Know these first
- L5.3.09: Calibration to caps and swaptions
- L5.4.03: Multi-factor affine models
Math Atoms
By the end you will be able to
- 1Calibrate a market model in layers: Rebonato's abcd shape, one Φ_k per forward to reprice every caplet, and the shape re-fitted to co-terminal swaptions.
- 2Estimate the correlation between forward rates from the Bank of England's curves, by regression (β = 0.0556) and by least squares, and reduce it to three factors.
- 3Explain why a model that reprices every caplet cannot reach some co-terminal vols here (the Cauchy-Schwarz bound), and choose between model A and model B.
- 4Compare the market model's fits with Hull-White's from L5.3.09: Calibration to caps and swaptions, in and out of sample.
Why it matters
Hull-White has two parameters and the market quotes dozens of options, so in L5.3.09: Calibration to caps and swaptions it could only be fitted to one part of the market at a time. That lesson promised market models that fit every caplet, or every co-terminal swaption, by construction, and both together only nearly. A market model has 40 forward rates, each with its own vol, and a 40 × 40 correlation matrix: parameters to spare. It can indeed fit every caplet exactly, or every co-terminal swaption exactly. With this module's quotes it can't do both, and the reason is in the quotes, not the model.