Sudden Stock Crashes Usually Caused by Human Error, SEC Says

Photographer: Scott Eells/Bloomberg

Traders work on the floor of the New York Stock Exchange. Close

Traders work on the floor of the New York Stock Exchange.

Close
Open
Photographer: Scott Eells/Bloomberg

Traders work on the floor of the New York Stock Exchange.

Concern that American stock markets have become more susceptible to split-second crashes due to computerization isn’t supported by the data, a Securities and Exchange Commission official said.

Most “mini-flash crashes,” a term sometimes applied when an individual U.S. stock briefly surges or plunges for no obvious reason, are the result of human errors, not broken software, said Gregg Berman, head of the SEC’s Office of Analytics and Research.

Scrutiny of market disruptions increased in the wake of malfunctions including the flash crash of May 2010, when the Dow Jones Industrial Average fell almost 1,000 points in minutes before rebounding. In September, the Senate Subcommittee on Securities, Insurance and Investment held hearings on the impact of computerized trading amid concern algorithmic and high-frequency strategies are contributing to investor uncertainty.

“A popular meme has emerged that, taken collectively, sudden price spikes indicate a broken market” and may be harbingers of another crash like the one in 2010, Berman said in New York today at a conference sponsored by the Securities Industry and Financial Markets Association. Critics who blame everything on electronic trading “may be looking in the wrong place,” he said.

Princeton Physicist

SEC staff found that swings in individual stocks are more often caused by human mistakes such as “fat finger” trades -- when a person enters the wrong number of shares to trade or some other typographical error -- or incorrectly entered limit orders, Berman said. While the errors reflect sloppiness and highlight a lack of checks, they can be fixed by better risk management and oversight, he said.

Berman, who trained as a physicist at Princeton University, was appointed to his SEC post in January. He joined the agency in 2009 from RiskMetrics Group and led the development of Midas, the SEC’s system for examining the U.S. stock market.

Sudden stock swings have spurred fluctuations in the paper value of some of America’s biggest companies. On May 17, shares of Anadarko Petroleum Corp. (APC), which had a market value of $45 billion at the time, briefly plunged 99 percent in the final minute of trading. Most of the transactions were later voided.

A week later, NYSE Euronext (NYX) let stand trades that sent American Electric Power Co. (AEP) and NextEra Energy Inc. (NEE) down at least 54 percent, while labeling them “aberrant” and excluding them from records showing the stocks’ lows of the day.

Bigger Moves

While crashes may be started by humans, today’s market structure means mistakes are more likely to snowball rapidly, said Sal Arnuk, a partner at Themis Trading LLC and a frequent critic of the way markets have evolved.

“I would ask Mr. Berman, how can you explain or justify that a large-cap or very liquid stock, when there is a fat-finger trade, sees the market widen out as much as it does,” Arnuk said in a telephone interview. The current structure of markets “is set up to extract the most amount of pain from any mistake.”

A study published by Credit Suisse Group AG on Jan. 17 found that few sudden swings are directly attributable to computer errors.

Ana Avramovic, an analyst at Credit Suisse Trading Strategy, examined mandatory halts prompted by volatility in individual stocks between June 2010 through December 2012. After excluding “extremely illiquid or cheap stocks,” she found that 85 percent were caused by news and 9 percent by human error.

Latency Arbitrage

Only 6 percent -- or 21 instances in 31 months -- were caused by a bad print, when a quote at an extreme price caused a halt, suggesting a computer algorithm was responsible.

Recent research from the University of Michigan explored another point of contention regarding high-frequency trading: whether investors generally benefit from the practice.

The report found that while a technique known as latency arbitrage -- in which traders exploit delays in sending market data among the 13 exchanges and about 40 alternative venues in the U.S. -- enriches some firms, overall market efficiency is harmed “with no countervailing benefit in liquidity or any other measured market performance characteristic.”

Berman said today that while the SEC continues to study computer trading, other measures should limit human mistakes. He highlighted the proposed Regulation Systems Compliance and Integrity, which seeks to limit technology breakdowns at venues handling stocks, options and bond trades and ensure they can withstand malfunctions that could jeopardize markets.

Risk Checks

Another initiative, the market-access rule adopted in 2010, requires risk checks on any order sent for execution. The two together are a message to market participants and venues to improve, Berman said.

In remarks to the Sifma audience, Berman also questioned critics who suggest computers have made buying and selling stocks too fast for the good of the market. He said the SEC’s analysis will look at the speed of trades, and he will reserve judgment until it’s complete.

“I’m not sure why, absent other facts, it should be a concern that trading take place faster than a blink of an eye,” he said.

To contact the reporter on this story: Sam Mamudi in New York at smamudi@bloomberg.net

To contact the editor responsible for this story: Lynn Thomasson at lthomasson@bloomberg.net

Bloomberg reserves the right to edit or remove comments but is under no obligation to do so, or to explain individual moderation decisions.

Please enable JavaScript to view the comments powered by Disqus.