Core Path · L5.3.10

Two-factor models

14 minG2++Two-factor Hull-WhiteCorrelationDecorrelationCalibration

By the end you will be able to

  1. 1Explain why a one-factor model makes every rate move together, and what a second factor adds.
  2. 2Write G2++ and two-factor Hull-White, and move between their parameters.
  3. 3Read the two-factor fit to all 54 grid points (RMS 2.27bp against one factor's 5.43bp) and the decorrelation it costs, corr(1y, 10y) = 0.153.
  4. 4Say which products need a second factor, and how desks fix parameters the vol grid cannot pin down.

Why it matters

In every model so far one random shock moves the whole curve: the 1-year and the 30-year rate always go the same way, so their correlation is exactly 1 (L5.3.01: Modelling the short rate). Real curves twist. The bank's own interest rate risk scenarios include a steepener and a flattener, short and long rates moving in opposite directions (L3.6.06: The standard shock scenarios), and the swaption grid hides a dip at short expiries that one factor misses by 17.3bp. A second factor fixes both. This lesson shows how, and what it costs.