Abstract

This paper formulates a conformity test for cointegration for a multivariate I (1) process obeying a VAR specification. The test statistic is a function of the characteristic roots of the sample covariance matrix of the cointegral vector; the latter is obtained from the unrestricted estimator of the underlying parameters of the VAR. It is further shown that this test procedure is applicable to the case where the I (1) process is a MIM (k ), i.e. a multivariate integrated moving average process, the moving average being of order k  < .

The test statistic, under null of cointegration, has a normal limiting distribution.