On the dynamics of treasury bond yields: From term structure modelling to economic scenario generation

Abstract

This study proposes an algorithmic framework that integrates the dynamic modelling of the term structure of treasury bond yields with the generation of market-consistent economic scenarios. The unscented Kalman filter (UKF) that works as a non-linear learning instrument for historical bond yields under the multi-factor models can facilitate the in-sample yield curve modelling, underpinned by statistical inferences, and further enhance the performance of the out-of-sample bond pricing and yield predictability. Moreover, market views that gauge the holistic assessments of macroeconomic prospects are incorporated into our framework. As such, the market-consistent economic scenarios are driven primarily by the dynamics of the term structure of bond yields and aggregate market sentiments among investors, offering a new instrument for interest rate risk management.