Forecasting Interest Rates Term Structure Based on Dynamic Four Shape Factors Model

Abstract

In this paper, the importance of curvature term structure movements on fitting and forecasting of interest rates term structure is analyzed. An extension of the exponential three-factor Dynamic Nelson-Siegel model-Dynamic Four Shape Factors model is proposed based on the Shape Factors Framework, where a fourth factor captures a second type of curvature. The new factor enhances model ability to generate volatility and to capture nonlinearities in the yield curve, which leads to a significant improvement of fitting and forecasting ability. The model is tested against the original Dynamic Nelson-Siegel model and some other benchmarks. Based on a fitting and forecasting experiment with Shanghai Security Exchange's fixed income data, it obtains significantly lower root mean square errors under three different forecasting horizons.