The Perils of Program Trading
On the afternoon of May 6, 2010, the Dow Jones Industrial Average (DJIA) plunged 800 points in less than 20 minutes before recovering most of its losses, creating a new term in the financial lexicon – the “flash crash.” The Dow’s intraday drop of 998.5 points or 9.2% was its largest points decline on record, while its intraday swing of 1,010 points was the second-largest in the history of the index, exceeded only by a 1,018-point swing on October 10, 2008. While the Dow’s temporary swoon during the flash crash was a harrowing reminder of the market collapse after the bankruptcy of Lehman Brothers in September 2008, the culprit this time was not overwhelmingly bearish sentiment but something altogether different – program trading.
What is Program Trading?
The NYSE has a rather dated definition of program trading, which it defines as (a) index arbitrage, or (b) the purchase or sale of a basket or group of 15 or more stocks as part of a coordinated trading strategy. Index arbitrage is one of the earliest program trading strategies, and essentially involves exploiting price discrepancies between, for example, futures on a stock index and the underlying stocks in that index.
More broadly, program trading can be defined as all aspects of computerized trading, including algorithmic trading, high-frequency trading (HFT) and quantitative trading. Program trading is generally undertaken by institutional traders at hedge funds, brokerages, and HFT firms, as well as institutions like mutual funds and pension funds. It typically involves the purchase or sale of various shares in very large quantities.
Not a New Concern, but a Growing One
Here’s an extract from a newspaper report on program trading: “It was a typical….day on Wall Street, a slow, nervous market, when suddenly the demand for big-name stocks jumped and started an apparent rally. Just as abruptly, prices tumbled and the Dow Jones Industrial Average ended the session with a loss.” The report goes on to say that at an investment firm, “executives knew the telephone calls would be coming in from irate clients demanding to know why the market had seesawed for no apparent reason. The executives also knew how they would answer: Computerized program trading done by a few big brokerages.” The report adds that program trading “has been widely blamed for injecting volatility into the stock market, frightening millions of ordinary investors and contributing to a malaise that has affected the securities business.” Sounds familiar? While the extract may appear to be from a 2013 or 2014 newspaper article, it is actually from the May 9, 1988 edition of the venerable Schenectady Gazette.
Program trading has come a long way in the quarter-century since then, thanks to the quantum leap in computing power and the proliferation of computerized trading strategies. These developments have also made disruptions caused by errant program trading more common. Apart from the flash crash, perhaps the best-known example of erroneous program trading is the one at market maker Knight Capital Group. On August 1, 2012, a technical glitch in Knight’s algorithmic trading systems caused misquotes in about 140 securities. The problem inflicted a loss of $440 million on Knight, taking it to the brink of bankruptcy and leading to its eventual acquisition by Getco.
Other prominent examples of problematic program trading include the cancelation of Bats Global Markets’ IPO in March 2012 on account of “technical issues,” and the May 2012 IPO for the Menlo Park, Ca.-based Facebook Inc. (Nasdaq:FB) that was dogged by technology problems and delayed trade confirmations. There have also been a number of stock-specific “flash crashes”, such as the one that affected the Woodlands, Texas-based Anadarko Petroleum Corp. (NYSE:APC) on May 20, 2013, when the stock plunged within seconds – for no apparent reason – from a price of about $90 to a penny, before recovering.
Like Anadarko, a number of stocks traded at absolutely crazy levels during the May 6, 2010 flash crash. For example, blue-chips like the Chicago, Il.-based energy company Exelon (NYSE:EXC), the Dublin, Ireland-based consulting firm Accenture PLC (NYSE:ACN), the Houston, Texas-based utility company CenterPoint Energy Inc. (NYSE:CNP) and the Boston-based brewing company Boston Beer Inc. (NYSE:SAM) briefly traded at zero on that day before rebounding, while the London, U.K.-based auction house Sotheby’s (NYSE:BID) briefly soared from the mid-$30s to $100,000 before closing at $33.
While the flash crash remains one of the most puzzling events in the history of theU.S.stock markets, a joint report by the Securities and Exchange Commission and Commodity Futures Trading Commission released in September 2010 shed some much-needed light on the subject. The report said that while volatility was unusually high and liquidity was thin on the morning of May 6, 2010 because of the European debt crisis, the flash crash itself was precipitated by a single $4.1 billion program trade generated by a trader at a mutual fund company.
What Caused the “Flash Crash”?
The trader used an automated execution algorithm to sell a total of 75,000 E-Mini contracts in the S&P 500 as a hedge to an existing equity position. The problem was that the algorithm was programmed to target an execution rate that was set to 9% of the trading volume calculated over the previous minute, but without regard to price or time. As a result, the sell program was executed in only 20 minutes, on a day when the markets were already under pressure. In contrast, a sell program by the same company on a previous occasion – which involved manual trading and several automated algorithms that took into account not just volume, but also price and time – required more than five hours to sell 75,000 contracts.
High-frequency traders and other intermediaries initially absorbed this selling pressure on May 6, 2010, building up temporary long positions. They then immediately and aggressively sold a number of E-Mini contracts to reduce their long positions. The report notes that during this time, HFTs traded nearly 140,000 E-Mini contracts or over one-third of the total trading volume, consistent with their strategy of trading a very large number of contracts but without accumulating an aggregate inventory of more than 3,000 to 4,000 contracts (long or short). The selling algorithm responded to this higher volume by increasing the rate at which it was feeding sell orders into the market. This adverse feedback loop created two liquidity crises – one at the E-Mini level, and the other in individual shares.