Stochastic-Volatility and Two-Regime LMM
The displaced-diffusion LMM shifts the rate to change how volatility responds to its level. Adding stochastic volatility or allowing two volatility regimes creates more ways for the smile to move. These additions can improve a fit while making parameters harder to identify. Compare models on the same quotes, test their sensitivities under changed inputs, and check what happens when a regime or volatility factor is removed.
How it works
A two-regime stochastic-volatility LMM lets the volatility state or its dynamics vary between regimes. It can represent periods where the smile behaves differently even when the underlying forward level is similar. A displaced diffusion provides an additional way to control how volatility depends on forward rates.
Worked example
Take a rough mixture with identical forwards and weights of 80% for a 15% volatility regime and 20% for a 35% regime. The average variance corresponds to about 20.62% volatility. Pricing each regime and blending payoffs generally gives a different price from plugging that average variance into a single Black formula.
Validation and limitations
Check how the fit depends on the probability of entering and staying in each regime. Base-maturity smiles may not identify transition rates. Compare pricing and hedging of held-out callable exposures with an otherwise similar one-regime model.
Further reading
- A Unified View of LIBOR Models
Kathrin Glau, Zorana Grbac and Antonis Papapantoleon · PDF
- The Market Model of Interest Rate Dynamics
Alan Brace, Dariusz Gątarek and Marek Musiela · Journal record