Purchasing power parity: three stakes through the heart of the unit root null

A recent influential paper (O'Connell 1998) argues that panel data evidence in favor of purchasing power parity disappears once test procedures are altered to accommodate heterogeneous cross-sectional dependence among real exchange rate innovations. We present evidence to the contrary. First, we modify two extant panel unit root panel unit root tests to eliminate the upward size distortion induced by contemporaneous cross-sectional dependence. Second, we exploit a recently-introduced test, based on SUR techniques, that also remains valid in the presence of cross-sectional dependence. Using the three new tests, we find overwhelming evidence in favor of real exchange rate stationarity during the post-Bretton Woods era among OECD economies, as well as among a larger group of “open” economies. We also find emphatic evidence of stationarity using O'Connell's GLS test. Bias-corrected parameter estimates indicate that deviations from PPP erode more quickly for real exchange rates defined using wholesale rather than consumer price indices. Monte Carlo experiments indicate that several of the tests discussed here have considerable power against the unit root null.