The pop culture view of high‐frequency traders (HFTs) has been formed primarily by books and movies. I previously reviewed Michael Lewis’s best‐selling book Flash Boys (“Finance According to Non‐Academics,” Spring 2015), which scrutinized the front‐running trading strategies of HFTs who, according to Lewis, “rigged” the market by shaving nanoseconds off their execution time. Sony planned to make the book into a movie, with the screenplay by Aaron Sorkin, but the project got stuck in development hell; Netflix is now giving it a shot. Another attempt to put HFTs on the big screen was 2019’s The Hummingbird Project, starring Jesse Eisenberg and Salma Hayek, though it failed to break $1 million at the box office.A less pop‐culture effort is sociologist Donald MacKenzie’s new book Trading at the Speed of Light. It applies a material political economy approach to the corporate history of trading. He previously authored 2006’s An Engine, Not a Camera: How Financial Models Shape Markets.MacKenzie’s research for this book extends back over a decade. He recollects an early New York Times story about HFTs: “It is the hot new thing on Wall Street, a way for a handful of traders to master the stock market, peek at investors’ orders and, critics say, even subtly manipulate share prices.” His primary research method was in‐person interviews. He conducted 337 in all, in all the trading centers of the world, including Chicago, New York, London, and Amsterdam. He spoke with founders and employees of high‐frequency trading firms; staff of exchanges and clearinghouses; algorithmic trading practitioners; suppliers of communications and technology; regulators, lawyers and lobbyists; and dealers and brokers. Nearly all of his quotations in the book are from anonymous sources, who are branded with initials for identification.Before automation and electronic trading / Before addressing today’s fast‐moving, high‐tech trading world, MacKenzie commits several early chapters to the bygone days of the trading pits and open‐outcry trading at Chicago’s Board of Trade and Mercantile Exchange (CBOT and CME), reminiscent of scenes from the movie Trading Places. As a bridge between the eras, the reader hears the story of Leo Melamed, who started his career as a runner in the trading pits in 1953 and whose first job entailed delivering orders to floor traders. MacKenzie notes that Melamed, who ultimately became chairman of the CME, “had not always been an enthusiast of electronic trading,” but a 2012 photo in the book shows an aging Melamed at the CME’s data center.The transition to automated trading was a more abrupt transition for some players in the industry than others. “A bank or even an institutional investor might no longer have to pay a broker simply to bring its orders to market but could itself enter its orders at a computer terminal,” MacKenzie writes. “In contrast to brokers, traders might hope to continue to flourish in electronic markets.”Globex and Aurora were two early competing systems for electronic trading of futures, led by the CME (managed by Melamed, jointly with Reuters) and CBOT, respectively. The visual representation of the two systems differed dramatically, with Globex having a mundane screen view showing prices, bids, and buttons, while Aurora, which ran on an Apple Macintosh, had traders visually represented by an icon to “simulate a trading pit.” Globex won the technology battle — though, interestingly enough, because Aurora “would overburden the then‐available bandwidth of global digital communication.”The beginning of the race / A parallel set of movements was afoot on the share‐trading side of the market. MacKenzie digs into the archives of the New York Stock Exchange to produce a sample of its order books, which “were until the 1980s handwritten on pre‐printed forms.” One pioneer on the share‐trading side was Instinet, which was one of the earliest electronic trading systems in the 1970s and focused primarily on allowing institutional investors to trade with each other via teletype machines. Another was Automated Trading Desk, which came along in 1989 and was founded by then–Rutgers University finance professor David Whitcomb after Instinet neglected to adopt some of his proposals for early‐stage trade execution algorithms.By the late 1990s, another key player in the development of the high‐frequency trading industry, Island, entered the fray by appealing to day traders and so‐called “bandit” traders who searched for and profited from stale price quotes. Island was known for perfecting its real‐time software called Watcher. MacKenzie explains that itbegan as just a program to watch for incoming executions and keep track of a trader’s position, [but later evolved into] a full‐blown trading system that gave traders market news and up‐to‐date information on NASDAQ dealers’ changing bids and offers, permitted traders to enter orders, and allowed them to send messages to other individual users or to all the traders using Watcher.Trades through Island’s system were executed in less than 10 milliseconds (thousandths of a second). Island was also a leader in co‐location, “encouraging trading firms to place their servers in its Broad Street building, even in the basement alongside [its] computer systems” so as to gain a trading advantage by having a shorter distance for its electronic commands to travel to the NYSE.For high‐frequency traders, tech‐ nology combined with algorithms became the all‐consuming focus of trading to achieve advantage.The race reaches the outer limits / For HFTs, technology combined with algorithms became the all‐consuming focus of trading to achieve advantage. MacKenzie puts the technology in perspective by comparing how quickly each of the evolving methods was able to transfer data between the CME’s data center in Chicago and the NASDAQ data center in New Jersey. For Spread Networks, known for buying up rights to lay direct fiber‐optic cables, the required time for transmitting a trade in 2010 was 6.65 milliseconds, which bested the previous technologies’ 8 milliseconds. Glass‐fiber technologies reduced that time to 5.79 milliseconds. Microwave technology improvements cut it to 3.98 milliseconds by 2016. The reference in the book’s title to “trading at the speed of light” refers to the theoretical Einsteinian constraint to transmit at the speed of light in a vacuum, which is 3.94 milliseconds. These dueling technologies were the focus of The Hummingbird Project.Algorithms are, in MacKenzie’s words, “a recipe that achieves a goal or solves a problem in a finite number of precise, unambiguous steps.” (See “Algorithms: The Life Blood of the FANGs,” Fall 2020.) In the case of high‐frequency trading in particular, algorithms go through the requisite steps for bidding to buy or offering to sell shares or other financial instruments, or for canceling or modifying an existing bid or offer. The algorithm is “taking in the world’s information and being able to translate that to predict the next [price movement].” The secret, according to MacKenzie, is revealed in an exemplary case of the “making” and “taking” strategies for all the choices to compete on price for bids to buy (ranging from $29.45 to $29.49) and offers to sell ($29.50 to $29.54) for a typical security. The strategy boils down thusto post bids at higher prices and offers at lower prices than others do … [and] getting good places in the queue … [because] an order at or near the back of the queue is at greater risk of being executed in adverse circumstances.This risk materializes when bids are being executed or c
Authors
- Published in
- United States of America
