Robust Estimation of the Term Structure

Abstract:     
Rarely in financial economics is the contrast between theory and reality more troublesome than in the problem of estimating the term structure of interest rates. The Nelson-Siegel-Svensson (NSS) functional form has become one of the most widely used models for doing so among academics, central bankers and practitioners. While many studies reported numerical difficulties when working with the NSS model, comparatively little attention has been paid to the practical problems of robust estimation of NSS. This paper conducts a thorough examination of the scope of these problems, the link between the methods employed to fit the NSS model, and the reliability of the estimated parameters and zero-coupon yields. Our investigation uses government bond portfolios of developed and emerging markets. We find that the NSS estimation problem confounds the commonly used gradient and direct search methods, but is amenable to global optimization methods, most particularly the hybrid particle swarm optimization introduced in this paper. Our results are consistent across the four countries, both in- and out-of-sample, and for perturbations in prices and starting values. For academics and practitioners estimating term structures, this study provides clear evidence of the noise that injudicious choice of optimization method can introduce in the estimated values, as well as suggesting and validating a method that works well for the NSS model.

Number of Pages in PDF File: 43

Keywords: Nelson-Siegel-Svensson, term structure, particle swarm optimization, robust estimation

JEL Classification: C13, C61, E43

working papers series