Abstract

This paper investigates the forecasting performance of macroeconomic fundamentals on exchange rate returns using a macro-finance approach. Exchange rate movements are endogenously determined by the ratio between domestic and foreign stochastic discount factors, through which the macroeconomic fundamentals nonlinearly model the exchange rate dynamics. Using three floating nominal exchange rates, i.e. DEM /USD, GBP /USD, and JPY /USD observed at the monthly as well as quarterly time frequencies, this paper has the following findings. First, five out of the six model-implied time-varying foreign exchange risk premiums satisfy the Fama conditions (Fama, 1984). Second, comparing to the random walk model, this no-arbitrage macro-finance model reduces the forecasting root mean square errors, especially for data observed at the quarterly time frequency.