Interest Rates Term Structure Forecasting and Bond Portfolio Risk Management

The dynamic Nelson-Siegel-style models, which are popular in the literature of interest rates term structure forecasting, may be unstable because of the potential existence of unit roots in the parameter series. In this paper, the dynamic Nelson-Siegel-style models are modified by modelling the first-order differenced instead of original parameter series. Empirical study shows that the modified models yield significantly smaller RMSE than the original ones when forecasting, meanwhile the forecasting RMSE of the modified models is much more stable even when the forecasting horizon increases. Besides, we interpret that the traditional duration of one bond is proportional to its expected return. Based on that, the information concerned with the forecasted bonds’ yields, which can be calculated by discounting their future cash flows after the term structure of interest rates is forecasted, are introduced to the constraint equations of the traditional duration matching model in interest rates risk hedging. Empirical study shows that incorporation of the forecasted yields information of bonds can improve the performance of interest rates risk hedging when applied to mid-term or long-term target bonds.