Application Of Multi-Factors Mixed Model For Researching Term Structure

Term structure has very important theory and reality significance. A well term structure model is the cornerstone for research. And it should not only fit to current spot interest rate, but also predict future spot interest rate. As I know, all multi-factors term structure models assumed the factors are follow the same instantaneous interest rate model. But " identity" will have a negative impact on describing term structure. This paper presents a new class of "Mixed" term structure models. Mixed model assumed some factors follow Vasicek model, others follow CIR model. Because of imperfectness, the data we get form market is subject to "noise". We should filter noise for following research. And Kalam filter method performs well in filtering noise. Then we introduce how to construct the state space by Kalman filter method, and present mix model state space. A Monte Carlo simulation is conducted and simulation results perform "well". What's more, we can improve estimate precision by increasing the number of simulation or extending the interest rate term. On the basis of real bonds market status, an empirical research is also conducted. The empirical study shows that the two-factor mixed model can fit current spot rate curve, but can not predict the future spot rate.