A Monte Carlo Study of the Forecasting Performance of Empirical Setar Models

In this paper we investigate the multi-period forecast performance of a number of empirical self exciting threshold autoregressive (SETAR) models that have been proposed in the literature for modeling exchange rates and GNP, amongst other variables. An indicator of when such models are likely to forecast well is suggested based on the serial dependence of regimes, as a means of distinguishing between types of nonlinearities that can be exploited for improved fit versus those that contribute to a better (relative to linear models) out-of-sample forecast performance. In our study the indicator provides a reasonable guide to those models which embody nonlinearities that may yield improved conditional mean forecasts.