Alternative Long-horizon Exchange-rate Predictors

Abstract

This paper employs quarterly observations on US dollar prices of the pound, Deutschmark, Swiss franc, and yen from 1973,2 to 1994,4 to sort out three broad issues raised by recent work showing that economic fundamentals have predictive power for exchange rates at long horizons. Three alternative fundamentals have been proposed in the literature: those implied by purchasing-power parity, uncovered interest parity, and the flexible-price monetary model. We first ask which of these three alternative fundamentals has the most predictive power. Secondly, we ask if pooling across currencies or if using multivariate statistical techniques improves prediction accuracy over standard regression techniques. Thirdly, we examine whether the conclusions drawn from statistical analyses of in-sample economet ric estimates concerning long-horizon convergence of exchange rates and their fundamentals coincide with those implied by analyses of out-of-sample forecasts. The short answers to these questions are; the monetary-model fundamentals, yes, and a qualified no.