Real versus nominal forecast errors in the prediction of foreign exchange rates

Abstract

Imposing a small number of common, but strong, assumptions allows the varianve of the ex post forecast error in forward exchange rates to be decomposed into its parts—that part due to errors in forecasting relative inflation (the ‘nominal’ error) and that part due to ‘real’ shocks. For a sample of large and small industrial countries we show that real shocks have been predominant in the post-Bretton Woods world. This result implies that models of exchange rate determination which do not explicitly incorporate real shocks are inappropriate for explaining recent exchange rate volatility.