Core Path · L5.4.03

Multi-factor affine models

14 minMulti-factor modelsDai-SingletonAdmissibilityPrincipal componentsCorrelationEigenvalues

By the end you will be able to

  1. 1Write an N-factor affine model, place models in Dai and Singleton's Aₘ(N) classes and check admissibility.
  2. 2Read the principal components of the Bank of England's OIS curve as the case for three factors.
  3. 3Compute the correlation of two yield changes from a model's loadings, and compare G2++, AFNS and the data.
  4. 4Use the eigenvalues of K under ℚ and ℙ to judge stability and half-lives.

Why it matters

In a one-factor model every yield moves with the short rate, so a 1-year and a 10-year rate are perfectly correlated. Module 5.3's two-factor G2++ fit swung the other way: a correlation of 0.153 (L5.3.10: Two-factor models). The Bank of England's OIS curve says 0.794. A model that hedges a 10-year swap with 1-year futures, or projects a curve for a stress test, lives or dies by that number. How many factors does the curve need, and of what kind?