Financial Markets

High-Frequency Trading: A Sociologist's Take

Professor Donald MacKenzie weighs the risks of models, algorithms, market concentration and jitter

Friday, May 7, 2021

By Katherine Heires


For Donald MacKenzie, individuals' interactions with machines, mathematics and technology in competitive settings - and the risks that may ensue - are subjects of great fascination, whether in financial markets, bitcoin mining or nuclear weaponry.

MacKenzie comes at it not as a technologist, mathematician or engineer, but rather as a social scientist.

“I am a sociologist of science and technology,” the University of Edinburgh sociology professor explains in a video on YouTube, “and we like stuff, we like machines, we like material objects, and we like looking at the role these play in relations among human beings.”

Donald MacKenzie Headshot
Donald MacKenzie, University of Edinburgh

MacKenzie's books include Inventing Accuracy: A Historical Sociology of Nuclear Missile Guidance (1990); An Engine, not a Camera: How Financial Models Shape Markets (2006); Material Markets: How Economic Agents are Constructed (2009); and, with four co-authors, Chains of Finance: How Investment Management Is Shaped (2017).

His latest, Trading at the Speed of Light: How Ultrafast Algorithms Are Transforming Financial Markets, out this month from Princeton University Press, appears particularly well timed. Frothy market conditions and the recent GameStop short squeeze have politicians and policymakers raising new questions about automated trading and market practices such as short selling and payment for order flow.

The new book is based on more than 300 interviews with high-frequency traders, technology suppliers, exchange executives, regulators and others. In a recent interview, MacKenzie discussed the evolution of his academic work and his insights on risk in market operations and behaviors.

Why are the social studies of finance important and what can we learn from them?

Let me discuss my earlier research on the subject of financial models. What I have studied is how and why people adopt models and what are the consequences for markets, for example, of the adoption of the Black-Scholes option pricing model. Are its effects performative - the idea that language can effect change or social action? Can it shift patterns of market prices toward the postulate of the model? One effect of Black-Scholes was precisely that. But did we also see counter-performative behavior? An example of that would be the market crash of 1987.

It is these types of mathematical and technical aspects of markets that interest me, such as financial models or regulations that push everyone in markets to do the same thing at the same time. Such developments can lead to a dangerous situation. I hope that my work has helped to make people more aware of these kinds of effects.

What are your concerns about high-frequency trading?

There are a couple, both arising from the race for speed.

One is the degree of market concentration. There are two classes of HFT algorithms, market-making and liquidity-taking. When the bids and offers placed by the market-making algorithms are suddenly stale, there is a race to try to cancel them while the liquidity-taking algorithms try to execute against them. We then have an expensive race between algorithms, forcing people to turn to microwave transmission trading systems rather than fiber optic transmission.

There is some evidence that as a result of this race and its costs, HFT is consolidating quickly, with 40% of all U.S. equity trading volume handled by just two firms, Virtu and Citadel. This market concentration is worrying.

Your other concern?

It is that the speed race creates a certain incentive for firms to do their trading privately and bilaterally. The reason why some investment firms like these bilateral relationships is that you are not exposed to being picked off as you are in public markets. But in doing so, you are taking liquidity away from the public markets and impacting the price discovery process.

What steps might regulators take in light of these issues?

I think that batch auctions - an accumulation of orders that are executed simultaneously - are a very interesting and useful way to manage high-frequency trading risk. Certainly, some in the HFT business might be happy with its introduction, because the arms race is a large cost for them.

But I would caution that any intervention in a complex technical system - and today's financial trading markets are a complex system - can have unexpected side effects. One would have to introduce batch auctions in a tentative, experimental model with a pilot program for some stocks. This is something the Securities and Exchange Commission has done several times before. It would be worth experimenting with it.

During the GameStop trading frenzy, questions were raised about the impact of payment for order flow. Your thoughts?

While there is a great deal of suspicion in the practice - even the term “payment for order flow” sounds like a Mafia term - what is not so clear is whether or not investors are being hurt. Some researchers are more concerned about rebates and brokers sending orders to HFT traders for rebates. There are issues here that require investigation, but it's not clear that investors, who are able to trade for free due to these activities, are being disadvantaged by these practices.

What is your view of cryptocurrency trading and bitcoin mining and their market impacts?

There are scenarios where blockchain applications could make money transfers far cheaper, but I am not an enthusiast of bitcoin trading. This is fundamentally because I am an environmentalist, the design of bitcoin has an energy-intensive system at its core, and the energy consumption required by bitcoin mining is truly shocking. It is different for the ethereum cryptocurrency, which has been trying to move away from a bitcoin-style proof-of-work algorithm for mining and toward a much less energy-intensive proof-of-stake approach.

What are your thoughts on the predominance of trading algorithms in the markets?

There are two risks. One is that algorithms don't have common sense, and thus, regulators' insistence on circuit breakers in the case of extreme price moves is sensible. There are moments where prices really jump and have to be policed. But the idea of a trading system with almost no humans in the loop is an inherently dangerous kind of thing.

The other risk is the growing disconnect between prices and the fundamental value of stocks. Many trading algorithms are essentially looking for patterns. These are relative things, not absolute things, and so there is always the possibility of a market where algorithms are just endlessly chasing each other.

Having a few, old-fashioned, Warren Buffett-type fundamental investors does seem to be a valuable part of the ecosystem of a healthy market. We should not aim to have too much of a monoculture of algorithms in our trading systems.

Are there less obvious factors that risk managers should be aware of?

One non-obvious thing has to do with jitter, or congestion triggered by the time lag between data being sent and arriving on trading networks. One aspect of the technological advancements we've seen in trading markets is the reduction of jitter, and that is an improvement. Nobody likes random delays. Everyone wants a system that is more deterministic, more predictable, and faster.

But jitter is also an equalizer in trading markets. In a jittery system, the fastest firm doesn't always get the trade. Instead, the second fastest will get the trade, or maybe the 10th fastest will. So jitter keeps competition alive.

When jitter declines, as it does today, we are moving toward a winner-takes-all situation, and that means a much more concentrated business and a far smaller number of trading firms. In some sense, what we're seeing with the reduction of jitter isn't an improvement, and its removal will bring about side effects that we need to worry about.

Katherine Heires is a freelance business journalist and founder of MediaKat llc.


BylawsCode of ConductPrivacy NoticeTerms of Use © 2024 Global Association of Risk Professionals