Wall Street crash exposes world of stock market electronic trading
by Andrew Beatty
Originally posted 11 May 2010
REGULATORS PICKING THROUGH THE RUBBLE OF LAST WEEK’S DRAMATIC WALL STREET crash have exposed a Byzantine system of electronic trading in the stock market that may have propelled the sell-off.
In ten bone-shaking minutes on Thursday the Dow Jones Industrial Average – representing the 30 most venerable US firms – briefly lost almost a tenth of its value. Open-jawed investors blanched as the pensions and savings of millions of Americans were decimated, along with the livelihoods of countless more. Days and a partial recovery later, the most fundamental question has still not been answered: What happened? Concerns about Europe’s debt crisis no doubt contributed, but few analysts believed worries about Greece’s fiscal malfeasance – as serious as it may be – would cause US equities markets to go through a near-death experience. The Securities and Exchange Commission, the New York Stock Exchange and even President Barack Obama have vowed to uncover the causes of the fall.
In the meantime, homespun theories have been shot down one-by-one. The major US stock markets said a glitch on their trading platforms was not to blame. Citigroup angrily brushed aside the notion that one of its traders had mistakenly hit the billion rather than million button on a sale. Some bolder television commentators speculated that cyber-terrorism may be to blame, although evidence appeared to be lacking. Whatever the trigger, blame for the severity of the crash is now aimed squarely at algorithmic trades.
At their simplest, “algos” are used to buy or sell shares at a certain trigger point, to limit losses or seek new profits. They are responsible for anywhere between 60 and 90 per cent of trading on a normal day. In a market that is driven by endless reams of data, algorithms are able to spot and act on opportunities at a speed investors could not hope to match.
“One program can trade thousands of stocks in milliseconds,” explained Terrence Hendershott, a professor of finance at the University of California, Berkeley. New figures for consumer spending or one stock’s deviation from the broader price trend can trigger a rash of buying and selling while a human trader is off getting coffee. But when the flow of data is disrupted or unusual, the effects can be dramatic. Hendershott speculated that Thursday crash could have been caused by a badly programmed algorithm, or a trader’s mistake.
That could have sparked automated selling that forced stock prices down and triggered a cascade of other automated sales. This domino effect could have occurred at such a pace that humans had no idea what was happening. It is a scenario that proponents of algorithmic trading, like Hendershott, had thought improbable, until Thursday. “What computers are particularly good at is processing the same information in the same ways and optimizing that very, very quickly,” he said. “What they might not be so good at is dealing with some new situations that have never occurred before.”
The response from regulators and Congress has been swift. “A temporary one trillion drop in market value is an unacceptable consequence of a software glitch,” said Senators Ted Kaufman and Mark Warner in a letter a colleagues. A Congressional hearing is planned for Tuesday and calls are growing for regulation. The Securities and Exchange Commission is urging exchanges to put better “speed bumps” in place that would slow down trade, turning more decisions back to human brokers. Many on Wall Street see resolving the quandary as crucial to restoring the trust of ordinary investors.
“America’s confidence in Wall Street was already low. It is now eroded even more than before,” worried David Kotok of Cumberland Advisors – a financial firm. But according to Hendershott, algorithms have also made markets more efficient, better able to assess the true value of an asset. They were also responsible for the equally rapid recovery seen in stock markets on Thursday, he said, as programs spotted ultra-low prices for some stocks.
“It was bad that it went down so fast, but there was no way the humans were reacting fast enough to make it come back.” Thanks to algorithms, 10 minutes is now a very long time in the stock market.