The Day Volatility Broke: 5 February 2018

The instrument: a fixed multiple, reset every day

The VIX is not a price that can be bought. It is a calculated number — the market’s expected thirty-day volatility of the S&P 500, derived from the prices of a strip of index options — so anyone who wants to trade it has to trade something else, usually VIX futures or options on them. That gap is where the trouble starts. A **VIX futures** contract settles to a future value of the index, so a fund holding it is not holding the index; it is holding a dated expectation, and when that contract approaches settlement the fund has to sell it and buy the next one out. If the futures curve is upward-sloping — **contango**, the normal state, because longer-dated uncertainty is usually priced higher than near-dated — each roll sells a cheaper contract and buys a more expensive one, which is a cost that recurs every month. That means a long-volatility fund loses money when nothing happens: the same mechanism as theta, restated in futures. The inverse of that is why short-volatility products were so popular for years, and why they looked so steady — the seller of that roll cost was collecting a rent that arrived almost every month. The second layer is the **daily reset**. A note that promises “−1× the daily return of the index” can only keep that promise by adjusting its exposure every day to the new level, because the compounding of a fixed daily multiple does not equal a fixed multiple over any longer period. That clause is what turns the position into something that grows as it loses. Suppose the index rises 5% on a day: a −1× note loses 5%, and its assets are now smaller, so the notional it must be short has fallen by more than the index rose — which means it has to *buy back* exposure. If the index rises again the next day, the same arithmetic requires more buying. The exposure chases the price in the same direction, which is precisely the definition of a positive-feedback strategy, and it is written into the prospectus rather than chosen by a trader. A −3× product has the same property with a three-fold multiplier, and a single-day index move of one-third wipes it out entirely — a number worth carrying because it converts “high risk” into an inequality that can be checked in advance. Why a daily reset is not a static short — Index +5% today: A −1× note loses 5% and must buy back some exposure to restore its target · Index +5% again tomorrow: The same rule requires another purchase, into a market that is rising ← · A −3× note: A 33.3% single-day index rise wipes it to zero — arithmetic, not scenario · A long-volatility fund in contango: Pays a monthly roll cost, so it bleeds whenever nothing happens The exposure chases the price. A rule-driven position whose size is set by price becomes part of the price.

The day itself, and the general rule it teaches

On 5 February 2018 the VIX rose from 17.31 to 37.32 — a 115.6% single-session increase — and the arithmetic of the reset clause did the rest. An inverse note delivering minus one times that day’s index return lost more than its entire value, and because a note cannot be worth less than zero, it was terminated: $10,000 became $0. The products that had spent years collecting the roll cost were gone in an evening, and the pattern is the one every short-volatility structure shares — the equity curve is a series of small steady gains punctuated by the loss of everything that came before. What made this case extreme was the *mechanism*, not just the payoff. Because the notes measured their exposure from the previous close, their required notional had to be restored at the close of the very day the index had risen, which meant buying volatility futures into a market already spiking. Around $600 million at $22,500 of notional per contract is roughly 26,700 contracts — some 8.9% of the day’s entire volume — arriving in a single closing window. Buying a tenth of a day’s volume in minutes lifts the price, and a higher close mechanically required *more* buying the following session. The feedback loop was created by an accounting rule, and no participant had to be wrong for it to happen. The case generalises past volatility products, and the generalisation is the most useful thing in the lesson. Whenever a position’s size is determined by a rule that responds to price — a daily reset, a stop-loss cascade, a margin call, an index rebalance, a risk-parity target, a covered-call overwrite that must sell calls when volatility rises — the response becomes a component of the market it is responding to. The question to ask before taking the other side of such a flow is therefore not “am I right about volatility?” but **“who is forced to trade if I am right, and in which direction?”** That question is answerable in advance for every one of the structures in this subject. A dealer short gamma is forced to buy as the market rises. A leveraged fund is forced to sell as its assets fall. An inverse volatility note is forced to buy as volatility rises. In each case, the size of the forced flow is estimable from public filings and methodology documents, and in each case the flow *adds* to the move that triggers it. That is how a learner moves from predicting a price to mapping a market, and it is the last piece of market structure this subject owes: not what will happen, but who will be made to act. • The VIX is a calculated expectation, so it is traded through futures and options rather than held directly. • Contango makes a long-volatility product bleed monthly — theta restated in futures. • A daily reset makes the exposure chase the price, which is why the arithmetic is compounding rather than static. • A −3× product is wiped out by a 33.3% single-day move in its index. • On 5 February 2018 the VIX rose 115.6% and inverse notes were terminated at zero; their required rebalance was about 8.9% of the day’s volume. • The general question: who is forced to trade if you are right, and in which direction? The failure mode of this case study is retelling it as a story about greedy retail traders. The instruments were used by institutions, the losses were concentrated in products with written reset clauses, and the mechanism was a rule rather than a sentiment — which is exactly why it can happen again in a different wrapper.

Two prices for the same index

The VIX close is not the number a VIX derivative settles to. Futures and options on the index expire on a fixed calendar — the Wednesday thirty days before the third Friday of the following month’s S&P 500 options — and their final value is a **special opening quotation (SOQ)**: a VIX calculated at the open on that Wednesday from the prices of one strip of SPX options. The index prints a continuous close every afternoon; the contract has a single settlement print at a single moment. The two are different numbers on any day, not merely on a violent one, and only one of them is what the contract pays. The reason is that a settlement has to be defined and hard to move, so it is computed from a stated basket over a stated window rather than from a continuous market. That makes it exact, and it makes it *other*. After 5 February 2018 the distinction stopped being academic: the front-month contract that looked like it was worth the 37.32 close settled days later to an SOQ well below it, because volatility had fallen back and because the settlement was struck from the option strip at a single open. A participant who was right about the panic — right about direction, right about the regime — could still be wrong about the number the contract paid. The same structure exists without any volatility index, which is what makes it worth learning. S&P 500 options come in two settlement flavours: the third-Friday series (SPX) settles to an **opening** quotation on its Friday, and the weekday series (SPXW) settles to the **closing** price that afternoon. Two contracts on the same index, exercisable at the same strike, pay out on different prices at different times of day, and the market prices them accordingly. A learner who watches only the index close will misprice both. What follows is a rule about definition rather than a rule about direction: before taking a derivative position, find out what it settles to, when, and from which basket, because that definition — not the price on the screen — is what the contract pays. In this case study the inverse note was terminated against the index close on the day the reset fired, while the futures that would have hedged it settled to an SOQ days later. Two instruments that track the same index and share its name are not the same as two instruments that pay the same number, and the gap between them is a risk the position carries whether or not anyone noticed it. • The VIX close is not what a VIX derivative settles to — futures settle to a special opening quotation. • SPX third-Friday options settle to the opening quote; SPXW weekday options settle to the close. • Read the settlement definition before taking the position it prices. A hedge is only a hedge against the thing it settles to. Owning a derivative whose settlement definition you have not read is owning an unmeasured basis, and it will be measured for you on settlement day.

Sources

• The 5 February 2018 VIX spike, inverse-volatility ETP terminations and the closing rebalanceSEC and CFTC staff reports on the February 2018 volatility event
• VIX futures term structure, roll yield and the decay of long-volatility productsCboe VIX futures and ETP methodology documentation
• Daily-reset leverage: why a fixed multiple of a daily return is not a static short positionCheng & Madhavan, “The Dynamics of Leveraged and Inverse Exchange-Traded Funds” (Journal of Investment Management)