This paper provides evidence of long run purchasing power parity by performing a recently developed method to test for unit roots in the presence of structural breaks. Data consist of real exchange rate series for 20 countries including developed and developing economies. Structural breaks are detected in 18 countries and real exchange rates are found to be stationary in all countries except Japan. Estimated linear trends are the result of cross-country total factor productivity differentials between tradable and nontradable sectors. Estimated breaks correspond to large and permanent total factor productivity shocks associated with historical events like wars, structural reforms or deep economic recessions. An exercise with total factor productivity data shows that the Balassa-Samuelson effect explains the estimated long run trends in most countries.