Simultaneity, forecasting and profits in the US dollar/Deutschemark futures market

One and a half decades ago critics emphasized the poor post-sample forecasting performance of traditional economic models of exchange rates, which it was claimed, could not outperform a naïve random walk. In the past decade the focus of exchange rate research has shifted to the study of microstructure issues with intraday data. Nevertheless, the forecasting failure of traditional models has not been fully addressed. This paper demonstrates that a trading routine, based on post-sample forecasts by the authors' simultaneous model (Goss and Avsar, 2000), can produce risk-adjusted profits for a seven-day holding period, but not for positions held for one month. While this result may be interpreted as evidence against market efficiency, it does not imply that agents should discontinue the use of currency futures markets for risk management purposes.