Abstract
This paper is a comment on Soren Johansen's (1994) paper on estimating systems of trending variables. The pedagogical and diagnostic value of using univariate time series methods is emphasized together with the use of small scale experiments that give insight into the sensitivity of unit root test procedures to misspecification of the deterministic components. The test statistics used in the likelihood approach advocated by Johansen are compared with several other test statistics, in particular, those of Box and Tiao (1977) and Stock and Vatson (1988). We also compare the corresponding methods to estimate pulling equilibria. We present the outcomes of two Zlonte Carlo experiments to illustrate some points.