Real exchange rates in the short, medium, and long run

This paper uses long-horizon autocorrelations and variance ratio statistics to test for long-term mean reversion in real exchange rates. Unlike most previous tests of this hypothesis, the tests do reject a random walk for monthly data in the post-Bretton Woods era; however, the statistics indicate that positively-correlated innovations, rather than mean reversion, are the source of the rejection. Tests using annual data for the twentieth century also reject the random walk. In this case, however, the rejection can be attributed to mean reversion and confirms that PPP is a long-term phenomenon.