Macro Risk In China's Bond Yield

The term structure of interest rate and risk premium have always been the research hotspots since the financial crisis.The macroeconomics and finance classics have different opinions on the role of macro information in the formation of the yield curve of government bonds.In the classic literature of macroeconomics,the yield curve All the macro information has been included,that is,the macro information only affects the current rate of return and does not include information predicting future rate of return and information on risk premium.Finance,on the other hand,believes that yield and risk premium are determined by several potential factors that may come from macro information.The latest view is that some macro information does not affect the current yield curve but will determine the future yield curve.This paper uses the macro financial yield curve model to model the yield of Chinese government bonds.The key of this model is to add the assumption that the macro information does not affect the current rate of return and affect the future rate of return.The paper method constructs a dynamic factor model,using three factors to extract the information of the rate of return,and then two factors to extract the information of the macro variable.And assume that these two factors do not determine the current rate of return,but in the dynamic process these two factors will affect the future rate of return three factors.In this paper,the likelihood ratio test is used to reject the macro-factor directly affecting the current yield curve and think that it has an impact on the dynamic process of the yield factor.It is believed that some macro information does not enter the current yield curve,but it will affect the yield.Expectations are also formed in the future yield.At the same time,the model improves the ability to predict the rate of return compared to the classic dynamic Nelson-Siegel model that does not contain macro factors.The risk premium implied by the model is characterized by a reverse cycle,and the correlation coefficient with the industrial added value is-0.5,which is more in line with empirical facts.The ability to interpret excess returns has increased to 40%-60%.In the out-of-sample forecast,the forecast of yield is significantly better than the dynamic Nelson-Siegel model,and the ability to predict excess returns is significantly improved.