Heston stochastic volatility
Models option smiles and skew using correlated spot and mean-reverting variance.
How it works
In Heston, instantaneous variance follows a mean-reverting square-root process. Spot and variance shocks may be correlated. Negative correlation tends to contribute to the downside equity skew. Mean reversion, the long-run variance and volatility of variance help determine the level, curvature and evolution of the implied volatility surface.
Worked example
Take variance initially 0.09, long-run variance 0.04 and mean-reversion speed 2 per year. The one-year mean of variance is 0.04 + 0.05 exp(−2), about 0.04677. This expectation is a property of the variance process. Pricing an option requires the distribution of the underlying and its joint evolution with variance.
Validation and limitations
Check calibration residuals by strike and maturity, test numerical pricing against the constant-variance limit and compare simulation schemes near zero variance. Different parameter sets can produce similar vanilla premiums. Use barriers or forward-starting options to see which unresolved dynamics matter.