Modelling Term-Structure Dynamics for Risk Management: A Practitioner's Perspective

Modelling term-structure dynamics is an important component in measuring and managing the exposure of portfolios to adverse movements in interest rates. Model selection from the enormous term-structure literature is far from obvious and, to make matters worse, a number of recent papers have called into question the ability of some of the more popular models to adequately describe interest rate dynamics. The author, in attempting to find a relatively simple term-structure model that does a reasonable job of describing interest rate dynamics for risk-management purposes, examines two sets of models. The first set involves variations of the Gaussian affine term-structure model by modestly building on the recent work of Dai and Singleton (2000) and Duffee (2002). The second set includes and extends Diebold and Li (2003). After working through the mathematical derivation and estimation of these models, the author compares and contrasts their performance on a number of in- and out-of-sample forecasting metrics, their ability to capture deviations from the expectations hypothesis, and their predictions in a simple portfolio-optimization setting. He finds that the extended Nelson-Siegel model and an associated generalization, what he terms the "exponential-spline model," provide the most appealing modelling alternatives when considering the various model criteria.