Flash Crash

A flash crash is an abrupt, severe decline in the price of a security, an index or a whole market that unwinds almost as quickly as it happened, driven by the mechanics of trading rather than by news about value. The name attaches first and foremost to the afternoon of 6 May 2010 in US equity and futures markets, and it has since been applied more generally to other sudden dislocations that reversed within minutes.

What separates a flash crash from an ordinary sharp fall is the reversal and the cause. Prices in a flash crash go to levels no participant would defend a minute later, because the orders that would normally have stood in the way were withdrawn, and orders that had to execute at whatever price was available were matched against whatever was left in the book.

The regulatory record describes the event tersely. In NYSE's own filing to make the current market-wide circuit breakers permanent, the exchange records that the pre-2013 trigger levels "were approached but not breached on May 6, 2010, when the U.S. securities and futures markets experienced a severe disruption, often referred to as the 'Flash Crash.' Between 2:32 p.m. and 2:45 p.m., the DJIA dropped about 9% and then rebounded within minutes. The decline never reached the 10% trigger, so securities trading continued unhalted."

The distribution of the damage is the part worth understanding. The same filing records that "approximately 86% of securities reached lows for the day that were less than 10% away from the 2:40 p.m. price. The other 14% of securities suffered greater declines than the broader market, with some trading all the way down to one penny." So this was not a uniform market-wide fall. It was a broad decline of ordinary severity with a tail of individual securities that detached completely from any plausible value.

Those penny prints are what exposed the structural problem. Market makers who did not want to trade at a given moment were satisfying their quoting obligations with placeholder quotes far away from the market, at a cent to buy or a very large number to sell. Those quotes were never intended to be hit. On 6 May they were, because orders that must execute immediately kept walking down a book that had nothing real left in it. The SEC subsequently banned such stub quotes outright.

The futures side of the market had a safeguard that worked. NYSE's filing notes that "at approximately 2:45 p.m., CME's Globex stop logic function initiated a five-second trading pause in the E-mini S&P 500 futures contract because of a rapid 5% decline in the contract's value." A pause of five seconds was enough to let resting orders be re-entered, and the recovery began shortly afterwards. The contrast with the equity market, which had no comparable mechanism and whose own circuit breakers were set too wide to fire, drove much of the regulatory response.

That response was substantial and is still with us. The SEC adopted or approved a ban on stub quotes; single-stock circuit breakers, which were later replaced by the limit up-limit down plan; a rewrite of the market-wide circuit breakers, including the switch to the S&P 500 and to daily recalculation; the Consolidated Audit Trail, which gives regulators an order-by-order record across markets; and Regulation Systems Compliance and Integrity, which sets requirements for the technology systems that run the markets. Our page on circuit breakers covers how the halts work today.

The durable lesson for an individual investor is narrow and practical. A flash crash is a liquidity failure, and the investors it actually costs money are the ones whose orders had to execute during it: market orders entered into the fall, and stop orders that converted into market orders when their trigger price was touched. An investor who did nothing lost nothing, because the prices came back.

The sequence is always the same in outline. Selling pressure arrives faster than resting buy orders can be replenished. Participants who normally provide quotes step back, either deliberately or because their own risk systems pull them out. The visible book thins to almost nothing. Any order that must be filled immediately then executes against whatever prices remain, however far away, and each such execution prints as the new market price, which can trigger further automated selling.

A hypothetical illustration of how a stub quote turned into a real loss. An investor holds 100 shares of a stock trading around $40 and sends a market order to sell, an instruction to accept whatever price is available. Ordinarily the best bid is a cent or two below the last trade, so the proceeds are close to $4,000.

In the middle of a flash crash the genuine bids are gone and the highest standing bid is a placeholder at $0.01. The market order fills there. The proceeds are 100 multiplied by one cent, which is $1.00. The order did exactly what it was told: it took the best available price, and the best available price was the problem.

A stop-loss order produces the same outcome by a different route, because a triggered stop becomes a market order. Our page on stop-loss orders covers that mechanism, and the limit order is the instruction that would have refused the fill.

What the episode established

• Displayed liquidity can disappear in seconds, so the depth visible on a screen is not a promise about the depth available a moment later.
• Orders that must be filled immediately carry a price risk that is invisible in normal conditions and enormous in abnormal ones.
• A very short, automatic pause can be enough to restore an orderly book, which is the principle behind the single-stock pauses adopted afterwards.
• Prices that reverse within minutes were never valuations, so an investor who took no action was not harmed by them.

What it did not establish

• It was not a verdict on the companies whose shares printed at a penny, and nothing about their businesses changed that afternoon.
• It is not evidence that a repeat is impossible now. The safeguards adopted since narrow the failure mode rather than removing it.
• It offers no guidance on what any market will do next, since the whole character of the event was that it undid itself.