Joint detection of unit roots and cointegration: Data-based simulation

This paper investigates two problems which a practitioner of Johansen's cointegration analysis faces: the pre-testing problem and the inconsistency problem. Johansen's cointegration analysis requires hypothesis testing at four stages, and thus Type I errors accumulate. In addition, it often occurs that the results of unit root tests and those of cointegration tests are inconsistent. In order to overcome these problems, an information-criterion-based model selection method for jointly detecting unit roots and cointegration is proposed. Through empirical applications and Monte Carlo simulations, the usefulness of this method is demonstrated.