Speech by SEC Staff:
Keynote Speech at the SIFMA 2008 Dark Pools Symposium
Erik R. Sirri
Director, Division of Trading and Markets
U.S. Securities and Exchange Commission
New York, NY
February 1, 2008
Thank you for the invitation to talk this morning about dark pools of liquidity and their role in equity market structure. Before I begin my remarks, let me remind you that the views I express are my own and not necessarily the views of the U.S. Securities and Exchange Commission, the individual Commissioners, or my colleagues on the Commission staff.
In the U.S., we have a culture that highly values transparency in nearly every aspect of life, from politics to financial markets to celebrity lifestyles. We feel uneasy when important events, particularly those that may affect our political or economic wellbeing, seem to be happening behind closed doors. Over the last few years, a large number of dark pools of liquidity have been launched in the U.S. equity markets. These distinct pools of liquidity are a relatively new phenomenon in the equity markets and have sparked considerable discussion, if not concern. This symposium offers a good opportunity to further that discussion.
I want to examine how dark pools fit with the fundamental policy objectives for the securities markets as set forth by Congress in the Exchange Act. These objectives include, among others, the efficient execution of transactions, price transparency, fair competition, and best execution of investor orders. Along the way, I hope not only to clear up some misplaced concerns, but also to highlight potential trouble spots that warrant close watching as U.S. equity market structure continues to evolve.
We need to be precise about what we mean when we talk about a "dark pool." It is a term that can refer to a wide variety of trading venues, or just one of many trading services that are offered by a particular venue. For my purposes this morning, I want to define "dark pool" very broadly and say it is a service offered by any trading venue that offers the potential to interact with liquidity beyond the venue's public quotes, if indeed the venue quotes at all. This broad definition should serve to highlight an important point — dark pools of liquidity have been around for a long, long time. The single largest dark pool in the world for many decades could be found on the trading floor of the New York Stock Exchange. The floor traders there manually represented a pool of undisplayed liquidity that could be accessed only by sending an order to the floor to probe buying and selling interest.
Dark pools are solutions to a perennial trading dilemma for anyone that needs to trade in substantial size, particularly institutional investors. They provide a mechanism for such transactions to interact without displaying the full scale of their trading interest. Today, nearly every equity trading venue in the U.S. offers some sort of dark liquidity. Consequently, I do not find that the term "dark pool" is all that helpful in analyzing market structure and want to suggest a more precise classification of trading venues.
First, I will divide trading venues into two categories based on whether they display quotes as an integral part of their business models. These I will call "quoting venues," and all others I will call "dark venues." The quoting venue category is made up of venues that display quotes as an integral part of their business model, even if they also offer one or more dark liquidity services. Exchanges, for example, long have sought to offer a total package of liquidity services that are as deep and wide as possible. Their dark liquidity services have included floor brokers on a trading floor, undisplayed orders in an automated trading system, and the undisplayed, or reserve, size of displayed orders in an automated trading system. The other type of quoting venue in today's market structure is the ECN, or electronic communication network. An ECN also is an ATS, or alternative trading system, but is distinguished from a dark ATS by its display of quotes.
The second, or non-quoting category of venues should be considered dark venues. These fall into two types — broker-dealer internalizers and dark ATSs. Obviously, many broker-dealers quote to some extent, but this quoting, and the resulting trades, generally occur on an exchange or ECN. In contrast, most of the executed volume of broker-dealer internalizers reflects liquidity that was never quoted to the marketplace. Finally, it is the last type of trading venue — the dark ATS — that has become so numerous in recent years and has received the most attention. This four-part categorization of trading venues reveals, if nothing else, that while the particular organizational structure of dark ATSs may be relatively new, the function of offering dark liquidity is not new at all.
The four categories of trading venue highlight a quite positive aspect of U.S. equity market structure. The U.S. has a wide range of trading venues, and the competition for order flow among them is exceptionally strong. This wide range of competing markets is the hallmark of the national market system approach to market structure that was mandated by Congress in the Exchange Act. Today, investors and traders have never had a wider range of choices of trading venues that compete to satisfy their particular trading needs, and the competition among these venues has never been stronger. Many dark ATSs exist, in large part, due to competitively-driven attempts to service the particular trading needs of different types of investors and traders.
I believe that competition among diverse trading venues is a tremendous strength of U.S. equity market structure. One of the common concerns expressed about the rise of dark ATSs as a trading venue, however, is that they threaten to supplant quoting venues and cause the equity markets to become less transparent. At least thus far, this concern does not appear to be well-placed. Rather than focusing too narrowly on the expanding trading volume of dark ATSs, a more telling measure of the health of quoting venues is to compare the cumulative trading volume of quoting venues with the cumulative trading volume of dark venues.
As I will discuss shortly, these volume percentages reveal that quoting venues collectively have maintained a remarkably stable percentage of total equity trading. Instead of a migration of trading volume from quoting venues to dark venues in recent years, most of the movement in trading volume has been within each of the two categories of quoting venues and dark venues; that is, volume has shifted among various quoting venues and among various dark venues, but has not shifted out of quoting venues into dark venues. The quoting venues — exchanges and ECNs — have engaged in a fierce battle for order flow for many years. Similarly, the increasing percentage of trading volume of dark ATSs largely has come as the percentage of trading volume of broker-dealer internalizers has declined.
This morning I do not intend to weary you with the fine points or difficulties of calculating trading volume statistics. Rather, I simply will compare some basic market share statistics from three years ago in December 2004 to those same statistics today. In December 2004, the total exchange and ECN share of trading was approximately 85% in NYSE stocks and 71% in Nasdaq stocks. By December 2007, total exchange and ECN share of trading was approximately 83% in NYSE stocks and 77% in Nasdaq stocks. In other words, NYSE volume percentage in quoting venues is roughly flat around 84%. Moreover, quoting venues increased their percentage share substantially in Nasdaq stocks. If these NYSE and Nasdaq percentages are rounded together loosely for simplification, the bottom line is that the volume percentage of dark pools of liquidity operated by dark ATSs and broker-dealer internalizers has remained at approximately 20% over the last three years.
The most noticeable shift in trading volume during this period is the decline in the NYSE's share of volume in its listed stocks from 78% in 2004 to 41% last month. Other quoting venues, however, have captured this large share of volume. Last month, these other quoting venues, particularly Nasdaq, NYSE Arca, BATS ECN and Direct Edge ECN, executed more than 40% of volume in NYSE stocks.
The volume statistics for any particular dark venue can be extremely difficult to assess. Broker-dealers generally do not publicly report their internalization volume in ways that can be attributed to them, and the varied reporting practices of the dark ATSs can render volume comparisons shaky at best. What does seem clear is that dark ATSs are increasing their share of trading volume, which necessarily means that the share of trading volume of broker-dealer internalizers is declining. This trend reflects an important market development that started with decimalization in 2001, when much narrower spreads led to a decline in profitability of the internalization business model. Broker-dealers still engage in proprietary trading, of course, but they are trading proportionally less with the internalized order flow of their customers and more through dark ATSs and quoting venues. For our purposes this morning, the important point to recognize is that this shift has been from one type of a dark venue to another type of dark venue. The rise of dark ATSs has not, contrary to what might be expected, led to a decline in the success of the business model for quoting venues in general or exchanges in particular. In fact, the two largest ECNs have both publicly stated their intention to become registered exchanges.
Of course, the fact that quoting venues collectively have maintained their share of trading volume does not necessarily refute the concern that the equity markets may be becoming less transparent, particularly with respect to size discovery. The proliferation of dark ATSs could reflect a more general shift in trading from displayed liquidity to dark liquidity at all types of trading venues. Indeed, the extent to which the public transparent markets that display quotes also offer substantial undisplayed liquidity often is overlooked. Most exchanges and ECNs offer order types to their customers that allow either all or part of orders to be undisplayed. Nasdaq, for example, recently said that approximately 18-20 percent of its executions are against undisplayed liquidity, and our analysis indicates these numbers may be even higher in some cases. Nasdaq also offers a historical data product that shows how much undisplayed size was available for execution on its order book on a T+10 basis. Undisplayed size can be quite substantial and can even equal displayed liquidity at the inside quote, particularly in stocks outside the very top tier of trading volume where displayed liquidity may not be very large. For example, for reserve orders on Nasdaq, only about one-tenth of the actual order size is displayed to the market. All of this suggests that as public exchanges move to adapt and compete with the services offered by non-quoting venues, they will offer large pools of dark liquidity in the U.S. equity markets.
Associated with the concern about decreasing transparency is the concern that institutional investors are finding it increasingly difficult to trade in large size effectively. One statistic often cited to support this concern is that the average trade size in the U.S. equity markets is now less than 300 shares. Another such statistic is the decline in block trading on the NYSE. Over the last three years, block share volume at the NYSE Group fell by more than 50%, even as its total share volume increased by 20%.
Both price transparency and the ability of institutional investors to trade efficiently in large size are vital components of healthy securities markets. I saw a comment the other day that institutional trading in size has become both more challenging and more costly in recent years. I believe this comment is only half correct. Institutional trading in size is definitely more challenging and has led to the rise of many dark pools, particularly the anonymous block crossing systems. Somewhat paradoxically, however, institutional trading also has become more efficient and therefore less, not more, costly.
This improvement in institutional transaction costs is evidenced by two sets of widely available measures of market quality: first, the publicly available statistics on consolidated displayed spreads and consolidated displayed depth that are publicly available on the ArcaVision website; and second, publicly reported studies that directly measure institutional transaction costs. All of these statistics run counter to the commonly heard impression that the current equity market structure had led to less transparency and less efficient institutional trading.
For example, the statistics on consolidated spreads between the inside quotes show that the average spread in Russell 1000 stocks decreased by approximately 20% from January 2005 to June 2007. The statistics on consolidated displayed depth at the inside quotes show a similar story. For example, consolidated displayed depth at the inside quotes in Russell 1000 stocks increased by more than 50% from January 2005 to June 2007.
Well, you might ask, how did these improved statistics on consolidated displayed spreads and depth affect the bottom line? Did they translate into more efficient trading in large size by institutional investors? Two of the most well-known public sources of information on the transaction costs of institutional investors are provided by Elkins/McSherry and by ITG, which now provides statistics previously published by the Plexus Group.
Elkins/McSherry has reported that, for the year ending June 2007, total transaction costs for institutional investors in NYSE-listed stocks fell by 15% from the previous year. Similarly, ITG's study of institutional trading costs for the 2nd quarter of 2007 found that institutional transaction costs in U.S.-listed equities had declined by 35% since the first quarter of 2005. Reduced commissions, lower price impact of large orders, and more timely execution of large orders all contributed to this improvement in the efficiency of institutional trading.
The studies suggest that there is no necessary correlation between the average trade size in the public markets and the efficiency of trading for institutional investors. Rather, the use of algorithms and other sophisticated trading strategies that search out the most efficient venues for executing different types of orders has enabled large investors not merely to deal with highly active, automated markets, but to benefit from them. Some of these venues, of course, are dark ATSs, particularly those block crossing systems that have found creative ways for "natural" buyers and sellers in size to find each other, while still preventing information leakage prior to actually executing the large trade. In this respect, technology has found more efficient ways to deal with the perennial trading dilemma of finding a way to trade in large size without tipping the market.
Despite the positive results thus far, current trends toward the use of dark liquidity do raise regulatory concerns that bear watching. I will briefly mention a few of these relating to transparency, fragmentation, fair access, and best execution.
First, with respect to transparency, the Commission long has emphasized the importance of displayed liquidity in promoting efficient equity markets. Displayed liquidity provides the price discovery that is the starting point for all types of trading, both on and off the quoting markets. Since the initiation of decimal trading in 2001, however, there have been a lot of reasons for market participants not to display their trading interest. One problem in particular is the ease with which other market participants can step ahead of displayed trading interest by an economically small amount. Over the years, the Commission repeatedly has acted to encourage the display of trading interest. Most recently, of course, it adopted Regulation NMS in 2005 to establish market-wide protection of displayed quotes against trade-throughs at inferior prices.
A common misunderstanding of the Commission's approach to price transparency is that it requires the display of liquidity and thereby drives trading off the public markets. This is not true. Attempting to force market participants to display their trading interest when they do not wish to do so would be both fruitless and counter-productive. Instead, Reg NMS gives market participants a choice of whether and how much trading interest they want to display. Its protection against trade-throughs is designed merely to provide an additional positive incentive for display.
A more accurate characterization of the Commission's approach to price transparency is that it seeks to encourage the display of trading interest on a fair and non-discriminatory basis. A good example of this approach is the Order Handling Rules of the mid-1990s. They addressed the practice of ECNs displaying to professionals quotes that were denied to retail investors. The Order Handling Rules strictly limited this two-tiered market structure. In particular, exchanges, exchange specialists, and OTC market makers are required to make their best bids and offers available to the public in the consolidated quote feeds. In addition, ATSs that display quotes to more than one person are required to make their quotes available to the public for each stock in which they exceed a 5% trading volume threshold.
This quoting requirement raises a number of potential issues for ATSs. First, the term "display" for purposes of Regulation ATS is not limited to visual display to a human being, but also includes display within other electronic systems. ATSs that provide this type of electronic display to more than one system are not truly dark and therefore are subject to the quoting requirement in Regulation ATS if they exceed the 5% threshold. Similarly, Regulation ATS does not permit ATSs to circumvent the display requirement by disseminating information on the availability of liquidity that effectively functions as a quote. Exchanges, as well, are subject to similar restrictions on the dissemination of information with respect to undisplayed liquidity.
Another important issue to monitor with respect to dark pools is fragmentation. The currently large number of separately organized dark pools poses challenges for market participants. These include the basic logistical task and cost of establishing connectivity to many different venues. In addition, multiple dark pools pose fragmentation problems more serious than multiple quoting venues precisely because of their darkness. The only way to know whether a dark pool has liquidity is to route an order to the pool. Routing this type of pinging order is a less efficient means to assess liquidity than viewing a consolidated montage of displayed quotes from all quoting venues.
Competitive forces, however, seem particularly apt to address the problem of fragmented dark pools. The ultimate users of dark pools — investors and traders — seem likely to pressure operators of the pools, particularly the less successful ones, either to consolidate with other pools or to cooperate with dark pool aggregators. These aggregators offer services that enable investors to check liquidity more efficiently at multiple dark pools. A key cost of fragmentation for traders is the opportunity cost of being out of the market on one venue when your search for a contraside on other venues. With latency dropping rapidly, such fragmentation costs are falling as well.
Another important issue to monitor with respect to dark pools is fair access. Section 6 of the Exchange Act requires an exchange to provide fair access to its services. Regulation ATS, as well, requires an ATS to meet fair access requirements with respect to any particular stock in which it exceeds a 5% trading volume threshold. Notably, the Commission reduced this threshold from 20% to 5% when it adopted Regulation NMS in 2005. As undisplayed liquidity becomes an increasingly useful tool for investors, the importance of assuring that this liquidity is available to the public on terms that are not unfairly discriminatory increases as well. Similar to the transparency problems that led to the Order Handling Rules, unfair access to significant pools of dark liquidity would raise serious concerns about two-tiered markets that disadvantage particular classes of market participants.
A final concern that I'll mention with respect to dark pools is best execution. One of the challenges presented by an evolving market structure that is continually generating new venues and new services is that best execution polices and procedures must evolve as well. Blindly following routing strategies that may have worked in the past but no longer fit the current market structure is not an appropriate best execution approach.
For example, when it adopted the Order Handling Rules in 1995, the Commission explicitly addressed the practice of simply routing to displayed liquidity and emphasized that it would not necessarily satisfy a broker-dealer's duty of best execution. Rather, brokers must undertake a regular and rigorous review of the execution quality that can be obtained at different trading venues, including the possibility that a venue may have liquidity available at an undisplayed price or additional undisplayed size beyond the displayed size of a quote. Effective routing, of course, can depend on a variety of factors that can change with the particular type of stock and type of order, among other things. It therefore is quite difficult to make any unqualified statements about how brokers should deal with dark pools when making routing decisions. Perhaps the one piece of advice I could offer, though, is for brokers to make sure that they have an informed basis for their routing decisions that is up-to-date with the current market structure, and not merely an unexamined theory or assumption about what may have worked in the past. The U.S. market structure is changing rapidly. Brokers should take the necessary steps to assure that their routing practices have kept pace.