Fred Tomczyk responds to a Wall Street analyst who cites leaked first draft of report written by three professors

Brooke’s Note: Fred Tomczyk could easily have brushed aside the Wall Street analyst or kept his comments ultra-lawyerly. Instead, he spoke some pretty plain English to an issue that has long since died down in the news — order routing. Originally, we were going to include those Tomczyk thoughts, abridged, in an earlier story we wrote about TD Ameritrade’s update to Wall Street analysts. See: How TD Ameritrade’s patient, patient bet on RIAs is finally paying off. But before doing so we checked with the professor, Robert Battalio, whose paper came up for some criticism by TD’s CEO. He, too, took some time to explain his side in this circumstance. Finally, we ran his comments past TD Ameritrade and got a handful of thoughts through its public relations veteran, Kim Hillyer. So this exchange of ideas from academic to corporation to academic grew into its own article as published here.

At first, Fred Tomczyk delivered a mild rebuke to the question posed by an analyst during TD Ameritrade Inc.'s July 22 business update.

“I mean to be honest, I’m a bit surprised at how much airtime this whole topic is getting,” he said.

But TD’s president and chief executive nevertheless spent 600 words responding to a question about a report that technically has never been completed or published, though apparently it has enjoyed a wide reading.

The query was lobbed in by Joel Jaffrey, a managing director of equity research at Keefe, Bruyette & Woods Inc. of New York.

“It’s kind of indicated that the practices of online broker activity isn’t necessarily consistent with best execution,” he said. “I was just wondering if you could give us your thoughts on this and anything else relevant to this topic.”

Routing at retail brokerages

The question related not to “Flash Boys, A Wall Street Revolt,” Michael Lewis’ non-fiction narrative on high frequency trading published in March, but rather to a more obscure draft of a paper that reached many influential eyes this year, and that was written by three professors in Indiana.

Shane Corwin and Robert Battalio of the University of Notre Dame Mendoza College of Business and Robert Jennings of Indiana University’s Kelley School of Business focused their reported research on the routing of trade orders at retail brokerages and claim to have found cases where those brokerages benefited to the detriment of clients. See: Chuck Schwab and Walter Bettinger issue call to squash flash traders but RIAs question hysteria over the issue.

In their study, “Can Brokers Have it all? On the Relation between Make Take Fees & Limit Order Execution Quality,” the academics specifically wag an accusing finger at Omaha, Neb.-based TD Ameritrade Holding; Boston-based Fidelity Investments; Scottsdale, Ariz.-based Scottrade Inc.; and New York-based E*TRADE Financial Corp. for allegedly routing all their non-marketable limit orders (those that are outside of that moment’s trading range) to maker/taker exchanges that paid the highest rebates for these orders. See: In wake of high frequency trading hoopla, Fidelity Investments is designing 'Sakura’ to secure better trades for its 'shareholders’.

Inopportune executions

This practice, according to the study, is not in the best interest of the brokerage’s clients because at these exchanges retail brokerage clients are put at the back of the queue at their order price — i.e. their orders always fill last. Should the price of a stock trade at a customer’s limit order, but not exhaust all orders at that price before moving away, a customer of these four retail brokerages won’t get to trade.

The report claims to have found empirical evidence that execution rates were much higher on exchanges with low — or even negative — rebates than on exchanges with high rebates. The professors’ paper also finds that this kind of order routing leads to executions at far less opportune times for clients than if these orders were routed to exchanges with low to negative rebates.

The paper came to light in February when it was leaked to the public and led to a hearing by the U.S. Senate’s Permanent Subcommittee on Investigations chaired by Sen. Carl Levin (D-MI) in June.

In the lead-up to the hearing, TD Ameritrade, under pressure, disclosed 2013 revenues of $236 million from order routing, although the company was keen to point out that less than half of this amount came from stock transactions. See: Why RIAs were the hot topic when 12 Wall Street analysts chatted up TD Ameritrade’s CEO Fred Tomczyk.

Anatomy of a leak

Robert Battalio: [The study] understates the
problem, not overstates it.

When asked about the study on the conference call, Tomczyk attacked its logic, methodology, conclusions, significance — and whether it even qualifies as a report at all.

“First off, the professors’ report is a draft report; it’s not even a final report. So for something to get so much coverage when it’s in draft form is a bit of a surprise.”

Indeed, the authors originally presented the draft of the paper to relevant industry representatives including the Securities and Exchange Commission and the National Association of Securities Dealers and to several brokerages including TD Ameritrade and others named.

“To their credit, TD Ameritrade was the only broker that talked with us,” Battalio says.

Releasing a draft of the study to industry insiders is “a common practice that is generally accepted by all parties to be a confidential forum to test and refine study hypotheses and findings,” according to a release from the University of Indiana on Feb. 5, after the paper’s leak.

The leak to the broader financial community was done without the authors’ knowledge or permission according to the release, which goes on to say that some unnamed people were so impressed by the report’s contents that they suggested its impact could equal that of a 1994 study by Bill Christie of Vanderbilt University and Paul Schultz of Notre Dame. That report showed implicit collusion among Nasdaq market makers and it led to sweeping reform of Nasdaq market (and a billion-dollar legal settlement in the lawsuit between a class of investors and the Nasdaq], according to a release from the University of Indiana.

The paper’s leak and the Financial Industry Regulatory Authority Corp.'s subsequent request for routing data from the 50 largest brokers — has the authors concerned that the findings about brokers’ maximizing liquidity rebates might be oversimplified, according to the University of Indiana’s release.

Subsequently, in March, the three professors submitted their study to The Journal of Finance, published by the American Finance Association, for review. That publication has recently asked the authors to do revisions before giving a final approval to publish. Prof. Battalio expects that revising task to take another two to four months.

“We’re just going to tighten the argument,” says Battalio, although he stresses that the original paper’s conclusions will not be modified. In fact, he and his colleagues have bolstered their claims with more evidence they have since discovered.

Tomczyk takes exception

At TD’s business update in July, Tomczyk downplayed the report’s application to TD Ameritrade, saying non-marketable limit orders are a relatively small percentage of TD’s total trades — about 12% — and explained the great lengths to which TD Ameritrade goes to ensure best execution for its clients’ trading orders.

Then he challenged the logic underpinning the paper as regards to where exactly TD’s economic incentives lie.

“When you think about it, you’ve got to remember the base commission of $9.99. If we don’t get a trade execution, not only do we miss the rebate, we also miss the $9.99 base commissions, unlike a lot of institutional trade volume, which is four to five times the rebate. Anybody that looks through that is going to see that we have an economic incentive to get as many trade executions on the non-marketable winnable orders as possible,” Tomczyk said.

But Battalio counters by reasoning that any retail investor whose order goes unexecuted is very likely to just cancel and replace it, giving their broker another chance to earn a commission.

TD Ameritrade declined to address Battalio’s argument related to canceling and replacing.

Not like the other

Battalio also expresses doubts that the economic calculus is as linear as Tomczyk describes, as brokerages earn 30 cents for every 100 shares routed to high-rebate exchanges. Thus, even if a few commissions are lost because orders aren’t executed, brokerages could hypothetically still earn more overall by routing to high-rebate exchanges than routing to low/negative-rebate exchanges with better execution rates.

TD questioned the assumptions behind this rebuttal.

“Our intent was not to engage in a public debate on an issue that we believe is best handled by regulatory bodies, like the SEC, and we don’t plan to do that now,” says Kim Hillyer, director of communications and public affairs for TD Ameritrade. “Rather, we were the subject of an academic study that used institutional trading data to draw conclusions about retail order flow — two very different things — and we felt it was appropriate to offer our view of the analysis. We have done so and continue to stand by what we said: that we take our best execution responsibilities very seriously and focus on it first and foremost when making order routing decisions.” See: Schwab’s website went down twice after two 'denial of service’ attacks — so what was up?.

Methodology in question

After criticizing the paper’s logic, Tomczyk next addressed the methodology of the academics.

“When you read into the details of it, the report is based on institutional order flow and, to make matters even worse, it includes institutional proprietary algorithmic trading, and this extrapolates conclusions from that flow to retail order flow. That’s like comparing apples to oranges. They are two very different things, and if you want to try to draw conclusions from one to another, you do so at your own peril.”

Hillyer goes further. “That study relies on data that should not be used to draw such conclusions.”

The retail brokerages’ academic accusers are willing to concede this point — with a few caveats.

“We don’t disagree,” says Battalio, referring to Tomczyk’s comments on the data. But Battalio goes on to note that the study does analyze data from the whole market in the final third of the paper. He points out, too, that regarding the section Tomczyk criticizes, the investment bank that supplied the data used a smart-order routing system to manage order flow. Because of this, Battalio and his colleagues believe their paper suggests an even more flattering picture of retail broker habits than exists. See: Preparing to appear before Senate subcommittee, TD Ameritrade discloses order-routing revenue.

“It understates the problem, not overstates it,” Battalio says.

The 7% conundrum

“Lastly,” Tomczyk concluded on the call, “we’ve looked at his findings and we’ve run tests on it with our own data to make sure we weren’t missing something, and we had Direct Edge recently examine all of our clients non-marketable limit order routed to their venue for a two-week period. And [in] the preliminary results, approximately 93% of the orders were executed on Direct Edge, provided there was a trade on any exchange at the limit price.”

Battalio questions the fairness of having Direct Edge, the high-rebating exchange to which Ameritrade routed all of its non-marketable limit orders and Ameritrade analyze and report their own numbers.

But he is pleased to hear Tomczyk say, however unintentionally, that there were problems with executions.

“In my view, what he said could be totally true but it still proves our point,” Battalio says. “What happened to the other 7%? They missed out on valuable trading opportunities.”

This is a significant question. As 12% of TD Ameritrade’s estimated 90,000,000 annual orders are non-marketable limit orders, if 7% of annual orders aren’t executed because they’ve been routed to an exchange, that doesn’t offer best execution and up to 750,000 client orders per year are lost.

But TD Ameritrade disputes that brokerages could get 100% execution on non-marketable limit orders that become marketable. Kim Hillyer says that is an “interesting assumption” and that “we feel good” about the 93% rate.

According to Hillyer, TD Ameritrade does not “have any plans to adjust our order routing strategy at this time.”

Further documentation

The study might be the first one in which retail brokerages have been accused of impropriety, but it isn’t the first documentation of the inferiority of high-rebate, maker/taker exchanges for non-marketable orders. There’s plenty of other research to support the fundamental assertions of Battalio, Corwin and Jennings. See: Unfazed by its misfire, BlackRock is taking a second shot at the 401(k) market, this time with a whiter hat.

In 2010, Sanford C. Bernstein & Co.'s research division found that for highly liquid stocks under $40, fill probability on inverted exchanges (i.e. exchanges that charge, instead of pay, brokerages for non-marketable limit orders) was 1.5 times to 3.5 times higher than on exchanges with the highest rebates.

And it’s not only fill rates that are improved when non-marketable limit orders are routed somewhere other than a high-rebate maker/taker exchange. Sanford Bernstein also found that value of the traded asset increases by about two basis points because the quality of the execution is much higher. See: Flash crash update: Why the multi-asset meltdown is a real possibility.

Goldman Sachs’ chief equities executions strategist, George Sofianos, independently confirmed these findings in a January 2011 Street Smart report. He found inverted exchanges had significantly higher fill rates and lower toxicity.

Open secret

And during the Senate hearing last month, Brad Katsuyama, the protagonist of “Flash Boys,” claimed that while he was working at RBC the bank ran tests that confirmed the findings of Corwin, Jennings and Battalio. RBC discovered, Katsuyama said, that “routing specifically with the goal of maximizing your rebate lowers the probability of getting filled and leads to adverse execution quality, or lower execution quality, for the clients.”

Later in the hearing, under questioning from Sen. Levin, TD Ameritrade’s senior vice president, trader group, Steven Quirk, admitted that, indeed, his firm had for years been sending every single non-marketable limit order to just a single exchange—and that this exchange just happened to pay the highest rebates.

The study can be found here.

Here’s the full exchange between Jaffrey and Tomczyk:

Joel Jaffrey – Keefe, Bruyette & Woods
Okay, great. And then just lastly, I appreciate the disclosure this quarter on the payment for order flow, but I wanted to ask you, just a question on some of the recent testimony that we’ve seen coming from some of the hearings, it’s kind of indicated that the practices of online broker activity isn’t necessarily consistent with best execution. I mean I think there was one of the people testifying, there was a professor at Notre Dame that basically said that you have a 25% less chance of being executed based on some of these practices. I was just wondering if you could give us your thoughts on this and anything else relevant to this topic.

Fred Tomczyk – president & chief executive, TD Ameritrade
Okay, thanks Joel. I mean to be honest, I’m a bit surprised at how much airtime this whole topic is getting and for a couple of reasons. First off, the professor’s report is a draft report; it’s not even a final report. So for something to get so much coverage when it’s in draft form is a bit of a surprise. And he does emphasize that it may be happening, he doesn’t say it is happening, and I’ll come back to that in the second.

Second, the focus of his report is on non-marketable limit orders. Non-marketable equity limit orders and that represents about 12% of our trades and a lot of people have sort of implied it across all of our trades, which is incorrect.

And the point that the paper is making and it was made at the hearing is that it’s based on their analysis, we and other brokers may be prioritizing rebates over best execution as you said. And I suggest then, because this is non-marketable limit orders, we may be missing out on trade executions we could have gotten if we were more focused on best execution.

Now I thought about this a fair bit and so let me make a few points. First off, as I said last quarter, we take best execution and our best execution responsibility seriously and focus on it first and foremost. Unless a market destination or venue meets our criteria for best execution, they don’t even get a second look. And after and only after we’ve satisfied best execution do we minimize trading cost and/or optimize revenue sharing.

The second point, the underlying economic logic in the paper we find wanting or flawed, suggests that we’re focused on maximizing rebates so much that we may be missing out on trade executions, and that you know from our perspective, that would be an entirely economically irrational thing for us to do, even though the paper is arguing that that’s in our economic interests.

When you think about it, you got to remember the base commission of $9.99. If we don’t get a trade execution, not only do we miss the rebate, we also miss the $9.99 base commissions, unlike a lot of institutional trade volume, which is four to five times the rebate. Anybody that looks through that is going to see that we have an economic incentive to get as many trade executions on the non-marketable winnable orders as possible.

And third, that report and the analysis, when you read into the details of it, the report is based on institutional order flow, and to make matters even worse, it includes institutional proprietary algorithmic trading, and this extrapolates conclusions from that flow to retail order flow. That’s like comparing apples to oranges, they are two very different things, and if you want to try to draw conclusions from one to another, you do so at your own peril.

Lastly, we’ve looked at his findings and we’ve run tests on it with our own data to make sure we weren’t missing something, and we had direct edge recently examine all of our clients non-marketable limit order routed to their venue for a two-week period. And the preliminary results, approximately 93% of the orders were executed on direct edge, provided there was a trade on any exchange at the limit price. So that confirmed our views and we are not surprised, and when you combine what we see as flawed logic and flawed methodology, it’s not surprising you get flawed conclusions and the reality is the facts just don’t bear out and that story is just not getting out.

Jaffrey
Great, thanks. I appreciate the color.