When speed kills

At the beginning of August, the Knight Capital Group, a broker of U.S. stocks, lost $440 million in about 45 minutes. Losses (and gains) of that magnitude are a part of the natural order for stock markets; winners and losers are made every hour.

But Knight’s losses were the result of an algorithm run amok — a new computerized trading program with a software glitch — and raises basic questions about the growing use of such programs and the benefits they deliver.

When the U.S. market opened on Aug. 1, a newly installed program in Knight’s computers entered incorrect bids for about 150 stocks. Other companies, using their own computer programs, recognized the errors and began trading against Knight. The Knight bids pushed up the value of many stocks and the company lost money when it had to resell those overvalued shares at a lower (and more accurate) price. By the end of the day, the company had lost nearly half a billion dollars — a loss of $10 million a minute.

It is not clear why the problem occurred. A firm is usually especially vigilant when it introduces new software. Yet inexplicably, Knight never hit the “off switch” to deactivate the program — even after market officials noticed the trades and notified the company. For a program trading in milliseconds, a half hour is an eternity.

The “Knight-mare” is not the only example of the danger posed by computer-driven trading. The May 2010 “flash crash” carved nearly 1,000 points — almost 10 percent of its value — off the Dow Jones Industrial Average in under an hour, but the Dow bounced back in the same period of time. To this day, no one is sure what triggered that episode.

In March of this year, Bats Global Markets Inc., a market making firm, was forced to withdraw its initial public offering after its system froze. Just two months later, the Facebook IPO was marred by glitches in a Nasdaq computer that forced a halt to many trades, and cost the Swiss Bank UBS nearly $350 million. Earlier last week, trading in Spain’s benchmark Ibex 35 Index was halted for nearly five hours as a result of a technical glitch.

The embrace of computer-driven programs reflects an obsession with speed among stock trading firms. Computers seek discrepancies in prices that exist for nano-seconds or even pico-seconds (one trillionth of a second). This creates absurd attempts to find advantage: Firms have moved to New York City from other parts of the United States to capture milliseconds; there is reportedly a real estate boom in blocks in NYC where cables come out of the ground to exploit the milliseconds of internet travel time. Mass volume trades can turn even the smallest differential into a profit. In theory, this is good for markets — a readiness to buy and sell provides liquidity. Moreover, argue proponents of high-frequency trading, the losses are felt by market participants who knowingly took the risk (if trader accounts are properly segregated).

As a result, it is estimated that 50 percent of stock-market volume in the U.S. is now based on high frequency trading. The phenomenon is relatively new to Australia, but it is already reckoned to account for 20 percent of trade volume there.

That potential for big gains is also a potential for big losses, and the expanding list of glitches that culminated most recently in the Spanish shut down is proof that some losses are real, not just potential. Thus far, we have been lucky and the losses contained. But the ubiquity of these programs and the money they are moving mean that a large disaster is only a matter of time.

And, as in the 2008 financial crisis, the scale will demand a government intervention and the use of taxpayer money. Moreover, one recent study argues that high frequency trading actually raises the price of some stocks for long-term investors. Finally, and most importantly, these incidents erode faith in markets and temper individual incentives to put their money at risk.

Market officials and regulators are increasingly skeptical of the notion that faster is by definition better. Ms. Mary Schapiro, chairwoman of the Securities and Exchange Commission, the chief U.S. regulator of stock trading, called the Knight incident unacceptable and has said that her staff would be discussing steps “to address these critical issues.”

That skepticism is good. There is little chance of a complete retreat from high frequency trading, but it can be better regulated. At the most basic level, that means better testing of software before it goes live. There needs to be renewed examination of “circuit breakers,” programs (ironically enough) that suspend trades when a particular share or a market behaves erratically. There also needs to be study of a financial transactions tax.

This last suggestion is likely to be most controversial. All calls for greater regulation meet instant criticism that they will reduce the amount of money made available to investors, and this reduction in liquidity will slow economic growth. That is the point. The lesson of the global financial crisis in 2008 was that some “growth” was not real growth at all. It was “vapor wealth”: fictional capital that vanished when scrutinized.

Unfortunately, that imaginary capital was the cornerstone of all sorts of other investments and institutions and when it was revealed to be fictional, the entire edifice crumbled. The world does not need that sort of growth.