Heston Model - Stochastic Volatility Options Pricing

Last reviewed: May 17, 2026 by Options Analysis Suite Research.

What Is the Heston Model?

The Heston model is a stochastic volatility extension of Black-Scholes published by Steven Heston in 1993. Where Black-Scholes assumes a single constant σ, Heston treats the variance itself as a random process that mean-reverts toward a long-run level and is correlated with the underlying price. The model captures two empirical features Black-Scholes can't: the volatility smile (and skew), and the volatility-of-volatility behaviour observed during regime transitions.

Heston is the closest closed-form-friendly stochastic volatility model in production use. It admits a semi-analytic price via Fourier inversion (the characteristic function is known in closed form), which makes it fast enough for vanilla options and reliable enough for calibration to listed volatility surfaces.

The Stochastic Differential Equations

The underlying follows dS/S = μ·dt + √v·dW₁ where v is itself stochastic: dv = κ·(θ − v)·dt + ξ·√v·dW₂, and the two Brownian motions have correlation ρ. The five parameters control different empirical features:

• κ (kappa): speed of mean reversion. Higher κ pulls variance back to its long-run level faster.
• θ (theta): long-run variance level. Often calibrated to match the long end of the term structure.
• ξ (xi, "vol of vol"): volatility of the variance process. Higher ξ thickens the tails and steepens the smile.
• ρ (rho): correlation between spot and variance. Negative ρ produces the leverage effect (selloffs spike vol), which gives equity smiles their characteristic asymmetric shape.
• v₀: initial variance. Set by the front-month ATM IV at calibration time.

What Does Heston Capture That Black-Scholes Doesn't?

• The volatility smile: OTM puts and OTM calls trade at different IVs than the ATM strike.
• Term structure: different expirations have different ATM IVs, with characteristic mean-reverting shapes.
• The leverage effect: equity vol rises when spot falls, captured by negative ρ.
• Vol of vol risk: the market's pricing of how much volatility itself can move.

What Doesn't Heston Capture?

• Jumps: Heston has continuous paths. Earnings gaps, FDA releases, and macro jumps need a Jump Diffusion or Variance Gamma model.
• Very short-dated smiles: overnight and weekly smiles are typically too steep for Heston to fit cleanly without compromising longer expirations.
• Co-movement of skew and term structure under stress: Heston's parameter set is rich, but real markets exhibit regime changes Heston can struggle to track in real time.

The Heston Characteristic Function

Heston's defining computational advantage is the closed-form characteristic function for ln(ST): a complex-valued function of frequency that encodes the full risk-neutral distribution of the underlying at expiration. Pricing a vanilla call under Heston reduces to evaluating a single-dimensional Fourier integral involving this characteristic function, typically via Carr-Madan FFT or numerical integration like Fourier-COS. This makes Heston pricing fast enough for production use, including during calibration loops where the model is repriced thousands of times per fit. Without the closed-form characteristic function, Heston would need Monte Carlo simulation, which is an order of magnitude slower and produces noisy prices that interfere with calibration stability.

Calibration in Practice

Heston calibration takes the listed volatility surface as input and finds the five parameters (κ, θ, ξ, ρ, v₀) that minimize the squared error between Heston-computed prices and listed market prices. The optimization is typically Levenberg-Marquardt or a similar gradient-based non-linear least-squares method. Best practice is to fit in Black-Scholes IV space rather than dollar prices, since that normalizes the error across strikes and expirations so deep-OTM options don't dominate the loss function with their tiny absolute prices. The calibrated parameters are not unique: ξ and ρ in particular trade off against each other, and bound-constrained optimization is necessary to keep parameters in economically-sensible ranges (κ > 0, ξ > 0, ρ ∈ [−1, 1]).

What Does Heston Capture That Black-Scholes Doesn't?

• The volatility smile: OTM puts and OTM calls trade at different IVs than the ATM strike, and Heston produces these systematically through ρ (skew direction) and ξ (smile curvature).
• Term structure: different expirations have different ATM IVs, with characteristic mean-reverting shapes governed by κ pulling variance back toward θ.
• The leverage effect: equity vol rises when spot falls, captured by negative ρ. This produces the characteristic asymmetric smile shape on equity options that Black-Scholes can't reproduce with a single σ.
• Vol of vol risk: the market's pricing of how much volatility itself can move. ξ controls the kurtosis of the implied distribution and the wing prices.
• Cross-strike consistency: a single calibrated Heston parameter set produces prices for all strikes and expirations, unlike Black-Scholes where each strike has its own σ.

What Doesn't Heston Capture?

• Jumps: Heston has continuous paths. Earnings gaps, FDA releases, and macro jumps need a Jump Diffusion or Variance Gamma model, or the Bates extension which combines Heston with Merton-style jumps.
• Very short-dated smiles: overnight and weekly smiles are typically too steep for Heston to fit cleanly without compromising longer expirations. SABR per-expiration is often the better tool for the front end.
• Co-movement of skew and term structure under stress: Heston's parameter set is rich, but real markets exhibit regime changes Heston can struggle to track in real time. Recalibration frequency matters.
• The forward-skew dynamics relevant to barrier options and other forward-dependent exotics. Local-Stochastic-Volatility (LSV) hybrids exist for this use case.

How OAS Uses Heston

The platform calibrates Heston to the live listed volatility surface using the semi-analytic Fourier inversion price. Calibration produces all five parameters; the model surface is then evaluated at any strike/expiration grid and compared to Black-Scholes on the same grid. The "model divergence" view exposes where Heston disagrees with Black-Scholes, typically on OTM strikes and the wings of the smile, which is exactly where Black-Scholes is known to misprice. The Heston-implied probability distributions also feed the per-ticker expected-move and probability views, where the smile-aware fat-tail behavior produces more honest tail probabilities than Black-Scholes' lognormal assumption.

Calibrate Heston in the pricing calculator

Related Concepts

Black-Scholes (vs) · SABR (vs) · Local Volatility (vs) · Jump Diffusion · Variance Gamma · Stochastic Volatility · Volatility Skew · Volatility Smile · Vol of Vol · Leverage Effect · IV Crush · Dealer Gamma · Tail Risk · Variance Risk Premium · Implied Volatility · Realized Volatility · Vanna / Charm / Vomma Exposure · Model Divergence · Calibration · eSSVI Parameterization · Butterfly Arbitrage · Monte Carlo · FFT Pricing · Greeks Reference · Heston vs Black-Scholes · SABR vs Heston · Model Landscape · Market-Structure Ontology

Try this model in the pricing calculator

This section is part of the Options Analysis Suite Documentation. Browse the full model index or compare alternatives in the pricing calculator.