Do panels help solve the purchasing power parity puzzle

While Rogoff (1996) describes the “remarkable consensus ” of 3 to 5 year half-lives of purchasing power parity deviations among studies using long-horizon data, recent papers using panel methods with post-1973 data report shorter half-lives of 2 to 2.5 years. These studies, however, do not use appropriate techniques to measure persistence. We extend median-unbiased estimation methods to the panel context, calculate both point estimates and confidence intervals, and provide strong evidence confirming Rogoff’s original claim. While panel regressions provide more information on the persistence of real exchange shocks than univariate regressions, they do not help solve the purchasing power parity puzzle