Report on the Practice of Preferencing

Table of Contents

Introduction

I. Background

A. History of Preferencing/Competing Specialist Programs on National Securities Exchanges

1. Competing Specialists on Exchanges: From the Late 1800s to 1991

a. Late 1800s to the 1975 Amendments to the Securities Exchange Act of 1934

b. Pacific Stock Exchange Competing Specialists (1976 - 1977)

c. New York Stock Exchange Competing Specialists (1976 - 1994)

B. The Cincinnati Stock Exchange's Dealer Preferencing Program (1991 - Present)

1. Background

2. Non-Preferenced Executions on the CSE

3. Description of the CSE's Preferencing Program

4. The Commission's Permanent Approval of the CSE's Preferencing Program in 1996

C. The Boston Stock Exchange's Competing Specialist Initiative (1994 - Present)

1. Background

2. Operation of the BSE's Competing Specialist Initiative

3. The Commission's Permanent Approval of BSE's Competing Specialist Initiative in 1996

II. Preferencing and Internalization

A. Internalization on National Securities Exchanges

B. Bypassing of Time Priority on Exchanges

1. Rules of Priority, Parity, and Precedence

2. Use of Precedence Based on Size to Bypass Time Priority

3. Crossing Orders

4. "Stopping" Stock Rules

5. Use of Marketable Limit Orders to Provide Price Improvement in Minimum Variation Markets

6. Percentage Orders

7. BSE's Competing Specialist Initiative and CSE's Preferencing Program

C. Off-Exchange Internalization: Third Market Trading

D. Summary

III. Preferencing and Agency/Fiduciary Obligations

A. Background - 1934 to 1975 Amendments to the Act

B. 1975 Amendments to the Act

C. Preferencing and Best Execution

1. CSE and BSE Permanent Approval Orders

2. Discussion

IV. Review of Order Handling Practices of Preferencing Firms

A. General Industry Order Routing Practices

B. Order Handling Practices of CSE Preferencing Dealers and BSE Competing Specialists

1. CSE Preferencing Dealers

a. Background

b. Order Handling Practices

c. Risk Management Procedures of CSE Preferencing Dealers

2. Boston Stock Exchange Competing Specialists

a. Background

b. Order Handling Practices

C. Execution Costs for Customers

V. Empirical Analysis of Preferencing Activity

A. Introduction

B. Data Description

C. The Quality of Market Quotations

D. Market Order Execution Costs and Price Improvement Statistics

E. Limit Order Execution Analysis and Dealer Routing Decisions

F. Conclusions

Tables

VI. Findings and Conclusions of the Study

Appendices

Appendix A: CSE Preferencing Pilot Program (1991 - 1996)

A. Commission Approval of the Initial CSE Preferencing Pilot Program in 1991

B. Extensions of the CSE Preferencing Pilot Program

C. CSE Request for Permanent Approval of the Preferencing Pilot Program

Appendix B: BSE Competing Specialist Initiative Pilot Program (1994 - 1996)

A. Commission Approval of the Initial BSE CSI Pilot Program in 1994

B. Extensions of BSE's CSI Pilot Program

Appendix C: Development of the CSE's National Securities

Trading System (1977 - 1990)

Appendix D: Detailed Data Description

Introduction

Section 510(c) of the National Securities Markets Improvement Act of 1996 ("NSMIA"), [Pub. L. No. 104-290, 110 Stat. 3416 (1996).] requires the Securities and Exchange Commission ("SEC" or "Commission") to conduct a study on "the impact on investors and the national market system of the practice known as `preferencing' on one or more registered securities exchanges" and specifies that the study include consideration of how preferencing impacts the execution prices received by retail securities customers whose orders are preferenced; the ability of retail securities customers in all markets to obtain executions of their limit orders in preferenced securities; and the cost of preferencing to retail securities customers. Section 510(c)(2) of the NSMIA requires the Commission to submit a report to the Congress on the results of the study by April 11, 1997. This report is submitted in accordance with that requirement.

For purposes of the study, the term "preferencing" is defined in Section 510(c)(3) of the NSMIA as "the practice of a broker acting as a dealer on a national securities exchange, directing the orders of customers to buy or sell securities to itself for execution under rules that permit the broker to take priority in execution over same-priced orders or quotations entered prior in time." The Commission notes that in its broadest sense, the term preferencing refers to the direction of order flow by a broker-dealer to a specific market maker or specialist, independent of whether or not some form of affiliation or inducement for thedirection of order flow exists between the broker-dealer and the market maker or specialist. [Internalization, a subset of preferencing broadly defined, is the direction of order flow by a broker-dealer to an affiliated specialist or order flow executed by that broker- dealer as market maker. See Securities Exchange Act Release No. 37619A (September 6, 1996), 61 FR 48290 (September 12, 1996) at n.357 (File No. S7-30-95) (Order Execution Obligations Adopting Release) ("OEO Adopting Release"). ] Moreover, because specialists acting as dealers on national securities exchanges currently may execute customer orders without regard to orders or quotations entered on other exchanges prior in time, given that time priority is not maintained across markets in the national market system, any activity of a specialist who trades for his own account when there are preexisting same-priced orders or quotations on another exchange would be considered preferencing activity under the NSMIA definition of preferencing. Thus, it is important that Congress recognize that preferencing occurs on every exchange.

Nevertheless, the Commission has made the assumption that Congress intended the Commission to focus on preferencing practices that allow a dealer to take priority in execution over same-priced orders or quotations entered prior in time on the same exchange. As a result, much of the Commission's study examines both the Cincinnati Stock Exchange's ("CSE") dealer preferencing program and the Boston Stock Exchange's ("BSE") competing specialist initiative ("CSI"). [See Securities Exchange Act Release Nos. 37046 (March 29, 1996), 61 FR 15322 (April 5, 1996) (File No. SR-CSE-95-03) (order permanently approving CSE's preferencing program) ("CSE Approval Order"); 37045 (March 29, 1996), 61 FR 15318 (April 5, 1996) (File No. SR-BSE-95-02) (order permanently approving BSE's CSI) ("BSE Approval Order"). ]

Despite focusing much of the study on the CSE's and BSE's rules permitting preferencing, the Commission also examined many types of preferencing practices that have taken place on nearly all of the national securities exchanges since the adoption of the Securities Exchange Act of 1934 ("Act"), [15 U.S.C. 78a et seq.] particularly those that have had competing specialists on their trading floors at one time or another.

The report is organized as follows. Part I begins with a brief history of competing specialists on U.S. stock exchanges from the inception of the specialist function shortly after the end of the Civil War, through their longstanding presence on the floor of the New York Stock Exchange ("NYSE"), and up to and including the Commission's permanent approval of the CSE's preferencing program and the BSE's CSI. In addition, Part I contains a detailed explanation of the operation of both the CSE and BSE programs. Part II places the CSE and BSE programs in context by providing an overview of preferencing and internalization in the national market system as a whole and a description of various other practices on exchanges by which a broker may take priority in execution over pre-existing customer interest. Part III reviews a broker-dealer's obligation to achieve best execution of its customers' orders and examines how this obligation is influenced by the CSE and BSE programs. Part IV examines the order handling practices of firmsparticipating in the CSE and BSE programs, as well as the general order routing practices of both preferencing and non-preferencing firms. Part V consists of an economic analysis of CSE and BSE market quality and executions received by customers on these exchanges, and compares the results to those of other markets.

The Commission has concluded from the study that preferencing is but one method to internalize order flow, a practice that exists on all markets in various forms. Moreover, the Commission has found that, to date, preferencing has not had a deleterious effect on the national market system. To the contrary, preferencing has furthered the CSE's and BSE's ability to compete for order flow. Furthermore, the Commission has concluded that while preferencing raises significant agency-principal concerns, if the appropriate protections are in place, preferencing is not necessarily inconsistent with the best execution of customer orders. Accordingly, the Commission will continue to review the effects of the practice of preferencing on the national market system on an ongoing basis to ensure that this practice is consistent with the maintenance of fair and orderly markets, the protection of investors and the public interest, and the furthering of the national market system goals of Section 11A of the Act. [15 U.S.C. 78k-1.]

I. Background

A.History of Preferencing/Competing Specialist Programs on National Securities Exchanges

A specialist is an exchange member that "specializes" in the tradi ng of one or more stocks and has market making responsibilities for these stocks , including the maintenance of a continuous two-sided market in such stocks. [Un der the rules of the various exchanges, only a natural person is eligible to be registered as a specialist. On most exchanges, individual specialists have orga nized themselves into specialist units. Each unit is composed of one or more sp ecialist firms, which are legal entities ( e.g. , partnerships, corporations, or joint ventures) formed by specialists. Further, the rules of most exchanges provide that allocations of stock are made to specialist units, not individual specialists or specialist firms. See , e.g. , NYSE Rule 103B.] A specialist acts both as broker and dealer with regard to his or her specialty stocks. As broker, a specialist generally holds and executes market, limit, and stop orders forwarded to him or her by other exchange members. The specialist maintains a "book" of limit and stop orders that other brokers have entered with the specialist in his or her specialty stocks. As dealer, a specialist may trade for his or her own account against incoming buy and sell orders or orders on the book. Moreover, a specialist is also responsible for setting a fair opening price to clear accumulated market interest. [A specialist has the same obligation with regard to reopenings following intra-day trading halts.] A specialist's activities are circumscribed by the federal securities laws and exchange rules, which impose certainaffirmative and negative obligations on a specialist's dealings. [See , e.g. , Section 11(b) of the Act, 15 U.S.C. 78k(b), and Rule 11b-1 thereunder, 17 CFR 240.11b-1.; Amex Rule 170; and NYSE Rule 104A. For example, Rule 11b-1(a)(2)(ii), 17 CFR 240.11b-1(a)(2)(ii), requires that exchange rules provide that specialists have the affirmative obligation to engage in a course of dealings for their own account to assist in the maintenance, so far as practicable, of a fair and orderly market. Rule 11b-1(a)(2)(iii), 17 CFR 240.11b-1(a)(2)(iii) requires that exchange rules provide that specialists have the negative obligation to restrict their dealings, so far as practicable, to those reasonably necessary to permit the specialist to maintain a fair and orderly market. In addition, Section 11(b) of the Act limits a specialist's discretion by providing that a specialist may not effect as broker "any transaction except upon a market or limited price order."]

Currently, with the exception of the BSE, CSE, and the Pacific Stock Exchange ("PSE"), [Currently, the BSE trades 87 of its 2221 stocks on a competing specialist basis, while the CSE is a competing dealer market for all of its 568 stocks. Because the PSE maintains trading floors in Los Angeles and San Francisco, it utilizes a variant of the unitary specialist model in which two co-specialists coordinate a single market in a particular stock. The PSE has recently changed its name to the Pacific Exchange, and its acronym to PCX. However, to prevent any potential confusion, PSE has been used throughout.] U.S. stock exchanges are organized on the unitary specialist model. In this model, each stock traded on the exchange is allocated to a single specialist or specialist unit. Though this model may be the predominant one in use today, the specialist function developed in the context of a competing specialist model, in which there was more than one specialist making a market in a particular stock. Competing specialists have been present on one or more exchanges almost continuously from the development of the specialist system to the present. Further, in almost all instances where there have been competing specialists onexchanges, these specialists have engaged in the practice of preferencing, as that term is defined in Section 510(c)(3) of the NSMIA.

1.Competing Specialists on Exchanges: From the Late 1800s to 1991

a.Late 1800s to the 1975 Amendments to the Securities Exchange Act of 1934

In the years following the Civil War, the specialist function developed on U .S. stock exchanges in conjunction with the supplanting of the call market tradi ng format by the continuous two-sided auction trading format. [See Report of Sp ecial Study of Securities Markets, 88th Cong., 1st Sess., H.R. Doc. No. 95, pt.2 ("Special Study") at 61. In auction market trading, investors or brokers repre senting investors trade directly with each other as opposed to a dealer market i n which most trades are made through a dealer at the dealer's bid or ask price. In a call market, each stock is traded only at specific times during the day an d at one particular price, which is determined from the application of pricing r ules to the buying and selling interest accumulated up to the time of a call. Id. Conversely, in a continuous two-sided auction market, trades in a stock may take place at any time during an exchange's trading hours through competition among both buyers and sellers, which enter bids and offers for the opportunity to trade against contra- side interest as it arrives at the exchange. See also Division of Market Regulation, The October 1987 Market Break, at 4-1 n.1. An exchange's rules of priority, precedence, and parity determine which bids and offers are entitled to take part in a particular execution. See infra Part II.B.1 for a discussion on the various exchanges' rules of priority, precedence, and parity. ] The first specialists were primarily brokers, handling limit and stop orders for other exchange members, while such members moved among the various trading posts to those that were the most active at any particulartime. [See Special Study, supra note at 61. In this regard, The Twentieth Century Fund stated in its 1935 examination of the securities markets that: "[b]y 1910 most floor members, other than those representing commission firms, followed the practice of moving to any post which then seemed attractive. There were a considerable number of specialists in round lots and odd lots, who would do brokerage and dealing for their own account in other issues if time or opportunity afforded itself." The Twentieth Century Fund, The Securities Markets ("Twentieth Century Fund") (1935) at 405. ] The dealer function of specialists in their specialty stocks became more prominent after 1910, both as a result of other exchange members' need for a continuous market in stocks, particularly in those that were thinly traded, and through the increased opportunities that were available to specialists to make profits by trading for their own accounts due to the sizeable amount of order flow routed to them. [See Special Study, supra note , at 61-62 and 99-101. ]

Prior to the adoption of the Act, the NYSE had adopted a number of rules tha t circumscribed the conduct of trading by its specialists. First, the NYSE proh ibited a member, including specialists, from buying (selling) a stock for its ow n account to the disadvantage of, or before the execution of, an unexecuted orde r to buy (sell) that the member held on behalf of a customer. [See Stock Excha nge Practices, Hearings on S.Res. 84 Before the Comm. on Banking and Currency, 7 3d Cong., 1st Sess., pt. 15, at 6635, 6660 (1934) (statement of Richard Whitney, President, NYSE); similar statement at Stock Exchange Regulation, Hearings on H .R. 7852 Before the Comm. on Interstate and Foreign Commerce, 73d Cong., 2d Sess ., at 218. After the passage of the Act, the Commission formulated 16 rules gove rning the conduct of specialists and other floor members ("1935 Trading Rules"), which each of the exchanges adopted into their own rules, with some modificatio ns, by the end of 1935. See Securities Exchange Act Release No. 179 (April 17, 1935); SEC, Report on the Feasibility and Advisability of the Complete Segregat ion of the Functions of Dealer and Broker ("Segregation Study") (June 20, 1936), at Appendix O-1 (copy of the text of rules as proposed). One of these rules ex panded upon the NYSE's restrictions on a specialist's trading while the speciali st held a customer order as agent. Specifically, this rule prevented a speciali st from buying (selling) a stock for its own account (1) while holding an unexecuted customer market order to buy (sell) as agent; or (2) at a price equal to or lower than (higher than) a customer buy (sell) limit order on the specialist's book. Id. The specialist, however, was permitted to buy (sell) at a price higher (lower) than the customer's limit order . Id. See also NYSE Rule 92 (current version of rule).] Second, in 1910, the NYSE adopted a rule prohibitingspecialists from trading as dealer with a customer order on the specialist's book without obtaining the consent of the customer. [See NYSE Rules, Ch. XI, Sec. 1 (1927); Special Study, supra note , at 65.] The NYSE amended this rule in 1922 to permit a specialist to trade with a customer order on his or her book provided that the price was justified by market conditions and that the broker who entrusted the customer's order to the specialist was notified that the transaction occurred on a principal basis and accepted the trade. [See Special Study, supra note , at 65.] In 1932, the NYSE further amended this rule to require that prior to trading as dealer with a customer's order to buy (sell), the specialist must publicly bid (offer) the customer's order at a price 1/8 lower than the specialist's own offer (bid). [See NYSE Rules, Ch. XI, Sec. 1 (1932). See also NYSE Rule 91 (current version of rule). In addition, prior to pairing off, or crossing, two customer orders for execution, the specialist was required to make a bid (offer) on behalf of the customer order to buy (sell) at a price 1/8 lower (higher) than the specialist's offer (bid). See NYSE Rules, Ch. III, Sec. 13 (1927).] Finally, in 1934 the NYSE adopted a ruleprohibiting a specialist from disclosing, except to NYSE officials, any information with regard to orders entrusted to him or her as a specialist. [See NYSE Rules Ch. XI, Sec. 2 and Ch. XIV, Sec. 12 (1934). Compare Section 11(b) of the Act, 15 U.S.C. 78k(b).]

On the NYSE, the number of specialists increased significantly from 1910 to 1930. [There were 123 specialists out of a total of 858 active NYSE floor members in 1910; 169 specialists out of 871 members in 1920; and 352 specialists out of 1188 members in 1933. See Twentieth Century Fund, supra note , at 403 and 406. ] At least as early as the mid-1920s, any NYSE member seeking to become a specialist in a stock was first required to receive the approval of the NYSE's Committee of Arrangements. [See Id. at 428-29.] The Committee required the member to have made the appropriate provisions for carrying on a specialist business (e.g., arranging for clearance and settlement of trades, clerical help, and telephone connections) before he or she would be assigned a space at the post where the stocks that the specialist sought to specialize in were traded. [In practice, a member would only become a specialist in a stock if he or she had some floor experience and only in stocks in which there appeared to be a need for better service. Id. Among the Commission's 1935 Trading Rules were a prohibition against a member acting as a specialist in a stock without registering as such with the exchange and a requirement that a specialist maintain, for twelve months, records of all orders placed with the specialist in his or her specialty stocks. See Securities Exchange Act Release No. 179, supra note ; Segregation Study, supra note , at Appendix O-1.]

By 1936, the specialist system was utilized by nine of the 23 exchanges [Segregation Study, supra note , at 25-26.] registered with the Commission as national securities exchanges pursuant to Section 6(a) of the Act. [15 U.S.C. 78f(a).] The rules of the NYSE permitted more than one specialist to trade a particular stock. Stocks that were actively traded had up to six competing specialists, whereas one specialist was the norm in less actively traded stocks. [Twentieth Century Fund, supra note , at 404-405. On the NYSE as of 1933: 11 stocks had six specialists; 22 stocks had five specialists; 159 stocks had three specialists; 195 had two specialists; and 740 stocks had one specialist. Id. See also Segregation Study, supra note , at 5-6.] The Special Study suggested that over time floor brokers probably tended to give orders to those competing specialists who were willing to deal as principal when it was necessary to provide a broker with an execution at a price close to the last sale. [See Special Study, supra note , at 61-62 (adding that at one time, commission firms invited specialists to start competing books).] Furthermore, the Segregation Study noted that, in selecting a competing specialist with whom to entrust their orders, floor members generally preferred the specialist who "st[ood] ready at all times to narrow the quoted market and preventprice fluctuations." [See Segregation Study, supra note , at 41. ] Accordingly, the Segregation Study concluded that the competing specialist model provided the specialist with "an important incentive to maintain a fair and orderly market, viz, the desire to attract commission business." [Id.]

Each competing specialist in a stock maintained his or her own book of limit orders. These separate limit order books did not interact with each other. As a result, while time priority was maintained between limit orders at the same price on any given specialist's book, if there was more than one specialist making a market in a particular stock, time priority would not be maintained among all same-priced limit orders at that price on the exchange. [As was noted previously, the 1935 Trading Rules and predecessor rules of the exchanges utilizing the specialist system only prevented a specialist from trading for his or her own account while holding a same-sided customer market order at a price equal or inferior to that of a customer limit order on the specialist's book. See supra note and accompanying text.]

On the NYSE, the number of stocks that had competing specialists steadily declined after the passage of the Act in 1934, from 466 in 1933 to 37 in 1963. [See Special Study, supra note , at 62, and Table VI-3.] The Special Study associated this decline with the effect of certain Exchange regulatoryrequirements [In 1939, the Exchange began a program to compel specialists to become dealers, which the Special Study believed possibly resulted from the need of brokerage houses that were "faced with difficulty in obtaining good executions in that period of low volume." Id. This program included the institution of a specialist net capital requirement, which although considered "nominal," was vigorously enforced at its inception, leading to the demise of a number of specialist firms. Id. Contemporaneously, the Special Study noted that the NYSE withdrew the registration of some firms "for their failure to deal adequately." Id.] and allocation policies, [The Special Study cited a 1953 policy of the Exchange which did not allow the allocation of new stocks to one-member specialist firms, and did not allow any such ventures to be started. Id.] which resulted in a drastic reduction in the number of individual specialists and the concentration of specialist units over this period, [Since 1933, NYSE specialists increasingly have combined to form fewer, and larger, specialist units:

Year	Total Number of Specialists	Number of Specialist Units	Number of Individual Specialists
1933	352	125	105
1963	360	109	1
1975	381	67	0
1987	422	55	0
1997	452	38	0

See Twentieth Century Fund, supra note 11, at 4 04; Special Study, supra note 10, at 67; Report of the Pr esidential Task Force on Market Mechanisms (January 1988), Study VI at VI-5. as well as a decrease in volume of trading per issue and the concentration of volume in the most active stocks. [The Special Study suggested the decrease in volume and concentration in active issues accentuated the importance of the dealer function, leading to thinner specialist books and less opportunity for specialists to earn brokerage income on their stocks, contributing to the decline in competing specialists on the NYSE. See Special Study, supra note , at 63-64.] By 1964, there were only twospecialists that competed with each other in two stocks on the NYSE, and there were no competing specialists at that time on the American Stock Exchange ("Amex"). [See Securities Exchange Act Release No. 7432 (September 24, 1964) (noting that the "absence of competing specialists [on the NYSE and Amex] makes an effective system of [specialist] regulation and surveillance particularly important").] In 1967 the last competing specialist disappeared from the NYSE floor. [Subcomm. on Securities, Comm. on Banking, Housing and Urban Affairs, Securities Industry Study, S. Doc. No. 13, 93d Cong., 1st Sess. (1973) ("Securities Industry Study") at 122 (citing Institutional Membership on National Securities Exchanges: Hearings on S. 1164 and S. 3347 Before the Subcomm. on Securities, Comm. on Baking, Housing and Urban Affairs, 92d Cong., 2d Sess. (1972) pt. 1, at 379 (statement of Robert Haack, President, NYSE)) ("Institutional Membership Hearings"). In 1972, the NYSE Board of Governors rejected an application by a specialist firm to be permitted to compete with another firm in the latter's 14 specialty stocks, maintaining that a recently-adopted NYSE initiative to improve specialist performance "should be given a trial period to see how effective it is before any alternative ways to achieve the same goals be adopted." See Securities Industry Study at 122 (citing Wall St. J. (January 21, 1972) at 4).]

b.Pacific Stock Exchange Competing Specialists (1976 - 1977)

In June 1976, the Commission approved a proposed rule change by the PSE to a dopt a six-month pilot program to permit competing specialists to trade on the E xchange in selected stocks. [See Securities Exchange Act Release No. 12578 (Jun e 25, 1976) (File No. SR-PSE-76-24) (order approving PSE competing specialist pi lot program on a six-month basis). As was noted above, by virtue of having two equity trading floors, the PSE already had (and continues to have) two co-specia lists in most of the stocks traded on the Exchange. Under PSE rules, a bid or o ffer that is clearly established as the first made at a particular price is enti tled to priority and has precedence up to the number of shares specified in the bid or offer, and each co-specialist is responsible for coordinating and synchro nizing orders and executions with the co-specialist on the other floor. See PS E Rule 5.8. The co-specialists coordinate the market in a stock and communicate with one another through dedicated telephone lines, and, as of October 1996, an electronic Consolidated Limit Order File, which allows each co-specialist to vi ew the orders on the other's limit order book.] ThePSE believed that consisten t with Section 11A(a)(1)(C)(ii) of the Act, [15 U.S.C. 78k-1(a)(1)(C)(ii). ] th e competing specialist pilot program would permit the PSE to test its ability to offer increased competition in selected securities. [The immediate impetus for PSE's adoption of its competing specialist program was to enable it to attempt t o establish itself as the primary market for BankAmerica stock, which began trad ing on June 28, 1976 and was initially listed on the Chicago Stock Exchange ("CH X"), NYSE, and PSE. On that date, the PSE had 17 competing specialists from 11 firms making markets in BankAmerica. See Wall St. J. (June 29, 1976) at 6. ] The PSE's program provided for the appointment of competing specialists, with r esponsibilities specified in the PSE rules, [See PSE Rule 5.35 (f) and (g) for a description of the responsibilities of PSE competing specialists.] in conjunc tion with the appointment of a book broker. [The proposal allowed for the appoin tment of a book broker on either or both of the Exchange's trading floors. See PSE Rule 5.35(c).] The book broker(s) would be responsible for operating the l imit order book and executing odd-lot orders and orders routed through the Exchange's automated order routing and execution system instocks traded on a competing specialist basis. [Under the PSE rules, a book broker was generally prohibited from engaging in principal transactions. See PSE Rule 5.35(e). Either at the request of a floor broker holding an order or whenever in the book broker's opinion the interests of a fair and orderly market were served, a book broker was obligated to call on the competing specialist(s) to make bids and/or offers or to narrow spreads in existing bids or offers or otherwise fulfill the responsibilities of a competing specialist. See PSE Rule 5.35(f).] The PSE's competing specialist program has been inactive since 1977, though the procedures pertaining to competing specialists remain in the PSE rules, and subsequently have been amended in conjunction with revisions to other PSE rules. [Securities Exchange Act Release Nos. 13458 (April 22, 1977) (File No. SR-PSE-77-6) (order approving PSE proposal to provide alternate capital requirements for competing specialists); 14549 (March 10, 1978) (File No. SR-PSE-78-3) (establishment of fee for participation in competing specialist or alternative specialist programs); 26878 (May 30, 1989), 54 FR 24616 (June 8, 1989) (File No. SR-PSE-89-05) (notice of proposal to increase capital requirements for competing and alternate specialists); and 26988 (June 28, 1989), 54 FR 28538 (July 6, 1989) (File No. SR-PSE-89-05) (order approving proposal to increase capital requirements for competing and alternate specialists).] It should be noted that the PSE continues to have competing specialists because the PSE has co-specialists in the overwhelming majority of stocks traded on its dual trading floors. However, as the PSE's rules require the maintenance of time priority between co-specialist quotes, the trading activity of PSE specialists does not strictly meet the definition of preferencing found in the NSMIA.

c.New York Stock Exchange Competing Specialists (1976 - 1994)

In adopting the 1975 Amendments to the Act, [Pub. L. No. 94-29, 89 Stat. 97 (1975) ("1975 Amendments").] Congress directed the Commission to facilitate the establishment of a national market system ("NMS") for securities. The legislative history to the 1975 Amendments asserted that the first priority in creating a NMS was to break down "unnecessary regulatory restrictions which now impede contact between brokers and market makers and which restrain competition among markets and market makers," [S. Rep. No. 75, 94th Cong., 1st Sess. 7-8 (1975) ("Senate Report") at 12-13.] and that developments in the markets had rendered a single specialist unable to serve the needs of both individual and institutional investors in a security. [Specifically, the Senate Report stated that "Committee investigations have adequately demonstrated that in our increasingly complex and institutional markets a single specialist...cannot provide adequate liquidity and continuity to the market for a security. To assure that markets are able to serve the needs of both individual and institutional investors, the Committee believes many types of market makers are necessary and that encouragement should be given to all dealers to make simultaneous competing markets within the [NMS]." Id. at 14.] As a result, the Commission undertook efforts to encourage the various exchanges to amend their rules to increase market making competition among their members.

Largely in response to the adoption of the 1975 Amendments and the Commissio n's subsequent efforts, in May 1976, the NYSE's Board of Directors endorsed a sy stem of competition between Exchangespecialists by reaffirming the ability of me mbers to register and act as specialists in stocks that were also assigned to ot her specialists. [Compare supra note .] In September 1976, the NYSE filed wi th the Commission certain procedures for the consideration of members' applicati ons to compete as specialists, which it characterized as a stated policy, practi ce, or interpretation not deemed to be a rule of the NYSE pursuant to Rule 19b-4 of the Act. [See Securities and Exchange Commission File No. SR-NYSE-76-43 (Se ptember 8, 1976).] The Commission believed that a number of the procedures wer e, in fact, changes in, or additions to, NYSE rules; accordingly, the Commission informed the NYSE that such procedures could not have any force or effect witho ut first receiving approval from the Commission under Section 19(b) of the Act. [Sections 19(b)(1)-(2) of the Act and Rule 19b-4 thereunder provide that a proposed rule change of a self-regulatory organization ("SRO") must be filed with and approved by the Commission. 15 U.S.C. 78s(b)(1)-(2) and 17 CFR 240.19b-4. However, Rule 19b-4 provides that a "stated policy, practice, or interpretation" of an SRO that is "reasonably and fairly implied" from an existing rule of the SRO is not considered a proposed rule change. Id. The specific NYSE procedures deemed to be a proposed rule change subject to the Commission's approval were: (1) an increased capital requirement for specialist units applying to compete; (2) a requirement that a unit must consist of at least five specialists in order to compete; (3) a review of the applicant unit's business history; (4) a provision that barred a competing unit that withdraws from competition from receiving new allocations for six months after such withdrawal; and (5) a training period applicable to new competing units. See Letter from George A. Fitzsimmons, Secretary, SEC, to James E. Buck, Secretary, NYSE, dated October 20, 1976 ("October 1976 Letter"). Further, the NYSE asserted that its existing rules regulating specialists were adopted contemplating the regulation of then- existing competing specialists, and therefore did not compromise the ability of specialists to compete. See Securities and Exchange Commission File No. SR-NYSE-76-43, supra note . In response, the Commission stated that while competition among specialists in the same stock was not technically prohibited under NYSE rules, the current NYSE rule structure may have imposed burdens on competition not justifiable in furtherance of the Act and "could ultimately frustrate efforts to introduce and maintain competition on the NYSE floor." See October 1976 Letter. ] However, theCommission's action did not preclude the NYSE from continuing to consider and approve applications to compete.

In October 1976, the NYSE approved an application by a specialist unit to compete in 18 stocks assigned to another unit. [A previous application to compete that was filed in June 1976 was later withdrawn following the break-up of the specialist unit facing competition and the division of its stocks among other units, including the applicant. The NYSE subsequently approved applications to compete in July 1977 and February 1978.] In February 1977, the NYSE submitted to the Commission a proposal to adopt procedures by which the Exchange would review and approve an application by a specialist unit to act as a competing specialist in stocks already assigned to another unit. [The NYSE's proposal to adopt its Procedures for Competing Specialists was noticed for public comment in Securities Exchange Act Release No. 13319 (March 1, 1977), 42 FR 13176 (March 9, 1977) (File No. SR-NYSE-77-6).]

Due to various regulatory issues raised by the proposal, the NYSE's competin g specialist proposal remained pending before theCommission until May 1986, [In the interim, the NYSE had rescinded NYSE Rule 114, which (1) required specialist units to be comprised of at least three active specialists, and (2) prohibited specialist units registered in the same specialty stock from entering customer l imit orders into a joint limit order book. The Commission believed that the res cission of this prohibition could contribute to an environment more conducive to competition among specialists as dealers, ameliorating physical constraints on the NYSE floor against such competition ( e.g. , new units might have been able to gain entry without the necessity of the Exchange providing additional facilit ies). See Securities Exchange Act Release No. 14147 (November 7, 1977) (File No . SR-NYSE-77-25). The Commission, however, requested the NYSE to furnish additi onal information as to its current rules governing the operation of multiple lim it order books in competing specialist situations, the adequacy of such rules an d whether additional rules were required and whether the NYSE believed that requ iring all competing specialists to enter limit orders into a single repository w ould further the purposes of the Act. Id. Additionally, in 1979 the NYSE adopte d a specialist evaluation program, under which a specialist unit that failed to meet minimum performance standards would be subject to a performance improvement action and, in the event of continued substandard performance, reallocation of one or more of its specialty stocks. See NYSE Rule 103A. Under Rule 103A, a s pecialist from whom a stock had been reallocated as a result of substandard perf ormance may register as a competing specialist in the reallocated stock. See S ecurities Exchange Act Release No. 15827 (May 15, 1979) (File No. SR-NYSE-77-24) .] when it was approved by the Commission on a six-month pilot basis to enable the NYSE to approve the registration of a competing specialist. [The NYSE stated that exigent circumstances caused the Exchange to approve the registration of a specialist unit to compete with another unit pursuant to the proposed procedure s, making it necessary for the Exchange to have Commission-approved procedures i n place to accommodate the unit's registration. See Securities Exchange Act Release No. 23202 (May 5, 1986), 51 FR 17424 (May 12, 1986) (File No. SR-NYSE-77-6) (order approving NYSE competing specialist procedures on a six-month pilot basis, effective May 2, 1986 through November 2, 1986).] However, the Commission required the NYSE to assess the adequacy of its rules as applied to competing specialists and develop during the pilotperiod appropriate amendments to its rules in order to accommodate competing specialist activity. [In this regard, the Commission delineated nine specific areas in which it believed such procedures should be developed: (1) the proper location on the floor of competing specialist units ( e.g. , whether they should be adjacent); (2) the routing of orders through the Designated Order Turnaround system to competing specialists; (3) opening prices and order imbalances at the opening; (4) methods for determining order imbalance trading halts; (5) trade and quote reporting and the display of quotes through the Intermarket Trading System; (6) the execution of limit orders in the same stocks held on each competing specialist's book, and application of the rules of priority, parity, and precedence under NYSE Rule 72 to the execution of such orders; (7) the participation of each competing specialist in block order executions under NYSE Rule 127; (8) the execution of various types of orders such as percentage and stop orders; and (9) the application of the specialist functions under NYSE Rule 104 to each competing specialist, including with respect to each specialist's affirmative and negative obligations. Moreover, the Commission anticipated that other NYSE rules and procedures may require amendments as the Exchange monitored competing specialist activity during the pilot. In addition, the Commission disagreed with the NYSE's statement that no further rule changes were necessary to accommodate competing specialist situations.]

The Commission subsequently extended the initial pilot program three times, through April 30, 1988. [See Securities Exchange Act Release Nos. 23869 (December 9, 1986), 51 FR 45417 (December 18, 1986) (File No. SR-NYSE-86- 31) (order extending pilot retroactively from November 1, 1986 for three months, through February 1, 1987); and 24183 (March 5, 1987), 52 FR 7721 (March 12, 1987) (File No. SR-NYSE-87-2) (order extending pilot through June 1, 1987). After an eight month lapse of the pilot, in February 1988 the Commission approved a proposal by the Exchange to reinstate the pilot until April 1, 1988. See Securities Exchange Act Release No. 25342 (February 11, 1988), 53 FR 5066 (February 19, 1988) (File No. SR-NYSE-88-01) (order reinstating pilot through April 1, 1988).] The Exchange requested the extensions to gain additional experience monitoring the activities of competing specialists before finalizing its responseto the questions posed by the Commission in the order originally approving the pilot program and submitting a request for permanent approval of its competing specialist procedures. [The Exchange's second request for extension of the pilot program indicated that specialist units had recently voluntarily withdrawn from a competitive situation. See Securities Exchange Act Release No. 24183, supra note .]

In September 1990, the NYSE submitted a proposal to permanently adopt procedures for the appointment and withdrawal of competing specialists. [The Commission noticed the NYSE's proposal for comment in Securities Exchange Act Release No. 28586 (October 29, 1990), 55 FR 46597 (November 5, 1990) (File No. SR-NYSE-90-46) (notice of NYSE proposal to adopt competing specialist procedures on a permanent basis).] The proposal included the Exchange's responses to the Commission's questions concerning the impact of competing specialist trading on certain NYSE rules. Moreover, the Exchange stated its belief that its responses were interpretations of existing rules that had been applied in previous competing specialist situations on the NYSE and that no formal modifications of its rules were necessary at that time to accommodate competing specialist trading. [In this regard, the Exchange believed that, in accordance with Section 19(b) of the Act and Rule 19(b)(4) thereunder, its responses were "stated policies, practices or interpretations" that were "reasonably and fairly implied" by existing NYSE rules and policies, and therefore were not deemed proposed rule changes under the Act. See supra note .] Consistent with its previous statements, the Commission disagreed with the Exchange's belief, given that the proposal would change significantly the existing trading procedures on the Exchange. [See supra notes and .]

In its response to the Commission's questions, the NYSE indicated that notwithstanding the existence of competing specialists, there would remain only one Exchange auction market in each stock subject to competition. Accordingly, the Exchange stated that its policies as to the execution of limit orders and the application of the rules of priority, parity, and precedence would remain unchanged. [See NYSE Rule 72. See also infra Part II.B.1.] The Exchange, however, would not require the maintenance of a unitary limit order book in a stock traded on a competing specialist basis. Therefore, limit orders at a particular price would not be executed on a strict time priority basis with regard to all such limit orders on the Exchange, but would only receive time priority with regard to other limit orders received by a particular competing specialist. [In addition, the Exchange stated that when acting as agent or principal or both, each competing specialist was to regard its competitor in the same light as it would a broker representing orders in the trading crowd, and that competing specialists may agree to "split" stock pursuant to NYSE Rule 72.20. See Id.] Further, Exchange policy provided that member organizations were to select (i.e., preference) the particular competing specialist unit that was to receive its Designated Order Turnaround ("DOT") system order business [If the member organization did not want to select a particular specialist unit, the NYSE would request the units to reach agreement on a means to handle the order flow. If such an agreement could not be reached, the NYSE would require the member organization to select a particular unit. ] and competing specialists in a stock were expected to disseminate a unitary quote through the Intermarket Trading System("ITS"), while the satisfaction of an incoming ITS commitment was the responsibility of the specialist making the bid or offer in question.

Commission staff, however, objected to the complete removal of a central limit order book on the Exchange for stocks that were to be traded on a competing specialist basis. The NYSE resisted the establishment of a unified limit order book for competing specialists, arguing that this would defeat the purpose of true competition between specialist units. In this regard, the NYSE argued that it was necessary for competing specialists to maintain separate limit order books so that each specialist could adequately represent its limit orders as agent. As the Commission and the NYSE were unable to resolve this issue, the NYSE subsequently withdrew its competing specialist proposal in September 1994 (three years after the initiation of the CSE's preferencing program). The NYSE's competing specialist program, if widely implemented, could have transformed the NYSE into a preferencing exchange as defined by Section 510(c)(3) of the NSMIA.

B.The Cincinnati Stock Exchange's Dealer Preferencing Program (1991 - Present)

1.Background

Established in 1885, the CSE is registered as a national securities exchange pursuant to Section 6(a) of the Act. [15 U.S.C. 78f(a).] Beginning in 1976, th e CSE changed its method of operation by phasing out its physical trading floor, automating its exchangeoperations, and adopting a competing market maker system . [See Securities Exchange Act Release No. 14674 (April 18, 1978), 43 FR 17894 (April 26, 1978) (File No. SR-CSE-77-1) (order approving CSE's National Securiti es Trading System on an initial nine-month pilot basis).] Trading on the CSE cu rrently is conducted through the National Securities Trading System ("NSTS& quot; or "System"), an electronic securities communication and executi on facility that interfaces with CSE member specialist workstations and order de livery systems in physically dispersed locations. [For a detailed discussion on the development of the NSTS from its inception in 1978 based upon predecessor systems to the initiation of the CSE's preferencing program in 1991, see Appendix C.] The NSTS combines a limit order file capable of being viewed by all NSTS "Users," [The CSE rules define a "User" as either a CSE member or an Approved Dealer, who may or may not be a CSE member. See CSE Rule 11.9(a)(5). An "Approved Dealer" is a Designated Dealer, a Contributing Dealer, or a specialist or market maker registered as such with another exchange with respect to a designated issue. See CSE Rule 11.9(a)(2). A Designated Dealer is a CSE member who maintains a minimum net capital amount and who has been approved by the CSE's Securities Committee to perform market maker functions by entering bids and offers into the NSTS. See CSE Rules 11.9(a)(3) and (b). During Exchange trading hours, a Designated Dealer is required to provide continuous bids and offers for round lots of his designated issues. See CSE Rule 11.9(c)(iii). A Contributing Dealer is a CSE member who maintains a minimum net capital amount and provides regular bids and offers for round lots of his designated issues. Currently, no CSE member is registered with the Exchange as a Contributing Dealer, nor are any NSTS terminals located on the floor of another exchange. See Securities Exchange Act Release No. 38117 (January 3, 1997), 62 FR 1480 at n.3 (January 10, 1997) (File No. SR-CSE-96-08) (order approving CSE Regulatory Circular re dealer quotation obligations) (citing Telephone Conversation between Jon Kroeper, Attorney, SEC, and Adam Gurwitz, Director of Legal Affairs, CSE, dated January 2, 1997).] the display of current CSE and national quotation and last saleinformation, and the electronic matching and execution of like-priced System interest according to programmed price/time and agency/principal priorities. [See CSE Rule 11.9(a)(1).]

CSE members approved to be Designated Dealers in NSTS are required to perform market making functions, including the provision of continuous two-sided markets in issues designated for trading in the System ("designated issues") in which they are registered. [See CSE Rule 11.9(a)(6).] Designated Dealers are also required to provide guaranteed executions for public agency market and marketable limit orders [A public agency order is defined as "any order for the account of a person other than a member, an Approved Dealer or a person who could become an Approved Dealer by complying with this Rule with respect to his use of the System, which order is represented as agent, by a User." See CSE Rule 11.9(a)(7). A marketable limit order is a limit order to buy (sell) that is immediately executable because the best offer (bid) available in ITS at the time the order is entered is equal to or better than the limit price on the order. See CSE Rule 11.9(c)(iv).] of up to 2099 shares at the ITS best bid or offer ("ITS/BBO"). [The ITS/BBO represents the highest bid and lowest offer available at a particular time among the market centers participating in the ITS that trade the stock.] If there is more than one Designated Dealer in a designated issue, the public agency guarantee obligation rotates among the Designated Dealers on a daily basis. [The Designated Dealer with the guarantee obligation is referred to as the "Designated Dealer of the Day." See CSE Rule 11.9(n). ]

As of April 1997, the CSE has 72 members and member organizations, including 13 Designated Dealers, seven of which participate in the CSE's preferencing program. With the exception of six stocks, the approximately 568 stocks eligible for trading on the CSE are stocks of NYSE-listed and Amex-listed companies that are traded on the CSE pursuant to unlisted trading privileges ("UTP") granted under Section 12(f) of the Act. [See 15 U.S.C. 78l(f).]

2.Non-Preferenced Executions on the CSE

Under the CSE's rules, Users may enter orders into NSTS in designated issues. NSTS executes orders that are not preferenced to a Designated Dealer based on price and time priority. [See infra Part I.B.3 for a description of trading by CSE preferencing dealers. ] All orders, and Designated Dealer bids or offers, entered into the System are queued for execution so that the highest priced bid or lowest priced offer entered earliest in time is the first to be executed. [The CSE rules, however, contain an exception to the time priority of public agency orders due to a Designated Dealer's guarantee to execute a customer limit order upon the occurrence of a transaction in another market at the price of the limit order. See CSE Rules 11.9(a)(10), (l), and (p).] In addition, all public agency orders are given precedence over professional agency orders, [The CSE rules define a professional agency order as an order entered by a User as agent for the account of a broker-dealer, a futures commission merchant, or a member of a contract market. See CSE Rule 11.9(a)(8).] and proprietary orders, bids and offers at the same price. [See CSE Rules 11.9(l) and (m). ]

Public agency market and marketable limit orders (other than preferenced ord ers) currently are executed in the following manner: [For a description of CSE d ealers' procedures for the handling of non-marketable limit orders prior to thei r execution, see infra Part IV.B.1.b.] A Designated Dealer or non-Designated Dealer User may enter a public agency market or marketable limit order into NSTS , which prices the order at the ITS/BBO at the time of its entry into the System . [A marketable limit order must be priced at the current ITS/BBO, because the e xecution of the limit order to buy (sell) at a price higher than (lower than) the ITS/BBO would constitute a trade through in violation of ITS trading rules. See CSE Rule 14.9 (ITS trade-through rule). NSTS automatically enforces the ITS trade-through rule for a marketable limit order to buy (sell) at a price higher than (lower than) the current ITS/BBO by pricing such an order at the ITS/BBO upon its entry into NSTS.] The order is first executed against any contra-side agency interest on the CSE's central limit order book at the ITS/BBO, and then against any principal interest at the ITS/BBO on the System (either in the form of a limit order on the CSE book or as represented in a dealer quote). If the order is of a size up to the lesser of 2099 shares or the quoted size at the ITS/BBO at the time of its entry into the System, any remaining portion of the order is executed against the Designated Dealer of the Day pursuant to the CSE's public agency guarantee. [See CSE Rule 11.9(n). Further, the Designated Dealer of the Day's obligation to execute such orders is reduced by the number of shares executed in the NSTS against any agency or principal interest, including that of the Designated Dealer of the Day, priced at the ITS/BBO at the time of entry of the order into the NSTS ( i.e. , the public agency order could be executed against both a Designated Dealer's quoted size at the ITS/BBO and in its capacity as Designated Dealer of the Day). See CSE Rule 11.9(n)(5).] If the order either islarger than 2099 shares [No portion of an order larger than 2099 shares is subject to the public agency guarantee. See CSE Rule 11.9(n)(4).] or the quoted size at the ITS/BBO, the remainder is exposed for 15 seconds to all Designated Dealers (whether or not they are registered in the designated issue involved) on their NSTS trading stations, priced at the ITS/BBO. Any unexecuted amount is formatted as an ITS outbound commitment and sent to the ITS participant displaying the ITS/BBO at that time.

Professional agency market or marketable limit orders are executed in the same fashion as public agency orders, with the exception that such orders are not eligible to be executed against the Designated Dealer of the Day in accordance with the public agency order execution guarantee. As with public agency orders, any amount of a professional agency order that is not executed against NSTS interest is formatted as an ITS outbound commitment and sent to the ITS participant displaying the ITS/BBO at that time.

A User who is a Designated Dealer may trade as principal in a designated issue in which it is registered [Designated Dealers may also trade on a principal basis in designated issues in which they are not registered, as long as such dealers comply with Section 11(a) of the Act, 15 U.S.C. 78k(a), and the rules and regulations thereunder. See CSE Rule 11.9(f).] by generating an order in the form of a commitment to trade for a specified numberof shares and entering it into the System. [NSTS automatically enforces the ITS trade-through rule for inter-dealer trades by rejecting any attempt by a Designated Dealer to enter a commitment to trade at a price inferior to the ITS/BBO. Compare supra note . In addition, a Designated Dealer currently may trade with another CSE dealer by changing either its bid or offer to lock the contra-side quote of another Designated Dealer.] These commitments to trade are first immediately matched with and fully or partially executed against any contra-side agency or principal limit orders on the CSE's central limit order book at the ITS/BBO. If no such interest exists, or there is any remaining unexecuted amount, the commitment then is executed against the contra-side bid or offer of any Designated Dealer(s) at the ITS/BBO, up to the size of the dealer's displayed quote. Any remaining amount would be rejected by NSTS and a notification to this effect would be sent to the dealer entering the commitment to trade.

3.Description of the CSE's Preferencing Program

The CSE preferencing program modifies the CSE's priority rules described abo ve to allow Designated Dealers participating in the program ("CSE preferenc ing dealers") to have priority over same-priced market maker or professiona l agency interest entered into the NSTS prior in time when the preferencing deal er is interacting with the public agency market and marketable limit orders it i s representing as agent. [See CSE Rules 11.9(l), (m) and (u).] Under the pref erencing program, CSEdesignated dealers send paired trades to the NSTS for execu tion. [The dealer may interact with public orders it represents as agent either by (1) taking the contra-side of the trade as principal ("paired order trade"), or (2) crossing the order with another customer order it represents as agent ("agency cross"). The majority of agency crosses are the result of a customer limit order resident in the dealer's proprietary system at the ITS/BBO, which is matched with an incoming contra-side market order. For example, if the ITS/BBO is 20 - 20 1/8, and a dealer holds a customer limit order to buy at 20, an incoming market sell order will be matched with that limit order because the dealer may not trade for its own account ahead of its own customer limit order. See CSE Rule 12.6(b).] These paired trades must be priced at or between the ITS/BBO. Before executing the paired order trade on the CSE, the NSTS replaces the dealer's side of the trade with any public agency limit order at the same price that is on the CSE's central limit order book. [NSTS will displace the dealer's side of the paired trade up to the size of the order(s) on the limit order book ( i.e. , the balance of the paired trade will remain intact and be executed). The NSTS follows the same procedure in the event of pre-existing public agency interest on the CSE central limit order book at the price of an attempted agency cross. Only one side of a paired order trade entered into NSTS is susceptible to being broken up by a public agency order, as NSTS would have automatically executed the buy and sell limit orders at the proposed trade price against one another (or contra-sided Designated Dealer interest) prior to the entry of the paired order trade. ] If there are no such public agency orders, the paired trade is executed, regardless of other prior CSE dealer quotations or orders at the same price. In this manner, the preferencing program ensures that preferenced orders still have an opportunity to interact with customer orders on the CSE's central limit order book, while permitting CSE preferencing dealers to match against their own customer orders when those orders wouldhave otherwise been executed against another professional.

For example, if Dealer A on the CSE is quoting at the ITS/BBO, preferencing Dealer B can still internalize its order flow (even if Dealer B is not quoting at the ITS/BBO) so long as Dealer B executes the order at the ITS/BBO (or better) and there is no contra-side public agency order in NSTS at that price. [In contrast, applying CSE's procedures for executing a non- preferenced trade on the CSE in this example, if there was no contra-side public agency order on the CSE's central limit order book, Dealer A would have time priority over Dealer B and customer order entered into NSTS would be executed against Dealer A. See supra Section I.B.2. ] If there is a public agency limit order in the CSE's central limit order book with priority, however, NSTS will automatically break Dealer B's paired order trade up to the size of the pre-existing public limit order on the CSE central limit order book. [But see discussion, infra Appendix A at note 26 and accompanying text (commenters' concerns regarding the lack of public customer interest on the CSE's central limit order book, and the CSE's response).] Dealer B's customer order then will be executed against the pre-existing limit order, with any remaining portion executed against Dealer B's side of the trade.

The CSE also has two order handling policies that apply specifically to preferencing dealers. The first policy is a price improvement policy applicable in greater than minimum variation markets, which requires preferencing dealers to either immediately execute a market order routed to it for execution on the CSE at an improved price or expose the order on the Exchange for a minimum of 30 seconds to allow other market participants an opportunity toprovide an improved price. [See CSE Rule 11.9(u), Interpretation and Policy .01. In exposing a market order on the Exchange for price improvement, a dealer "stops" the order ( i.e. , guarantees the execution of the order at the then-best inter-market price) in the event that the order does not receive price improvement. As part of its Order Execution Obligations proposals, the Commission had proposed a rule requiring that in markets where the difference between the best bid and offer is greater than the minimum trading variation, all market orders receive an opportunity for price improvement. See Securities Exchange Act Release No. 36310 (September 29, 1995), 60 FR 52792 (October 4, 1995) (File No. S7-30-95) ("OEO Proposing Release"). At the present time, however, the Commission has deferred taking action on the price improvement proposal. See OEO Adopting Release, supra note .] To comply with the policy, a preferencing dealer may either represent the order at an improved price in his or her CSE quote, or place the order on the CSE's central limit order book at an improved price.

The second policy is a limit order price protection policy that is designed to ensure that limit orders routed to CSE dealers for execution on the CSE receive timely executions relative to same-priced limit orders on the primary market for the security. [See CSE Rule 11.9(u), Interpretation and Policy .02. Under this policy, a public agency limit order routed to a CSE preferencing dealer for execution on the CSE would be filled if (i) the bid or offer at the limit price has been exhausted in the primary market; (ii) there has been a price penetration of the limit order in the primary market; or (iii) the stock is trading at the limit price on the primary market unless it can be demonstrated that such order would not have been executed if it had been transmitted to the primary market. The customer and the CSE dealer may agree to a specific volume related or other criteria for requiring execution of limit orders. ] This policy is substantially similar to limit order

price protection policies that have been adopted by other regionalexchanges. [The CHX and the BSE currently have limit order price protection policies nearly identical to the CSE's. See BSE Rules, Ch. II, Sec. 33; CHX Art. XX, Rule 37(a).]

4.The Commission's Permanent Approval of the CSE's Preferencing Program in 1996 [The CSE's preferencing program initially was approved on a six-month pilot basis in February 1991. See Securities Exchange Act Release No. 28866 (February 7, 1991), 56 FR 5854 (February 13, 1991) (File No. SR-CSE-90-06). Over the next five years, the Commission granted approval to a series of extensions of the CSE's pilot program. See Appendix A at note 5. For a detailed discussion on the CSE's preferencing pilot, including a summary of commentary received on the pilot and the CSE's response thereto, see Appendix A.]

On March 29, 1996, the Commission approved the CSE's preferencing program, as amended, on a permanent basis. [See CSE Approval Order, supra note .] In considering the CSE's request for permanent approval, the Commission had analyzed thoroughly the issues presented by the CSE's program, including the potential effect of preferencing on the execution of customer orders, competition among markets, and CSE market quality. This analysis involved a careful consideration of concerns raised by commenters, data submitted by the CSE and commenters, as well as the Commission's undertaking of its own data collection and examination. In approving the CSE's proposal, the Commission concluded that preferencing had improved CSE quotations and had added to the depth and liquidity of the CSE market. In addition, the Commission believed that the preferencing program, as supplemented by three order handling policies proposed by theCSE, [See Appendix A at notes 19-24 and accompanying text.] was not necessarily inconsistent with best execution of customer orders. [For a detailed discussion of the Commission's rationale for finding that the CSE's preferencing program is not necessarily inconsistent with best execution of customer orders, see infra Part III.C.1.] As a result, the Commission believed that the proposal, as amended, was consistent with the Act, particularly Sections 6(b)(5) and 11A. [15 U.S.C. 78f(b)(5) and 78k-1.]

The Commission believed that the CSE's adoption of an amended limit order display policy [See Appendix A at notes 22-24 and accompanying text.] and a primary market print protection policy [See supra note and accompanying text. See also Appendix A at notes 20-21.] addressed commenters' concerns that the ability of CSE preferencing dealers to internalize their own customer order flow provides them with an incentive to delay sending limit orders to the Exchange until they are marketable, thus depriving all orders on the CSE from the benefits of order interaction. In this regard, the Commission believed that these limit order policies should promote order interaction on the CSE through improved quotations and increased volume on the Exchange's central limit order book, as well as add to the quality of information displayed to the NMS. [The Commission noted that such orders will be included in the CSE consolidated quote and disseminated to the national market system, regardless of whether they were represented in the dealer's quote or placed on the Exchange's central limit order book. ]

Further, the Commission noted that while preferencing, and the resulting internalization of order flow by broker-dealers, may reduce order interaction on the CSE, preferencing does not inhibit dealers from providing executions of customer orders at or better than the ITS/BBO. In addition, the Commission noted that the CSE's proposed market order price improvement policy would prevent CSE preferencing dealers from internalizing market orders in greater than minimum variation markets without first executing such orders at an improved price or providing them with the opportunity for price improvement. [See supra note and accompanying text.]

The Commission also addressed the question of whether the CSE had made t he required demonstration that its preferencing program had added depth and liqu idity to the CSE market, and improved quotations. In response to the differing assertions made by the commenters and the CSE, [For a detailed discussion on the assertions and data submitted by the CSE and commenters, see Appendix A, at Sec tion C.] the Commission's Office of Economic Analysis ("OEA") evaluat ed CSE quotations and transactions before approval of the CSE filing. OEA's ana lysis found that CSE preferencing dealers often matched the NYSE BBO, and that t he percentage of time that the CSE quotes matched those on the NYSE was greatest for those stocks in which preferencing takes place. [During the time period con sidered in OEA's study, preferencing dealers accounted for more than 90% of trad es and 66% of share volume on the CSE. For the 281 stocks in which preferencing dealers accounted for 80% to 99% of total CSE trades, the CSE quote on average matched the NYSE best bid approximately 54% of the time and the NYSE best offer nearly 61% of the time. When matching at least one side of the ITS/BBO, the CSE' s quotation depth in these 281 stocks averaged over 720 shares. For all quotes in the 281 stocks, the CSE quotation depth averaged close to 900 shares.] The Commission concluded, based on the results of itsanalysis of the data submitted by commenters and the CSE, as well its own data analysis, [In response to concerns that CSE dealers' quotes do not add depth and liquidity to the NMS because they change too quickly for other market participants to react, the Commission noted that OEA analyzed the data regarding CSE quotations on a time- weighted basis, so that, unlike the figures provided by the commenters and the CSE, the results took into consideration whether the quotes at the NYSE BBO were short in duration relative to quotes outside of the NYSE BBO. In this regard, the OEA figures indicated that CSE dealers often quote at the NYSE BBO.] that the CSE's quotes compared favorably to those of the other regional exchanges and indicated that preferencing dealers had added depth and liquidity to the CSE market. [The Commission also believed that the CSE's amended limit order display policy could further add to the depth and liquidity of the CSE market. Specifically, display of limit orders could produce, among other benefits, quotes that more fully represent buying and selling interest in the market and enhance an investor's ability to monitor execution quality. The Commission concluded that this, in turn, should increase competition among CSE dealers based on their respective quotations.] The Commission, however, recognized that the quality of the CSE's market relative to other markets may change over time, and indicated that it would periodically review the practices of broker-dealers participating in the CSE's preferencing program. If the Commission found that a deterioration in their performance was evident, it would consider whether the CSE would need to discontinue the preferencing program, or take other actions toimprove the quality of market making on the CSE.

Finally, the Commission also removed the restrictions placed on the preferencing program during the pilot that limited the number of stocks that a single preferencing dealer could trade and prevented preferencing dealers from making cash payments for order flow. In this regard, the Commission concluded that these restrictions, which were put into place to allow the Commission to assess the effects of preferencing, were no longer necessary given the decision to permanently approve the preferencing program. [Contemporaneous with and after permanently approving the CSE's preferencing program, the Commission responded to numerous inquiries from Congress regarding the Commission's action. See Letter from Senator Alfonse D'Amato, Chairman, Committee on Banking, Housing, and Urban Affairs, to Arthur Levitt, Chairman, SEC, dated March 22, 1996 ( re consideration of proposal to approve CSE preferencing); Letter from Congressman Dingell, Ranking Member, Committee on Commerce, to Chairman Levitt, dated April 1, 1996 ( re permanent approval of CSE preferencing); Letter from [19 Congressmen], to Chairman Levitt, dated May 9, 1996 ( re permanent approval of CSE preferencing); Letter from Congressman Dingell, Ranking Member Committee on Commerce, to Chairman Levitt, dated April 18, 1996 ( re CSE trading in Lucent Technologies on April 4, 1996); Letter from Senator Christopher J. Dodd, Ranking Member, Senate Subcommittee on Securities, to Chairman Levitt, dated May 15, 1996 ( re CSE trading in Lucent Technologies on April 4, 1996); and Letter from Congressman Edward J. Markey, Ranking Member, to Chairman Levitt, dated June 12, 1996 ( re approval of preferencing).]

C.The Boston Stock Exchange's Competing Specialist Initiative (1994 - Present)

1.Background

The BSE is registered as a national securities exchange pursuant to Section 6(a) of the Act. Prior to the adoption of the Competing Specialist Initiative ("CSI") in May 1994, the BSE generally accounted for the smallest percentage of Consolidatedtape trades and volume of any regional exchange, with the exception of the CSE. As of April 1997, there were 30 specialists and 18 specialist units on the BSE, trading a total of 2,221 stocks. Six BSE specialist units participate in the CSI, each trading from four to 31 stocks on a competing specialist basis. Overall, 91 stocks currently are traded as part of the CSI.

The Boston Exchange Automated Communication Order-routing Network ("BEACON") is the BSE's securities communication, order routing, and execution system. [See BSE Rules, Ch. XXXIII.] The BSE rules provide for the guaranteed automatic execution of market and marketable limit orders from 100 shares up to and including 1,299 shares at the ITS/BBO. [See BSE Rules, Ch. XXXIII, Secs. 3 and 5(a). BSE specialists may alternatively provide an execution guarantee for orders of up to 2500 shares in specific stocks. See BSE Rules, Ch. XXXIII, Sec. 5(a). ] BEACON will expose a market or marketable limit order on the specialist's BEACON terminal for 15 seconds prior to automatically executing the order at the ITS/BBO to give the specialist the opportunity to improve the execution price of the order. [See BSE Rules Ch. XXXIII, Sec. 3(c). Moreover, a market or marketable limit order that would be executed outside of the primary market price range for the day in the stock will be manually handled by the specialist and receive an execution at the ITS/BBO or better as subsequent trades occur in the stock. See infra Part IV.B.2.b.]

2.Operation of the BSE's Competing Specialist Initiative

The BSE's CSI permits the appointment, in addition to theregular specialist, of competing specialists in stocks traded on the Exchange [Under the CSI, competing specialists have the same affirmative and negative market making obligations as regular specialists. However, the regular specialist remains responsible for coordinating all openings, reopenings, and trading halts to ensure that they are unitary.] and enables members to route order flow to a designated specialist for execution. A designated specialist may execute such order flow only if there are no limit orders then on the BSE central limit order book at the execution price and none of the other specialists are quoting at the ITS/BBO with time priority.

Under the CSI, BSE rules governing auction market principles of priority, parity, and precedence remain unchanged for quotes at the ITS/BBO. Specialist quotes that represent customer orders have priority over specialists' own quotes at the same price, and specialists quoting at the ITS/BBO have priority over specialists not quoting at the ITS/BBO. If two or more specialists are quoting at the ITS/BBO, the earliest bid/offer at that price has time priority and will be filled first up to its specified size. [This is in contrast to the CSE's preferencing program, which suspends time priority to allow preferencing dealers to have priority over same-priced professional interest, even at the ITS/BBO, when the dealer is interacting with public agency orders that it represents as agent. See supra Part I.B.3.] If the specialists are on both price and time parity at the ITS/BBO, all bids/offers equal to or greater than the size of the contra-side order are on parity and are entitled to precedence over smaller orders.

All limit orders transmitted to the BSE floor must be entered into BEACON, which maintains one consolidated limit order book for the Exchange and ensures that limit orders at the same price are kept in strict time priority, irrespective of routing designations. [Each specialist in a security has the ability to execute limit orders on the Exchange's consolidated limit order book through its BEACON terminal. ] Market and marketable limit orders routed through BEACON are automatically executed by the system against any contra-side orders on the consolidated limit order book. In accordance with the BSE rules governing BEACON noted above, however, before market and marketable limit orders are automatically executed by the BEACON system against contra-sided limit orders, they are exposed to the designated specialist for 15 seconds to give that specialist an opportunity to improve the price.

Under CSI rules, when there are no customer limit orders on the Exchange's consolidated limit order book at the ITS/BBO and none of the other specialists in the stock are quoting at the ITS/BBO with time priority, orders may be executed at the ITS/BBO or better by the designated specialist. Orders not directed to a particular specialist are automatically routed to the regular specialist for execution, except that the orders of a routing firm that is affiliated with a competing specialist are deemed to be designated to that member firm's affiliated specialist. This prevents member firms affiliated with a specialist from routing non-profitable orders to a non-affiliated specialist when marketconditions are unfavorable.

Until March 1997, a competing specialist had not been able to enter quotes directly into BEACON. Instead, the competing specialist had to orally communicate its quotes to the regular specialist, who would then enter the quotes into BEACON on the competing specialist's behalf. Further, BEACON presently routes orders that are not executed against BSE's consolidated limit order book to the designated specialist without systematically determining whether another specialist may be quoting at the ITS/BBO with priority. In addition, BEACON routes orders not designated to a particular specialist to the regular specialist, even though a competing specialist may be quoting at the ITS/BBO with priority.

To encourage competitive quoting by all specialists making markets in a stock, the BSE made a commitment to implement two BEACON systems enhancements. [See Letter from John I. Fitzgerald, Exec. Vice President, BSE, to Howard Kramer, Associate Director, SEC, dated February 29, 1996 (BSE agreement to complete system enhancements within one year from permanent approval of the CSI). See also Securities Exchange Act Release No. 36323 (September 29, 1995), 60 FR 52440 at n.26 (October 6, 1995) (File No. SR-BSE-95-14) (order extending BSE CSI pilot through March 29, 1996).] The first enhancement, implemented in March 1997, was to enable BSE competing specialists to enter quotes directly into BEACON. The second enhancement was to reprogram BEACON to route incoming orders to the specialist with priority on the Exchange at the ITS/BBO, or if no such priority has been established, to the designated specialist. The BSE has represented to the Commission that it will need additional time to

implement the second enhancement due to significant programming obstacles.

The Commission believes that the BSE's quote performance should improve as a result of its implementation of quote entry capability for competing specialists. In addition, the Commission will work closely with the BSE to ensure that the automatic routing of orders to the specialist at the ITS/BBO with priority is implemented as soon as practicable.

3.The Commission's Permanent Approval of BSE's Competing Specialist Initiative in 1996 [The BSE's CSI initially was approved on a one-year pilot basis in 1994. See Securities Exchange Act Release No. 34078 (May 18, 1994), 59 FR 27082 (May 25, 1994) (File No. SR-BSE-93-12). In 1995, the Commission granted approval to two extensions of the BSE's pilot program. See Securities Exchange Act Release Nos. 35716 (May 15, 1995), 60 FR 26908 (May 19, 1995) (File No. SR-BSE-95-07) (extension of BSE CSI pilot through October 2, 1995); and 36323 (September 29, 1995), 60 FR 52440 (October 6, 1995) (File No. SR-BSE-95-14) (extension of BSE CSI pilot through March 29, 1996 with clarification that BEACON limit orders are to be executed in strict time priority, allowing up to four competing specialists per stock, and increasing to 100 the maximum number of stocks in which each competing specialist could register). For a detailed discussion on the BSE's CSI pilot, see Appendix B.]

After careful review of the BSE's competing specialist program, including da ta submitted by the BSE and other sources, [See BSE, Competing Specialist Initi ative Report, submitted to the Commission on February 13, 1995 ("BSE February 19 95 Data and Report"); Letter from Karen Aluise, Assistant Vice President, BSE, t o N. Amy Bilbija, Attorney, SEC, dated April 28, 1995 ("BSE April 1995 Data"); Letter from Karen Aluise, BSE, to Glen Barrentine, Senior Counsel, SEC, dated Fe bruary 14, 1996 ("BSE February 1996 Data"). See also Letter from Robert Jenni ngs, Faculty Fellow and Professor of Finance, Indiana University School of Business, to Jonathan G. Katz, Secretary, SEC, dated June 30, 1995; and Letter from Robert Battalio, Asst. Professor, University of Notre Dame, to Jonathan G. Katz, SEC, dated March 6, 1996 (collectively, "IU Study").] the Commission determined in March 1996 thatpermanent approval of the BSE's program was consistent with the Act and the rules and regulations thereunder applicable to a national securities exchange. [See BSE Approval Order, supra note . The Commission received four comment letters on the BSE's proposal, two of which asserted that preferencing programs denied orders the benefits and protections of auction market trading, the other two included preliminary drafts of the IU Study. See IU Study, supra note . For a more detailed discussion of the IU Study, see Appendix A at note 29. ] In particular, the Commission believed that the program was consistent with Sections 6(b)(5) and 11A of the Act. The Commission also believed that while the BSE's program may increase internalization, it was not necessarily inconsistent with a broker-dealer's obligation to seek best execution of customer orders. [See infra Part III.C.1 for a detailed discussion of the Commission's rationale for concluding that the BSE's CSI was not necessarily inconsistent with best execution of customer orders.] In this regard, the Commission noted that the BSE's program provides incoming market or marketable limit orders with an opportunity to interact before a specialist can execute the orders against itself, while also providing an opportunity for price improvement for such orders before they are automatically executed. [These were the same enhancements that the BSE committed to implement pursuant to the SEC's permanent approval of the CSI. See supra note and accompanying text. Further, the Commission stated that the BSE program had enhanced order interaction on the BSE by increasing the volume of limit orders sent to the Exchange.]

The Commission noted that the data provided by the BSE during the course of the pilot indicated that trade and share volume for the Exchange overall increased during the CSI pilot. [The BSE reported that while for the month of December 1994 there were 171,075 trades and 106,753,284 shares traded on the Exchange, these numbers increased to 195,272 trades and 120,665,485 shares for December 1995. See BSE February 1995 Data and Report, supra note . The BSE also reported that the number of reports for all BSE trades to the Consolidated Tape Association also increased slightly since the beginning of the CSI. See BSE April 1995 Data, supra note .] The data indicated that competing specialists had received and executed a substantial amount of order flow in CSI stocks. [For example, the BSE's report for the last nine months of 1995 indicated that orders directed to a competing specialist accounted for 46% of total trades and 32% of total shares executed in CSI issues in April 1995, and increased steadily to 58% of total trades and 43% of total shares by December 1995. See BSE February 1996 Data, supra note . ] Moreover, the data showed that the depth of the BSE's consolidated limit order book in CSI stocks generally increased during the pilot. [Approximately 25% of the orders directed to competing specialists were limit orders. BSE April 1995 and February 1996 Data, supra note . In addition, between January - December 1995, between 12% and 21% of the orders directed to a competing specialist were executed against limit orders on the Exchange's consolidated limit order book. Id.]

In addition, the Commission stated its belief that the BSE's CSI was reasonably designed to facilitate competition among BSE specialists. In this regard, the Commission noted that a specialist could not execute directed order flow against itself if a competing specialist was quoting at the ITS/BBO, thus providing an incentive for specialists desiring to attract order flow to enter competitive quotes.

The Commission recognized, however, that the data was mixed in regard to whether the CSI had increased competition on the BSE. In this regard, BSE data had indicated that regular specialists executed less than 1% of the orders directed to a competing specialist. [See BSE April 1995 and February 1996 Data, supra note (data indicating regular specialist intercepts less than one percent of the order flow directed to competing specialists). ] The rules applicable to the CSI would have allowed regular specialists to execute a higher percentage of such orders if they had been quoting aggressively at the ITS/BBO. On the other hand, BSE data also indicated that the CSI may contribute to BSE's competitiveness within the NMS; specifically, during November 1995 BSE quotes matched at least one side of the ITS/BBO more often than in a comparable sample of BSE-traded issues in which there was a unitary specialist. [See Letter from Karen Aluise, Assistant Vice President, BSE, to Glen Barrentine, Senior Counsel, SEC, dated March 5, 1996.]

While the Commission noted that the data collected during the pilot indicated a lack of quote competition, the Commission anticipated greater quote competition once BEACON system enhancements were completed that would enable competing specialists to enter their own quotes directly into BEACON and provide for the routing of orders either to the specialist with priority at the ITS/BBO or if no specialist is at the ITS/BBO, to the designated specialist. [See supra note and accompanying text.]

For these reasons, the Commission determined to permanently approve the BSE's preferencing program. The Commission also removed the restrictions placed on the BSE CSI program during the pilot, [Specifically, these restrictions limited to four the number of competing specialists in each stock and to 100 the number of stocks in which each competing specialist could trade and prevented specialists from making cash payments for order flow in stocks in which they were registered as competing specialists.] concluding that such restrictions, which were put into place to allow the Commission to evaluate the effects of competing specialists on the BSE and the NMS, were no longer necessary.

II.Preferencing and Internalization

The NYSE's proposal for competing specialists, the CSE's preferencing program, and the BSE's CSI are three examples of programs that fit the definition of "preferencing" found in the NSMIA. The CSE and BSE programs, however, are not anomalous in the national market system. Indeed, many forms of preferencing have existed for quite some time across the various market centers. To understand the effect of the CSE and BSE programs, it is important to view them in the broader context of preferencing and internalization in the national market system and the various other practices by which a "broker [takes] priority in execution over same-priced orders or quotations entered prior in time."

As was noted in the Introduction, in its broadest sense, the term preferencing refers to the direction of order flow by a broker-dealer to a specific market maker or specialist, independent of whether or not some form of affiliation or inducement for the direction of order flow exists between the broker-dealer and the market maker or specialist. The definition of preferencing in the NSMIA appears to include only directed order flow that is internalized, [As was stated in the Introduction, internalization is the direction of order flow by a broker-dealer to an affiliated specialist or order flow executed by that broker-dealer as market maker. See supra note .] and to exclude the myriad of other arrangements to direct order flow to a particular market center.

The Act, the rules and regulations promulgated thereunder, and the rules of the various exchanges enable specialists to tradeunder certain circumstances as dealer with customer orders. [See supra note . ] Consistent with this regulatory framework, broker-dealers have established various means to internalize their customer order flow on exchanges, of which the CSE's preferencing program and the BSE's CSI are only two examples. Moreover, a number of large broker-dealer firms internalize their customer order flow in Rule 19c-3 securities, [See infra notes - and accompanying text.] third market makers internalize customer order flow in listed stocks, and internalization is commonplace in the over-the-counter market.

In addition, while the CSE's preferencing program may result in pre-existing customer interest that is represented in a preferencing dealer's quote at the ITS/BBO being bypassed by orders directed to other preferencing dealers, such a result is not unique to this program. In this regard, a number of practices under the rules of exchanges allow broker-dealers representing proprietary and customer orders to receive executions of such orders while preexisting customer interest in the trading crowd or the limit order book is bypassed.

The following discussion provides an overview of some of the means by which specialists or dealers can receive preferenced order flow, internalize customer order flow, and how broker-dealers can utilize exchange trading rules to bypass customer interest entered on the exchange prior in time.

A.Internalization on National Securities Exchanges

Specialists on all the national securities exchanges actively trade for their own account as dealer. The level of a specialist's participation as dealer in any particular stock by volume traditionally has been measured by two methods. The first method measures specialist trading in each stock by dividing the sum of a specialist's purchases and sales in the stock by twice the trading volume in the stock effected on the exchange. The resulting figure is commonly referred to as the twice total volume ("TTV") rate. [As the specialist can participate as dealer on only one side of each transaction, the maximum TTV rate is 50 percent.] The second method, referred to as the specialist participation rate ("SPR"), represents the total percentage of a stock's volume in which the specialist participates as either buyer or seller, and can be calculated by doubling the TTV rate.

In 1996, NYSE specialists participated as dealers at a TTV rate of 9.0% and a SPR of 18%, while Amex specialists participated as dealers at a TTV rate of 12% and a SPR of 24%. [In 1995, NYSE specialists had a TTV rate of 8.6% and a SPR of 17.2%. See NYSE 1995 Fact Book (May 1996) ("NYSE 1995 Fact Book") at 19-20. For the same year, Amex specialists had a TTV rate of 10.7% and an SPR of 22.14%. See 1996 Amex Fact Book ("1996 Amex Fact Book") at 20. ] The TTV rate and SPR of both the primary and the regional exchanges for 1996 were as follows:

Table II-1:Rates of Specialist Dealer Participation on Primary and Regional Exchanges - 1996

Exchange	Total Specialist Trading Volume	Total Exchange Trading Volume	TTV Rate (%)	SPR (%)
Amex	1,353,161,825	5,628,187,680	12.0	24.0
NYSE	18,881,674,879	104,636,179,522	9.0	18.0
BSE	1,055,180,219	1,457,880,354	36.2	72.4
CHX	1,867,381,236	3,927,600,302	23.8	47.5
CSE	1,703,855,065	2,091,037,631	40.8	81.5
Phlx	843,838,284	1,399,138,119	30.2	60.3
PSE	2,544,648,236	3,250,052,853	39.2	78.3

As the foregoing illustrates, specialist dealer activity is commonplace on both the primary and regional exchanges, though significantly more prevalent on the regional exchanges as measured by the percentage of trading volume effected by regional specialists. However, the amount of specialist dealer activity on the NYSE on an absolute share basis exceeds the total amount of shares transacted on all of the regional exchanges combined. In 1995, for example, NYSE specialists' purchases and sales amounted to approximately 15.05 billion shares, whereas the total combined regional exchange trading volume in NYSE-listed issues amounted to 10.61 billion shares. [Derived from data found in NYSE 1995 Fact Book, supra note , at 20 and 27. ]

Trading as dealer can be very profitable to specialists. It enables them to capture all or a portion of the spread between the bid and ask price for a stock. The more order flow a specialist attracts, especially retail orders which do not strain a specialist's liquidity, the more profitably a specialist can perform its dealing activity. Hence, all specialists, both on the primary and regional exchanges, attempt to attract as much retail order flow as possible. One of the most direct methods to accomplish this is for a broker-dealer to internalize order flow by routing customer order flow to an affiliated specialist unit. [In this context, an affiliated specialist unit is a unit in which a broker-dealer firm has a proprietary interest.] While the precise arrangements between broker-dealer firms and affiliated specialist units may vary across exchanges and firms, [See infra Part IV.A for more detailed descriptions of several existing affiliations between broker-dealers and specialist units.] such arrangements can result in significant benefits for the integrated firm as a whole and the affiliated specialist unit in particular. [It should be noted that various exchanges have rules pertaining to the adoption, maintenance and enforcement of Chinese Wall procedures to establish functional separation between upstairs firms and their affiliated specialist units. See , e.g. , NYSE Rule 98; Amex Rule 193; BSE Rules, Ch. 2, Sec. 36; and CSE Rule 5.5.] In this regard, a retail firm is likely to route most of its order flow in a particular security to the exchange on which it has an affiliated specialist unit dealing in the security. This directed order flow facilitates the marketmaking and, accordingly, the profitability of the specialist unit. [It is likely that a specialist will trade as dealer with orders represented by an affiliate at the same rate as that of its overall specialist participation on the exchange.]

So-called "upstairs" broker-dealers have affiliated specialist units on both the primary and regional exchanges. With regard to the primary exchanges, of NYSE's 38 specialist units, ten units, or 26.3%, are affiliated with upstairs broker-dealers. The ten affiliated specialist units on the NYSE trade 1061 of the 3308 securities traded on the NYSE (32.1%). As for the Amex, of its 13 specialist units, six units, or 46.2%, are affiliated with upstairs broker-dealers. These six affiliated units trade 403 of the 896 securities traded on the Amex (45%).

As the following table indicates, affiliated specialist units are even more prevalent on some of the regional exchanges than on the primary exchanges:

Table II-2:Affiliated Specialist Units on the Primary and Regional Exchanges - February 1997

Exchange	Number of Specialist Units	Number of Affiliated Specialist Units	Percentage of Affiliated Specialist Units	Number of Stocks Traded on Exchange	Number of Stocks Traded by Affiliated Specialist Units	Percentage of Stocks Traded by Affiliated Specialist Units
Amex	13	6	46.2	896	403	45.0
NYSE	38	10	26.3	3308	1061	32.1
BSE	18	12	66.7	2221	1684	75.8
CHX	33	6	18.2	2702	363	13.4
CSE [As the CSE utilizes a competing dealer system, each stock is counted only once, regardless of the number of designated dealers that trade a particular stock.]	13	8	61.5	515	456	88.5
Phlx	22	5	22.7	2768	740	26.7
PSE [PSE specialists are not organized into specialist units. Instead, the PSE has 82 specialist posts, distinct for allocation and review purposes, which are operated by 24 specialist firms. For purposes of this analysis, the 24 firms are considered as equivalent to specialist units.]	24	16	66.7	2550	2100	82.4

Table II-3:Affiliated Specialist Units on the Primary and Regional Exchanges - 1994 [Division of Market Regulation, SEC, Market 2000: An Examination of Current Equity Market Developments ("Market 2000 Study") (January 1994), at Exhibit 29.]

Exchange	Number of Specialist Units	Number of Affiliated Specialist Units	Percentage of Affiliated Specialist Units	Number of Stocks Traded on Exchange	Number of Stocks Traded by Affiliated Specialist Units	Percentage of Stocks Traded by Affiliated Specialist Units
Amex	15	3	20.0	958	327	34.1
NYSE	40	3	7.5	2089	259	12.4
BSE	19	11	57.9	1863	1681	90.2
CHX	42	13	31.0	2306	582	25.2
CSE [See supra note .]	12	7	58.3	543	500	92.1
Phlx	26	4	15.4	2047	545	26.6
PSE/LA [See supra note .]	19	13	68.4	1758	1213	69.0
PSE/SF [Id.]	17	13	76.5	1758	1458	82.9

Since 1994, the total number of specialist units on all of the exchanges combined has decreased, while the total number of affiliated specialist units on the exchanges has increased. The increase in the number of affiliated specialist units is especially marked on the primary exchanges. The overall number of stocks traded by affiliated units on the exchanges has increased, although the overall percentage of stocks traded that are traded byaffiliated units on the regional exchanges has decreased slightly. [This is due to the sizeable growth in the number of issues traded on the exchanges since 1994.] However, both the number and percentage of stocks traded by affiliated specialists on the primary exchanges has increased substantially.

Though the gap has closed significantly, on an overall percentage basis affiliated specialist units continue to be more prevalent and account for a higher percentage of stocks traded on the regional exchanges than on the primary markets. The nature of the trading environment on the regional stock exchanges provides significant incentives for large broker-dealer firms, particularly NYSE member firms, to purchase specialist units on these exchanges and direct their order flow to such affiliated specialists for execution. Specifically, the lower cost of obtaining a regional exchange membership and operating a regional exchange specialist unit, as well as the lower execution fees on the regional exchanges, has facilitated the purchase of regional specialist units by upstairs firms. The directed order flow from the upstairs firms to the affiliated specialist units consists mainly of small retail customer orders. [See infra Part IV.A.] Such order flow provides the affiliated specialist unit with a stream of orders that are easy to handle and which bolster the overall order flow of the specialist. Accordingly, the firm has the opportunity to make markets on a large percentage of these trades.

Moreover, off-board trading restrictions of the primary exchanges provide an additional incentive for the members of these exchanges to purchase regional specialist units. [See Market 2000 Study, supra note , Study III at III-10.] For example, NYSE Rule 390 prevents NYSE members from effecting certain transactions in NYSE-listed securities off of an exchange. [See NYSE Rule 390. See also Amex Rule 5. With the exception of the CSE, each of the regional exchanges also has off-board trading restrictions. See BSE Rules, Ch. 2, Sec. 23; CHX Rules, Art. VIII, Rule 9; PSE Rule 5.43; and Phlx Rule 132. ] The Commission has narrowed the scope of Rule 390 through the adoption of Rule 19c-3 under the Act, [17 CFR 240.19c-3.] which permits NYSE members to execute proprietary trades off of an exchange in securities listed after April 26, 1979 ("Rule 19c-3 securities"). [Prior to the adoption of Rule 19c-3, the Commission had narrowed the scope of exchange off-board trading restrictions by Rule 19c-1 under the Act, 17 CFR 240.19c-1, which enables exchange members to execute agency trades with a market maker who is not an exchange member.] On a practical basis, the purchase of a regional specialist unit allows a NYSE member firm to internalize its small retail order flow without violating the NYSE's off-board principal trading restrictions. [NYSE Rule 390. See also infra Part II.C for a discussion on NYSE member firms' internalization of their customer order flow in Rule 19c-3 securities and non-NYSE member firms' third market trading.]

In addition to broker-dealer proprietary relationships with affiliated specialist units, broker-dealers have entered into a variety of contractual joint venture arrangements with specialistfirms on most of the regional exchanges. In a typical joint venture arrangement, a broker-dealer will agree to route its small retail customer order flow to a primary or regional exchange specialist in return for a guaranteed percentage share of the specialist unit's trading profits. Additionally, in some joint venture arrangements, the broker-dealer also may agree to provide capital to the specialist unit in return for a percentage of revenues. [See infra Part IV.A. ] As a result, joint venture arrangements are functionally equivalent to affiliations through an actual proprietary relationship between the broker-dealer and the specialist unit. The following table illustrates the extent of joint venture arrangements on the primary and regional exchanges:

Table II-4: Joint Venture Specialist Units on the Primary and

Regional Exchanges - February 1997

Exchange	Number of Specialist Units	Number of Joint Venture Units	Percentage of Joint Venture Units	Number of Stocks Traded on Exchange	Number of Stocks Traded by Joint Venture Units	Percentage of Stocks Traded by Joint Venture Units
Amex	13	0	0.0	896	0	0.0
NYSE	38	0	0.0	3308	0	0.0
BSE	18	2	11.1	2221	279	12.6
CHX	33	3	9.1	2702	169	6.3
CSE [See supra note .]	13	0	0.0	515	0	0.0
Phlx	22	4	18.2	2768	119	4.3
PSE [See supra note .]	24	1	4.2	2550	250	9.8

B.Bypassing of Time Priority on Exchanges

1.Rules of Priority, Parity, and Precedence

Section 510(c)(3) of the NSMIA defines preferencing to include practices when a dealer takes priority over same-priced orders or quotations entered prior in time. As discussed below, there are different ways on the various exchanges in which priority is bypassed. Each exchange has rules of priority, parity, and precedence that determine the bids and offers that participate in a transaction and the sequence in which they do so. [See Amex Rule 126; BSE Ch. II, Sec. 6; CHX Article XX, Rules 14-18; CSE Rule 11.9; NYSE Rule 72; Phlx Rules 118-121; and PSE Rule 5.8. ] For those exchanges utilizing a physical trading floor, [With the exception of the CSE, each of the national securities exchanges uses a physical trading floor. ] once a trade in a particular stock occurs on an exchange, a new auction begins in that stock. All bids and offers must be reestablished, [There is an exception in the event that stock remains after the execution of the order. See infra note .] including any bid (offer) representing interest on the limit order book. Priority generally refers to the sequence of eligibility of bids and offers, while precedence determines the execution of bids and offers (i.e., orders with priority take precedence on the next sale at that price). Parity refers to a situation where there are simultaneous bids (offers) or an inability to determine time priority among bids (offers). Such bids (offers) are said to be atparity, or "on par" with one another. [It should be noted, however, that the NYSE and other exchanges utilize the term precedence to distinguish among bids at parity.]

Generally, priority and precedence among bids and offers are determined first by price (i.e., the bid (offer) at the highest (lowest) price has priority over bids (offers) at inferior prices, regardless of its time of entry, and is entitled to be executed first on the next sale). If there are two or more bids (offers) at the same price, priority generally is based on time of entry. The first clearly established bid (offer) at a superior price attains priority over subsequent bids (offers) at the same price, up to the size bid (offered). Because each trade closes the auction, all bids and offers that are reestablished are considered to be reentered at the same time, regardless of whether a bid or offer was previously announced but not executed. Thus, in this situation, a previously entered limit order will not take priority over a subsequently announced order in the crowd.

The exchanges take different approaches in determining which bids (offers) at parity are to be executed. Under the rules of the NYSE, BSE, and Phlx, once the bid (offer) having time priority has been executed, or when for some reason no bid (offer) has time priority, precedence is determined by the size of the bids (offers) in question ("precedence based on size" or commonly, "sizing out"). [Precedence based on size applies in two specific situations. First, any bid (offer) equalling or exceeding the amount of shares in an incoming contra-side order has precedence over all other bids (offers) at that same price and is entitled to the first execution at that price. If more than one bid (offer) equals or exceeds the size of the incoming order, all such bids (offers) are at parity. If the time sequence of their entry can be clearly determined, the bid (offer) made first in time is entitled to precedence, and will be filled. Second, if all bids (offers) are for amounts less than the size of the incoming order, precedence is accorded to the largest-sized bid (offer), and if any balance remains, the second largest-sized bid (offer) then is entitled to precedence, and so forth. If any two bids (offers) are for the same size and, if the time sequence of their entry can be clearly determined, they will be filled in that order. Otherwise, such bids (offers) will be on parity.] The Amex, CHX, and PSE do not utilize precedence basedon size, with an exception in the case of the Amex for orders to cross 25,000 shares or more. [One or both of the orders must be for the account of a member or member organization. See Amex Rule 126.01. See also infra notes - and accompanying text (Amex "clean cross" rule). ] Under the rules of these

exchanges, if no bid (offer) has price or time priority, all bids (offers) at the best price are at parity, regardless of size. [Pursuant to Amex Rule 126, a specialist who is on parity with the crowd and has an accumulation of agency orders on his or her book is entitled to a specific percentage of transactions that take place at the bid or offer price: 60% when one broker is entitled to match; 40% when two to five brokers are entitled to match; and 30% when six or more brokers are entitled to match. Until the Commission's recent approval of its codification into Amex Rule 126, this procedure was applicable specifically to openings pursuant to Amex Rule 108, but informally was applied throughout the trading day. See Securities Exchange Act Release No. 38238 (February 4, 1997), 62 FR 6591 (February 12, 1997) (File No. SR-Amex-96-39).] Whether or not an exchange utilizes precedence based on size, in the event there is no price or time priority and size precedence, bids (offers) at parity "match." In order to determine which bid (offer) is entitled to execution against an incoming order, those matching would "flip a coin (or match coins)," thewinner being entitled to the order up to the size of its bid, the loser receiving any remaining amount. Alternatively, the rules of most exchanges allow the members on parity to agree to split the lot of stock represented by the incoming order among themselves. In practice, this is what usually occurs. However, if any member of the trading crowd objects to a split, the rules of some exchanges require the members on parity to match for the lot.

2.Use of Precedence Based on Size to Bypass Time Priority

The practice of precedence based on size has a significant effect on the sequence in which orders are executed. Precedence based on size allows members with the largest-sized bid (offer) to receive an execution ahead of same-priced market interest that

otherwise would be on parity. The NYSE has justified this practice on the grounds that:

in the NYSE market it is often a practical impossibility to keep track of successive time priorities. To keep the market moving, giving bids and offers precedence based on size is often a practical necessity. [NYSE Floor Official Manual (June 1996) ("NYSE Floor Official Manual") at 38.]

At the same time, it should be recognized that "sizing out" confers a significant advantage to members representing large orders. In this regard, "sizing out" allows for such orders to receive an execution ahead of smaller market and marketable limit orders routed to the exchange. In addition, as all bids and offers must be reestablished after a trade, including the limit order book,

"sizing out" allows a broker-dealer to trade ahead of preexisting i nterest on the limit order book. [Under NYSE Rule 108.10, the specialist can com bine a bid (offer) for his or her own account with orders on the book to take precedence based on size; if the specialist is establishing or increasing a position, he or she must ask other members in the crowd if they have public orders at that same price. If so, the specialist must fill that member's order before he or she may retain any stock. The specialist does not have to give up stock to the crowd if it is liquidating a position. It also should be noted that when the limit order book has priority, precedence among orders on the book at the same price is accorded strictly with regard to their time of entry on the book. There are a few exceptions to this rule: (1) when a day limit order is changed to a good-til-cancelled ("GTC") order, it is considered to be a new order; (2) the simultaneous election of percentage orders results in parity; and (3) if there is a change in the last sale price which alters the lowest price at which a "short" order may be executed, the short order is treated as a new order. See NYSE Rules 123A.20; .30; and .71. See also infra Part II.B.6 for a detailed discussion of percentage orders.] For example, assume the limit order book consists of five 100 share customer orders (entered earlier in the day) bid at 20. A transaction occurs in which 100 shares of the book are executed against a 100 share offer. The transaction clears the floor, and the following bids are entered simultaneously: the specialist bids 400 at 20 on behalf of the book (consisting of the remaining four 100 share limit orders), Broker A bids 400 shares at 20, and Broker B bids 1000 at 20 on behalf of an order it had just received from a non-affiliated broker-dealer. If an offer for 1000 shares at 20 is made, Broker B would get the stock, being the only bidder equalling the offer size. Pursuant to precedence based on size, it would be irrelevant that the four limit orders at 20 were received on the exchange atan earlier time. [See infra text accompanying notes - .]

3. Crossing Orders

Each of the exchanges has rules permitting a broker-dealer to execute an ord er to buy and an order to sell against one another at a particular price (a &quo; t;cross" trade). These rules permit cross trades to occur at or between an exchange's currently disseminated quotation or outside of the quotation. In pr actice, however, the overwhelming majority of crosses occur at or between an exc hange's quotation. In addition, each of the exchanges' cross trading rules requ ire, with the exception of the CSE, [As discussed above, however, Interpretation and Policy .01 to CSE Rule 11.9(u) requires that in greater than minimum variat ion markets CSE preferencing dealers must either immediately execute a market or der at a better price than the ITS/BBO or expose the order either in its CSE quo te or the CSE central limit order book at an improved price for at least 30 seconds prior to executing such orders at the ITS/BBO. See supra note and accompanying text.] that prior to executing a cross, a member must first make a public bid and offer on behalf of both sides of the cross, offering at a price one minimum variation higher than its bid. [Alternatively, the member may bid one side of the cross at a price one minimum variation lower than its offer. These rules provide that if a member is attempting to cross a customer order with itself ("agency-principal cross"), the customer's side of the cross must be bid (offered) at the minimum variation lower (higher) than the cross price. See , e.g. , Amex Rule 152(a)(2)(i) and (ii); NYSE Rule 91. Further, such rules require a member who is attempting an agency-principal cross to ensure that the cross price is justified by the condition of the market and the customer is promptly notified of the cross trade and accepts the trade. See , e.g. , Amex Rule 151(a)(2)(iii) and (iv); NYSE Rule 91.] However, most of the exchanges also have rules designed to facilitate the execution ofcrosses without interference by other interest. In this regard, the primary and regional exchanges generally take slightly different approaches: (1) the regional exchanges restrict the ability of particular exchange members to interfere with attempted crosses between (and in some cases at) the exchange's disseminated bid and offer; and (2) the primary exchanges accord priority at the cross price to attempted crosses of orders of 25,000 shares or more where neither order is for the account of a member or member organization ("clean cross" rules). [See infra notes - and accompanying text.] As detailed below, in some cases, these practices may facilitate the internalization of customer orders on the regional exchanges and the bypassing of pre-existing customer orders at the attempted cross price on the primary markets.

As a means of increasing their ability to compete with each other and the pr imary exchanges for cross trade order flow, particularly with regard to block tr ades (i.e., trades of 10,000 shares or more), the regional exchanges have adopted rules to facilitate the execution of such crosses. [Indeed, the rules o f the PSE state that "[i]t is the policy of the Exchange that all efforts should be made to facilitate the execution of such crosses." See PSE Rule 5.14(b), C ommentary .01.] Specifically, these rules are intended generally to limit the ability of regional specialists and/or other exchange members to interfere with cross transactions priced at or between the quotation disseminated by theExchang e at the time the cross is announced. [See CHX Art. XX, Rule 23; Phlx Rule 126; and PSE Rule 5.14(b).] At the same time, the regionals' crossing rules do not grant priority, parity, or precedence to the order of a member over customer in terest at the cross price that is either represented in the specialist's book or by another member; as a result, these rules are consistent with Section 11(a) o f the Act and the rules promulgated thereunder. [15 U.S.C. 78k(a); 17 CFR 240.11 a-1 et seq. ] Given the lack of limit orders on the regional specialists' b ooks, [See , e.g. , Securities Exchange Act Release Nos. 15770 (April 26, 1979), 44 FR 26692 (File No. S7-32-92) (Rule 11Ac1-3 Proposing Release); 31343 (Octobe r 21, 1992), 57 FR 48645 (October 27, 1992) (File No. SR-NYSE-90-39); 34089 (May 19, 1994), 59 FR 27301 (May 26, 1994) (File No. SR-Amex-92-41); Market 2000 Stu dy, supra note , Study II, at II-9 to II-10.] the effective absence of trading crowds on the regionals, [Id.] and limitations on the ability of specialists a nd other members to interfere with crosses brought to these exchanges, the regio nal exchanges have created facilities that allow for the efficient routing and e xecution of cross transactions without a significant likelihood that such transa ctions will be broken up by other interest. [Indeed, the regional exchanges' suc cess in competing for block cross executions led to the adoption of the "clean c ross" rules by the NYSE and Amex discussed below, and the Amex's adoption of siz e precedence for agency-principal/professional or principal-principal crosses of 25,000 shares or more. See infra notes - and accompanying text. ] Moreove r, due to the absence of intermarketlimit order priority or price protection, [I n 1978, the SEC requested the SROs to take joint action to develop a central lim it order file that would maintain price and time priority for public limit order s, regardless of the market on which they were entered. See Securities Exchang e Act Release No. 14416 (January 26, 1978), 43 FR 4354. In 1979, the SEC determ ined that because affording absolute time priority to limit orders could have "a radical and potentially disruptive impact on the trading process as it exists t oday," SEC and industry efforts should be concentrated on achieving intermarket price protection for all public limit orders. See Securities Exchange Act Rele ase No. 15671 (March 22, 1979), 44 FR 20360. Later that year, the SEC proposed R ule 11Ac1-3, which would have required any broker-dealer executing a transaction in a covered security at a price inferior to the price of any displayed public limit order to satisfy those limit orders either simultaneously with, or immediately after, such execution. See Securities Exchange Act Release No. 15770, supra note . Due to a lack of interest from the markets, the partial achievement of intermarket price protection through the adoption of the ITS trade through and block protection policies, and potential difficulties in implementing a system for intermarket price protection, the SEC withdrew the proposal in 1992. See Securities Exchange Act Release No. 31344 (October 21, 1992), 57 FR 48581 (October 27, 1992) (File No. S7-32-92) (Rule 11Ac1-3 Withdrawal Release). The ITS trade-through policy requires ITS participants to avoid initiating a purchase (sale) of an ITS security at a price that is higher than (lower than) the price at which the security is being offered (bid for) on another participating ITS market center. With various exceptions, the ITS Plan provides remedies for an aggrieved party that makes a timely and valid complaint. The ITS block trade policy, a specific application of the trade-through policy, provides that a member representing a block-sized order(s) shall, at the time of execution of the block, send through ITS to each participating ITS market center displaying a bid (offer) superior to the execution price, a commitment to trade at the execution price and for the number of shares displayed with that market center's better-priced bid (offer). ] block positioning firms and broker-dealers seeking to cross trades can be assured that crosses routed to a regional exchange will not be broken up despite pre-existing limit orders on the books of the otherexchanges, particularly the primary markets, where limit order books are generally much deeper than those of regional exchanges. [The converse holds true in some instances ( i.e. , a broker- dealer "shopping" a proposed cross may bring the cross to the NYSE or Amex to avoid interest on a regional exchange's limit order book).]

In addition, the crossing rules of the regional exchanges also provide an incentive for proprietary matching networks developed by several broker-dealers [E.g. , Investors Liquidity Network; Lattice Network.] to route agency and principal non-block sized crosses to these exchanges for execution. [See Market 2000 Study, supra note , Study II, at II-10.] These networks match customer and broker-dealer orders entered into the network either against each other or against the network's parent firm. [Some of these proprietary crossing networks even allow other broker-dealers to enter their customer orders into the network.] Crosses generated by these networks are priced either at or between the ITS/BBO and are sent to a predetermined regional exchange for execution. The crossing rules and other characteristics of regional exchange trading cited previously permit most of these cross trades to be executed without being interfered with by other interest. [Indeed, the Market 2000 Study referred to "the willingness of regional specialists to refrain from interfering in [crosses from such networks]." Id.] Accordingly, a firm routing a network-generated cross trade to a regional exchange can represent to its customer that the transaction received an exchangeexecution, and the exchange will receive a print for transaction reporting purposes. [The exchanges, as participants in the Consolidated Tape Association ("CTA"), derive income from the fees charged to CTA vendors and subscribers, which is divided based on each participant's annual share of last sale transactions in CTA securities as a percentage of such transactions by all participants. This provides each exchange with an incentive to maximize the number of trades conducted through its facilities.]

The primary markets also have procedures to facilitate block crosses on their floors. On the exchanges that utilize precedence based on size, such as the NYSE, [The Amex (only for crosses of orders of 25,000 shares or more), BSE, and Phlx also utilize precedence based on size. Amex Rule 126.01 provides that agency-principal and principal- principal crosses may establish precedence based on size as long as the two orders were represented at the post when a sale removing all bids and offers from the floor takes place. Once precedence is established, the member handling the cross must bid (offer) one side of the proposed cross at a price one minimum variation higher than (lower than) the proposed cross price.] a member can prevent its attempted cross trades from being interfered with by other members and/or the limit order book by taking advantage of the practice of sizing out. For example, assume a market quoted 20 - 20 1/8, 2000 shares bid and 3000 offered. [See NYSE Working Paper 93-01 (1993) at 38.] Both the bid and offer consist of customer limit orders. If Broker A desires to cross 5000 shares at 20 1/8, he or she must bid 20 for 5000 shares and offer 5000 shares at 20 1/8. In order to bring its bid and offer to parity with the pre-existing orders at the quote, Broker A may purchase (on behalf of the buy side of the cross) 100 shares from the 3000 shares offered at 20 1/8. The trade clears the floor of priority,effectively placing Broker A's offer for 4900 shares at 20 1/8 and the book offer at 20 1/8 for 2900 shares on parity, as both would be considered entered simultaneously. Broker A may now assert precedence based on size, given that it has the only bid equal to the size of the offer, and may take his or her 4900 share offer at 20 1/8.

In a competitive response to the regional exchanges' attempts to attract block cross business away from the primary exchanges, the NYSE and Amex also adopted "clean cross" rules to facilitate the execution of block cross trades of 25,000 shares or more. [The NYSE and Amex originally had proposed to make the clean cross rule applicable to agency crosses of 10,000 shares or more. See NYSE Rule 72(b); Amex Rule 126.02. In response to the adoption of the NYSE and Amex provisions, the PSE subsequently adopted a substantially similar clean cross rule. See Commentary .05 to PSE Rule 5.14(b).] Specifically, these rules allow a member who has an order to buy and an order to sell 25,000 shares or more of the same security, where neither order is for the account of a member or member organization, [There is a further restriction in the case of the PSE that neither order can be for the account of a broker-dealer. Id.] to cross those orders at a price that is at or within the prevailing quotations without being broken up at the cross price, irrespective of pre-existing bids and offers at that price. [Although it appears that the clean cross rule is limited to individual customer orders of 25,000 shares or more, the NYSE has stated that a member firm can aggregate customer orders in order to take advantage of the clean cross rule. See NYSE Information Memo 96-32 (October 14, 1996); NYSE Floor Official Manual, supra note , at 180-181.]

Accordingly, the clean cross rules enable a member to cross large blocks, consisting almost exclusively of institutional orders, on the floor and to bypass pre-existing customer and professional interest on the limit order book at the proposed cross price, as the cross will have priority at that price. The clean cross rules do require that a member must bid or offer one side of the cross at an improved price prior to executing the cross. The cross may be broken up at this price, but only after all pre-existing interest at this price has been satisfied. The latter requirement has the result of greatly increasing the likelihood that the cross will be effected at the cross price without being interfered with by interest either in the book or in the trading crowd. [In 1995, there were 1,963,889 block transactions on the NYSE, accounting for 49.7 billion shares, or 57% of NYSE volume. The NYSE stated that 27% of its 1995 block volume was facilitated by upstairs firms, i.e. , brought to the floor as a cross trade. See NYSE 1995 Fact Book, supra note , at 17. The average size of a NYSE block transaction in 1995 was 25,307 shares, slightly above the minimum amount required for a clean cross transaction to take place. Id. However, previous studies have indicated that the percentage of upstairs-facilitated block volume increases with block size. In this regard, on one day in January 1993, the percentage of upstairs facilitated volume was 32% for blocks from 25,000 to 100,000 shares, and 57% for blocks over 100,000 shares. See NYSE Working Paper 93-01, supra note , at Table 5. As for the Amex, in 1995 there were 88,365 block transactions, accounting for 2.51 billion shares, or 49.4% of Amex volume. The average size of an Amex block transaction in 1995 was 28,370 shares. See 1996 Amex Fact Book, supra note , at 10.]

In approving the adoption of clean cross rules by the NYSE, the Commission recognized that such rules might not have been theideal means to address the situation prompting their adoption, i.e., the incentive for broker-dealers to effect block transactions on a regional exchange to completely avoid the limit order books and trading crowds on the primary markets. [See Securities Exchange Act Release No. 31343 (October 21, 1992), 57 FR 48645 (October 27, 1992) (File No. SR-NYSE-90-39) (order approving NYSE clean cross rule). See also Securities Exchange Act Release No. 34089 (May 19, 1994), 59 FR 27301 (May 26, 1994) (File No. SR-Amex-92-41) (order approving Amex clean cross rule).] Instead, the Commission believed that a preferable approach would have been to establish a means of intermarket limit order price protection. [See supra note .] As such a means was not available, the Commission reluctantly approved this initiative by the NYSE to circumvent time priority rules in order to compete with the regional exchanges for block orders.

In conclusion, the cross facilitation rules and unique characteristics of the trading environments of the various exchanges, coupled with the lack of limit order price protection among the market centers comprising the NMS, provide a broker-dealer with the ability to route its cross transactions to the exchange where the likelihood of the cross being interfered with is the lowest at any particular time (i.e., preference such trades to a particular exchange). As was demonstrated above, however, allowing broker-dealers' cross transactions to bypass other interest on the same or other exchanges may result in the loss of an opportunity for an execution by orders entered prior in time,both on the exchange where the cross is effected and on the other exchanges in the NMS. Given the absence of intermarket limit order price protection, the result is akin to preferencing as defined in the NSMIA, but on an intermarket scale.

4. "Stopping" Stock Rules

Aside from precedence based on size and the clean cross rules, there are mec hanisms by which a specialist or other professionals can trade ahead of pre-exis ting limit order or trading crowd interest. These include the practice of stopp ing stock, the use of marketable limit orders, and percentage orders. The pract ice of "stopping" stock refers to a guarantee by a specialist that an order received by the specialist will be executed at no worse a price than the prevailing bid (offer) when the order was received, with the understanding that the specialist will attempt to get a better price for the order than the prevailing bid (offer). ["Stopping" stock should not be confused with a stop order, which is an order designated as such by the customer that requires the specialist to buy (sell) a security once a certain price level has been reached. Once the market reaches that price, the stop order is "elected" and becomes a market order to buy (sell), with no guarantee by the specialist that the order will be executed at the stop price. For example, a sell stop order at 25 for 100 shares becomes a market order to sell 100 shares on the specialist's book when the stock trades at or below 25, and is treated by the specialist as any other order to sell stock at the market.] For example, assume a market quoted at 20 bid, 20 1/4 offered; the bid representing 1000 shares is for the specialist's own account, the offer of 3000 shares is made up of customer limit orders. If the specialist receives an order to buy 1000 shares, the specialist may stop the order at the offer (i.e., guaranteeing that thestopped order will not pay more than 20 1/4) "against the book," allocating limit orders on the book to fill the stopped order in the event a better price cannot be achieved. [The specialist can also grant a stop against its own account, but cannot attain priority over limit orders on his book by doing so.] The specialist will then change its quote to 20 1/8 - 20 1/4, 1000 shares bid and 3000 offered, the bid representing the stopped order. If a market order to sell 1000 shares subsequently reaches the post, the specialist will execute the sell order against the stopped order at 20 1/8, improving the price for the stopped order.

In this example, the sell limit order at 20 1/4 with priority on the book is bypassed and does not receive the execution it would have (against the stopped order) if the stop had not been granted. Further, if the market turns away from the limit price (e.g., moves to 20 - 20 1/8 or lower), the limit order ma y never be executed. For these reasons, the Commission historically has had mixe d reactions about the practice of stopping stock. [See Special Study, supra no te , at 150-154. Indeed, the Special Study recommended that specialists should be prohibited from stopping stock at any price at which a specialist holds an unexecuted customer order capable of execution at such price. Id. at 169. The NYSE subsequently amended its rules to restrict a specialist's ability to stop stock. Specifically, under the NYSE's amended rule, a specialist was permitted to stop stock: (1) in connection with an opening or reopening; (2) when a broker in the trading crowd represented another order at the stop price; (3) when the specialist did not have an executable order at the stop price; and (4) upon an unsolicited request of a broker in a greater than minimum variation market if the bid-ask spread was reduced, the size of the market was not reduced after the stop, and on the election of the stop, book orders with priority would be elected against the stopped stock. See Securities Exchange Act Release No. 7432, supra note ; NYSE Rule 116.30 (1965). The NYSE subsequently amended Rule 116.30 to delete exception (3), as it claimed that the exception is "clearly implied within the language of paragraph 116.30." See Securities Exchange Act Release No. 19839 (June 2, 1983), 48 FR 26579 (June 8, 1983) (File No. SR-NYSE-80-16). With regard to exception (4), SuperDot market orders are presumed always to have a "try to stop" request. See NYSE Working Paper 93-01, supra note , at n.90. ] As a result,until 1991 the practice was only allowed in markets where the bid-offer spread was at least twice the minimum variation because the harm to existing orders on the book was believed to be offset by the resulting reduced spread and possibility of price improvement for the stopped order. [See , e.g. , Securities Exchange Act Release No. 36399 (October 20, 1995), 60 FR 54900 (October 26, 1995) (File No. SR-NYSE- 95-14) (order permanently approving NYSE's pilot program for stopping stock in minimum variation markets). ]

The Commission in 1995 granted permanent approval to the pilot programs of various exchanges that permit the stopping of limited quantities of stock in minimum variation markets under certain specified conditions, the most important being the existence of a significant disparity between the bid and ask size (on the opposite side of the market from the order being stopped) that suggests the likelihood of price improvement far exceeds the possibility of harm to customers' orders on the limit order book. [ See , e.g. , Amex Rule 109; NYSE Rule 116.20-30.] For example, assume a market quoted 20 - 20 1/8, 1000 shares bid and 20,000 shares offered, with both the bid and ask representing customer limit orders, and the large imbalance on the offer side suggesting that subsequent transactions will be on the bid side. A specialiststopping an incoming buy order for 1000 shares at 20 1/8 against the book would be required to increase its bid to 2000 shares. If a market order to sell 2000 shares subsequently reaches the post, the specialist will execute it along with the limit orders at the bid against the stopped order at 20, improving the price for the stopped order. [Under the procedures pertaining to stopped stock in minimum variation markets, the existing limit orders at the bid would have priority and precedence for execution over stopped orders at the bid. See Securities Exchange Act Release No. 36399, supra note , at n.10.]

However, if another order to buy is executed at the offer before the stopped order is executed at the bid, the stopped order is executed at the stopped price. [Id.] As with the practice of stopping stock in general, it remains the case in minimum variation markets that the contra-sided limit order with priority on the book at the time the order was stopped continues to lose an opportunity for an immediate execution against the stopped order, and may never receive an execution if the market subsequently moves away from the limit order. [In the NYSE's fourth report to the Commission on its stopping stock in minimum variation markets pilot program, the NYSE stated that for stops in minimum variation markets, approximately one-third of the contra-side limit orders did not receive an execution by the end of the day, approximately 40% were subsequently cancelled, and approximately 30% were executed by the end of the day. See Id. ] Thus, in a primary market such as the NYSE, with limit orders often residing on the bid and offer side, a market order that is stopped usurps the time priority of pre-existing agency orders on the specialist's book.

Specialists stop stock as a means to provide price improvement. When the spread is greater than the minimum variation, the possibility of an execution between the spread provides an opportunity for price improvement. Price improvement in minimum variation markets, however, is problematic. Theoretically, with no ability to trade between the spread, price improvement can only occur by bypassing the interest represented on the contra-side of the market. If the contra-side interest consists solely of the specialist's quote, then it may be appropriate to stop market orders and give them a better price than if executed against the quote, because the specialist is giving up his own quote to advantage a customer order. However, if a stopped order is executed only because the specialist has refrained from executing the limit order entrusted to him, the limit order is denied an execution. The Commission believes that the practice of stopping market orders and executing them in advance of previously-entered limit orders seriously disadvantages the limit orders and should be reconsidered in the context of minimum variation markets. [See also infra Part V.D. (discussion evaluating results of analysis of price improvement in minimum variation markets).]

5.Use of Marketable Limit Orders to Provide Price Improvement in Minimum Variation Markets

In analyzing the rates of price improvement provided by the various exchanges for purposes of the study, the Commission has found a pronounced disparity in the rates of price improvement received by market orders versus that received by marketable limitorders. [See infra Part V.D. and Tables V-8A to V-10C.] This disparity is prevalent across exchanges and markets. A primary cause of this disparity may be the use of marketable limit orders by specialists to provide price improvement to subsequently arriving contra-side orders. [There may other reasons that account for a portion of this disparity. Certain regional exchange automated execution protocols do not provide the opportunity for price improvement to marketable limit orders. For example, the CHX's SuperMAX and Enhanced SuperMAX price improvement algorithms are only applicable to market orders. See CHX Rules, Art. XX, Rule 37. Alternatively, specialists on both the primary and regional exchanges may simply execute marketable limit orders at their limit price.] In minimum variation markets, this practice deprives preexisting customer limit orders represented in both sides of the specialist's quote from the opportunity of receiving an execution, and calls into question whether the specialist is fulfilling his fiduciary duties with regard to the orders on the limit order book. [Further, the Commission notes that the practice results in the avoidance of time priority and precedence in execution among limit orders on the exchange. The use of marketable limit orders to provide price improvement in greater than minimum variation markets raises the same issues with regard to the bypassing of contra-sided book orders as discussed in the previous section on stopping stock. See supra Part II.B.4.] Therefore, the Commission has serious concerns as to whether marketable limit orders should be used to provide price improvement in minimum variation markets.

For example, assume a market 20 - 20 1/8, 200 bid and 1000 offered, the bid and the offer representing customer limit orders. If the specialist receives a marketable limit order to buy 500shares at 20 1/8, this order is immediately executable against the outstanding offer at 20 1/8. The specialist does not have the option of improving the bid quotation to 20 1/8, as this would create a locked market. However, instead of executing the marketable limit order at the offer, the specialist may decide to hold the marketable limit order off of his book, and utilize this order to provide price improvement to sell orders that subsequently enter the market. Specifically, if the specialist then receives a market order from Broker A to sell 500 shares, Broker A's sell order will be executed not against the preexisting customer limit order at 20 represented in the bid at 20, but against the marketable limit order at 20 1/8, thereby providing an improved execution price to Broker A's sell order.

Accordingly, in utilizing marketable limit orders to provide price improvement, preexisting limit orders on both sides of the book are disadvantaged. [Although the marketable limit order may receive a delayed execution as a result of this practice, it is not otherwise harmed, as the limit order receives its limit price and is executed at the same price it would have been had it been executed against the preexisting sell limit order on the book.] First, a preexisting contra-side limit order is disadvantaged by being deprived of an opportunity for an execution when the specialist decides to hold the marketable limit order. Second, the preexisting limit order on the same side as the marketable limit order is deprived of an opportunity for an execution against the incoming order to sell. If after the marketable limit order is executed the market subsequently moves away from the price of either limit order on the book, that ordermay never receive an execution.

The Commission believes that if the contra-side interest consists solely of the specialist's quote, then it may be appropriate to use incoming marketable limit orders to give subsequently arriving contra-side interest a better price than if executed against the quote, because the specialist is giving up his own quote to advantage a customer order. However, if a marketable limit order is executed against subsequently arriving interest only because the specialist has refrained from executing it against a limit order that was previously entrusted to him, the preexisting limit order is denied an execution. Therefore, the Commission believes that it may be inappropriate to utilize marketable limit

orders to provide price improvement under these circumstances, as it seriously disadvantages preexisting limit orders.

6.Percentage Orders

The rules of the NYSE and Amex allow members to enter a percentage order wit h a specialist, [See NYSE Rules 13 and 123A.30; Amex Rules 131 and 154.] which is a limited price order to buy (sell) up to 50% of the volume of a specified s tock after its entry. A percentage order is essentially a memorandum entry left with a specialist all or part of which becomes an order capable of execution ei ther by being: (1) "elected" into a limit order on the specialist's bo ok through subsequent transactions [Under the election process, as trades occur at the percentage order's limit price or better, an equal number of shares of th e percentage order are "elected" and become a limit order on the specialist's bo ok at the price of the electing sale. See NYSE Rule 123A.30; Amex Rule 154. Fo r a description of the three types of elected percentage orders, which differ on ly in the manner in which their terms of execution are determined, see NYSE Rule 13 and Amex Rule 131.] or (2) "converted" by the specialist into a bi d or offer or to participate directly in a trade. On the NYSE, substantially al l of the percentage orders placed with a specialist contain the conversion instr uction, which also enables the specialist to trade on parity for its own account with the converted percentage order. [If the specialist is holding more than on e percentage order capable of execution at the same price, each floor broker mus t agree to allow the specialist to trade on parity with its order. In addition, any limit orders on the book ahead of the converted percentage order must be filled before the specialist can trade for its own account. ] Such orders commonly are referred to by the acronym "CAP" ("convert and parity") or "CAP-D," the "D" indicating that conversions are permitted to occur on both stabilizing and destabilizing transactions. The use of CAP-D orders are subject to a number of restrictions. [Among the most significant of these restrictions are: (1) an order may be converted on a destabilizing tick for the purpose of participating in a trade of 10,000 or more shares; (2) the execution effected by the conversion may occur no more than 1/4 point away from the last sale (subject to waiver by a Floor Official); and (3) the specialist cannot convert percentage orders for consecutive, or contemporaneous, trades on destabilizing ticks without Floor Governor approval. See Securities Exchange Act Release No. 24505 (May 22, 1987), 52 FR 20484 (June 1, 1987) (File No. SR-NYSE-85-1) (order approving amendments to NYSE percentage order rules to permit conversion on destabilizing ticks under certain restrictions).] These restrictions were intended to minimize the specialist's discretion in handling such orders and to ensure that the specialist cannot, through its use of theconversion process, unduly influence market trends. [Id. While the Commission was concerned whether the amendments to the NYSE's percentage order rules permitting CAP-D orders to be converted for execution or quotation purposes provided the specialist with discretion in violation of Section 11(b), see supra discussion at note , the Commission believed that the restrictions imposed would provide sufficiently stringent guidelines to ensure that the specialist will use the conversion provisions only in a manner consistent with its market making duties and Section 11(b). See Securities Exchange Act Release No. 24505, supra note .]

Despite such restrictions, a specialist can utilize the conversion process to enable the percentage order and the specialist trading for its own account to receive an execution while bypassing pre-existing trading crowd and limit order book interest. In addition, the conversion process provides the specialist with the opportunity to trade as dealer with a percentage order. Specifically, if a transaction occurs in a stock, requiring unexecuted bids (offers) to be reestablished, and an order to sell (buy) enters the market, a specialist may convert a percentage order at the same price as the reestablished bids (offers) in the crowd, aggregating the percentage order with a bid (offer) for its own account, and gaining precedence based on size over the pre-existing, but reestablished, bids (offers). Moreover, as a percentage order to buy (sell) can be converted at a range of prices up to (down to) and including its limit price, the specialist can convert a percentage order at a price higher than (lower than) existing limit order and trading crowd interest, and achieve priority over such interest for a bid (offer) on behalf of the percentage order and its own account. While in this instancethe converted percentage order (and any specialist bid (offer) on parity) is providing a better price to the incoming contra-side order, as the percentage order is capable of execution at the prevailing bid (offer) prior to its conversion, the percentage order does, in a sense, bypass pre-existing limit order book and trading crowd interest.

Further, the conversion process provides a specialist with the opportunity to trade as dealer with a percentage order by converting a percentage order to buy (sell) to execute against himself as principal. For example, assume a market quoted 20 - 20 1/4, 5,000 bid and 10,000 offered, the bid representing customer limit orders and the offer representing the specialist's own interest. If the specialist receives a CAP-D percentage order to buy 10,000 shares with a maximum limit of 20 1/2, the specialist

may convert the percentage order at 20 1/4 to execute against it as principal for his own account.

7.BSE's Competing Specialist Initiative and CSE's Preferencing Program

The BSE's CSI and the CSE's preferencing program both permit the internalization of customer orders and meet the definition of preferencing found in Section 510(c) of the NSMIA. Specifically, the BSE's CSI permits an upstairs firm to route its customer orders to an affiliated BSE specialist and execute such orders as dealer unless there is either contra-side interest on the BSE's consolidated limit order book at the execution price (i.e., the ITS/BBO) or another BSE specialist in the stock is quoting at the ITS/BBO with time priority. In practice, neither of theseconditions poses much of an obstacle to the upstairs firm's internalization of its customer order flow. As with the other regional exchanges, the BSE's limit order book is generally thin, which reduces the likelihood that the book will interfere with a member's internalization of its order flow. Moreover, as BSE specialists are at one or both sides of the ITS/BBO only approximately five percent of the time, [See infra Part V at Table V-5.] it also is highly unlikely that a BSE specialist's quote will prevent a member from internalizing its order flow on a regular basis.

The CSE's preferencing program facilitates members' internalization of their customer order flow to a greater extent than the BSE's CSI does. First, the CSE rules eliminate time priority among dealers for preferenced trades. In this regard, even if Dealer A has time priority while bidding (offering) at the ITS/BBO, Dealer B could send a preferenced cross into the NSTS to be executed at the ITS/BBO without it being broken up by Dealer A. Further, non-marketable limit orders routed to BSE competing specialists are automatically entered into the BSE's consolidated limit order book, and thus are eligible for execution against incoming contra-side interest when they become marketable. [The practical impact of this difference between the CSE and BSE may be mitigated to a certain extent by the BSE specialist's option to route a limit order to another exchange.] Non-marketable limit orders routed to a CSE dealer are not required to be entered into the CSE's consolidated limit order book;instead, limit orders not entered into the limit order book must be represented in accordance with the Commission's Display Rule. [See CSE Rule 12.10, Interpretation and Policy .01. ] As a result, the CSE dealer can internalize such orders when they become marketable by sending a preferenced trade to NSTS without a significant probability that the trade will be broken up by interest on the CSE central limit order book.

C.Off-Exchange Internalization: Third Market Trading

The over-the-counter ("OTC") trading of exchange-listed securities is commonly referred to as third market trading. Third market trading is conducted both by NYSE member firms and non-NYSE member firms. The NYSE member firms must restrict their third market trading as principal in listed stocks to Rule 19c-3 securities. [See supra notes - and accompanying text.] Most of the securities traded by the NYSE firms have either been underwritten by the firm or recommended by the firm's research department. The NYSE firms receive most of their order flow from the accounts of their own retail customers, which are mainly orders for 5,000 shares or less. The NYSE firms generally execute the customer orders against their own account as dealer at the current ITS/BBO, but attempt to achieve price improvement for their customers' orders under certain circumstances.

With regard to the OTC firms' third market trading, the largest firms make markets in over 2,500 securities, many of which are listed stocks. The OTC firms receive their order flow inlisted stocks from other broker-dealers, especially discount firms and institutions, often paying for such order flow or providing some non-cash form of inducement for it. The OTC firms generally trade for their own account as dealer with this order flow at the current ITS/BBO, but may attempt price improvement for their customers' orders under certain circumstances.

It is important to note that if the third market as a whole were treated as an exchange for purposes of the definition of preferencing found in Section 510(c)(3) of the NSMIA, given the lack of time priority between the quotes of third market dealers, third market trading would fall squarely under the definition of preferencing.

D. Summary

The preceding discussion illustrates a number of the means by which broker-dealers may preference customer orders. Moreover, the discussion demonstrates that there are numerous practices by which a broker-dealer may obtain time priority over pre-existing customer orders, both with regard to orders generally in the NMS and orders on a particular exchange. Accordingly, while the BSE and CSE programs are perhaps the prime examples of preferencing, this practice occurs often, and in many forms, on all of the exchanges. Therefore, the important questions that arise out of the operation of these programs are whether preferencing harms the particular investors whose orders are preferenced or the markets in the national market system.

The Commission will consider these questions in the remainderof the study by: (1) examining a broker-dealer's obligation to achieve best execution of its customer order and how it is influenced by preferencing; (2) examining the order handling practices of preferencing firms; and (3) providing an economic analysis of market quality and executions received by customers on the CSE and BSE, and comparing the results to those of other markets.

III.Preferencing and Agency/Fiduciary Obligations

As the discussion in the previous section indicates, preferencing is just one of many market practices which accommodates a broker desiring to internalize order flow by dealing as principal with its own customers. [See generally supra Part II.] Inherent in any such practice is the concern that the broker-dealer's interests as principal may conflict with the interests of the customers to whom the broker-dealer owes a fiduciary duty of best execution.

A broker-dealer's duty of best execution derived originally from common law agency principles and fiduciary obligations, [A broker-dealer's duty to seek to obtain the best execution of customer orders derives from the common law duty of loyalty, which obligates an agent to act exclusively in the principal's best in terest. Restatement 2d Agency � 387 (1958). Thus, when a broker-dealer acts as agent on behalf of a customer in a transaction, the agent is under a duty to ex ercise reasonable care to obtain the most advantageous terms for the customer. R estatement 2d Agency � 424 (1958). The duty of best execution "does not dissolv e when the broker-dealer acts in its capacity as a principal." Merrill Lynch Securities Litigation , 911 F. Supp. 754, 760 (D. N.J. 1995). Accord E.F. Hutton & Co. , Exchange Act Rel. No. 25887, 49 S.E.C. 829, 832 (1988); Opper v. Hancock , 250 F. Supp. 668, 673-74 (S.D.N.Y.), aff'd 367 F.2d 157 (2d Cir. 1966).] and subsequently were incorporated both in SRO rules and, through judicial and Commission decisions, in the antifraud provisions of the federal securities laws. [See Market 2000 Study, supra note , Study V, V-1, 2 and sources cited therein.] The duty of best execution requires a broker-dealer to seek the most advantageous termsavailable under the circumstances for a customer's transaction. [The Commission, however, has not promulgated a separate best execution rule or explicitly defined best execution. While price is the predominant element of the duty of best execution, a broker-dealer is not bound exclusively by price considerations in satisfying best execution obligations. In this regard, the Commission has stated that among the factors to be considered by a broker-dealer in satisfying its best execution obligations are: the size of the order; the trading characteristics of the security involved; the availability of accurate information affecting choices as to the most favorable market in which execution might be sought; the availability of technological aids to process such data; the availability of economic access to the various market centers; and the cost and difficulty associated with achieving an execution in a particular market center. See SEC, Second Report on Bank Securities Activities: Comparative Regulatory Framework Regarding Brokerage-Type Services, at 97-98, n.233, as reprinted in H.R. Rep. No. 145, 95 Cong., 1st Sess. 233 (Comm. Print 1977). See also OEO Adopting Release, supra note , at Section III.C.2. ] As discussed below, concerns have arisen as to whether practices such as preferencing are consistent with a broker-dealer's duty of best execution.

A.Background - 1934 to 1975 Amendments to the Act

Questions concerning the conflicts of interest inherent whenever a broker de als as principal with its own customer order flow predate the establishment of t he Commission. Congress wrestled with these conflicts in drafting the Securitie s Exchange Act of 1934. An early draft of the Act included a provision that wou ld have prohibited any member of a national securities exchange or any broker tr ansacting business through an exchange member from acting as a dealer in securit ies. [See Section 10 of S. 2693 and H.R. 7852, 73d Cong., 2d Sess. (1934).] D uring hearings on this provision, parties argued that the segregation of the bro ker anddealer functions would seriously disrupt the financial system and that su fficient information was not available to recommend such far-reaching legislatio n. [See Stock Exchange Practices: Hearings on S. 2693, 73d Cong., 2d Sess. (193 4); and Stock Exchange Regulation: Hearings on H.R. 7852, 73d Cong., 2d Sess. (1934). Congress, however, was presented with contrary views on the matter, particularly with regard to the dual broker-dealer function of specialists. For example, the Twentieth Century Fund's study of the security markets concluded that the services of a specialist were not of sufficient value to warrant the continuation of their preferred position in the market. See Twentieth Century Fund, supra note , at 438-39. Moreover, the Twentieth Century Fund noted that a specialist was constantly in a position where his own interests come into conflict with those of his customers: "[c]ertain of his customers' orders are executed with other customers while some are executed with himself. His primary duty is to serve all of his customers at all times with due diligence and care and to regard opportunities to profit from his own trading as of distinctly secondary importance. In view of the fact that a large part of specialists' profits is derived from their personal trading, it is expecting too much to ask that their trading be given a secondary place at all times." Id. Accordingly, the Twentieth Century Fund recommended that specialists, as well as other exchange members, should not be permitted to function as both brokers and dealers. Id. at 439-440. ] As a result, the provision for complete segregation of the broker and dealer functions was deleted from the final version of the Act; however, in Section 11 of the Act the Commission was granted the authority to accomplish such segregation by rule or regulation, [See 15 U.S.C. 78k(a)-(b) (1934). If specialists were permitted to act as dealers or limited to acting as dealers, the Commission was supposed to limit their dealings by rule as far as practicable to those reasonably necessary to maintain a fair and orderly market. Id. ; see also supra note and accompanying text.] and was directed to study

and report to Congress on the feasibility and advisability of suchsegregation. [15 U.S.C. 78k(e) (1934), repealed by Pub. L. No. 94-29 (June 4, 1975). ]

The Commission's 1936 report to Congress concluded that while the combination of broker and dealer functions in one firm "involved a conflict of interest that was provocative of abuse of the fiduciary relationship in the brokerage function," it was not advisable to separate completely the two functions by legislation. [See Segregation Study, supra note . ] Rather, the Commission believed that a regulatory approach, directed towards curbing abuses resulting from the conflict of interest, while avoiding the unforeseeable impact such a separation may have on market liquidity and allowing for the development of the markets, was the most prudent course of action. [Id. at 99-114.]

Over the ensuing years, Congress and the Commission have opted for a regulatory scheme that attempts to balance the need for efficiency and liquidity in the markets against the potential harm to investors from the inherent conflict between the principal and agency roles of a broker-dealer. Moreover, in adopting the 1975 Amendments to the Act, Congress emphasized the encouragement of competition among dealers in the national market system as being essential to the achievement of best execution of customer orders.

B.1975 Amendments to the Act

In the 1975 Amendments to the Act, Congress gave a prominentrole to best exe cution concepts in encouraging the development of a national market system. [In Section 11A(a)(2) of the Act, 15 U.S.C. 78k-1(a)(2), Congress directed the Commi ssion to facilitate the establishment of a national market system in accordance with the objectives contained in Section 11A(a)(1) of the Act, 15 U.S.C. 78k-1(a )(1). These included: (i) economically efficient execution of securities transa ctions; (ii) fair competition among brokers and dealers, among exchange markets, and between exchange markets and markets other than exchange markets; (iii) the availability to brokers, dealers, and investors of information with respect to quotations for and transactions in securities; (iv) the practicability of broker s executing investors' orders in the best market; and (v) an opportunity, consis tent with the provisions of clauses (i) and (iv) of this subparagraph, for inves tors' orders to be executed without the participation of a dealer. 15 U.S.C. 78 k-1(a)(1)(C)(i)-(v). ] Congress granted the Commission the authority to facilitate the development of the NMS and to implement rules as necessary to ensure that customers receive best execution of their securities transactions regardless of the type or physical location of a particular trading market. [Senate Report, supra note , at 7-8.] Congress declined endorsing a single market through which all orders must flow, or even a single market structure. [In this regard, Congress stated that it approached the problem of encouraging development of the NMS from the point of view of preserving the competing markets for securities that have developed; breaking down all barriers to competition that do not serve a valid regulatory purpose; encouraging maximum reliance on communications and data processing equipment consistent with justifiable costs to enhance competition; and allowing economic forces, interacting within a fair regulatory field, to arrive at appropriate variations in practices and services. Id. at 8. ] Indeed, Congress concluded that best execution may only be achieved in an environment of opencompetition among markets and market makers, [Congress believed that it would be "contrary to the purpose [of an NMS] to compel the elimination of differences between type of markets or firms that might be competition-enhancing." Id. Moreover, Congress stated that the first order of priority in creating an NMS was to break down the unnecessary regulatory restrictions which impeded contact between brokers and market makers and which restrain competition among markets and market makers, as the cost in creating system facilities will be wasted if brokers were prevented by restrictive rules and practices from using them to search out the best price for their customers or if dealers were prevented or hindered from engaging in market making activities. Id. at 12-13.] in which investors would be ensured a central, liquid market for the execution of their orders. [Id. at 12. ]

Congress also found two important benefits to investors when trading in an idealized auction-type market as opposed to a pure dealer market: (1) customer limit orders would have to be satisfied before any transaction at the same price by a broker-dealer for its own account could be executed; and (2) market orders could be executed against another customer limit or market order at a better price than that currently quoted by any dealer for his own account. [Id. at 16.] Congress believed that with respect to securities suitable for auction trading, every effort should be made to design the NMS so that public investors received such benefits and protections associated with auction-type trading. In this regard, Congress cited the need to develop trading rules and procedures to tie the individual market centers together in order that suchbenefits and protections of auction-type trading could be achieved. [Id. at 17.] Since 1975, these benefits largely have been addressed in the listed markets through the establishment of consolidated quotation and last sale reporting, the development of ITS as an intermarket trading linkage with trade-through protection, the implementation of a proprietary trading scheme pursuant to Section 11(a) and the rules promulgated thereunder, and, most recently, the adoption of the Display Rule.

Notwithstanding these developments, Congress in the 1975 Amendments to the Act, and the Commission consistently over the years, have grappled with the means of achieving best execution in a market structure comprised of dispersed market centers, linked by technological mechanisms, but marked by the practical necessity of automating the routing of customer order flow in a security to a particular market or market maker. The Commission has been particularly concerned that practices such as payment for order flow and internalization, while facilitating sources of competition to dominant market centers, could cause broker-dealers to make automated order routing decisions for reasons other than their customers' best interests. Accordingly, as discussed below, the Commission consistently has addressed this problem by stating that a broker routing retail orders in a particular security to a single market (by automated or other means) must at least make periodic assessments of the quality of competing markets to assure that it is taking all reasonable steps under the circumstances to seek outbest execution of customers' orders. [See Securities Exchange Act Release Nos. 15671 (March 22, 1979), 44 FR 20360 (April 4, 1979) (Status Report on the Development of a National Market System) ("1979 Status Report"); 15926 (June 6, 1979), 44 FR 36912 (June 15, 1979) (Rule 11Aa2-1 Proposing Release); 16590 (February 19, 1980), 45 FR 12391 (Rule 11Ac1-2 Adopting Release); 17583 (February 27, 1981), 46 FR 15713, 15715 n.16 (March 9, 1981); 26870 (May 26, 1989), 54 FR 23963, 23973 n.127 (June 5, 1989); 34902 (October 27, 1994), 59 FR 55006 (November 2, 1994) (File No. S7-29-93) (Payment for Order Flow Adopting Release); CSE and BSE Approval Orders, supra note ; OEO Adopting and Proposing Releases, supra notes and . See also Market 2000 Study, supra note , Study V at V-4.]

In 1994, the Division's Market 2000 Study reiterated previous Commission statements that an automated order routing environment can be consistent with best execution, as long as the broker-dealer periodically assessed the quality of competing markets. [Market 2000 Study, supra note , Study V, at V-4.] However, the Division asserted that the presumption of some market participants that their best execution duties were satisfied under all circumstances by routing their small customer orders to a market guaranteeing execution at the ITS/BBO might not be consistent with the then-existing market alternatives for listed securities. Specifically, the Division noted that orders routed to an exchange are exposed to other public orders or interest in the trading crowd, with the possibility that the order may receive a price better than the existing quotations. [Id. The regional stock exchanges' automated order routing and execution systems incorporate a feature that displays certain orders that are eligible for automatic execution at the ITS/BBO to the specialist on the specialist's trading screen for at least 15 seconds to give the specialist the opportunity to provide an improved execution price to such orders. See BSE Rules, Ch. XXXIII, Secs. 3 and 5; CHX Rules, Art. XX, Rule 37; Phlx Rule 229; and PSE Rule 5.25.] Therefore, theDivision concluded that, for transactions in listed stocks, the possibility for price improvement bears on the question of whether a broker-dealer is fulfilling its best execution obligations, and that the existence of such a possibility, even if the price is not actually improved, could be a factor in determining whether best execution had been achieved.

Later in 1994, the Commission adopted rules relating to the disclosure of broker-dealer payment for order flow arrangements, which generally involve the automated routing of a broker's small retail order flow to a specialist or market maker for execution in return for compensation. [See Payment for Order Flow Adopting Release, supra note . See also Securities Exchange Act Release No. 33026 (October 6, 1993), 58 FR 52934 (October 13, 1993) (File No. S7-29-93) (Payment for Order Flow Proposing Release). In the Payment for Order Flow Proposing Release, the Commission preliminarily stated that the position taken by some commenters that best execution would be obtained by executions for small orders at the NBBO may not be consistent with previous Commission statements regarding best execution.] The Commission's primary concern with such arrangements was whether a broker-dealer receiving payment for order flow had met its best execution obligations if it routed its order flow solely based on rebates from a market maker rather than on the quality of executions it received from the market maker.

Considering, among other things, the Division's statements in the Market 2000 Study with regard to price improvement and best execution, the Commission concluded that the opportunity for priceimprovement, while not the exclusive factor, bears on the question of whether a broker-dealer is fulfilling its duty to seek best execution, especially when payment or other inducements are received for guaranteeing order flow. The Commission noted that a broker-dealer must assess periodically the quality of the markets to which it routes packaged order flow absent specific customer instructions, and believed that a broker-dealer must not allow a payment or an inducement for order flow to interfere with its efforts to achieve best execution. In light of this ongoing assessment, the Commission did not believe a broker-dealer violates its best execution obligation merely because it receives payment for order flow.

In September 1996, the Commission adopted the Display Rule to require the display of customer limit orders priced better than a specialist's or OTC market maker's existing quote or that add to the size associated with such quote. The Commission also adopted amendments to Rule 11Ac1-1 ("Quote Rule") under the Act to require transparency of market maker and specialist prices entered into electronic communications networks ("ECNs"). [See OEO Adopting Release, supra note .] The Commission believed that the Display Rule in particular would have a number of beneficial effects, including facilitating the ability of broker-

dealers to obtain the best available executions for their customer orders.

In conjunction with the adoption of the Display Rule and the amendments to the Quote Rule, the Commission also issued astatement as to the best execution obligations of broker-dealers in a market system driven by evolving technology. The Commission stated that the scope of the duty of best execution must evolve as changes occur in the market that give rise to improved executions for customer orders, including opportunities to trade at more advantageous prices. [See Id. at 163.] Accordingly, the Commission believed that broker-dealers' procedures for seeking to obtain best execution must be modified to consider price opportunities that become "reasonably available." [Id. ]

The Commission noted that it traditionally had recognized the practical necessity of automating the routing and execution of small retail orders and had indicated that such practices are not necessarily inconsistent with best execution. [See supra sources cited at note .] Nevertheless, the Commission emphasized that best execution obligations require that broker-dealers routing orders for automatic execution must regularly and rigorously assess the quality of competing markets to assure that order flow is directed to markets providing the most beneficial terms for their customers' orders. [See OEO Adopting Release, supra note , at 163.]

The Commission acknowledged the increasing incorporation of order exposure a nd execution algorithms that afford the opportunityfor price improvement [The Co mmission noted that the term "price improvement" referred to the difference betw een execution price and the best quotes prevailing in the market at the time the order arrived at the market or market maker. See OEO Adopting Release, supra note , at n.357. Further, the Commission stated that any evaluation of price improvement opportunities would have to consider not only the extent to which orders are executed at prices better than the prevailing quotes, but also the extent to which orders are executed at inferior prices. Id. ] into the automated order routing and execution systems of broker-dealers and exchanges. [See Id. at 165; Market 2000 Study, supra note , Study V at V-4 and n.19. See also supra note . In this regard, the Commission had stated in the OEO Proposing Release that "the availability of sophisticated order handling systems has made it possible for some broker-dealers and market centers to provide an opportunity for price improvement for their customer orders. The use of these efficient routing and execution facilities by firms and exchanges suggests that price improvement procedures and other best execution safeguards in an automated environment are increasingly practicable and are setting new standards for the industry." See OEO Proposing Release, supra note , at 9-10.] As a result, the Commission asserted that the development of these efficient new facilities has altered what broker-dealers must consider in seeking best execution of customer orders; namely, the importance of the opportunity for price improvement as a factor in best execution, in the context of aggregate order handling decisions for both listed and OTC stocks. [See OEO Adopting Release, supra note , at 164-168. See also OEO Proposing Release, supra note ; Payment For Order Flow Adopting Release, supra note , at n.31-33; and CSE and BSE Approval Orders, supra note . ] The Commission therefore stated its belief that routing order flow for automated execution, or internally executing order flow on an automated basis, at the best bid or offer quotation, would not necessarilysatisfy a broker-dealer's duty of best execution for small orders in listed and OTC securities. [See OEO Adopting Release, supra note , at 164.]

Furthermore, the Commission made clear that in conducting the requisite evaluation of its internal aggregate order handling procedures, a broker-dealer must "regularly and rigorously" examine execution quality likely to be obtained from the different markets or market makers trading a security. [See Id. at 165; OEO Proposing Release, supra note . See also CSE and BSE Approval Orders, supra note .] To appropriately carry out such an evaluation, the broker-dealer must carefully examine the extent to which directed order flow would be afforded better terms if executed in a market or with a market maker offering price improvement opportunities. In the event that material differences exist between the price improvement opportunities offered by markets or market makers, the Commission asserted that these differences must be taken into account by the broker-dealer. [With regard to limit orders, the broker-dealer must take into account any material differences in execution quality ( e.g. , the likelihood of execution) among the various markets or market centers to which limit orders may be routed. The Commission added that for all orders, a broker-dealer could continue to consider the traditional best execution factors such as suitability of particular markets for different types of orders or particular securities ( i.e. , the cost or efficiency of executions on a particular market). See OEO Adopting Release, supra note , at 164 and 166. See also supra note (list of various non-price related best execution factors). ] Moreover, if different markets may be more suitable for different types of orders or particular securities, the broker-dealer willalso need to consider such factors.

In addition, the Commission cautioned that broker-dealers must not allow an order routing inducement, such as payment for order flow or the opportunity to trade with that order as principal, to interfere with its duty of best execution. [See OEO Adopting Release, supra note , at 166.] However, the Commission also reemphasized that in light of a broker-dealer's obligation to assess the quality of the markets to which it routed packaged order flow absent specific instructions from customers, the Commission does not believe that a broker-dealer violates its best execution obligation merely because it receives payment for order flow or trades as principal with customer orders. [See Id. ; Payment for Order Flow Adopting Release, supra note .]

C.Preferencing and Best Execution

1.CSE and BSE Permanent Approval Orders

In permanently approving the BSE's CSI and the CSE's preferencing program, the Commission determined that these practices, while they may enable broker-dealers to internalize order flow on these exchanges, were not necessarily inconsistent with best execution of customer orders. [See supra Parts I.B.4 and I.C.3 for a detailed description of the Commission's rationale for approving these programs on a permanent basis.] With regard to the BSE, the Commission noted that under the BSE's CSI, all limit orders directed to the BSE are represented in its consolidated limit order book and executed according to strict time priority. Further, all incoming market and marketable limit orders are scanned against the BSE limit order book prior to being routed to a designated specialist, and exposed to the specialist for possible price improvement prior to being matched against contra-side orders on the book. Accordingly, the Commission noted that the BSE's CSI provides the opportunity for order interaction before a specialist can execute the order against itself, as well as an opportunity for price improvement. [Moreover, the Commission noted that the CSI had enhanced order interaction on the BSE through increasing the volume of limit orders sent to the BSE. See BSE Approval Order, supra note .]

As for the CSE, the Commission recognized commenters' concerns that the ability of CSE preferencing dealers to internalize order flow on the CSE without interruption from other dealers provided an incentive for dealers to delay sending limit orders to the CSE until marketable, thus depriving all orders on the CSE of the benefits of order interaction. However, the Commission believed that the limit order display and price protection policies adopted by the CSE in conjunction with the permanent approval of the preferencing program would promote order interaction on the CSE through improved quotations and increased volume on the CSE's central limit order book. [See CSE Approval Order, supra note . ] The Commission, however, recognized that the holding of customer limit orders that are routed to a CSE dealer for execution on the CSE outside of the CSE's central limit order book raised concerns about whether such order handling practices are consistent with a CSE dealer's best executionobligations. Therefore, the Commission emphasized that a CSE dealer choosing to represent a customer limit order in his or her quote instead of on the CSE's central limit order book must ensure that the customer is not disadvantaged as a result of that decision. [Id.] In this regard, the Commission further noted that a CSE dealer choosing to represent a limit order in his or her quote has the obligation to monitor executions on the CSE to ensure that the limit order receives an appropriate execution. [Id.]

Moreover, the Commission noted that the CSE was also adopting a market order price improvement policy which would prevent CSE dealers from internalizing market orders in greater than minimum variation markets without first providing them with an improved price or the opportunity for price improvement through exposure on the CSE and to the national market system. [ See supra note and accompanying text.]

Further, in both the BSE and CSE Approval Orders, the Commission once again noted that a broker-dealer choosing where to automatically route orders must assess periodically the quality of competing markets to assure that its order flow is directed to markets providing the most advantageous terms for its customer's orders. Thus, the Commission made clear in the BSE and CSE Approval Orders that broker-dealers sending orders to these exchanges must satisfy themselves that this routing decision is consistent with their best execution obligations. In performingthis task, the broker-dealer must rigorously and regularly examine the executions likely to be obtained for customer orders in the markets trading the security, together with any other relevant considerations in routing such orders.

2.Discussion

As can be gleaned from the prior sections, preferencing is but one facet of the larger principal/agency debate. Like the specialist system, the combination of brokerage and money management, and the internalization of listed and OTC securities, preferencing is a manifestation of the tension between the duty of an agent and the need for efficiency and competition in the market. Like a large integrated broker-dealer firm that internalizes Rule 19c-3 securities, a preferencing dealer is attempting to gain economies of scope through efficient use of vertically organized distribution networks; achieve economies of scale with order flow it has generated; avoid sending business to its competitors; and capture the dealer's turn, or spread. [See Payment for Order Flow Proposing Release, supra note at Section F. ]

Preferencing also exists because of the need for firms to automate the routi ng of their retail/bulk order flow. Where no practical means exists for a broke r-dealer to route on an order-by-order basis, it must establish automatic defaul ts for the routing of retail order flow. Preferencing is one means by which a firm seeks to extract value from its packaged order flow. The key question is whether such a practice is consistent with bestexecution of customer orders. Because preferencing is so similar to other well-sanctioned market practices, it is not apparent that it should be held to a different standard than stopping stock or internalization and the broad affiliation of many specialist units with upstairs firms. In theory, if appropriate protections are in place to avoid compromising the customers' interests, and the broker meets its best execution obligations, preferencing should be no more disadvantageous to the customer than the dealer activities of the traditional specialist and should provide similar liquidity and quality of execution. In this regard, the execution quality on preferencing exchanges is in large part dependant on the diligence of those exchanges' members in handling customer orders. While this is true on all markets, it is of particular significance in markets where dealers execute customer orders as principal. Accordingly, it is incumbent on the preferencing exchange, as well as the Commission in its oversight capacity, to ensure that preferencing dealers provide best execution of customer orders. Moreover, the Commission examined the preferencing programs of the CSE and BSE during their pilot phase and found that both were not necessarily inconsistent with a broker-dealer's duty of best execution. To reach this conclusion, the Commission's OEA examined a variety of market data. Nevertheless, in response to the NSMIA, the Commission undertook a fresh round of data analysis and reviewed the order handling practices of preferencing firms. As discussed more fully in the sections below, new data continues to support the Commission's earlier determination that the CSE'spreferencing program and BSE's CSI are not necessarily inconsistent with a broker-dealer's duty of best execution.

IV. Review of Order Handling Practices of Preferencing Firms

A. General Industry Order Routing Practices

Commission staff ("staff") reviewed the order routing practices of 13 broker-dealers, [The following staff observations were reached by interviewing broker-dealer and exchange employees. ] including seven that participate in the CSE’s preferencing program, [The seven firms participating in the CSE � s preferencing program are Dain Bosworth Inc.; National Financial Services Corp. (Fidelity's brokerage arm); Olde Discount Corp.; PaineWebber Inc.; Pershing Trading Co. (a division of Donaldson, Lufkin & Jenrette Securities Corp.); Piper Jaffrey Inc.; and Prudential Securities Inc.] six that participate in the BSE’s CSI, [The six firms that participate in the BSE � s CSI are Dean Witter Reynolds Inc.; National Financial Services Corp.; Garden State Securities Inc.; Jefferies & Company Inc.; PaineWebber Inc.; and Pershing Trading Co.] and three that participate in neither program. [These three firms are Merrill Lynch, Pierce, Fenner & Smith Inc.; Charles Schwab and Co., Inc.; and Smith Barney Inc. ] These firms include many of the largest retail broker-dealers in the United States. Most of the firms receive their order flow either from their proprietary network of retail brokers or through correspondent relationships with other broker-dealers. [Correspondent broker-dealers maintain their own account base and route their orders to another broker-dealer for clearing and/or execution services.] Some firms pay for their order flow or receive other forms of inducement for order flow. [For a detailed description of payment for order flow practices, see Payment for Order Flow Adopting and Proposing Releases, supra notes and .]

Several of the firms reviewed own and/or have entered into joint venture arrangements with specialist units on the regional exchanges. The joint venture arrangements usually provide that a firm will route its order flow to a regional specialist unit in exchange for a portion of the unit's profits. [Many of these firms contribute capital to the joint venture specialist unit without acquiring a proprietary interest in the unit. Typically, this is done by lending cash or securities to the unit as an interest-free capital contribution.] In addition, one firm operates a specialist unit on the NYSE but does not own or have any other specialist arrangements on the regional exchanges. As described below, the order routing practices of most firms do not vary significantly. In general, all firms which own and/or have joint venture arrangements with specialist units on the regional or primary markets will direct retail customer order flow to those regional affiliates. The firms generally do not distinguish between market and limit orders for purposes of order routing. [One firm's CSE preferencing desk routes most of its non- marketable limit orders to the primary market for execution and retains its market and marketable limit orders for pairing and execution on the CSE .]

Brokers at all but one firm can enter retail orders into their firm's automated order routing system which will automatically route the orders to a designated market. The overwhelming majority of retail orders are routed this way. Most firms also provide the registered representative with other routing options. For example, the broker may telephone a large-sized retail order to the firm'sspecial handling or block desk. [A firm's block desk will attempt to match or execute block orders ( i.e. those for 10,000 shares or more) placed by institutions against contra-sided interest. In these transactions, the block trader usually acts as agent on both sides of the transaction.] The broker may also forward retail orders for Rule 19c-3 securities to the firm's 19c-3 desk if the firm is a 19c-3 dealer. [See supra notes - and accompanying text; Part II.C.] Finally, pursuant to a customer's request, the broker may override the firm's automated allocations and direct the retail order to a designated market for execution. [One firm does not provide its customers the option of routing their orders to a particular market center .]

Retail orders entered into a firm's automated system are automatically routed to a pre-determined market according to the parameters contained in the firm’s order routing algorithm. Each of the firms reviews and updates its order routing parameters depending on certain factors. First, a firm will prefer to route most retail orders in a stock to the market on which it has an affiliated or joint venture specialist unit that handles that stock. If the retail order is for a stock in which the firm does not make a market (either directly or through a joint venture arrangement), then the order is routed to the primary market or to an unaffiliated specialist subject to the factors described below. One firm of the firms reviewed does not route any retail orders to the regional exchanges. Instead, this firm routes all of its

retail orders to the primary market, or to its block desk for execution. [The firm does route retail orders to its affiliated specialist unit on the primary market. ]

Second, firms take into account the size of the retail order when considering a regional exchange for execution. The order size limit varies from firm to firm and, within each firm, from stock to stock. The order size limits of the firms reviewed varied from 2,000 shares to 10,000 shares. Several firms set higher size limits for retail orders sent to their affiliated specialists than for retail orders sent to unaffiliated specialists. For example, one firm routes all retail orders up to 10,000 shares to its affiliated specialists while at the same time only routing retail orders up to 5,099 shares to unaffiliated specialists on the regional exchanges. The firms explained that they generally set higher order size limits for their affiliated specialists than unaffiliated specialists because they believe that their affiliated specialists are more likely to have the resources to handle the larger-sized orders. [Some firms stated, however, that they may raise their order size limits for some unaffiliated specialists who have demonstrated that they can handle larger-sized orders and who are actively seeking more business.]

Generally, an order exceeding the firm’s automated parameters is handled manually. A large sized retail order may be sent to a special handling or block desk where it is crossed with a contra-sided institutional order. Usually, the blocks will then be

executed at the NYSE, although the blocks might be executed at another market center.

Third, the firms claim to review execution quality on the primary and regional exchanges. One way firms review execution quality is by comparing price improvement statistics they receive from the regional and primary exchanges. [Each exchange has developed the capability to provide price improvement statistics to its members. Generally, most exchanges provide firms with price improvement statistics for market orders received in non-minimum variation markets. In addition, some exchanges also provide price improvement statistics for market orders received in minimum variation markets. Further, most exchanges can also break down price improvement statistics on a specialist unit by specialist unit basis.] Firms also review proprietary price improvement statistics for their affiliated specialists. Because the criteria for determining price improvement vary across exchanges, some firms have constructed, or are in the process of constructing, their own internal systems to evaluate primary and regional price improvement statistics based on the firm's own criteria. Some firms also monitor for "price disimprovement" which is generally described as an execution at an inferior price to the ITS/BBO at the time the order was received by the exchange. Some firms compare execution quality by comparing the speed of execution of limit orders on the various exchanges. [In general, these firms have found that smaller sized retail limit orders receive a faster execution on a regional exchange than on the NYSE because of the smaller size of the regional exchanges' limit order books .] Some firms also consider the exchanges' technological capacity and the ability of their systems to quicklyexecute orders as a significant routing factor. Finally, some firms consider the capitalization of the specialists trading the stock and specialists' ability to execute large orders.

Fourth, many firms stated that they are advocates of the National Market System and/or encouraging competition among the exchanges. These firms asserted that they attempt to support the various regional exchanges by routing orders to many different markets. [As noted previously, however, one firm exclusively routes its orders to the primary market . See supra note and accompanying text.] Fifth, the firms look at the overall costs of executing orders on each exchange. Finally, the firms consider intangible factors such as the integrity of a particular specialist and whether or not the firm has a positive working relationship with the specialist.

B.Order Handling Practices of CSE Preferencing Dealers and BSE Competing Specialists

1. CSE Preferencing Dealers

a. Background

The CSE makes markets in approximately 568 stocks. With the exception of six dually-listed stocks, all of these are listed on the NYSE or the Amex, and are traded on the CSE on a UTP basis. CSE's preferencing firms make markets in 422 stocks, accounting for 80% to 99% of the total CSE trade volume in those stocks.

In the six years since its preferencing program was first approved on a pilot basis, the CSE has experienced rapid growth intrading. [Compare Appendix C at Table C-2 (CSE trading and share volume 1985-1990).] In 1991, the CSE had an average daily volume of 2,268,000 shares and 1,983 trades. By the first half of 1996, the average daily volume had grown to 7,520,945 shares and 16,911 trades with 75% of the share volume and 84% of the trade volume attributable to the CSE's preferencing program.

Of the 13 CSE member firms that are Designated Dealers on the CSE, seven participate in the preferencing program. [See supra note . Two other firms that are no longer Designated Dealers on the CSE, Smith Barney and Redwood Trading Inc., were participants in the preferencing program until July 15, 1996.] These seven firms trade from 28 to 266 stocks within the preferencing program. The five largest preferencing dealers account for approximately 95% of the preferencing program trade volume and 80% of the aggregate trade volume on the CSE. In addition, the five largest preferencing dealers account for approximately 97% of the preferencing program share volume and 73% of the aggregate share volume on the CSE.

b. Order Handling Practices

Staff visited four of the five largest participants in the CSE preferencing program to gather information on how their CSE preferencing dealer desks operate. Each CSE preferencing dealer desk receives its order flow from its respective firm's automated system. Each desk operates a proprietary screen-based trading system that handles incoming retail orders. Retail orders received from the firm’s automated system are automatically entered into thefirm’s proprietary trading system. These proprietary trading systems also function as the firms' proprietary limit order books. The four firms reviewed generally do not use the CSE’s central limit order book. Instead, the firms prefer to either place limit orders on their proprietary limit order books or route the limit orders to the primary market. [One firm routes all limit orders away from the market to the primary market .]

The proprietary trading systems allow certain retail orders to be automatically paired and entered into NSTS in certain market situations. For example, some firms’ proprietary trading systems automatically pair the firm against market orders and enter these into NSTS in minimum variation markets when the primary market is printing at both the bid and the offer.

As a general rule, each CSE preferencing dealer claims that it attempts to provide each customer’s order with the same type of execution as if the order were executed on the primary market or better. [If the NYSE is not quoting at the ITS/BBO, a CSE preferencing dealer will still pair itself against a customer � s order at the ITS/BBO or better. Generally though, CSE preferencing dealers find that the NYSE quotes at the ITS/BBO most of the time.] Preferencing dealers will check the consolidated quotes and trades immediately after receiving an order. The preferencing dealer will then attempt to execute the order through NSTS at the ITS/BBO price or better. Furthermore, a preferencing dealer will not execute a customer order that would establish a new high or low price in the stock for the day. Creating a new high orlow is avoided because it is believed to be a trade at a price "inferior" to that of the primary market. To avoid establishing a new high or low, the preferencing dealers generally stop [In stopping the order, the specialist is guaranteeing that the order will be executed at no worse a price than the prevailing ITS/BBO when the order was received, with the understanding that the specialist will attempt to execute the order at a better price. See supra Part II.B.4.] the order at the ITS/BBO and match the price of the next transaction on the primary market. [For example, assume the market is 20 - 20 1/8 and the CSE preferencing dealer receives a market order to buy. The most recent print on the primary market is at 20, which is also the highest print of the day. The CSE preferencing dealer will not execute the buy order at the offer of 20 1/8 since this would result in the highest execution price for the day. Therefore, the preferencing dealer will stop the order at 20 1/8 and execute the order in NSTS at the next primary market print so long as the print is at the stop price or better.]

Similarly, a preferencing dealer will not execute a transaction at a price that would cause the market to trade up or down on consecutive prints (the "double tick" rule). [A double uptick occurs after two consecutive transactions are executed at prices higher than the previous transaction, while a double downtick occurs after two consecutive transactions are executed at prices lower than the previous transaction. For example, a double uptick occurs if the next two consecutive trades in a stock which had previously been trading at 20 occur at 20 1/8 and 20 1/4.] As with establishing a new high or low, the preferencing dealer wants to avoid trading where the primary market has yet to trade. Before effecting a trade that would lead to a double uptick or double downtick, the preferencing dealer will stop its customer order at the ITS/BBO and wait for a primary market trade. The customer order will then be executed at the same price as the next trade onthe primary market. [For example, assume the market is 20 - 20 1/8 and the CSE preferencing dealer receives a market order to sell. The most recent print on the primary market is at 20 1/8 on a downtick. The CSE preferencing dealer will not execute the sell order at the bid of 20 since this would result in an execution on a double downtick. The CSE preferencing dealer will stop the order at 20 and execute the order at the same price as the next primary market print.] Some CSE preferencing dealers have computer system safeguards that prevent them from inputting paired orders into NSTS at a price that would create a high or low for the day or produce a double uptick or double downtick.

If the spread between the ITS/BBO is greater than the minimum variation at the time a customer market order is received, preferencing dealers may handle the order in one of two ways to comply with the CSE's price improvement policy. [The CSE's price improvement policy, applicable in greater than minimum variation markets, requires dealers either to execute a customer market order at an improved price or to display the order on the CSE for a minimum of 30 seconds. See supra note and accompanying text for a description of this policy.] First, the preferencing dealer immediately may execute the order at an improved price. Second, the preferencing dealer may attempt price improvement by stopping the order on the bid (for a sell order) or on the offer (for a buy order) and displaying an improved, representative bid or offer in either the CSE's central limit order book or the preferencing dealer’s CSE quotation for at least 30 to 60 seconds. [Id. ] Some firms also have adopted computer system safeguards to prevent the preferencing dealer from trying to execute the order without attempting price improvement. Inpractice, preferencing dealers generally will represent a portion of the order in their CSE quotation. [Several firms will display a representative portion of the order in its CSE quote, which is usually the greater of 500 shares or ten percent of the entire order.]

All preferencing dealers will generally give their customers price improvement during this time period if one of the following three conditions occurs. First, either the primary market or the CSE executes an order at the improved CSE bid or offer. Second, the improved bid or offer is hit either by a CSE dealer or another market’s ITS commitment. Third, the preferencing dealer receives a contra-side market or marketable limit order. For example, assume the market is 20 - 20 1/4 and the CSE preferencing dealer receives a market order to buy 500 shares. The preferencing dealer may stop the existing order at 20 1/4, and then improve the quoted market by entering a new bid at 20 1/8 for 500 shares. If the primary market or another CSE Designated Dealer prints an execution at 20 1/8, the CSE preferencing dealer will execute the order at 20 1/8. If the bid is hit either by another CSE dealer or via ITS, the stopped order will receive an execution at 20 1/8. If the CSE preferencing dealer receives a market or marketable limit order to sell, then the stopped order will be matched against the sell order at 20 1/8.

If none of these events occur by the end of 30 seconds, then the preferencing dealer will usually execute the order through NSTS based on the last consolidated trade. For example, if the stopped order from the above scenario has not been executed within 30seconds, and the next primary market print is at 20 1/4, then the stopped order will receive an execution at 20 1/4. The preferencing dealer, however, might not execute the stopped order at 20 1/4 if it believes market conditions suggest that the stopped order may still receive an execution at 20 1/8. In addition, some CSE preferencing dealers may send a proprietary market buy order to the primary market. The preferencing dealer will wait for the proprietary order to be executed and then give the customer order an execution at the same price as the proprietary order.

When a market order is received in a minimum variation market, the order is either automatically executed at the ITS/BBO or sent to the preferencing dealer for manual handling. Usually a market order to buy will be executed on the offer and a market order to sell will be executed on the bid. On some occasions, a customer may be able to buy on the bid or sell on the offer and thus receive price improvement. This generally occurs when the primary market is printing on one side of the market. For example, if the market is 20 - 20 1/8 but the primary market is printing at 20 and has not printed at 20 1/8 for some period of time, the preferencing dealer will not automatically execute a market buy order at 20 1/8. Some preferencing dealers will stop the order at 20 1/8 and then send a proprietary market buy order to the primary market. The preferencing dealer will then wait for the proprietary order to be executed and then give the customer order an execution at the same price as the proprietary order. Other CSE preferencing dealers will attempt to achieve the same result by stopping the customerorder and then matching the price of the next trade on the primary market so long as the print is at the stop price or better. [Another instance in which the customer may receive price improvement is if an execution would establish a new high or low in the stock for the day or if the execution would cause a double uptick or double downtick.]

When receiving a limit order that is away from or at the current market, each CSE preferencing dealer claims that it attempts to provide each customer's order with the same type of execution as if the order were sent to the primary market. Preferencing dealers fulfill this guarantee one of three ways. First, a preferencing dealer may send a proprietary limit order for part of the limit order to the primary market at the customer's limit order price. [The theory behind sending a proprietary limit order to the primary market is that the order will act as a " marker " on the limit order book. The proprietary order stands in line behind previously entered limit orders at the same price. When the proprietary order is executed, the customer order is also immediately executed. This procedure ensures that the customer order will be executed no later than if it had been sent to the primary market. The proprietary order may also serve a risk management function by acting as a layoff trade for the preferencing dealer.] This "marker" order marks the corresponding place of the customer’s limit order on the primary market's limit order book. After the preferencing dealer receives an execution for the entire "marker" limit order on the primary market, the preferencing dealer will pair itself against the customer’s limit order and enter the trade into NSTS.

Second, a preferencing dealer may send the entire customer limit order to the primary market. Simultaneously, the preferencing dealer takes steps to ensure that it does not tradeahead of the limit order. Most of these limit orders are executed on the primary market. However, if the CSE preferencing dealer receives a contra-side market order while the limit order is at or better than the existing market, the preferencing dealer will match the existing limit order with the incoming market order and send the orders to the CSE for execution, while simultaneously canceling the existing limit order on the primary market.

Third, a preferencing dealer may place the order in the firm's internal limit order book and take a "snapshot" of the marketplace when the primary exchange first quotes at the limit order price. The snapshot includes the size of the limit order book at the limit order price. The system will then monitor primary market executions at the limit order price and will generally execute the limit order when the aggregate size of the prints on the primary exchange is equivalent to the size of the primary exchange’s limit order book when the snapshot was taken. [One advantage of using the snapshot method is that it generally yields a faster execution than the marker method since the snapshot method counts all market, limit and block orders executed on the exchange. One disadvantage of the snapshot method is that because a firm cannot see the entire primary market limit order book, a snapshot cannot be taken until the stock is quoted at the limit order price. Therefore, the snapshot provides, at best, an approximation of what the execution on the primary market would be. One firm recently replaced the snapshot method with the marker method and another firm is considering doing the same.] If an incoming contra-side market order arrives in the interim, then the CSE preferencing dealer must match the existing limit order against the incoming order.

The firms contended that customer limit orders routed to a CSEpreferencing dealer have a significant trading advantage. Specifically, the firms execute customer limit orders against incoming contra-side customer market orders while guaranteeing the limit orders an execution no worse than if the limit orders were sent to the primary market. This interaction allows some customer limit orders to receive an execution earlier than if the orders had been sent to the primary market. For example, if a preferencing dealer receives a limit order to buy at 20 when the primary market is 20 - 20 1/8 with 50,000 shares at the bid and 10,000 shares at the offer, the limit order sent to the primary market presumably would have to wait at least until 50,000 shares traded at 20 before receiving an execution and perhaps much longer if interest in the trading crowd takes precedence based on size over the limit order book. [See supra Parts II.B.1 and B.2.] If the preferencing dealer, however, receives a market order to sell during the time that the limit order to buy at 20 is on the primary market's limit order book, the dealer will cross the limit order with the market sell order and enter the paired trade into NSTS for execution.

For limit orders that improve the market (i.e., an order to buy above the current bid or an order to sell below the current offer), the preferencing dealer will either immediately enter a paired transaction with the limit order into NSTS or display the limit order in its CSE quotation. An order reflected in the firm’s CSE quotation may be executed against by another CSE Designated Dealer, or by an incoming ITS commitment from another market, orthe firm will execute the order in NSTS if the primary market or CSE prints a trade at that price. CSE preferencing dealers generally will fill such a limit order in its entirety even if the primary market trade is for fewer shares than the limit order held by the preferencing dealer. Some preferencing dealers may send the limit order to the primary market, rather than executing or displaying it in their CSE quotation.

Before the implementation of the Display Rule, [The Display Rule requires the full display of most customer limit orders in the market maker's quote. See OEO Adopting Release, supra note .] CSE preferencing dealers would show either a representative sample of a limit order in their CSE quotation or send the full size of a non-marketable limit order to the primary market. For example, CSE preferencing dealers would display a representative portion of a limit order in their quotation in order to meet the CSE's limit order display policy. [See Appendix A at notes 22-24 and accompanying text for a description of the CSE's limit order display policy prior to the adoption of the Display Rule. ] Currently, the Display Rule "require[s] the display of customer limit orders priced better than a specialist's...quote or that add to the size associated with such quote." [See OEO Adopting Release, supra note . ] The CSE preferencing dealers reviewed by the staff fulfill this obligation in various ways. One CSE preferencing dealer automatically routes all customer limit orders priced away from or at the current market to the primary market. Another CSE preferencing dealer automatically displays the full size ofcustomer limit orders in its CSE quote. Other CSE preferencing dealers permit their traders to either display the full size of the limit order in the dealer's CSE quote or route the order to the primary market.

The CSE preferencing dealer must display the full size of a limit order in the CSE quote even if the dealer has sent a marker order to the primary market. CSE preferencing dealers differ in the size of the limit order that they will display in their quotation. For example, some CSE preferencing dealers will show the full size of a customer limit order in the quote, while one CSE preferencing dealer will not display a limit order if it is less than or equal to 10% of the displayed quote, in accordance with the de minimis guidelines set out in the Display Rule. CSE preferencing dealers also differ in the amount of time they take to display the full size of a limit order in one market center. As discussed in an interpretive letter regarding the Display Rule, "an exchange specialist...should either display the full size of the order in its own quote or send it to a market center where the order will be fully displayed." [See Letter from Richard R. Lindsey, Director, Division of Market Regulation, SEC, to Richard Grasso, Chairman and Chief Executive Officer, New York Stock Exchange, dated November 22, 1996.] Some CSE preferencing dealers will display the entire limit order in one market center immediately upon receipt. Another CSE preferencing dealer will immediately display a portion of the limit order on both the primary and CSE markets; however, within 30 seconds of its receiptof the limit order, the dealer will manually update the existing quote in one market so that the dealer displays the full size of the limit order. [Commission staff has asked the CSE to examine whether the practice of this dealer is in compliance with the requirements of the Display Rule.]

c.Risk Management Procedures of CSE Preferencing Dealers

Each individual CSE preferencing dealer manages its proprietary position according to the firm's guidelines for risk exposure. Under these guidelines, if a CSE preferencing dealer wishes to exceed the guidelines, the preferencing dealer must first receive internal supervisory authorization. Generally, most preferencing dealers will attempt to limit large proprietary positions in stocks by laying off their larger positions at the primary market in order to minimize their risk exposure. In this regard, some preferencing dealers will attempt to lay off their positions throughout the trading day, while others will attempt to unwind their positions towards the end of the trading day.

2.Boston Stock Exchange Competing Specialists

a. Background

The BSE's CSI program first became active in July 1994 with two specialists competing in two stocks. As of April 1997, six firms participate in the CSI, representing 15 specialists. [See supra note .] These six firms act as competing specialists in 91 stocks, ranging from four to 31 stocks per firm. In 1996, the BSE's average dailytrade volume was 10,636 trades, of which approximately 2,183 trades (21%) were executed by competing specialists. Moreover, the BSE's 1996 average daily share volume was 6,794,000 shares, of which approximately 748,000 shares (11%) were executed by competing specialists.

b. Order Handling Practices

As a general rule, each BSE specialist claims that it attempts to provide each customer's order with an execution as if the order were executed on the primary market or better. In order to ensure this, specialists check the quotes and trades on the primary market immediately after receiving an order, and then attempt to execute the order at the price of the last primary market print or better. Further, all specialists believe that to be consistent with this obligation, they must avoid establishing a new high or low for the day and avoid executing an order at a price that would cause a double uptick or double downtick. [See supra Part IV.B.1.b (comparable CSE preferencing dealer policies).]

All BSE CSI orders are entered into the BEACON system before they are routed to a designated specialist. BEACON automatically executes any incoming market or marketable limit orders against any existing contra-side order on the BSE's consolidated limit order book before routing the order to the designated specialist. [However, as with the other regional exchanges, there are few customer orders on the BSE's limit order book. See supra sources cited in note . ] In addition, any specialist may "intercept" order flow directed toanother specialist if the specialist is quoting at the ITS/BBO with priority. In this regard, if two or more specialists are at the ITS/BBO, the specialist that entered the quote earlier in time will receive the order. [See BSE April 1995 and February 1996 Data, supra note (data indicating regular specialist intercepts less than one percent of the order flow directed to competing specialists). See also infra Part V at Table V-5 (BSE specialists quote either one or both sides of the ITS/BBO only five percent of the time). ]

All market and marketable limit orders received in an 1/8 point or 1/4 point market are automatically displayed through BEACON for 15 seconds on the BSE specialist's screen for possible price improvement prior to automatic execution at the ITS/BBO. During this time, the specialist typically will execute the orders at the ITS/BBO or better if the specialist receives an incoming contra-side order or the order is hit by an ITS commitment or another BSE member. After 15 seconds, the execution price of the order will depend largely on prints occurring at the primary market. For example, if the market is 20 - 20 1/8 and the primary market is printing at both the bid and the offer, BEACON's automatic execution function immediately will execute all customer buy orders on the current offer and all customer sell orders at the current bid. However, if the execution would result in either a new high or low for the day (or a double uptick or double downtick), BEACON will automatically forward the order to the BSE specialist for manual execution. For example, if the market is 20 - 20 1/8 and the highest primary market print for the day is 20,BEACON will forward any market buy orders to the specialist. If, however, the market is 20 - 20 1/4 and the primary market is printing at 20 1/8 and 20 1/4, then the BSE specialist will most likely execute customer buy orders at 20 1/4 and customer sell orders at 20 1/8, thus improving the price of the customer’s sell order.

All limit orders received by the specialist reside in the BEACON consolidated limit order book and are represented and executed according to strict time and price priority. When the limit order is either at or better than the market, the entire limit order will be automatically displayed in the BSE quotation. For example, if the market is 20 - 20 1/4 and BEACON receives a 5,000 share limit order to buy at 19, the limit order will be placed in the limit order book. If the bid subsequently moves to 19, BEACON will automatically display the full 5,000 share buy order in the BSE bid.

Each BSE specialist has the discretion to determine the best method for executing its customer's limit orders. Various trading methods may be used. First, whenever a limit order is at or better than the market, the BSE specialist may execute the limit order immediately at the limit order price, provided that the execution does not result in a double uptick, double downtick, a new high or new low for the day. Second, the BSE specialist may route the entire limit order to the primary market for execution. Third, the BSE specialist may send a proprietary marker order to the primary market’s limit order book. Upon execution of the marker order atthe primary market, the BSE specialist will execute the existing customer limit order at the primary market execution price. Fourth, the BSE specialist may execute the customer limit order based on prints on the primary market. For example, assume the market is 20 - 20 1/4, and the BSE specialist receives a limit order to buy 1,000 shares at 20. If the primary market prints an execution of 500 shares at 20, the BSE specialist will also execute 500 shares of the customer limit order at 20.

C. Execution Costs for Customers

All preferencing firms charge retail customers a transaction commission based on the firms’ fee schedules. [The actual fee schedule is usually a fixed commission set by the firm. The commission is subject to negotiation by individual clients and registered representatives.] The firms do not pass through actual exchange execution costs for a particular transaction to the customer. Whenever a firm executes a customer's order through one of its affiliated specialists, the firm will disclose on its customer order confirmation that the order was executed by an affiliated specialist on a particular exchange. [Firms disclose joint venture arrangements on their confirmations as either a joint venture transaction or as a payment for order flow transaction.]

Every exchange charges different fees for various types of services. However, all firms reviewed in this study generally concluded that overall order execution costs were higher on the primary market than on the regional markets. [In this regard, the firms stated that order execution costs at the NYSE are generally higher due to the NYSE � s SuperDOT user fee, which is charged on all automated orders, and to the NYSE � s specialist order handling fee, which applies to all limit orders on the NYSE limit order book that remain unexecuted after a certain period of time.]

One firm, which provides clearing and execution services for its correspondent broker/dealers, charges its correspondent broker/dealers based on a negotiated fee schedule that includes transaction charges and specialist order handling fees. Because correspondent clients may route their orders to any exchange, the firm incurs higher execution costs for orders that are sent to the primary market and charged a specialist order handling fee. The firm may directly charge its correspondent clients for these primary market specialist order handling fees. The firm will waive this fee if the correspondent customer agrees to execute its orders on a regional exchange.

V. Empirical Analysis of Preferencing Activity

A.Introduction

In Section 510(c) of the NSMIA, Congress required the Commission to examine the costs and the effects of the practice of preferencing on the execution quality of retail investors’ trades. In particular, Congress directed the Commission to examine whether retail trades in both market and limit orders are disadvantaged by the presence of preferencing dealers in the national marketplace.

This section of the study addresses Congress’s concerns from an empirical perspective by analyzing retail trading activity on the NYSE, CSE, BSE, CHX, Phlx, and the PSE. The goal of the section is to quantify the execution quality of market and limit orders in each of those markets.

There are potentially many ways in which comparisons between these markets can be made and the choice of method can have a strong effect on the outcome. To minimize such effects, this section presents comparisons of market and limit order executions in a variety of ways, discussing in turn the benefits and the drawbacks of each method. Note that because this study was to be completed within six months, it was not feasible to collect data on any trade dates that occurred after January 20, 1997, when the Commission's new order handling rules went into effect. [See OEO Adopting Release, supra note and accompanying text.]

B.Data Description

Transaction and quotation data were obtained from each of the regional exchanges for the four weeks of October 28 to November 22, 1996 and from the NYSE for the one week of October 28 to November 1, 1996. The analysis is conducted only for NYSE-listed issues. Shares that are listed on the Amex or exclusively listed on a regional exchange are not included in the study.

The study analyzes executions that took place during normal trading hours (i.e., from 9:30 a.m. to 4:00 p.m. EST). For an order to be included in the study, it must have arrived at the exchange after the exchange opened. Opening orders, tick sensitive orders, and market orders with price qualifiers (for those datasets that identify such orders) also are excluded. As a result of these screens, the study only considers three types of orders: regular-way market orders, marketable limit orders, and non-marketable limit orders. [See Appendix D for complete details on the screens applied to each data set.] In the tables that follow, results are partitioned according to these three order types where appropriate.

The inclusion of NYSE trading in the study arises from the need for a consistent benchmark against which to compare the trades and quotes of regional exchanges. It is difficult to assess the absolute quality of a market, and one cannot know a priori the cost of a "good" versus a "poor" execution. For this study, the NYSE has been chosen as the benchmark against which each regional exchange is compared. The NYSE is the primary market and receivesconsiderable order flow in the securities under study. [A more detailed description of the data for each of the six exchanges is provided in Appendix D.] Further, not every regional exchange trades the same stocks. Therefore, to facilitate comparison the study was designed so that the NYSE trades every security in the sample.

Table V-1 presents a summary of the trading data used in the study. The table indicates the number of trades included in the study and the distribution of those trades by size of order for each regional exchange and the NYSE. For the sample of approximately 2,300,000 trades, almost 1,300,000 were executed on the NYSE. The PSE and CHX receive the most trades among the regional exchanges per unit of time, followed by the CSE, Phlx, and BSE. [Data in Table V-1 represents trades included in the study and do not necessarily represent all trades executed on the exchanges.]

The proportion of small orders (i.e., orders under 500 shares) is significantly larger on each of the regional exchanges than it is on the NYSE. Orders for 500 shares or less account for 44% of the NYSE’s trades, as compared to 70% of the BSE’s orders, and roughly 80% of the trades on the other regional exchanges. This reflects the use of the regional exchanges primarily for smaller orders. Nonetheless, the regional exchanges are also used to a certain extent by the upstairs market to execute large block cross

trades, thereby circumventing the priority of other orders on the NYSE floor. [See supra Part II.B.3.]

Table V-2 describes the distribution of market, marketable limit, and non-marketable limit orders across the sample. The table shows that the relative order mix is roughly constant across the regional exchanges. However, for all order size groupings, the NYSE has a larger fraction of non-marketable limit orders than any of the regional exchanges.

Tables V-3A, V-3B, and V-3C partition the same data along two other dimensions: (1) the bid-ask spread quoted in the market at the time the trade was executed; and (2) market orders versus marketable limit orders. Table V-3A describes 1/8 point (or minimum variation) markets; Table V-3B describes 1/4 point markets; and Table V-3C describes 3/8 points or wider markets. Each table further partitions the trades between market orders and marketable limit orders.

Turning first to Table V-3A, it is clear that a substantial majority of trades on both the NYSE and each regional exchange take place in 1/8 point markets: 82% of all trades on the NYSE were in 1/8 point markets, as compared to 83% on the Phlx, 81% on the CHX, 83% on the BSE, and 90% on the CSE. This general observation holds true across the various order size categories and for both market orders and marketable limit orders (although the percentages vary).

Marketable limit orders account for roughly two-fifths of all trades executed on the NYSE in 1/8 point markets, but asignificantly smaller fraction of trades on the regional exchanges. The proportion of the NYSE’s trades accounted for by marketable limit orders increases from 18% for 100-share orders to 79% for orders of more than 5,000 shares. The comparable proportions for the regional exchanges are about half as large as those for the NYSE when order size is taken into account. This may be indicative of different order routing strategies by investors and/or their brokers for market and marketable limit orders. It also may be due, in part, to the practice of sending orders to the NYSE by some regional dealers and specialists to "lay off" positions accumulated in trades that they execute as principal. [See supra Part IV.B.1.b.]

The data for trades in 1/4 point markets shown in Table V-3B are much sparser than that for 1/8 point markets, so the conclusions that can be drawn from comparisons among exchanges are more tenuous. Nevertheless, it appears that the proportion of trades accounted for by marketable limit orders is significantly smaller in 1/4 point markets than it is in 1/8 point markets. On the NYSE, only 7% of 100-share trades executed in 1/4 point markets were marketable limit orders; whereas, the comparable figure for 1/8 point markets is 18%. On the regional exchanges, only 6% of 100-share trades executed in 1/4 point markets were marketable limit orders as compared to 10% in 1/8 point markets.

The data for trades executed when the bid-ask spread was 3/8 or more are presented in Table V-3C only for the sake of completeness. Since these data are very sparse, caution should beexercised in drawing conclusions from these limited data, particularly with respect to the results for transaction costs presented later in this section.

C.The Quality of Market Quotations

Before considering trading activity, it is instructive to look at the quotations placed in the marketplace by the individual exchanges. These quotations indicate the prices at which exchanges stand willing to trade, and thus contribute importantly to price discovery and market efficiency. There are two basic measures to consider when evaluating the quality of market quotations. The first is the quoted price. The second is the depth, or number of shares, for which the dealer is willing to trade at the quoted price. Other things being equal, a market with a small difference ("spread") between the best price bid and the best price offered is generally more liquid than a market with a larger spread. Similarly, other things being equal, a market whose quotes are for a larger number of shares is generally more liquid and deeper than an identical quote for a smaller number of shares.

An important caveat here is that other things are frequently "not equal." Judging a market strictly by the size of its quoted spread or the depth of its quotes is problematic because it fails to consider the complexities of a market’s structure. Quote behavior information is more useful when used in conjunction with other measures of market quality. More sophisticated market quality measures are presented later in the section. It is

important to consider these qualifications when interpreting the tables in this subsection.

Table V-4 compares the bid-ask spreads and associated quotation depths of each of the five regional exchanges with those of the NYSE market. This comparison is done only for those NYSE stocks quoted on that particular regional exchange. For example, for those stocks that trade on both the NYSE and the CSE, the average NYSE quoted spread is 16.2 cents and the average quoted CSE spread is 27.8 cents. However, for the NYSE to BSE comparison, the average NYSE quoted spread is 18.5 cents and the average BSE spread is 45.8 cents. The reason the NYSE has two different spreads, 16.2 and 18.5 cents, is that the CSE and BSE trade different stocks. The stocks common to the NYSE and CSE trade at a narrower spread on the NYSE than the stocks common to the NYSE and BSE trade on the NYSE. This pair-wise comparison is constructed for all five regional exchanges.

The number of NYSE stocks included in the study for each regional exchange varies considerably. The CSE sample included only 333 stocks (consisting only of preferenced stocks). In comparison, the BSE sample included 1,800 stocks (including 81 CSI stocks), and the CHX and Phlx samples each consist of about 2,500 stocks.

The first two lines of Table V-4 compare mean and median spread quotations. The NYSE's quotations are narrower than those of the regional exchanges. This is related to the greater amount of order flow received by the NYSE. The ratio of the spreads --that of the regional divided by that of the NYSE -- ranges from 1.7for the CSE to over 2.9 for the Phlx. If spreads are compared among the regional exchanges, the CSE has narrower spreads than the other regionals, and the BSE's spreads are generally comparable to those of the PSE and CHX.

Comparing spreads based on quotations alone may be inappropriate, however. Consider an exchange that places a single quotation in an hour that is for a spread of .25. Further suppose that same exchange then issues three quotations for a spread of .75, where the time between each spread quotation was one minute. In such a case, the average spread over the 63 minutes as computed in the first two lines of Table V-4 would be about .70, while the actual spread for 97% of the time under study would be .25. For this reason, the next lines of the table V-4 compute quote measures based on time-weighted averages. Such a weighting scheme makes an adjustment to better illustrate the true quotation activity of an exchange.

Lines three and four of Table V-4 show that each exchange's time-weighted bid-ask spread does not differ materially from its quotation-weighted spread. The NYSE still quotes the narrowest market by a factor of between 1.7 to 3, and the CSE's quotes are narrower than those of the other regionals. If the depth of the exchanges' quotes is compared, the next lines of the table show that on average the NYSE's depth is about 20 to 80 times greater than the depth of the corresponding regional exchanges. This is most likely due to the large number of limit orders received by the

primary market. The ratio is similar whether the comparison is based on means or medians.

It is still possible that the comparison of Table V-4 understates the quote quality of the regional exchanges. In particular, because the regional exchanges have much lesser volume than the NYSE, it may not be profitable for their dealers to quote actively all of the time, or to quote competitively on both sides of the market. If the regionals have fewer limit orders than the NYSE, this could also contribute to the wider quotes. [See supra sources cited in note .] However, a regional exchange may have a competitive quote on one side of the market even though its bid-ask spread is large. In addition, though its average quoted depth may be low, a regional exchange may provide considerable liquidity to the market by quoting a greater depth when it has the best quotation price than it does when its quoted price is less competitive. Alternatively, a regional exchange specialist or dealer may try to control risk by decreasing its quoted depth when it narrows its spread.

Table V-5 explores these possibilities on a matched stock basis as in Table V-4. The top panel of the table computes the fraction of the time the regional exchange is at the national best bid and offer ("NBBO") on both quotation sides, on one side of the NBBO (but not both), and on neither side of the NBBO.

The measures of quote quality and depth vary considerably across exchanges as shown in Table V-5. The BSE and Phlx were at one or both sides of the NBBO only about 5% and 8.4% of the time,respectively. In comparison, the CHX and the PSE were at one or both sides of the NBBO about 41% and 54% of the time, respectively. The CSE has by far the best quotation performance among the regional exchanges based on the data in Table V-5. It is at one or both sides of the NBBO over 80% of the time. These results suggest that the CSE provides competitive price quotes a substantial portion of the time. [The data in the tables report the composite CSE quote, including both preferencing and non-preferencing dealers. When looked at separately , preferencing dealers are found to be on both sides of the NBBO 19.6% of the time, on only one side of the NBBO 62.8% of the time, and on neither side 17.6% of the time. The sample used for this calculation differs slightly from the one used in Table V-5.]

To explore quote quality a bit further, the bottom panel of Table V-5 examines quote depth when an exchange is on both sides of the NBBO, when it is on only one side of the NBBO, and when it is not on either side of the NBBO. The comparison is done on a matched stock basis.

In general, quotation depth on the regional exchanges is larger when a regional exchange’s price quote is at the NBBO than when it is not. This effect is particularly pronounced for the BSE and Phlx. The BSE’s average quote depth is 3,493 shares when its quoted prices are at the NBBO on both sides of the market, and 831 shares when it is on one side of the NBBO; however, its average quote depth is only 111 shares when it is not quoting on either side of the NBBO, which is 95% of the time. Similarly, the Phlx’s average quote depth is 1,211 shares when it is on both sides of the NBBO, and 547 shares when it is on one side of the NBBO; however,its average quote depth is only 110 shares when it is not quoting on either side of the NBBO, which is about 92% of the time.

The NYSE also exhibits slightly larger average quoted depth when it is on both sides of the NBBO than it does when it is on only one side (or in those rare instances when it is not at either side of the NBBO).

The CSE is the only exchange that does not have a larger average quoted depth when it is on both sides of the NBBO. The CSE’s average quote depth is 667 shares when it is on both sides of the NBBO; its average quote depth is 841 shares when it is on one side of the NBBO, which is only slightly higher than its average quoted depth when it is not on either side of the NBBO. Thus, the CSE’s quoted depth appears to be fairly insensitive to whether its quoted price is either away from the NBBO or on only one side of NBBO. However, the CSE’s average quote depth is 20% lower when it is on both sides of the NBBO rather than on only one side.

In summary, the quotation quality of the regional exchanges, though showing considerable variation, is generally lower than that of the NYSE. The NYSE quotes tighter spreads and deeper markets than the regionals, on average. In addition, two of the regional exchanges, BSE and Phlx, are seldom on one or both sides of the NBBO. While the PSE and CHX are on at least one side of the NBBO about half the time, only the CSE is on at least one side of the NBBO more than 80% of the time. On the other hand, the regional exchanges do, with the exception of the CSE, provide more quote depth when they are at the NBBO than when they are away from themarket. The CSE's quoted depth when it is at the NBBO is comparable to the depth provided by other regional exchanges. Note, however, that with the implementation of the Commission's new order handling rules, the characteristics of a regional exchange's quote depths are likely to change.

D.Market Order Execution Costs and Price Improvement Statistics

This subsection of the study analyzes the execution quality of market orders. Execution quality is measured in two ways:

(1) effective spread; and (2) price improvement.

For a customer market buy order, the effective spread is calculated by doubling the difference between the trade price and the midpoint of the bid-ask spread (measured at the time the order arrives). Thus, the effective spread ("ES"), for a market buy order can be calculated as follows:

ES_buy = 2 x [ trade_price - .5 x (bid_price + ask_price) ]

For a customer market sell order, the effective spread is calculated by doubling the difference between the midpoint of the bid-ask spread (measured at the time the order arrives) and the trade price.

ES_sell = 2 x [ .5 x (bid_price + ask_price) - trade_price ]

If all trades were executed on either the quoted bid or ask price, then the effective spread would precisely equal the quoted spread. [This can be seen by substituting "ask_price" for "trade_price" in the above equations, and observing that ES would then equal the quoted bid-ask spread.] But, in general, some fraction of trades occur at prices inside the quoted spread. The effective spread captures this effect.

It should be noted that the effective spread can be negative for any trade in which a customer buys below or sells above the spread midpoint. For example, in a 1/4 point market quoted

20 - 20 1/4, customer buy orders executed at 20 1/4 have an effective spread of 1/4, buy orders executed at 20 1/8 have an effective spread of 0, and buy orders executed at 20 have an effective spread of -1/4.

A second measure of execution quality is price improvement. Price improvement occurs when an incoming customer market order is executed at a price superior to the applicable quote. For a seller, price improvement is obtained when the execution price is higher than the best bid price quoted. For a buyer, price improvement is obtained when the execution price is lower than the best ask price quoted. In the above example, any customer buy order that is executed at a price lower than 20 1/4 receives price improvement. The only exception to this is if the customer buy order was paired against an incoming, and thus undisplayed, customer limit order to sell at 20 1/8. The non-display of limit orders will likely decrease as a result of the Commission's recentadoption of the Display Rule. Nonetheless, it may still be possible for a specialist to receive a limit order, fail to reflect it for several seconds, and then pair it against an incoming market order. Though limit orders generally are reflected in the specialist's quote, a specialist will, on occasion, choose not to do so (e.g., if the customer requests that an order not be displayed in the quote). [The Commission notes that the Display Rule circumscribed the ability of specialists to refrain from displaying limit orders in their quote.] Though the market order will transact at a price superior to the quote, it should not be deemed to have received price improvement because the superior price was due to the specialist’s decision not to display the contra-side limit order. However, if the specialist crossed two incoming market orders on opposite sides of the market at the midpoint of 20 1/8, for purposes of these calculations both orders would have received price improvement.

In the tables discussed below, the amount of price improvement received by market orders may be expressed as a percentage of the purchase price, as the per share improvement in trade price over the relevant quote, or simply as a binary variable indicating that a particular order either was or was not price improved.

Tables V-6A, V-6B, and V-6C report the effective spread for market orders of various sizes in 1/8 point markets, 1/4 point markets, and 3/8 point or greater markets, respectively. Within each order size grouping, orders are also partitioned into market orders and marketable limit orders.

Turning first to Table V-6A, which analyzes 1/8 point markets, the table shows that across all exchanges the effective spread for market orders ranges from about nine cents per share to 12 cents per share for orders of less than 1000 shares. Such orders are, presumably, mostly retail customer trades. For orders up to 500 shares, the NYSE shows a slightly lower effective spread, averaging about 9.5 cents per share versus an average of about 11.5 cents per share for the regional exchanges. The table also indicates that marketable limit orders have wider effective spreads than market orders.

The effective spread for trades executed on the NYSE increases with order size. That is, larger orders are traded at less favorable prices than smaller ones. This is consistent with the predictions of adverse selection models in the economic literature on market making. However, the effective spread for trades executed on the regional exchanges does not increase with order size.

It may appear surprising that the effective spread in a minimum variation 1/8 point market could be less than the minimum variation of 12.5 cents. This arises because some market orders receive price improvement in 1/8 point markets. Some market orders to buy are executed at the bid; some market orders to sell are executed at the ask. This causes the average effective spread to fall below 12.5 cents.

The bottom line of Table V-6A indicates that the NYSE, CSE,

and BSE generally have smaller effective spreads than the CHX,Phlx, and PSE. The average effective spread in 1/8 point markets across all order sizes is around 11 cents per share for the NYSE, CSE, and BSE. It is about a penny per share higher on the CHX, PSE, and Phlx. However, the relative rankings of exchanges in terms of effective spreads varies considerably depending on what size category is examined, or whether marketable limit orders are included in the comparison.

Table V-6B, which analyzes spreads in 1/4 point markets, shows more variation in the estimates of average effective spread. The average effective spread on the NYSE still increases with order size: 100 share market orders have an average effective spread of 8.8 cents per share; 1000 to 5000 share market orders execute at an average effective spread of 18.6 cents per share.

The average effective spread of the CSE in Table V-6B is lower than the NYSE and any other regional exchange. In fact, across all market orders in 1/4 point markets, the CSE has an average effective spread of 9.1 cents per share versus 12.2 cents per share for the NYSE. The BSE and the CHX have similar effective spreads of about 14 cents per share, while the PSE has an effective spread of 16.5 cents per share. The Phlx shows a somewhat higher effective spread of 18.2 cents, almost double that of the CSE. However, when interpreting this table and Table V-6C, it should be kept in mind that the sample sizes are much smaller than those of Table V-6A. [See Tables V-3A, V-3B, and V-3C for a display of sample sizes by spread and order size.] Also, the differences may be affected by the factthat different exchanges trade different securities.

A comparison of Table V-6A and Table V-6B also presents a curious anomaly pertaining to the size of discrete spreads on U.S. exchanges. [Stocks trade on prices whose increments are usually an 1/8, such as 20, 20 1/8, 20 1/4. Because spreads are the difference of two prices, they too are discrete multiples of an 1/8, such as 1/8, 1/4, 3/8, etc.] In Table V-6A, which analyzes markets with a quoted spread of 1/8 point, the CSE's effective spread across all orders is 10.9 cents. However, in Table V-6B, which analyzes wider 1/4 point markets, the CSE's effective spread across all orders is a narrower 9.4 cents. Intuition suggests that effective spreads should increase when quoted spreads increase.

A likely explanation for the counter-intuitive result is the presence of a mid-point at which to trade in a 1/4 point market that serves to narrow the effective spread. In a minimum variation 1/8 point market, dealers may have difficulty providing price improvement (and narrowing the effective spread) because they are constrained by exchange rules concerning time priority [See supra Part II.B.1 (discussion of rules of priority, parity, and precedence).] and the priority of the limit order book over a specialist trading for its own account. [See , e.g. , NYSE Rule 92(b).] Thus, customers can receive better trade prices, on average, in markets with a quoted spread of 1/4 point than in an 1/8 point market, but only because the 1/4 market has a feasible mid-point for a trade price and the 1/8 point market does not. The lower effective spread in 1/4 point markets is not caused by thewider spread, per se, but is a consequence of a spread midpoint on which to trade. A smaller minimum tick size would create a spread midpoint for the current minimum variation 1/8 point markets. The results of Tables V-6A and V-6B suggest that the presence of such a spread midpoint would lower effective spreads, and thus, trading costs. [The small sample sizes analyzed in Table V-6C cause the results to be uninterpretable. The Table is included only for reasons of completeness.]

In making comparisons of execution costs between exchanges, an important tradeoff must be made. In order to maximize the sample size of the analysis, one may look at all trades on a particular exchange. The problem is that different exchanges trade different securities, which may in turn have inherently different trading costs. To control for this effect, one could analyze only those securities that are traded both on a given regional exchange and on the benchmark exchange. The cost of this comparability is a decreased sample size.

Table V-7 compares the effective spreads of five regional exchanges using the benchmark approach outlined above. The table is broken down into six panels (BSE is broken down into CSI and non-CSI stocks), each of which compares effective spreads on a regional exchange to spreads on the NYSE (for those securities that trade on both exchanges). For example, in the top panel of the table, the NYSE and the BSE have 45 CSI stocks in common for which there is sufficient trading to calculate reasonably reliable estimates of the average effective spreads for orders in the 100 to500 share range. The effective spread for such orders is 8.7 cents per share on the NYSE and 11.2 cents per share on the BSE. In the previous three tables, Tables V-6A, V-6B, and V-6C, the observed differences between the effective spreads of the NYSE and the BSE might have been affected by differences in the stocks traded on the two exchanges. The differences in Table V-7 do not have this limitation because only those securities that trade both on the NYSE and on a regional exchange are compared.

Across all exchanges in this matched sample, the NYSE displays a lower effective spread than the corresponding regional exchange in almost every case. This is indicated by the negative value in the column labeled "Difference" in the table. The exception to this is for large trades, which is likely due to brokers crossing institutional trades on the regionals.

It should be noted that the number of stocks in the actual comparison, the figure in the column "Number," may be lower than the number of stocks in common between the two exchanges. The reason is as follows. To provide for a reasonable level of statistical reliability, a stock was excluded if it had less than 100 trades overall or less than ten trades in an order size category on the regional exchange. Accordingly, a stock that had less than ten trades on a regional exchange, but perhaps 500 trades on the NYSE, is excluded from the comparison.

Turning next to the price improvement data, Tables V-8A, V-8B, and V-8C present the percentage of trades in a given trade size group that are price improved in 1/8 point, 1/4 point, and 3/8point markets, respectively. For example, the upper left corner of Table V-8A shows that 15.4% of 100 share market orders on the NYSE received price improvement. [Price improvement for the BSE is reported for both CSI and non-CSI stocks unless otherwise noted in order to conserve space and increase the sample size. An analysis of CSI versus non-CSI price improvement in Table V-23 show that there is little difference in price improvement rates between the two groups of stocks.]

In Table V-8A, most of the interesting variation in price improvement rates takes place between exchanges rather than across order sizes. It is sufficient to look at the statistics for the "All Orders" results at the bottom of the table. In 1/8 point markets, it can be seen that the NYSE "price improves" about 13% of all market orders. The regional exchanges price improve considerably fewer market orders. In this regard, the CHX's price improvement rate is affected by the fact that orders in minimum variation markets are not eligible for its SuperMax and Enhanced SuperMax automated price improvement programs. [See CHX Rules, Art. XX, Rule 37(c) and (e). ] Thus, less than 4% of all market orders are price improved on the CHX in 1/8 point markets.

Recall from Table V-1 and Table V-3A that on average, about 80% of the trading activity on regional exchanges takes place in 1/8 point markets. The CHX, Phlx, and the PSE noticeably underperform the CSE, and especially the NYSE, at such times. As discussed in Part II, the phenomenon of price improvement may notbe well defined in minimum variation markets. [See supra Parts II.B.4 and B.5.] If a market buy order is stopped against a specialist’s ask quote that represents a customer limit order (as opposed to a specialist’s proprietary quote) and is subsequently executed at a lower price, then the price improvement granted to the market order comes at the expense of the non-execution of the customer limit order at the ask. [Further, in a minimum variation market, it is difficult to characterize an order that is stopped against other customer limit orders on the specialist's book as "price improved." See supra Part II.B.4.] Whether the specialist’s quotation is composed of customer limit orders or is proprietary is an empirical matter that could not be addressed within the time constraints of this study.

Tables V-8B and V-8C show that this picture changes considerably in 1/4 point and greater markets. The disparity between the regional exchanges and the NYSE closes considerably. Looking at "All Orders" in 1/4 point markets, one can see that the CSE’s price improvement rate exceeds that of the NYSE.

Recall that price improvement is defined as the difference between the execution price of an order and the quotation measured at the time the order arrived at the specialist’s post. On occasion, the specialist will not be able to act on the order before quotations change. In that case, it is possible that when the trade is executed at the revised quotations, negative price "improvement" occurs. For example, suppose the market was quoted 20 to 20 1/4 when an order to buy arrived at the specialist post. Five seconds later, the quote changes to 20 1/8 to 20 3/8. The specialist then looks over at the display screen and executes the order to buy at the prevailing ask price, 20 3/8. This order would receive negative price "improvement" relative to the 20 1/4 offer prevailing at order arrival time. [The delay between order arrival and order execution time leads to negative price "improvement." Notice also that this delay could result in positive price improvement as well. If the negative is not netted against the positive, the total price improvement rate will be inappropriately overstated for the market as a whole.] Tables V-8A, V-8B, and V-8C all have the general trait that marketable limit orders receive less price improvement than normal market orders. This is somewhat of a puzzle. For the CHX, this may be due to the exclusion of marketable limit orders from the SuperMAX and Enhanced SuperMAX automated price improvement programs. Due to time constraints, the only exchange on which the effect was examined directly was the NYSE. The data show that both the specialist and the trading crowd provide less price improvement to marketable limit orders than they provide to market orders. Also, market orders appear to provide less price improvement to marketable limit orders than to other market orders.

Tables V-8A, V-8B, and V-8C only calculate positive price improvement figures. To supplement those tables, Tables V-9A, V-9B, and V-9C calculate the net price improvement rate. To do this, the rate of negative price improvement is subtracted from the positive price improvement rates of Tables V-8A, V-8B, and V-8C. The results indicate that the general relative patterns of priceimprovement do not change, but net price improvement rates are lower by two to three percentage points than those based only upon positive price improvement.

Tables V-10A, V-10B, and V-10C calculate the net dollar value of price improvement granted to orders of a given size in markets of various spreads. The numbers in the table are expressed in dollars per trade. These are the average dollar benefits per trade received by customers from price improvement. Thus, in 1/8 point markets, customers receive on average $4.66 of price improvement for market orders of 401 to 500 shares on the NYSE, while they receive $1.72 of price improvement for such market orders on the CHX.

Table V-10B shows that the dollar value of the price improvement rises sharply as the spread increases. In 1/8 point markets, the NYSE again shows superior execution costs, as expected, but this advantage is eliminated in 1/4 point and greater markets.

As done for effective spreads above, Table V-11 presents a matched sample analysis of price improvement rates for executions on the regional exchanges in comparison to the NYSE benchmark. In such a calculation, there is little variation among the regional exchanges: they all give price improvement around one-half to one-third as often as the NYSE. As discussed above, each comparison is done only for those securities that trade on the two exchanges being compared.

There is also a possibility that the regional exchanges are concentrating all of their price improvement in a few stocks, while largely giving executions at the NBBO to all other shares. To investigate whether this is the case, Table V-12 calculates the number of stocks for which the price improvement rate on a regional exchange is greater than the price improvement rate on the NYSE. If price improvement is evenly distributed among stocks on each exchange, then this number should be 0%. If the price improvement is clustered into a small number of securities, then that, too, should be reflected in the table.

Table V-12 shows that 13 of 98 BSE non-CSI stocks have a greater price improvement rate on the BSE than on the NYSE for trades in the 100 to 500 share range. Recall from Table V-11 that the NYSE has roughly twice the price improvement rate as the BSE in such a size category. Thus, there appears to be considerable non-uniformity in the distribution of this price improvement. One explanation for this may be that the regional exchanges specialize in price improving a subset of the securities they trade in common with the NYSE. Another possible explanation may be that these distributions result from a random process and, therefore, have no special meaning.

Table V-13 presents yet another dimension of price improvement, looking at the number of transactions and the fraction of an exchange’s total price improvement that is achieved in 1/8 point and in 1/4 point markets. In this regard, Table V-13 shows that almost 90% of the trading on the CSE takes place in 1/8 pointmarkets. Fully half of the CSE’s price improvement occurs in such markets, with the rest occurring in 1/4 point markets. For the NYSE approximately 82% of trading takes place in 1/8 point markets but less than half of the price improvement occurs in those markets. The CHX, BSE, Phlx, and PSE data show that most of their price improvement is provided in 1/4 point markets. The table indicates that the conditions under which price improvement is granted to orders varies across markets.

To achieve maximal comparability, Tables V-14, V-15, and V-16 report effective spreads and price improvement statistics for only those stocks that trade in common across all exchanges. This analysis is done at considerable sacrifice to sample size, but it results in an "apples to apples" comparison of trading. The most important thing to notice is that the primary messages of the previous tables do not change. The NYSE has lower effective spreads and higher price improvement rates than the regional exchanges. The result in Tables V-8B and V-9B that showed higher price improvement on the CSE than on the NYSE does vanish, however. Of the regional exchanges, the CSE appears to offer the highest price improvement rate and lowest effective spread, though often the magnitude of the differences in effective spread among the regional exchanges are not economically significant. Note, however, that the absolute level of price improvement does have economic value to the trades executed on those exchanges.

E.Limit Order Execution Analysis and Dealer Routing Decisions

Estimating the quality of limit order executions on a particular exchange is a more complex problem than that posed by market orders. In particular, because limit orders are priced orders, the execution price cannot be used to assess a limit order’s execution quality. The trade is required to occur at the limit price. [Technically, for the exchanges, the trade must occur at the limit price or better. As a practical matter, there are few executions of limit orders at prices better than the limit price.]

In this study, execution quality of limit orders is measured in three ways. The first method is analyzed in Table V-17, which calculates the probability that a limit order is executed. Of course, the probability of a limit order execution depends primarily on the price of the limit order relative to the market price for the shares. If the market is currently quoted at

20 - 20 1/4 and a limit order to buy is placed at 19, there is a significant chance the order may not execute for a very long time. As the limit price is raised, the likelihood of an execution rises. In the extreme, if the limit buy price is set at 20 1/4 or greater, the limit order becomes a marketable limit order and should be executed immediately, just as would be the case for a market order. [Of course, there are other factors that determine the likelihood of a limit order executing. These factors include the amount of market order activity, the information asymmetry present in the market at a given time, and time of day and volume effects. Complete analysis of all such factors is beyond the scope of this study.]

To control for this effect, Table V-17 reports the results of limit order fill rates by breaking out limit orders into three groups. The first is marketable limit orders. The second group is quote improving limit orders, which are orders whose price lies inside the quoted bid-ask spread at the time the order was placed. The third group is at-the-quote ("ATQ") limit orders, which would be a buy limit order with a limit price set equal to the prevailing bid or a sell limit order with a limit price set equal to the prevailing ask.

For quote improving limit orders, there is a striking difference in the probability of execution among the exchanges. On the CSE, over 90% of quote improving limit orders are filled, while on the NYSE this number is 79%. The high fill rate of limit orders on the CSE carries over to ATQ limit orders as well. For marketable limit orders, fill rates on some exchanges are lower than expected. This may be due to data errors, such as incorrectly coding a stop-loss order as a limit order.

One reason for the striking difference between the CSE and NYSE fill rates is the depth of the NYSE limit order book. If relatively few limit orders are placed on the CSE and there is sufficient market order traffic, these orders have a very good chance of executing. If routed to the NYSE, however, the orders would have to take their place behind all other (same-priced) limit orders entered prior in time, as well as any trading interestrepresented on the floor of the exchange. In addition, the CSE preferencing dealers may be forced to "take out" a limit order to enable them to trade as principal with their customers’ market orders. [See supra Part I.B.3.]

Table V-18 analyzes only ATQ limit orders and breaks the results of Table V-17 out by order size. The differences between the exchanges are striking. Fill rates for all orders range from 30% to 80%, depending on the exchange and the quoted spread, with the CSE providing the best fill rates of any market.

Table V-19 examines a second aspect of limit order execution quality: time to order execution. To the extent that, conditioned on the order executing, limit order customers prefer rapid to slow executions, average execution time is a valid measure of market quality. The top panel of Table V-19 shows that marketable limit order execution time is generally very short. However, it ranged to seven minutes for some orders on the CHX. The bottom panel of the table shows the execution time for ATQ limit orders. Note that the NYSE, BSE CSI stocks, and the CSE have the quickest time for execution of such orders. From previous tables, it is known that the likelihood of such an execution is low on the NYSE. But those limit orders that are executed, are executed quickly. The slowest limit order executions appear to occur on CHX. In this case, ATQ limit orders took over 45 minutes to execute, on average. Table V-20 breaks out the time to execution for ATQ limit orders by order size.

The last of the three measures of limit order market quality examined in the study is the ex-post cost of a limit order execution. The price the limit order pays is established by the limit price. Limit orders are naturally buying low and selling high, competing with the specialist for this right. However, conditioned on a limit buy order executing, there is a substantial probability that the market price for those shares will continue to fall through the limit price and keep falling for some period of time. Thus, having bought shares on the bid with a limit buy order, one might be curious to know what the price is for those shares at some time in the future. It is well known that placing a limit order with a specialist is tantamount to granting the specialist an option to lay off or purchase shares from the limit order trader. The specialist may profit from that opportunity as follows. If the specialist knows that the price will fall in the near future, the specialist may allow market sell orders to hit bids on the limit order book instead of buying for its own account. This leads to a kind of winner’s curse associated with limit order execution and commonly is known as the adverse selection cost.

One measure of the adverse selection cost of limit orders is the difference between the limit price of an executed order and the price of shares at some time in the future. If the adverse selection problem is severe, the buy (sell) limit order trader will only buy (sell) shares before a market decrease (increase). The difference between the future market price and the limit price will be negative. If there is no adverse selection and prices arerandom walks, then the price difference will be closer to zero. It is difficult to determine how large this difference should be in a fair market, but it is true that the more negative the difference the worse the adverse selection problem. Therefore, the difference between the future market price of shares and the limit price can be used to develop a ranking of limit order execution quality among exchanges.

Tables V-21 and V-22 look at the adverse selection problem. It is very difficult to draw any consistent conclusions from these two tables. Note that all ex-post execution costs are negative, indicating the adverse selection problem does exist. There does not appear to be any clear difference between the exchanges in terms of the adverse selection costs faced by limit orders.

The results of the analysis of adverse selection costs are not particularly sensitive to the length of time used to measure the cost. Though the prices used as proxies for market price are taken by measuring the same side quote five minutes after the trade, the results do not change appreciably if prices are observed 15 minutes or 60 minutes after the trade.

In summary, there appear to be no notable differences in ex-post limit order execution costs between exchanges.

F.Conclusions

This section of the study analyzed trading activity on the NYSE (the primary market), the CSE and the BSE (regional exchanges with formal preferencing programs), and on the CHX, PSE, and PHLX (regional exchanges without formal preferencing programs). Thepurpose of the section was to quantify, to the extent possible:

1) the effect of preferencing on the overall market quality of the preferencing exchanges; 2) the effect of preferencing on the costs of market orders executed on a preferencing exchange; and 3) the effect of preferencing on limit orders executed on all exchanges. The market quality statistics based on quoted spread and quoted depth indicate better performance for the NYSE than the regional exchanges. Because there is two to three times more trading activity on the NYSE than on all regional exchanges added together, this result is not surprising. The regional exchanges show considerable dispersion in their quoted spreads and the quality of quotations on the preferencing exchanges is not uniform. The CSE quoted the tightest spreads of all regional exchanges, and as measured by the percentage of the time the quotes were at the NBBO, the CSE provided more competitive quotes than the other regional exchanges. The BSE, however, did not quote tight spreads, and, relative to the NBBO, it rarely provided competitive quotations to the market. With regard to quote depth, the CSE provides greater depth than other regionals most of the time, but unlike the BSE, its depth did not increase when quoting at the NBBO.

For market order executions, the two preferencing exchanges, along with the NYSE, were found to have smaller effective spreads than the other exchanges. Moreover, the CSE was found to have the best price improvement rate of the regional exchanges, followed by the BSE, CHX, PSE, and PHLX. Although most trades occur in 1/8point markets where the NYSE has a slight price improvement advantage, [As noted earlier, price improvement is suspect certain types of executions in minimum variation markets. See supra notes - and accompanying text.] as spreads widen to 1/4 point or more, the regional exchanges offer comparable executions to the NYSE, and in some cases, offer superior executions. The number of trades at spreads wider than an 1/8 is small, composing only about 17% of the sample across all exchanges.

The picture is slightly different for limit orders. The NYSE is the primary market and receives most of the limit order traffic. Limit orders generally are executed subject to price and time priority and the NYSE execution rate of limit orders submitted at the prevailing quote is about 45%. On a regional exchange, the probability of an at-the-quote limit order executing ranges from about 50% on the CHX to over 80% on the CSE. There is no evidence that limit orders are subject to greater adverse selection (i.e., a likelihood of being "picked off") by receiving an execution on a regional exchange rather than on the NYSE.

The data thus far analyzed do not support the view that preferencing has harmed the execution of market orders or limit orders. Market orders traded on preferencing regional exchanges tend to trade more favorably relative to the NYSE than market orders placed on non-preferencing regional exchanges. In addition, limit orders have a greater probability of executing on regional exchanges than on the NYSE. Further, the data do not indicate that preferencing has harmed the market quality of the preferencingexchanges. Indeed, the CSE often provides competitive markets. However, these results do not conclusively demonstrate that preferencing has improved the marketplace because one cannot observe directly the effect of preferencing on quote-based competition.

Tables

Table V-1: Number and Distribution of Trades in Sample by Order Size for Each Exchange

Table V-2: Number and Distribution of Market and Limit Order Executions by Order Size for Each Exchange

Table V-3A: Number of Sample Trades in 1/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-3B: Number of Sample Trades in 1/4 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-3C: Number of Sample Trades in 3/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-4: Average Quoted Bid-Ask Spreads and Depths by Exchange for Stocks that Trade on Both the NYSE and Regional Exchanges

Table V-5: Regional Bid-Ask Quotes and Depths Compared to NYSE Quotes and Depths

Table V-6A: Average Effective Spreads in 1/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-6B: Average Effective Spreads in 1/4 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-6C: Average Effective Spreads in 3/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-7: Average Effective Spreads (Stock Weighted) for Market Orders in Stocks Traded on the NYSE and the Regional Exchanges Broken Down by Order Size

Table V-8A: Frequency of Price Improvement in 1/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-8B: Frequency of Price Improvement in 1/4 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-8C: Frequency of Price Improvement in 3/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-9A: Net Price Improvement Rates in 1/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-9B: Net Price Improvement Rates in 1/4 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-9C: Net Price Improvement Rates in 3/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-10A: Average Dollar Per Trade Net Price Improvement in 1/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-10B: Average Dollar Per Trade Net Price Improvement in 1/4 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-10C: Average Dollar Per Trade Net Price Improvement in 3/8 Point Markets by Exchange and Order Size for Market and Marketable Limit Orders in NYSE-Listed Issues

Table V-11: Average Price Improvement Rates (Stock Weighted) for Market Orders in Stocks Traded on the NYSE and Regional Exchanges

Table V-12: Number and Percent of Stocks with Greater Price Improvement on a Regional Exchange than on the NYSE

Table V-13: Distribution of Market and Marketable Limit Order Trading and Price-Improved Trades by Quoted Spread at Order Entry Time for Each Exchange

Table V-14: Matched Sample (Stock Weighted) Price Improvement Statistics for Market Orders in Stocks Traded on the NYSE and on Each Regional Exchange Broken Down by Order Size

Table V-15: Average Effective Spreads (Stock Weighted) for Market Orders in Stocks Traded on the NYSE and on Each Regional Exchange Broken Down by Order Size

Table V-16: Percent of Stocks with More Price Improvement on the Regional Exchange Compared to the NYSE for Stocks Traded on the NYSE and on All of the Regional Exchanges

Table V-17: Fill Rates for Limit Orders for Each Exchange

Table V-18: Fill Rates for At-the-Quote Limit Orders for Each Exchange

Table V-19: Average Time from Order Entry to Execution for Limit Orders Entered on Each Exchange

Table V-20: Average Time from Order Entry to Execution for At-the-Quote Limit Orders Entered on Each Exchange

Table V-21: Average Ex-Post Transaction Costs for Limit Orders Entered on Each Exchange

Table V-22: Average Ex-Post Transaction Costs for At-the-Quote Limit Orders Entered on Each Exchange

Table V-23: Boston Stock Exchange: Percentage of Trades Price Improved for CSI Stocks and Non-CSI Stocks

VI.Findings and Conclusions of the Study

After undertaking a thorough analysis of the practice of "preferencing," as defined in Section 510(c) of the NSMIA, the Commission has concluded that preferencing has not had a deleterious effect on the national market system. To the contrary, market data for preferencing exchanges is at least as good as that of other regional exchanges, and in some cases, is better. Moreover, preferencing has furthered the ability of the CSE and BSE to compete in the national market system. Furthermore, the Commission has concluded that preferencing is not necessarily inconsistent with the best execution of customer orders.

Through a review of the trading practices permitted under the rules of the various exchanges, the Commission has found that preferencing is but one method to accommodate the internalization of order flow. As demonstrated in Part II of the report, internalization exists in all markets and has manifested itself in many variations over the years. Even now, both primary exchange and regional exchange rules permit a specialist to engage in a significant amount of dealer activity and provide a number of means through which time priority among orders can be avoided. Further, preferencing is merely one means utilized by broker-dealer firms to capture market maker profits from their customer order flow.

Internalization is made possible by the lack of time priority for orders across exchanges. The absence of such intermarket time priority allows broker-dealers to route their customer order flow to a particular exchange and, in many instances, execute such orderflow as principal regardless of "same-priced orders or quotations entered prior in time" on other exchanges. Preferencing programs merely allow this to occur on an intra-exchange basis. CSE's preferencing program and BSE's CSI are the latest manifestations of internalization. As these preferencing programs have made both exchanges more competitive, they have attracted criticism from competitors.

In recent years, the Commission has examined practices such as preferencing and other inducements to order flow and found that such practices are not necessarily inconsistent with best execution of customer orders. Nevertheless, a broker-dealer that automatically routes order flow to a particular exchange has the responsibility to regularly and rigorously evaluate the executions received by customers in that market. Such an evaluation takes on particular importance when preferencing or other inducements for order flow are involved.

The Commission undertook an analysis of CSE and BSE firms' order handling practices and compared those to other arrangements between broker-dealers and either their affiliated or joint venture specialist units. The Commission found no significant variance among the order handling procedures of preferencing and non-preferencing firms. Broker-dealer firms frequently route their retail order flow to captive market makers. However, firms generally review execution quality across various market centers on a regular basis. As noted above, such a review is a necessary prerequisite to ensure that preferencing is not interfering withthe firm's best execution obligation.

The Commission also undertook a thorough examination of market quality and execution data to respond to both the general and specific issues posed by Congress in Section 510(c) of the NSMIA. Section 510(c)(1) generally requested an analysis of the impact of preferencing on investors and the national market system, along with setting forth specific matters for the Commission to consider in its analysis. The Commission could find no evidence from the measures of market quality it analyzed to suggest that the CSE's preferencing program and the BSE's CSI have had a harmful impact on investors and the national market system. The Commission's analysis found that the CSE's quotation spreads were narrower than those of the other regional exchanges. The BSE's quotation spreads for stocks were wider than the CSE's but generally comparable to those of the other regionals. The CSE's quotation depth was found to be better than that of the other regionals, except when the CSE was at both sides of the NBBO. In such instances, it was comparable to that of most regionals. The BSE's quotation depth was equivalent to, or exceeded that of, other regionals when the BSE was at one or both sides of the NBBO. However, the BSE was only at one or both sides of the NBBO 5% of the time. The Commission believes that the BSE's quote performance should improve as a result of its implementation of quote entry capability for competing specialists in March 1997, and its future implementation of automatic routing of orders to the competing specialist at the ITS/BBO with priority. The Commission will continue to monitor theBSE's progress in this area. In light of the foregoing, the Commission concludes that the adoption of preferencing programs generally has not harmed the market quality of both the CSE and BSE, while making these exchanges more competitive.

In Section 510(c)(1)(A)(i), Congress requested the Commission to consider how preferencing impacts the execution prices received by retail securities customers whose orders are preferenced. The data indicates that preferencing has not necessarily diminished the ability of customers to acquire quality executions of their orders. [It must be noted that nothing in this section should suggest that the execution quality provided by any particular market center necessarily satisfies a broker-dealer's best execution obligation for its order flow.] Both the CSE and BSE guarantee the execution of orders up to a specified size at the ITS/BBO or better. The size of this execution guarantee is significantly higher than the average size of an individual investor's order. Importantly, the rate of price improvement (i.e., the percentage of time an order is executed at a better price than the best quotes prevailing in the market at the time the order arrived at the market or market center) on the CSE surpasses that of the other regionals. In addition, the CSE's rate of price improvement compares favorably to that of the NYSE. The BSE's rate of price improvement in preferenced stocks is as good as or better than that of most other regionals. [It should be noted, however, that the Commission evaluated data for each exchange in the aggregate. It is likely that market making performance varies (perhaps substantially) among the specialists or dealers at each exchange. Consequently, our findings should not be taken to mean that every specialist or dealer at a particular exchange performs at the overall aggregate level for that exchange. Similarly, in making order routing decisions, broker-dealers need to consider that execution quality may vary across market makers at an exchange.]

In Section 510(c)(1)(A)(ii), Congress requested the Commission to consider the ability of retail customers in all markets to obtain executions of their limit orders in preferenced securities. The Commission's analysis found that the limit order fill rates (i.e., the percentage of time limit orders on an exchange receive an execution) on the CSE and BSE are higher than those of the NYSE. Further, the CSE's fill rate is usually better than that of other regionals. Moreover, the Commission found that limit orders sent to the CSE and to the BSE in CSI stocks usually were executed as quickly as those sent to the NYSE and more quickly than those sent to the other regionals. The Commission also found that, prior to the adoption of the Commission's Display Rule, the CSE and BSE had in place policies with regard to the exposure of customer limit orders to trading interest on their respective exchanges and to the national market system. Moreover, the Commission believes that the Display Rule will further enhance the exposure of customer limit order interest not only on the CSE and BSE, but on all exchanges, thereby leading to greater order interaction within the NMS. This increased order interaction will result in quicker and more frequent executions of customer limit orders in all markets.

In Section 510(c)(1)(B), Congress requested the Commission to consider the costs of preferencing to customers whose orders are preferenced. The Commission found no evidence of adverse costs for preferenced orders in the aggregate vis-a-vis other markets in the national market system. Further, the Commission's analysis has found no evidence that customers necessarily receive inferior executions of their orders on the CSE. Indeed, in almost all measures of market quality examined, the CSE and BSE performed as well as other regional exchanges and, in some cases, better than other markets. Accordingly, our analysis did not find that preferencing has harmed the quality of executions on the CSE or BSE. [The data collected by the Commission was used for the preferencing study, and should not be relied upon by broker- dealers in making order routing decisions. The Commission's data is already several months old and predated the effective date of the Order Execution Rules. In addition, the Commission analyzed selected measures of market quality in the aggregate. Broker-dealers, in regularly and rigorously evaluating execution quality across markets, should not use this report as a substitute for their own monitoring, but instead should conduct their own, current analyses of execution quality based on the types of orders they handle.]

Although the Commission does not believe that the preferencing of customer orders necessarily results in inferior executions, the Commission notes that such programs do raise significant agency-principal concerns. As the Commission stressed in the orders permanently approving the CSE's preferencing program and the BSE's CSI, the Commission believes that exchanges operating preferencing programs and broker-dealers participating in such programs need to continuously monitor the executions received by customers whoseorders are preferenced. Specifically, broker-dealers must regularly and rigorously evaluate the executions received by customers whose orders are preferenced.

While to date the Commission has found that these programs in aggregate do not appear to have had an adverse effect on either the executions received by customers whose orders are preferenced or the national market system as a whole, these findings should not be taken to mean that the Commission believes that such adverse effects may not arise in the future. Indeed, changed circumstances may result in a determination by the Commission that the findings it has made in this study require reconsideration. For example, a significant increase in the amount of preferencing activity as a percentage of overall national market system activity that results in a decline in execution quality on the national market system, or a significant deterioration in the surveillance and regulatory programs of an exchange operating, or broker-dealers participating in, a preferencing program would require reconsideration. Therefore, the Commission will continue to review the effects of the practice of preferencing, as well as those of other exchange practices that allow for time priority to be avoided, on the national market system on an ongoing basis to ensure that this practice is consistent with the maintenance of fair and orderly markets, the protection of investors and the public interest, and the furthering of the national market system goals of Section 11A of the Act.

(Go to Appendices)

http://www.sec.gov/news/studies/prefrep.htm