Understanding forecast failure in ESTAR models of real exchange rates

The forecast performance of the empirical ESTAR model of Taylor et al. (2001) is examined for 4 bilateral real exchange rate series over an out-of-sample evaluation period of nearly 12 years. Point as well as density forecasts are evaluated relative to a simple AR(1) specification, considering horizons up to 22 steps head. The results of this study suggest that no forecast gains over a simple AR(1) model exist at any of the forecast horizons that are considered, regardless of whether point or density forecasts are used. Using simulation and non-parametric techniques in conjunction with graphical methods, this study shows that the non-linearity in the point forecasts of the ESTAR model decrease as the forecast horizon increases. Multiple steps ahead density forecasts of the ESTAR model are approximately normal looking, with no signs of skewness or bimodality. For an applied forecaster, there do not appear to exist any gains in using the non-linear ESTAR model over a simple AR(1) specification.