A0739
Title: Observation-driven filtering of the term premium
Authors: Dennis Umlandt - University of Innsbruck (Austria) [presenting]
Abstract: Investors in long-term sovereign bonds are compensated for interest rate risk by a term risk premium, which is represented by the higher expected return on long-term bonds relative to short-term bonds. The spanning hypothesis would suggest that the yield curve contains all the information necessary to forecast the time-varying term premium. However, standard affine term structure models often perform poorly without additional macroeconomic data, possibly due to their restrictive linear risk pricing assumption. We propose a new class of dynamic term structure models that incorporate non-linear observation-driven dynamics to capture richer term premium variations. These models offer improved predictive power and sparse parameterization, while still allowing for direct likelihood-based inference. In an empirical application, we find that the dynamics of US risk premia can be well captured by a three-factor model if observation-driven risk premia are taken into account.
