The Predictability of Exchange Rates

Abstract

This study investigates the predictability of exchange rates by using the long horizon regression approach (or sometimes called the Error Correction Model (ECM)) derived from the Vector Error Correction Model (VECM) for Canada, Japan and Switzerland for the period between 1973:Q2 and 2013:Q4. The predictive ability of the exchange rate models were evaluated according to in-sample analysis and out-of sample analysis. The in-sample analysis results suggest that the fundamentals are useful to explain the long horizon changes in the logarithm of the exchange rates under the assumption of country specific income elasticities for Canada and Japan but not Switzerland. On the other hand, the out-of sample analysis presents the evidence that whether the ECM or the random walk (RW) explains the nature of exchange rates is time varying. During recessions, the RW explains the exchange rate changes better while, the ECM works better in expansions.