Chicago Booth Review Podcast How to Fix the Stock Market

Eric Budish: Providing that useful service to the rest of the market is made more complicated and made more expensive by the fact that I have to constantly be vigilant about being a millionth of a second too slow to react to some pricing signal. So, I'm constantly getting picked off. I'm constantly getting traded against at a slightly stale price. That's an expense of my business that I ultimately pass on to the rest of the market.

Hal Weitzman: In recent decades, stock markets have increasingly been speeding up with large numbers of high frequency traders executing trades at speeds up to billionths of a second. Does that reward the most powerful computers at the expense of retail investors? Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I'm Hal Weitzman, and today I'm talking with Chicago Booth's Eric Budish, who argues that the incentives for trading stem from the flaws in the way financial markets are designed. At the same time, it's become cumbersome and complicated for institutional investors to execute trades as they would like to. With both challenges in mind, Budish outlines two simple ideas from his research that could transform stock markets for the better. Eric Budish, welcome to the Chicago Booth Review Podcast.

Eric Budish: Oh, it's fun to be with you, Hal.

Hal Weitzman: Let's talk about high frequency trading. Everyone knows that high frequency means fast. How fast is high frequency trading?

Eric Budish: So, high frequency trading, I like to distinguish between trading advantage that comes from smarts versus trading advantage that comes from speed. So, smarts, knowing something the rest of the market does not know, speed being faster than the rest of the market to act on a piece of information. And in modern financial markets, speed means millionths of a second, sometimes even billionths of a second. It's astonishingly fast .every time there's a photo finish and a hundred-meter dash like there was in the Olympics last summer, I like to remind my students and colleagues that high frequency trading is a thousand times faster than that. It's really astonishing,

Hal Weitzman: But less fun to watch probably.

Eric Budish: It probably is.

Hal Weitzman: And if we look at the stock market, US stock markets, how big is high frequency trading? How much of the market is high frequency?

Eric Budish: It's hard to measure. So, the best measurement I've been able to do is in the UK financial markets. And we're able to do that with help from the UK Financial Conduct Authority, which facilitated a study I did jointly with two researchers there at the time, Matteo [inaudible 00:02:36].

Hal Weitzman: That's the regulator.

Eric Budish: ... and Peter O'Neill. Yeah, they're like the analog of the Securities and Exchange Commission in the US, but with some other kind of an amalgam of SEC, the CFTC that we were talking about-

Hal Weitzman: Regulates commodities. Yeah.

Eric Budish: Yeah, the Futures and Trading Commission. They're the all-encompassing UK financial markets regulator. And we got this kind of data that's never been available before, including in the United States, that allowed me as the researcher to see races playing out in the UK equity markets. What I mean by a race is multiple trading participants, multiple market participants trying to execute the same trade at essentially the same time.

Hal Weitzman: So, that is watching the a hundred meter-

Eric Budish: It's exactly like watching the a hundred-meter dash. And in the UK financial market, I could see in this data set, I could see the winner of the hundred-meter dash and I could also see all the losers. Because I could see the firms that were trying to get the trade and just missed. The modal time between the winner and the first loser, the modal time difference was between five to 10 millionths of a second. And this was back in 2016, so now it'd be even faster.

In the US financial market, the kind of data that's more readily available, you see the winner of the race, but you don't actually see that there were lots of other firms trying to do the same trade at the same time because their trade requests wouldn't leave an empirical signature that a researcher could see. Essentially would get bounced back with an error message that says you're too late for the trade. But as a researcher you don't see that. So, we were able to see the races, see just how fast they are, and it was about 20% of trading volume in the UK. And so, I would speculate that-

Hal Weitzman: So, you're talking there about the executed trades.

Eric Budish: Yeah, 20% of every trade executed in the UK stock market was a high frequency trading race where the modal race again, so the most common race, the difference between the winner and the first loser was between five and 15 millionths of a second. Astonishingly [inaudible 00:04:44].

Hal Weitzman: That's a lot. It was one in five trades, but it's not as dominant as you might've thought. A few years ago, Michael Lewis had his book out about Flash Boys and about high frequency trading. And there was a moment when there was uproar about high frequency trading, but you would've assumed from that that it was bigger than 20%.

Eric Budish: Oh, that's fascinating. So, 20% to me is a big freaking number.

Hal Weitzman: It is.

Eric Budish: I won't curse on air. Michael Lewis quoted a different number that high frequency trading firms touch 50% about trading volume. Roughly speaking, every trade involves a counterparty. Most trades involve a counterparty that is a trading firm as opposed to an investor. The modeling paradigms, researchers like myself and others in the field use to think about how financial markets behave, involve investors trading based on their long horizon view or maybe even pretty short horizon view of where stock's going to go, and trading firms that make markets for them. And trading firms are providing a useful service in financial markets. They're offering to buy and sell, make a market to all comers. And my research is about ways in which the provision of trading services has gone awry in the last couple of decades. Or at least partly awry because of the underlying market structure, which incentivizes a particular form of market making based excessively on speed. And we can talk about that.

Hal Weitzman: I'm interested just on a general level, what's wrong with high frequency trading? Or is there anything wrong with it? Because the implication of the Lewis book is. So, let's forget about Michael Lewis for a second, but just what's your view? Is it negative for the market?

Eric Budish: So, my view on high frequency trading and algorithmic computerized trading more generally, so I'll use the phrase algorithmic trading, is that there's a positive and a negative aspect to it. And the positive aspect is in lots of swaths of the economy, computers are really helpful. They make services and make lots of aspects of the economy more efficient. Computers are smarter than humans with lots of stuff. There's this additional dimension of modern financial markets that's very excessively focused on speed. And I guess the distinction I'd make going back, I used the phrase smarts versus speed earlier, the component of proprietary trading firms activity that's based on being smarter than the rest of the market and making money from being smarter than the rest of the market. That's, I think, a productive part of capitalism, a productive part of just stock market efficiency. The part of high frequency trading that is about being a millionth of a second faster than the next guy to capitalize on an obvious piece of information, that part I think is counterproductive.

And what my research argues is that that aspect of proprietary trading, you can think of as a tax on market liquidity, a small tax on the order of a basis point, but it's a tax on market liquidity. And that's the 20% of trading volume that I was talking about earlier. That 20% of trading volume looks like it's a race to be slightly faster than the next guy to trade on a slightly stale price. And high frequency trading, I think before my research seemed kind of mysterious and into that vacuum can also come a lot of conspiracy theorizing.

What I showed was that a lot of high frequency trading was that you have a lot of assets across finance that are highly correlated with each other, that move together. That could be your Apple stock and Google stock, or it could be the Nasdaq 100 index, and the S&P 500 index. Or could be the S&P 500 index expressed as an ETF versus the S&P 500 index expressed as a futures contract. There's correlated stuff all throughout financial markets. Ten-year treasuries, two-year treasuries, tons of correlations across finance. And when one asset's price goes up, there's a lot of correlated assets whose prices are momentarily stale and a lot of high-frequency trading is arbitraging those [inaudible 00:09:16]-

Hal Weitzman: It sounds like it's a fulfillment of the efficient market hypothesis.

Eric Budish: Yeah, so it's correcting our mispricings, but correcting obvious mispricings and getting paid a lot to do so. And I guess the-

Hal Weitzman: Rewarding speed not smarts to go back to-

Eric Budish: Rewarding speed, not smarts. And-

Hal Weitzman: What's wrong with that though?

Eric Budish: Well, that's subtle. So, a lot of the high-frequency trading firms that I interacted with over the years would say, "Well, if I'm faster than the next guy or if I'm smarter than the next guy, either way I make a buck and that's capitalism. What's wrong with that?" And from a moralistic perspective, I don't have any objection to trying to be faster or smarter than the next guy and trying to make money from doing so. What I showed is that the underlying market structure had, I view as a design flaw that overly incentivized speed. And the design flaw was that even in the logical extreme where there's a trading strategy that's so obvious that the whole market perceives it... it's obvious to grandma trading strategy. And even if multiple trading firms try to execute that trade at literally the same time still one of them is going to earn an economic profit.

So, rather than the obvious mispricing getting corrected for free, if you will, it gets corrected at a rent to the arbitrageur. And those rents add up to serious money and they become like a tax on liquidity provision. So, another way to think about it is that if I'm trying to perform the useful service of making a market to investors, I'm trying to offer bids and asks that I'm willing to sell to you at a price, I'm willing to buy from you at a price, I'm providing that useful service to the rest of the market. Providing that useful service to the rest of the market is made more complicated and made more expensive by the fact that I have to constantly be vigilant about being a millionth of a second too slow to react to some pricing signal. So, I'm constantly getting picked off. I'm constantly getting traded against at a slightly steel price. That's an expense of my business that I ultimately pass on to the rest of the market.

Hal Weitzman: So, the costs are born by everyone. But as you say, what's interesting about your research is you're not blaming as maybe other people have the high frequency traders themselves, but saying this is part of the design. So, just briefly, what is that flaw in the design? And then what's your proposal that might change it?

Eric Budish: So, the flaw in the design, it's a subtle technical Flaw, which is the serial processing. Serial, S-E-R-I-A-L, not like the breakfast cereal. The serial processing of requests to trade, meaning that if multiple requests to trade arrive even at the same time, they're processed one at a time in order of arrival. And if you think back to the human era of financial market trading back when there were humans yelling at each other and trading pets, serial processing then was, well, if Hal asks for the trade before Eric asks for the trade, Hal gets to trade before Eric. And that sort of has a certain justice, fairness, simplicity to it.

But now imagine a thousand computers reacting to the same signal at exactly the same time up to human perception, all saying, "I want the trade at exactly the same time." In a human trading pit you'd say, "Hang on a second. You're all asking for the same thing. Let's figure out what the new price is and who ought to get the trades." You'd run a improvised auction. In the electronic market as currently exists, you can't run that improvised auction. You have to ascertain, well, which of those thousand voices yelling was a millionth or a billionth of a second earlier than the rest?

Hal Weitzman: So, a classic example as you just said, of rewarding speed rather than the best price.

Eric Budish: Yeah. And so, what my proposal is to just transform competition at the microstructure level, at the level of a second or below a second ,by if there are multiple participants trying to do the same trade at the same time, run an auction. We call that frequent batch auctions or frequent just means auctions run frequently throughout the day.

Hal Weitzman: How frequent?

Eric Budish: I think they could be as frequent as once per millisecond, so once per thousandth of a second and be quite effective because a milliseconds a long time for a computer.

Hal Weitzman: It's still slow.

Eric Budish: Yeah. So, that would be 23 million auctions per symbol per day. It's a lot, very frequent. It could be slower than that. There's cost and benefits.

Hal Weitzman: But just by introducing that change, that breaker, even at these very high speeds, we would be able to get better prices. Is that what it would end up [inaudible 00:14:11]?

Eric Budish: Yeah, effectively we'd get better prices, really better liquidity. It would become easier for institutional investors to trade larger quantities of stock at cheaper prices directly and at cheaper prices over the course of executing a large trade over a period of time. And that's actually what's been driving. There's been some innovation in the US stock market and European stock markets geared towards helping institutional investors realize those savings from trading in a different market architecture that isn't vulnerable to the negative parts of high-frequency trading. So, there've been a few trading venues in the United States that have innovated along the line suggested by my research. In Europe, there's several venues that are running variations of batch auctions. And they're drawing trading volume from institutional investors that are trying to trade essentially at a cheaper price, trying to trade large volumes [inaudible 00:15:16]>

Hal Weitzman: What are the times? Is it success?

Eric Budish: They're gaining volume. So, it's not like an overnight success where boom, the first person to be inspired by my research suddenly got a 100% of market share. But it's creeping upwards. It's easily trillions of dollars cumulatively of trading volume. So, seen from one perspective of an academic, is my research having impact on the world. That's great.

Hal Weitzman: You're changing, you're changing the market.

Eric Budish: So, that feels very gratifying and fulfilling, whatever positive words you want to insert there. But has it become the market norm overnight? No. One of the market tensions is that exchanges actually make a decent amount of money from selling speed technology to market participants. So, if you are NASDAQ or you're the New York Stock Exchange, you can uniquely sell the right to locate your computers next to NASDAQ computers or uniquely sell the right to locate your computers next to the New York Stock Exchange's computers, which if there's a speed sensitive trading opportunity on the New York Stock Exchange, you need that to be able to be competitive in that race. And so, the exchanges have found that they can make a lot of money from selling fast access.

Hal Weitzman: So, they don't-

Eric Budish: And the rest of their business is pretty lousy.

Hal Weitzman: So, they don't really want to change.

Eric Budish: It's a tension that gets in the way of market design reform, which is the exchanges aren't exactly jumping up and down to adopt my ideas. The tax created by high frequency trading by what we call latency arbitrage, it's on the order of a basis point. So, if you have a $100 share of stock, 1% is a dollar, a basis point is a penny. And in the research study out of the UK was like half a basis point. The thing about a basis point is it's really hard to get worked up about a market design reform that saves one 100th of a percent on trading volume. So, even if you're a large institutional investor like a Vanguard or a BlackRock, you should care about saving your investors a basis point or a half a basis point. But it's your investors, not you personally. If you're an ordinary retail investor or an ordinary wealthy participant in the stock market investing their own funds, a basis point doesn't register to my [inaudible 00:17:49]. If you accumulate a hundred million dollars of stock, which is a lot of money, a basis point of that is $10,000.

So, it's not nothing, but it's not existential. What a one basis point tax on trading creates in the political economics of the situation is a concentrated dispersed issue where on the one hand, a basis point adds up to we estimated on the order of $5 billion a year in global stock markets, easily that much more in other financial assets that trade on similar market architecture like futures, markets, options, bonds. You take a net present value of that, it's easy to get to the hundreds of billions of dollars of social welfare gains from my academic market design reform. So, that's a lot of money.

But on the other hand, the gains from that reform are very dispersed, a basis point at a time to investors. Whereas the harm from that reform would be borne disproportionately on a small number of market participants. The very most sophisticated high-frequency trading firms and the exchanges that sell speed to the very most sophisticated high-frequency trading firms. So, it's really a classic example of what political scientists call a concentrated dispersed problem, where one side of the policy debate has very concentrated issues, lobbies very vigorously. And the other side of the policy debate while larger and dollar terms, and it should win the argument on the social merits is so dispersed as to be less powerful.

Hal Weitzman: If you're enjoying this podcast, there's another University of Chicago podcast network show that you should check out. It's called Entitled and it's about human rights. Co-hosted by lawyers and law professors, Claudia Flores and Tom Ginsberg, Entitled explores the stories around why rights matter and what's the matter with rights.

Okay. Eric, we talked in the first half about high-frequency trading and how fast the market is, and some of your concerns about that. But there's other stuff about how stocks trade, how they actually trade, and you think this doesn't work from the perspective of these institutional customers. Just say what that complaint is.

Eric Budish: So, what I'm doing in this research on, we call it flow trading, and this is joint with a large team of collaborators at Pete Kyle who's at the University of Maryland, and is one of the foremost researchers in this whole area of market and microstructure. Peter Crampton, who's my collaborator on the frequent batch auctions research. Meena Lee and David Malek. Our vision for how institutional investors could trade in an updated for the 21st century market structure, is that they could express directly to the market, "Here's the portfolio of assets I want to trade. Here's the time horizon with which I want to trade it. And here are some pricing parameters for that trading." So, let me contrast with the present. A.

Nd the present, if I'm a mutual fund and I want to buy a million shares of Apple stock, I buy a hundred shares at a time, a hundred shares here, a hundred shares there, a few hundred shares there. I'm trying to disguise my trading, trading across lots of different venues. So, I'm not buying a million shares all at once. I'm trading that million shares gradually over a long period of time and trying to be very strategic in how I disguise disguise my intent to prevent essentially being front run by market participants who detect that there's a large investor buying a lot of Apple stock and that's going to push the price up. I don't mean front run and in the sense of it being an illegal activity, I mean more in the statistical sense of trying to trade in front of the rest of the market.

So, what we're trying to do is build directly into the market. Well, if I want to trade a million shares of Apple, I could express to the market I want to buy a million shares up to a hundred shares per second, for example, over the next 10,000 seconds of trading. And you could just say that, express that trading demand directly to the market in a clean-

Hal Weitzman: In one order.

Eric Budish: In a single order, and the market would clear once a second. So, a lot of the strategic gaming about disguising my trading intents actually goes away or becomes a lot cleaner because for example, let's say in the current market design, I want to trade a hundred shares per second for the next 10,000 seconds. That's going to leave a very statistically obvious footprint because whatever specific millionth of a second I happen to be trading on consistently over the next 10,000 seconds, the computer program will be able to pick up like, "Oh, someone's buying a lot of Apple, let me trade in front of them." And a lot of high-frequency trading or algorithmic trading is picking up the patterns in our very rich data, continuous time stock market to trade in front of institutional investors.

But if the market were clearing once a second, you could batch together with all of the other institutional investors and all the other market participants. So, you don't leave as obvious a footprint. And you can express to the market exactly what you want to express, which is I want to buy a million shares of Apple over the next few trading days. So, you need the batch auction trading as a base layer to enable that type of, we call it flow trading, trading as a flow over time rather than very discreetly 100 share block at a time.

And then we're able to layer on top of that also trading and portfolios. And this is useful for institutional investors and also useful for market makers. So, I'll give you an example of each. As an institutional investor, I could define for the market to the market, "Well, I'd like to buy a portfolio that consists of some Apple, and some Google, and maybe is short some Tesla if I think that's overvalued and maybe as long..." you name it, whatever basket of securities you're interested in. And you could directly express to the market, "I would like to buy this long short portfolio of assets over the next..." a period of time the user specifies with pricing parameters. The user specifies and directly execute that portfolio trading strategy in the market without having to use sophisticated algorithms provided by a broker dealer like a Goldman Sachs or JP Morgan, without having to strategically disguise my trading in the modern market design.

As an arbitrageur, as a trading firm, I could rather than... like I mentioned in the discussion of the previous work on high frequency trading, a lot of high frequency trading is you got two assets that are highly correlated. Let's say A and B. A goes up, buy some B. Or A goes down, sell some B to correct the mispricing. With portfolio trading and the way we envision in this flow trading paper, I could directly express to the market I'm willing to buy A and sell B at a net price. Or I'm willing to sell A and buy B at a net price. So, that allows an arbitrageur to directly express to the market design, "Here are the pricing relationships that I'm willing to maintain at a price, at a fee." And by sucking out a lot of the cost of engaging in that kind of arbitrage, we can then enhance the competitiveness of that kind of arbitrage for a very simple arbitrage like you got two assets that clearly should move in perfect lockstep, buy A, sell B, let's say they're the same thing.

Hal Weitzman: They could actually move at the same time.

Eric Budish: Yeah, they could actually just be tied together like a string. Jane Street, they famously hired Sam Bankman-Fried, they make a lot of their money from what are called ETFs or exchange traded funds. ETFs are complicated to make markets for because they consist of baskets of lots of different assets. And so, you have to be very sophisticated algorithmic of like, okay, if I buy the S&P 500 ETF, I got to sell the 500 components of the S&P 500 ETF, and I've got to do it in exactly the right proportions, and I got to be weary of price impact and all that stuff. And they make a lot of money from doing that quite smartly and effectively. They hire a lot of really exceptional talent. We're building in the capability to mark and make an ETFs directly into the market design. So, like Hal or Eric could say, "I'm willing to buy S&P 500, ETF and sell the 500 components at a net price." It's going to make it a lot more competitive and efficient to keep prices in line with each other.

So, the whole proposal in float trading, it's motivated by sucking out a lot of, or reducing is probably a more polite phrase, reducing a lot of rent in modern financial markets. There's a lot of rent associated with arbitrage, a lot of rent associated with speed, a lot of rent associated with the trading technology that allows investors to trade gradually over a period of time disguising their intents. And our theory of the case is, well, with a better market design, we could actually build into the market the ability to engage in these functions pretty cheaply and efficiently. And that would suck out or reduce a lot of the excess costs in the modern financial system.

Hal Weitzman: You talked in the first half about how batch auctions are being tried in Europe and tried in the US. What's been the reaction to your proposal about flow trading?

Eric Budish: So, flow trading, I think that's a great question and a fair question. When I first put out the batch auctions work over 10 years ago at this point, I think I naively expected exchanges to pick up the phone and call me up and, "Oh, let's try this thing. This sounds like a good market design." And at first I felt like I was getting some traction and then ran into a lot of headwinds that we talked about, and a lot of which relate to that it would reduce the ability of an exchange to make money from selling speed technology. With flow trading, I didn't come into with the delusion that someone would pick up the phone and try-

Hal Weitzman: You're already jaded.

Eric Budish: I'm already a little jaded or just more patient. So, I think of it very much as an intellectual thought piece of a proposal for what a market architecture of the future could look like. At the same time, there's been a lot of interest in the ideas in it, both in traditional financial markets and also in crypto financial markets for that matter. In traditional financial markets, the closest is an alternative trading system called OneCronos that is running a version of batch auctions as a structurally as what's called a dark pool in the US stock market. They're getting some traction. I don't have the numbers for their latest volume off the top of my head. But they're doing something that's in the same spirit as what we're proposing in the flow trading paper. And they seem to be getting a lot of traction and interest from institutional investors, enabling this kind of portfolio trading and batch auctions a little slower, a little bit more expressive and complicated, and enhance better matching how institutional investors actually want to trade.

And then in crypto markets, there's been a lot of interest in this batch auctions work. And the part of the reason there is crypto markets... I have a lot of complicated views on cryptocurrency that we can talk about some other time. But crypto markets by nature, because the blockchain progresses in blocks, it's built into the terminology, there's a lot of intrinsic batching to the way blockchain data arrives. But the earliest crypto trading mechanisms nevertheless processed all of the requests to trade that arrived in a block serially, one at a time in order of receipt. And that's led to a version of high-frequency trading in crypto markets that is like high-frequency trading on steroids.

And the reason I say that is in crypto markets, you can literally reorder time in a way that you can't in traditional financial markets. So, if I want to buy some Bitcoin and you want to sell some Bitcoin, or I want to buy some Ethereum and you want to sell some Ethereum, and those orders are waiting to trade, as a market maker, a market participant, or what the crypto people call a block builder, I could literally reorder the trades where like, "Okay, let me buy some Bitcoin in front of Eric and then let me sell some Bitcoin in front of Hal, and reorder the trades to maximally extract value from the crypto trading participants."

So, it's like a version of high-frequency trading rent extraction, but just magnified because there's this ability to control time that's not present, and to reorder time in a way that's not present in traditional financial markets. And so, crypto folks have had a lot of interest in my research for that reason. They see solutions in my research to a lot of the negative features of crypto trading that they're seeing in those markets. So, we'll see. I try to have a patient academic attitude while also engaging with practice. And whenever there's interest from the real world, I try to engage vigorously and also try not to lose sleep over as anyone trying my stuff tomorrow?

Hal Weitzman: Well, Eric, thank you very much for coming on the podcast and talking to us about two ways to improve the stock market.

That's it for this episode of the Chicago Booth Review Podcast, part of the University of Chicago Podcast Network. For more research, analysis, and insights, visit our website at chicagobooth.edu/review. When you're there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research. This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe and please do leave us a five-star review. Until next time, I'm Hal Weitzman. Thanks for listening.