Currency forecasting based on an error components-seemingly unrelated nonlinear regression model

This paper proposes to forecast foreign exchange rates by means of an error components-seemingly unrelated nonlinear regression (EC-SUNR) model and, simultaneously, explore the interrelationships among currencies from newly industrializing economies with those of highly industrialized countries. Based on the empirical results, we find that the EC-SUNR model improves on the performance of forecasting foreign exchange rates in comparison with an intrinsically nonlinear dynamic speed of adjustment model that has been shown to outperform several other important models in the forecasting literature. We also find evidence showing that the foreign exchange markets of the newly industrializing countries are influenced by those of the highly industrialized countries and vice versa, and that such interrelationships affect the accuracy of currency forecasting.