What Is eSSVI?

What eSSVI Is

The implied-volatility surface is the function that maps every (strike, expiration) pair to an IV value. Real-market surfaces have three regularities that any surface fit must respect: (1) per-expiration smile shape, (2) term structure across expirations, (3) absence of arbitrage (calendar arbitrage between two tenors and butterfly arbitrage across strikes). The eSSVI parametrization (Gatheral and Jacquier 2014) satisfies all three with a five-parameter functional form.

The eSSVI total variance function is:

w(k, theta) = theta/2 · (1 + rho · phi(theta) · k + sqrt((phi(theta) · k + rho)^2 + 1 - rho^2))

where k is log-moneyness (log(K/F)), theta is the at-the-money total variance, phi(theta) is a parametric function of ATM variance controlling smile curvature, and rho is the smile asymmetry (correlation-like parameter). Two key extensions over the original SSVI (Gatheral 2004) make eSSVI fit better:

• Multiple expiration parameters. phi(theta) allows the smile width to vary across tenors flexibly rather than scaling rigidly with sqrt(theta).
• Calendar arbitrage constraints. The parameter set is constrained to ensure that variance grows monotonically with tenor at every strike (no calendar arbitrage).

Why eSSVI Matters

Three reasons eSSVI is the institutional choice for full-surface fitting:

• Arbitrage-free under parameter constraints. Naive surface fits (cubic splines, polynomial regressions, kernel smoothers) routinely produce surfaces with internal arbitrages: butterflies that imply negative probability mass, or calendars where short-tenor vol exceeds long-tenor. Trading on such surfaces leads to systematic losses against any market-maker who actually trades the wing or calendar. With its calibration constraints satisfied, eSSVI rules out static butterfly and calendar arbitrage on the fitted surface.
• Smooth across strike and tenor. The five-parameter form produces analytically smooth surfaces. This is essential for downstream analytics (RND extraction by twice-differentiating the price function, Greek calculations, no-arbitrage interpolation between listed expirations).
• Compact representation. Five parameters per surface is small enough to fit reliably even with sparse data (e.g., single-name names with 20-50 listed contracts) and to compare across days, names, and regimes.

How eSSVI Is Calibrated

Calibration finds the parameter values (theta_grid, phi_params, rho) that minimize the squared distance between observed and modeled implied volatilities across all listed (strike, expiration) pairs, subject to no-arbitrage constraints. Practical fitting:

1. Pre-clean the surface. Remove illiquid contracts (zero OI, wide bid-ask), apply put-call parity to derive consistent IVs from both sides, normalize to forward-moneyness.
2. Fit per-expiration ATM theta. For each listed expiration, fit theta = (ATM IV)^2 · T directly from the ATM strike(s).
3. Fit smile parameters. Across the full surface, jointly minimize squared residuals of (k_i, T_j) IVs against the eSSVI functional form, with penalty terms for arbitrage violations.
4. Validate. Check butterfly and calendar arbitrage on a dense grid; check that the implied call price function is monotonic in strike and convex.

Daily refits typically converge in seconds for index surfaces and 10-30 seconds for full single-name names. Production systems cache surface parameters and refit incrementally rather than from scratch.

Worked Example

SPY surface on a representative date. Calibrated eSSVI parameters:

• theta(30d) = 0.0027 (corresponding to ATM 30-day IV = 14.4%)
• theta(60d) = 0.0061 (ATM 60-day IV = 14.5%)
• theta(90d) = 0.0099 (ATM 90-day IV = 15.0%)
• phi parameters fitting smile widening with tenor
• rho = -0.65 (asymmetry: put-side IV elevated relative to call-side)

This compact parameter set reproduces the entire SPY IV surface across hundreds of listed contracts with average residual ~30 basis points. When the eSSVI no-arbitrage parameter constraints are enforced and the calibration validator confirms no butterfly or calendar arbitrage on a dense check grid, downstream analytics derived from the surface (RND, expected move, model-divergence aggregation) inherit the same property.

How eSSVI Compares to Alternatives

• vs SVI (Gatheral 2004): SVI is the per-expiration ancestor. eSSVI extends it to a full surface with calendar-arbitrage constraints. SVI fits each smile independently; eSSVI fits the whole surface jointly with cross-tenor consistency.
• vs SABR: SABR is also a per-expiration smile model with closed-form (Hagan) approximation. SABR is widely used for interest-rate options; eSSVI is preferred for equity-index surfaces because it handles the full surface natively without per-expiration recalibration.
• vs Heston: Heston is a stochastic-volatility process model that produces an IV surface as output. eSSVI is a direct surface parametrization that does not assume any underlying stochastic process. eSSVI fits the observed surface tightly; Heston fits the surface dynamics across regimes.
• vs local volatility: Dupire local-vol fits any surface exactly by construction. eSSVI fits with five parameters and a small residual, but the parametric form constrains the surface to a tractable family. Trade-off: local-vol gives exact fit but unrealistic forward-smile dynamics; eSSVI gives near-fit with stable forward-smile evolution.

Operational Use

• Risk-neutral density extraction. Twice-differentiating eSSVI synthetic call prices produces clean, smooth, arbitrage-free RNDs that can be compared across dates, names, and tenors.
• Expected move and probability cones. The forward-smile dynamics implied by eSSVI feed expected-move calculators that respect cross-tenor consistency.
• Model-divergence baseline. eSSVI provides a market-fit reference against which structural models (Heston, SABR, jump-diffusion) can be compared. Differences between calibrated structural-model IVs and eSSVI IVs are the model-divergence signal at the surface level.
• Synthetic strike interpolation. When traders want a price on a strike that is not listed (e.g., weekly options on a single name with sparse listings), eSSVI gives a no-arbitrage synthetic IV that can be priced into a model.

Limitations

• Parametric. eSSVI assumes the surface follows the five-parameter form. Real surfaces sometimes deviate (single-name names with very thin listings, illiquid strikes). The fit residual at problematic points can be 50-100 basis points.
• Static. eSSVI fits a snapshot. Capturing surface dynamics (how rho moves through regimes, how smile term-decay evolves) requires either re-fitting daily or pairing eSSVI with a dynamic model.
• Single-asset. eSSVI is a single-underlying parametrization. Cross-asset surface modeling (e.g., joint SPX-VIX surface) requires more elaborate frameworks.

Related Concepts

Volatility Smile · Volatility Skew · Term Structure · SABR Model · Heston Model · Local Volatility · Risk-Neutral Density · Pricing Model Landscape

References & Further Reading

• Gatheral, J. and Jacquier, A. (2014). "Arbitrage-Free SVI Volatility Surfaces." Quantitative Finance, 14(1), 59-71. The eSSVI paper.
• Gatheral, J. (2004). "A Parsimonious Arbitrage-Free Implied Volatility Parameterization with Application to the Valuation of Volatility Derivatives." Presentation at Global Derivatives. The original SVI.
• Roper, M. (2010). "Arbitrage Free Implied Volatility Surfaces." Working paper, University of Sydney. Cross-tenor arbitrage constraints.
• Gatheral, J. (2006). The Volatility Surface: A Practitioner's Guide. Wiley. Practitioner background on volatility-surface modeling and SVI's market context.

View the live calibrated eSSVI surface for SPY ->

This page is part of the Pricing Model Landscape and the canonical reference set on options market structure. Browse all documentation.