Sophisticated and expensive computers allow high-frequency traders to take advantage of minuscule differences in price among the many exchanges where securities are bought and sold. Some firms pay to place their computers on the site of a stock exchange to be sure their access to price data is as fast as possible, a practice known as colocation; others will use technology to obscure their trading intentions for a few crucial thousandths of a second. Lewis's book tells the story of Brad Katsuyama, a former trader at the Royal Bank of Canada in New York. Katsuyama opened a new stock exchange last year to give investors protection from HFT.
Lewis is not the first to cry foul on these strategies. Eric Scott Hunsader, the founder of Nanex, has made himself immensely unpopular in some circles for his outspoken and persistent criticism of HFT, which he first encountered during the "flash crash" of 2010. Bloomberg called him the "nemesis" and "scourge" of the HFT world.
I asked Hunsader to talk about the book, the new stock exchange, and his long career in financial technology. The conversation focused on the Securities and Exchange Commission ruling in 2007 that allowed what we now know as high-frequency trading. The transcript, edited for length and clarity, is below.
Wonkblog: When I was a kid, I can remember my grandpa showing me how to look up stock prices in the newspaper. And there were exactly three exchanges. Oftentimes the prices were in fractions -- one-eighth, three-quarters, and so on. And then, it wasn’t long after that that I was showing him how to look up stock prices online. I’m wondering if you can talk about the transition into electronic trading -- and you think that at least initially, it was good for everybody. Is that right?
Eric Hunsader: Oh yes. Definitely, just like computerization improved the efficiencies of all industries, the same thing happened in the beginning. I’ve been doing this since ‘86. It wasn’t really until 2008 and 2009 that we noticed the explosion of quotes, of price changes for stock, without an accompanying explosion of actual trading. Normally, you’d see the price change once or twice a second -- now, a hundred times, or a thousand times. Just a little quick calculation on the calculator, and a knowledge of the speed of light, tells you, “Hey, wait a minute, you guys are changing the prices faster than anyone can physically access them, so why are you blasting that all over the country on the consolidated feed?”
This is, in your view at least, the result of Regulation NMS issued in 2007 by the Securities and Exchange Commission.
With Reg NMS, it essentially took x amount of available stock at one place, and made it one tenth of that x at ten different places. The regulations, as they were written and they were understood, were that routing of orders for stock would all be based on what they called a SIP, a security information processor. The SIP would tell each exchange if there were a better price somewhere else. If so, the order had to route it to that other exchange, which guarantees you get the best price available on any of the exchanges. Where does the best price come from? It comes from the SIP, also known as the consolidated feed. It’s really what goes along the bottom of CNBC or financial sites.
All would have been fine and well, or at least, it would be a lot better than it is today, if they actually used the SIP. Instead, the exchanges created these direct feed connections, which is a way to get the prices faster, which is actually illegal under Reg NMS. That’s the big issue I have. Reg NMS is very clear: exchanges can’t give this market data to anyone else in any other way faster than they give it to the SIP. High-frequency traders are taking advantage of the exchanges illegally giving them this data sooner. If anybody is going to get arrested or busted or fined over this, it probably won’t be the high-frequency traders. It’s going to be the exchanges. That’s what the NYSE got fined $5 million for in 2012. The SIP is the core of that piece of regulation. Without the SIP, all of Reg NMS falls apart.
For a non-pro, SIP costs a dollar a month. For a pro, its $127 a month. A direct feed from one exchange could be $10,000 to $60,000 a month. If the SIP is no longer real-time, it means that people who want to get real-time data have to pay, well, significantly more. It’s no exaggeration here to say you go from $1 a month to well over $100,000, with your own networking staff and a room full of computers just to get real-time pricing. The regulator and the industry foresaw this. This was debated. It was talked about. It’s written in Reg NMS. This is why we did it, so investors wouldn’t have to go to all these exchanges and contract with each of them individually. They could get pricing information from one source. We’d be able to audit it. It would be inexpensive, and it would be the reference price. There would be no question as to what the price of, say, IBM is right now.
But that got buried or lost somewhere early on, and that really is the key to why things have fallen apart. So when Michael Lewis used the word "rigged," he’s right. It’s rigged, because all of these gains depend on receiving the information faster from an alternate source, which is specifically and clearly, with precedent, forbidden by Reg NMS. Period. It’s such a simple thing, such a simple thing to fix. I don’t know why it’s still going on.
Let’s talk in more detail about those gains.
It ends up being a race. The ones who are colocated, the ones who are spending millions of dollars a year on colocating network-engineer guys, fast, direct lines, whatnot -- the fastest of them will be able to buy everything available when they detect an institutional order coming in, for example, and the others will lose. Which is all fine by me, as long as it doesn’t impact everybody else. As long as these guys, at the end of whatever they’re doing, are using the SIP as the true reference price, then it doesn’t bother me how fast trading goes. As long as its transparent, and everybody’s following the same rules, I’m all for that. But the problem is we have rules, and they’re not being enforced. That means an honest businessman is going to say, “Well, I can’t really do that, because that’s against regulations--and even though the regulator isn’t doing anything about that now, they might, and that’s not a risk I want to take.” So you only have people involved who are willing to skirt that rule, so they’re likely to be the same kind of people who will break other rules, too. That’s not how our society is built. It’s built on law. Enforce it.
If you’re trading and you’re trying to sell a stock, say, and someone else has a bid that’s perhaps higher than the reference price that everyone sees, you might be able to make a small amount of money on that. If you did that every day for a year, you might make quite a bit of money, but, it’s not immediately obvious why there’s anything wrong with that. The person who’s buying the stock from you is willing to give you the money.
It’s when they don’t. It’s when they put in the order, and it shows that there were 10,000 shares offered, and then all of the sudden -- the buyer only gets 6,000, and they’re wondering, “Why didn’t I get all 10,000?”
[To explain this point, Hunsader brings up this simplified diagram shows three exchanges and the cables running between them.]
Mahwah, at the top, is just 40 miles north of Nasdaq at the bottom. Nasdaq is down there at Carteret. Mahwah is north New Jersey, where NYSE is. Weehawken is right across the Hudson, pretty much smack in the middle of these two.
When you want to buy a lot of stock, you’ve got to go to multiple exchanges to get it. That’s one of the things Reg NMS did. It took all of the liquidity in one spot and split it up. The order, if it comes into BATS first, a high-frequency trader there will see that trade execution. Immediately, on the blue line, up to Mahwah, they will signal up to their other machine to say, “Buy everything that’s available.” And they can do that faster than the order gets to the NYSE. It’s the fastest network there is. They’re able to remove that order on the other side. The only reason they made that purchase in the first place is because they saw this other order coming in. That is not something that humans would tolerate face-to-face.
This is just one example. Having that speed advantage--it’s the equivalent of having tomorrow’s Wall Street Journal in your hand today.
You’ve written about some other approaches to high-frequency trading, like using the speed to create volatility, one technique that stood out to me as seeming particularly dangerous.
All the other algorithms out there are based on reading what the orders coming into and out of the book are. When they suddenly see prices moving, and they know that they’re not the fastest, but they’re faster than the general public, they know that if they jump on the trend right away, they will be able to dump it off on the other slower movers who will start reacting a few seconds later -- or a second later, or a hundred milliseconds later. So it’s very easy to influence a lot of other algos to join you. If you have enough money, you can cause what are known as momentum ignition events, where you create enough price movement in the right direction and in the right way to get the other algos to believe, “Hey, something’s going on. I want to participate.” They all jump in.
All of a sudden, what happens -- and we see this in the market every day -- it will take the price up 5 percent in a second, on hundreds of trades. The machines are just feeding off themselves. They all have their different limits, so the fire kind of puts itself out, and then when no real news about the stock materializes, it’s a race back down to the bottom. We get these 5-percent moves inside of a second in these big-cap stocks.
A lot of the criticism of Lewis’s book seems to be that it is already out of date, that the practices it describes were very common a few years ago but are less so now.
That’s not true. I certainly don’t see a change. This all comes to how Brad Katsuyama created this program called Thor, and then decided to create the exchange IEX. Thor was designed to time the trades so they would hit, say, BATS, the NYSE and NASDAQ all at the same time, so that a high-frequency trader at each one of those facilities couldn’t see it fast enough to react and get to the other exchange. And so that’s what Thor did. But Thor only works when the networks are clear and uncongested. if one of the networks, on one of the exchanges, has to process another, even 50 quotes -- that’s just enough time, enough delay, that it will expose what Thor is trying to do. It doesn’t always work, but it works often enough. It doesn’t cost anything to stuff these quotes in the system.
That’s why we see these high quote rates in stocks without trading. It doesn’t even have to be the same stock. It can be the same group of stocks that processed by the same computer. That’s what quote stuffing is, and that still goes on today. That should just disappear, if they weren’t doing these games any more, but they are. They’re going to roll out this new laser technology that gets from Carteret to Mahwah in an obscenely small amount of time. It’s getting on the low-microsecond range. There’s a reason why they’re willing to pay for that.
This has not changed at all. They’ve just gotten more sophisticated at hiding it. Instead of doing it solid for a second, they’ll do it ten times as much, but they’ll do it in a tenth of a second, so that the exchanges, which measure traffic on a one second interval, won’t see it. It will be invisible, but it has the same impact.
Getting back to Lewis’s book, IEX just opened in December. They were still pushing Thor around a year ago.That book is not old at all. It is spot on the money.
Your own company, Nanex, in Winnetka, Ill.--when and why did you start it? How do you see your place in all of this as a critic of high-frequency trading?
Back in '86, I was collecting everything traded on the CME onto a floppy desk, and selling that historic data from the Chicago Mercantile Exchange via Bulletin Board System. (That was the precursor to the Internet).
We actually had cell phones back then. We even had a real-time system once. We were transmitting real time futures charts into your car -- you’d ride around, and on a laptop hooked up to your cell phone, you’d see real-time, streaming charts. I’ve been here since the beginning.
People who know me know that I’m all about data. I’m very scientific about it. I’m not going to publish anything or say anything unless I’m 100 percent sure and I’ve verified that that’s what it is. I’m a very quiet person. I am not one who -- I’m a programmer at heart. That’s what I enjoy doing. It wasn’t until the flash crash came along: May 6, 2010, at 14:42:44 is when it overloaded.
We ended up processing over a billion quotes that day. It seemed to me the SEC was having trouble assembling data. We had it. My company, we can replay any market day that we want. I made the fateful mistake of saying to one of my programmers out in California, “Hey, let’s see what we can do with our data. Maybe we can find something here.” If I could go back to that day, I would, and I would not have made that decision.
But what you saw when you did--
The system overloaded. The capacity had not been upgraded in a long time for the SIP. It couldn’t process the information fast enough, which meant that a whole group of people, instead of being just a little bit behind, got 30 seconds behind. We lined up where the prices were versus what CNBC was seeing, and CNBC was almost a full minute behind. They were screaming, “It’s the end of the world. It’s the end.” They thought the Dow was down to 10,000. The market had already rallied almost already to the starting point. That’s how wrong they were.
Looking at the data closer, we started seeing all these bizarre quote patterns, where there was 1,000 quotes on the stock, but no trading was going on. We plotted it out, and it actually made quite beautiful pictures, because it was mathematical. It was geometric. My programmer out in California called them crop circles. And that name kind of stuck.
High-frequency traders say that their trading makes it easier for investors to buy and sell stock, that it improves liquidity or decreases spreads.
First of all, their definition of liquidity is not the same thing as what is commonly held in the industry up to this time. It’s almost in a way saying, “ ‘Up’ depends on who you are. ‘Down’ depends on who you are.” People in the industry are like, “What are you talking about?” Liquidity doesn’t depend on who the person is. It depends on whether they’re willing to buy or sell at the current price. The whole business of trying to cancel your order on another exchange, because you’re faster--how is that liquidity? That’s not liquidity. That’s phantom liquidity. That’s bunk.
The lowering spreads: Reg NMS started in 2007, and that’s when high-frequency trading was born, this kind of high-frequency trading that we’re talking about in Lewis’s book. It started in 2007. Spreads were at their lowest in 2006. They have never gotten back down to that point since.
What about Katsuyama's new exchange, IEX, which hopes to eliminate the networking advantage of some firms party by slowing down orders? Do you think the exchange will be successful?
What they’ve done is they have 38-mile coils of fiber. If those coils were 4,000 miles long, that I could get behind. You could actually simulate 4,000 miles’ worth of cable with software, and that would mean that everybody in the United States would be on equal footing. It seems to me like IEX is equal in speed for anyone in a small, 30-mile radius.
Anyone in New York, basically. That’s probably frustrating for you.
Have we learned nothing from September 11, 2001? We were supposed to have redundant facilities in other places. The financial system shouldn’t all be concentrated like that one in space.
Last question: if it’s only Wall Street traders and major hedge funds who are losing money--
Then who is harmed? And why should ordinary people care?
Society is harmed when a group of people don’t follow the rules, and the regulator knows about it. Then all of society loses. We don’t want people to have to make a decision whether or not they want to go into business if they have to make a decision about -- there’s a term for this on Wall Street -- regulatory risk. That should not be, ever, part of the American dream.
We don’t want this wink-wink-nod-nod business. That drives smart people out. I could quantify it--everybody is losing a small amount of money. The real cost here is that we live in a society based on rule of law. I’m sorry, but you’re not getting a free pass on this. This is wrong and it needs to be corrected.