An Expert Explains High-Frequency Trading

Editors note: This podcast was recorded at the 21st Annual Financial Markets Conference [link], hosted in May by the Atlanta Fed. Mani Mahjouri was a speaker at the conference.

Kris Gerardi: Hello, my name's Kris Gerardi. I am a financial economist at the Atlanta Fed, and I'm very happy to be joined by Mani Mahjouri, the chief strategist and chief investment officer at Tradeworx, a financial technology company that is based in Redbank, New Jersey, which specializes in high-frequency trading. Mani, thank you very much for joining us today.

Mani Mahjouri: Thanks for having me, Kris.

Gerardi: Can you define for our audience high-frequency trading in sort of basic terms?

Mahjouri: Sure. If you look online, you're going to find things like a computerized trading using proprietary algorithms. You're going to find things like executing a large order by computerized algorithm by splitting that order into smaller pieces and executing them according to some schedule, typical VWAP [volume-weighted average price] or some other format of getting a position in a security. The second type is really not executing a set order but you're looking for other opportunities in the market, sort of more what a proprietary trading firm does.

Man pointing to clear screen with trading data

The way I think about it is that high-frequency trading is the logical disruptive evolution of what was open outcry. At the end of the day, these opportunities that these proprietary traders are really looking at arise by providing liquidity through a warehousing of risk between buyers and sellers. So I like to emphasize that there's really nothing new about that particular function. Whether you're a specialist in a stock on the exchange floor or you're a local in a futures pit, basically the function is take the other side of a trade and hold it in hopes of getting out at a better price and use all the information at hand to make the most competitive price on both ends, and that's the service you're providing to the market.

The thing about high frequency is that it kind of infers a very short holding period, and that's one of the big things that come into people's minds. The reason for that is it just reflects the average time it takes to match a buyer and a seller in predominately what is the U.S. equities market. By comparison, the real estate market is a low-frequency market. So characteristics are extremely high transaction costs and the need to locate a buyer if you're home owner to get liquidity. All the supply-and-demand issues, bubbles and crashes and things like that, happen in those markets just as well as high frequency, and I think it's more a function of the concept of warehousing risk in our markets.

Gerardi: Excellent. Can you give us some background about high-frequency trading, in terms of how it started and what were the technological innovations that made it possible?

Mahjouri: Sure. High-frequency trading arose from three primary simultaneous happenings. One of them was, as you mentioned, advances in technology. Basically, for the first time, it was actually possible to meticulously record every meaningful atomic action associated with executing a stock, and that's kind of where it started. That led to great efficiencies to be gained in terms of being able to assimilate lots of information into a trading decision. So, you went from a scenario where you're a specialist and your job is really to understand the micro supply-and-demand forces that are governing that particular name or set of names that you're responsible in, to being able to do that in the context of lots and lots of names and being able to leverage the fact that we understand that a lot of these securities are subject to a lot of the same forces of risk underneath. So, by being able to develop strategies that were able to very efficiently offset and neutralize volatile risk factors.

So that's where technology helped us, because it allowed us to get access to the data very, very quickly, very efficiently, and very impartially. But that alone wouldn't have been enough; it came along with major changes in regulation. Primarily, the way I like to think about the changes in regulation is that there was a sense of if somebody has a better price out there, our intent is going to be to honor that better price, provided they meet certain minimal requirements that promote stability.

Then finally, we just had market conditions that were very ripe for the proliferation of this type of trading. What I mean by that is we had an unprecedented demand for liquidity on the tail end of the banking crisis. We had a very underbuilt supply of liquidity following years of relatively low volatility and banks that were capital-constrained. So you had very low supply and very high demand, and so the price of liquidity as a commodity was very high, so it did massively favor firms that had ventured into this more efficient way of providing that service and, hence, became a major force in the markets.

Gerardi: So HFT [high-frequency trading] is obviously proven to be a pretty lucrative, profitable business. From a social standpoint, though, can you speak a little bit about the benefits to society of high-frequency trading? High-frequency trading has often gotten knocked in the popular press, but at least from an economic standpoint, there is an argument that it makes markets more efficient and more liquid; it provides more liquidity to the market. What's your view?

Mahjouri: I think that at its core, technology made it possible to efficiently access orders of magnitude, more information about stocks and securities. That allowed you to build significantly more competitive pricing models for those securities, and the result of that was a massive collapse of bid-offer spreads, helped by the fact that the minimum tick increment was lowered through regulation. In that sense, it's made markets significantly more liquid. I think that it's also promoted a more transparent paradigm than we had before. That doesn't mean that every order is scrutinized, because maybe there are just too many for the system we built for today to be able to go through and methodically characterize—but the fact is that all that information is there and can be studied now, or years from now, and that's something that we never had before. Ultimately, I think that leads us a much more efficient and better market in the future.

Gerardi: So the flip side of that is that there are some potential costs to HFT. One of them is that it could potentially contribute to additional financial instability by making episodes like the flash crash in 2010 and the recent August 2015 stock market plunge, much more likely. Do you believe this view has any merit and would you be worried about something like that in the future?

Mahjouri: First of all, I believe it's pretty widely accepted at this point that the flash crash arose from a fat-fingered error. So while there was some electronic [error] going on there, it wasn't really necessarily an algorithm, as much as a system that we hadn't fully understood yet and hadn't built adequate safeguards around. Having said that, the current state of markets is that they've become so connected and technology's allowed them to become so much more connected, that even in the state of an unbelievable shock—I think that a trade that was the size of the trade that set off the flash crash would never have been allowed in a trading floor dominated by humans. The fact that it was there, the fact that that risk was so quickly taken into all these other neighboring assets, and starting out with all the different equities, but then moving into all different types of risk factors. It's like dropping a pebble into a pond, the ripples go out, but if the pond is big enough, that doesn't result in a tsunami for anyone. Eventually that energy dissipates. In my opinion, the availability of all these algorithms and their intense search for coming up with profitable trades led to a dampening of these types of effects, rather than a...

Gerardi: Interesting. So you're saying that it actually probably went the other way, in the sense of developing...

Mahjouri: Yes. We look at individual stocks trading, and we see this in individual stocks. People like to use the word "flash crash" but on any given day, you're going to be able to find the stock that drops by a lot or goes up by a lot. Sometimes, that's in response to news, but sometimes that's just in response to a big investor who decided to change their preference. When you see that and you look at how much liquidity is actually in the order book and where the order book finishes after the perturbation has dissipated, what we usually find is that the price impact is smaller than what we would have expected to see if there were no algorithms laying off that risk to other correlated names. If you think about it, it makes a lot of sense. What the algorithm does is it says this stock is plummeting in price—or soaring in price—and I can't figure out any reason why that is, so I'm going offset this risk on to things that are almost the same. So, effectively that's a really, really powerful way to warehouse risk. I think that the benefits of that do pretty routinely materialize themselves in a dampening of volatility.

Gerardi: So the share of U.S. stock trades accounted for by high-frequency trading has increased fairly dramatically over the past decade to more than 50 percent by many accounts. Do you think there's a saturation point, whereby the benefits of high-frequency trading in the financial markets maybe begin to decline? You think we've reached that point, I guess.

Mahjouri: Yes. Here some things I look at that make me think that we've reached that point. I think that the benefits of an innovation start to decline when the new incumbents pursue entrenchment strategies designed to discourage newcomers from entering the arena. So, prior to turning on these mics, we were talking about speed and how we went from piecing together existing fiber routes to digging trenches across the country to lay straight fiber to now sending signals through microwave towers and laying new trans-Atlantic cables, doing these very, very expensive technological investments. When you look at that, it makes sense if you're a leader in that business because you definitely want to build a moat around your business; that's understandable. But if you look at it from the perspective of society, that's a tremendous amount of capital that's being invested by a lot of places. In my view, maybe that would be better spent making a more competitive price, as opposed to spending it on speed so that you're first in line when your opportunity arises.

Gerardi: I see. So you feel like the cost of entering the HFT business has now become so great, that you're a little bit afraid that there might be some monopoly and the industry is becoming less competitive, and that's...

Mahjouri: Well, there are two competing forces. The technology itself becomes commoditized through time, so having a tier 2 system is a lot cheaper now than it used to be. Tier 2 is a lot closer than tier 1, but if you're solidly entrenched in tier 1, then you have a very, very clear budget that's coming from the revenue that you're creating from this advantage that you have that goes back into sustaining that advantage. My view on that is that, to the extent that's going on, that feels like waste. There's other signs of that, outside of what people are spending on technology. If you look at any exchange, they're going to offer tiers to the biggest participants.

There are a lot of reasons that are stated for that, but here's an economic one, which is: if you're in the business of being an exchange, you make money off of transactions. So you're in a position where you encourage your biggest customers to focus is on doing the maximum amount of transactions. If you're the biggest customer, by virtue of doing so, you're artificially increasing the cost that a newcomer would have to pay. So by doing more volume and perhaps printing trades that you might not otherwise have any reason to print you're solidifying an advantage.

The third thing is that data costs are getting more and more and more expensive. I think that the exchanges are finding more and more "premium products" that are a "must have" if you want to be on tier 1, so you see those costs escalating as well.

It feels to me—and I'm not the economist, but—when everyone's on the same footing, the best idea wins. So people focused on creating the best idea and that leads to lots of good ideas and that leads to everyone's benefit. But when people start to focus on solidifying a particular advantage, then you're heading into a two-Tiered market, where there's an inside group and an outside group. For what it's worth, that's how stock exchanges were a hundred years ago, before HFT, and it looks like it is our natural tendency to try to build a moat around our castles, and so there is a danger that we could head towards that direction again.

Gerardi: As HFT technology becomes more widespread, and the cost of acquiring greater and greater speed becomes higher, do you think that this is going to adversely affect HFT profits and do you see this driving more HFT firms out of business, and if so, how is a firm like Tradeworx, for example, going to adapt?

Mahjouri: So our strategy is to try to find the things about the market that are very stable and then try to be the best at them. So for us, that's alpha; for us, that's like, "Let's come up with the absolute best pricing model," and assume that the market's going to head towards some direction like that. That's also talking our book, because that is what I believe is our core competence. We don't have the same sort of engineering budget—billions of dollars or hundreds of millions of dollars—that can go into muscle. We have to focus on other elements of what we're looking at.

The one thing that's pretty interesting is that the universe gave us the speed limit, and that's the speed of light. That's how fast you can transact and that's the speed limit everyone knows about. There's another speed limit, which is the speed at which exchanges are able to assimilate all this data. It took me awhile to realize that it may not be a fully deterministic system, given how many different...if you think about how NASDAQ generates a billion records a day that they print and there's other records that they don't even print that are also pretty big. So, if you think about how the entire world wants to do something in the U.S. equity market and it goes through one of these 12 exchanges.

The amount of machinery that's in between the decision to buy yourself a security and then to the point where it's matched. Even if you optimize—and believe me, we have, we understand elements of hardware and how processors work that I never even thought about. We do understand that, but there's a part of that which is nondeterministic and once you break into a certain level of the speed, then you can't predict that you're always going to be first unless you have a significant advantage. Innovation does get out and so people do get to access it. There is a first-mover advantage but the thing is that there is a compression. So, we're not at the speed of light yet, but we're inside of perfectly predictable exchange technology. There's a probability distribution governing if I send my order in at the same time as you send your order in, who's going to win that trade, even if we go down to understanding everything we know about the systems.

Gerardi: Let's wrap it up with one more question. I'm curious to get your take about what role, if any, you see in terms of regulation. Do you think there is room for regulation? Is it necessary, and if so, what kind of form would you like to see it take?

Mahjouri: Sure. I'm a huge, huge proponent of data-driven decision making. I think that that hasn't always been possible in the past because of how you gathered that much data from an open outcry pit. How do you even quantify all the observables? But now we have data, so we see things like a tick size pilot that the SEC is planning to do in October, where you can actually ask the question, "What is going to lead to the best trading experience?" In some ways, the reason why we have 60 different dark pools is a reflection of that as well, which is that different investors have different preferences for how they want to meet each other in the marketplace. We have Reg ATS [alternative trading systems], which sort of says basically you could try things because we want you to innovate, so we do see that continuing as well.

Gerardi: All right. Well, that wraps it up. Thank you very much, Mani, for joining us here today and giving us some very unique perspective and insight about the high-frequency trading business.

Mahjouri: Well, thanks a lot for having me. It's been a pleasure and this conference is actually really, really interesting and I'm really, really happy that I came.