| In recent years, with the continuous development and improvement of the financial market, as well as the deep innovation of market-oriented of the interest in our country, the term structure of national bonds is becoming more reasonable and the trading scale bigger. Therefore, on one hand, it requires us to carry out deeper research into the term structure, providing useful information for investors and helping our country to establish currency policy; On the other hand, rich statistics of national bonds will also offer us a basis for the research into interest term structure.This paper will first introduce several classical models of static term structure and then analyze the advantages and disadvantages of these models when fitting yield curves. After comparing, the writer thinks N-S model, which could better fit the yield curves of the national bonds with its better stability, is suitable to be used to estimate interest term structure in our country. Thus, this paper will use N-S model to estimate the hidden interest term structure of the national bonds. Judging from the test result, N-S model perfectly fits the hidden interest term structure of the submitted national bonds. It also shows that our national bonds market is relatively effective and investors are rational.Based on the daily interest term structure that N-S model estimates t, this paper works out the immediate yield of diverse mature term national bonds which is1,2,3,4,5,7and10years respectively. After correlation analysis to different lists of immediate yield, the writer finds that no one of the coefficient of correlation is equal to one. That means single factor can’t depict the dynamic features of instantaneous interest but by several factors. It also shows the necessity of establishing multi-factor models. Then, this paper analyzes the principal component of different lists of immediate yield. The result indicates that the first four components are better at explaining abilities and the cumulative abilities is up to99.941%. So, the paper holds the view that employing the first four factor to depict the dynamic features of instantaneous interest is suitable.Since the single factor model is inconsistent with the reality, the paper enlarges it from one to four based on single factor CIR model, establishing a four-factor CIR model. In terms of the estimate of four-factor CIR model, this paper first make use of regressing method to compute the initialize data of each parameter, and then uses Calman filtering method to calculate the best estimate of model parameter. After that, the writer chooses to use the root mean square error as the index to evaluate whether the fitting effect of four-factor CIR model is good. The outcome indicates that four-factor model can fit the sample perfectly. |
