Exchange rate forecasting with DSGE models

Abstract

We run an exchange rate forecasting “horse race”, which highlights that three principles hold. First, forecasts should not replicate the high volatility of exchange rates observed in sample. Second, models should exploit the mean reversion of the real exchange rate over long horizons. Third, they should account for the international price co-movement seen in the data. Abiding by the first two principles an open-economy dynamic stochastic general equilibrium (DSGE) model performs well in forecasting the real but not the nominal exchange rate. Only approaches that conform to all three principles tend to outperform the random walk.