ZICKLIN SCHOOL OF BUSINESS
Department of Economics and Finance
17 Lexington Avenue, Box-E-0621
New York, NY 10010
Daniel G. Weaver
Associate Professor of Finance
Voice: (212) 802-6363 Fax: (212) 802-6353
April 27, 2000
Jonathan G. Katz
Securities and Exchange Commission
450 Fifth Street N.W.
Washington, D.C. 20549-0609
Re: File No. SR-NYSE-99-48;
Commission Request for Comment on Issues Relating to Market Fragmentation
Dear Mr. Katz:
I am an active teacher and researcher in the area of market microstructure. I have done research on many of the topics covered in the Commission's concept release. Therefore, I feel I am qualified to opine on the issue of market fragmentation.
First let me state unequivocally that penny ticks will not eliminate payment for order flow or internalization in a fragmented market. Thinking that penny ticks will eliminate it assumes that all market participants have similar costs. Perhaps an example will help. Assume that the true equilibrium spread for a security is $.11. This spread compensates market makers for the cost of processing orders (including a fair profit), for the cost of carrying inventory, and for the risk that they will be trading with someone with superior information (and hence market makers may incur a loss). Any spread greater than $.11 for this security will result in the market maker earning an excess profit (excess economic rents). Spreads can be wider than $.11 if the spread is not a multiple of the tick for the security. In today's environment, with a general 1/16 tick, spreads can be $.0625 or $.125. No one will make a market if spreads are $.0625, so spreads on this security are set at $.125 and market makers earn excess profits of $.015 per roundtrip transaction. Internalizing allows firms to earn an extra $.015. If a market maker paid $.005 per share for order flow, she would still earn a $0.005 excess profit. One might be then led to believe that enacting a penny tick would eliminate the excess profits and hence make internalization and payment for order flow unprofitable. This line of thinking is wrong.
The above example assumes that all orders are internalized or paid for. If a market maker can restrict the type of orders they internalize then they can still earn excess profits even if spreads equal $.11. Simply put, if a market maker can reduce the probability that they will receive informed trades then they can avoid the spread cost associated with trading with informed traders. Then that market maker's real spread is less than $.11 and they can earn excess profits even if the NBBO spread is $.11. As long as market participants can fragment the market by sending small uniformed orders to one market participant and large informed orders to another. The empirical evidence supports this notion. Porter and Weaver (1997) find that a reduction in tick size on the Toronto Stock Exchange had no impact on the rate with which member firms internalized trades.1 Under the current level of market fragmentation in the U.S., a smaller tick will have very little impact on the market practices of payment for order flow and internalization.
I next respond to the questions listed in the SEC concept release. For readability, I precede each answer with the original question in bold-type.
a. Fragmentation in General
To what extent is fragmentation of the buying and selling interest in individual securities among multiple market centers a problem in today's markets? For example, has fragmentation isolated orders, hampering quote competition, reducing liquidity, or increasing short-term volatility? Has fragmentation reduced the capacity of the markets to weather a major market break in a fair and orderly fashion?
Fragmentation reduces liquidity and hampers price competition. Retail providers of liquidity are reluctant to submit orders if they stand little chance of execution. The issue is a lack of system-wide priority rules. Retail liquidity providers are more willing to submit limit orders (provide liquidity) if others cannot step in front of them. Nasdaq has some priority rules (agency orders before principal), but they only apply to an individual dealer. There are no rules across dealers.
Think about queuing up to buy tickets to a concert. If there is a single line and people line up according to when they arrive, then people will try to arrive early. If instead, new arrivals have the choice of either going to the back of the current line, or forming a new line with an equal probability of getting tickets first, they will invariably choose the latter option. It doesn't take much imagination to see that after a short period of time, ticket seekers will stop arriving early. The current lack of priority rules across market centers (and within market centers in the case of Nasdaq) similarly discourages retail investors from providing liquidity to the market. If investors withhold liquidity from the market then there is an increased chance that an incoming order with soak up all available liquidity at a price and then move to the next price level. This, in turn, increases volatility in the market.
Although instituting system-wide price/time priority rules would stimulate public limit order submission, the down side is that a strict implementation of such a system does not allow for price improvement. Recent NYSE statistics indicate that 38% of orders arriving through Super DOT receive price improvement relative to the NBBO.2 Market orders and marketable limit orders are attracted by the potential for price improvement. On the NYSE, 55% of orders execute at the NBBO, therefore public limit orders also have a chance to interact with this order flow. Investors will send limit orders to venues if they have a chance of executing. Price/time priority rules is one way to attract public limit orders, but so is a large number of market orders. Giving investors information that may attract market orders, will also attract limit orders. It also allows market centers to compete for order flow from empowered investors, rather than requiring an onerous system be imposed on market systems. As I will argue later, creating informed investors, couple with a suggested change in internalization and payment for order flow rules, can have the same impact as system wide priority rules.
Is fragmentation in the listed equity markets likely to increase with the elimination of off-board trading restrictions, such as NYSE Rule 390?
Yes. There is no reason to expect that non-19c3 stocks will not follow the pattern of 19c3 stocks. I expect, in very short order, for a good portion of smaller orders in the affected stocks to be routed to Nasdaq.
In the existing over-the-counter market, what are the incentives for investors and dealers to quote aggressively?
The best feature of the Order Handling Rules was the inclusion of anonymous ECN quotes in the NBBO. Simaan, Weaver, and Whitcomb (2000) illustrate that Nasdaq dealers are reluctant to improve on quotes on the Nasdaq quote montage since doing so requires that they identify themselves. 3This leaves them open to retaliation from other market makers who may want quotes to remain wide. Allowing market makers to narrow quotes by placing orders on ECNs that are then attributed to the ECN (and not the market maker) has removed any fear of retribution.
If fragmentation is a problem, are competitive forces, combined with the existing components of market structure that help address fragmentation (price transparency, intermarket linkages to displayed prices, and a broker's duty of best execution), adequate to address the problem?
No because brokers usually decide where orders are routed and the rules governing a broker's best execution duty generally allow orders to be executed at the NBBO without adequately taking into account the probability of price improvement. Generally investors are told/shown the NBBO and are given no indication as to where that quote came from. Investors are also not provided with information on the probability of receiving price improvement, or the implicit cost of speed of execution. Provided with this additional information (and given the ability to route orders to the market venue of their choice) investors could make intelligent choices. This would in turn unleash the competitive forces in the market.
Will the greater potential provided by advancing technology for the development of broker order-by-order routing systems, or for informed investors to route their own orders to specific market centers, address fragmentation problems without the need for Commission action?
No because informed investors do not have sufficient information to make a decision as to where an order should be routed.
b. Internalization and Payment for Order Flow
What proportion of order flow currently is subject to internalization and payment for order flow arrangements in the listed equity and Nasdaq equity markets? Will the proportion increase in the listed equity markets as a result of the elimination of off-board trading restrictions?
Yes, it is an obvious point.
Is it possible for a non-dominant market center to compete successfully for order flow by price competition, without using internalization and payment for order flow arrangements? If not, is the inability to obtain access to order flow through price competition a substantial reason for the existence of internalization and payment for order flow arrangements?
No. The essence of internalization and payment for order flow is that dealers can pick off the small profitable orders and send the unprofitable larger orders to another market center. Recall that if a dealer can increase the probability that she is not trading with an informed trader, then her adverse selection costs are smaller than a market center that does not discriminate. Therefore, the discriminating dealer will always earn more than the non-discriminating dealer. Structuring of order flow amounts to redlining larger trades. We don't allow redlining in lending - why do we allow it in securities markets.
A good analogy is the common comparison of public versus private education. Public schools are often criticized for costing more per pupil than private schools. Public schools are open to everyone, including high-cost developmentally challenged students. In contrast private schools tend to attract only the low-cost non-challenged students, since they offer little special education. Imagine that we enacted a voucher system and that private schools received the same payment per pupil that public schools spend. Since the public school per pupil cost is the average of high-cost educationally challenged students and non-challenged students - private schools would earn a very high profit - at taxpayer expense.
A similar situation exists in our equity markets. As long as one market center exists that is required to accept all non-conditional orders while others can just pick the most profitable orders, true price competition cannot exist. The easiest solution is to allow for internalization and payment for order flow, but do not allow participants to discriminate among orders. Assume a situation where there is a market center that accepts all orders and multiple internalizing firms. If the internalizing firms where required to either internalize all orders or none, they would do so if their cost of processing orders was lower than the market center's. This will lead the market center to become more efficient to lower its costs. Equilibrium would be reached when processing costs would be minimized and firms would then be indifferent between internalizing and sending orders to the market center. Payment for order flow could not exist and investors would be paying the smallest possible transaction costs.
To what extent can brokers compete as effectively for retail business based on execution quality (or implicit transaction costs), as opposed to commissions (or explicit transaction costs) and other services?
While brokers can potentially compete on implicit (as opposed to explicit) transaction costs there is little investor awareness of the dimensions of implicit costs. For example, while executing an order quickly in a fast moving market can improve the trade price, there are no measures of what this implicit cost is for retail orders. Until studies are done that quantify the price of execution time for retail orders, this will remain a nebulous concept. Also, although brokers are required to include the possibility of price improvement into their order routing decision, there are no easily applied standards. In other words how are brokers to incorporate the probability of price improvement? Perhaps a system should be put in place whereby market centers, and dealers are required to make available to the public, the percentage of orders (by size) that received price improvement as a result of being routed to them. Investors would take this into account when deciding where their orders are to be routed. I firmly believe that investors not brokers should have control of orders, including the decision to where to route the order. The Ontario Securities Commission has adopted this stand, and I urge the U.S. Commission to do the same.
Do investor market orders that are routed pursuant to internalization and payment for order flow arrangements receive as favorable executions as orders not subject to such arrangements? Even if these orders subject to internalization and payment for order flow arrangements receive comparable executions, does the existence of such arrangements reduce the efficiency of the market as a whole (by, for example, hampering price competition) so that all market orders receive less favorable executions than they otherwise would if there were no internalization or payment for order flow?
Yes - see my public versus private education example above.
Even if internalization and payment for order flow arrangements increase the fragmentation of the markets, are any negative effects of increased fragmentation outweighed by benefits provided to investors, such as speed, certainty, and cost of execution?
It is possible that internalization benefits retail investors by executing trades quicker in a fast moving market. However, that assumes that brokers internalize quicker than an order can be sent to another market for execution. There is no economic reason for brokers to do this. What incentive would an internalizing broker have to buy stock quickly as prices are dropping? If internalizing brokers could demonstrate that retail orders receive better executions than those, of similar size, routed to other venues for execution, then I would say there is a benefit to internalization. However, I do not believe such evidence exists.
c. Best Execution of Investor Limit Orders
Does increased fragmentation of trading interest reduce the opportunity for best execution of investor limit orders? Are brokers able to make effective judgments concerning where to route limit orders so as to obtain the highest probability of an execution?
For this question, I assume the Commission is referring to non-marketable limit orders (marketable limit orders are those that are effectively market orders). Fragmentation does hurt limit orders, since it reduces the flow or orders interacting with any one limit order. Fragmentation is forcing limit order traders to have yard sales as opposed to gathering together at flea markets. The point is that the amount of limit order trading is directly related to the amount of total order flow. If all non qualified limit orders could gather in the same spot as all the market orders then we have a flea-market effect - vendors gather where the customers are. Allowing brokers to decide where limit orders are routed results in a yard-sale effect - a vendor only has access to the traffic on their road. Customers are also disadvantaged in a yard sale since there is no price competition (as compared to a flea market).
Does the opportunity for brokers to share in market maker profits through internalization or payment for order flow arrangements create an economic incentive to divide the flow of investor limit orders from investor market orders among different market centers? If so, does this adversely affect the opportunity for investor limit orders to be executed fairly and efficiently?
I am not sure what this question is asking. However, I have addressed the economic incentives earlier.
Is it consistent with national market system objectives (such as efficiency, best execution of investor orders, and an opportunity for investor orders to meet without the participation of a dealer) for market makers to trade ahead of previously displayed investor limit orders held by another market center (that is, trade as principal at the same price as the limit order price)? Does this practice significantly reduce the likelihood of an execution for limit orders by reducing their opportunity to interact with the flow of orders on the other side of the market? Does the practice offer any benefits that outweigh whatever adverse effects it might have on limit order investors?
I think the important point here is that limit order traders are far less likely to trade if there is a small chance of their order being executed. Increasing the probability of limit orders executing will increase the amount of limit order trading. This will provide protection against volatility in markets. Since the Actual Size Rule has led many Nasdaq dealers to quote for 100 share sizes as opposed to the previous 1,000 minimum share size, the need for public limit order participation is greatly increased. I am afraid that without increased incentives for public limit order placement, a downturn like we had in 1987 could be far worse.
Having answered the questions posed in the concept release, I would like to return to the issue of who should control an order. I firmly believe that the investor placing the order should have control over it. If that person decides they don't want to expose the entire order to the public - they should be allowed to hide part of it and still remain in a queue. If the investor wants to execute all of an order at the same price, or none of it (called an all-or-none order qualification) - they should be allowed to. No one can force investors to reveal what they do not want to. In a previous concept release the Commission raised the idea that no part of an order should be hidden from public view. I think it is far better for an investor's order to be allowed to remain in a queue with some liquidity not revealed than to have an investor withhold from the queue that portion they want hidden. The latter will exacerbate market movements, since there will be a reduction in the liquidity necessary to absorb incoming order flow.
The fact that some investors wish to hide part of their order from other market participants as well as place all-or-none orders, suggests that we should not consolidate markets to the point where all orders are traded in a single venue (either physically the same venue or virtually the same venue through system wide rules). I was careful in my answers above to only refer to non-qualified orders. If an investor is willing to accept any part of their desired size at a price, then cost savings can accrue if those orders were consolidated. However, the desire of institutions to trade in large blocks with minimum price impact suggests that there will always be a need for separate trading facilities for institutions. So I am not an advocate for completely consolidated order flow. I believe that orders without qualifications could benefit from system-wide priority rules. However, the same effect can be achieved if the following occurs:
I urge the Commission to remember that one-size-fits-all will not work in the increasingly global securities industry. Trying to force investors to all trade in the same system will drive traders off shore. Bermuda could quickly become a haven for trading outside of the SEC's reach.
Daniel G. Weaver, Ph.D.
1 See "Tick Size and Market Quality", Financial Management, vol. 26, no. 4, 1997, pp. 5-26.
2 See the April, 2000 issue of The Exchange, page 3.
3 See "The Quotation Behavior Of ECNs And Nasdaq Market Makers." Working paper. Baruch College. March 22, 2000.