Non-Linear Filtering in the Estimation of a Term Structure Model of Interest Rates

The methods of the class of Kalman filters have recently been used in the estimation of the term structure of interest rates. These methods can employ both time-series and cross-sectional aspects of term structure models. This paper compares the performance of two kinds of non-linear Kalman filter algorithms Extended Kalman Filter (EKF) and Square-Root Unscented Kalman Filter (SRUKF) in estimating one popular exponential-affine term structure model. Simulation results show that SRUKF is of higher approximation accuracy and stronger numerical stability than EKF is. Key-Words: Non-linear filtering, Square-Root Unscented Kalman Filter, Extended Kalman Filter, Term structure of interest rates, Exponential-affine term structure model, One-factor Vasicek model