forecasts of the dollar-mark rate which perform somewhat better than forward rates constructed on covered interest rate parity in predicting monthly changes in the spot rate. It is claimed that the estimated model is more successful than the forward rate as a predictor of changes in exchange rates challenges the view that exchange risk premiums are non-existent. On the other hand, their simplifications of the portfolio balance model are too severe to explain the major portion of observed changes in exchange rates, since observed changes in exchange rates have been predominantly unexpected. They argue that if it is true that observed changes in exchange rates have been predominantly unexpected, then the mileage may lienot so much in refining the assumptions about portfolio behaviour but rather in fuelling the rational expectations framework with better measures of the expected future values of asset stocks and wealth variables. Questions are hence raised