A market event with several clocks
On May 6, 2010, prices in U.S. equity-linked markets fell and rebounded extraordinarily quickly. The September 30 joint SEC–CFTC staff report describes major indices, already more than 4% below the previous close, falling another 5–6% within minutes. More than 20,000 trades in over 300 securities occurred over 60% away from recently prevailing prices. The broad indices ended the day down about 3%, so recovery from the most extreme dislocation did not mean an unchanged daily close. Exchanges and FINRA later agreed to cancel trades at least 60% away from their 2:40 p.m. prices; extreme reported prints were therefore not all final investor outcomes. [1]
Three clocks matter: the deterioration before the abrupt break, the minute-by-minute transfer of selling across instruments, and the much longer legal and regulatory aftermath. Combining them into a single story about one erroneous order or one convicted trader loses important distinctions. This article treats the staff reconstruction as evidence about market functioning, not a judicial determination of responsibility.
A large hedge entered a thinning market
The staff report identifies a mutual-fund complex’s program to sell 75,000 E-mini S&P 500 futures, approximately $4.1 billion, beginning at 2:32 p.m. Eastern. Its algorithm targeted 9% of the previous minute’s trading volume without a price or time constraint. At 2:45:28, CME Stop Logic paused trading for five seconds; buying increased and prices subsequently recovered even while the program continued. [1]
The mechanism is understandable without assuming a software malfunction. A volume-participation algorithm can interpret rapid trading as capacity to execute. But repeated transfers among intermediaries can raise measured volume without adding investors willing to carry the ultimate exposure. Under those conditions, the execution schedule responds to activity rather than durable risk-bearing demand. That is an analytical explanation of the feedback, not a claim that every such algorithm produces a crash.
Why futures and individual stocks behaved differently
In an October 2010 account, SEC staff member Gregg Berman emphasized that E-mini futures and SPY lost roughly 5% in five minutes and recovered during the following ten. Some individual-stock and ETF prices reached extreme levels during that recovery. He also cautioned that the report was not a general verdict on high-frequency trading. The speech expresses his views rather than a Commission position. [2]
Arbitrage links prices across a stock basket, its ETF and futures on the basket. When one becomes relatively cheap, a trader can buy it while selling an economically similar exposure elsewhere. This normally narrows discrepancies, but it can also transmit an urgent desire to shed risk. If reliable quotes disappear in one leg, the apparent arbitrage becomes harder to execute and remaining providers face greater uncertainty. A common underlying market does not guarantee identical recovery times across every instrument.
Fast turnover was not the same as a buyer of last resort
Kirilenko, Kyle, Samadi and Tuzun’s May 2014 research version, hosted by the CFTC, argues that the high-frequency traders studied did not initiate the crash but amplified the price movement. Their account stresses demand for immediate execution, limited inventories and rapid recycling of contracts. It is an empirical research interpretation of a particular futures market, not proof that every automated trader followed one strategy. [3]
The distinction between a trigger and an amplifier matters. A participant can absorb early selling and later compete to reduce exposure. Gross purchases can be large while net inventory hardly changes. This explains how a tape full of transactions can coexist with a shortage of counterparties willing to retain risk. It also makes a simple count of messages, trades or cancellations an incomplete measure of economic .
The Sarao prosecution is a separate evidentiary record
On November 9, 2016, the Justice Department announced Navinder Singh Sarao’s guilty plea to wire fraud and spoofing. He admitted placing orders he did not intend to execute to create false supply or demand; the plea covered a scheme lasting more than five years. DOJ said that on May 6 he entered at least 85 spoof sell orders that at times represented more than 20% of visible E-mini sell orders. [4]
That record establishes admitted manipulative conduct, rather than merely an unresolved accusation. It does not turn the September 2010 staff report into a prosecution of Sarao, or by itself quantify what prices would have done without his conduct. Nor is an authentic institutional hedge the same conduct as an order entered with no intention of execution. The market reconstruction and the plea can both be informative without being interchangeable causal tests.
Safeguards changed the route through a price shock
The SEC announced in June 2012 that it had approved a limit-up/limit-down mechanism and changes to market-wide circuit breakers. The former was designed to prevent individual-stock executions outside specified price bands, replacing the single-stock circuit-breaker pilot introduced after May 2010. The latter addressed broad-market interruptions. These are historical descriptions of that approval, not a complete statement of today’s operating parameters. [5]
A pause can provide time for orders and information to catch up, while a price band constrains where executions occur. Neither creates an obligation for investors to buy at the old price. Analysis of such designs therefore separates erroneous or disorderly executions from a genuine change in valuation. Preventing an extreme print and preserving uninterrupted trading are different objectives and can conflict during stress.
What the event can and cannot establish
The strongest conclusion is narrower than a slogan about computers: displayed prices and high trading volume can be fragile when many intermediaries simultaneously avoid inventory risk. The event also shows why daily closing returns miss important execution losses and why an index rebound does not reverse every participant’s completed transaction.
Its limits are equally important. One exceptional session cannot identify the average contribution of high-frequency trading to spreads or . Reconstructed order books do not provide a clean experiment with the large hedge, spoofing, macroeconomic anxiety or safeguards independently switched off. Historical attribution is therefore a comparison of documented mechanisms and evidence, rather than a precise allocation of blame among a handful of percentages.
Sources
- SEC–CFTC staff, Findings Regarding the Market Events of May 6, 2010; September 30, 2010, executive summaryFiling / report · PDFBack to text: ↑1↑2
- Gregg Berman, SEC staff speech, Market Participants and the May 6 Flash Crash; October 13, 2010Filing / reportBack to text: ↑
- Kirilenko, Kyle, Samadi and Tuzun, The Flash Crash; May 5, 2014 research versionOfficial source · PDFBack to text: ↑
- U.S. Department of Justice, Sarao guilty plea; November 9, 2016Official sourceBack to text: ↑
- SEC, approval of extraordinary-volatility proposals; June 1, 2012 announcementFiling / reportBack to text: ↑