equity market micro structure

8/3/2019 Equity Market Micro Structure

1/16

E q u i t y M a r k e t M ic r o s tr u c t u r e :T a k i n g S t o c k o f W h a t W e K n o wIIR E T O F R A N C I O N I , S O N A L I H A Z A R I K A , M A R T I N R E C K ,ASD RO BER T A. SCH W ART Z

H A Z A R I K A

of Business, Baruch@ baruch.cuny.edu

R T I N R E C K

E R T A . S C H W A R T Z

School of Business,

m quity market microstructure addressesB i issues that involve the implementationI ^ of portfolio, or investment, decisions- ^ ^ ^ in a marketplace . Implementat ionenfils the placement and hand ling of orders ina securities market and the translation of theseorders into trades and transaction prices. Theprocess links fundamental information con-cerning equity valuation of primary conce rn toportfolio managers to prices and transaction vo l-umes that are realized in the marketplace. Thequality of the link depends on the rules, proce-dures, and facilities of a securities market, andon the broader regulatory and competitive envi-ronment w ithin which the market operates.Widespread interest on the part of thesecurities industry, government, and academiais testimony to the importance of marketmicrostruc ture analysis. The subject addressesissues that concern investors, broker/dealerintermediaries, market regulators, exchanges,and other trading venues, as well as the broadeconomy. Interest in microstructure hasincreased sharply over the past three and a half

decades, spurred in particular by three eventsissuance of the U.S. Securities and ExchangeConimission (SEC) Institutional Inuestor Reportin 1971, passage by the U.S. Congress of theSecurities Acts Am endments of 1975, and thesharp stock market drop on Octo ber 19, 1987.Further, the advent of computer-driven tradingin recent years has enabled researchers to cap-ture electronically the full record of all trades

and quotes, and this has provided empiricalresearchers with far richer data, namely highfrequency data, for analyzing trading and pricesetting.Over the years, microstructure analysishas expanded and exchange structure hasstrengthened. We consider both of these devel-opments in this article. First, we set forth themajor challenges that the microstructure lit-erature addresses. Second, we present a broadview of the direction in w hich microstructureanalysis has been and is evolving. Third, weturn to one applicationthe design of anactual marketplace, D eutsche Brse s electronictrading system Xetra. The German market wasthe last of the major European bourses to in tro-duce an electronic trading platform, and theresult, Xetra, is state of the art, which makesDeu tsche B rse a particularly interesting sub-ject of analysis. Last, we consider the bumpyand hazardous road that takes us from theoryto the development of an actual marketplace.M ICROSTRUCTURE'S CHALLENGE

Microstructure analysis has four broadapphcations. First, and the key focus of thisarticle, is that microstructure analysis givesguidance to market structure development.Th e hn k betw een the two areas is straightfor-ward. The critical factor that drives microstruc-ture analysis is friction in the marketplace (i.e.,the explicit and implicit costs of implemen tingFALL 2008 THE JOURNAL OF PORTFOLIO M ANAGEM ENT 5 7


2/16

portfolio decisions) and trading costs depend on the archi-tecture of the marketplace, which determines how ordersare handled and turned into trades. The flipside of fric-tion is illiquidity, and a primary function of a marketcenter is to amass Hquidity.The second application of microstructure analysis isto facilitate the development of trading strategies and algo-rithms for asset managers and broker/dealer intermedi-aries. The importance of this application is evident in thecurrent development of computer-driven algorithmictrading. Algorithms can be fine-tuned to take into accounta number of factors, such as the probability of a limit orderexecuting; time-of-day effects, such as market openingsand closings; the search for liquidity in a fragmented env i-ronment; and the choice of trading modality (e.g., con-tinuous limit-order-book market, quote-driven dealer

market, periodic call auction, block trading facility, or ahybrid combination of these modalities).Th e third application of microstructure analysis con -cerns tests of market efficiency. In the 1970s, the efficientmarkets hypothesis (EMH) was widely accepted by finan-cial economists as a cornerstone of modern portfoliotheory and it continues to receive broad academic sup-port today. The hypothesis addresses informational effi-ciency as distinct from operation al efficiency, the latter ofwhich refers to the containment of transaction costs bysuperior market design. According to the EM H, a market

is informationally efficient if no participant is able toachieve excess risk-adjusted returns by trading on cur-rently available information. Many o f the EM H tests haveconsidered a single, although major, part of the informa-tion setmarket information (e.g., recent quotes, tradingvolume, and transaction prices). If prices properly reflectall known information, then, in a frictionless market atleast, they must change randomly over time; hence theterm random walk. Earlier studies, based on daily data,generally supported the random walk hypothesis.However, with the advent of high frequency data, thefootprints of complex correlation patterns have beendetected. This observation, along with superior knowl-edge of the impact of trading costs on returns behavior,is casting a new light on market efficiency. Regardless ofwhether inefficiency is thought of in operational or infor-mational term s, the EM H is not as stellar as it once was.In its fourth application, microstructure analysis shedslight on how new information is incorporated into secu-rity prices. In a zero-cost, frictionless environment, sharevalues would be continuously and instantaneously updated

with the release of new information. In actual marketshowever, information must be received and assessedtraders' orders must be placed and processed, executionmust be delivered, and accounts cleared and setded. Costsbo th explicit (e.g., commissions) and implicit (e.g., markeimpact), are incurred throughout this chain of eventsHighlighted in much microstructure literature are thecosts that some participants incur wh en, in an asymmetriinformation environment, other participants receive information first and trade on it to the disadvantage of theuninformed.

Asymmetric information is not the only reality, however. In light of the size, complexity, and imprecision omuch publicly available information, one might expecthat investors in possession of the same (large) information set will form different expectations about future riskand return configurations. This situation is referred to adivergent expectations.^ Asymmetric information and dgent expectations together reflect a rich set of forces thaimpact the dynamic behavior of security prices.

This overview of microstructure's four b road app lications underscores that trading frictions are the subjectraison d'tre. Participant orders cann ot be translated intotrades at zero cost (markets are not perfectly liquid), andtrades typically are not m ade at ma rket-clearing, or somemight say, equilibrium, prices. Trading decision rule(algorithms) are needed because the costs of implem entingportfolio decisions can sharply lower portfolio performance. In fact, much algorithmic trading is designed tocontrol trading costs, rather than to exploit profitabletrading o ppo rtunities. Today, trading is recognized as anactivity that is both distinct fi-om investing and equivalently professional. Market structure is of con cern to thbuyside desks precisely because markets are not perfecdyliquid, nor are they perfectly efficient, informationally ooperationally. Consequently, better market structurdelivers superior portfolio performance for p articipants

What is the economic service, one might ask, thaan equity market provides? The fuzzy link that connectinformation and prices in the nonfrictionless environment underscores two major market functions price discovery and quantity discovery. Price discovery refersparticipants collectively searching for equilibrium pricesQuantity discovery refers to the difficulty that participants, who would be wuling to trade with each otheractually have finding each other and trading when markets are fragmented. This difficulty is accentuated becaussome participants (primarily institutional investors) d

5 8 EQUITY MARKET MICROSTRUCTURE: TAKING STOCK OF W HAT W E KNOW FALL 200


3/16

Inot immediately reveal the total size of their orders,because doin g so would, undu ly drive up th eir m arketimpact costs.

Market structure afFects both the accuracy of pricediscovery and the completeness of quantity discovery. TheUnk between market structure and price discovery dependson the environment within which participants are oper-ating. At one end of the spectrum, investors can be equallyinformed and form hom cigeneous expectations based onthe information. At the other end, they can be differen-tially informed and form divergent expectations withregard to com mon ly shared information. W hen investorswh o share comm on information all agree on share values(i.e., have homogeneous expectations), prices can be "dis-covered" in the upstairs offices of research analysts. Butwh en investors are not equally informed, and whe n theyform different expectations based on common informa-tion, prices must be discovered in the m arketplace. In thelatter environme nt, the econom ic service provided by theexchange is clearit "produces the price."

In handling the orders of large institutional cus-tomers, quantity discovery is a challenge. It is not at alluncommon for an institution to want to buy or sell, forexample, 500,000 shares of a company that has an averagedaily trading volume of 300,000 shares. Executing anorder of this size can easily drive prices away from thetrader before the trade has been completed. The adverseprice move is a market impact cost. Institutions attempt tocontrol their marke t impact costs by trading patiently and,as much as possible, invisibly Go od market structure canhelp. To this end, a num ber of alternative trading systemshave been formed in recent years, and dark, (nontrans-parent) liquidity pools have emerged.

When price discovery occurs in the marketplace,participants employ market trading strategies to imple-ment their portfolio decisions. Participants, who have dif-ferential information that will soon become public,determine how best to mete their orders into the marketso that prices will move to new levels at minimal speed.Other questions that a trader might in any event askinclude the following: If I trade now, at the c-urrent mom ent, h ow w ill theprice that I receive compare with the average pricethat shares are trading at today? Is price curren tly at a sustainable, validated level, oris it likely to move higher or lower in the cominghours, minutes, or even seconds?

Would I do better to be patient and place a limitorder, or submit a market order and get the jo b doneright away? Should I attempt to trade now in the continuousmarket, or wait for a closing call?

The orders that a set of participants reveal to the marketdepend on how questions such as these are answered, andthe prices that are set and the trading volumes that are real-ized depend on the orders that are revealed.The categories of trading costs that receive the mostattention on the part of exchanges, regulators, and acade-micians are generally those that are the most straightforwardto measurecommissions and bid-ask spreads. With increas-ingly precise measures of market impact becom ing available,however, the market impact cost is also now being widely

taken into account. But because the opportunity cost of amissed trade is far more difficult to quantify, it is often over-looked. An even greater challenge is quantifying a cost thathas received little formal attention: realizing executions atpoorly discovered prices. The problem, of course, is thatequilibrium values are not observable and appropriate bench-mark values are not easily defined.WHAT MICROSTRUCTURE ANALYSIS HA STO OFFER: PERSONAL REFLECTIONS

In this section we review the development ofmicrostructure analysis. Ou r objective is not to provide acomprehensive survey of the literature, but to highlightsome of the important themes that can give guidance tomarket structure development. More detailed informa-tion can be obtained from Cohen, Maier, Schwartz, andWhitcomb [1979] who provided an early survey of thefield; O'Hara [1997] who discussed important theoreticalmicrostructure models; Madhavan [2000], Biais, Glosten,and Spatt [2005], and Parlour and Seppi [2008] who pro -vided more recent surveys; and Hasbrouck [2007] whodealt with empirical microstructure research and researchmethodology. We first focus on the early literature, thenturn to more recent developments, and lastly present ourthoughts concerning an important direction in whichfuture microstructure research ought to head.

The Early FocusThe first contributions to the new field in finan-cial economics that came to be called microstructure were

FALL 2008 THE JOURNAL OF PORTFOLIO MANAGEMENT 59


4/16

made by a couple of people w ho participated in the SEC sInstitutional Investor Report [1971]. A handful of othersindependently started to focus on microstructure topicsin the early 1970s. Eventually a few of the early researcherscame to recognize the commonality of their interests and,adopting the title of Garman's [1976] well-known article"Market Microstructure," gave the field its name.

Mu ch of the early hterature focused on dealers andexchange speciahsts. These market makers were viewedas the suppliers of immediacy to investors, and the spreadwas considered the price they charged for providing thisservice in an environment where order arrival is non-synchronous. Of key importance was the relationshipbetween spreads and the costs of market making.Th e earher m arket-maker studies were in large partmotivated by a desire to determine whether intermedi-aries were realizing monopoly profits and, if so, whethertheir profits were attributable to market making being anatural monopoly. Spreads that are greater than the costsof market making would be taken as an indication ofmonopoly power on the part of the dealers, and spreadsthat were negatively related to trading volumes wouldindicate economies of scale in market making, whichcould imply a natural monopoly (Stigler [1964]). Spreadswere indeed found to decrease with transaction volume,but reasons other than market making being a naturalmonopoly were advanced (Smidt [1971]; Tinic [1972]).The general picture which emerged was that thetrading costs incurred by investors could be lowered bystrengthening com petition between market-maker inter-mediaries. In particular, competition in the New YorkStock Exchang e (NYSE) market was deemed inadequatebecause specialists and the exchange itself were viewedas having monopoly positions. Each stock was assignedto just one specialist according to the NYSE order con-solidation rule (Rule 390), which precluded in-houseexecutions by requiring that exchange members sendtheir orders for NYSE-listed securities to an exchange.

In addition, commissions were fixed and unjustifiablyhigh (Tinic and West [1980]).^Not surprisingly, the focus on the market-makerfirms led several researchers to model market-makerpricing decisions (i.e., the setting of their bid and askquotes). These studies include Bagehot [1971], StoU[1978], Amihud and Mendelson [1980], Ho and Stoll[1980, 1981 , 1983], and M ildenstein and Schleef [1983].With the one exception of Bagehot, the early formula-tions focused on inventory considerations. A market-

maker firm holding an undesirably long position wouldlower its quotes; that is, it would lower the offer, or ask,in order to sell more shares, and reduce the bid to discourage others from selling shares to it. Reciprocally, amarket m aker wh o was short would raise its quotes. Thisresponse on the part of the pubHc buy m ore shares whe nthe market maker's offer is lower and sell more shareswhen the market maker's bid is higheris evidence thathe public's demand to hold shares of any specific stockis downw ard sloping. A variety of mathem atical tools wereused to solve for optimal market-maker quotes. Thesemodels also gave further insight into the cost componentsof the market maker's spread (Stoll [1989]).

While insightful, the early inventory-based pricingmodels suffered several shortco mings. First, the early formulations typically assumed monopoly market makerseven though some of these models were applied to markets, such as the NY SE, wh ere exchange specialists werecompeting with other floor brokers and customer limiorders (Demsetz [1968]). The apphcation of theory further suffered from the reality that the price of immediacyfor an investor is not the spread of an individual markemaker, or even an average market-maker spread, but theinside spread (i.e., the lowest ask across all market makminus th e h ighest bid).-^ It is important to note that dealespreads can individually remain relatively invariant withrespect to transaction vo lume , wh ile the inside spread fallappreciably.A further shortcom ing of most of the earher modelis that they did not take into account a major cost incu rredby market makers, the losses generated by trading withbetter-inform ed investors. Re cog nition of this reality leto a development that did much to establish microstructure as an important new field in financial economicsThis development was the introduction of market-makemodels that were basedrather than on inventory managementon controlling the costs incurred when sominvestors possess information that the market maker anothe r investors have not yet received. Bageho t [1971] wathe first to embark on this hne of thought. He was latefollowed by, amo ng othe rs, Gloston and M ilgrom [1985and Kyle [1985].

With information asymmetries, the market makealways loses when trading with a better-inform ed parti cipant. For microstructure theorists at that time, this meanthat, for the dealer market not to fa, some investors mustrade for reasons unrelated to info rmation.'' Liquidity (i.ean investor's personal cashflowneeds) was one such motiv

6 0 EQUITY MARKET MICROSTRUCTU RE: TAKING STOCK OF WHA T W E KNOW FALL 200


5/16

for public buying and selling. A third type of participantwas also introduced along with the liquidity tradersnoise traders. A noise trader trades on price moves as ifthe moves contain information, whe n in fact they do not.This trio of informed traders, liquidity traders, and noisetraders was used to shovv how markets could functionand, in so doing, enable new information to be incor-porated into security prices (Grossman and Stiglitz [1980],Milgrom and Stokey [1982], Kyle [1985], Glosten andMilgrom [1985], Copeland and Galai [1983], and Easleyand O'Hara [1987, 1991,1992]).

At this early stage of development, the m icrostruc-ture pricing models werej predominantly market-makermodels. One exception should be noted, however. ANational Book System prjoposed by Mendelson, Peake,and Williams [1979] contained a comprehensive descrip-tion of an order-driven automated trading system thatprovided guidance for designing the first exchange-basedelectronic trading system.^ Most equity markets around theglobe are now order-driven, limit-order-book markets,which might include market makers in a hybrid struc-ture, such as the N YSE , but are no longer basically quo te-driven, or dealer, markets, as were the old NASDA Q andLondon Stock Exchange. The limit-order-book marketsare driven by orders placed by investors themselves, notby market-maker intermediaries.

I

The Current Focus iOver the years, microstructure analysis has grownextensively on both theoretical and em pirical fronts. Co n-comitantly, the security m arkets themselves have evolved,becom ing ever more technologically developed and m oreglobal in outreach, but al^o more fragmented betweendifferent trading facilities. One important new directionthat microstructure research has taken is to model theorder-driven market. Th e clrder-driven market is an envi-ronm ent in wh ich natural buyers and sellers provide im me-diacy to each other; those who are patient are willing topost limit orders, while thpse who demand immediacychoose to submit market orders that execute against theposted limit orders. Understanding the costs of, andmotives for, placing limit orders, as distinct from marketorders, was necessary. iWith limit orders, ie very existence of the bid-ask spread has to be explained. W he n a sufficiently largenum ber of participants pla.ce priced orders, one mightexpect that orders would be posted at virtually every

FALL 2008

available price point in the neig hborh ood of equilibrium,and that consequently the bid-ask spread would disap-pear. Cohen et al. [1979, p. 827] made this point (andCohen et al. [1981] analyzed the existence of thespread),^ asserting that[w]ith regard to modeling the market spread, wesuggest that a straightforward aggregation from indi-,vidual spreads is not possible in a system where thereis no clear distinction between demanders and sup-pliers of immediacy, and where traders meet in adynamic, interactive environment that incorporatesthe impact of investor order placement strategies.

Strategic order placem ent clearly requ ired further analysis.The task, however, was not simple. Some of the firstarticles in this area assumed, as is true for a dealer market,that limit-order and market-order participants are twoseparate, exogenously fixed group s that are separated bya firewall (Glosten [1994 ]). This assum ption, wh ile sim-plifying mathematical modeling, unfortunately distillsout much of the richness of an order-driven market.Mo re recen t models have eliminated the firewall (Handaand Schwartz [1996a], Foucault [1999], Parlour [1998],Handa, Schwartz, and Tiwari [2003], Faucault, Kaden,and Kandel [2005], and Goettler, Parlour, and Rajan[2005]). With the choice between a limit order and amarket order m ade endogenously, in order for the marketto function participants must divide naturally into fourgroups reflecting two dichotomiesone between buyersand sellers, and the other between limit-order andmarket-order placersrather than the standard twogroups (buyers and sellers).

With order-type selection endogenous in the order-driven market, the balance between immediacy suppliersand imm ediacy sellers becomes a second equilibrium thatmust be understood. O ne needs to recognize the con di-tions under which some participants choose to be liq-uidity demanders (place market orders), while otherschoose to be hquidity supphers (place hmit orders). If areasonable balance is not achieved between these twogroups, the order-driven market v ^ fail, as indeed it doesfor thinner, small-cap stocks. Increasingly, these issueshave been handled, and some sophisticated limit-ordermodels have been developed.^The microstructure analysis of trading systems hasexpanded to include periodic call auctions.* The eco-nomics ofa call auction are quite different from those of

THE JOURNAL OF PORTFOLIO MANAGEMENT 6 1


6/16

continuous trading and, consequently, so too are the orderplacement strategies that participants should em ploy in acall market. Call auctions do not, by their very nature,supply immediacy. Orders that are entered during a call'sbook -buildin g phase are held for a periodic crossing at asingle clearing price at the generally predeterm ined timeof the market call. Buy and sell orders submitted to a calldo not execute when they arrive even if they match orcross in price. Th is contrasts to a continu ous trading env i-ronment in which matching and crossing orders executeimmediately. Limit and market orders, therefore, have adifferent meaning in a call; Hmit orders do not supplyimmediacy to market orders and market orders are simplyextremely aggressively priced limit orders; that is, a m arketorder to sell in a call effectively has a Umit price of zero,and a market order to buy effectively has a Umit price ofinfinity.Today, virtually all modern, electronic exchangesopen and close their continuous markets with call auc-tions. Consequently, participants face further decisionswhe n operating in a caU-plus-continuous h ybrid m arket;for example, the decision of how to submit an order to acall auction that is followed by continuous trading (anopening call), and how to submit an order to a continuoustrading environment that is followed by a call auction(a closing call). Taking these tactical decisions into accountis part of the complexity of microstructure analysis.

Technological developments have simultaneouslyenabled new trading venues to emerge, which can frag-ment markets, while providing connectivity betweenthem, which can consolidate markets. Concurrendy, reg-ulatory initiatives have been motivated by the desire tointensify intermarket competition. Questions can be raised,however, c oncern ing fragmentation of the order flow.The conventional w^isdom has been that the consolida-tion of order flow improves liquidity, and exposing eachorder to all othe r displayed o rders gives investors the bestprices for their trades.On the one hand, consolidating trading in a singlemarket provides incentives to liquidity suppliers to com -pete aggressively for market orders by revealing theirtrading interests, and by being th e first to estabhsh a morefavorable price (if time is used as a secondary priorityrule). On the other hand, arguments in favor of tradingon multiple markets include the benefits of intermarketcom petition and the fact that traders wdth disparate motivesfor trading may want different marketplaces to trade in (i.e.,the one-size-does-not-fit-all argument).

Thus, different markets develop to serve diversinvestor needs, such as achieving a faster execu tion versuobtaining a better price. One growing need among larginstitutional investors, the ability to trade large orders w itminimal market impact, has led to the advent of darkpool, block trading facihties, such as Liquidnet, P ipelineand ITG's Posit that aid in quantity discovery. This industrdevelopment has spawned a related line of research ooff-exchange and upstairs trading (Grossman [1992], Sepp[1990], Madhavan [1995], Keim and Madhavan [1997]and Madhavan and Cheng [1996]).

A spectrum of market-quality issues have been olong and continuing importance to microstructurresearchers. These include market transparency,' both preand post-trade (Porter and Weaver [1998]); accentuatioof intraday price volatility; and correlation patterns, w hichave been observed in high frequency data (Engle andGranger [1987]). Other important issues include pricclustering and tick sizes (Harris [1991, 1994]; Ange[1997]). Applications, such as transaction cost analysis analgorithmic trading, have received increasing attentio(Domowitz, Glen, and Madhavan [2001]). The relativperformance of floor-based versus electronic trading ianother important issue (Domowitz and Steil [1999]).

A major line of empirical research was pioneereby Hasbrouck [1993] who decomposed transaction priceinto two componentsa random-walk component ana stationary component. The random-walk componenis identified with an efficient price, which the market itrying to discover. The stationary component is vieweas microstructure noise. Microstructure noise is commonly explained by features such as the bid-ask spreadmarket impact, and discreteness of the pricing grid. Thnoise component has also been attributable to price discovery being a dynamic process (Menkveld, Koopmanand Lucas [2007], and Paroush, Schwartz, and Wol[2008]).'"

Numerous empirical studies have focused on twof the world's premier markets, the NY SE and NA SDA Q(Hasbrouck [1991, 1995], Hasbrouck and Sofianos [1993Christie and Schultz [1994], Christie, Harris, and Schult[1994], Bessembinder and Kaufman [1997a, 1997b]Bessembinder [1999, 2003], and Barclay, Hend ersho tt, anMcCormick [2003], among others). Many other studiehave considered European, Asian, and other markets arounthe world (e.g.. Biais, Hillion, and Spatt [1995], Sanda[2001], and Ozenbas, Schwartz, and Wood [2002]).Across all of these markets, structural and performanc

62 EQUITY MARKET MICROSTRUCTURE: TAKING STOCK OF W HAT W E KNOW FALL 200


7/16

differences have heen noted, hut major similarities havealso been observed. It is apparent that, despite the influ-ence of historic and cultural considerations, trader behaviorand market performance around the globe depend largelyon microstructure realities. Alternatively stated, tradingrooms and markets around the world bear a striking resem -blance to each other.

Another recent linei of research has considered howsearch costs affect bid-ask! spreads in fmancial markets. Tothis end, Duffie, Pederson, and Grleanu [2007] presenteda dynamic model of market makers under the assumptionsof no inventory risk and of information being sym metri-cally distributed. They showed that sophisticated investorswh o have better search and bargaining abilities face tighterbid-ask spreads. This contradicts traditional information-based m odels which imply that spreads are wider for moresophisticated (i.e., better-informed) investors.As we have noted. Unlike in the frictiorJess marketarena of the capital asset pricing model, the amassing ofliquidity is a primary flincdon of a marketplace, and marketstructure features are generally designed with liquidityimplications in lnind. Asset managers also take liquidityinto account along with the two other standard variablesof mo dern portfolio theoryrisk and return. Difficultiesin denning, measuring, and modeling liquidity are for-midable, however, and the hterature that deals directlywith liquidity is relatively sparse (Bernstein [1987]). Nev -ertheless, liquidity considerations have permeated themicrostructure literature, both explicitly and implicitly.'^

Recalling the development of microstructureanalysis, two observations' stand ou t. First, microstructurestudies have given direction in multiple ways to marketstructure development. Second, to a remarkable extent,the various theoretical rnicrostructure models that arecenter stage today, and rrtany empirical analyses that arebased upon them, share a comm on structural framew orkthe asymmetric information paradigm. This consistencyis desirable in that it implies that the field has grown byaccretion rather than by replacement. Consequently, newinsights are more apt to refine than to contradict old con-clusions.Consistency, however, is not desirable if the c omm onstructural framework becomes overly rigid and restric-tive, and if it yields incomple te a nd /o r m isleading answersto questions involving tra'der behavior, market structure,and regulatory policy. At times, a literature will begin toadvance along a new front. Next, we consider this pos-sibility for the rnicrostructure literature.

Future DirectionsAs we have noted, the current focus in the literatureis on asymmetric informationbased models, which arecharacterized as follows. Trading is driven by informational

change, liquidity needs, and noise trading. The informa-tional motive for trading is the first mover of the three,and hquidity and noise trading are required so the marketwill not fail. Further, order arrival in the continuous envi-ronment is generally taken to be asynchronous. For a con-tinuous trading regime to function with asynchronousorder arrival, the presence of a limit order book and/or amarket-maker intermediary is required.Information trading is of keen interest because itrepresents the process by which new information isreflected in share values. In the standard asymmetric infor-

mation models, it is assumed that all participants in pos-session of the same information form equivalentexpectations concerning future risk and return configu-rations. When information changes, however, all partic-ipants may not receive the news at the same time; somereceive it before others , a reality that, at any point in time,can divide traders into two groupsthe informed andthe uninformed. Informed participants will never tradewith each other and consequently, liquidity and noisetraders m ust be present for a market to function. As noted,asymmetry of information, for the most part, lies at theheart of the standard microstructure models of today.The standard homogeneous expectations assump-tion has been tem pered of late. It is being recognized thatsome participants produce private information; that is, cetain participants further process information so as to gaininsights that are not immediately available to others.Whether participant expectations differ because of theactual production of private information, or simplybecause different people have varying interpretations ofthe same information or news announcement, the expec-tations of a group of investors can diverge.Also at the heart of an asymmetric information model

is the presumption that a stock has a fundamental value,which bears a unique relationship to the fundamentalinformation that informed traders possess rather than totrader activity in the marketplace. The process of infor-mation being fully reflected in prices under asymmetricinformation is the act of informed and uninformed agentstrading with each other until any discrepancy between amarket price and a fundamental value is eliminated. Theprocess can be viewed as arbitrage. In the earher dealer

FALL 2008 THE JOURNAL OF PORTFOLIO M ANAGEM ENT 63


8/16

models, the market m aker was assumed to know a stock'sfundamental value. In later models, informed traders (butnot the market maker) were assumed to know the fun-damental value (Kyle [1985]). Especially in the latermodels, price discovery is not instantaneous, but is a pro-tracted process that depends on the individual strategiesemployed by the informed and uninformed agents.

In recent years, an alternative paradigm has beenemerginga divergent expectations env ironment (MiEer[1977]). While institutionally realistic, this paradigm hasmet with considerable academic resistance. One reasonfor this is that homogeneous expectations environmentsare far easier to deal with mathematically, and homo-geneity has, in many applications, proven to be a usefulmodehng assumption. The homogeneous expectationsassumption has also been retained for ano ther reason. Asan attribute of individual rationality, it is presumed thatintelligent agents facing the same information andapplying the same (correct) analytical techniques willreach the same conclusions and, therefore, will havehomogeneous expectations.

Fundamental information, however, is enormou s inscope. It is complex and imprecise, and our tools for ana-lyzing it are relatively crude. In the presence of fuzzyinformation, expectations can be divergent. Allowing fordivergent expectations opens another path for microstruc-ture analysis and introduces new questions concerningagent behavior, market structure, and regulatory policy.Moreover, a further element can enter the analysis in adivergent expectations environmentalong with formingtheir ow n op inions, agents may also respond to the opi n-ions of others; that is exhibit adaptive valuation beh avior(Paroush, Schwartz, and Wolf [2008] and Davis, Pagano,and Schwartz [20 07])." Just how agents co mm unicatewith each other and respond to each others' opinions isa subject of ongoing research. Th e topic also opens anotherinterface with behavioral nance.

Price discovery acquires a different meaning in adivergent expectations environment, and this newmean ing has imp ortant implications for market structure.When asymmetric information characterizes a commu-nity of investors, the strategic behavior of informed agentscan affect the path that price takes when news moves ashare value from one equilibrium value to another. In thiscase, however, the new equilibrium is path in depe nden t.With divergent expectations, on the other hand, the newequuibrium is path dependent because it depends on how^the opinions of a diverse set of agents are integrated

(Paroush, Schwartz, and Wolf [2008]). Alternatively statedwith divergent expectations, price discovery is a coordi-nation process and, as such, it is directly affected by markestructure.In the standard asymmetric information environment, the key dichotomy isbetween informed and uninformed participants. But a second d ichotom y alsoexistsone that separates large institutional customersfrom small retail customers. One might expect that theinformed investor set would largely comprise institutionainvestors. After all, the institutions are professional andcan afford to continuously m onito r information and theyrespond to news, and the institutions' very size, all elseconstant, reduces the per share cost of doing so. Withdivergent expectations, however, there is no presu mptionthat institutional custom ers can, because of their size, con

sistently evaluate shares more accurately. On the contraryinstitutions commonly disagree with each other and, asa consequence, comm only trade with each other.In the divergent expectations environment, institutional investors do not necessarily have an advantage overetail customers as fundamen tal analysts. In fact, their sizmakes trading m ore difficult and they inc ur high er trans-action costs. So what accounts for their popularity? Thevalue added by mutual funds, pension funds, and otherinstitutional investors comes largely from their ability tofacilitate diversification. Further, they can b ring a systematic, professional, and disciplined approach to portfoliomanagement (Davis, Pagano, and Schwartz [2007]).FROM THEORY TO APPLICATION

Microstructure analysis is inherently involved withanalyzing the detailed functioning of a marketplace. T heliterature has a strong theoretical component and, to alarge extent, is structured to yield insights into the effecof market design (structure and regulation) on markeperformance. Theory should be able to provide a broadroadmap for real-world market architects to followHaving considered the development of microstructure theory, we turn now to the designing of an actuamarketplace. O ur focus is on D eutsche Brse and its electronic trading system Xetra. Deutsche Brse is the dominant stock exchange in Germany, the last of the majorEurope an bourses to go electronic, and it has technologythat is state of the art. Important insights were gainedfrom the microstructure literature during Xetra's planning period and the system's implem entation has marked

6 4 EQUITY MARKET MICROSTRUC TURE: TAKING STOCK OF WH AT W E KNOW FALL 200


9/16

a huge step forward forour roadmap, which is un Germany's equity markets. Butdoubtedly still incom plete today,was even more limited ih the 19941997 period whenXetra was being designed. And , there is always the dangerthat the cartographer, whose map is being used, holdscertain m isconceptions; for example, beUeving in the exis-tence of the No rthw est Passage.

The G erman Equity Market in the Mid-1990sAs recently as the niid-1990s, the German markethad major structural defects that would underm ine its com -petitiveness in the European arena. In recognition of this,Deutsche Brse, the newly founded exchange operator ofthe Frankfurter Wertpapierbrse (FWB), became theleading force for change in the German equity market.'''In the m id-1990s, Frankfurt's trading floor was themajor marketplace for German stocks, but the Germanmarket had become badly fragmented. Kursmaklers, theequivalent of specialists, concentrated much of the mar-ket's liquidity in their order books. By today's standards,a primitive electronic trading system, IBIS (which wasowned by FWB), operated in parallel with the market'sfloor trading. T he central icomponent of IBIS was an openlimit order book that had hit-and-take functionality, butdid not match orders automatically. The electronic systemcaptured about 40% of the trading volume in the 30 large-

cap DAX stocks, but no link existed between IBIS andthe floor. Seven other floor-based regional exchangeswere also operating in Germany with technical infra-structures that were similar to those in Frankfurt. In total,the regionals at that time were attracting roughlylO% ofGerm an exchange-based trading volume. Moreover, off-board trading has been, and still is, prevalent in Germ any(Davis, Pagano , and Schvvartz [2006]).Transparency for floor tradingin particular, pre-trade transparencywas low. Quotes were not distrib-uted publicly and were available on the floor only. Pricepriority between different trading venues was not enforcedand orders executed in one market commonly tradedthrough orders waiting to be executed in another market.Market m anipulation and, other abuses of power and posi-tion were believed to be rife on the old Frankfurt floor.Given the appreciable market fragmentation, poo r trans-parency, imperfect intermarket linkages, and dubious floo rbehavior, transaction costs were high. C hanges, bo th struc-tural and regulatory, were called for. The result was thedevelopment of Xetra} an electronic order-driven

FALL 2008

trading system that comprises two principal modali-tiesa continuous-order-book platform and periodicsingle-price call auctions.Des ignin g a Ne w Trading SystemXetra's development started in 1994, and the systemwas launched in 1997. Strong external forces also moti-vated this reengineering of the Deutsche Brse marketstructure. These forces included regulatory reform,soaring trading volumes, pan-European harmonizationof the exchange industry, vibrant cross-border compe-tition for order flow, and rising concerns of market par-ticipants about the future performance of Germany'sfmancial markets.Throughout the design stage of Xetra, microstruc-

ture theoryeven as it existed at the timewas an indis-pensable guide. This new field in fmancial economics,with its origin in issues concerning the competitive andarchitectural structure of an equity market, would beexpected to guide the development of an actual market-place such as Xe tra. To som e ex tent, it has fulfilled thispromise. The microstructure literature provided a broadroadmap to Deutsche Brse and highlighted underlyingrelationships and other important considerations that amarket architect should be aware ofBuilding the Xetra model involved specifying the

principles that the new market should implement anddefining the system's functionality. Most importantly, thenew market system was to provide equal and decentral-ized access to all of its participants. Further, the system'sfunctionality and the market information delivered to users,both pre - and post-trade, were to be the same for all traders,because a trader's location should be irrelevant. With thisin mind, the fundamental architectural decision of theDeutshe Brse was to structure a hybrid market thatincluded two major modalitiesa continuous electronicorder-driven platform and periodic call auctions that wereused primarily for market openings and closings.'^An absolutely critical attribute of an order-driventrading system is its ability, vis--vis its competitors, towin the battle for hquidity. The earlier microstructureliterature provided some guidance on this matter, but liq-uidity is a complex attribute. Because liquidity is not easyto define and measure, it has been very difFicult to modeland assess. However, as previously noted, the measure-ment and analysis of liquidity are currently attracting con-siderably more attention in the microstructure literature.

THE JOURNAL OF PORTFOLIO MANAGEMENT 6 5


10/16

Price discovery and transparency are two o ther areasin which the microstructure literature provided valuableguidance to the architects at Deutsche Brse. DeutscheBrse recognized that price discovery is a primary func-tion of a market center, and thus introduced the call auc-tions to sharpen the accuracy of price discovery, particularlyat the m arket's open and close. Understanding that trans-parency is important, but recognizing that it should notbe excessive, Deutsche Brse made the decision to dis-close only the indicative clearing price, rather than thefull bo ok of orders, in the pre-call, book -building period.

Microstructure literature has provided insight intothe operations of the public limit order book for con tin-uous trading. At the time Xetra was being designed, therecognition of periodic call auctions as a modality that wasclearly differentiated from, but could effectively be usedvwth, the con tinuous m arket, was just beginning to em erge.In terms of continuous trading, the m icrostructure analysesof the use of limit and market orders and of the interac-tion between these two order types proved to be mostvaluable. Today, a deeper understanding of the economicsof an order-driven market exists than was the case in the1994-1997 period when Xetra was being designed.

Another im portant con tribution of microstructuretheory has been the classification of traders according totheir needs for immediacy and their propensities to beeither givers or takers of liquidity. Th e differentiationbetween informed and uniformed traders also proved tobe valuable, particularly with respect to the market-makerrole that has been incorporated into Xetra. Specifically,market makers, referred to as designated sponsors, wereincluded to bolster liquidity provisions for smaller-capstocks. A balance had to be achieved between the obhg-ations imposed on the designated sponsors and the priv-ileges granted to them. To accomplish this, informationhad to be assessed co ncernin g the role of dealers in g en-eral (e.g., NASDAQ-type market makers) and specialistsin particular (e.g., NYSE-type specialists). That balancedefined the designated sponsors' role in Xetra and securedtheir willingness to accept it. Market microstructureinsights also yielded the understanding needed to trans-form the specialist role into the newly designed desig-nated-sponsor role.

But designing an automated trading system is indeeda complex task, and the gap between theory and imple-mentation is both large and intricate. Trading decisionscan be made in a large variety of ways that run the gam utfrom humans interacting directly with humans without

computers, to humans trading via electronic order handling and execution systems, to computers making tradingdecisions that are sent electronically to a computerizedmarket (e.g., computer-driven algorithmic trading). Sincethe mid-199 0s, market structure developm ent has mainlyinvolved the design of ari electronic trading facility.Deutsche Brse took into account the fact thaautomation impacts both the way in which trading decisions are made and the process by which prices are determined and trades executed in a market center. Anelectronic market requires the specification of an array ocritical features, for example, the trading modalitieemployed, rules of price and quantity determ ination, andbasic features, such as order types and trading parametersWith an electronic market, the software that implementa desired market structure must be specified on a level odetail which far exceeds what is required for human-intermediated trading.

For instance, a human agent, or specialist, has historically handled price determination at NYSE openingsThis function is performed with reference to various rulesbut the specialist is also fi-ee to exercise reasonable judgmentFurther, hum an-to-h um an interactions can evolve naturallyas problems, opportunities, and new competitive pressurearise. In contrast, with a fuUy electronic opening, everypossible condition that can occur must be recognized anda rule for dealing with it specified; electronic interactioncan be changed only by rewriting the code that specifiewith step-by-step precision just how orders are handledand turned into trades and transaction prices.

Ho w does one achieve the precise specifications thaa computerized trading system must have? In 1994, thmarket architects at Deutsche Brse could study the operations of othe r electronic platforms (e.g., CATS in Torontoand CA C in Paris). Doin g so was helpful, but of limitedvalue, given that Deutsche B rse was looking to developa distinctive system.W hen moving into new territory, market structur

development is a venture. Ho w does one know in advancif it will work? H ow can one determine if the new systemwill be viable from a business perspective? Neverthelessdesign decisions have to be made, technical requirementmust be specified, and the system must be built. Th e d ecisions involved represent huge financial bets abou t w heth ea new market structure will attract sufficient liquidityPrototyping a new market in the design phase helps theassessment process, but doing so in 1994 was considerablymo re difficult than it \vould be today with the advent o

66 EQUITY M ARKET M ICROSTRUCTURE: TAKING STOCK OF WHA T WE KNOW FALL 200


11/16

superior inform ation technology and testing capabilities.In 1994, the architects were forced to rely primarily ontheir own educated judgrhents and the insights they mayhave gained from microstructure research.Those who are involved in the design of an actualmarket realize that the devil is in the details. Consider, forinstance, the specification qf a call auction. A call has excel-lent theoretical properties, but ho w should an actual auc-tion be designed? It is straightforward to say that the m arketclearing price in a call auction should be the value thatmaximizes the number of shares that trade. But what shouldthe specific rule be for selecting the clearing price if twodifferent prices result in the same maximum trade size?Additionally, how transparent should the book be in thepre-call, order-entry period? Are further design featuresneeded to counter the possibihty of gaming? And so on.Other considerations that for the most part wereoutside the scope of the microstructure literature alsocame into play during the design of Xetra. Informationtechnology issues, such as'scalability, open architecture,and system reliability, were of critical importance. So toowere procedures for post-trade clearing and settlement.O ne of the final steps in the structural design of the newGerman market was the introduction in 2003 of a cen-tral counterparty so that counterparty risk managementwould be centralized and trading would become fullyanonymous, both pre - and post-trade. Electronic trading

was also a prerequisite for highly efficient straight-throughprocessing, w hich involves all stages ofa trade's life cycle.Information technology has further facilitated the timelycapture of market data (all trades, quotes, market indexvalues, and so forth) and has expedited its delivery tousers. With regard to these diverse applications, DeutscheBrse achieved a closer integration between trading onXetra and the broader market infi-astructure.THE ROADMAP AND THE ROAD

A m arket architect m ust have a roadmap that, broadlyspeaking, explains where to head and roughly ho w to getthere. The microstructure literature has added clarity,articulation, and intellectual support to the marketroadmap. Briefly stated, the roadmap's objective is toreduce trading frictions (costs), sharpen price discovery,and facilitate quantity discovery. The means of achievingthis broad objective involve the amassing of liquidity,which is done through the appropriate use of limit orderbooks for both continuous and call auction trading and.

where appropriate, the inclusion of broker/dealer inter-mediaries. Further insights are gained from microstruc-ture s in-dep th analyses of trading motives, which includenew information, liquidity needs, and technical tradingsignals. The hterature has also provided guidance withregard to issues such as transparency and the consolida-tion (fragmentation) of order flow.But theory, even if it does provide a goo d roadmap,can take one only so far. The closer one gets to the designof an actual system, the more apparent the complexitiesof trading and trading systems beco me. Th e road actuallytraveled is indeed bumpy and it is hazardous. Systemdesigners know that the devil is in the details. They haveto grapple with issues ranging from scalability, reliability,and other IT requirements, to business considerations thatconc ern the ultimate profitability ofa trading venue. T hemarket architects at Deutsche Brse recognized theseissues and their new system, Xetra, has marked a hugestep forward for the German equity market.Today, important problems persist with regard tomarket design in Germany, as well as in all other marketsaround the w orld. Wh at is the best way to deal with large,institutional orders? How is liquidity creation best han-dled for mid-cap and small-cap stocks? These are two fun-damental questions concerning market architecture thathave yet to be adequately answered. At the same time,important microstructure topics continue to emerge atacademic research desks. Are there limits beyond whichmicrostructure theory cannot provide guidance? Are therelimits to the level of efficiency that a real-wo rid market canever achieve? Undoubtedly, both answers are "yes" but,without question, neither of these Hmits has as of yet beenreached. Quite clearly, microstructure research and thedesign of an actual marketplace remain works in progress.ENDNOTES

'For a current discussion, see Davis, Pagano, and Schw artz[2007].Another major issue addressed by the niicrostructure lit-erature at that time was the impact of information on tradingvolume and price (Copeland [1976], Beja and Hakansson[1977], and Beja and Goldman [1980]).'For further discussion, see Co hen et al. [1979].""A market supported by informational trading only canfunction if agents trade with each other because theirexpectations are divergent. When the information that triggerstrading is common knowledge, the condition may be thoughtof as one in which agents are agreeing to disagree.

FALL 2008 THE JOURNAL O= PORTFOLIO M ANAGEM ENT 6 7


12/16

a more recent discussion of automated trading sys-tems, see Domowitz and Steil [1999].''Cohen et al. [1981] described the tradeofFbetween exe-cution probability and price improvem ent in the optimal choicebetween limit and market orders.^See Back and B aruch [2007] for a recent discussion andfurther references.*See Economides and Schwartz [1995] for a descriptionof alternative call market structures.'Trading systems differ in their degree of transparency.Pagano and Rell [1996] investigated whether greater trans-parency enhances market liquidity by reducing the opp ortun i-ties for taking advantage of uninformed participants."Also see Hasbrouck [1995], Harvey and Huan g [1991],and Jones, Kaul, and Lipson [1994]. Furthe r references are pro-vided by Menkveld, K oopman, and Lucas [2007]."Also see Bessler [2006] for discussion and further ref-

erences.'^For further discussion and references regarding liquiditysee Amihud and Mendelson [1986], Chordia, Roll, and Sub-rahmanyam [2000, 2008], Hasbrouck and Seppi [2001], Amihud[2002], and Pastor and Stambaugh [2003]."Adaptive valuation behavior refers to individual agentsbecoming more bullish (bearish) when they learn of the rela-tively bullish (bearish) attitudes of others.'' 'FWB also owned the futures and options exchangeDeutsche Termine Brse. After the 1997 merger with SOFFEX ,DTB became Eurex."Interestingly, the rnicrostructure literature on call auc-tions was relatively sparse at that time. For an early discussion,see Handa and Schwartz [1996b].REFERENCESAmihud, Yakov. "Illiquidity and Stock Returns: Cross-Sectionand Tim e-Series Effects." Jowrna/ of Financial Markets, 5 (2002),pp. 31-56.Amihud, Yakov, and Haim Mendelson. "Dealership Market:Market-Making with Inventory." Jowma/ of Financial Economics,8 (1980), pp. 31-53.

. "Asset Pric ing and th e Bid-Ask Spread." JoMma/ of Finan-cial Economics, 17 (1986), pp. 223-24 9.Angel, James. "Tic k Size, Share Prices, and Stock Splits."of Finance, 52 (1997), pp. 655-681.Back, Kerry, and Shmuel Baruch. "Working Orders in LimitOrd er Markets and Floor Exchanges." Journa/ of Finance, 62(2007), pp. 1589-1621 .

Bagehot, Walter. "The Only Game in Town." Financial AnlystsJournal, Vol. 27, No. 2 (March/April 1971), pp. 12-14Barclay, Michael, Terrence Hendershott, and TimothMcCormick. "Competition Among Trading Venues: Information and Trading on Electronic Communications Networks.Journal of Finance, 58 (2003), pp. 2637-2666.Beja, Avraham, and M. Barry Goldman. "On the DynamiBehavior of Prices in Disequilibrium." Journal of Finance,(1980), pp. 235-248.Beja, Avraham, and Nils H. Hakansson. "Dynamic MarkeProcesses and the Rewards to U p-to -D ate Information." JOMWof Finance, 32 (1977), pp. 291-304.Bernstein, Peter. "Liquidity, Stock Markets, and MarkeMakers." Financial Management, Vol. 16, No. 2 (Summer 19pp. 54-62.

. "The Surprising Bond between CAPM and the Meaninof Liquidity." A Guide to Liquidity. New York: InstitutionInvestor, 2007.Bessembinder, Hendrik. "Trade Execution Costs on Nasdaand the NYSE : A Post-Reform Comp arison." Jowrna/ of Finacial and Quantitative Analysis, 34 (1999), pp. 387-407.

"Quote-Based Competition and Trade Execution Costin NYSE-Listed Stocks." Jowrtta/ of Financial Economia, 70 (2pp. 385-422.Bessembinder, Hendrik, and Herbert Kaufman. "A Comparison of Trade Execution Costs for NYSE and Nasdaq-ListeStocks!' Journal of Financial and Quantitative Analysis, 32 (pp. 287-310.

. "A Cross-Exchange Comparison of Execution Costs anInformation Flow for N YSE-listed Stocks." Jowrna/ of FinancEconomia, 46 (1997b), pp. 293-319.Bessler, Wolfgang, ed. Bsen, Banken und Kapitalmrkte. BDuncker & Humblot, 2006.Biais, Bruno, Larry Glosten, and Chester Spatt. "MarkeMicrostructure: A Survey of Microfoundations, EmpiricaResu lts, and Policy Implications." JoMma/ of Financial Mark8 (2005), pp. 217-264.Biais, Bruno , Pierre HiUion, and Chester Spatt. "An EmpiricaAnalysis of the Limit Order Book and the Order Flow in thParis Bourse." JoMma/o/Finance, 50 (1995), pp. 1655-1689.

68 EQUITY M ARK ET M ICROSTRUCTURE: TAKING STOCK OF WHA T W E KNOW FALL 200


13/16

Chordia, Tarun, Richa rd R oll, and Avanidhar Subrahmanyam."Commonality in Liquidity." JoMwa/ of Financial Economics, 56(2000), pp. 3-28.

. "Liquidity and Market Efficiency." Journal of FinancialEconomics, 87 (2008), pp. 249-268.Christie, William, Jeffrey Harris, and Paul Schultz. "Why DidNASDAQ Market Makers Stop Avoiding Odd-Eigh th Q uotes?"Journal of Finance, 49 (1994)|, pp. 1841-1860.Christie, William, and Paul Schultz. "Why Do NASDAQMarket Makers Avoid Odd^Eighth Quotes?" Journal of Finance,49 (1994), pp. 1813-184 0. 'Cohen, Kaiman, Steven Maier, Robert Schwartz, and DavidWh itcomb. "Market Makers and the Market Spread: A Reviewof Recent Literature." Journal of Financial and QuantitativeAnalysis, 14 (1979), pp. 8131-835.

I. "Transaction Costs,i Order Placement Strategy, and

Existen ce o f the Bid-A sk Spread ." JoMma/ of Political Economy,Vol. 89, No. 2 (1981), pp. 2'87-305.Copeland, Thomas. "A M'odel of Asset Trading under theAssum ption of Seq uential]In form ation Arrival." Jowrwij/ ofFinance, 31 (1976), pp. 1149-1168.Copeland , Tho mas, and Dan, Galai. "Information Effects on theBid -Ask Spread." JoMWij/ ofknance, 38 (1983), pp. 1457-146 9.Davis, Paul, Michael Pagano, and Robert Schwartz. "LifeAfter the Big Board G oes Electronic." Finandal AnalystsJournal,Vol. 62, No. 5 (2006), pp. l|^-2 0.

ortfolio Management,"Divergent Expectations."Vol. 34, No. 1 (2007), pp. 84-9 5.Demsetz, H arold. "Th e C ost of Transacting." Quarterly Journalof Economics, Vol. 82, No. 1 (1968), pp. 33-53.IDom owitz, Ian, and Benn Steil. "Automation, Trading Costs,and the Structure of the Securities Trading Industry." Brookings-Wliarton Papers on Financial Services, 1999, pp . 33-92.Dom owitz, Ian, Jack Glen, and Ananth Madhavan. "Liquidity,Volatility, and Equ ity T radingiCosts Across Co untries and OverTime." International Finance, 4 (2001), pp. 221-256.

IDufFie, Darell, Nic olae Grleanu, and Lasse Pedersen. "Valua-tion in Over-the-Counter Markets." Review of Financial Studies,20 (2007), pp. 1865-1 900. 1

Easley, David, and Maureen O'Hara. "Price, Trade Size, andInform ation in Se curities M arkets." JoMr/w/ of Financial Eco-nomics, 19 (1987), pp. 69-90.

. "Order Form and Information in Securities Markets."Journal of Finance, 46 (1991), pp. 905-928.

. "Time and the Process of Security Price Adjustment."Journal of Finance, 47 (1992), pp. 577-606.Economides, N icholas, and Robe rt Schwartz. "Electronic CallMarket Trading." Jowma/ of Portfolio Management (Spring 1995pp . 10-18.Engle, Robert, and Chve Granger. "Co-Integration and ErrorCorrection: Representation, Estimation, and Testing." Econo-metrica, 55 (1987), pp. 251-276.Foucault, Thierry. "Order Flow Composition and TradingCosts in a Dynamic Order Driven Market." Joi/rna/ of FinancialMarkets, 2 (1999), pp. 99-134.Foucault, Thierry, Ohad Kaden, and Eugene Kandel. "TheLimit Order Book as a Market for Liquidity." Review of Finan-cial Studies, 18 (2005), pp. 1171-1217.Garman, Mark. "Market Microstructure." Journal of FinanciaEconomics, 3 (1976), pp. 33-53.Glosten, Lawrence. "Is the Electronic Open Limit Order BookInevitable?" Jonma/o/F/ijce, 49 (1994), pp. 1 127 -116 1.Glosten, Lawrence, and Paul Milgrom . "Bid , Ask, and Trans-action Prices in a Specialist Market with HeterogeneouslyInformed Agents." Journa/ of Financial Economics, 14 (1985)pp . 71-100.Goettler, Ronald, Christine Parlour, and Uday Rajan. "Equi-librium in a Dynamic Limit Ord er Market." Jowma/ of Finance,60 (2005), pp. 2149-2192.Grossman, Sandford. "The Information Role of Upstairs andDownstairs Markets." JoMma/ of Business, 65 (1992), pp. 50 9-529 .Grossman, Sandford, and Joseph Stiglitz. "O n the Impossibilityof Informationally Efficient Markets." American Economic RevieVol. 70, No. 3 (1980), pp. 393-408.Handa, Puneet, and Robert Schwartz. "Limit Order Trading."Journal of Finance (December 1996a), pp. 1835-1861.

FALL 2008 T HE J OURNAL OF PORT FOLIO M ANAGE M E NT 6 9


14/16

. "How Best to Supply Liquidity to a Securities Market."Journal of Portfolio Management (Winter 1996b) , pp . 4 4 - 5 1 .Handa, Puneet, Robert Schwartz, and Ashish Tiwari. "QuoteSetting and Price Formation in an Order-Driven Market."

Journal of Financial Markets, 6 (2003) , pp . 461-489.Harris, Larry. "Stock Price Clustering and Discreteness." Reviewof Financial Studies, 4 (1991), pp. 389-4 15.

. "M inim um Price Variations, Discrete Bid-Ask Spreads,and Quotation Sizes." Review of Financial Studies, 1 (1994),pp. 149-178.. Trading and Exchanges: Market Microstructure or P ractitioners.New York: Oxford University Press, 2003.

Harvey, Campbell, and Roger Huang. "Volatility in the For-eign Currency Futures Market." Review of Financial Studies,4 (1991), pp. 543-5 69.Hasbrouck, Joel. "Measuring the Information Co nten t of StockTrades." JoMwa/ of Finance, 46 (1991), pp. 179-207.

. "Assessing the Quality of a Security Market: A NewApproach to Transaction-Cost Measurement." Review of Finan-cial Studies, Vol. 6, No. 1 (1993), pp. 191-21 2.

. "O ne Security, Many Markets: Determ ining the Co n-tribution to Price Discovery." Jowma/ of Finance, 50 (1995),pp. 1175-1199.

. Empirical Market Microstructure. New York : Oxfo rd Un i -versity Press, 200 7.Hasbrouck, Joel and Duane Seppi. "Co mm on Factors in Prices,Ord er Flows, and Liquidity." Jowma/ of Financial Economics, 59(2001), pp. 383-411.Hasbrouck, Joel, and George Sofianos. "The Trades of MarketMake rs: An E mpirical Analysis of NYSE Specialists." Jouma/ ofFinance, 48 (1993), pp. 1565-159 3.Ho, Thomas, and Hans StoU. "O n Dealer Markets under Co m-petitio n." Journa/ of Finance, 35 (1980), pp. 259-267.

. "Optimal Dealer Pricing under Transactions and R etu rnUncertzinty" Journal ofFinancial Economics, 9 (1981), pp. 4 7 - 7 3 .

. "Th e Dynamics of Dealer Markets under C ompetition."

H o, Thomas, Robert Schwartz, and D. Whitcomb. "TheTrading Decision and Market Clearing under TransactionPrice Uncertainty." JoMWfl/ of Finance, Vol. 40, No. 1 (1985pp. 21-42.Jones, Charles, Gautam K aul, and Marc Lipson. "In formationTrad ing, and Volatility." JoMrnii/ of Financial Economics, 36 (1pp. 127-154.Keim, Donald, and Ananth Madhavan. "The Upstairs Markefor Large Block Transactions: Analysis and Measurement oPrice Effects." Review of Financial Studies, 9 (1996), pp. 1-3Kyle, Albert. "Co ntinuo us Auctions and Insider Trading." Econmetrica, 53 (1985), pp. 1315-1335 .Madhavan, Ananth. "Consolidation, Fragmentation and theDisclosure of Trading Inform ation." Review of Financial Stud8 (1995), pp. 579-603 .

. "Market Microstructure." of Financial Markets,(2000), pp. 205-258.Madhavan, Ananth, and Minder Cheng. "In Search of Liquidity: An Analysis of Upstairs and D ownstairs Trades." Revieof Financial Studies, 10 (1997), pp. 175-204.Mendelson, M orris, Junius Peake, and T Williams. "Toward aModern Exchange: The Peake-Mendelson-Williams Proposafor an Electronically Assisted Auction Market." In ImpendinChanges or Securities Markets: Wliat R ole or the Exchange? E . B l o c hand R . Schwartz, eds. Greenwich, CT : JAI Press, 1979.Menkveld, Albert, Siem Koopman, and Andr Lucas. "Modeling Around-the-Clock Price Discovery for Cross-Listed StockUsing State Space Methods." JOMCHJ/ of Business an d EconoStatistics, Vol. 25, No. 2 (2007), pp. 213-225.Mildenstein, Eckart, and Harold Schleef. "The Optimal PricinPolicy of a Monopolistic M arketmaker in the Equity Market.Journal of Finance, 38 (1983), pp. 218-231.Milgrom, Paul, and Nancy Stokey. "Information, Trade, anCom mon Knowledge." Jowma/ of Economic Theory, Vol. 26, N(1982), pp. 17-27.Miller, Edward. "Risk, Uncertainty, and Divergence oOpinion." JoMrJj/ of Finance, 32 (1977), pp. 1151-1168.

Journal of Finance, 38 (1983), pp. 1053-1074. O'Hara, Maureen. Market Microstructure Theory. CambrMA: Basil Blackwell, 1997.

7 0 EQUITY MARKET MICROSTRUCTURE: TAKING STOCK OF WHAT W E KNOW FALL 200


15/16

Ozenbas, Deniz, Rob ert, Schwartz, and R obe rt Wood."Volatility in U.S. and European Equity Markets: An Assess-ment of Market Quality." International Finance, Vol. 5, No. 3(2002), pp. 437-461. ,

Pagano, Marco, and Ailsa RpeO. "Transparency and Liquidity:A Comparison of Auction and Dealer Markets with InformedTrading." JoMwa/ of Finance, 51 (1996), pp. 579-611.Parlour, Christine. "Price D ynamics in Limit Order M arkets."Review of Financial Studies, l i (1998), pp. 789-81 6.Parlour, Christine, and Duane Seppi. "Limit Order Markets:A Survey." In Handbook of Financial Intermediation and Banking,A.W.A. Boot and A.V. Thako r, eds. Amsterdam: Elsevier, 2008.

IParoush, Jacob, Robert Schwartz, and Avner Wolf. "TheDynamic Process of Price Discovery in an Equity Market."Working paper, Baruch College at CUNY, 2008.Pastor, Lubos, and Robert Stambaugh. "Liquidity Risk andExpected Stock Retu rns. " Jowrna/ of Political Economy, 113 (2003),pp. 642-685.Porter, David, and Daniel Weaver. "Post-Trade Transparencyon Nasdaq's National Market System." JoHma/ of Financial Eco-nomics, Vol. 50, No. 2 (1998), pp. 231-252.Sandas, Patrick. "Adverse Selection and Competitive MarketMaking: Empirical Evidence from a Limit Order Market."Review of Financial Studies, 14 (2001), pp. 705-734.Schwartz, Robert. Reshaping ihe Equity Markets: A Guide or the1990s. Ne w York: Harper Business, 1991.Schwartz, Ro bert, and Reto Francioni. Equity Markets in Action.Hoboken, NJ: John Wiley and Sons, 2004.

Seppi, Duane. "Equihbrium Block Trading and AsymmetricInform ation." Jowma/ of Finance, 45 (1990), pp. 73-94 .Smidt, S. "Which Road to an Efficient Stock Market: FreeCompetition versus Regulated Monopoly." Financial AnalystsJournal, 27 (September/October 1971), p.l8.Stigler, Ceorge. "Public Regulation of the Securities Markets."Journal of Business, Vol. 37, No. 2 (1964), pp. 117-142.Stoll, Hans. "The Supply of Dealer Services in Securities Mar-kets." Journal of Finance, 33 (1978), pp. 1133-1151 .

. "Inferring the Components of the Bid-Ask Spread:Th eor y and Em pirical Tests." Jonma/ of Finance, 44 (1989),pp. 115-134.

Tinic, Sneha. "The Economics of Liquidity Services." QuarterlyJournal of Economics, Vol. 86, No . 1 (1972), pp. 79-93.Tinic, Sneha, and Richard West. "The Securities Industry underNegotiated B rokerage Com missions: Changes in the Structureand Performance of New York Stock Exchange MemberFirms." Bell Journal of Economics, 11 (Spring 1980), pp. 29-41U.S. Securities and Exchange C omm ission. Institutional InvestoStudy Report. Washington, D.C.: U.S. Government PrintingOffice (1971).

To order reprints of this article, please contact Dewey [email protected] or 212-224-3675.

FALL 2008 TH E JOURNAL OP PORTFOLIO MANAGEMENT 7 1


16/16

equity market micro structure

Documents