dark pools

7/30/2019 Dark Pools

1/86


2/86

2012 CFA Institute

CFA Institute is the global association of investment professionals that sets

the standard for professional excellence. We are a champion for ethical

behavior in investment markets and a respected source of knowledge in the

global financial community.

Our mission is to lead the investment profession globally by promoting thehighest standards of ethics, education, and professional excellence for the

ultimate benefit of society.

ISBN 978-0-938367-58-1

October 2012


3/86

DARK POOLS,INTERNALIZATION,

AND EQUITY MARKETQUALITY


4/86


5/86

2012 CFA INSTITUTE iii

Acknowledgments

Among the many contributions to this report, we would like to extend particular thanks toJerey Smith and Frank Hatheway, CFA, at NASDAQ OMX, or the provision o data orthis study and or helpul comments; Pro. Larry Harris, CFA, rom University o SouthernCaliornia Marshall School o Business, or helpul comments and suggestions; and DennisDick, CFA, ony an, CFA, Kapil Khetan, CFA, and Yasuhiro Oshima, CFA, who serveon the CFA Institute capital markets policy council, or their insightul input.


6/86


7/86

2012 CFA INSTITUTE v

Contents

Executive Summary 1

1. Introduction 8

2. Market Structure 10

3. Regulatory Framework 25

4. Literature Review 32

5. Data 39

6. Descriptive Statistics 43

7. Analysis 51

8. Conclusions and Policy Considerations 60

Appendix A 63

Appendix B 65

Appendix C 67

Appendix D 72

Bibliography 74


8/86


9/86

2012 CFA INSTITUTE 1

Over the past decade, the U.S.equity market has been transormed bythe orces o technology, regulation, andglobalization. oday, the way in whichinvestors, market participants, interme-diaries, and trading venues interact ishighly automated and critically depen-dent on speed. Signicantly, the domi-nance o the incumbent exchanges hasbeen eroded and liquidity has rag-

mented over numerous trading venuesas competition has intensied.

Within this ragmented environment,o-exchange trading, including broker/dealer internalization and dark pools in

which prices are not displayed priorto execution, has grown signicantly.Undisplayed or dark trading awayrom public exchanges is estimated toaccount or approximately 31% o con-

solidated volume as o March 2012agrowth o around 48% since the starto 2009.1

From a market integrity perspective, thegrowth in dark trading raises poten-tial concerns, ranging rom a perceiveddecline in the transparency o markets

1Figures based on data presented in able 1 inSection 2 o this report or March 2012 andon data rom Tomson Reuters Equity MarketShare Reporter or January 2009.

Executive Summary

to a reduced willingness o investors todisplay quotes i dark venues ree rideo those quotes and privatize order ow.Indeed, CFA Institute members haveraised concerns that the incentive todisplay orders in public markets is beingundermined by certain o-exchangetrading practices. In turn, these con-cerns have implications or public pricediscovery, liquidity, and the quality and

integrity o markets.

Regulators around the world havealso voiced concerns about dark trad-ing, including the U.S. Securities andExchange Commission (U.S. SEC)in its equity market structure conceptrelease (2010), the European Commis-sion in its review in 2011 o the Marketsin Financial Instruments Directive,2the International Organization oSecurities Commissions (IOSCO) inits report on dark liquidity (2011), as

well as regulators in Canada and Aus-tralia in their reviews o their respectivemarket integrity rules. Certain regula-tory proposals, including those o theU.S. SEC, remain under consideration.

2Visit http://ec.europa.eu/internal_market/secu-rities/isd/mid_en.htm or more inormation.


10/86

Dark Pools, Internalization, and Equity Market Quality

WWW.CFAINSTITUTE.ORG2

Tis report presents an examination o the relationship between dark trading and mar-ket quality in order to inorm public policy issues related to undisplayed liquidity and to

address the aorementioned market integrity concerns. Specically, we examine the rela-tionship between dierent types o undisplayed trading volumes and market quality mea-sures, including bidoer spreads and top-o-book market depth.

Te results o our analysis show that increases in dark pool activity and internalization areassociated with improvements in market quality, but these improvements persist only up toa certain threshold. When a majority o trading occurs in undisplayed venues, the benetso competition are eroded and market quality will likely deteriorate.

o protect market integrity, we recommend that (1) internalization o retail orders berequired to oer meaningul price improvement, (2) regulators monitor the growth in dark

trading and take appropriate measures i it grows excessively, and (3) dark trading acilitiesimprove reporting and disclosures around their operations to enable investors and regula-tors to make more inormed decisions about their use.

Summary of Findings

Market structure Te U.S. equity market structure is ragmented. oday, equity trading is dispersed across

13 exchanges, at least one electronic communications network (ECN), approximately

16 reporting dark pools, and more than 200 broker/dealers who internalize order ow.Exchanges collectively account or approximately two-thirds o consolidated volume, ando-exchange transactions account or approximately one-third o total volume.

Most exchanges are structured as electronic limit order book markets. Exchanges are gen-erally both pre-trade and post-trade transparent: Prices and trading interest are displayedprior to execution, and transaction details are publicly disseminated in real time.

ECNs operate similarly to exchanges in terms o secondary market trading o equity secu-rities, generally being both pre-trade and post-trade transparent. Tere is only one signi-cant ECNLavaFlowwhich accounts or approximately 1% o consolidated volume.

Dark pools are systematized execution acilities that operate with limited pre-trade trans-parency. Te prices o orders entered into the dark pool are not displayed to other marketparticipants and are matched anonymously against contra-side orders.


11/86


Executive Summary

Several dierent types o dark pools operate in the United States, ranging rom continuous-crossing systems operated by the large broker/dealers to independently operated block-

cross platorms. In aggregate, dark pools have accounted or between 8% and 13% o consolidated volume

over the past three years.

Internalization involves broker/dealers internally executing client order ow against theirown accounts on a systematic basis. Broker/dealer internalization is not subject to pre-tradetransparency.

Internalization and other over-the-counter (OC) transactions represent approximately18% o consolidated volume. Internalization is also thought to account or almost 100% oall retail marketable order ow.

Retail internalization is driven by the purchase o order ow by wholesale OC marketmakers rom retail brokerage rms. Tis practice enables broker/dealers to preerence thoseorders that are protable to arbitrage and route unwanted orders to other market centers.

Broker/dealer internalizers must match or beat the national best bid and oer (NBBO).Broker/dealers can provide price improvement to their customers in the orm o sub-pennyexecutions. Although this approach can provide savings to retail customers, it carries anopportunity cost to liquidity providers that post orders on public exchanges.

Other OC transactions include ad hoc, large, or irregular transactions between broker/dealers and other counterparties seeking to execute client order ow in the most efcient

manner possible.

Te process by which orders are handled, routed, and executed can be very complex becauseo the ragmented nature o the equity market and the reliance on advanced technologyand speed. Firms use algorithms to route dierent portions o an order to dierent venuesin various sequences, taking into account such actors as minimization o market impact,minimization o inormation leakage, immediacy o execution versus cost o execution in

various pools, and other actors to provide the most efcient executions possible.

Once trades are executed, they are immediately reported to the consolidated tapethemechanism or the provision o public post-trade transparency.

O-exchange transactions are reported to one o two main trade reporting acilities (RFs)that are registered and overseen by the Financial Industry Regulatory Authority (FINRA).

Te largest RF is operated by NASDAQ (the FINRA/NASDAQ RF), and the otherRF is operated by the NYSE (the FINRA/NYSE RF). rades reported through theRF are then printed on the consolidated tape.


12/86



Regulatory framework Te regulatory ramework surrounding the operation o exchanges is the Regulation

National Market System (Reg. NMS). Te key aspects o Reg. NMS include the AccessRule, which ensures that market participants have air and nondiscriminatory access to mar-kets and prices; the order protection rule, which protects displayed quotations at the bestbid or best oer rom being traded through; the sub-penny rule, which prevents exchanges,broker/dealers, and other market centers rom displaying, ranking, or accepting any ordersthat are priced in an increment o less than 1 cent or stocks priced above $1 (although itallows broker/dealers to execute transactions in sub-penny increments); and market datarules, which govern the allocation o revenues to market centers that contribute data to theconsolidated quote and tape.

Non-exchange-trading modalities that include dark pools, ECNs, and certain other bro-

ker/dealer systems are captured under Regulation Alternative rading System (Reg. AS).Alternative trading systems (ASs) are not required to publicly display price quotationsand are able to restrict access to their crossing systems and internalization pools.

Te issues associated with dark liquidity are prevalent internationally. A number o regulatorybodies in other jurisdictions have developed rameworks governing how dark pools and undis-played orders are allowed to operate within their markets. Additionally, IOSCO has establisheda broad set o international best practices or regulatory treatment o dark pools and dark orders.

Literature review

Overall, the academic literature most relevant to this study is at best mixed. Out o the studiesreviewed, the most applicable are Weaver (2010, updated 2011); Degryse, de Jong, and vanKervel (2011); Buti, Rindi, and Werner (2010a); and OHara and Ye (2011). Te rst twosupport the notion that undisplayed trading harms market quality, whereas the latter twosuggest undisplayed trading is associated with improvements in market quality. Tis divisionindicates that the relationship between undisplayed liquidity and market quality is complex.

Empirical analysis

Data

A sample o 450 stocks stratied across listing market and market capitalization wasselected by CFA Institute. For each stock, data on bidoer spreads, top-o-book depth,o-exchange volumes, and other variables were obtained or a selection o dates over theperiod rom the rst quarter o 2009 through the second quarter o 2011.


13/86


Executive Summary

O-exchange trades reported to the NASDAQ RF, which account or approximately95% o all o-exchange trading in our sample, have been subcategorized by NASDAQ

according to the type o trading modality used.

Descriptive statistics

Te data show that relative bidoer spreads have declined by approximately 50% over thereview period o January 2009May 2011. Te decline in spreads is evident among large-capitalization, medium-capitalization, and small-capitalization stocks. Te median quotedspread or large-cap stocks in our sample over the review period is 1 cent, or in relativeterms, 4 basis points (bps). For medium-cap stocks, the median spread is 2 cents or 9 bps,and or smallcap stocks, it is 9 cents or 83 bps.

op-o-book depth, measured by the average size (in shares and in dollars) at the best bidor best oer across all markets displaying at the NBBO, is relatively at over the reviewperiod. Median depth or large-cap stocks is 1,663 shares or $66,905, compared with 454shares or $5,964 or small-cap stocks.

Te market shares o internalization and dark pools have trended upward over the reviewperiod or the total sample and or each o the large-, medium-, and small-cap subsamples,reecting growth in undisplayed trading or all stocks. Internalization is higher amongsmall-cap stocks relative to large- and medium-cap stocks, whereas dark pools are moreactive in large- and medium-cap stocks relative to smallcap stocks.

Regression analysis o analyze the relationship between dark trading and market quality, internalization and

dark pool volumes (the independent variables o interest) and other explanatory variablesare regressed against bidoer spreads and depth (the dependent variables o interest),respectively. We test the hypothesis that there is no relationship between the proportions odark trading and market quality.

Te results or the total sample suggest that increases in dark trading are initially associated withimprovements in market quality. Bidoer spreads decrease and depth increases as internaliza-tion and dark pool activity increase. Te regression results illustrate an association between darktrading and market quality, but they do not denitively prove the direction o the relationship.

Te relationship between dark trading and market quality is likely quadratic. Tat is, beyonda certain threshold, it reverses such that market quality initially improves but then declinesas dark trading increases. Specically, we estimate that when a majority o trading in a stockoccurs in undisplayed venues, market quality will likely deteriorate.


14/86



One possible explanation is as ollows: Initially, competition or order ow among on-and o-exchange venues causes more aggressive quoting in the limit order book to obtain

incoming order ow. When lit markets dominate (i.e., dark market share


15/86


Executive Summary

2. Monitor growth in the proportion of dark trading volume, and take appropriate measures.

Regulators should monitor developments with respect to internalization and dark poolactivity. Regulators should consider introducing measures to restrict the use o dark ordersand dark trading acilities i such activity becomes excessive, such as i the share o darktrading exceeds 50%. One possible measure would be to lower the threshold at which ASsmust display orders and meet general access requirements rom the current level o 5% othe trading volume in a given stock.

Tis proposal is consistent with the IOSCO principles and is analogous to the recommen-dations o CSA/IIROC and ASIC to monitor growth in dark liquidity.

3. Improve reporting and disclosure around the operations of dark trading facilities.

Insufcient inormation about the operations o dark pools, internalization pools, thetypes o orders that are accepted within those systems, and the process by which ordersare matched makes it difcult or investors to make inormed decisions about whether orhow to utilize dark trading acilities. It also makes it harder or regulators to monitor theirgrowth (the second consideration) and to evaluate how dark pools aect price discoveryand liquidity. Dark trading acilities should, thereore, voluntarily reveal greater inorma-tion about their operating mechanics and report more inormation on the volumes theyexecute. Such disclosures would improve transparency and enable all stakeholders to betterunderstand their relative benets and drawbacks.

Implementation o these considerations would help protect displayed orders while oer-

ing meaningul savings to retail investors executing away rom public markets, maintaincompetition, and urther transparency. More undamentally, these measures would enhancemarket integrity and underpin investor condence in the equity market structure.


16/86


1. IntroductionSince the establishment o ormalized stock exchanges in the United States more than 200

years ago, equity markets have continuously evolved. But over the past decade, the paceo evolution has accelerated as a result o the combined orces o technology, regulation,globalization, and competition. ogether, these orces have transormed the structure andunctioning o equity markets. Te way in which investors, market participants, intermedi-aries, and trading venues interact is now highly automated, extremely ast, intensely com-petitive, and dependent on scale more than ever beore. In essence, the U.S. equity markettoday represents a vast, decentralized electronic network that is critically dependent ontechnology to generate and match order ow at great speed.

Te main trends characterizing the evolution o the market over the past decade4 include a

decrease in average trade sizes and a signicant increase in overall quote trafc and transactionvolumes (largely owing to the adoption o electronic systems that have improved operationalefciency and network capacity); a reduction in trading costs (both bidask spreads and com-missions); some pronounced periods o volatility; and increasing ragmentation o liquidity.

Regarding ragmentation, one o the most conspicuous trends has been the shit in tradingvolume rom the primary listing markets to new trading venues and, in particular, to o-exchange business as competition has intensied. Te market share o NYSE Euronext5 inNYSE-listed stocks, or example, ell to 25.1% o the consolidated share volume in October2009 rom 79.1% in January 2005.6 More generally, market shares or almost all exchangescontinue to trend downward while o-exchange volume increases.

O-exchange business can be broadly categorized into transactions that occur in alterna-tive trading systems (ASs), such as electronic communications networks (ECNs) and darkpools (undisplayed liquidity pools that acilitate the execution o orders in a systematized

way); broker/dealer internalization (internal execution o client orders against the broker/dealers own account on a systematic basisinternalization is the dominant trading modal-ity or retail orders, in which brokerages route retail order ow to an o-exchange marketmaker that may or may not be afliated with the brokerage); and other over-the-counter(OC) transactions. A common denominator o dark pool transactions, broker/dealerinternalization, and other OC transactions is their absence o pre-trade transparency. Tatis, orders are not publicly displayed prior to execution. Consequently, these types o trans-actions can be considered to represent dark liquidity. In contrast, on-exchange business

4See, or example, Angel, Harris, and Spatt (2010).5NYSE Euronext operates the NYSE (New York Stock Exchange) and NYSE Arca in the United States. Italso operates the NYSE MK, ormerly the Amex (American Stock Exchange), which caters to small growthcompanies.6See U.S. SEC (2010).


17/86


Introduction

is largely (though not exclusively) pre-trade transparent.7 Dark liquidity is estimated tohave grown by approximately 48% between January 2009 and March 2012, to account oraround 31% o consolidated volume.8

A number o concerns about the current market structure have been raised by both inves-tors and regulators. Tese concerns relate to a perceived degradation o transparency arisingrom the growth in dark trading and the corresponding all in the market share o lit ven-ues, an uneven playing eld between dierent types o trading venues and dierent classeso investors that could potentially distort competition and airness, and a greater potentialor systemic risk propagation as a result o the dependence on automation and the intercon-nectedness o markets.

A more undamental concern associated with the growth in dark liquidity is that the will-ingness o investors to post displayed limit ordersthe building blocks o price discoverycould be potentially harmed i a signicant proportion o orders are lled o-exchange atthe expense o the limit order submitter. At its worst, this scenario could disincentivizeinvestors rom displaying orders altogether.

Te question thus arises: What exactly do these market structure concerns mean or overallmarket integrity? More precisely, and more quantiably, what is the relationship betweendark liquidity and market quality? Tis issue is central to the debate on market structureand is the ocus o this report.

Specically, we examine the relationship between dierent types o undisplayed, o-exchange volumes and market quality measures, including bidoer spreads and top-o-book market depth.

Te objective o this report is to inorm public policy issues related to undisplayed liquidityand to address the aorementioned market integrity concerns. Te report contributes to theliterature by attempting to separate o-exchange volume into its component parts, therebyenabling one to isolate the eects on market quality rom dark pools and rom internalization.

Te rest o this report is structured as ollows. Section 2 provides an overview o the U.S.equity market structure. Section 3 addresses the key aspects o the prevailing regulatoryramework. Section 4 examines existing academic literature. Section 5 explains the data usedor this study. Section 6 presents descriptive statistics. Section 7 presents the results o theanalysis, and Section 8 concludes and oers policy considerations on the basis o the ndings.

7Most exchanges operate electronic limit order books, in which the top-o-the-book trading interest is pub-licly displayed in the consolidated quote stream i it constitutes the national best bid and oer (NBBO) ora stock, whereas the depth o trading interest beneath is displayed to participants o the exchange. Note thatexchanges do acilitate undisplayed liquidity as wellsee Section 2 or details.8Figures are based on data presented in able 1 in Section 2 o this report or March 2012 and on data rom

Tomson Reuters Equity Market Share Reporter or January 2009.


18/86


2. Market Structure

oday, equity trading in the United States is dispersed over 13 exchanges, at least one ECN,approximately 16 reporting dark pools, and more than 200 broker/dealers who internalize orderow. rading can be distilled into exchange transactions and o-exchange transactions, withthe latter encompassing ECNs, dark pools, internalization, and any other OC transactions.

Te ollowing sections present a review o each o these dierent types o trading venuesand trading modalities.9 Later in this section, the processes by which orders are handled,executed, and reported are also set out.

2.1. ExchangesMost registered exchanges are structured as electronic limit order book markets. Such mar-kets are multilateral (meaning that multiple buyers and sellers can trade against each other

within the system), oer nondiscriminatory access, and operate according to nondiscretion-ary rules and procedures.

Limit order book markets result in trades when an acceptable match between buy andsell orders occurs. Tis requires either a buy or a sell order to cross the spread between thehighest bid and lowest oer submitted so that acceptable prices or both buyer and seller

overlap. ypically, limit order books operate according to price and then time priority orthe sequencing o order execution. Pricetime priority ensures the air treatment o orders.

Te role o brokers in limit order book markets is limited to acilitating the execution oclient orders. But many exchanges include market makers who provide liquidity and ensuresmooth market unctioning. Te activities o liquidity providers supplement the interactiono customer orders and acilitate the operation o a continuous market.

Exchanges are generally both pre-trade and post-trade transparent: Prices and tradinginterest are displayed prior to execution, and transaction details are publicly disseminatedin real time. But exchanges do permit hidden order unctionality, meaning that undisplayed

or dark liquidity can reside in exchange systems. Such hidden liquidity on exchanges,

9See also U.S. SEC (2010) or a thorough account o U.S. equity market structure.


19/86


Market Structure

however, is reasonably small. For any given exchange, it typically accounts or less than 1%o the consolidated share volume. For all exchanges in aggregate, hidden liquidity accounts

or approximately 34% o the consolidated volume.10

Figure 1 illustrates the market share trends or the seven largest exchanges over the periodo January 2009April 2012. As noted previously, nearly all exchanges lost market shareover this period. Te corollary is that o-exchange trading has grown. Te only exchange

10Figures obtained rom Rosenblatt Securities.

Figure 1. Exchanges Market Share of Consolidated Volume,

January 2009April 2012

NASDAQ

NYSE Arca

NYSEBATS (BZX)

Boston (NASDAQ OMX BX)

Direct Edge EDGX

Direct Edge EDGA

Percent

30

25

20

15

10

5

0

1/09 4/127/094/09 1/1010/09 7/104/10 1/1110/10 7/114/11 1/1210/11

Source: Based on data from Thomson Reuters Equity Market Share Reporter.


20/86



that appears to have gained market share over this period is Direct Edge (EDGX andEDGA). Note, however, that prior to July 2010, Direct Edge was registered as an ECN, so

it does not appear in the market share statistics prior to this month. Consequently, its seriesarticially starts rom a base o zero and hence is not ully comparable.

A ull list o exchanges and their approximate average market share is provided inTable 1.

2.2. Electronic Communications NetworksECNs operate similarly to exchanges in terms o secondary market trading o equity secu-rities. Tey are multilateral electronic trading venues, typically structured as limit orderbook markets that are generally pre-trade and post-trade transparent. Te primary dier-

ence rom exchanges is that ECNs are typically regulated as alternative trading systems. Assuch, they do not provide the ull range o unctions that exchanges do (or example, theyare not listing venues) and are not subject to the same regulatory responsibilities, includingsurveillance and oversight requirements, as registered exchanges.

Most o the ECNs that competed with the primary exchanges in the 1990s have since beenacquired or merged with the large exchange operators. For example, Instinet, the largest ECNat one time, was partially spun o and merged with the Island ECN to create Inet, which waslater acquired by NASDAQ. Similarly, Archipelago was acquired by NYSE Euronext andnow operates as the exchange NYSE Arca. Consequently, today there is only one signicantECNLavaFlowwhich accounts or less than 2% o the consolidated share volume.

2.3. Dark PoolsDark pools are systematized execution acilities that operate without ull pre-trade trans-parency. Tat is, orders entered in the dark pool are not displayed to other market partici-pants and are matched anonymously against contra-side orders. Many dark pools do, how-ever, send out indications o interest (IOIs) to select market participants; such IOIs mayindicate the security, number o shares, and side (buying or selling interest), but no price.In that regard, dark pools can be considered to operate with a limited degree o pre-tradetransparency. Dark pools are registered as ASs or regulatory purposes.

Te primary economic purposes o dark pools, and dark orders in otherwise transpar-ent venues, are to reduce inormation leakage and minimize market impact costs. Darkpools also provide the possibility o price improvement and reduced transaction costs more


21/86


Market Structure

generally. wo ways in which dark pools realize these benets are by crossing osettingcustomer orders at the midpoint o the national best bid and oer, thereby saving on both

exchange ees and the bidoer spread, and by restricting access to undesired market par-ticipants (such as high-requency trading rms rom the perspective o the institutional buyside). For these reasons, such acilities have been historically popular or execution o largeblock orders, enabling investors to obtain efcient, low-cost executions or nonstandardtypes o business. However, today, order and transaction sizes in dark pools are, on average,broadly equivalent to those on transparent exchanges.

Tere are several dierent types o dark pools operating in the United States. RosenblattSecuritiesa well-established source o dark pool statisticscategorizes dark pools intoour groups: pools operated by bulge-bracket brokerage rms, pools operated by marketmakers, independent or agency pools, and consortium-sponsored pools. An alternative

classication is provided by ABB Group, which categorizes dark pools according to theirunction, namely: block-cross platorms, continuous-cross platorms, and liquidity-providerplatorms. Te last category corresponds to the dark pools run by the large market makers,such as Knight and Getco.11

Based on the Rosenblatt Securities and ABB Group classications, independent/agencyand consortium-sponsored dark pools are mostly block-cross platorms, meaning that peri-odic auctions take place within the system in which the volumes crossed at each pointare o a large size. Examples include Liquidnet (independent/agency) and LeveL AS(consortium sponsored). In comparison, broker/dealer-operated dark pools are most com-monly continuous-cross platorms, in which crossings are more requent and are typically

o smaller sizes. Examples o continuous-cross platorms include Credit Suisse Crossnderand Goldman Sachs Sigma X, the two largest dark pools in operation.

Crossing systems are automated systems that match order ow in an orderly or systematizedashion between counterparties using the system or network. Orders are typically crossed at apoint within the spread o the best bid and oer reerence prices. Some dark pools acilitate onlythe matching o customer-to-customer order ow, whereas others (typically those operated bybroker/dealers) allow customer order ow to also execute against the brokers own account.

Some dark pools acilitate the aggregation o liquidity rom dierent sources to deepen thepool o available liquidity, which makes them attractive or executing large orders. Addi-

tionally, access to certain dark pools may be limited to only certain counterpartiesorexample, clients o the bank that sponsors the dark pool or buy-side-only institutions.

11See also Zhu (2012) or a good overview o dark pools and their classications.


22/86



In aggregate, dark pools have accounted or between 8% and 13% o the consolidated vol-ume over the past three years.12 A list o dark pools and their approximate market shares,

along with other trading venues, is provided in able 1.13

Table 1. U.S. Equity Trading Venues Market Share of Consolidated Volume as

of March 2012

Venue

Average Market Share of

Consolidated Volume (%)

Exchanges

NASDAQ Stock Market 18.1

New York Stock Exchange 12.3

NYSE Arca 11.7

BAS BZX Exchange 8.3Direct Edge EDGX Exchange 6.3

NASDAQ OMX BX (ormerly the Boston Stock Exchange) 2.8

Direct Edge EDGA Exchange 2.7

BAS Y-Exchange (BYX) 2.6

NASDAQ OMX PSX (ormerly the Philadelphia StockExchange) 1.0

National Stock Exchange (NSX) 0.4

Chicago Stock Exchange (CHX) 0.4

NYSE MK (ormerly NYSE Amex/ American StockExchange) 0.2

CBOE Stock Exchange 0.2

otal exchanges 67.0

ECNs

LavaFlow 1.8

otal ECNs 1.8

Dark pools

Credit Suisse Crossnder 1.9

Goldman Sachs Sigma X 1.5

Knight Link 1.4

12Figures are based on data primarily rom ABB Group, Rosenblatt Securities, and NASDAQ.13Tere are more dark pools in operation than those presented in able 1, which only shows gures or those

venues that report their volume to ABB LiquidityMatrix. Any volume rom non-reporting dark pools is,thereore, accounted or in the residual OC category.


23/86


Market Structure

Table 1. U.S. Equity Trading Venues Market Share of Consolidated Volume as

of March 2012 (continued)

VenueAverage Market Share ofConsolidated Volume (%)

Getco GEMatched 1.2

Barclays LX 1.1

Deutsche Bank SuperX 0.8

UBS PIN 0.7

Knight Match 0.7

Morgan Stanley MS POOL 0.7

LeveL AS 0.6

Liquidnet 0.6

BIDS rading 0.5

Instinet Cross 0.5

Citi Match 0.5

ConvergEx Millennium 0.3

ConvergEx VortEx 0.2

otal dark pools 13.2

Broker/dealer internalization/OC

More than 200 rms 18.0

otal broker/dealer 18.0

otal 100.0

Note: LavaFlow ECNs executed volume includes FLOW matched shares and GOTO routed shares.

Approximately half of LavaFlow volume relates to matched shares on its system (FLOW), and

half relates to routed shares (GOTO). Therefore, FLOW matched shares are estimated to account

for 0.9% of consolidated volume. Estimations are based on data from www.lavatrading.com.

Sources: Based on data from TABB LiquidityMatrix, Thomson Reuters Equity Market Share

Reporter, www.lavatrading.com, and U.S. SEC.

2.4. Internalization/Retail Market MakingInternalization involves broker/dealers internally executing client order ow against theirown accounts on a systematic basis. Internalization is a bilateral orm o trade execution in

which the broker/dealer (the OC market maker) acts as the counterparty to all incomingorders, trading as principal and using its own risk capital. It is, thereore, classied as an o-exchange activity. Internalization represents a orm o dark liquidity because OC market


24/86



makers are not subject to any requirements to display quotes prior to execution. Further-more, similar to dark pools, these market makers may discriminate among the counterpar-

ties that they will accept orders rom.

Te practice o internalization is not regulated as such. However, rms conducting internal-ization are required to register with the SEC and adhere to applicable conduct o businessrules as established by the Financial Industry Regulatory Authority (FINRA), the indus-trys sel-regulatory body.

Broker/dealer internalization represents a signicant proportion o overall trading activity,as shown in able 1. Internalization is also thought to account or almost 100% o all retailmarketable order ow, whereby brokerages route marketable retail orders to a wholesaleOC market maker.

Broker/dealer internalization is typically driven by the purchase o order ow by wholesaleOC market makers rom retail brokerage rms. Many wholesalers have standing payment ororder ow agreements with retail brokerages, typically paying 0.1 cent per share or less to theretail brokerage or the order ow.14 Another variant o this model is one in which the marketmaker and brokerage are vertically integrated, afliated entities under the same corporate group.

Upon receipt o the order ow, the wholesale market maker typically lls orders internallyagainst its own account or, i an order is undesirable, routes it to other wholesalers, otherinternalization pools, other market centers, or exchanges or execution. Te routing algorithmis determined by such actors as the depth o liquidity available in other pools, other pay-

ment or order ow arrangements, trading venue access ees, and so orth.15 In this manner,internalization is also reerred to as preerencing, reecting the act that the retail marketmaker executes the orders it chooses under the terms o its pre-arranged agreement withthe retail brokerage rm and routes any unwanted orders elsewhere. Because retail investorsare typically less well-inormed than proessional or institutional investors, retail order owis very desirable to wholesale market makers. Teir inormation advantage means that theycan preerence those retail orders that are on the wrong side o the market (that is, againstthe direction o expected market movements in the short-term and hence protable to themarket maker) and route any other orders to other market centers. Tis practice has also beenreerred to as cream skimming in academic literature (see Section 4 or urther discussion).

14See U.S. SEC (2010).15Te standard take eethe ee charged when market orders execute against resting liquidity in exchangebooksranges between $0.0018 and $0.003 per share (ASIC 2011, p. 109; Rosenblatt Securities; U.S. SEC2010).


25/86


Market Structure

Table 2 shows the proportion o market orders routed to wholesale OC market makers or aselection o retail brokerages based on rms SEC Rule 606 public disclosures on order routing

practices. Te table illustrates that nearly all retail market orders are internalized and that suchinternalization activity is dominated by a handul o wholesalers. Although market orders arealmost exclusively internalized, some customer limit orders are routed to exchanges.16

Table 2. Proportion of Market Orders in NYSE EuronextListed Stocks Routed to Wholesale

OTC Market Makers for a Selection of Brokerages, 1Q2012 (%)

Brokerage

OTC Market Maker E*TRADE

TD

Ameritrade Scottrade

Vanguard

Brokerage

Services

Charles

Schwab

Edward

Jones

Raymond

James &

Associates

Citigroup Global Markets/Automated rading Desk 5.2 20.9 20.2 0.1 18.2 15.8

Knight Capital Americas 8.4 40.7 22.0 13.4 15.1

Citadel 17.1 76.0 18.6 35.7 0.2 30.0 16.1

UBS Securities 7.4 18.0 13.9 17.0 99.7 6.1 6.0

E*RADE Capital Markets 61.9 32.2 35.2

Other wholesalers 11.8

otal 100.0 94.0 94.1 94.9 100.0 99.9 99.9

Source: Based on information in SEC Rule 606 reports for each company ( means not disclosed).

Broker/dealers have to provide best execution or their clients, so the prices at which they

internalize order ow must match or beat the NBBO. Because there are no quoting obliga-tions or internalization, broker/dealers can provide price improvement to their customersin the orm o sub-penny executions, oten to the magnitude o $0.0001 per share.17 Tere-ore, although internalization can reduce trading costs, such savings are oten only nominal.For example, on a 100-share marketable buy order (most orders are or a ew hundred

16For example, the D Ameritrade Rule 606 report or 1Q 2012 shows 52% o limit orders in NYSE Euro-nextlisted stocks routed to NYSE Arca. Similarly, the Scottrade Rule 606 report shows 44.7% o limit ordersrouted to LavaFlow ECN and 31.5% o limit orders routed to Direct Edge EDGX Exchange or NYSEEuronext-listed stocks. However, E*RADE, Vanguard Brokerage Services, Edward Jones, and Raymond

James & Associates routed all o their limit orders in NYSE Euronextlisted stocks to wholesalers. Te rela-tive proportions o market orders and limit orders handled by these brokerages vary. In some cases (e.g., Scot-trade, Vanguard, Raymond James), the proportion o market orders is broadly equivalent to the proportion olimit orders or orders in NYSE Euronextlisted stocks.17See Section 3 or details on the sub-penny rule. See also CFA Institute (2010b) comment letter to the U.S.SEC on sub-penny trading.


26/86



shares or less) or a stock trading at $19.99 bid$20.01 oer, the OC market maker mightll the order at $20.0099, saving the investor $0.0001 per share rom the oer price or a

total saving o 1 cent on the $2,000 transaction.

One way in which broker/dealers may prot rom retail transactions is by sweeping themarket with immediate-or-cancel (IOC) contra-side orders to execute against hiddenliquidity at exchanges residing between the NBBO. Continuing our example rom theprevious paragraph, suppose that there are hidden limit sell orders at $20.00 resting atexchanges while the displayed NBBO remains $19.99 bid$20.01 oer.18 Having sold at$20.0099 to the retail investor, the OC market maker may send out IOC orders to buyall the undisplayed liquidity oered at $20.00, thereby replenishing its inventory (assumingthe hidden size oered is sufcient to cover the market markers short position) and prot-ing by $0.0099 per share on the round-trip transaction. Internalization can become very

protable when one considers the requency with which the process occursa round-triptransaction takes a matter o milliseconds.

Other sources o potential prot are ickering quotes and latency arbitrage.19 Te order pro-tection rule o Regulation NMS (see Section 3) contains a provision exempting ickeringquotesdened as the least aggressive ask and bid quotes over the previous one secondrom the evaluation o a trade through. Broker/dealers with aster data eeds than the publicconsolidated quote stream may be able to arbitrage between the stale NBBO seen by slowtraders, such as retail customers, and the updated NBBO seen by ast traders with proprietarydata eeds within the permitted one-second window beore a trade through is enorced.

Using the previous example, suppose two exchanges are at the displayed NBBO o $19.99bid$20.01 oer. Suppose urther that a broker/dealer has a resting limit order oering$20.01 at Exchange A but the best oer quickly updates to $20.00 at Exchange B. A retailinvestor with a slower data eed sees the best oer still at $20.01 and submits a market-able buy order within the one-second window (making the least aggressive oer o $20.01icker compliant). An OC market maker can match the oer o $20.01 at Exchange

A to ll the incoming marketable buy order and then quickly lit the oer o $20.00 atExchange B, making a prot o $0.01 per share without violating trade-through rules. Tisexample suggests that regulators should consider the appropriateness o the one-secondtime window under the icker quote exemption because o the ease it aords relativelymore sophisticated market participants to prot rom it.

18Numbers are or ease o illustration only.19See, or example, McInish and Upson (2011).


27/86


Market Structure

Moreover, although internalization does provide price improvement, it carries an opportu-nity cost to liquidity providers that must be balanced against the aorementioned savings

provided to retail investors. Wholesale broker/dealers are eectively provided with an optionto trade against incoming marketable order ow. Only i they choose not to exercise thatoption does the order ow get routed elsewhere. Consequently, the order ow eventuallyreaching the exchanges is ltered, meaning that there are ewer opportunities or displayedliquidity providersthose posting limit orders on exchangesto trade against retail orderow.20 Investors who submit passive limit orders on exchanges take on risk by displayingtheir trading intentions; however, that risk may not be commensurately rewarded i theinterception o marketable order ow by retail market makers leaves limit orders unlled, orlled only because those limit orders are on the wrong side o the market. At its worst, thissituation could disincentivize liquidity providers rom displaying quotes altogether, whichmight have negative repercussions or price discovery and overall market integrity.

Te process o internalization or marketable retail orders is summarized in Figure 2. Solidarrows represent the routing o order ow, and dashed arrows represent possible payments.

2.5. Other OTCAny other o-exchange transactions between broker/dealers and their counterparties that donot belong to the aorementioned activities can be simply classied as OC. Such transactionsinclude ad hoc, large, or irregular transactions between broker/dealers and other counterpar-ties seeking to execute client order ow in the most efcient manner possible. Broker/dealers

conducting such OC transactions may use their own inventory to meet their clients specicneeds and then lay o their risk exposure in subsequent trades with other counterparties.

Many OC transactions represent technical trades that are non-price-orming. Such tradesdo not represent real or addressable liquidity and thus are reported with a special identi-er that excludes them rom ofcial volume gures. An example is a guaranteed volume-

weighted average price (VWAP) trade in which the executing broker undertakes a serieso transactions on the open market and then ips the shares back to the customer at theguaranteed price. Only the original open market transactions are included in the publicconsolidated tape inormation.

20In July 2012, the SEC approved the NYSEs Retail Liquidity Program, in which retail orders sent to theexchange will execute against undisplayed retail price improvement orders residing within the exchange sys-tem. Te Retail Liquidity Program is akin to internalization in that retail orders will largely not interact withdisplayed orders within the limit order book o the exchange.


28/86



Figure 2. Retail Market Making

Retail investor

Retail marketableorder flow

Order flow

Execute internally

Yes

Commission (typically $10 per trade)

Payment for order flow ( $0.001 per share);

Possibly passed through to client via reduced commission

No

Retail brokerage firm

OTC market maker

Other market centers(dealers, dark pools,

exchanges)

Order

desirable?


29/86


Market Structure

2.6. Institutional Order Processing

As the preceding discussion suggests, the process by which orders are handled, routed, andexecuted can be very complex, such is the ragmented nature o the equity market and thereliance on advanced technology and speed.

At the institutional level, the chain o events rom order origination to execution is partlydependent on the scale and resources o the buy-side institution originating the order. It iscommon or buy-side rms to send their orders to their broker/dealer to handle the rout-ing and execution, although many large buy-side institutions operate their own systems tohandle this process as well.

Te vast majority o institutional orders are processed using computerized algorithms

(algos). Tese algos take the institutional order (the parent order) and spray pieces o it(child orders) into the market to minimize the price impact that would otherwise occur ithe ull order were dropped into the marketplace too quickly.

Algos determine the parameters o the order (what, when, where, how much, etc.), taking intoaccount such actors as minimization o market impact, minimization o inormation leakage,immediacy o execution versus cost o execution in various pools, and other market-speciccircumstances to provide the most efcient executions possible. In this respect, algos perormthe role ormerly ullled by oor brokers or the upstairs trading desks o sell-side rms.

Sell-side rms provide a sophisticated range o order management and execution services to

their clients. In addition to various algos, these services include smart order routing systems(SORs) that seek out liquidity across multiple venues and help optimize execution as wellas direct electronic access to exchanges, dark pools, and other market centers. Downstream,exchanges have also developed their own SORs and other technological innovations, both tosell to market participants and to handle the order ow routed to the exchange in question

when it is not immediately or ully executable within the exchanges own pool o liquidity.

Te use o sophisticated algos to slice orders into small sizes and spread them across multi-ple venues to eect transactions partly reects the need to compete with other sophisticatedalgorithmic traders, including high-requency trading (HF) rms that dominate tradingactivity in the equity market. An increasingly common eature o the algos developed by the

sell-side or their clients is the inclusion o anti-gaming logic and allocation optimization.Such eatures are designed to minimize the risk o predatory algorithms snifng out largerblocks o liquidity that typically reside behind child orders and almost instantaneouslyadjusting public quotes to prot against the incoming order ow.


30/86



A simplied example o the process by which orders are routed is illustrated in Figure 3.Te exact routing process is dependent on such actors as the type o order, price, timing

(such details are omitted here or simplicity), and minimization o inormation leakagealong each step o the process. Te chain o events depicted in the gure begins with thebuy-side institution sending the order to its broker; as noted earlier, however, large buy-sideinstitutions that employ their own algos may skip this step o the process and route directlyto the various market centers, oten using broker-sponsored platorms that provide directelectronic access to dierent trading venues.

Figure 3. Example of Order Routing

Buy-side desk:

Sell 15,0000 sharesXYZ stock

Sell-side desk:

Sell 15,000 sharesXYZ stock

800 shares 700 shares 500 shares

15,000shares

2,000 shares

1

Dark pool 1:XYZ 5,000available



Exchange A:XYZ bid size

2,200

Exchange B:XYZ bid size

1,500

Exchange C:XYZ bid size

800

Routing logic:1. Dark pools2. Exchanges

2


31/86


Market Structure

Tis simplied description highlights the efciencies to be gained rom the use o algosand the critical importance o technology in navigating across the ragmented marketplace.

2.7. Trade Reporting and Public Price TransparencyOnce trades are executed, they are immediately reported to the consolidated tapethemechanism or the provision o public post-trade transparency.

Tere are three networks under the consolidated tape system: Networks A and B or thereporting o trades in NYSE-listed and NYSE MKlisted securities, respectively, andNetwork C or the reporting o trades in NASDAQ-listed securities. Networks A andB are governed by the Consolidated ape Plan and are operated by NYSE echnologies,

whereas Network C is overseen by NASDAQs UP (Unlisted rading Privileges) Plan.

Exchanges and other market centers send details o executed trades in real time to the rel-evant central data consolidator o the respective network, which then publishes the tradeson the consolidated tape.

O-exchange transactions are reported to one o two main rade Reporting Facili-ties (RFs) that are registered and overseen by FINRA. Te largest RF is operated byNASDAQ (the FINRA/NASDAQ RF), through which approximately 93%21 o all o-exchange transactions are reported. Te other RF is operated by NYSE (the FINRA/NYSE RF). FINRA also maintains the Alternative Display Facility (ADF) and the OC

Reporting Facility (ORF) or OC trades, but these acilities are largely unused.

Importantly, all transactions reported through a RF are sent to the central data consoli-dator or inclusion in the consolidated tape. Tis means that all transactions, whether exe-cuted on- or o-exchange, are reported via the same public mechanism.22 In this respect,the consolidated tape has a linking eect on the equity market, acting as a repository or alltransaction inormation across the trading network.

21Based on data rom www.nasdaqtrader.com.22It should be noted that all on-exchange transactions are required to be reported in real time, whereas

o-exchange transactions must be reported (through FINRA) within 30 seconds. A second point o note isthat an exchange is required to send trade inormation to the central consolidator at the same time it dis-seminates inormation on its proprietary data eed. But because o aggregation, inormation published on theconsolidated quote and tape is visible approximately 510 milliseconds later than proprietary exchange eeds.Proprietary trading rms urther minimize latencies by utilizing co-location acilities provided by exchanges.See U.S. SEC (2010) or urther discussion.


32/86



Similarly, public pre-trade transparency is realized via the consolidated quote system.Exchanges send quotation inormation to the central data consolidators (as previously

described) that assimilate bids and oers rom across the displayed market centers andproduce the NBBO or dissemination through the consolidated quote stream. But as notedelsewhere in this report, ASs, such as dark pools, do not publish quotes in the public con-solidated quotation eed.23

23Alternative trading systems (including dark pools) are exempt rom having to display quotation inormationunless the trading volume or a stock on their system exceeds 5% o consolidated volume or that stock. SeeSection 3 or more details.


33/86


3. Regulatory Framework

Te key regulatory development shaping the oundation o the current U.S. equity mar-ket structure was the establishment o the National Market System (NMS) ollowing theSecurities Acts Amendments in 1975. Te aim o the amendments was to create a marketsystem that provided or greater competition between market operators and participants,that promoted transparency in terms o the visibility and accessibility o prices, and thatmade markets interconnected.

In 2005, the rules governing the National Market System were consolidated and updatedby the SEC under Regulation NMS (Reg. NMS). Te rules require the protection o dis-played quotations, air and nondiscriminatory access to market prices, and improvementsto the consolidation o data.

3.1. Regulation NMSFollowing are summaries o the key eatures o Reg. NMS.

1. Rule 610Access Rule

Rule 610 establishes standards or access to quotations in NMS stocks. It is designed to ensure

that market participants have air and nondiscriminatory access to markets and prices, therebyacilitating competition. As noted in Section 3.2, the provisions or air and open access underRule 610 only apply to exchanges; they do not generally apply to ASs unless an AS executesmore than 5% o the volume in a given stock. ASs are thus able to operate as dark venues.

A key eature o Rule 610 is that it enables market participants to access dierent marketcenters via private linkages, replacing the previous linkage system, IS (Intermarket rad-ing System). Such private linkages include, or example, the type o SOR unctionalityprovided by broker/dealers and exchanges outlined in Section 2. Importantly, to enable thistype o access, Rule 610 prohibits trading venues rom imposing unairly discriminatoryterms that would prevent or inhibit efcient access by any market participant.


34/86



Te other pertinent element o Rule 610 is that it caps the ees that trading venues cancharge or accessing protected quotations (i.e., the take ees or hitting resting liquidity) at

$0.003 per share. Te ee limit is designed to ensure that market participants are not undulyrestricted rom accessing displayed quotations.

2. Rule 611Order Protection Rule

Rule 611, also known as the trade-through rule, protects displayed quotations at the topo the order book. Specically, incoming orders cannot execute against resting liquidity atprices inerior to the NBBO beore rst exhausting the trading interest at the best bid oroer. In other words, marketable orders cannot trade through the best prices. By extension,undisplayed trading venues do not benet rom trade-through protection because they donot display quotes in the public consolidated quotation system.

Rule 611 provides protection to market participants that post displayed limit orders, therebyupholding inter-market price priority or displayed and accessible quotations. Signicantly, Rule611 eectively removed the protections aorded to the manual quotation systems o incumbentoor-based exchanges, thereby accelerating the shit toward ully electronic markets.

It is worth emphasizing, however, that orders below the top o the order book are notcovered under Rule 611. Accordingly, once a market center has satised its requirementsunder Rule 611, it can execute the remainder o the order within its system even thoughits remaining order book quotations (below its best-priced orders) may be inerior to othermarket centers. Tat is, a customer may receive an inerior average price or his or her order

than would otherwise be the case with through-the-book trade-through protection.24

3. Rule 612Sub-Penny Rule

Rule 612 prevents exchanges, broker/dealers, and other market centers rom displaying,ranking, or accepting any orders that are priced in an increment o less than $0.01 or, orstocks that trade at less than $1, in an increment o less than $0.0001. Te purpose o thesub-penny rule is to prevent liquidity rom ragmenting over excessively diuse pricingpoints and to protect limit orders rom yielding execution priority to other limit orders thatprovide incremental but economically insignicant price improvement.

24Te U.S. SEC raised the issue o extending trade-through protection below the top o the order book in its2010 concept release. No rules have been passed at the time o this writing.


35/86


Regulatory Framework

It is notable, however, that although Rule 612 prevents sub-penny quoting, it does notprohibit sub-penny executions by broker/dealers. As the explanatory text to Rule 612

notes, In addition, a broker/dealer could, consistent with the proposed rule, provide priceimprovement to a customer order that resulted in a sub-penny execution as long as the bro-ker/dealer did not accept an order priced above $1.00 per share in a sub-penny increment.

Tis rule has resulted in a signicant proportion o retail marketable orders being executedin sub-penny increments. Indeed, because broker/dealers are not required to publish quota-tions, they are not constrained in their ability to execute in sub-penny increments to thesame extent as exchanges. As noted in Section 2.4, sub-penny transactions could generateadverse consequences or market integrity relative to the economically small benets accru-ing to the investor receiving price improvement rom the sub-penny transaction.

4. Rules 601 and 603Market Data RulesRules 601 and 603 reect renements to the plans governing the consolidation o mar-ket data. In particular, they revise the ormulas or the allocation o revenue to the marketcenters that contribute data to the consolidated quote and consolidated tape. Te rules alsostrengthen plan governance arrangements and authorize market centers to distribute theirown data independently (in addition to their submissions to the consolidated quote and tape).

Finally, Reg. NMS aims to improve the transparency o order routing and execution prac-tices by requiring market centers to publish monthly execution quality statistics (Rule 605).It also requires broker/dealers to publish quarterly reports on their order routing practices,including details o the venues to which they route orders (Rule 606). Collectively, the

transparency provided under Rules 605 and 606 is designed to strengthen competitionamong market centers and broker/dealers.


36/86



3.2. Regulation ATS

Regulation AS (Reg. AS) was introduced in 1998. It was established to capture multi-participant non-exchange-trading modalities, including dark pools, ECNs, and broker/dealer systems.25 Because these entities are not considered to meet the specications oan exchange, they are not bound by the same requirements as exchanges, such as marketsurveillance obligations and other sel-regulatory responsibilities.

Fundamentally, U.S. securities market regulation is premised on a distinction betweenexchanges and broker/dealers. Because ASs are not considered to be exchanges, they mustregister with the SEC as broker/dealers and adhere to the business conduct rules applicableto broker/dealers established by FINRA.

Te pertinent aspects o Reg. AS relate to two types o access requirements: order displayand execution access, and air access to AS services.

Generally, ASs, including dark pools and other broker/dealer systems, are not required topublicly display price quotations and are able to restrict access to their crossing systems andinternalization pools to their clients or to only certain types o investors. In other words,

ASs can be discriminating about who has access to their systems and to their prices, incontrast to the air access principles required o exchanges.

Te AS access requirements, however, change once a 5% trading volume threshold isexceeded. Specically, i an AS executes more than 5% o the trading volume in a given

stock and displays orders to more than one person, then it must adhere to the general orderdisplay and execution access requirements applicable to exchanges. According to U.S. SEC(2010), there are no ASs that currently exceed this 5% threshold. Te SEC has proposedlowering the threshold at which these order display and access requirements would applyrom the current level o 5% to 0.25%. Te SEC has also proposed amending the denitiono bid and oer to include actionable IOIs. Tis would require IOIs, which are currently

25Although Reg. AS was introduced in 1998, volume transacted on undisplayed ASs, such as dark pools,did not begin to grow signicantly until the middle o the next decade. One actor that may have contrib-uted to the growth o ASs was an amendment to ederal legislation on pensions. Te Employee Retirement

Income Security Act (ERISA) o 1974, under Section 408 (b) (29 U.S.C.1106), prohibits a duciary rom,among other things, dealing with the assets o a pension plan in his own interest or or his own account.But the Pension Protection Act (PPA) o 2006, Section 611, includes an amendment to the prohibitionsunder ERISA. Specically, the PPA exempts transactions . . . executed through an electronic communicationsnetwork, alternative trading system, or similar execution system. . . rom the prohibited transactions underERISA.


37/86



only transmitted by ASs to select customers, to be subject to the public quoting require-ments o exchanges. As noted in Section 3.3, these proposals remain under consideration

as o this report s publication date.

Looking ahead, as the equity market evolves and trading venues and trading modalitiescontinue to innovate, the historical regulatory distinction between exchanges and broker/dealers will increasingly blur. Tis transition suggests that a unctional approach to regula-tion would be most appropriate to maintain a level playing eld, in which all unctionallysimilar trading venues are treated in the same way.

3.3. International Regulatory Developments

regarding Dark LiquidityTe issues associated with dark liquidity are prevalent internationally. A number o regulatorybodies in other jurisdictions have developed rameworks governing how dark pools and undis-played orders are allowed to operate within their markets. Tese initiatives provide useul context

when considering the U.S. regulatory ramework and possible improvements to that ramework.

In 2011, the International Organization o Securities Commissions (IOSCO) publisheda report on principles or dark liquidity, which establishes a broad set o international bestpractices or regulatory treatment o dark pools and dark orders in otherwise transparentmarkets. According to IOSCO (2011), the issues surrounding dark liquidity include

the potential impact on the price discovery process when there is a substantial number odark orders (in dark pools or otherwise) that may not be published,

the potential impact o ragmentation on inormation and liquidity searches, and

the potential impact on market integrity arising rom dierences in access to markets andinormation.

IOSCO sets orth ve broad principles designed to mitigate these issues. Tese include theollowing:

1. Regulators should generally require and promote public pre-trade and post-trade transpar-ency. When allowing exemptions rom pre-trade transparency, regulators should considerthe potential impact on price discovery, ragmentation, airness, and overall market quality.


38/86



2. Market rules and regulations should provide or priority o transparent orders over darkorders at the same price.

3. Reporting regimes should enable regulators to access inormation on dark orders and trans-actions.

4. Inormation should be made available to market participants about dark pools and dark ordersso that they are able to understand the manner in which their orders are handled and executed.

5. Regulators should periodically monitor the development o dark pools and dark orders intheir jurisdictions and take appropriate action as needed.

A number o national regulators have adopted standards to regulate dark liquidity withintheir respective markets that are broadly consistent with these principles.

In the European Union (EU), revisions to the Markets in Financial Instruments Direc-tive (MiFID) and Regulation (MiFIR) propose handing responsibility or and oversighto the pre-trade transparency waiver process to the European Securities and Markets

Authority.26 Tis move is designed to ensure that exemptions rom pre-trade transparencyare applied consistently throughout the dierent EU member states.

In Canada, the Canadian Securities Administrators (CSA), in conjunction with the Invest-ment Industry Regulatory Organization o Canada (IIROC), announced a revised rame-

work or dark liquidity that will become eective in October 2012. Te ramework is basedon three elements. First, visible orders will have execution priority over same-priced dark

orders on the same marketplace. Second, to trade with a dark order, smaller orders mustreceive a minimum level o price improvement, dened as one trading increment (gener-ally 1 cent) or hal a trading increment or securities with a bidask spread o one tradingincrement. And third, IIROC will be permitted to introduce a minimum size threshold orthe use o dark orders at a later date. CSA and IIROC will monitor market developmentsto consider whether and when IIROC should implement a minimum size.

In 2011, the Australian Securities & Investments Commission (ASIC) undertook a serieso consultations on Australias equity market structure. In April 2012, ASIC provided anupdate on the direction o its market structure reorms on the basis o the consultation pro-cess. In the area o pre-trade transparency, ASIC has proposed, rst, replacing the existing

26MiFID permits waiving the obligation to publicly display orders or orders and trading systems satisyingone o the ollowing criteria: (1) orders that are large in scale; (2) reerence price systems, which passivelymatch orders at a reerence price rom the lit market; (3) systems that ormalize negotiated transactions,such as VWAP trades; and (4) orders held in an order management acility, such as iceberg orders.


39/86



AU$1 million threshold or block trades (which are exempt rom pre-trade transparency)with a tiered model to more closely link large trades with liquidity actors. Second, the pro-

posals would require dark orders to provide meaningul price improvement o at least onetick size or the midpoint o the quoted best bid and oer prices. And third, although ASIChas not proposed introducing a minimum size threshold or dark orders at this stage, it willconsider potential triggers or possible uture application o a minimum size. Te latter twomeasures are consistent with those announced in Canada by CSA/IIROC.

In the United States, in addition to considering lowering the volume threshold at whichASs would be required to display orders and in addition to considering including action-able IOIs in the consolidated quote data (as noted in Section 3.2), the SEC has consideredintroducing a trade-at rule. Tis rule would prevent a market center or broker/dealer romtrading at the NBBO unless it had been displaying at that price at the time it received

an order. Te rule would require a market center or broker/dealer not displaying at theNBBO to either execute the order with signicant price improvement or route the orderto a venue displaying at the NBBO. Separately, the Joint CFCSEC27 Advisory Com-mittee on Emerging Regulatory Issues, in a report published in 2011, has also suggestedthat the SEC analyze urther the impact o broker/dealer internalization, with the idea oconsidering whether to adopt a rule requiring that internalized or preerenced order owbe executed at a price materially superior to the best bid or oer.

As o this report s writing, no rulemaking decisions have been made in the United Stateson any o the proposals.

27CFC is the Commodity Futures rading Commission.


40/86


4. Literature Review

Tis section presents a summary o the academic literature most relevant to this study. Intotal, our studies suggest an adverse relationship between dark trading and market quality(meaning that undisplayed trading is associated with a reduction in market quality) andeight suggest a neutral or positive relationship.

O the most recent literature directly related to this study, two studies suggest that darktrading is associated with a deterioration in market quality (Weaver 2010, updated 2011;and Degryse, de Jong, and van Kervel 2011) and two studies indicate that dark trading isassociated with improved market quality (Buti, Rindi, and Werner 2010a; and OHara and

Ye 2011). Overall, the results rom the academic literature can be viewed as mixed.

4.1. Studies Suggesting an Adverse Relationship

between Dark Trading and Market QualityWeaver (2010, updated 2011). In this study, Weaver examines o-exchange trading andmarket quality in the U.S. equity market. He examines the market share o trading volumereported through a RF and nds evidence that o-exchange reporting is associated with areduction in market quality. In particular, Weaver nds that stocks with higher levels o o-exchange reporting have wider bidoer spreads (quoted, eective, and realized). Weaveralso nds that increased o-exchange reporting is associated with greater price impact pertrade and higher volatility.

Te study is based on cross-sectional regressions o various market quality measures againstthe proportion o volume reported through a RF, based on data rom October 2009.

Weaver updated the study in 2011 using data rom October 2010, ater Direct Edges con-version to an exchange in July 2010, in order to more cleanly capture undisplayed tradingvolume. Te ndings are the same as those in the 2010 version.

Tere are some limitations to this study. First, trades reported through a RF are treated

in aggregate and are collectively considered as internalization. Tereore, Weaver does notexplicitly distinguish between dierent types o o-exchange trading. Second, the 2010version revealed some exceptions to the overall conclusion that o-exchange reportinghas a harmul eect on market quality. Specically, or NASDAQ stocks, dollar-quotedspreads and RF volume are negatively related (higher RF volume is associated with


41/86


Literature Review

lower spreads). Te same is true or NYSE stocks with regard to percentage quoted spreadsand RF volume. In the 2011 version, however, the relationship between the proportion

o RF volume and spreads is uniormly positive or NASDAQ and NYSE stocks (moreRF volume increases spreads), although there are some exceptions or Amex-listed stocks.Weaver iners that the removal o displayed ECN volume rom the sample in the updatedstudy could explain these dierences. In other words, the results are stronger in the absenceo displayed volume in the RF data.

Overall, Weaver concludes that increased o-exchange trading is associated with a degra-dation o market quality.

Degryse, de Jong, and van Kervel (2011). Te authors study the impact o dark trading andvisible ragmentation on market quality. Tey nd that ragmentation in visible order books

(in other words, competition among displayed venues) improves global liquiditythat is,liquidity consolidated across limit order books. But dark trading has a detrimental eect onliquidity measures, resulting in higher spreads and lower depth. Te authors posit that thenegative impact o dark trading is consistent with a cream-skimming eect between darkand visible markets because the inormativeness o trades strongly increases with dark activ-ity. Tat is, the degree o well-inormed order ow (rom proessional traders) on exchangesis positively related to the level o less well-inormed order ow preerenced in dark markets.

Tere is much literature on order preerencing and adverse selection. In theory, high levelso internalized (particularly retail) order ow may cause the adverse selection componento the bidoer spread on exchanges to increase, reecting the greater risk o market par-

ticipants trading against inormed order ow. In other words, more internalization o retailorders results in a greater proportion o inormed order ow routed to exchanges. Marketparticipants, thereore, widen the quoted spread to compensate or the increased risk oadversely selecting inormed or toxic order ow to trade against.

Easley, Kiefer, and OHara (1996). Te authors examine the concept that purchased orderow (whereby brokers pay to acquire and internalize retail order ow) is used to cream skimuninormed liquidity trades, leaving the inormation-based trades to established exchanges.In their study, they ocus on trades executed on the NYSE versus purchased order ow on theCincinnati Stock Exchange. Tey nd a signicant dierence in the inormation content othe dierent types o stock trades and conclude that this dierence is consistent with cream

skimming. Te authors note that this practicethat is, ragmentation o orders by typecanimpose wider costs on the market. Te more successul the practice o preerencing is, themore protable the strategy is or its pursuers. As more uninormed orders are preerenced


42/86



(or internalized), prices become worse on the remaining orders. Wider prices on-exchangegive greater scope or o-exchange order matching to be more protable; it enables brokers to

extract higher rents rom customers because they can deal within a wider spread.

Zhu (2012). In this study, Zhu examines the impact o dark pools on price discovery in atheoretical setting and nds that execution risk is a key actor. Specically, because ordermatching in a dark pool depends on the availability o other counterparties, some orderson the heavier side o the marketthat is, the side with more orderswill ail to beexecuted. Well-inormed proessional traders are more likely to cluster on the heavier sideo the market and hence suer lower execution probabilities in a dark pool. In contrast,liquidity-seeking traders (those demanding immediacy) are less likely to cluster on theheavy side o the market and hence enjoy higher execution probabilities in dark pools.

Tis dierence in execution risk pushes relatively better-inormed traders onto exchange

order bookswhich contain the presence o market makers that can absorb excess orderowand relatively uninormed liquidity-seeking traders into dark pools. Consequently,although the presence o better-inormed order ow on exchanges improves price discovery(in an efciency sense), it can also lead to wider spreads because market makers widen theirquotes to compensate themselves or the increased adverse selection risk associated withthe higher proportion o inormed order ow on exchanges.

aken together, these studies suggest that internalization and dark pool activity can leadto wider quoted spreads and generally poorer market quality. However, this conclusion isnot uniorm across the internalization literature; several studies suggest a benign eect onmarket quality rom order preerencing.

4.2. Studies Suggesting a Neutral or Positive

Relationship between Dark Trading and Market

QualityLarrymore and Murphy (2009). Te authors explore the link between internalization andmarket quality by conducting an event study around the implementation o the orontoStock Exchanges Price Improvement Rule in 1998. Te rule required orders o 5,000 or

ewer shares to receive price improvement or be routed to the limit order book or execu-tion. Te authors again note that in markets with high levels o internalized retail orderow, the adverse selection component o the bidoer spread increases to reect the greaterrisk o market makers trading against inormed order ow. Larrymore and Murphy ndthat in the post-implementation period o the Price Improvement Rule, market quality


43/86


Literature Review

improved as reected by declining spreads, reduced variance o pricing error, and greaterdepth, among other things. Te outcome o improved market quality implies that market

makers could compete more aggressively or order ow ollowing the rule change becausemore limit orders came back on to the order book, thereby reducing adverse selection riskas reected in the narrowing o spreads. In other words, brokers who previously internal-ized order ow could no longer earn abnormal prots and so either improved on the pricesquoted in the displayed market or routed orders or execution on the exchange.

It is notable, however, that internalization rates (measured by the same counterparty executingboth sides o a trade) increased or small transactions ollowing the rule change. Specically, asinternalization rates increased or trade sizes between 1,200 shares and 5,000 shares, quotedspreads and volatilities declined and market depth increased. Consequently, Larrymore andMurphy conclude that together, these ndings suggest that increases in internalization rates

are not necessarily associated with deteriorating market quality but may, in some instances, beassociated with superior market quality (p. 358). In other words, judiciously regulated inter-nalization practiced under avorable market conditionsin this case, under the terms o the

oronto Stock Exchanges Price Improvement Rulecan increase market quality.

Chung, Chuwonganant, and McCormick (2006). Te authors examine order preerencingand adverse selection costs. Teir ndings are consistent with the hypothesis that brokersselectively internalize orders based on the inormation content o orders. But Chung et al.nd that order preerencing is not necessarily harmul. Tey note that brokers are likely tocharge uninormed traders lower commissions as compensation or the high spreads thatthey pay. o the extent that execution costs o uninormed traders are reduced by lower

commissions, the net eect o order preerencing on overall execution quality can be posi-tive, although accurate quantication o these benets is difcult.

Other studies suggesting a neutral relationship. wo other notable studies that suggest abenign eect on market quality rom order preerencing are Hansch, Naik, and Viswana-than (1999) and Peterson and Sirri (2003).

Hansch, Naik, and Viswanathan study execution quality, preerencing, and internalization ora sample o stocks on the London Stock Exchange and do not nd any relationship betweenthe extent o preerencing or internalization and spreads across stocks. Peterson and Sirriexamine execution quality o retail order ow on primary and regional U.S. equity exchanges.

Tey examine preerencing on regional exchanges, where dealers either are allowed to directorders to themselves or pay to obtain order ow. Although execution quality is generally


44/86



ound to be superior on the NYSE, the authors nd that market quality is superior on thepreerencing regional exchanges compared with the nonpreerencing regional exchanges. In

sum, they do not nd evidence that preerencing harms market quality.

Te next our studies all indicate that dark trading is associated with improved market quality.

Buti, Rindi, and Werner (2010a). Te authors analyze the relationship between dark poolactivity and market quality based on a sample o 11 dark pools that voluntarily report theiractivity to SIFMA (Securities Industry and Financial Markets Association). Teir samplecaptures approximately 5060% o all dark pool activity reported by Rosenblatt Securities.

Looking rst at the determinants o dark pool activity, Buti et al. nd that more-liquidstocks with high volume and low volatility are generally associated with relatively high

levels o dark pool activity. Tey also nd that dark pool activity is higher or stocks withnarrow quoted and eective spreads and high inside bid depth, suggesting that dark poolsare more active when the degree o competition in the limit order book is higher.

Buti et al. also examine the relationship between dark pool activity and market qualityacross stocks and over time. For the cross-sectional analysis, the authors use a regressionspecication analogous to Weaver (2010, 2011), with market quality measures on the let-hand side and the proportion o dark pool volume on the right-hand side, alongside aquadratic term o the proportion o dark pool volume (to account or nonlinear eects) andother explanatory variables. Teir results show that a higher amount o dark pool activity isassociated with lower quoted and eective spreads, lower price impact, and lower volatility.

Tat is, more dark pool activity is generally associated with higher market quality. Note thatin the quadratic term, however, spreads decline initially but increase beyond a dark poolmarket share o approximately 8%, which may suggest a curvilinear relationship.

In the time-series analysis o dark pool activity and market quality, the authors nd thathigher dark pool activity is associated with lower spreads, greater depth, and lower volatilityas measured by the standard deviation o returns. In sum, Buti et al. nd no evidence thathigh levels o dark pool activity are associated with a worsening o market quality.

Buti, Rindi, and Werner (2010b). In this paper, the authors model the competition betweena dark pool and a visible limit order book. Tey demonstrate that dark pool market shareis higher when limit order book depth is high, when spreads are narrow, and when the ticksize is larger. Tey also show that the initial level o liquidity determines the eect o thedark pool on spreads. For liquid stocks, the quoted spread remains very tight as dark pools


45/86


Literature Review

compete or, and attract, order ow rom the displayed limit order book. But or illiquidstocks, the competition induced by the dark pool reduces the execution probability o limit

orders, causing the spread to increase.

OHara and Ye (2011).Te authors examine RF volumes to measure ragmentation levelsin individual stocks and use a variety o empirical approaches to compare execution qual-ity and efciency o stocks with dierent levels o ragmentation. OHara and Ye nd thatmore-ragmented stocks have lower transaction costs. In particular, they nd a negativerelationship between eective bidoer spreads and the proportion o RF volume, imply-ing that higher proportions o RF volume are associated with narrower spreads and hencebetter market quality.

Gresse (2006). Te author investigates the relationship between the trading activity o a

dark pool crossing network and the liquidity o a traditional dealer marketin this case,the quote-driven segment o the London Stock Exchange. A cross-sectional analysis obidoer spreads shows that dealer market spreads are negatively related to crossing-network executions, which supports the notion that dark trading improves market quality.

Te author nds that risk-sharing benets rom crossing-network trading dominate rag-mentation and cream-skimming costs.

4.3. SummaryOut o the studies reviewed here, the most applicable to the ollowing analysis are Weaver

(2010, 2011); Degryse, de Jong, and van Kervel (2011); Buti, Rindi, and Werner (2010a);and OHara and Ye (2011). Te rst two support the notion that undisplayed trading harmsmarket quality, whereas the second two suggest undisplayed trading is associated withimprovements in market quality.

Out o these studies, it should be noted that Degryse, de Jong, and van Kervel (2011) isbased on Dutch stocks whereas the others are based on U.S. stocks. Te European equitymarket structure diers in two key respects rom the U.S. market: Tere is no single con-solidated tape or the reporting o trades, and there is no trade-through rule enorcingexecution at the best price across all trading venues. Consequently, the European marketis not virtually consolidated in the same manner as the U.S. market (in which the consoli-

dated tape has a linking eect across market centers), and thereore, the dynamic betweenragmentation and market quality may dier between Europe and the United States. For


46/86



example, Degryse et al. nd that visible ragmentation (competition between displayedvenues) improves liquidity but dark trading does not, which diers rom the conclusions o

OHara and Ye (2011), who nd benecial eects rom o-exchange trading.

I

dark pools

Documents