Could Computerized Trading Cause Another Market Crash?

Unless regulators put the brakes on Wall Street's robot traders, more stock-market tumbles like the May 6, 2010, "flash crash" are "absolutely certain" to occur, according to a knowledgeable trading expert.

Fewer and fewer Wall Street traders are human beings. Instead, they’re computers that execute trades in milliseconds (a millisecond is one thousandth of a second). A forerunner of today’s robotic trading, computerized program trading, was largely responsible for the stock market crash of October 19, 1987, when the Dow Jones industrial average plunged 22.6%.

No one knows for sure, but the best estimates are that these high-speed computers -- I prefer to call them robots -- account for 60% to 70% of the volume on the major U.S. stock exchanges. Their activities are shrouded in mystery. Some are run by well-known firms, such as Goldman Sachs. Others, such as Jump Trading, Citadel, Getco and Tradebot, are known mainly by traders. Revenues from robotic trading totaled an estimated $7.2 billion in 2009, according to Tabb Group, a New York City market-research company.

Subscribe to Kiplinger’s Personal Finance

Be a smarter, better informed investor.

Save up to 74%

Sign up for Kiplinger’s Free E-Newsletters

Profit and prosper with the best of expert advice on investing, taxes, retirement, personal finance and more - straight to your e-mail.

Profit and prosper with the best of expert advice - straight to your e-mail.

Sign up

Some of the activities of these high-frequency traders are benign. Robots have taken the place of market makers, buying stocks at one price and selling them at a slightly higher price. Spreads are typically a penny or two on widely traded securities -- much lower than when market makers were humans. It’s hard to complain about this practice. Indeed, one recent report argued that robotic trading decreases market volatility. I think that’s hogwash.

In addition to making markets, robots also practice momentum strategies -- buying stocks, exchange-traded funds and other securities that have already gained in price and selling securities that have fallen in price. Momentum strategies exacerbate volatility. The robots are programmed with complex algorithms that tell a computer to be on the lookout for patterns in trading. When it spots a certain pattern, the algorithm tells it to buy or sell a security, and what to look for next.

During wild swings in the market, robots can make bigger profits for their human masters, so they step up their activities. Not only does this practice further increase market volatility, it causes markets to move more closely in sync with each other during periods of turbulence, says Eric Hunsader, president of Nanex, a Chicago-area firm that collects and sells market data. Because robots operate on all kinds of exchanges around the world, they increase the tendency of markets to move in lockstep, so that futures on stock indexes, commodities and bonds all tend to move in the same direction, along with stocks and exchange-traded funds.

Many market watchers, including Hunsader, are convinced that robotic trading caused the 2010 flash crash, when the Dow tumbled 1,000 points, then recovered -- all in a matter of minutes. He adds: “I am absolutely certain there will be more flash crashes.”

Just a few days before the flash crash, on April 28, 2010, a robotic trader simultaneously sold hundreds of millions of dollars of the SPDR S&P 500 ETF (symbol SPY) and a similar amount of S&P futures, Hunsader says. That caused spreads on both, normally a penny, to balloon for one or two seconds, enabling the robot to buy back the ETF and the futures contract at a lower price and allowing it to pocket quick profits.

Robots executed the exact same trades -- again with hundreds of millions of dollars -- three times within ten seconds immediately before the Dow plunged more than 600 points in the flash crash, Hunsader says. He’s convinced that the same robotic trader sold both the ETF and the futures because the trades happened at the same millisecond.

The Securities & Exchange Commission, which had earlier reported that the flash crash was due to other factors, has subpoenaed the records of high-frequency traders as part of a renewed investigation.

Hunsader and others are also highly critical of a robotic practice called “quote stuffing.” Robots distribute hundreds of thousands of bids on securities, then withdraw them in milliseconds before anyone can act on them. The practice seems designed to slow the traffic of market data slightly so that humans and other robot traders fall just a bit behind. High-speed traders “can ignore the bogus quotes, but nobody else can,” says Randy New, who works for Abel/Noser, a New York City brokerage firm. “It’s like running the New York marathon and throwing marbles behind you.”

Quote stuffing has surged. Just a year ago, for every executed trade, there were four or five limit orders for a stock that were quickly withdrawn without their price being hit. “Now it’s 50, 60 or 70 orders per trade,” says Hunsader.

That has led to an explosion in the volume of trading data on the exchanges, slowing the speed at which market participants can receive the information. The day of the flash crash, Hunsader says, the data hit a record high of half a trillion bytes. Then it subsided. But for the five trading days starting August 5, the volume surged to an average of one trillion bytes per day.

Instead of fostering liquidity, Hunsader argues, robots are increasingly forcing human traders to the sidelines because they can’t compete. That decreases trading volumes and liquidity, as well as the number of different strategies. “Anyone whose algorithm isn’t totally based on speed is unable to trade,” says Hunsader.

Eugene Noser, who heads the Abel/Noser brokerage, says the stock exchanges are increasingly hostage to robotic traders because they account for so much of the exchanges’ income. To gain tiny speed advantages, robotic traders have been allowed to locate their computers right next to the exchanges’ computer servers.

More damaging, says Noser, is that some exchanges are providing robotic traders with master order numbers that enable the robots to determine what firm is buying or selling a stock. For example, if Fidelity wanted to sell its holdings in Apple (AAPL), the robots would know from the first trade that Fidelity was the seller. “So how smart do the high-frequency traders have to be to make money from that information?” Noser asks rhetorically. “This is privileged information that a firm like Fidelity thinks belongs to them.”

The solution? It’s really pretty easy. Robotic traders never hold a stock for more than two seconds. Make it illegal to hold a stock less than 60 seconds. To deal with quote stuffing, just require that any limit order be required to stay open for one second. And master order numbers should be confidential.

The problem, Noser says, is that the SEC “is absolutely overwhelmed. It doesn’t have the personnel who know what’s going on.” All we can do for now is hope that we don’t get another flash crash.

Steven T. Goldberg (bio) is an investment adviser in the Washington, D.C. area.