Empirical Study On The Dynamic Term Structure Model With Macro Factors Of Interest Rates

The term structure of interest rates is a bridge between macroeconomics and microeconomics,which not only reflects the economic changes of the whole society,but also plays an important role in pricing financial assets and risk management.The research of the term structure of interest rates is one of the most important research directions in the financial field.Based on the HJM framework,this paper establishes a dynamic term structure model with time-varying risk prices.By assuming that the risk price is an affine function of potential state variables,the model can be transformed into a state space form in order to estimate the parameters using the maximum likelihood estimation method based on Kalman filtering.The empirical results of the yields of treasury bonds in Shanghai Stock Exchange show that the cumulative variance contribution rate of the three factors to interest rate changes is as high as 99.45%.The cumulative variance contribution rate of the first three factors to interest rate changes is as high as 99.45%.The three-factor term structure model of the interest rates can explain the dynamic changes of the yields well,and the model fits the short-term bonds better than the long-term bonds.The dynamic trend graphs of the factors show that the first factor has a small change range and maintains a horizontal trend in the sample interval as a whole.The second factor fluctuates more frequently than the first factor,which is related to the change of the term structure in the direction of the slope.The third factor has the same effect on short-term and long-term bonds,but the opposite direction on the medium-term bonds.Then,this paper expands the latent factors model,introducing three macro factors into the model,and analyzing the dynamic correlation between potential factors and macro factors through the VAR process.The correlation between the horizontal factor,the slope factor and the macro factor is strong,the curvature factor and the macro factor are weakly correlated,and the slope factor is more sensitive to the impact of the macro factor.The empirical results show that using observable macro variables to replace some unobservable latent factors has a considerable effect on improving the effectiveness of the model,but the setting of pure latent factor models or pure macro factor models can not reflect the changes in returns well.Among the models in this paper,the model setting effect with two latent factors and two macro factors is the best.