Forecasting Bond Yields with Segmented Term Structure Models

Recent empirical analysis of interest rate markets documents that bond demand and supply directly affect yield curve movements and bond risk premium. Motivated by those findings we propose a parametric interest rate model that allows for segmentation and local shocks in the term structure. We split the yield curve in segments presenting their own local movements that are globally interconnected by smoothing conditions. Two classes of segmented exponential models are derived and compared to successful term structure models based on a sequence of out-of-sample forecasting exercises. Adopting U.S. interest rates data available from 1985 to 2008, the segmented models present overall better forecasting performance suggesting that local shocks might indeed be important determinants of yield curve dynamics.