Real-Time Exchange Rate Predictability: High Frequency Forecasts with Low Frequency Data

Abstract

This paper evaluates short-run out-of-sample exchange rate predictability with real-time data for 15 OECD countries from 1973 to 2013. We consider the Taylor rule fundamentals model, where the variables that enter the Taylor rule are used to forecast exchange rate changes, and the Taylor rule differentials model, where a Taylor rule with postulated coefficients is used in the forecasting regression. We find evidence of predictability with the Taylor rule fundamentals model for 9 out of 15 countries. The Taylor rule differentials model performs worse, and the evidence of predictability is the weakest with the conventional monetary and PPP models.