Risk-Neutral Densities and Gaussian Mixtures
A smooth option price curve implies a distribution used to price payoffs at expiry. A Gaussian mixture is one way to fit that distribution to a sparse option chain. A usable fit needs non-negative weights, total probability one, the correct forward mean and acceptable repricing errors. Sparse quotes leave room for substantially different tails, and support restrictions need to suit the underlying.
How it works
Under standard discounting assumptions a smooth call price curve reveals a risk-neutral density through strike curvature. One can fit a positive mixture of component densities and price the observed options by integration. This imposes probability constraints directly and provides an interpretable way to vary tail shape while keeping the central option fit.
Worked example
Take a mixture with component means 90 and 110 and weights 0.5 each. The mixture mean is 100, which agrees with a forward of 100. Changing the weights to 0.75 and 0.25 moves the mean to 95. The fit would then need adjusted components or weights to preserve the same forward.
Validation and limitations
Check for negative densities, total mass, the forward mean and the payoff re-pricing errors. Normal components assign some mass to negative prices. If the asset’s domain excludes them, choose compatible components or constraints. Distinguish the pricing density from a frequency forecast under the physical measure.
Further reading
- Prices of State-Contingent Claims Implicit in Option Prices
Douglas Breeden and Robert Litzenberger · PDF
- Arbitrage-Free SVI Volatility Surfaces
Jim Gatheral and Antoine Jacquier · PDF