Rough Volatility Models
Takeaway
Empirical evidence suggests volatility is “rough” (Hurst H≈0.1), better captured by fractional stochastic processes than classical diffusions, improving fits to implied volatility surfaces.
The problem (before → after)
- Before: Heston/Black–Scholes miss microstructure-driven roughness and term-structure of skew.
- After: Rough fractional models reproduce short-maturity smiles and long-memory effects.
Mental model first
Volatility is like a jagged coastline—zooming in reveals more irregularity rather than smoothness. Fractional processes encode this persistence of roughness.
Just-in-time concepts
- Fractional Brownian motion with H in (0, 1/2).
- Rough Bergomi and related models; forward variance as primary object.
- Calibration via characteristic functions and Monte Carlo with hybrid schemes.
First-pass solution
Model log-variance as a Volterra integral driven by fBM; price options with Fourier or simulation; calibrate to implied vol surfaces across maturities.
Iterative refinement
- Microstructure foundations from order flow.
- Efficient sampling of fBM (Cholesky, circulant embedding).
- Risk management: Greeks under rough dynamics.
Principles, not prescriptions
- Match empirical scaling laws; avoid over-smoothing volatility.
- Use forward variance for stability.
Common pitfalls
- Naive discretizations bias dynamics.
- Overfitting short-term smiles without robust out-of-sample checks.
Connections and contrasts
- See also: [/blog/black-scholes], local/stochastic vol literature.
Quick checks
- Why “rough”? — H small implies irregular sample paths.
- Why better smiles? — Short-maturity skew and term structures align with data.
- Why forward variance? — More stable calibration target.
Further reading
- Gatheral et al. (source above) and follow-ups