He Empirical Research Of Zero-coupon Bonds Term Structure Of Interest Rates

Interest rates forecasting is very important for bond portfolio management,derivatives pricing and risk management. For a long time, researchers mainly focuson the theory and modeling of the interest rates term structure, and have pro-posed kinds of parametric or semi-parametric rate models. Functional Data Analy-sis (FDA) has been widely used in financial, engineering, pharmaceutical and otherfields, since it was proposed in1982. The paper combines interest rates and FDAby using the Nonparametric Functional Data Analysis (NP-FDA) to do empiricalresearch for the interest rates term structure of Chinese zero-coupon bonds for thefirst time.Considering the monthly Chinese zero-coupon bonds’ yield to maturity as sam-ple data, we try to forecast the yield to maturity of Chinese zero-coupon bondsby using three NP-FDA methods and one traditional time series analysis method:the kernel estimator with automatic bandwidth selection, the kernel estimator withglobal choice of the number of neighbour, the kernel estimator with local choice ofthe number of neighbour and one-order vector auto-regression.We consider two forecasting horizon to evaluate the prediction accuracy of theproposed estimators by their RMFSE. See the one-step rolling forecast (h=1) andthe12-months long term forecast (h=12). The result shows that:First, VAR(1) performs better than NP-FDA;Second, the methodology of NP-FDA do well in trends forecasting;Third, when h=12the prediction error decreases as the maturity growth.Most of these jobs are carried out in R, source code see Appendix A.