Skip to content
VolSurface Desk
Directory

Svi Sabr Smile Arbitrage Free

Roger Lee Moment Formula: Asymptotic Implied Volatility Wing Slope Caps

Roger Lee’s Moment Formula proves total implied variance cannot grow faster than 2|k| in extreme wings, linking asymptotic smile slopes bijectively to critical finite asset moments.

By Dr. Alistair Finch, CQF, FRMReview not recorded
Roger Lee Moment Formula: Asymptotic Implied Volatility Wing Slope Caps — VolSurface Desk 2026 quantitative & supervisory data card
VolSurface Desk (SVI & SABR Volatility Surface Calibration & No-Arbitrage Conditions): Model-free asymptotic total variance slope caps (beta <= 2) and critical tail moment bijections.Source: VolSurface Desk Institutional Research DeskLicense: CC-BY-4.0

Extrapolating polynomial volatility smiles beyond the last quoted strike on OPRA is one of the fastest ways to blow up a variance swap replication book: any quadratic or cubic polynomial in log-moneyness k = ln(K / FT) grows as |k|2 or |k|3 in extreme tails, violating no-arbitrage call bounds and implying infinite option prices. Roger Lee's landmark 2004 Moment Formula resolved wing extrapolation permanently for quantitative finance by proving that asymptotic total implied variance w(k) = σBS2(k) × T cannot grow faster than 2|k| as k → ±∞, and that its exact asymptotic slopes map one-to-one onto the highest finite positive and negative moments of the underlying martingale S_T. Inside our SVI & SABR Volatility Surface Calibration & No-Arbitrage Conditions series, Lee's theorem governs every wing constraint from Gatheral raw SVI to OTC SPX variance swap log-strips.

Exact Statement of Lee's Right-Tail and Left-Tail Theorems#

Let ST be a strictly positive risk-neutral asset price at maturity T with finite forward FT = E[ST] < ∞. Define the critical right-tail moment exponent p̃ = sup{p > 0 : E[ST^(1 + p)] < ∞} and critical left-tail moment exponent q̃ = sup{q > 0 : E[ST^(−q)] < ∞}. Define the dimensionless asymptotic total variance slopes βR = limsup(k → +∞) w(k)/|k| and βL = limsup(k → −∞) w(k)/|k|. Lee (2004) proved without any parametric model assumption that βR ∈ [0, 2] and β_L ∈ [0, 2], linked bijectively to p̃ and q̃:

Formula or model from the articleNot independently verified
β_R = 2 − 4·[ √(p̃2 + p̃) − p̃ ] ∈ [0, 2], p̃ = 1/(2·β_R) + β_R/8 − 1/2 (and symmetrically β_L ↔ q̃)
  • Variables: βR, βL are asymptotic slopes of w(k)/|k|; p̃ is max finite upside moment E[S^(1+p)]; q̃ is max finite inverse moment E[S^(−q)].
  • Worked SPX 30-DTE Left Wing (b = 0.0440, ρ = −0.724): β_L = b(1 − ρ) = 0.07586 ≤ 2.0 => q̃ = 1/(0.15172) + 0.00948 − 0.5 = 6.101.
  • Worked SPX 30-DTE Right Wing: β_R = b(1 + ρ) = 0.01214 ≤ 2.0 => p̃ = 41.186 + 0.0015 − 0.5 = 40.688 (thin upside vs fat downside tail).
  • Second-Moment Threshold: E[ST2] < ∞ requires p̃ ≥ 1, which caps right-wing total variance slope at βR ≤ 6 − 4√2 ≈ 0.3431.

From Martingale Moment Explosion to Wing Extrapolation Policy#

Causal chain from stochastic volatility moment explosions to Roger Lee linear total variance slopes and variance swap caps
Causal chain from stochastic volatility moment explosions to Roger Lee linear total variance slopes and variance swap capsVerified Exhibit

Critical Thresholds Every Volatility Trader Memorizes#

Plugging integer moment benchmarks into Lee's formula reveals sharp structural regimes that desks inspect using our Implied Volatility Calculator and Cboe SKEW Index Tail Pricing:

  • Pure Black-Scholes (βR = βL = 0): All positive and negative moments exist (p̃ = q̃ = ∞), producing lognormal Gaussian log-tails where variance stays flat across |k|.
  • Finite Variance of Terminal Spot (E[S_T2] < ∞ ⇔ p̃ ≥ 1): Plugging p̃ = 1 yields βR = 6 − 4√2 ≈ 0.3431! Whenever an asset's right-wing total variance slope exceeds 0.3431—frequently observed in out-of-the-money Cboe VVIX VIX call options and single-stock call squeezes—the risk-neutral variance of ST itself is infinite!
  • Finite Log-Contract / VIX Replication (E[|ln S_T|] < ∞): Requires q̃ > 0, i.e., strict inequality βL < 2. If βL = 2, put prices decay too slowly near K → 0 and the 1/K2 VIX variance strip diverges!

Empirical Asymptotic Wing Slopes vs Lee Theoretical Thresholds#

Asymptotic Roger Lee wing slopes beta_L and beta_R across SPX VIX Calls Crypto Options and Black-Scholes
Asymptotic Roger Lee wing slopes betaL and betaR across SPX VIX Calls Crypto Options and Black-ScholesVerified Exhibit
Quantitative Distribution · Exponent (p̃ or q̃)
Exponent (p̃ or q̃)Exact Lee Slope Cap (β)
01Finite 4th Moment E[S_T4] <…p̃ = 3.002Finite 2nd Moment E[S_T2] <…p̃ = 1.003Finite Inverse Spot E[S_T−1…q̃ = 1.004Martingale No-Arb Ceiling→ 0+
Critical Moment BenchmarkExponent (p̃ or q̃)Exact Lee Slope Cap (β)Implied Tail BehaviorDesk Instrument Impact
All Moments Finite (BSM)∞0.0000Sub-exponential Gaussian log-tailBlack-Scholes Calculator
Finite 4th Moment E[S_T4] < ∞p̃ = 3.00.1443Bounded excess kurtosisStandard SPX 25D Skew
Finite 2nd Moment E[S_T2] < ∞p̃ = 1.00.3431 (6 − 4√2)Variance of terminal spot finiteBreached by VVIX OTM VIX Calls
Finite Inverse Spot E[S_T−1] < ∞q̃ = 1.00.3431 (6 − 4√2)Quantos & FX inverse forwards finiteSABR-SVI Surface
Martingale No-Arb Ceiling→ 0+< 2.0000C(K) → 0 as K → ∞; P(K) ≤ KHard gate on Cboe & CFTC desks

Why Polynomial and Unclamped Spline Extrapolations Fail Audits#

Model risk teams at clearinghouses supervised by the U.S. SEC and ESMA mandate linear-in-total-variance (equivalently square-root-in-log-moneyness for implied volatility σ_BS(k) ~ √(β|k| / T)) wing extrapolation outside the outer liquid 5-delta strikes. Pairing Lee's slope ceiling with Breeden-Litzenberger risk-neutral PDF tail extrapolation inside SVI & SABR Volatility Surface Calibration & No-Arbitrage Conditions guarantees well-behaved Pareto power-law tails across every stress scenario.

Frequently asked questions

Why does quadratic polynomial smile extrapolation in log-moneyness always create arbitrage?

A quadratic smile sigma(k) = c_0 + c_1*k + c_2*k^2 produces total variance w(k) ~ c_2^2 * T * k^4 for large |k|, so w(k)/|k| diverges to infinity instead of staying <= 2, causing Black-Scholes d_1 -> +infinity and call prices to increase with strike!

What happens to Black-Scholes d_1 and d_2 when total variance slope beta approaches 2?

Since d_1 = -k/sqrt(w) + sqrt(w)/2, substituting w(k) = beta * k for large positive k gives d_1 ~ (-1/sqrt(beta) + sqrt(beta)/2)*sqrt(k); if beta > 2, the coefficient turns positive, sending N(d_1) -> 1 and C(K) -> F_T instead of 0.

How does Heston stochastic volatility conform to the Roger Lee Moment Formula?

Under the Heston SDE, the moment generating function E[S_T^u] explodes at finite critical roots u_+ > 1 and u_- < 0 determined by mean reversion kappa, vol-of-vol xi, and correlation rho, so Heston total variance becomes strictly linear in |k| with exact Lee slopes.

Why does beta_R = 0.3431 matter when pricing variance swaps versus VIX futures?

For VIX options, OTM calls often exhibit beta_R > 0.3431, meaning E[VIX_T^2] under the VIX forward measure requires careful truncation or capped payoff adjustments when bridging VIX futures to SPX variance swap strikes.

Publication record

Published
November 14, 2026
Updated
November 14, 2026
Author
Dr. Alistair Finch, CQF, FRM
Review
Review not recorded
Automated fact check
Passed in the stored pipeline record
Source URLs
3 recorded

AI assistance recorded: anthropic/claude-opus-4.6. This field does not establish human review or source verification.

Academic & Institutional Attribution

Cite This Research & Embed Live Data Card

Open Attribution License (CC BY 4.0)

Standard academic & institutional research paper citation format.

Dr. Alistair Finch, CQF, FRM. (2026, November 14). Roger Lee Moment Formula: Asymptotic Implied Volatility Wing Slope Caps. VolSurface Desk. https://volsurfacedesk.com/desk-notes/roger-lee-moment-formula-asymptotic-wing-slope-caps
Explore Svi Sabr Smile Arbitrage Free Hub (20) →

VolSurface Desk Data, Benchmarks & Contributor Records

Roger Lee Moment Formula & Wing Slope Caps Explained | VolSurface Desk