Using panel data methods, we investigate long-run purchasing power parity by testing for unit roots in real exchange rates of industrial countries under the current float. The evidence against the unit root hypothesis is stronger for larger than for smaller panels, for monthly than for quarterly data, and when the German mark, rather than the United States dollar, is used as the base currency. While we find that accounting for serial correlation considerably weakens the evidence against the unit root null, the results as a whole are consistent with long-run purchasing power parity