Purchasing power parity and Markov regime switching

We revisit the evidence on PPP in the twentieth century allowing for Markov regime switching in the ADF regression. Using Lopez et al.'s (2005) extension of Taylor's (2002) original data set, our results are: (1) For most countries, there are periods over which the real exchange rate is stationary and PPP holds, and periods over which the real exchange rate is non-stationary and PPP does not hold, namely, regime-dependent stationarity. Thus, real exchange rate stationarity is a stochastic event itself. (2) The probability that the real exchange rate is stationary is less than 50% for most countries. (3) There is evidence of non-stationarity during both the Bretton Woods and the recent float periods for the majority countries. The comparative performance of PPP is slightly better during Bretton Woods than the recent float. Copyright 2006 by The Ohio State University.