U. of Michigan Comment CFTC Comment Letter About Automated Trading



Three months ago the CFTC proposed rules designed to maintain the integrity and modernize the regulatory framework of derivatives markets. Titled 78 FR 56542, this Concept Release recognizes that:

The operational centers of modern markets now reside in a combination of automated trading systems (ATSs’) and electronic trading platforms that can execute repetitive tasks at speeds orders of magnitude greater than any human equivalent. Traditional risk controls and safeguards that relied on human judgment and speeds, and which were appropriate to manual and/or floor-based trading environments, must be reevaluated in light of new market structures. Further, the Commission and market participants must ensure that regulatory standards and internal controls are calibrated to match both current and foreseeable market technologies and risks.

The deadline for comments on the CFTC’s Concept Release just passed on December 11th, 2013. This morning we want to highlight one specific comment letter written by a team at the University of Michigan at Ann Arbor, which you can view here.

The team at the University of Michigan is diverse, and includes computer scientists, law professors, and finance professors. In fact, Professor Michael Barr pepreviously served as Assistant Secretary of the Treasury in 2009-2010, and was an architect of Dodd Frank. The team approached their study of the workings of our modern automated markets in a detailed way, and note that our markets have evolved into an arms race, where the race is destructive and involves many interconnected asset classes, with the potential for unforeseen dangers.

We knew we were going to enjoy the University of Michigan paper, as it was immediately attacked as flawed by none other than the Chairman of the FIA European Principal Traders Association (HFT Lobby) – Remco Lenterman, who wrote a blog post detailing The Myths Around Latency Arbitrage. Mr. Lenterman assures you that there is no latency arbitrage, as:

The fact that the people watching the race from their TV set are seeing the result later than the people physically located at the track does not mean that the latter are able to profit from this information advantage. Why not? Because they are both observing an event that already has taken place, i.e. history. The one that is physically located at the track is merely observing it sooner after the event has taken place than the one sitting at home. They cannot miraculously make a bet on the winning horse in between the time that they have observed the result and the time that the people at home have observed it.

Of course Remco has it wrong; there absolutely is latency arbitrage, as we have written about many times – including when we also wrote about horse races, “past posting”,  HFT, and the movie, The Sting. And in case you don’t believe Themis, you can pay attention to HFT firms like Jump Trading, who are specifically setting up groups to practice latency arbitrage by name! Anyhow, we are digressing…

The University of Michigan comment letter discusses their research, where they specifically found that in continuous trading mechanisms, automated traders that arb the pricing and latency differences that exist among fragmented trading venues actually degrade market quality. They instead advocate for  periodic call markets, where orders are matched to trade at fixed periodic intervals, and not continuously. Their research shows that such call markets align the rate of trading with the rate of economic information; they propose call markets that take place every second.  From their comment letter:

The call market veils all submitted orders within each bidding interval, as in a sealed-bid auction. Clearing prices and price quotes are available only after each market clear event. This opacity guarantees that all market participants ultimately receive|and respond to|the same information regardless of latency, thereby nullifying the benefit of relative speed within the interval. Periodic clears every second, for example, would be virtually imperceptible to investors but would substantially curb HFT activity. We submit that there is a range of settings for this clearing interval that would effectively deter latency races while providing sufficient market timeliness for high-functioning price discovery.

Please read the University of Michigan’s comment letter. It is written with the expertise of policy makers, experts in law, engineers, and computer scientists who have studied the operations of markets for dozens of years.