When the Dow Jones Industrial Average plunged more than 1 000 points on May 6, there was no lack of obvious culprits. Between Greeks bearing petrol bombs, a giant oil slick off the east coast and an unemployment rate stubbornly stuck in the double digits, traders on the New York Stock Exchange had plenty to be twitchy about.
But the biggest intraday drop ever (bigger even than Black Thursday) had nothing to do with the real economy and everything to do with the virtual one. It seems that the massive wave of selling that dragged the market down was almost completely automated — driven not by human panic but by computerised logic.
A huge majority of trading on modern stock markets is now computerised. The popular image of brokers yelling orders from the floor of an exchange has been largely anachronistic since the late 1980s.
At first the May 6 crash was thought to be a human error — a so-called “fat finger” trade where a trader accidentally adds a nought or two to a big order. This happens with more regularity than brokerages would like to admit.
But an ongoing inquiry into the sudden crash (and equally sudden recovery) is turning up evidence that the new breed of “high-frequency trading” (HFT) firms may have fuelled the crash.
These traders don’t just use computers to trade — they teach computers to do the trading for them. Using powerful algorithms these computers tirelessly scan bourses around the world, looking for tiny differences in prices between exchanges.
The differences are often less than a US penny in value, and the trades typically take only a few millionths of a second to execute. That sounds like a lot of trouble for a small amount of money, but if you repeat the trick several million times a day, those pennies start to add up. Although they’ve only existed for a few years, HFTs now account for more than two-thirds of all trades on most exchanges.
Call for the banning of HFTs
So what, exactly, happened on May 6? Regulators are still sifting through records of millions of trades to determine the exact cause, but the logic goes as follows: a very large trade order hit the market, triggering hundreds of HFT drones, which swarmed all over the opportunity.
In doing so they exacerbated the natural drop in prices, triggering yet more drones, including the simpler variety used by traditional brokerages to ensure they are shielded from large drops in the market. This triggered yet more HFT drones, and so on, with the algorithms chasing each other to the bottom of the market.
As a result some people are calling for the banning of HFTs, claiming they pose too much of a risk and add too little value. Traditional brokerages are among the loudest advocates of the idea. They dislike HFTs because they cut into their margins by exploiting their inefficiencies.
Of course they are conveniently ignoring the fact that their own automated systems were probably equally blameworthy. And while HFTs may seem to be making money for nothing, they actually perform a valuable service: they make the markets faster, more accurate and more efficient.
They also drive the big brokerages to invest in their own technology, and their innovations will undoubtedly ripple out into the general computing world, helping to do everything from predicting the weather to monitoring for epidemics.
Luckily a more sensible safeguard than banning is already on its way. Major exchanges are setting up “circuit breakers” that will halt trading on a share if its price suddenly rises or falls by a significant margin in a short period of time.
We may not like the idea of computers furiously buying and selling our pension funds, but they are probably better suited to it than we are. At the very least they are logical, dispassionate and their flaws can be fixed. And they’ll never put your entire life savings with Bernie Madoff, or burn down banks when they don’t get their way.