the Engle and Granger cointegration test for two variables. Simulation results indicate that (nearly) integrated GARCH processes, as well as GARCH processes that are not covariance stationary, change the critical values. An application to testing for cointegration between spot and futures prices illustrates the practical relevance of using the appropriate critical values. © 1994 Kluwer Academic Publishers.