Brake needed for lightning-speed stock trading

The United States has the most sophisticated financial markets in the world, yet they can unaccountably spin out of control at a moment’s notice.

The latest case involves Knight Capital Group Inc., a securities-trading company in Jersey City, N.J., that was laid low by one of its inadequately tested computer-trading programs. In less than an hour on Aug. 1, the program entered incorrect bids for about 150 stocks into the interconnected electronic marketplace. Computer programs at other firms sniffed out the errors and traded against Knight. By the end of the day, the company was out $440 million, forcing it to seek outside financing to survive.

This debacle highlights a market weakness that regulators have been unable to address because high-frequency trading has raced ahead of the humans who try to contain the mischief it can cause. Industry resistance to improved regulation hasn’t helped. If any good comes out of the Knight episode, it will be a commitment by Wall Street’s trading firms to help regulators design systems that can track lightning-speed transactions and tap the brakes when the computers spin out of control.

By now, it’s pretty clear that computer-driven trading can be a major source of volatility. In May 2010, the so-called flash crash knocked almost 1,000 points off the Dow Jones Industrial Average in less than an hour, only to bounce back just as fast.

Then in March of this year, Bats Global Markets Inc., another computer-trading firm, had to withdraw its initial public offering after its system froze, forcing a halt in shares of Apple Inc. Two months later, Facebook Inc. couldn’t trade on time in its first day as a public company because of hang-ups in a Nasdaq computer, feeding doubts about the reliability of markets and costing the Swiss bank UBS about $350 million. UBS was able to absorb the loss; a smaller firm might have failed.

The pursuit of speed is an obsession among Wall Street trading firms. They can profit by taking advantage of minuscule discrepancies in prices that might exist for no more than a few microseconds or by getting a sneak peek at orders before they are executed. The risks of this are becoming more obvious, even to some market leaders.

Markets now believe that "if something is faster, then by definition it’s better," says Duncan Niederauer, chief executive officer of NYSE Euronext, which owns the New York Stock Exchange. "We are understanding that speed is not always better."

Amen. It is also true that making markets more resilient doesn’t mean a return to telephone orders, paper order tickets or stock exchanges with only human traders bellowing out bids and offers.

What we can do, at a minimum, is ensure that monitoring systems keep up with the traders’ algorithms.

The Securities and Exchange Commission last month took a half-step by requiring markets to build a $4.1 billion system that can generate audit trails of all transactions. The trouble with this system is that trading data won’t be generated until the next day, a feature the industry insisted on. Day-old data might not help regulators much when they are called upon the instant a market blows up. (It took five months to confirm the cause of the flash crash.) Nor is it clear that the system will be able to pinpoint the identity of every party in a transaction.

The SEC might need to consider whether more work is needed on market circuit breakers, a device introduced after the 1987 market crash that temporarily stops trading when an individual stock or an entire market behaves erratically. New rules, scheduled to take full effect next August, set lower thresholds for triggering a trading halt. One idea the SEC might consider is applying circuit breakers to specific firms that issue an excessive number of trade orders.

The SEC and stock exchanges should also require major trading firms to demonstrate that their software programs are reliable before letting them go live. Now, firms are simply urged to adhere to an industry-recommended set of best practices. Unleashing a flawed program, as Knight seems to have done, is unacceptable.

Market apologists have said Knight’s errant trades caused no harm to anyone other than Knight and its shareholders, who saw the value of their investment shrink by about $600 million in a few hours. Yet who can be so certain the next bug-infested program won’t inflict much more damage? And what might have happened if Knight, which handled about 11 percent of all U.S. stock trading before the errors, had shut down?

Plus, the argument that Knight only hurt itself is bogus: Investors withdrew $127 billion from stock mutual funds in the 12 months ended in June. Repeated computer-trading misfires — not to mention the financial crisis of 2008 — erode confidence in U.S. markets. At some point, regulators and Wall Street have to decide whether the quest for speed is worth the chaos that can result.