Empirical Researches On Term Structure Models Of Interest Rates With Macroeconomic Factors

Macroeconomists want to understand the effect of macroeconomic factors on interest rates,while financial economists look for the factors that drive the yield curve dynamics.To shed light on both issues,we describe the joint dynamics of bond yields and macroeconomic variables in a Vector Auto Regression,where identifying restrictions are based on the absence of arbitrage.We use three models to explore the relationship between macro factors and the term structure of interest rates,such as Structure Vector Auto Regression model,ATSM without macro factors(Yield-Only Model) and ATSM with macro factors(Yield-Macro Model).The Yield-Macro Model presents an empirical macro-finance model that allows us to identify inflation shocks, real activity shocks and monetary policy shocks,and traces the effect of these shocks on the prices of bonds of different durations.Compared to a standard VAR,this approach has the advantage of incorporating the information embedded in a large cross-section of bond prices.Moreover,the pricing equations provide new ways to assess the model's ability to capture risk preferences and expectations.The results of SVAR suggest that many parameters in the model are not significant, but it does not matter when we explore the interaction between macro factors and interest rates.Interest rates themselves can explain up to 70%of the variation in bond yields,and Variance Decomposition tells us latent factors are primary factors of the movement of yield curves.Impulse Responses tell us that three macro factors have different effects on zero-coupon rates with different terms.Their impacts on yield diminish when time goes on,but real activity will still have impact on yield curves when time goes on.The results of Yield-Only Model suggest that three latent factors we assume are the traditional three factors.They are "level","slope" and "curvature".The primary factor is "level" factor.The three factors have a greater impact on short-term rates,while they have a smaller impact on long-term rates.The results of Yield-Macro Model suggest that the coefficients of macro factors in pricing equations are smaller than the coefficient of latent factors.Latent factors are still the primary factors that can explain most of the movements of yield curve.The result of Impulse Responses is somewhat different with the results in SVAR and it has explanation power on the economic situation.We also find that the short rate we forecast from the short rate process in Affine Term Structure Model is highly consistent with the real time series of short rate,so we can formulate monetary policy on the basis of this process.Moreover,we find that imposing the cross-equation restrictions from no arbitrage helps in out-of-sample forecasts.Incorporating macro factors in a term structure model further improves forecasts.The best performance model is Random Walk when we forecast short end rates,while the best model is Yield-Macro model when we forecast middle end and long end interest rates.We can conclude that Yield-Macro model has a best performance than the other models.The innovations of this paper are as follow.First,we use Dynamic Term Structure Model developed by Dai and Singelton recently,compared with traditional model VAR.Secondly,our short rate dynamic is consistent with Taylor rule,and we can formulate monetary policy on the basis of this process.Finally,this paper tries to prove that our model have a better performance than other models.