Forecasting the term structure of Korean government bond yields using the dynamic Nelson-Siegel class models

In this paper we propose and examine various extensions of the three-factor dynamic Nelson-Siegel model with the purpose of forecasting. We enhance the flexibility of the model by adding an additional driving factor or allowing for regime shifts in the model parameters. The regime changes are modeled through a recurring regime switching process or a change point process. Out-of-sample one through 6-months ahead forecasts are generated and evaluated using monthly Korean government bond yield data at sixteen different maturities. This paper finds that the three-factor model performs best for both short and long forecast horizons. Incorporating additional factor or multiple regimes does not seem to improve the out of- sample predictive accuracy of the yield curve forecasts.