Previous findings of long-run purchasing power parity come mainly from data for industrial countries, raising the issue of whether the results suffer sample-selection bias and exaggerate the general relevance of parity reversion. This study uncovers substantial cross-country heterogeneity in the persistence of deviations from parity. The results show that it is more likely, rather than less likely, to find parity reversion for developing countries than industrial countries. Although some persistence variations may partly reflect country differences in structural characteristics such as inflation experience and government spending, a considerable portion of those variations seems unaccounted for. (C) 2000 Elsevier Science B.V. All rights reserved.