Does Market Microstructure Matter?

The Securities and Exchange Commission has decided to delay for a second time ruling on the application by IEX to register as a national securities exchange. This time they did so without seeking or receiving permission from the applicant, on the grounds that a decision requires clarification of their own order protection rule. Accordingly, they have posted a notice of proposed interpretation and invited the general public to submit comment letters.

The key passage in the notice is the following:

Specifically, the Commission preliminarily believes that, in the current market, delays of less than a millisecond in quotation response times may be at a de minimis level that would not impair a market participant’s ability to access a quote, consistent with the goals of Rule 611 and because such delays are within the geographic and technological latencies experienced by market participants today... permitting the quotations of trading centers with very small response time delays, such as those proposed by IEX, to be treated as automated quotations, and thereby benefit from trade-through protection under Rule 611, could encourage innovative ways to address market structure issues.

Accordingly, the Commission today is proposing to interpret “immediate” when determining whether a trading center maintains an “automated quotation” for purposes of Rule 611 of Regulation NMS to include response time delays at trading centers that are de minimis, whether intentional or not.

If this proposed interpretation is sustained, it seems to me that the application would have to be approved. But perhaps I'm not being cynical enough. There will certainly be a flurry of comment letters from those whose current business models are threatened by the entry of IEX, and it's possible that the delay is intended to provide cover for a change in interpretation on the basis of which the application will eventually be rejected.

But one thing I find encouraging about the notice is that it seems to find persuasive two excellent comment letters by RT Leuchtkafer (I flagged the second when it was submitted but had missed the first). Among the many points made in these letters is the following: if an intentional delay in allowing traders access to quotes is a violation of Regulation NMS, then the entire system of co-location services, differential access speeds, and proprietary data feeds would need to end. Here's the logic of the argument:

If deliberately slower access is an "intentional device that would delay the action taken with respect to a quotation," as the IEX Critics' reasoning certainly implies, the problem isn't just that all the major exchange groups use "delay coils" to equalize access within their data centers. The problem is that you have to pay to get into their data centers in the first place, and if you don't it sure looks like you are intentionally delayed compared to those who can and do pay.

It gets worse! Even within exchange data centers, exchanges charge fees depending on the speed of your connections. A 10gb connection is certainly delayed with an "intentional device" you know, routers, switches and the like relative to a much more expensive 40gb connection, especially when the faster connection is priced out of all proportion to its actual cost, especially when the public SIP feeds have average delays of 500 microseconds to one millisecond and the SEC's own statistics show that billions of quotes are stale before they are ever broadcast by the SIPs.

Where does this logic take us? I naturally started to wonder that if the IEX Critics are right, by their own reasoning the exchanges will have to dismantle their co-location facilities and stop offering tiered high-speed network facilities. They are selling faster access to their markets, and if you don't pay, aren't you slower than you could be, aren't you intentionally delayed?

The critics might want to be careful what they wish for.

I am on the record in support of the IEX application, and hope that the interpretation proposed in the notice is indeed sustained. The IEX design prevents trading based on information from an order that has been partially filled but not fully processed. It therefore moves us closer to a true national market system, in the sense that orders are processed in full in the sequence in which they make first contact with market.

But does all this really matter, except to those whose interests are directly at stake? I believe it does, because the rules governing transactions in asset markets affect the relative profitability of different trading strategies, and this in turn has consequences for share price accuracy and volatility, the allocation of capital across competing uses, the costs of financial intermediation, and the returns to ordinary investors.

There is an extremely diverse set of participants in the secondary market for stocks, with significant differences in goals, investment horizons, and trading strategies. It is useful to group these into three broad categories: (a) long-term investors, who save during peak earning years and liquidate assets to finance consumption during retirement (b) information traders, who seek to profit from deviations between prices and their private estimates of fundamental values, and (c) high-frequency traders, who combine a market-making function with arbitrage and short-term speculation based on rapid responses to incoming market data.

There is clearly a lot of overlap between these categories. For instance, actively managed mutual funds and some hedge funds belong to the second category but often manage money for long-term investors, pension funds, or university endowments.

The traditional market making function involves the placement of passive orders that provide liquidity to the rest of the market. Such passive order placement is subject to adverse selection: if a posted offer to buy or sell is met by an information trader the market maker will suffer losses on average. In order for a market making strategy to be profitable, these losses have to be matched by gains elsewhere. Where do these gains come from?

In standard models of market-making, the bid-ask spread is determined by a balance between losses from transactions with information traders and gains from transactions against those with price-insensitive demands. But this is not the balance that exists in markets today. Instead, high-frequency traders combine passive liquidity provision with aggressive liquidity-taking strategies based on the near instantaneous receipt, processing, and reaction to market data. The posting of bids and offers is motivated less by profiting from the spread than by fishing for information, which can then be used to take and quickly reverse directional positions. The relative weights on passive liquidity provision and aggressive short-term speculation varies considerably across firms, but there is evidence that the most aggressive and profitable among these are able to effectively forecast price movements over very short horizons.

A transition to a truly national market system will affect the competitive balance between information traders and high-frequency traders. It is in the interests of the former to prevent information leakage so that they can build large positions with limited immediate price impact. It is in the interest of the latter to extract this information from market data and trade on it before it has been fully incorporated into prices. Other things equal, the ability to extract information from a partially filled order and trade ahead of it at other exchanges benefits high-frequency traders at the expense of information traders. A truly national market system would mitigate this advantage.

This means, of course, that high-frequency traders would be more vulnerable to adverse selection and would place a lower volume of passive orders to begin with. But the orders would be genuinely available, and not subject to widespread cancellation or poaching if one of them were to trade. Visible bid-ask spreads may widen but there would be no illusion of liquidity.

The shift in competitive balance between these trading strategies would have broader economic implications. The returns to investment in fundamental information would rise relative to the returns to investment in speed, which should result in greater share price accuracy. Furthermore, there is a real possibility that the aggregate costs of financial intermediation would decline, as expenditures on co-location, rapid data processing and transmission, equipment, energy, and programming talent are scaled back. This would be a desirable outcome from the perspective of long-term investors. After all:

It is the iron law of the markets, the undefiable rules of arithmetic: Gross return in the market, less the costs of financial intermediation, equals the net return actually delivered to market participants.

Finally, extreme volatility events should arise less often. Algorithms making short-term price forecasts may predict well on average but they will sometimes mistake a random fluctuation for a large order imbalance. Such false positives can give rise to a hot potato effect, of the kind that is believed to have been in play during the flash crash. Of course, such events can occur even in the absence of market fragmentation, and cannot be prevented entirely, but a transition to a true national market system should reduce their amplitude and frequency.

For these reasons and more, approval of the IEX application would be a modest but meaningful step in the right direction.