Evolving yield curves in the real-world measures: a semi-parametric approach

Abstract

In this paper we show how to evolve a yield curve over time horizons of the order of years using a simple but effective semiparametric method. The proposed technique preserves in the limit all the eigenvalues and eigenvectors of the observed changes in yields. It also recovers in a satisfactory way several important statistical features (unconditional variance, serial autocorrelation, distribution of curvatures, eigenvectors) of the real-world data. A simple financial explanation can be provided for the methodology. The possible financial applications are discussed.