The Changing Role of The Broker
Tue, 28 Sep 2010 16:53:00 GMT
Interview with Joe Gawronski, Rosenblatt Securities
Joe Gawronski is the President of Rosenblatt Securities, an agency-only brokerage founded in 1979. In addition to its reputation as an innovator in the execution services arena, the firm has carved out a niche as one of the leading providers of analysis on global exchanges and market structure and investment banking services in the fin tech/capital markets space. Joe is formerly a securities lawyer with Sullivan & Cromwell, a Vice President in the equities division with Salomon Smith Barney and COO of Linx, a block trading ATS. He received his B.A. in Public and International Affairs at Princeton’s Woodrow Wilson School and his J.D. from Harvard Law School. He is an Allied Member of the NYSE, a member of the Advisory Boards of both the Journal of Trading and Wall Street & Technology magazine, a term member of the Council on Foreign Relations, and a member of NOIP.
High Frequency Trading Review: The first question is the same question I ask everyone and that’s what’s your definition of high frequency trading?
Joe Gawronski: That’s a very good question to start with because I think different people mean different things. People throw out figures like, “high frequency trading or algorithmic trading now accounts for X % of the market”– well what are you including in that definition? The term “algorithmic trading” can be used to mean the tools sell-side brokers distribute to clients and these institutions then use for order placement, such as VWAP and implementation shortfall algos. It’s not just high frequency trading, but pretty much all institutional trading, that is using automated trading techniques and algorithms to some extent to slice up their orders. So I think you’re wise to start with a definition of your terms.
If you look specifically at high frequency trading, we divide it up into two categories for ease of categorization. One is electronic market making and the other is stat-arb.
Stat-arb is where firms are really looking for short term anomalies, price inefficiencies, looking at relationships between instruments. So the classic example (which I guess I can no longer use because of what’s happened in the market but I’ll use it anyway!) is buy GM, short Ford.
HFTR: Pairs trading, you mean?
Joe: Yes, pairs trading or even cross-asset class trading, like the ETF versus the underlying, or the futures versus the underlying. Now the stat-arb players tend to be net takers of liquidity as opposed to providers of liquidity posting bids and offers.
The second group of high frequency traders are the electronic market makers, which is the group people mostly are referring to when they talk about high frequency trading. They are the bigger of the two groups and the ones who are making two-sided markets. If they’re doing it right, then they’re capturing spread and exchange rebates and entering and exiting positions within microseconds, milliseconds or (worst case) seconds. Not only are they trying to end each day flat, but they really don’t want any significant long or short exposure to any instrument at any time if they can avoid it; whereas stat-arbs actually are taking a position and making a directional bet. They may buy Ford and short GM because they think the prices are going to converge or they think there’s something in the trade. They don’t tend to have the holding periods of big institutions who hold the stock for three years or a hedge fund for three months, but they may hold it for three weeks, they may hold it for three days, they may hold it for three hours. Electronic market makers don’t really want to be holding positions at all, never mind at the end of the day.
These two groups aren’t mutually exclusive by the way and the groupings are over-simplications. There are some stat-arbs that will also be employing what looks like electronic market-making strategies and providing liquidity. But overall those are the two groups.
HFTR: Are there any common threads between those two groups?
Joe: Well, with both groups you have the automated entry of hundreds or even thousands of orders over a period measured in seconds, with mass use of cancel & replace, i.e. canceling orders and putting in new ones. They’re both very opportunistic. They both need very reliable low latency, high capacity systems, the electronic market making category even more so because they play a very ultra-low latency game.
Some of the stat-arbs play more what I would call a low latency game. It doesn’t necessarily have to be the fastest, it just has to be fast enough. The other commonality is they make very little money per trade. We estimate that the average revenue per share traded is only between a tenth and one-twentieth of a penny.
HFTR: So they have to trade in huge volumes to make it work?
Joe: That’s right. The last thing I’ll say on this for clarity is what high frequency trading isn’t. And that is sell-side and institutional algorithms, what I started with. These are used by brokers and institutions to execute orders. They may have some commonalities with high frequency trading. Some of these algos, for example, will use co-location to minimize the latency disadvantage. But there are some key differences. The first is that they have instructions behind them and investment goals driving them so in a lot of cases they need to get aggressive and cross the spread rather than sit back, wait and post shares on the book hoping to earn the spread and/or a rebate. That’s one of the reasons you can say that high frequency traders are “rebate hogs,” because they tend to crowd out the sell-side and the institutions in terms of being able to earn spreads and rebates. Another key difference is that the high frequency guys are entering orders hundreds of thousands of times in a several second period. The algos that the sell side use on behalf of institutions, on the other hand, are entering orders a few times a minute, or once every five minutes, not hundreds of thousands of times a second. So the actual frequency isn’t nearly what it is for the high frequency trading, not surprisingly given the label of high frequency. They don’t have the same end goal.
HFTR: Staying with the sell-side for the moment, in these days of multiple execution venues, fragmented liquidity across the traditional exchanges and dark pools, how do you think the traditional role of the broker is changing?
Joe: Well, the broker’s interaction with the market is much more complex than it was ever before, right? Back in the day, we had two market structures in the US– the NYSE floor for trading NYSE-listed stocks and Nasdaq market makers (and ECNs eventually) for trading the Nasdaq-listed stocks. It was pretty clear what you had to do and you had a limited number of venues. There were a few dark pools like POSIT & Liquidnet that came along, but it was pretty manageable.
Now, you’re looking at dozens of destinations. The New York market structure was really simple before. To get good execution you had to go down to the floor. You could use DOT, an automated delivery mechanism, but even with that, there was a manual process requiring specialist interaction. So when you sent your order to the NYSE, you knew what happened and if you ever visited the floor you knew exactly how it worked.
Today, while in some sense the buy-side is more empowered than they ever have been because they have the algos on their desk, they have DMA, and really all the tools the sell-side has, at the end of the day they have a lot less understanding and transparency as to where their orders are even going.
And that’s true even for brokers, right? Brokers are using exchanges and dark pools that go to a lot of different places. The destination chosen, even if labeled a dark pool, implying limited information leakage by its very name, is not often the end of the line. When you send an order to a major broker’s so-called dark pool or use one of its algos, unless you really understand the market structure and the tools that you’re using, you have no idea where the order is going. The market has become much, much more complex to manage.
At Rosenblatt, we try independently to look at the market and give our honest opinion about things. The need to help the buy-side navigate the market, as well as making sure your own trading desk can navigate the market properly, has become an increasingly harder job. To be a true execution consultant to the buy-side, you certainly have to be expert in all these venues and understand the complexity And although the concept of best execution is almost a joke now, striving for best execution still exists among a few brokers. We’re one of them.
HFTR: Why do you say that the concept of best execution is almost a joke?
Joe: Because most brokers today make their routing decisions based upon how expensive one venue is relative to another. There are only a handful of brokers that really make those decisions genuinely, based on best execution and available liquidity. And in truth, it’s very difficult, even with all the transaction cost analysis out there, to measure best execution because you are now faced with the situation that at any particular time you have an order, you have a dozens of choices, and thus how can you be a hundred percent certain that you’ve achieved best execution for that order? There’s no way a regulator is going to say, after the fact, you did or did not get or at least strive for best execution. It’s too hard to prove with the amount of fragmentation that exists. I know we strive for best ex and as one of the best trading firms out there I think we come as close as anyone does. But you can never say on any given trade with certainty that you achieved best execution because you don’t know where all the hidden orders are. It’s impossible to even be connected to every venue out there, because you can only have one or two OMS and EMS systems and they don’t connect to everyone.
So I guess the quick and dirty answer to your question about how the role of the broker is changing is that it’s a much more challenging role today than it ever was because of the complexity and the need to navigate a very fragmented market. Our motto is to try to become “execution consultants”, which is an overused term, but then most people who use it don’t really live it. You have to understand that most of the time the buy-side firms use a long list of brokers as they’ve got lots of obligations deriving from research usage and IPO allocations, so if you can help them with all of their trading you’re going to be in a much better position and be much more valuable to them.
HFTR: OK. Can we talk about risk for a couple of minutes?
Joe: Sure.
HFTR: As we all know, electronic and automated trading has gone through an explosion in growth over the last 10 or 15 years. What are some of the new risks associated with that growth?
Joe: While the overall market structure seems to have benefited from the growth in automated trading, the systemic risk and the potential threat to the health of our capital markets that speed and the frequent absence of human judgment as a backstop must be acknowledged. Let me explain.
It’s our firm belief that high frequency trading has been a net benefit to lowering transaction costs, reducing spreads, things like that. I think all the evidence points in that direction. True it’s not without blemishes and there are some problems associated with high frequency trading. For instance, the evidence suggests that the benefits have been concentrated in the most liquid stocks as opposed to the less liquid stocks. There are other issues around costs to institutional investors in some instances, the “rebate hogging” I mentioned earlier. But this cost is very much outweighed by the amount of spread reduction and the reduction in transaction costs in our opinion.
Though we, as a firm, are net positive on high frequency trading from a day-to-day perspective, the events of May 6 illustrate that the pervasiveness of speed in the marketplace, fragmentation, and automation have created certain risks, right? The bottom line is sometimes—namely when there are dislocations or anomalous market data happenings– things like price discovery are better off going slow than fast! When they’re going fast it accelerates, compounds the problem.
From a systemic perspective, I think one could argue that we live in a riskier market structure than that. I say “arguably” because I still think the market structure that we have here in the States is a phenomenal market structure. I do think it results in the lowest transactions costs, which means that it does help the end goal of our secondary market. The trading market has one goal, to make people feel comfortable that there is liquidity to get in and out of positions so that the primary market, the market for raising capital, functions properly. And I think our secondary market is a fantastic market.
I would say however that when speed is put at a premium to all else and you don’t have things like circuit breakers, then you do run the risk of having a confluence of events that could lead to a systemic problem, which then tends to outweigh the day-to-day benefits of high frequency trading and the automation, because you scare off investors. They say “oh it’s a rigged game” or “there’s a systemic risk” and that has to be addressed in both perception and reality. But I would say that day-to-day, high frequency trading has helped the market structure and the overall capital markets. Yes, May 6 demonstrates that there are some systemic risks that need to be evaluated and carefully considered. But I think actually the SEC is doing a good job at looking at that, to be honest with you.
The overall evidence would suggest that our markets have continually improved. Now high frequency trading may be coincident with that and not the cause of it, I can’t be sure. But it certainly hasn’t ruined the positive direction of our markets from a market structure perspective. May 6 certainly highlights that it’s not all roses. But you have to look at the fact that we’ve changed our market structure dramatically over a decade plus period and it really accelerated over the past few years, especially since Reg NMS. I’m comfortable that we’re heading in the right direction.
HFTR: So in terms of the review that the SEC is currently doing regarding the US equities market structure, what do you personally think needs to change?
Joe: I come at it from a very pragmatic point of view. May 6th revealed a shortcoming in where the market structure has brought us and needs to be addressed. We need to look at things that will protect the systemic integrity of the market like circuit breakers, elimination of stub quotes, and making sure that we can slow down the markets when necessary to prevent a disaster. Outside of that, I actually think our market structure works pretty well. I subscribe to the maxim “if it ain’t broke don’t fix it.” You don’t really want the politicians and the regulators trying to fix a market if it’s not broken as the unintended consequences can be severe.
That said, where do I see additional issues that perhaps should be considered but around fairness. Broker routing practices for example, need to be much more transparent.
HFTR: In what way?
Joe: We talked earlier about the complexity, and I said people hit the send button,without really knowing where things are going. That’s not right and it’s not consistent with the best-ex responsibilities of brokers. So I believe that we’d be in a better world if there were less opaqueness and more transparency on the broker routing front. Now it’s another question whether that needs to be driven by regulation or whether the buy-side just has to speak up more and say, “damn it if I’m using your algorithm you’ve got to tell me where everything is going and I want real time executions by venue and I want full transparency, and I don’t want to hear you tell me you can’t do it because you have a non-disclosure agreement in place with someone”. People need to know where their orders are going and it would benefit the overall market structure and the end investors if there were more transparency there.
The second thing I would point to is I personally don’t believe that most internalization by broker-dealers is a good thing. I think a decade plus ago, before we really had competition in the exchange market, internalizing brokers performed a valuable competitive element to the exchanges. But now that the old NYSE/Nasdaq duopoly has gone away, I don’t think brokers should be able to internalize trading because it just creates further fragmentation and impedes price discovery, unless they are actually offering price improvement. True price improvement, that is. They should not be allowed to internalize at the national best bid or offer (NBBO), as it discourages display of limited orders. If someone that has put themselves out there and has a limit order offering stock at $20, the broker can trade at $20 and ignore that quote. That’s something I would change.
HFTR: Ok then, to wrap things up, how do you see the markets evolving over the next few years? What you see coming along, in terms of technology, in terms of changes in regulation, or in any other way the markets might evolve, have you got any predictions for us?
Joe: I don’t actually expect wholesale changes in the market structure. Look at the history of the SEC. They like automated trading. They adopted Reg ATS and all the order handling rules back in the ’97-’98 time period. That’s what created the fragmentation that exists today. Then they brought in 16ths and ultimately decimalization. That got rid of the traditional makers. So by 2006 Nasdaq was already dominated by high frequency trading, even though it didn’t get to be a hot topic in the press until 2009. It’s not a new phenomenon.In adopting Reg NMS at in the 2006 timeframe, the SEC basically endorsed the high frequency trading dominated Nasdaq market structure and eliminated the traditional NYSE market structure.
In short, if you look at the SEC’s history, they believe in automated trading, in part because they like the surveillance capability that automation provides. The SEC is actually pretty sophisticated despite the popularity of bashing the regultors. The politicians may jump up and down and shout and say, “oh high frequency trading is ripping off the public institutions”, but if you talk to the SEC, they look at the statistics and say, “no market structure has actually improved”. And every step along the way they’ve encouraged more automation. I don’t think all of a sudden they’re going to stand up, do a 180 and take a mea culpa saying, ”Our bad. The past 15 years of rulemaking has been a mistake. Let’s go back to a manual market”. It isn’t going to happen.
So the changes I foresee are just on certain fronts: circuit breakers, maybe some minimal market maker obligations, broker routing transparency, and some rules on dark pools– basically a few tweaks around the edges of the market structure. We’ll eventually get all this stuff but it’s going to take longer and it’s going to be less intense than people think. They’re not going to outlaw things like co-location, for example or have minimum quoting times. And all the rulemaking being necessitated by Dodd-Frank financial reform will assure that the SEC will not even have the time to mess around too much with market microstructure issues that may be fascinating to someone like me, but will need to take a backseat on this major piece of legislation.
In my view, the changes will be modest. The doomsayers and the people predicting big changes will be proved wrong.
HFTR: Great, thanks Joe.
Related content
