Most economists intuitively consider purchasing power parity (PPP) to be true. Nevertheless, quite surprisingly, the empirical literature is not very supportive for PPP. In this paper, however, we find evidence in favor of PPP using a new test. The test is embedded in a Markov regime-switching model for the exchange rate, because earlier papers have shown that this model describes the data better than the popular random walk. We allow for PPP by making the regime-switching probabilities depend on the PPP deviation. Our second result is that PPP disequilibria have become shorter- lived for some exchange rates, which may be due to an increase in the trade openness of the countries involved.