The effect of spillover on the Johansen tests for cointegration: a Monte Carlo analysis

This paper investigates the effect of spillover (i.e., causality in variance) on the Johansen tests for cointegration by conducting a Monte Carlo experiment where 16 different data generating processes (DGP) are used and a number of factors that might affect the properties of the Johansen cointegration tests are varied. The result from the simulation study clearly shows that spillover effect leads to an over-rejection of the true null hypothesis. Hence, in the presence of spillover it becomes very hard to make inferential statements since it will often lead to erroneous claims that cointegration relationships exist.