Testing for Stationarity and Cointegration in an Unobserved-Components Framework 1 James Morley and Tara M. Sinclair Murray Weidenbaum Center on the Economy, Government, and Public Policy and Department of Economics Washington University in St. Louis One Brookings Drive St. Louis, MO 63130 April 8, 2005 JEL Classifications: C32, C15 Keywords: Unobserved Components, Cointegration, Common Trend, Unit Roots VERY PRELIMINARY! PLEASE DO NOT QUOTE OR CITE. Abstract While tests for stationarity and cointegration have important econometric and economic implications, they do not always offer conclusive results. In this paper we suggest that exploiting the parametric structure of the multivariate correlated unobserved components framework can provide a more powerful way to test for stationarity and cointegration than the non-parametric, asymptotic tests currently available. The parametric test nests a partial or restricted unobserved components model within a more general unobserved components model. Then we estimate both the general and the restricted models and determine the likelihood ratio test statistic. The distribution of this likelihood ratio test statistic is nonstandard, but a Monte Carlo simulation provides proper error bands for use in inference. We then compare these results to the asymptotic, non-parametric KPSS test and the common trends test of Nyblom and Harvey (2000). 1 The authors gratefully acknowledge the support of the Murray Weidenbaum Center on the Economy, Government, and Public Policy for this project. We wish to thank Tom King, Michael Owyang, and Christoph Schleicher for helpful discussions and comments. All remaining errors are our own. 1