Is purchasing power parity overvalued?

The purpose of this paper is to reexamine the hypothesis of long-run purchasing power parity among industrial countries under the current float using the panel data approach. We propose a new finite-sample panel data unit root test procedure that allows for a General serial correlation structure and arbitrary contemporaneous correlation in model innovations across countries. This procedure is then applied to the recently developed panel data unit root tests of Im, Pesaran, and Shin (1996) and of Maddala and Wu (1999). The test results show sufficient evidence against the unit root null for most samples. By comparing our results with the existing ones in the literature, we found that the method of assuming a restrictive serial correlation structure tends to weaken evidence against the unit root null.