Forecasting exchange rates using principal components

We introduce a novel atheoretical approach to forecasting bilateral exchange rates. We first obtain principal components for a set of up to 20 bilateral exchange rates for a set of major currencies of interest. We then fit autoregressive processes to get one-period-ahead forecasts of each of the principal components and use these to forecast individual bilateral exchange rates. We focus on six major currencies, including the US dollar, Japanese yen, British pound, euro, and the Australian and Canadian dollars. Based on the daily data from 01/02/1999 to 08/03/2018, our results suggest this approach can be useful in forecasting some one-day and one-month ahead bilateral exchange rates, particularly for the pound and Australian dollar. The optimal number of principal components that should be included in the forecasting procedure depends on the currency of interest.