SEC ROUNDTABLE ON MARKET STRUCTURE
WASHINGTON DC, OCTOBER 29, 2002
TESTIMONY OF DAVID K. WHITCOMB
Founder & Chairman, Automated Trading Desk, LLC and
Professor Emeritus of Finance, Rutgers University
Background & Point of View
I recently retired from Rutgers University, where I spent most of my academic career and was one of the earliest researchers in the then-new field of "market microstructure". Much of my research centered on the role of limit orders, the provision of marketmaking services, and the determinants of the bid-asked spread1. Around the time of the 1987 Crash, I had the radical idea that computers could be taught to think like a human trader, in order to price, enter and "work" limit orders for institutions. When Wall Street's fear of computerized trading and my own lack of practical experience made me unsuccessful at peddling the idea as a consulting project, I started Automated Trading Desk (ATD).
ATD is a financial services technology firm founded in 1988 and focused almost exclusively on developing and operating expert systems for fully automated limit order trading. ATD's two active brokerage subsidiaries2, ATD Brokerage Services and Chicago Securities Group, trading for its own-account and for institutional and brokerage firm customers, together account for about 4% of Nasdaq volume and 2% of listed volume, roughly 80 million shares/day in nearly 300,000 individual transactions. Nearly all of our Nasdaq and some of our listed trading is done on ECNs, and we account for perhaps 10% of all ECN volume.
Successful Regulatory Reform: Order Handling Rules & "Tick" Size Reduction
It may be instructive to begin with the two most important, and in my view, most successful market structure reforms in a generation: the Order Handling Rules and the reduction in the minimum "tick" size (culminating in decimalization3). The Order Handling Rules were widely derided by academics and others as "micromanagement" by the SEC, but they had the desirable effect of bringing competition and transparency to the Nasdaq market and fueling the growth of purely electronic trading systems. Decimalization has decimated the profits of marketmakers (excepting NYSE Specialists), and is forcing institutional investors to change their trading strategies4. However, decimalization is greatly reducing the trading costs faced by individual investors, and I believe that in the long run it will bring desirable change to the entire securities market.
The common thread in both these structural reforms is the liberating effect of competition: Competition for traditional SRO markets from private ECNs not beholden to special interests and not locked into perceived ways of doing things, and competition for marketmakers and Specialists from dynamic limit orders. Spreads have collapsed, in some cases to below the operating costs of traditional marketmakers, clearing costs have fallen dramatically due to the enormous efficiencies of locked-in electronic trading, and volumes (measured in shares, trades, and "message traffic") have soared.
Regulation of Electronic Trading
One of the most desirable aspects of the explosion of electronic trading on private ECNs is that many of the issues on which we might need regulation in the traditional SRO markets will largely settle themselves with ECNs. Some examples:
(1) ECNs have voluntarily chosen to make their "deep book" available to investors (via direct feed or the Internet) because the ECN owners gain no advantage from suppressing this information, because disseminating it is easy and cheap, and because it bestows a marketing advantage. By contrast, consider how long it took the NYSE to decide to display its five best prices and how grudgingly they are displayed (with "refreshments" only every 10 seconds)5.
(2) In a market like the NYSE where limit orders are not automatically and instantaneously displayed (and many are never displayed at all), the NBBO is indeed meaningless as information and as a standard for "best execution", but in electronic markets whose best quotes are sorted to form the NBBO, the NBBO has real meaning as the standard for best execution of small orders6.
(3) "Fragmentation" arising from purely electronic venues that are not linked by consolidated quotes or a CLOB becomes nearly a non-issue due to the ease of arbitrage. Suppose the bid on one non-linked electronic market exceeds the offer on another by one cent. Profit-seeking traders who have invested in bilateral links to both markets will instantly sell to the one and buy from the other, and low systems and clearing costs will allow arbitrage profits until market center differences are driven down to mill levels. The ready availability of "smart routing" technology makes such arbitrage profits widely available and drives the "crossing time" down to milliseconds. In other words, network technology is the Twenty-First Century alternative to forced consolidation7. We don't need a CLOB. I believe the marketplace, not regulation, will drive the wide availability of inter-market linkages, both between brokerage firms and multiple market centers and between market centers.
Regulatory & SRO Impediments to Limit Order Trading
Competition among ECNs and between ECNs and the Nasdaq dealer market has revolutionized the Nasdaq marketplace. It is impossible to overemphasize the role of limit orders in this revolution. It was the required display in the NBBO of dealers' "layoff" limit orders which closed the window on the mechanism by which dealers maintained artificially wide spreads prior to 1997. It was then the growth of limit order trading by proprietary traders, institutions and individuals, combined with smaller tick sizes, that drove typical spreads for the most liquid Nasdaq stocks down to one cent8 and gave the ECN markets good (if somewhat "distributed") liquidity.
Given the central role of limit orders posted on electronic markets, I feel that any regulatory impediments to the use of limit orders would take the markets in a retrograde direction, both with respect to spreads and to liquidity.
(1) The fact that it makes business sense for the owners of ECNs to charge less for limit orders (which provide liquidity) than for marketable orders (which "take" liquidity) means that it also makes policy sense to allow them to. Whatever encourages limit orders encourages the competitive provision of liquidity and narrows spreads. It would be bad public policy to mandate "one size fits all" pricing of limit and market orders. It might even be a good reform to end the various forms of price discrimination against limit orders that exist at SRO exchanges (and that benefit their marketmakers).
(2) SRO rules that seek to reduce "quote flickering" by preventing immediate canceling of limit orders also impede their use and are anti-competitive. Traders whose limit orders "flicker" to the distraction of other traders will learn that their orders rarely trade except when there is adverse selection against them9.
(3) Charges for entering/canceling limit orders seriously hamper quote competition by discouraging their use. The only way a trader, seeking to compete with conventional marketmakers by using dynamic limit orders, can protect himself against "adverse selection" is by trying to detect adverse market movements and canceling orders. Exchange and Nasdaq charges for entering/canceling limit orders should not be allowed to exceed the incremental cost of expanding "bandwidth" to accommodate the extra message traffic. When an SRO dominated by marketmakers or Specialists assesses the charges, the SEC must be vigilant to ensure that order entry/cancellation fees are not being used to hamper competition. It is significant that private ECNs price limit order entry/cancellation as if the incremental cost of adding bandwidth were zero.
(4) SRO rules that permit Specialists to compete against limit orders by delaying (or even declining) the display of limit orders are a serious impediment to competition and the evolution of a more electronic listed market structure. It is no accident that NYSE Specialists are the only marketmakers who have earned good profits following decimalization. Specialists have a unique ability to "penny" limit order traders and not be "pennied back" due in part to this incomplete display. This limits price improvement, hurting both market order traders and limit order traders10.
Listed Market Reform Proposals
I believe it is unfortunate that the Order Handling Rules, which, together with decimalization, have driven such remarkable changes in the Nasdaq market, have not really been implemented fully in the listed market. As my scattered comments above imply, dynamically and aggressively-priced limit orders are unable to compete on a level playing field with Specialist quotes11.
There are four specific reforms which I believe would allow competition from limit orders to narrow spreads and increase listed market liquidity:
(1) Require Exchanges to display limit orders instantaneously upon receipt. Like a market-maker's quote, limit orders are an advertisement to trade. If they are not displayed promptly, they lose their advertising value. When they are displayed, they attract market orders to the trading venue. As we have seen in the post-Order Handling Rules experience in the Nasdaq market, when the makers of limit orders can see their quote-improving orders displayed immediately upon submission they gain confidence that their orders will be handled properly and submit more. Thus, the instantaneous display of limit orders will cause more limit and market orders to be placed, enhancing competition.
(2) Require Exchanges to permit instant cancellation of limit orders. It is hard to over-emphasize the "adverse selection" cost to limit order traders of delayed order cancellation. The largest trading cost faced by limit order traders is "adverse selection" - the fact that the probability that your order will execute is greater if the market is about to move against you. Limit order traders mitigate this cost by monitoring the market carefully and canceling their orders when they detect evidence of an impending adverse move. When they cannot predict short-term market direction well or when they are unable to cancel quickly, they adjust by pricing less aggressively. This impedes limit order use and results in higher quoted spreads (and less liquidity).
(3) Require Specialists to display their quotes prior to transacting and to fill electronically-delivered market orders instantaneously at the best posted price. Specialists are able to "jump ahead" of ("penny") limit orders by filling market orders at an "improved" price. While this appears on the surface to be good for market order traders, it is not, because they would get even better prices if limit order competition were not inhibited. The rationale offered for delaying execution of market orders (and marketable limit orders12) is that the Specialist may find "price improvement" by exposing the order to the "floor". However, whether "price improvement" comes from floor traders jumping ahead or the Specialist himself jumping ahead, the result is to chill competition from limit orders. Limit orders are unable to compete on a level playing field. If two floor traders wish to trade with each other, or if one floor trader wishes to trade with the Specialist, fine. But market orders coming in via electronic means ought to trade against the displayed book. This will require Specialists to post their interest as firm quotes and floor traders to post their interest as limit orders if they wish to capture market orders coming in via electronic means (which are usually the most desirable "dumb" orderflow). The resulting "level playing field" competition among Specialists, floor traders and limit order traders will increase displayed liquidity and reduce displayed spreads, to the benefit of small public investors using market orders13.
(4) Make ITS instantaneous. The Intermarket Trading System (ITS) is notoriously slow. This discourages its use and makes Regional Exchanges and ATSs that link via a Regional Exchange (as ARCA is trying to do) unable to compete with the NYSE as genuine alternative pools of liquidity14. If there is any rationale for this situation in 2002, it is that orders coming into a market center electronically have to interact with a manual trading system in which the Specialist finds liquidity for incoming orders (whether from ITS or from the Exchange's own system) via an "auction". Since the "auction" is frequently more myth than reality (especially on Regional Exchanges - often all liquidity is either on the Specialist's book or "in his pocket"), it seems counterproductive to force a "lowest common denominator" approach. With Reform #(3) above, ITS could easily be made instantaneous, truly linking market centers.
The Role of SROs
I have argued for years that the SROs are an unfortunate and anti-competitive holdover from early New Deal policy in which industry groups were encouraged to manage prices and production. While the NRA and the other elements of this policy were quickly dismantled as a failed experiment, securities dealers and Specialists have continued to be allowed to make and enforce the rules for trading. There is a natural conflict between their profit interest and the public interest, that often results in problems such as those that plagued Nasdaq in the mid-nineties. Competition among profit-motivated market centers (such as that introduced by the Order Handling Rules, which spurred the growth of private ECNs) is one element of the solution.
Demutualization is the other. If the NYSE and Nasdaq become private firms with a diverse set of shareholders (i.e., not majority owned by the current member firms), their incentive will be to create profitable trading systems, rather than to restrain trading in ways that benefit certain constituencies15. The profit motive will also foster innovation. The clearest evidence for this is the degree of innovation among and the level playing field offered by ECNs.
Demutualization provides an opportunity for the reform of regulatory rule-making and enforcement. Rule-making and enforcement should really be a government role, one which can be performed best by the SEC. This would end the practice of "regulatory arbitrage" and it would eliminate conflicts of interest in both rulemaking and enforcement. The savings from eliminating overlapping and layered regulation could more than pay for the additional SEC staff that would be needed, many of whom could come from among the many honest and hard-working people who work in SRO enforcement positions today.
|1|| See, for example Chapter 5 of Cohen, Maier, Schwartz & Whitcomb, The Microstructure of Securities Markets, Prentice-Hall, 1986; and Schwartz & Whitcomb, Transaction Costs and Institutional Investor Trading Strategies, Salomon Brothers Center Monograph Series, 1988.
|2|| For convenience in using my firm's experience to illustrate comments, "ATD" will refer to either the parent R&D company or to the B/D subsidiaries, as appropriate in context.
|3|| Some of you are aware that, following the lead of Chairman Oxley, an activist SEC Commissioner, and a few tireless academics, I was a strong proponent of decimalization. You may find it amusing that the day penny increments became universal, ATD's proprietary trading profits were reduced by more than 50%.
|4|| It is commonly asserted that penny increments have reduced "liquidity". This is not strictly accurate. ATD trades substantial blocks for its institutional customers, and we find that by posting small, dynamically-priced limit orders on ECN or exchange "books" in sequence over minutes or hours, we can find quite adequate liquidity while leaving tiny "footprints".
|5|| When you look at the NYSE deep book you often see aggressively-priced, good-sized limit orders that never are displayed in the NYSE quote due to the 30-second delay window and the display discretion Specialists are allowed. I believe this makes it clear why the NYSE was so grudging in its deep book display.
|6|| Retail orders of several thousand shares cannot always be expected to be executed at the NBBO in a market with penny increments and "distributed" liquidity. The DOE statistics which the SEC now collects from retail brokers provide more appropriate performance standards.
|7|| This assertion is not true when limit orders/quotes are not instantaneously displayed or are not accessible instantaneously by all players. Hence it applies to electronic markets but not to traditional SRO auction or dealer markets.
|8|| I have to admit to being caught off-guard by the collapse of the modal spread for liquid stocks down to one cent (vs. the 2-3 cents I expected). This was fueled by dramatically diminishing clearing and system costs and by the economic fact (which I should have known from my own research) that traders who use limit orders simply to effect position changes (rather than to "make markets" electronically and profit from the spread) gain from substituting limit orders for market orders so long as the probability of order execution is high, even if the spread is just a penny.
|9|| Clearly, rapid entry and deletion of quotes cannot legally be used to "paint the tape". Fortunately, ECN and Exchange/Nasdaq electronic trading systems provide an audit trail that allows electronic surveillance and existing rules to detect and sanction violators. Also, the fact that limit orders are "firm" so long as they exist (as opposed to dealer quotes in a market without instantaneous automated execution) means that the order maker must consider the probability of execution in his order placement strategy and is limited in his ability to "game".
|10|| Credit is due to the SEC's Chief Economist for being the first to point this out (in his role as a major contributor to the academic microstructure literature). The Specialist has the option to "step ahead" of a limit order at the cost of just one cent/share. While this "price improvement" is good for the market order customer, competitive price improvement is limited by the fact that the limit order trader is unable to reply by improving price further.
|11|| When my colleagues in ATD's Specialist Unit read these comments, they will be almost as happy with me as my colleagues in our proprietary trading unit were following decimalization!
|12|| Delay of marketable limit orders is an especially vexing problem on the AMEX. ATD's traders and trading models find that their marketable limit orders rarely execute against the displayed quote (or receive "price improvement"). What makes trading electronically truly awful on the AMEX is that we almost never get a trade confirm back quickly. Our order disappears into a "black hole", and, more often than not, we get no response and simply observe the quote moving after our order is sent. When we try to cancel, we have to wait again for a "UROUT". This makes us wonder if the AMEX Specialist is observing the Firm Quote Rule, and it makes us reluctant to send orders to the AMEX. A for-profit Exchange whose objective is to make profits on trading volume would not have practices like the AMEX.
|13|| When the NYSE Direct+ system was introduced, I had high hopes that it would give market order traders a fair chance. However, ATD's traders observe that their Direct+ orders execute against the displayed quote even when a limit order at a better price is displayed on the NYSE deep book. NYSE Direct+ promised no price improvement, and it delivers it (even when price improvement requires no "auction"). This is hardly an encouragement to market order traders to use the system.
|14|| As everyone knows, the Regional Exchanges are today really only dealer markets, competing via directed order flow. They do not offer significant competing pools of liquidity to the general investing public or participate in meaningful price discovery. Reform of ITS could make Regional Exchanges a meaningful source of listed-market competition and give them a legitimate raison d'être.
|15|| Clearer evidence of the need for demutualization cannot be found than the statement Friday by Nasdaq Chairman and CEO Hardwick Simmons seeking to roll back decimalization because of the current profit woes of marketmakers. [See Nasdaq Head: Move to Decimals 'Misguided' Friday October 25, 5:49 pm ET (Reuters).] As Chairman of a firm whose profits remain substantially lower than they were just prior to decimalization, I wish Chairman Hardwick success, but as a person who looks for future growth and profits through reforms that benefit the public, I do not.