The 2003 Rewrite

In 2003 Cboe, working with Goldman Sachs, replaced the ATM OEX construction with a formula that aggregates a strip of SPX option prices weighted by the inverse square of strike. The public explanation was that the S&P 500 had become the core U.S. equity index. The deeper explanation was that variance had become the OTC claim the listed market wanted to shadow.

The replication argument

A continuously monitored variance swap can be replicated, in a frictionless market with a continuum of strikes and no jumps, by a static portfolio of out-of-the-money calls and puts plus a dynamic position in the underlying that keeps the portfolio aligned with the log contract. The value of that portfolio is independent of the volatility path and of the level of the underlying; it isolates quadratic variation. Taking the square root and scaling to a 30-day year-fraction produces a volatility-like index number.

The listed market does not have a continuum of strikes, does not have zero jumps, and does not trade a pure log contract. The VIX methodology is therefore an estimator: it sums midquotes of listed SPX puts and calls, excludes zero-bid wings, interpolates between two expiries that bracket 30 days (including Friday weeklys that fall inside the 23–37 day window), and annualizes. The published number is 100 times the square root of that estimated 30-day variance.

What changed in the object

Three substitutions define the rewrite. The underlier of the options moved from OEX to SPX. The estimator moved from model-dependent ATM implied vol to a model-light strip of OTM option prices. The economic story moved from “average implied volatility” to “square root of expected 30-day variance under the risk-neutral measure, as inferred from a replicating portfolio.”

The third substitution is the one that licensed a futures market. A futures contract on an abstract ATM average is difficult to hedge. A futures contract on a quantity that is itself the price of a portfolio of listed options can, in principle, be hedged in those options. Cboe’s later settlement design—the Special Opening Quotation—makes that hedge operational on expiration morning by opening the relevant SPX series in an auction whose prints feed the final settlement value.

What the rewrite did not fix

1. Jumps. The log-contract replication for variance is exact for continuous semimartingales. Equity indexes jump. The VIX is therefore not a pure claim on quadratic variation.
2. Discrete strikes and the wing cutoff. Zero-bid exclusion truncates tail mass precisely when tail mass is expensive.
3. Midquote convention. The live index uses mids, not executable sizes. The SOQ uses opening trade prices. Live VIX and settlement VIX are different legal quantities.
4. The square-root transform. Markets that want variance, not volatility, must undo the transform and accept convexity.
5. A single tenor. Thirty days is a convention, not a sufficient statistic of the surface.

Family expansion after 2003

Once the strip methodology existed, Cboe cloned it across tenors and underliers: VIX9D, VIX3M, VIX6M, VIX1Y, VIX1D, VVIX (the same estimator applied to VIX options), and a roster of asset-class volatility indexes. SKEW, which is not a VIX clone, extracts a tail-shape statistic from the same SPX surface. The existence of this family is evidence that the market already knew a single 30-day second moment was incomplete. It is also evidence that Cboe’s legal strategy was proliferation under one methodology and one mark family rather than replacement of the flagship ticker.

Cboe’s February 2026 methodology booklet (version 6.0) remains the operative public specification as of this writing. A successor that cannot point to a comparably auditable booklet will not clear a DCM or NMS listing review.