Mary Jo White is going to think about some things.  Photographer: Andrew Harrer/Bloomberg
Mary Jo White is going to think about some things. Photographer: Andrew Harrer/Bloomberg

Securities and Exchange Chair Mary Jo White gave a big speech today about market structure and high frequency trading in which she said that the SEC is planning to do a lot of looking into things. For my money the most important sentence in her speech is this:

For institutional investors, the costs of executing large orders, measured in terms of price, were more than 10 percent lower in 2013 than in 2006.

There are I think two important and hard to dispute empirical facts about high-frequency trading:

  1. Explicit trading costs (commissions, etc.), for both individual and institutional investors, are lower than they were before the rise of algorithmic trading.1
  2. Individual retail investors get a much better deal than they did before the rise of algorithmic trading.2

But the fact that White leads with -- that all-in costs of executing large orders are lower now than they used to be -- is, I think, a disputed fact.3 And White takes a side. And it's the side of the high frequency traders.4

A lot of Weltanschauung follows from the side you take on that empirical question. If you think that high-frequency trading is fundamentally bad for institutional investors, then you probably think that HFT is about stealing from institutional investors. When institutional investors can't execute large orders at displayed prices, because HFT firms quickly change their bids and offers to get ahead of those orders, you probably think that that's "front-running." You think markets are rigged against real investors and in favor of value-destroying unwanted middlemen.

If you think that high-frequency trading is fundamentally good for institutional investors, then you take a much more benign view of their tactics, and of the ecosystem that supports them. You think of them, not as thieves, but as efficient market-makers who quite properly replaced older, less efficient, market-making businesses. When HFT firms quickly change their markets to get ahead of large institutional orders, you think that it's just good sensible risk management for a market maker to try to avoid adverse selection by moving its markets quickly. You're glad that the market-makers' quick reaction time enhances market efficiency and allows them to offer much tighter spreads than slower, old-school market makers could have.

Now, your view on this question doesn't determine every decision that you make. But it determines the big ones. New York Attorney General Eric Schneiderman, for instance, pretty clearly believes that high-frequency trading is Bad. He thinks the whole thing is "Insider Trading 2.0." So his goal seems to be to eradicate it entirely, by going after direct feeds from exchanges and co-location and even just the continuous trading of stocks.5 In Schneiderman's ideal world, speed would simply be eliminated as a trading advantage.

That is not Mary Jo White's world, and it shows in her proposals for improving the markets. If you think that the current HFT business model is basically good for the world, then you will be hesitant to make fundamental changes to it. And if you're the SEC, and you're under a lot of Michael-Lewis-driven pressure to make fundamental changes to market structure, what do you do?

Disclosure. You do disclosure. To be fair, disclosure is the SEC's answer to most questions, but it's especially the answer to questions that the SEC doesn't especially want to talk about.

So, for instance, one concern that investors have is that high-frequency traders use faster direct feeds from exchanges to pick off investors who rely on the slower, official consolidated data feed.6 One way to resolve this would be to ban speedy direct feeds, which is more or less what Eric Schneiderman wants.7 Here's Mary Jo White's approach:

I am also asking the exchanges and FINRA to consider including a time stamp in the consolidated data feeds that indicates when a trading venue, for example, processed the display of an order or execution of a trade. With this information, users of the consolidated feeds would be able to better monitor the latency of those feeds and assess whether such feeds meet their trading and other requirements.

And I am asking the exchanges to develop proposed rule changes to disclose how -- and for what purpose -- they are using data feeds. For example, which data feeds are used to execute and route orders? And which feeds are used to comply with regulatory requirements, such as trade-through rules? Brokers and investors could use the enhanced transparency to better assess the quality of an exchange’s execution and routing services.

So exchanges can still use slower feeds to price trades while selling faster feeds to HFT firms. They just have to tell tell investors what they're doing, so the investors can judge for themselves if they're getting ripped off.8

Do you worry that brokers are acting against their customers' interests in order to take advantage of exchanges' maker-taker fees and rebates? Maybe you could ban those fees and rebates. Or not. You could just tell people more about them:

I have asked the staff to prepare a recommendation to the Commission for a rule that would enhance order routing disclosures. Rule 606 of Regulation NMS currently requires some public disclosure of broker order routing practices, but it does not cover the large orders typically used by institutional investors. The rule proposal would address this gap by requiring disclosure of the customer-specific information that a broker is expected to provide to each institutional customer on request.

Are you suspicious of dark pools?

Just this week, FINRA began disseminating aggregate information on trading volume of ATSs. This is a useful first step, but ATSs represent less than half of dark venue volume. To remedy this gap, I fully support FINRA in considering an expansion of its trading volume disclosure regime to off-exchange market makers and other broker-dealers.

I also have asked the SEC staff to prepare a recommendation to the Commission to expand the information about ATS operations submitted to us and to make the information available to the public. As you have seen in the recent media, some operators of dark venues began offering greater transparency to their operations this week, but a broader effort is needed.

You get the idea.9 The SEC's core view is that the fundamental business model of high frequency trading is fine. There are probably some abuses at the margins, and shedding some light on those margins will be enough to correct those abuses. But for the most part, White thinks, our market structure is nothing to be embarrassed about, so there's no reason to fear broader disclosure.

To be fair, White's speech is not just about disclosure. In particular, whatever you think about the ordinary-course economic impact of high-frequency trading, you might worry that unchecked algorithms can occasionally go haywire and destabilize markets. White says, "as I have said from the day I took office, one of the most serious concerns about today’s equity markets is the risk of instability and disruption," and she describes a variety of circuit-breaker-type rules that the SEC has already implemented.

Now she wants unregistered proprietary trading firms, and their algorithms, to be subject to more SEC oversight to reduce the risk of the algorithms running wild. She also proposes new rules against "disruptive trading," which would "apply to active proprietary traders in short time periods when liquidity is most vulnerable and the risk of price disruption caused by aggressive short-term trading strategies is highest." Again: "Aggressive short-term trading strategies" are not a fundamental problem to be stamped out; they're just something you might want to keep an eye on when things are crazy.

White also recognizes that the basic structure of U.S. equity markets -- a structure due largely to SEC rules -- has come in for some criticism. And the SEC will be doing some serious reflection in the coming months:

We also are continuing to consider whether more fundamental changes are needed to bring our regulatory structure in line with the significant market changes of the last decade. Importantly, we will be considering whether the SEC’s own rules, such as the trade-through rule of Regulation NMS, have contributed to excessive fragmentation across all types of venues.

We also will be considering whether the current regulatory model for exchanges and other trading venues makes sense for today’s markets.

So we'll see what comes of that. Certainly a lot of people on all sides of the HFT debate have their doubts about the current trade-through rules.10 But that passage about possible fundamental changes comes pretty far down in the speech, and doesn't seem all that heartfelt.11 Fundamentally, the SEC thinks things are great.

1 I think literally no person disputes this fact. However, critics of high-frequency trading point out that a lot of the cost reductions are due to improving technology, not to the business of actually existing algorithmic trading firms. I suppose that's true, though they're not easy to disentangle. You don't get cost reductions from hypothetical computerized trading; you get them from the computerized trading that actually exists.

2 Hahahahahahaha this is a crazily disputed fact! In particular, Michael Lewis himself runs around saying that high-frequency trading is rigged to hurt the average investor. But the disputes about this fact seem to be entirely uninformed. Really:

  • Spreads, commissions, etc. are way down -- you can trade all the shares you want for under 10 bucks.
  • The market is not going to run away from you if you put in a 100-share order: Big guys have to worry about being "front-run" or whatever, but small investors can just trade at the displayed price.
  • In fact they can often trade at better than the displayed price, because order internalizers offer price improvement for the right to trade against uninformed retail order flow.

As Felix Salmon says:

I’ll tell you what happens when the little guy presses that key: his order doesn’t go anywhere near any stock exchange, and no HFT shop is going to front-run it. Instead, he will receive exactly the number of shares he ordered, at exactly the best price in the market at the second he pressed the button, and he will do so in less time than it takes his web browser to refresh. Buying a small number of shares through an online brokerage account is the best guarantee of not getting front-run by HFT types.

The problems of high frequency trading are about its interaction with big, informed orders. We have discussed this before, when Cliff Asness wrote that "there has been one unambiguous winner" from high frequency trading, "the retail investors who trade for themselves." My own view is that a regime where retail investors (rich hobbyists) win at the expense of institutional investors (pension funds and my 401(k)) is a bad regime, though that seems not to be the view of the SEC. But that's not important right now; what's important is that individual investors definitely win in the current regime. All of the real action is in figuring out whether institutional investors win or lose.

Incidentally whenever I hear people say things like "retail brokers sell your orders to high-frequency traders, who look at each order, decide whether it's informed or not, and then either trade against it or pass it on to public exchanges," I kill a kitten. If you are ever tempted to write a sentence like that, let me recommend Larry Harris's book "Trading and Exchanges," the most sustained sensible writing you're ever likely to read about market structure.

3 I mean, for instance, "Flash Boys" itself quotes a lot of money managers complaining about high frequency trading. They at least perceive that they're getting ripped off, and there are some studies that support this view.

It's worth spending a minute to understand why this question doesn't have an obvious answer. When you buy 100 shares of a liquid stock, you pay a $10 commission and buy at (or below) the displayed inside offer. When you buy 10,000 shares, though, you probably don't get all of those shares at the displayed offer. Maybe you get 1,000 at the offer, and then all of a sudden all the displayed bids and offers move up, and you end up paying an extra 10 cents for the other 9,900 shares. Here some obvious first-order measures of trading cost (bid/offer spread) suggest that your costs are low, but the fact that the market moves so quickly makes your actual cost high. But, of course: If the market makers couldn't move so quickly, they'd have displayed a wider quote to begin with. So it's hard to know whether you're ahead or behind, on net.

4 She cites the ITG database of trading costs, which measures "implementation shortfall," the difference between the price when institutions send an order and the price they get. This is not a perfect measure! Market structure affects how you send an order, which affects measurement of implementation shortfall. But it's a sensible measure, and much better than just displayed spreads or whatever. (White also cites academic studies that support her conclusion that overall costs are down.)

5 Really!

To address this imbalance in the markets, which now tilt in favor of high-frequency traders, Attorney General Schneiderman today called on the exchanges and other regulators to review the feasibility of certain market structure reforms that could help eliminate some of the fundamental unfairness in our markets. Currently, securities are traded continuously, so that orders are accepted and matched by price, with ties broken by which order arrives first. This system emphasizes speed over price, rewarding high-frequency traders for flooding the market with orders. One detailed proposal would seek to correct this imbalance by processing orders in batches in frequent intervals, to ensure that price – not speed – is the deciding factor in who obtains a trade.

6 The problem is basically that an exchange or dark pool can't trade at a worse price than the consolidated national best bid or offer. But if, say, Nasdaq gets a bid that is better than the previous national best bid, there's a fraction of a second where HFT firms who have the direct Nasdaq feed know that but the consolidated data feed doesn't. So the HFT firms can go to a dark pool or another exchange and trade at the stale price before it updates, ensuring themselves a profit.

7 More specifically, he thinks that co-location and "ultra-fast connection cables," which allow HFTs to take advantage of this latency arbitrage between feeds, "in the hands of predatory high-frequency traders distort our markets."

8 The time-stamp thing seems like such an obviously good idea that it's hard to see why they don't do it now. The idea is that if you can tell (1) when you traded, (2) what the consolidated-feed market looked like at the time, and (3) what the actual market in each exchange looked like at the time, you can tell whether you were being taken advantage of by HFTs who have direct feeds (and, if so, by how much).

9 What about the proliferation of order types?

I am asking the exchanges to conduct a comprehensive review of their order types and how they operate in practice. As part of this review, I expect that the exchanges will consider appropriate rule changes to help clarify the nature of their order types and how they interact with each other, and how they support fair, orderly, and efficient markets.

Etc.

10 Those rules, very roughly speaking, require traders to buy at the lowest price (sell at the highest price) offered on any displayed venue, which can force investors to split up orders, use many different venues, and generally be more vulnerable to predation than they would be if they could just decide for themselves where they wanted to trade.

11 Certainly it's not "hey let's rethink the whole concept of a continuous limit order book," ha.

To contact the writer of this article: Matt Levine at mlevine51@bloomberg.net.

To contact the editor responsible for this article: Tobin Harshaw at tharshaw@bloomberg.net.