The Performance of SETAR models by Regime: A Conditional Evaluation of Interval and Density Forecasts

The aim of this paper is to analyse the out-of-sample performance of SETAR models using daily data for the Euro effective exchange rate. The evaluation is conducted on point, interval and density forecasts. The benchmark used for the comparison is a linear AR model for point forecast evaluation and a GARCH model for interval and density forecasts. In each case the models are evaluated unconditionally, over the whole forecast period, and conditionally, on the regimes of the SETAR models. The results show that, in general, the performance of the SETAR models improves significantly for the forecasts governed by the regime(s) with fewer observations. However, overall the GARCH model is better able to capture the distributional features of the series and to predict higher ordered moments.