Diebold Plunge May Show How ‘Dueling Algorithms’ Roil Markets

Diebold CEO Thomas Swidarski
Minutes before a June 2 meeting with investors, Diebold CEO Thomas Swidarski saw his company's share price plummet more than 30 percent in six seconds. Source: Diebold Inc. via Bloomberg

Just as a June 2 meeting with investors was starting, Diebold Inc. Chief Executive Officer Thomas Swidarski saw his company’s share price plummet more than 30 percent in six seconds.

As Swidarski and his colleagues tried to figure out what to say at their meeting, the price shot back up. By the end of the day, more than 6 million shares of the Canton, Ohio-based maker of ATM machines had traded, about 11 times the usual volume. The stock closed at $29.08, 3 percent higher than its opening.

The plunge is under review by U.S. regulators, according to people with direct knowledge of the investigation. It may have occurred when a shred of Diebold news bounced into the growing network of computer programs that search out financial information, analyze it and trade instantly without human involvement, market analysts said. The incident highlights concerns about high-speed trading like those surrounding the May 6 selloff that erased $862 billion from the stock market in less than 20 minutes.

“The days when humans traded with humans have long since been replaced by my computer trades with your computer,” said James Angel, a finance professor at Georgetown University in Washington. “What we see here are dueling algorithms. Some algorithm triggers, the stock falls, the other investors look and say that’s old news and start buying.”

Diebold’s June 2 freefall started after the Securities and Exchange Commission announced that the company would pay $25 million to settle allegations of fraudulent accounting. The accord was old news; a year earlier Diebold had announced it agreed in principle to pay that amount to end the investigation of its discontinued practice of recording some sales before machines were delivered. The SEC on June 2 also filed suit against three former Diebold executives.

Six Seconds

At 12:22 p.m., Diebold plunged $8 in six seconds to $18.26, according to Bloomberg data. The stock recovered and by 12:23 p.m. was trading above $25.

Company spokesman Michael Jacobsen said Diebold is considering whether and how to investigate the plunge.

While regulators haven’t ruled out the possibility that the Diebold trades were executed manually, they are racing to catch up with an industry that executes transactions by the millisecond.

The SEC, staffed mainly with attorneys and accountants, is adding statisticians and former hedge-fund employees to its ranks. Last month, the agency proposed a $4 billion system to track unusual market events and suspicious trades in real time.

The agency is “worried about the innocent bystanders,” said Terrance Hendershott, an associate professor of finance at the University of California at Berkeley’s Haas School of Business. “Everyone wants to believe that when they send their order in, they are going to get the fair price.”

Historical Events

Use of software to analyze and trade within milliseconds on news from press releases, headlines and news stories has taken off in the past three years, said Timothy Sargent, chief executive officer of Naperville, Illinois-based Quantitative Services Group LLC. The software is based on algorithms that process historical data, events or keywords that have pushed stock prices up or down in the past.

“There is a parsing activity to look at the flavor of the words chosen,” said Sargent, whose company sells research to institutional investors. “Obviously, there will be negative words and positive words.”

News Corp.’s Dow Jones, Thomson Reuters Corp. and Bloomberg LP supply news for algorithmic trading. Bloomberg LP is the parent company of Bloomberg News.

‘Arms Race’

Temporary price fluctuations have led some firms to install software that looks to profit by buying at the low point of a swing. Pipeline Trading Systems LLC assists institutional investors in implementing algorithms that aim to identify the end of a selloff, said Fred Federspiel, the company’s founder.

“It’s an arms race between institutional traders who are moving positions driven by portfolio managers and traders who are deploying these tools to figure out which direction to trade a particular stock,” said Federspiel, who earned a doctorate in nuclear physics from the University of Illinois in Champaign and previously worked at Los Alamos National Laboratory in New Mexico. “If one party comes to this interaction without a well-programmed machine on their side, they are likely to lose a lot of money.”

Traders don’t care if algorithms occasionally misinterpret data as long as computers keep mistakes to a minimum, said Sargent of Quantitative Services Group.

“The dollars and the money to be made being right a majority of the time is enough to pursue these strategies,” he said. “We are going to get these glaring circumstances, but for every one of those, there are a number of trades where it works very well.”

Circuit Breakers

To respond to the May 6 crash, the SEC and exchanges are implementing curbs on a trial basis to halt trading of Standard & Poor’s 500 Index stocks when their prices rise or fall 10 percent in less than five minutes. The circuit-breaker test, which is scheduled to last through Dec. 10, will pause trading for five minutes to let market participants assess news about the company and submit new buy and sell orders.

While Diebold isn’t listed on the S&P 500, SEC Chairman Mary Schapiro said last week that she plans to “rapidly” expand the six-month pilot program to “thousands” of additional companies.


Trading specialists and enforcement attorneys at the SEC and the Financial Industry Regulatory Authority are examining what triggered the Diebold trades and whether the selloff could have resulted from manipulation or inside information, according to people familiar with the matter who spoke on condition of anonymity because the probes aren’t public. NYSE Arca and Nasdaq OMX Group Inc., which reviewed the trades for misconduct and pricing errors, determined they should stand.

Four hundred twenty-seven trades occurred in Diebold below $23 on June 2 covering about 113,600 shares, according to Bloomberg data. All of the trades took place on electronic venues such as Nasdaq Stock Market and Bats Exchange. No transactions occurred on the New York Stock Exchange for 70 seconds starting at 12:22 p.m.

Diebold CEO Swidarski was on the road that day with another company official to speak with investors, according to a person with direct knowledge of the trip.

NYSE spokesman Ray Pellecchia said a so-called liquidity replenishment point was triggered in Diebold that switched electronic trading to automated auctions overseen by humans. Other trading venues don’t use the curbs.

On May 6, liquidity replenishment points caused sell orders to flow to other platforms. The SEC and Commodity Futures Trading Commission said in a May 18 report that they are analyzing the LRP system to determine whether it exacerbated the crash.

‘On the Lookout’

The Diebold plunge is not the first to be reviewed by regulators. News reports in March said the SEC is also examining why Dendreon Corp.’s shares fell 69 percent in less than two minutes in April 2009. Minutes after trading was halted, the Seattle-based drug developer said its Provenge cancer treatment extended the lives of men in a clinical study, spurring a rally of more than 130 percent after hours.

“We’re constantly on the lookout for aberrational trading,” said Scott Friestad, an associate director in the SEC’s enforcement division, declining to comment on any particular case. “When we find trading that looks suspicious, it’s our normal practice to take steps to determine whether any improper conduct occurred.”

Diebold’s sudden spike in trading may have hurt investors who used so-called market orders and stop-loss orders, said Hendershott, who researches electronic markets and does consulting for firms that engage in high-frequency trading.

Stop-Loss Orders

A client placing a market order is requesting that a broker buy or sell a stock as quickly as possible at any price. During the Diebold plunge, a broker who received a market order when shares traded for $22.73 might not have executed the transaction until shares sold for $18.99 less than one second later. The client would have lost more than 16 percent of the investment, only to see the stock immediately recover.

A stop-loss order tells a broker to sell shares for a client at a specific price to cap losses. Trading in Diebold may have triggered stop-loss orders, adding to the downward momentum.

“If people think the market is rigged or does all these weird things, why would they have any faith that they will ever get their money back,” Hendershott said. “Why would people choose to put money in the markets to help firms grow and raise capital?”

Before it's here, it's on the Bloomberg Terminal. LEARN MORE