Part II · Forms
Chapter IV
The Mathematical Form
Estimators, measures, and what a single number cannot carry
A working statement of the VIX estimator, its relation to the risk-neutral second moment, the term structure, convexity, and the information the index discards.
Graduate work on the VIX begins by naming the functional. The published index is a particular map of a particular option surface. Name the map, or you cannot say what a successor conserves. Versioned I_t is that naming, extended: a density print, a realized clock, or dual F, each a published member.
The risk-neutral second moment
Let Q be a risk-neutral measure consistent with the observed SPX option surface at a fixed expiry T. Let F be the T-forward on the index. Breeden–Litzenberger identifies the risk-neutral density of S_T from the second derivative of the call price with respect to strike. Integrating K^{-2} against OTM option prices recovers the price of the log contract and therefore a model-light estimate of E^Q[∫_0^T (dS_u / S_u)^2] when paths are continuous. That expectation is the variance-swap rate. The VIX is one hundred times the square root of a 30-day interpolation of two such rates.
−(2/T) E^Q[log(S_T / F)] = (2/T) ∫_0^∞ (Q(K) / K²) dK
which equals expected quadratic variation of log S on [0, T].
With jumps the identity prices a specific combination of quadratic variation and a jump compensator. The listed VIX does not correct for that compensator.
Discretization choices that are legal facts
Every operational choice in the Cboe booklet is a legal fact because settlement uses those choices, not the integral. The strike grid is the listed SPX grid. The contribution of each strike is ΔK / K². The option price is a midquote for the live index and an opening print for the SOQ. Series with a zero bid are dropped, which endogenously shortens the wing when liquidity withdraws. Only Friday-expiring SPX options inside a 23-to-37 calendar-day window enter the flagship VIX; other tenors have their own windows. Interest-rate inputs are specified by the administrator. These are not “implementation details.” They are the definition.
A replacement methodology that changes the wing rule, the midquote rule, or the expiry window has changed the claim, even if it keeps the 100√(·) display. Counsel should treat methodology amendments the way they treat index transition events in equity derivatives: as potential market-disruption and fallback problems.
Term structure and basis
The cash VIX is a 30-day constant-maturity statistic. VIX futures are claims on the then-prevailing cash VIX at listed expiries. The futures curve is therefore a curve of expected future 30-day variance rates (under a futures-implied measure, with risk premia), not a curve of 60-day or 90-day variance rates. VIX3M is the closer cousin of a three-month variance rate. Confusing the VX curve with the variance term structure is a standard error in both trading books and offering circulars.
In ordinary time the VX curve sits in contango: longer futures print above spot VIX. A long-only short-dated futures product then bleeds through roll. In stress the curve inverts (backwardation), short-dated futures rally toward or through spot, and the same product makes money until the roll normalizes. This path dependence is not a defect of “the VIX.” It is the geometry of a futures curve on a mean-reverting, uninvestable cash index.
Convexity, VVIX, and vol of vol
Because the listed products are written on √variance rather than on variance, a movement in the variance rate does not map linearly into VIX points. Options on VIX exist because the market wants convexity on that already-convex transform. VVIX applies the VIX estimator to the VIX option surface and is therefore a second-moment statistic of the VIX itself—vol of vol in the same compressed sense. A full-distribution view would replace both VIX and VVIX with explicit functionals of the joint law of (S, σ) or with a documented set of moments and tail probabilities.
What is discarded
- Skew and kurtosis of the 30-day risk-neutral law (partly recovered by SKEW, not by VIX).
- Pathwise vol-of-vol and rough-vol structure.
- Overnight versus regular-hours quadratic variation.
- Single-name and sector variance that does not survive index diversification.
- Realized-versus-implied premia; VIX is a Q-measure object, not a forecast of subsequent realized vol.
- Liquidity state of the wing, except insofar as zero bids drop out.
Variance risk premium
The difference between the VIX-style Q-variance rate and subsequent realized variance is the variance risk premium. It is usually positive: option markets price more variance than is later realized. That premium is why short-vol strategies have a long-run drift and why they die in clusters. A legal analysis of “VIX products as hedges” that does not mention this premium is incomplete. A hedge that collects the premium is an insurance underwriting book. A hedge that pays the premium is insurance. Most retail VIX products do neither cleanly, because they hold futures, not variance swaps, and because they roll.
Problem IV.1
Show, using Jensen’s inequality, why the fair strike of a volatility swap is less than or equal to the square root of the fair strike of a variance swap. Then explain in two sentences why a note that “tracks the VIX” is tracking neither swap.
Hint. VIX is a cash index; listed products track VX futures or a futures index.
Problem IV.2
A methodology amendment drops all strikes more than 20% from spot. Who is harmed in a crash, and which document—index booklet, futures rulebook, or ETN supplement—governs the claim?