Stochastic correlation and risk premia in term structure models

This paper analyzes a term structure model that allows for both stochastic correlation between underlying factors and an extended market price of risk specification. We show that significant improvement in bond fitting and portfolio performance is obtained by the model. However, the restriction on market price of risk has a more negative impact on bond price fitting and forecasting, whereas the restriction on correlated factors has a more negative impact on hedging performance. The model has good predictive power for bond risk premia. Once our model factors are taken into account, other predictive factors become insignificant.