Computerised trading systems make it possible to shift stocks at the speed of light - but with potentially disastrous consequences as those investigating the circumstances of the 6 May 'Flash Crash' have discovered.
Leonard Amoruso, general counsel at broker-dealer Knight Capital Markets, recalls clearly the scene of confusion on the afternoon of Thursday 6 May 2010, when the US stock markets, suddenly and without warning, nosedived.
'I was on the trading floor among our traders. Half of them were looking at the TV screens to see if there was some worldwide catastrophe that had not been broadcast yet. The other half were looking behind the computers to see if the cables were plugged in,' he told New York financial regulators in the wake of an event that earned the title 'Flash Crash'.
The Dow Jones Industrial Average plummeted by close to a thousand points around 2:45pm Eastern Time on 6 May. In a matter of minutes, $1tr was wiped from the value of shares traded on Wall Street.
Shares in consumer products giant Procter & Gamble alone fell almost 40 per cent, although they would bounce back before long. Some shares traded for a single cent - a fraction of their value only minutes before the crash occurred.
The cause of the Flash Crash was a mystery - and it remains so.
A 'fat finger' trade - jargon for a keying error that might have caused a huge number of shares to be dumped on the market at a low price - has now been all but ruled out. One abiding theory is that it could have been the result of a confluence of automated trades and market rules that somehow synchronised to push prices down. Managers from a number of the US-based equities exchanges told a Securities and Exchange Commission (SEC) hearing that the likely trigger was a rush of bad news from Europe. The Euro looked to be in trouble as rioters took to the streets of Athens amid Greece's economic woes; and the oil spill from BP's Deepwater Horizon rig was fast turning into an environmental disaster for the US states along the coast of the Gulf of Mexico.
CEO of electronic stock exchange Direct Edge William O'Brien describes 'an environment of widespread investor skittishness'. Investors had many reasons to sell first and analyse later, he says. 'Sharp market declines are as old as markets themselves,' O'Brien points out; but that has not stopped widespread suspicion falling on computer-initiated trades for turning what was probably a simple mistake to a near catastrophe for US stocks. The fact is, humans just cannot trade that fast.
The transitory penny stocks provided a clue to what happened. In Congressional hearings, SEC chair Mary Schapiro pointed to computerised trading as a potential driver for stock prices to drop to just one cent - placeholder offers to buy, called stub quotes, that only come into play when a price plummets to such low levels. No one expects those orders ever to be made - no one bar a computer.
Schapiro told congressmen: 'Automated trading systems will follow their coded logic regardless of outcome, while human involvement likely would have prevented these orders from executing at absurd prices.'
Gary Gensler, chair of the US Commodity Futures Trading Commission, told politicians: 'Algorithms don't think smart. They are programmed to do what they do. Repeatedly.'
Then Senator Charles Schumer went further: 'Something bad happened when machines took over. I worry that we have let the technology get too far ahead of us. These machines are both very smart - and very dumb.'
Roots of the crash
Schumer's opinion has a number of backers on Wall Street, among more traditional brokerages that feel that 'whizzkids with computers' have taken away their profits and destabilised the markets, all in the quest for 'alpha', the ability to make money even when markets are not shooting skyward. Jeffrey Engelberg, principal and senior trader at Southeastern Asset Management, told regulators: '6 May is the result of a situation where computers are left to misread complex market signals, and run amok before intelligent human thought can regain formerly relinquished control. Our psychology has fallen victim to the notion that, without qualification, technology and speed should be embraced.'
Senior adviser at accountancy firm Grant Thornton David Weild went on to claim: 'A computer arms race has been unleashed on Wall Street. It has created new forms of systemic risk.'
Companies that use algorithmic trading and high-frequency trading strategies argue that the core technologies are not to blame, pointing to the fact that markets have collapsed before without any assistance from computers. The difference with the Flash Crash lies in market structure. Matt Schrecengost, chief operating officer of financial technology firm Jump Trading, argues: 'Even at the epicentre of' the Euro, those markets held solid. If you thought any market would fall apart, you would think it would be in Europe.'
In reality, the Flash Crash happened after the European markets closed, restricting the worst of the damage to the US. Some fear what might have happened had the Flash Crash occurred at 10am Eastern US Time. But even then, Schrecengost points out that it was only US equities or stock markets that saw the oddest behaviour.
'I believe 6 May was a stress test of all the ecosystems out there... And US equities came out the worst,' he says. Traders can buy and sell equities at no less than five venues around New York. 'It is amazing how complex the US equities market is,' he says, adding that the circuit breakers each market was using were not harmonised in their behaviour.
Tabb Group analyst Kevin McPartland points out that only a few hundred firms employ trading strategies that depend on execution speed to make money; but, because they can pick up on opportunities caused by temporary mis-pricing and sudden changes in valuation, other firms have been forced to follow suit. Anoop Prasad, managing director at DE Shaw, explains: 'Even if I want to hold a position for a month. I will break it down into small pieces and do it algorithmically,'because I don't want to stand out.'
Trading 'arms race'
Meanwhile, David Cummings, chairman of liquidity provider Tradebot Systems, argues that the role technology plays in financial trading is a self-evidence reality that the sector will have to get used to: 'Technology is a fact of life,' he avers. 'If you don't like the prices, get a better broker. Not every grandma can get a collocation centre, but she can hire a good broker.' Cummings and his ilk point to the way that trading costs have declined with the rise of automated trading and the death of the 'good old boy' regime, when open-outcry trading ruled.
'People sometimes characterise this as an 'arms race',' says Giles Nelson, deputy CTO at software vendor Progress Software. The company provides much of the complex event processing (CEP) software that drives algorithmic trading and, increasingly, its regulation.
'Over the past year, high-frequency and algorithmic trading have been getting a hard time. A lot of issues have been conflated together, and high-frequency trading appears to be the team that people like to blame. With an incident like the Flash Crash, no one really knows what went on. There were some movements that were legitimate and then the algorithmic engines started following those.'
'It seems to be more of a market structure issue,' says Tabb Group's Kevin McPartland. He points out that the collapse was fast, but the recovery from the situation where valuable shares were offered for a penny each came equally quickly.
'Did it happen quickly because of the technology? Of course. But the rebound was also fast. Algorithms were able to spot what was going on, and to see a buying opportunity. In previous years, [in a situation like 6 May] the markets would have had to close for the rest of the day.'
Given the potential advantages it brings, it seems that beyond tighter regulation of standard operating procedure, the scaling back of high-frequency trading - relying on ever-faster, ever-more-capacious computer systems - is negligible.
'Markets do get abused; traders do try,' admits Progress Software's Giles Nelson, 'and they employ various strategies aided by technology to manipulate the market. That's not acceptable. But that's not technology per se doing the damage there.'
There is no going back - high-frequency trading is here to stay, Nelson believes: 'But what does it actually mean? This is one of the problems: the lack of a common definition. And, if high-frequency trading is bad, then what frequency is good? Who is going to decide that?'
Exchanges such as the London Stock Exchange, and regulators such as the US Securities and Exchange Commission (SEC), and the UK's Financial Services Agency, are looking to use the same technology as the algorithmic traders to watch for potential abuses. The SEC is trying to 'replay the tape' of the afternoon of May 6th to try to work out what happened, recreating the events on computer.
'You can take all the data from the market, and look in hindsight - but if you are looking in real-time, you are in a better position to prevent that behaviour occurring,' Progress Software's Giles Nelson concludes. 'Regulators need to catch. You can't catch a Ferrari in a three-wheeler.'