Abstract
Purchasing power parity is re-examined using a new data set containing long time span data for thirty-four economies, half of them industrial and half of them from Latin America. This is the most comprehensive CPI based data set ever assembled for the purposes of testing purchasing power parity. It is also the first time that this kind of data is available for a significant number of developing economies.The empirical results differ from those of previous studies in virtually all respects. First, in constant-only specifications the hypothesis of nonstationarity is rejected for one quarter of the five hundred and sixty one real exchange rates in the sample. The number of rejections increases in constant-trend specifications. However, with proper accounting of significant time trends the evidence supports nonstationarity in just eighty two percent of the sample. Previous studies, by contrast, overwhelmingly reject nonstationarity. Second, half lives of convergence to purchasing power parity for stationary real exchange rates average a little over two years. This is a much faster rate of convergence to purchasing power parity than previous work would suggest indicating that there is no "purchasing power parity puzzle" after all. Finally, it is shown that there is no survivorship bias since the performance of PPP is the same for developed and developing economies.