Mean reversion in long-horizon real exchange rates: Evidence from Latin America

This paper examines mean reversion in real effective exchange rates in six leading Latin American economies during the XXth century using a new data set. A unit-root approach is complemented by an error-correction model including key fundamentals such as terms of trade, trade openness and relative productivities. Unit-root testing shows a very slow process of reversion - if any - to a constant mean in the original series, rejecting the strict PPP hypothesis; however, mean reversion is found after allowing for trends and structural breaks with a half-life average of 1½ years for the six countries. We also found reversion to a conditional mean defined by the co-integrating relationship with an average half-life of 2½ years. Our estimates, although lower than the 3-5 year range that motivated the Rogoff's puzzle, still indicate the presence of important obstacles to the adjustment process that need further investigation.