Dynamic Relationship Between Term Structure Of Interest Rates And Macro Economy Based On Markov-VAR Perspective

The term structure of interest rates,used to analyze and forecast economic conditions based on the relationship between interest rates or bond yields and different terms,has been a major focus for macroeconomists.The study of the term structure of interest rates is of great significance against the backdrop of China's marketization reform of interest rates and macroeconomic regulation and control increasingly relying on price rather than quantity.On the basis of China's financial market reform and macroeconomic conditions,this thesis makes a comparison of several models' fitting effect with China's term structure of interest rates after the 2008 financial crisis,exploring various theories of the term structure of interest rates and its relevance to macroeconomic variables.It is theoretically and practically important of this thesis to discuss the basic features of the term structure of interest rates,the economic meaning of latent variables and its nonlinear relevance to macroeconomic variables.Based on the dynamic Nelson-Siegel model proposed by Diebold and Li in 2006,the paper uses the method of maximum likelihood estimation to get the three factors which are horizontal factor,slope factor and curvature factor.And using the linear combination of the yield of different maturity bonds as the proxy variable to fit the three potential factors.From the fitting effect of the three factors,the horizontal factor is similar to the trend of the slope factor and the proxy variable,and the fitting effect of the curvature factor is poor.On the basis of three potential factors,we set ?as a time-varying parameter and use the same data to carry out load analysis and factor estimation.The results show that the fitting of three potential factors and agent variables is improved.The three factors of DNS and time-varying DNS method are respectively VAR regression,and the dynamic fitting effect of time-varying DNS model was better.Considering the actual situation of China's economic environment,Markov transformation is introduced,Markov-VAR regression is carried out by three methods,and four regression models are used respectively.Markov-VAR regression model without arbitrage condition is considered,Conditional Markov-VAR regression model,regardless of the non-arbitrage conditional time-varying Markov-tvtp-VAR regression model and the no-arbitrage condition Markov-tvtp-VAR regression model.By comparison,it is found that the three factors of the DNS model and the time-varying DNS model perform well in the Markov-VAR model which does not include the no-arbitrage condition,and the time-varying DNS model is better.In order to examine the relationship between interest rate term structure and macroeconomic,the three macroeconomic variables of GDP,CPI and bank overnight interest rate are brought into the DNS-VAR model.The results show that the three potential factors and macroeconomic variables are related to their own lagging items.Significant relationship.It is further found that the correlation between the level factor and the CPI and GDP is high,which reflects the expectation of the economic growth and inflation level of the market participants.The slope factor is highly correlated with the bank overnight interest rate,and more is the reflection of the economic cycle Interest rate information;but the curvature factor and the correlation between the three macro variables.The results of the impulse response further confirm the regression model derived from the conclusion.Because of the introduction of Markov transform in the model of adding macroeconomic variables,the empirical study of DNS-Markov-VAR and time-varying DNS-Markov-tvtp-VAR model is carried out,and impulse response analysis is carried out.The results show that most of the parameters on the main diagonal are more significant,the state 1 is longer,the regression parameters are mostly negative,so we define state 1 as a recession state,state 2 as a prosperous state,The correlation coefficient is smaller than that in the prosperous state,and the three macroscopic variables show the opposite characteristics in two states,which is in line with the actual situation of China's economy.Finally,using DNS-Markov-VAR and time-varying DNS-Markov-tvtp-VAR model to predict the rate of return with the expiration time is 6,12,24,36,48,60,72,84,96,120,180,240 months.It is found that the short-term forecasting effect of the short-term yield is not good,and the fitting effect is improved with the increase of the forecast interval.But the accuracy of long-term forecasting of the long-term yield is better.