From Black-Scholes to Stochastic Volatility: Why Market Reality Requires Better Models

I · The model

What Black-Scholes claims

The 1973 model makes one structural bet: that a single number, \(\sigma\), describes the randomness of an asset for the life of an option. Under geometric Brownian motion with constant volatility, continuous frictionless hedging, and a constant rate, every European option on the same underlying and expiry must be priced by the same \(\sigma\). This is not a minor implication — it is the testable implication. Invert the formula on market prices and the implied volatility should come back flat across strikes:

Until October 1987, index option markets roughly obeyed this. Then they stopped.

II · The refutation

What any trading day violates

Take the model's assumptions in order and hold them against an ordinary day of market data. Returns are not Gaussian: daily index returns show excess kurtosis and negative skew; moves the model rates as once-in-millennia arrive within careers (the arithmetic is in Note VI). Volatility is not constant: it clusters — quiet weeks follow quiet weeks, violent days follow violent days — and it moves inversely with price, the leverage effect. Prices jump: earnings, policy surprises, and liquidity spirals produce discontinuities no continuous diffusion generates. Each violation leaves a fingerprint in option prices, and the market, which prices repricing risk whether or not the model does, records them all in one curve:

Implied volatility across strikes for a 6-month option (S₀ = 100). Curve: implied vols recovered from Heston-model prices calibrated to equity-like dynamics — illustrative of the post-1987 index skew. Dashed line: what Black-Scholes says the curve must be. The slider in section IV moves the dashed line.

Out-of-the-money puts trade at implied volatilities several points above at-the-money; out-of-the-money calls below. Traders are paying up for crash protection and the market is quoting that premium through the Black-Scholes formula, using implied volatility as the unit. The smile is the market's standing annotation on the model: your Gaussian is too thin on the left, and your constant σ is not constant.

III · The repair

Heston's intuition: let variance breathe

Heston (1993) promotes variance from parameter to process:

Three mechanisms do the work. Mean reversion (\(\kappa\), \(\theta\)): variance is pulled toward a long-run level, so volatility shocks decay — this produces clustering, and it makes long-dated smiles flatter than short-dated ones, as observed. Vol-of-vol (\(\xi\)): variance itself is risky, which fattens both tails of the return distribution and gives the smile its curvature. Correlation (\(\rho < 0\)): the leverage effect, wired in explicitly — negative return shocks arrive with rising variance, skewing the distribution left and tilting the smile downward toward high strikes. With \(\rho\) near \(-0.7\), the model generates the equity index skew almost as a matter of course. And because the model has a closed-form characteristic function, whole volatility surfaces can be calibrated in seconds — a large part of why Heston, and not a dozen worthier candidates, became the industry workhorse.

IV · The comparison

Black-Scholes vs. Heston, strike by strike

Below, the flat-σ model meets smile-consistent prices. The Heston call values are precomputed offline from the Heston characteristic function (Heston 1993, "little trap" formulation, Simpson-rule integration of the P₁/P₂ integrals; parameters in the table note). The Black-Scholes prices are computed live at whatever flat volatility you choose. No single σ reconciles the two columns — that is the entire point.

Six-month European call prices across strikes: precomputed Heston values (points) against live Black-Scholes at the chosen flat σ (line).

Pricing error, BS minus Heston. Matching at-the-money leaves the wings mispriced — cheap crash protection on the left, overpriced upside on the right. Slide σ and watch the error seesaw: fixing one region breaks another.

Heston parameters: S₀ = 100, r = 2%, q = 0, T = 0.5y, v₀ = 0.04, κ = 2.0, θ = 0.05, ξ = 0.6, ρ = −0.7. Values computed offline in Python (Simpson rule, 4,000 nodes on [0, 200]) and hardcoded; implied vols recovered by bisection.

V · The residue

What stochastic volatility still misses

Heston is a better description, not a true one. Its diffusive variance cannot move fast enough to explain the steep smiles of short-dated options — a market pricing overnight jump risk needs jumps, which is why practice moved to Bates (Heston plus Merton jumps) and beyond. Its single variance factor forces the whole term structure of skew to move in lockstep, while real surfaces twist. Its calibrated parameters, re-fit each morning, drift in ways the model says they should not — a slow-motion confession that the model is a curve-fit to a deeper process. And the modern frontier (rough volatility, with Hurst exponents near 0.1) suggests volatility's memory is structurally unlike anything a Markovian diffusion can produce. Each generation of models recapitulates the same pattern: capture the last generation's known failure, inherit an unknown one.

VI · Conclusion

Model risk and epistemic humility

The deepest lesson of the smile is not that Black-Scholes is wrong — every model is wrong — but that markets metabolized its wrongness. The formula survived its own refutation by becoming a quoting convention: implied volatility is the market's chosen coordinate system for expressing exactly how non-Black-Scholes the world is. That is the relationship with models this lab advocates. A model is an instrument for locating structure in uncertainty — for saying precisely where reality diverges from an idealization, and by how much. Used that way, Black-Scholes remains among the most useful wrong ideas ever written down. Used as prophecy — as a claim that the tails are thin because the formula needs them to be — the same equation has helped sink funds and banks. The mathematics did not change between those two uses. The humility did.