Core Path · L5.4.04

Nelson-Siegel and Svensson

14 minNelson-SiegelSvenssonLevel, slope, curvatureLeast squaresYield curve fitting

By the end you will be able to

  1. 1Write the Nelson-Siegel and Svensson curves and read their factors as level, slope and curvature.
  2. 2Fit them as least squares at a fixed decay rate λ (M-D03: Linear regression) plus a one-dimensional search over λ.
  3. 3Read the factors month by month from 2009 to 2025, and say why the level before 2016 is an extrapolation.
  4. 4Compare the fits with the Bank of England's own curve on 31 December 2025, spot and forward: 1.53bp on spot rates hides 36.3bp on forwards.

Why it matters

A curve has fifty points, but its moves from month to month come in about three shapes. L5.4.03: Multi-factor affine models found that on the Bank's OIS panel three principal components explain 99.67% of monthly changes: a shift, a tilt and a bend. Nelson and Siegel built a curve from exactly those three shapes, with one weight on each. Three numbers then describe a whole curve, a month of curves becomes three time series, and a model of three factors becomes possible. That is the idea the rest of this module rests on.