Spread Networks, a Mississippi-based telecom provider, announced with fanfare in June 2010 that it had completed a high-speed fiber optic cable between Chicago and New York. The company reportedly spent $300 million constructing an 825-mile route through the Allegheny Mountains, to make the cable as short as possible. The Herculean achievement cut 0.003 seconds from the time it takes a message to make the round trip, from 16 milliseconds to 13 milliseconds.
Such investments are part of a technology arms race in financial trading. Critics say the high-frequency trading (HFT) this technology enables causes market instability—the 2010 Flash Crash, for instance—and costs long-term investors billions of dollars. But Eric Budish, associate professor of economics; John Shim, a PhD student at Chicago Booth; and the University of Maryland’s Peter Cramton argue that the HFT arms race is a symptom of market instability and not a cause—and that the public debate has overlooked a fundamental flaw in the structure of financial markets: continuous trading.
Most securities exchanges operate continuously— traders can buy or sell any time during trading hours. Orders that provide liquidity are governed by price-time priority—the highest bid or lowest offer at any given time is executed first. When two identical offers arrive, as tracked by a continuous limit order book, the one that arrived first is executed first.
But the continuous limit order book doesn’t actually work in continuous time, say the researchers. Pricing relationships between related securities that function at what the researchers define as “human time intervals”—a second, minute, or hour—break down at millisecond speeds.
The researchers look at a pair of securities that track the S&P 500 index: the SPDR S&P 500 exchange traded fund, which trades in New York, and the Chicago-traded E-Mini Future. Using 2005–11 data from the New York Stock Exchange and the Chicago Mercantile Exchange, they find that the correlation between the two securities is nearly perfect at human time intervals, but essentially zero at higher frequencies.
Arbitrage opportunities are, therefore, inherent in the continuous limit order market. However fast computers get, prices for related securities cannot change at exactly the same moment. When one security’s price changes and the related security’s price has not yet responded, a profitable opportunity emerges: buy the cheap one and sell the expensive one. Such opportunities drive the technology arms race.
High-frequency traders argue that speed competition ultimately benefits fundamental investors. The researchers conclude otherwise, emphasizing which high-frequency traders are on the wrong side of these trades: liquidity providers, trading firms that stand willing to buy from or sell to all comers, including fundamental investors and trading firms.
Liquidity providers have as much incentive to invest in speed as other high-frequency traders, yet they often suffer when prices jump and their quotes become stale. In the race to react, a liquidity-providing trader tries to adjust his prices quickly, while others simultaneously try to exploit his stale quotes. Because there are more of the latter than the former, the lone trader often loses out. To accommodate those losses, he widens his bid-ask spreads, making fundamental investors pay more to trade.
The researchers propose instead that exchanges run batch auctions at frequent time intervals, such as once per second. In batch auctions, exchanges collect and aggregate orders, and execute those at the price where the most bids and offers match—where supply meets demand. Most exchanges already run these types of auctions for the market open.
In a batch market, the researchers write, “if multiple traders observe the same information at the same time, they are forced to compete on price instead of speed.”
The result, they argue, would be narrower bid-ask spreads, improved liquidity, increased market stability, and billions of dollars of investors’ money saved.