The Power of the "Objective" Bayesian Unit-Root Test

Abstract

Abstract Some researchers, for example, Koop (1992), and Sims (1988), advocated forBayesian alternatives to unit-root testing over the classical approach using theaugmented Dickey-Fuller test (ADF). This paper studies the power of what Koop(1992) has called the ”Objective” Bayesian approach to unit-root testing. Koop’s”objective” Bayesian test is interesting in light of the call by Phillips (1991a,1991b) for more objective Bayesian analysis of time series. We apply the ”ob-jective” Bayesian unit-root test to a study of long-run purchasing power parity(PPP) in the post-Bretton Woods era and also Monte Carlo simulations. Overall,our results suggest that the ”objective” Bayesian test is biased in favor of trend-stationarity. We conclude that, at least for the ”objective” Bayesian test, it is notbetter than the classical ADF approach in unit-root tests, and because of its bias,the ”objective” priors suggested by Koop is not appropriate.Journal of Economic Literature Classification: C11, C22, F31I wish to thank Professor Stephen M. Miller, participants at the Southern Eco-nomic Association Annual Meetings, and Department of Economics, Universityof Connecticut, brown-bag seminar for comments on an earlier draft of this paper.Remaining errors are my sole responsibilities.