Days after the so-called flash crash on May 6, veteran high-frequency trader and exchange entrepreneur Dave Cummings sat down with the entire Traders Magazine staff to discuss a bevy of issues facing the equities markets. Cummings started the pioneering high-frequency trading firm Tradebot Systems out of a spare bedroom in his house 11 years ago. He remains its owner and chairman. In 2005, Cummings launched the electronic communications network BATS Trading. Today it’s the third-largest equities exchange. Here is part two of our interview with Cummings.
Traders Magazine: How about "trade at?" This would require internalizers to have more skin in the game, have a quote up. If they do want to internalize the order rather than send it to the public markets, they have to price improve or sweep the limit orders on their books?
Dave Cummings: I think that has some problems. You need to think about that very, very carefully. There’s a balance in the markets today, and it’s not a bad balance. In a lot of ways, dark pools are a cleaner form of internalization. So, having a broker-dealer match their customers off using an ECN-like dark pool, that’s better than having it done manually on someone’s desk. It does a better job of maintaining information barriers among the customers, and so the market structure has allowed some thoughtful new models. Maybe there should be some discussions about how to bring back a little more of the liquidity to exchanges. But you have to be careful. There are multiple forces that are each trying to get a competitive edge. You have to ask yourself if they’re trying to squeeze out these dark pools on theoretical grounds or for competitive advantage. It doesn’t seem to be horribly broken. Maybe there can be some small tweaks. We do not internalize. We don’t deal with customer flow. All these comments are in the context of Tradebot.
TM: But firms like yours feel that they’re being put at a disadvantage because good retail flow is being stopped at big brokers. It’s not reaching the public markets.
DC: It’s an old argument. It would certainly be good for me if they’d just inject all these orders in the public markets. But it’s a long-standing practice. The advantages and disadvantages have been debated over the years. And they’ve allowed it to be the way it is. Interestingly, in futures, it’s not that way. They don’t allow internalization in futures. …
TM: They say the retail customer would end up paying more because you have to pay access fees as they pass through, and you wouldn’t get price improvement. And under the SEC’s idea, the internalizers would be forced to pay a tick price improvement. Now they get a quarter-tick or half-tick, or something.
DC: Paying a full tick effectively bans internalization, or bans a fair amount. So let’s not call it something it isn’t. A lot of these markets are a penny wide. The increments are big–you can talk to the wholesalers and see what they can do–but I don’t think economically they can price improve by a full penny, in very few cases. So, I think you’re effectively talking about a ban on internalization. But, again, that’s not my war.
I don’t think the public is clamoring for price-improvement. I think it’s more about liquidity and certainty.
TM: Are there some issues out there that you feel need to be addressed?
DC: The industry needs to come up with thoughtful circuit breakers. We need to prevent bad trades. Anytime you break a trade, you’re going to have an upset party; it damages public confidence in the markets. It really, really matters. Breaking trades after the fact–sometimes by several hours–is never good. You’ve got to define what is a good trade that will stand and what is a trade that is out of bounds. You’ve got to have collars and circuit breakers that prevent those trades in the matching engines …
TM: … so, if something moves by X percent, then in milliseconds … it’d be a time thing, too, right?
DC: People need to put together a concrete proposal and thoughtfully think it through.
In futures, they have the concept of limit down and limit up. We need collars in both directions; it’s silly to pay a penny for Accenture, but it’s also silly to pay $1,000. The public will be very outraged either way. And so, sometimes it’s easier to see the fury on the downside. But the rebound [May 6] could have been many trades on the upside, too–I haven’t looked at the data closely. The issue is, when there’s a news event–whether it’s stock-specific or it’s the whole market–people need time to digest the news. The algos are generally built for normal market conditions. And once you’ve got a move that tells you you’ve got abnormal market conditions, and people need more time to think … when you ask somebody to step up in milliseconds, you’re going to get a different answer than when you give them a few minutes. I’m only talking 10 to 15 minutes; I don’t think we need an hour shutdown, or anything.
TM: Is that a halt, though?
DC: That’s where it comes into play, whether it’s a collar or a limit down that adjusts lower and lower or higher and higher, and you never actually shut the trading down. And you can bounce off of the limit down.
Maybe there’s a way you can structure it so that it avoids a complete halt.
TM: Talking about the markets, in general, didn’t a lot of market orders execute at a penny, or something?
DC: I don’t want to mischaracterize any particular order. This is on an order-by-order basis. I don’t think any of the current systems were the complete answer. The other thing that you had going on … you also need to rethink the market-wide circuit breakers. It’s 10 percent during the day, and 20 percent during the last hour and a half. I pretty firmly believe they need to tighten that in to 5 percent. But the halt needs to be 10 to 15 minutes, rather than an hour now, which is way too long. What we found was a 992-point move is too much. We need some intermediate things to slow that down–if you go to a market-wide 5 percent collar and you make the halt period shorter. That kind of stuff you could implement, because it has the same form as the existing rules. You’re changing the definition of the trigger point, and you’re shortening the length. But I think those changes could be implemented very quickly. On top of that, you need the stock-specific collar, but it’ll take longer to implement because a lot of people will have to change their software. But the combination of those two will be robust markets.
When the whole markets are crashing down, people need time to analyze the news and say: was this a real event, a terrorist activity that will really, meaningfully, fundamentally change the value of the stock, or was this a random spike?
TM: Wasn’t the problem that all of the buy interest was used up? There was no liquidity on the buy side, right, and that’s why prices plummeted?
DC: Most of the liquidity when you look at the book is in a few price levels around the market. And that’s just the way, practically speaking, people trade. Few people put in limit orders way off the markets that are unlikely to get executed. And I don’t know that we’d really want them to, because that would just cause a lot of needless noise in the matching engines and cost the industry to collect and redistribute those quotes off the markets. The answer is defining a collar, or a limit, and just saying: time out. You can’t trade below this price before a few seconds have passed.
There are thousands of value asset managers who will buy stocks, but they may not have their order in the millisecond the stock is plunging; they’re going to thoughtfully step back and say: where is it going to re-open?
TM: And when they re-open, it’s just like with the market-on-close, and all that, they know what the market is–they can see that.
DC: I think all people should see; I really hesitate to see a structure where only certain people can see everybody else’s orders and then they can front-run them by a penny. I’m really worried about a specialist, or anybody, having the license to front-run people by a penny on the re-opening.
TM: You were quoted in a WSJ article [shortly after May 6] saying that you pulled your orders during the flash crash. Why did you stop trading then?
DC: … to limit our risk. The buyside needs better technology. The markets need better circuit breakers or collars.
TM: But can’t you blame the brokers for what happened, too? They have a duty to get best-execution. And if they’re selling what used to be a $50 stock at $10 …
DC: But realistically, the brokers are in a no-win situation. Hypothetically, if a $50 stock drops to $30, and you’re holding a market order, should you sell it? Or, what happens if it goes to $10, and you had an opportunity to sell it but you didn’t because you thought it would bounce back? Again, the market needs to have a collar. The broker never should be second-guessing the customer, because inevitably they’re going to get sued if they guess wrong. It’s just not fair. The fair thing is for the market to say: that move is extreme; we need a few moments to think about it. And then, the next print will be a reasonable print. Sometimes stocks move hard for good, fundamental reasons. But the price you get in a few milliseconds … If you just delete that 15-minute window, where the market was 15 minutes earlier was probably pretty reasonable. There were fundamental things going on [May 6]. That’s part of what caused the situation–it wasn’t just people being weird. There were about five serious news stories going on. And that’s going to cause fundamental buyers to be a bit more timid–for good reason. But the value guys aren’t going to stand there and say: Yes, if it drops 2 percent, back up the truck. You’ve got riots in Greece, the British election, Euro contagion stories, a potential bombing in Times Square, an oil spill in the gulf and flooding in Memphis–about five or six good reasons why buyers would be getting skittish.
TM: If you’ve got more selling interest than buying interest in an era where there’s no NYSE specialist to slow things down and add his own liquidity, what happens?
DC: "His own liquidity" is a myth. There’s no market participant that can hold up a trillion dollar market. The high-frequency trading guys can’t. The specialist can’t. Slowing down [the market] gives the whole community time to react, not one guy deciding where it’s in his interest to re-open the stock.