Bond Risk Premia and Gaussian Term Structure Models

Abstract

Cochrane and Piazzesi (2005) show that (i) lagged forward rates help predict bond returns and that (ii) modern Markovian dynamic term structure models (DTSMs) cannot match the evidence. We develop the family of Conditional Mean DTSMs where the dynamics depend on current yields and their history through a moving-average component. Our preferred Conditional Mean model combines one moving-average with the usual three Gaussian risk factors, closely matches the bond risk premium measured from predictive regressions and provides better forecasts of bond returns. Our framework nests Duffee (2011) models with a small “hidden” factor and our results compare favorably with his 5-factor model. Conditional Mean models are easier to estimate than state-space term structure based on Kalman estimates of latent factors.

Keywords: term structure models, bond risk premia, Markovian dynamics, Kalman filter

JEL Classification: E43, E47, G12