In this paper a high-quality disaggregate database is utilized to exam
ine whether individual forecasters produce efficient exchange rate pre
dictions and also if the properties of the forecasts change when they
are combined. The paper links a number of themes in the ''change rate
literature and ''amines various methods of forecast combination. It is
demonstrated, inter alia, that some forecasters are better than other
s, but that most are not as good as a naive no-change prediction. Comb
ining forecasts adds to the accuracy of the predictions, but the gains
mainly reflect the removal of systematic and unstable bias.