Static-Arbitrage Constraints on the SVI Parameterization

Gatheral and Jacquier characterize precisely the conditions under which the Stochastic Volatility Inspired (SVI) parameterization of the implied volatility smile, and its surface extension across maturities, is free of static arbitrage. Three types of static arbitrage are relevant: butterfly arbitrage, which requires the total implied variance as a function of log-moneyness k to be convex (∂²w/∂k² ≥ 0 everywhere) so that the risk-neutral density implied by the smile is non-negative; calendar-spread arbitrage, which requires that the total implied variance w(k, T) be non-decreasing in T at every fixed k; and the large-moneyness (Lee) moment formula constraints, which require the right and left wings of the smile to grow at most linearly in |k| as |k| → ∞. The raw SVI parameterization, with five parameters (a, b, ρ, m, σ), does not automatically satisfy any of these constraints, and the paper derives the explicit parameter-space conditions under which each constraint holds, providing a practical algorithm for butterfly-free and calendar-spread-free calibration. The SSVI (Surface SVI) sub-parameterization introduced in the paper imposes a specific functional relationship between the ATM total variance θ = w(0, T) and the remaining shape parameters, yielding a two-parameter family φ(θ) (plus the correlation ρ) such that the resulting surface is arbitrage-free by construction for any choice of φ satisfying three explicit inequalities. The power-law and Heston-compatible choices of φ(θ) are analyzed in detail, providing both a flexible fitting parameterization and one that is anchored to the Heston model's structural predictions for the term structure of the ATM variance. The Lee moment formula — which constrains the maximum slope of the smile wings in terms of the moments of the risk-neutral distribution — is recovered as a special case of the SSVI constraints. For the desk, the significance of this paper is primarily methodological: any quantitative analysis of implied volatility surface features — skew level, term structure slope, wing behavior — is meaningful only if conducted on a surface that has first been verified or fitted to be free of static arbitrage. An arbitrage-laden surface is not the output of any consistent pricing model, and conditioning signals derived from its features are therefore conditioning on numerical artifacts rather than on genuine market information. The SSVI parameterization is accordingly adopted as the desk's canonical surface representation for any analysis that conditions on surface shape.

Raw SVI total implied variance at log-moneyness k is w(k) = a + b{ρ(k − m) + √((k − m)² + σ²)} with parameters a ∈ R, b ≥ 0, ρ ∈ (−1, 1), m ∈ R, σ > 0. Butterfly freedom requires ∂²w/∂k² ≥ 0 for all k, which translates into explicit constraints on (a, b, ρ, m, σ) that are satisfied on a non-trivial subset of parameter space but not everywhere. Calendar-spread freedom across slices T_1 < T_2 requires w(k, T_1) ≤ w(k, T_2) for all k. The SSVI parameterization sets w(k, θ) = θ/2 · {1 + ρ φ(θ) k + √((φ(θ)k + ρ)² + (1 − ρ²))} where θ = w(0, T) is the ATM total variance and φ: (0, ∞) → (0, ∞) is the surface function to be specified. Butterfly freedom of SSVI holds if and only if φ(θ) ≤ 4/(θ(1 + |ρ|)) and (1 + |ρ|)φ(θ) ≤ 2, equivalently θ φ(θ)(1 + |ρ|) ≤ 4. Calendar-spread freedom holds if additionally θ ↦ θ φ(θ) is non-decreasing and θ ↦ θ/φ(θ) is non-decreasing. The power-law choice φ(θ) = η θ^{−γ} with γ ∈ (0, 1/2] and η > 0 satisfies these conditions under the same constraints on ρ, and the Heston-compatible choice is φ(θ) = (1 − (1 − e^{−κθ})/(κθ)) / (κθ/2).

The separation of fitting from arbitrage-consistency is the core methodological contribution of Gatheral–Jacquier: prior practice often calibrated the raw SVI parameters to market mid-quotes via least-squares minimization and then checked ex post whether the calibrated surface was arbitrage-free, which frequently failed and required ad hoc parameter adjustments that degraded the fit. The SSVI sub-parameterization inverts this order by making arbitrage-freedom a structural property of the parameterization, so that any calibrated (θ, φ, ρ) satisfying the explicit inequality constraints produces an automatically consistent surface, and the fitting problem is reduced to finding the best-fitting member of the arbitrage-free family rather than fitting a general five-parameter model and then projecting onto the arbitrage-free subset. The power-law and Heston-compatible φ(θ) choices represent different priors on the term structure of skew. The power-law choice gives a parametric family that can fit the empirically observed T^{−H} decay of ATM skew consistent with rough volatility, while the Heston-compatible choice anchors the surface shape to what Heston dynamics would produce at each ATM variance level, providing a natural benchmark for identifying surface dislocations relative to the model. For the desk, the Heston-compatible SSVI is used as the baseline fit against which deviations are measured and reported as conditioning inputs, while the power-law SSVI is used for the rough-vol calibration where the skew term structure is itself a quantity of interest. Practical calibration stability of SSVI versus raw SVI is substantially better: the SSVI parameter space is lower-dimensional (three parameters rather than five) and the constraints are explicit inequalities rather than implicit conditions, making constrained optimization more tractable and the resulting calibration less sensitive to initialization and numerical precision. This stability is particularly important for the desk because calibration is performed daily and any instability in the surface fit introduces noise into the surface features used as conditioning inputs.

The desk methodological prescription from this paper is unambiguous: any surface feature — ATM skew level, term structure slope, butterfly spread, wing asymmetry — that is used as a conditioning input for signal generation or risk analysis must be extracted from an arbitrage-free surface fit, not from raw market data or from a surface calibrated without explicit arbitrage constraints. Using raw market mid-quotes as inputs introduces bid-ask noise, expiry-specific liquidity premia, and potential static-arbitrage violations that contaminate the signal, while an SSVI fit provides a smooth, consistent representation that extracts the structural information in the surface while discarding the noise. Specifically, skew as a conditioning input is meaningful only if the surface satisfies butterfly-freedom (otherwise the skew corresponds to a negative-density region of the smile that is not a valid risk-neutral distribution), and term-structure inversion is meaningful only if calendar-spread-freedom is imposed (otherwise the inversion may be a calibration artifact rather than a genuine market signal). The desk's signal-generation pipeline enforces SSVI arbitrage-free fitting as a pre-processing step before any surface feature extraction, and fits that fail the inequality constraints are flagged for manual review before being used in conditioning.

Single-name surface calibration below liquidity thresholds is the most common failure mode in practice: for names with fewer than twenty to thirty liquid strikes per expiry, the SSVI fit is underdetermined and the calibrated φ(θ) is highly sensitive to which strikes are included, producing parameter estimates that are unstable across small perturbations of the input data. The desk applies a minimum-liquidity filter before attempting SSVI calibration on single names and uses a simpler one-parameter smile approximation for illiquid names. Parameter instability in raw SVI under perturbation of bid-ask midpoints is well-documented and motivates the move to SSVI, but even SSVI parameters can exhibit regime-dependent instability: during stress periods when the smile shape changes rapidly across consecutive trading days, the smooth parameterization may underfit the actual market-observed kinks and mode shifts, particularly in the near-term wings where liquidity can temporarily evaporate. The tension between parsimony and expressivity is fundamental: SSVI with three parameters imposes a strict functional form on the smile that may not be flexible enough to fit multi-modal or kinked smiles observed during large-move events, and in those cases the desk may need to temporarily switch to a more flexible but less constrained fitting approach, accepting the risk of calibration instability in exchange for a better representation of the observed market.