Streetwise Professor

It is certainly the case that the controversies and conflicts over market structure have continued unabated since 2005, and show no signs of letting up. (Cf. Flash Boys.) Chairman Piwowar’s call for a new Special Study is testament to that.

More specifically, the major prediction of my article has been fully borne out. I predicted that the Order Protection Rule in particular would break the network effect that resulted in the dominance of the NYSE in the securities it listed. Since RegNMS was passed, the highly concentrated listed stock market (where virtually all price discovering transactions in NYSE stocks occurred on the NYSE) has been utterly transformed, with four exchanges now splitting most of the business, with no exchange doing more than a quarter of the volume.

I further predicted that this would result in the disintermediation of traditional intermediaries–like specialists–and the substantial erosion of economic rents. This too has happened. This is best illustrated by the trajectory of Goldman’s investment in specialist firm Spear, Leeds & Kellogg. Goldman paid $5.4 billion for it in 2000 (before RegNMS) and sold it for a pittance–$30 million–in 2014. I didn’t foresee exactly the nature or identity of the new intermediaries–HFT–but I was broadly aware that there would be entry into market making, and that this would reduce trading costs and undermine incumbents with market power. Further, as I’ve written about recently, the new intermediaries don’t appear to be making rents in the new equilibrium.

The years since RegNMS have seen a dramatic decline in trading costs for investors, and it is likely the case that this decline is largely attributable to the increase in competition. Much of the controversy that has raged since 2005 relates to disputes over trading practices that were an inevitable consequence of the breaking of the NYSE near-monopoly–a process pejoratively referred to as “fragmentation.” In particular, multiple markets necessitate arbitrageurs, who effectively enforce the law of one price. The strategies and tactics arbitraguers use often appear unsavory, and strike many as unfair: arbitrageurs get something even though they appear to do nothing substantive. Moreover, arbitrage uses up real resources. That’s costly, and it would be nice if this could be avoided, but that’s unlikely ever to be so. The trade-off between much greater competition (and reduced welfare losses due to the exercise of market power) and the expenditure of real resources to enforce the law of one price seems to be a great bargain.

Much of the criticism of RegNMS relates to the Order Protection Rule, which requires that no order can be executed on market X if a better price is displayed at market Y. The critics (e.g., the Principal Traders Association which ironically represents some of the biggest beneficiaries of RegNMS) argue that this rule (a) has led to a proliferation of order types intended to ensure compliance with the rule, which make the market far more complex, and (b) requires traders to maintain connections with and monitor all trading venues displaying quotes, no matter how small.

These complaints have some merit. The crucial question is whether the equity trading marketplace will be as competitive without the Order Handling Rule as it is with it. This is an open question, and one which should be the focus of the SEC’s inquiry. For if the Order Handling Rule is a necessary condition for robust competition, the costs that the PTA and others identify are likely well worth paying in order to realize the benefits of competition.

My prediction that competition would intensify post-RegNMS was based on my analysis of the effects of the Order Handling Rule, which was in turn based on my work on liquidity network effects done in the late-90s and early-00s. Specifically, in the formal models I derived (e.g., here), the self-reinforcing liquidity effect obtains when investors decide which trading venue to submit an order to on the basis of expected execution cost (i.e., bid-ask spread, price impact). The market with the bigger fraction of trading activity typically offers the lowest execution cost. Therefore, traders submit their orders to the bigger market. This creates a self-reinforcing feedback loop (and a self-fulling prophecy) in which trading activity “tips” to a single exchange. (There are some complexities here, relating to cream skimming of uninformed order flow. See the linked paper for a discussion of that issue.)

Mandating something akin to to the order handling rule forces order flow to the market offering the best price at a particular moment, not the one that offers the best price in expectation. As I phrased it in my Regulation paper, such a rule “socializes order flow”: even if an order is directed to a particular exchange, that exchange does not control that order flow and must direct to any other exchange offering a better price.

I think that both theory and the post-RegNMS experience show that the Order Handling Rule is sufficient to break the liquidity network effect because it socializes order flow. But is it necessary? Maybe not, but it is important to try to find out before jettisoning it.

Here’s a story which suggests that the rule is not necessary in the modern electronic trading environment. One reason why traders may choose to submit orders to where they expect to get the best execution is because of search costs. In a floor-based environment in particular, it is costly to verify which market is offering the best price at any time.  Moreover, since it takes time get quotes from two floor-based markets, by the time that you actually submit your order to the one giving the best quote, the market will have moved and you won’t get the price you thought you were going to get. So economize on search costs and the risks associated with delay by submitting the order to the market that usually offers the best price. Ironically, the inevitable result of this process is that there is only one market left standing.

Search is cheaper and faster–and arguably far cheaper and far faster–in the modern electronic environment. Based on feeds from multiple markets, an electronic trader (and in particular an automated trader) can rapidly compare quotes and send an order to the market offering the best quote, or by viewing depth (something pretty much impossible in the floor days, where much of the liquidity was in the hands of floor brokers) split an order among multiple venues to tap the liquidity in all of them.

In other words, the natural monopoly problem was far more likely in a floor-based environment where pre-trade transparency was so limited that search costs were very high: it was nigh on impossible to know precisely what trading opportunities were or to move fast enough to exploit the one that appeared best at any point in time, so traders submitted their orders to where they expected the opportunities to be the best. In contrast, electronification and automation have created such great pre-trade transparency and the ability to act on it that it is plausibly true that in this environment traders can and will submit their orders to whatever venue is offering the best trading opportunity at a point in time, regardless of whether it usually does so. In this story, technology eliminates the uncertainty and guesswork that created the liquidity network effect.

Maybe. Perhaps even likely. But I can’t be certain. Note that one complaint about the existing market structure is that even though everything has vastly speeded up, some traders are still faster than others. As a result, those who submit a market order in response to seeing a particular displayed price are often dismayed to learn that the market has moved before their order actually reaches the trading venue, and that their order is executed at a worse price than they had anticipated. Freed of the obligations of the Order Handling Rule, these traders may choose to submit their order to where they usually get the best price: if enough do this, the liquidity network effect will reemerge.

Further, the PTA and others have complained that it is costly to monitor and maintain connections with all trading venues as is necessary under the Order Handling Rule. If the Rule is relaxed or eliminated, one would expect that they will disconnect from some venues. If enough do this, the smaller venues will become unviable. After this happens, there will be fewer venues–and some traders may choose to disconnect from the smallest remaining one. This dynamic could result in another feedback loop that results in the survival of a single dominant exchange that exercises market power.

It is therefore not clear to me that elimination of the Order Handling Rule will result in traders having their cake (intense inter-exchange competition) and eating it too (less complexity, lower connection cost). Given the substantial benefits of greater competition that have been realized in the past dozen years, changes to the cornerstone of RegNMS should not be taken lightly. The Special Study, and the SEC, should pay close attention to how competition will evolve if the Order Handling Rule is eliminated. This analysis should take into account the existing technology, but also try to think of how technology will change in the aftermath of an elimination and how this technological change will affect competition.

Most importantly, any analysis must be predicated on an understanding that there are strong centripetal forces in securities trading. Any time traders have an incentive to direct order flow to the venue that is expected to offer the best price, the likely outcome is that only one venue will survive. The incentives of traders in a high speed, largely automated, and electronic market in the absence of an Order Handling Rule need to be considered carefully. It should not be assumed that technology alone will eliminate the incentive to direct orders to the market that is usually best, not the one that is best at any particular instant. This hypothesis should be probed vigorously and skeptically.

Experience in futures markets suggests that liquidity network effects can persist even in high speed, automated, electronic markets: futures contracts in a particular instrument exhibit a strong natural monopoly tendency, and strong tendencies towards tipping. It is arguable that the vertical integration of clearing, and the resulting non-fungibility of otherwise identical contracts traded on different venues, could contribute to this (though I am skeptical about that). But it could also mean that something like the Order Handling Rule (which is not present in futures markets) is necessary to create strong competition between multiple venues even in a highly computerized and automated trading environment.

This is the big issue in any revamping of RegNMS. It should be front and center of any analysis, including in the impending Special Study. The intense competition in the post-RegNMS world is a remarkable achievement, particularly in comparison with the near monopolistic market structure that existed before 2005. It would be a great shame if this were thrown away due to an incomplete analysis of what competition in a modern computerized market would be like in the absence of something like the Order Handing Rule.