ABSTRACT
Predicting the future exchange rates is a difficult task due to various reasons. The most important econometric reasons are the nonlinearity of the exchange rates with a meandering nature. That is to say, a typical foreign exchange rate series would be non-stationary in the short run and mean reverting in the long run. As for the economic reasons, exchange rate series are likely to be regime-switching series either due to central bank policies or issues related to the microstructure of the financial markets, etc. Then choosing a good switching variable, and deciding on the true nature of the switches, become the issues to tackle. In short, due to various reasons – as explained in the previous chapters of this book – predicting the foreign exchange rates is a difficult task, and almost all the prior attempts in the literature have been unsuccessful in beating even the out-of-sample prediction accuracy of a simple driftless random walk process. Given these difficulties and the disappointing empirical results, focusing on how to predict the direction of change rather than making point estimations might be a good idea.
