So now we know that the “Flash Crash” of May 6, when the S&P plunged almost 1000 points in a few minutes was not caused by a “fat finger” error after all. A fat finger error, we should explain, is where a trader means to sell, say, one billion shares and keys one too many zeros before hitting the “send” button, thus selling 10 billion shares, and in the process triggering a briefly self-sustaining selling wave by both traders and automated trading programmes.
The joint report into the May 6 flash crash by the Securities and Exchange Commission (SEC) and the Commodity Futures Trading Commission (CFTC) did not name the firm responsible, but the report’s analysis shows that a computer sell order of $4.1 billion, which executed in a few minutes, precipitated the crash. The media have identified the firm involved as the Kansas-based firm, Waddell & Reed.
According to the SEC report, the order was properly made but the way it was executed - the mechanics of the software sell program, as it were, and the constraints on it - left quite a bit to be desired. Those who got hit by the steamroller that crunched through the markets on May 6 would put it more strongly than that.
What the report goes on to uncover is how in today’s high frequency trading environment a negative feedback loop can be created and sustained, triggering unintended and very damaging effects. Markets are supposed to have “circuit breakers” that trip to stop bizarre ripples from destabilising trading activity, but on May 6, none of these worked for a while because the ripple came in a way that had not been anticipated by those who designed the market circuit breakers. Which goes to show that you can’t figure out everything in advance and experience can be a painful teacher.
On the day the sell order was executed, the markets were already having a bad time of it and many stocks were down as much as 4% from their prior day close. When the sell order hit they lost a further 5-6% of their value in minutes, and then rebounded almost as quickly. Some lost 15% or more and at the extreme end of the scale, over 20,000 trades, involving more than 300 securities were, as the report puts it, “executed at prices more than 60% away from their values just moments before.”
It is worth reflecting on that for a second or two. Anyone who had placed a long position on a stock with a sensible stop would have been stopped out, losing the difference between the price at which the stop was executed and the price at which they entered the market, multiplied by the scale of their position. If it was a large position with a sensible deep stop, that would have hurt. The same thing would have happened to anyone who tried to react to the crash by going short, once it bounced back up again.
The howls from small traders could be heard across America, and in these days of global trading, not just America. The point of course is that if markets cease to reflect real supply and demand pressures and instead show a tendency to just bounce around whimsically, it’s goodnight and goodbye to trading and farewell to the liquidity that markets provide.
Traders will tolerate losing because of unexpected market volatility. That’s just part of the game and is covered by the technical phrase “stuff happens”. But they won’t put up with being screwed by a computer glitch. And they look to the regulators, at both the Exchange level and the Federal level, to make sure this doesn’t happen – which is why the SEC and the CTFC put so much work into understanding the May 6 flash crash phenomenon.
The report is an excellent forensic exercise and is a must-read for anyone who wants to understand where high frequency trading and computerised dealing is taking the markets. There is no space here to provide a detailed account of what happened, and anyway, the report does that very nicely in 104 pages. In a nutshell, however, it seems that the failure to make the sell order sensitive to price or time and only sensitive to volume, created a wild feedback loop.
As the underlying liquidity in the market drained away in the face of huge selling, it looked to the market as if buyers for stocks were falling away so prices kept falling. Where the £4.1 billion sell order was presumably supposed to execute over several hours, the programme looked at the volume of orders flowing through the market and responded by selling more – and more – and more – in minutes. Finally one of the market buffers kicked in and stopped all trading for a few minutes. That kind of reset things.
Then the tide came flowing back in and a buying frenzy started, pushing stocks back to pretty much where they had been before the $4.1 billion sell programme did its little number. And then everyone looked around and said: “What the hell just happened?” The report provides a very detailed answer…
Will it happen again? The “Lessons Learned” section of the Report is unequivocal:
"One key lesson is that under stressed market conditions, the automated execution of a large sell order can trigger extreme price movements, especially if the automated execution algorithm does not take prices into account. Moreover, the interaction between automated execution programs and algorithmic trading strategies can quickly erode liquidity and result in disorderly markets. As the events of May 6 demonstrate, especially in times of significant volatility, high trading volume is not necessarily a reliable indicator of market liquidity."
The report also points out that 6 May was a blindingly clear demonstration of the “inter-connectedness of our derivatives and securities markets, particularly with respect to index products.” The SEC and the CTFC are now working flat out, along with the markets, on the question of recalibrating existing market circuit breakers.
One of the subtler lessons though, which the report also picks up, is that if one market participant looks at what appears to be weird behaviour in the market and withdraws their participation, that’s fine. If everyone says, “Hey, that’s weird,” and withdraws, market liquidity crashes and prices fall through the floor.
What is needed is a circuit breaker to pause things before events get to such a stage. The hope is that the pause gives participants a moment to collect their wits, and because they are professionals, it tends to work. Just as importantly, an enforced pause serves to jolt robotic algorithms into a reset, so they stop pouring fuel on the fire at speeds that only turbocharged computerised systems can achieve. High frequency trading running amok. When the machines aren’t with you, they surely are against you, even if they don’t mean to be.
Further reading on the flash crash and market risk:
- Fat fingers and the log-log law, a blog post by Anthony Harrington
- Investing in a Volatile Environment: A Black Swan Perspective, by Javier Estrada
- Risk Management at a Crossroads, by Maureen J. Miskovic
Tags: financial crisis , fiscal stimulus , regulation , SEC , sovereign debt , trading , transparency , US