MORE UNCERTAINTY ABOUT THE UNIT-ROOT IN US REAL GNP

The new KPSS test is used to test the null hypothesis that U.S. real G NP is a trend stationary process. When appropriately sized, the test f ails to reject the trend stationary null. This provides an important c ounter-example to the generic inability to reject the difference stati onary null hypothesis for output by classical hypothesis testing. The evidence in favor of the trend stationary representation is weakened, however, by showing that such size correction dramatically reduces pow er. Reversing the null and alternative hypotheses, this paper compleme nts Rudebusch's (1993) recent analysis.