What's faster than light-speed trading? Inside Wall Street's sketchy early access schemes
In February of this year, a group of UC Santa Cruz physicists made a surprising discovery. Apparently someone had shortened the distance from New York to Chicago. In August of 2012, information made it from Wall Street to Chicago's commodities exchange three milliseconds faster than it had just two years before. Someone was making the trip in just over 4 milliseconds, edging up to the hard limit set by the speed of light. They had ditched sluggish fiber optics for line-of-sight microwave transmitters, a network of towers sending electromagnetic signals through the air, reaching Chicago just fast enough to always have the first trade. It still isn’t clear who got there first, but high-speed services like McKay Brothers, Tradeworx, and Epsilon Networks are prime suspects. The research team estimated the five-year cost of the system at half a billion dollars.
They had ditched sluggish fiber optics for line-of-sight microwaves
In markets where the fastest trade often wins, the race for speed isn't surprising — but it's had some unpleasant consequences. When information travels too fast for human monitoring, it opens markets up to more predatory deals, operating while regulators are blinded by the speed of the trades. While some traders trust that they'll arrive first because of high-tech systems, another class are paying for early access or outright leaks, gaming the market in secret.
One example occurred earlier this month, when the Federal Reserve announced its plans to continue a long-running bond-buying program. The news was announced in a lockdown room at Fed headquarters in Washington, DC, and allowed to leave the room at exactly 2:00PM EST, timed to the millisecond. Somehow, markets in Chicago began trading on the information instantaneously, without waiting the few milliseconds that it would have taken for information to travel via microwave. Clearly someone leaked the information — but so far, they're getting off scott free.
Just like poker, a good hand isn't worth much without a sucker betting into it
It’s a potentially lucrative cheat. The Fed news was bound to make the market shoot up, so anyone who thought they had it first would buy big, wait for the bounce, and then sell at the higher price. It’s a good trick, but it only works if you’re the first one in, before any other big buys have driven up the market price. And for the scheme to really shine, you’d like someone else buying big a few seconds after you, so you can unload your shares quickly at the higher price. Just like poker, a good hand isn't worth much without a sucker betting into it.
"I have a feeling we're not going to ever know."
In the case of the Fed scheme, that’s exactly what happened. A huge volume of trades were placed in the milliseconds after the announcement, presumably from firms who’d shelled out hundreds of thousands of dollars for light-speed info transit to Chicago. They had every reason to think they were the first ones on the commodity market with the information — but they weren’t. Someone else had gotten there first, bidding up the market in advance and then selling off to the light-speed traders, making hundreds of thousands of dollars in a matter of milliseconds. For days, no one even knew it had happened.
Eric Hunsader, founder of the market analysis firm Nanex, noticed the discrepancy and sent it to CNBC as evidence of fraud, but so far the response from regulators has been minimal. "It does not take this long to figure out who did what, and we still don't know. My guess is they just want this to die," Hunsader says. "I have a feeling we're not going to ever know." Since the 2:00PM embargo had lifted, it's hard to make an insider trading case, and when the entire case hinges on a few milliseconds, the SEC is often wary of getting too involved.
In a light-speed market, some investors simply don't know they're being beaten to the punch
Economist Ciamac Moallemi, who studies latency in financial markets at Columbia, finds the idea of early access scandals puzzling, saying the limited time frame should limit the available profit. "One might initially be able to make money with such information," Moallemi says, "but over time other investors would become aware and liquidity would dry up." Moallemi attributes the established instances of firms paying for early access as yet another irrationality of the market, saying, "it's an open question as to why this is valuable." The most likely explanation is that, in a light-speed market, some investors simply don't know they're being beaten to the punch.
The crucial thing is not just getting there first, but doing it without other traders realizing they’ve been beaten to the punch. Secrecy is a necessary part of the scheme, and that protective layer of secrecy has proven hard to break through, especially since the rare whistleblowers on early access schemes have found little protection from the industry. In January, the FBI announced an investigation into financial newswires, on evidence that someone was giving paying clients early access to the University of Michigan's consumer confidence index. The tip came from a Reuters employee named Henry Rosenblum, who had tipped off the FBI after he became concerned the company was engaged in insider trading.
"People who were getting that number thought they were getting it first."
The Reuters scheme was more complex than usual, offering three full tiers. The man on the street would hear the consumer confidence numbers at 10:00, but documents revealed that Reuters was actually selling the information twice, both to paying subscribers at 9:55 and two seconds earlier to "ultra low-latency" subscribers who shelled out for the head start. Crucially, most of the clients paying to get the number at 9:55 had no idea there was an ultra low-latency service running, which made them easy pickings for anyone who got the number early enough. Financial reporters scoffed at the idea that Reuters’ data service could be illegal, chalking it up as business as usual for a financial news service. But the numbers tell a different story. Hunsader's Nanex reporting shows that trading on the University of Michigan numbers dropped nearly 100-to-one after the scandal. Once the ultra low-latency secret was out, Michigan’s consumer confidence numbers seemed too risky to bet on.
By then, Rosenblum was a distant memory. He was fired by Reuters some time after the allegations came out, and is currently suing for whistleblower protections in Manhattan Federal Court. For many working his case, it's a sign that the regulations covering information transit are hopelessly outdated. "People who were getting that number thought they were getting it first," says Jesse Rose, one of the lawyers working on Rosenblum's case. "If the people who are running these things are going to claim that this is the free market, then I think they need to provide an explanation to their investors."
Unlike the Fed announcement, the University of Michigan numbers are private data, so there are few restrictions on how it's released. At the same time, it's hard to imagine how similar schemes could be uncovered without work like Rosenblum's and Hunsader's. Two seconds is a lot of machine time, but it's just a blip in human time, easily explained away as the work of overeager traders, or a clever algorithm trying to anticipate the market. Some, like Rose, think the answer is better disclosure, but for others, it may not even be a problem worth fixing. Moallemi doesn’t think financial regulators should bother with early access violations, since they don’t have much effect on individual investors: "it's basically a game between high-frequency traders." But for those playing the game, having the fastest tech might not mean getting there first.