Department of Economics,.pdfVIP

Wojciech W. Charemza*) and Ewa M. Syczewska**) Joint application of the Dickey-Fuller and KPSS tests *) University of Leicester, U.K. **) Central and East European Economic Research Center, University of Warsaw, Poland Please send any communication to: Wojciech W. Charemza, University of Leicester, Department of Economics, University Road, Leicester LE1 7RH, UK. e-mail: wch@le.ac.uk Acknowledgement: The financial support of the A.C.E. Project Structural change and spillovers in the East European reform process is gratefully acknowledged. We are indebted to Derek Deadman and Len Herk for their helpful comments. We are also grateful to Alan G. Isaac and David Rapach for sharing their GAUSS procedures with us. We are solely responsible for any remaining deficiencies. Abstract For the joint Dickey-Fuller and KPSS test the probabilities of joint confirmation have been defined as those which jointly confirm the stationarity hypothesis. It is suggested that, where the Dickey-Fuller and KPSS statistics are jointly used, the conventional critical values for those tests should be replaced by symmetric critical power values. This corresponds to the probability of type 1 error for the Dickey-Fuller test and power of the KPSS tests obtained in the case where both marginal cumulative distributions are equal. In Monte Carlo experiments the symmetric critical power test values have been computed. Suggested keywords: Unit roots, nonstationarity, joint hypotheses testing Joint application of the Dickey-Fuller and KPSS tests It is often implied in the literature (e.g. by Amano and van Norden (1992) and Schlitzer (1996)) that testing the null hypoth