Research Article A Good Beginning Makes a Good Market: The Effect of Different Market Opening Structures on Market Quality

Transcription

1 e Scienific World Journal Volume 2015, Aricle ID , 14 pages hp://dx.doi.org/ /2015/ Research Aricle A Good Beginning Makes a Good Marke: The Effec of Differen Marke Opening Srucures on Marke Qualiy Gerno Hinerleiner, 1 Ulrike Leopold-Wildburger, 1 Roland Mesel, 2,3 and Sefan Palan 2,4 1 Deparmen of Saisics and Operaions Research, Karl-Franzens Universiy Graz, Universiassraße 15/E3, 8010 Graz, Ausria 2 Insiue of Banking and Finance, Karl-Franzens Universiy Graz, Universiassraße 15/F2, 8010 Graz, Ausria 3 Deparmen of Banking and Insurance, Universiy of Applied Sciences Joanneum, Eggenberger Allee 11, 8020 Graz, Ausria 4 Deparmen of Banking and Finance, Universiy of Innsbruck, Universiassraße 15/4, 6020 Innsbruck, Ausria Correspondence should be addressed o Sefan Palan; sefan.palan@uibk.ac.a Received 13 June 2014; Acceped 8 Ocober 2014 Academic Edior: Sarah Brown Copyrigh 2015 Gerno Hinerleiner e al. This is an open access aricle disribued under he Creaive Commons Aribuion License, which permis unresriced use, disribuion, and reproducion in any medium, provided he original work is properly cied. This paper deals wih he marke srucure a he opening of he rading day and is influence on subsequen rading. We compare a single coninuous double aucion and wo complemen markes wih differen call aucion designs as opening mechanisms in a unified experimenal framework. The call aucions differ wih respec o heir levels of ransparency. We find ha a call aucion no only improves marke efficiency and liquidiy a he beginning of he rading day when compared o he sand-alone coninuous double aucion, bu also causes posiive spillover effecs on subsequen rading. Concerning he design of he opening call aucion, we find no significan differences beween he ransparen and nonransparen specificaion wih respec o opening prices and liquidiy. In he course of subsequen coninuous rading, however, marke qualiy is slighly higher afer a nonransparen call aucion. 1. Inroducion The opening of markes is a crucial ime in he rading day, because i consiues he saring poin of he price discovery process and he beginning of he incorporaion of overnigh informaion ino prices. The uncerainy abou he fundamenal value of he asse is relaively high, leading o lower marke qualiy in erms of mispricing and liquidiy compared o inraday rading [1 3]. Therefore, he quesion of which rading srucure is bes suied for he opening carries paricular relevance. The sudy a hand focuses on he effec of opening call aucions on several measures of marke qualiy. The objecive is o deermine which marke srucure is ideal for he marke opening and wheher here exiss a spillover effec from an opening call aucion on subsequen coninuous rading. As will be made clear below, he relevan lieraure on hese issues hascomeodivergenresuls.onereasonforhisfacmigh be he differences in he applied research mehodologies. Our sudy aims a bridging he gap beween he resuls of empirical and heoreical research by using a sable environmen. For hree reasons we op for an experimenal approach. (1) When using field daa, informaional efficiency can only be measured using esimaes of he underlying fundamenal value. (2) The disribuion of informaion is no known in such empirical sudies. (3) Comparing hree kinds of marke srucures in one framework, which is a prerequisie for our research quesion, is no possible on real sock markes. In conras o earlier sudies, our examinaion exends he exising lieraure in wo imporan respecs. Firs, our sudy no only invesigaes he marke qualiy of call aucions a he marke opening, bu also analyzes he effec of an opening call aucion on he subsequen coninuous rading phase. We achieve his by comparing a coninuous double aucion afer an opening call aucion wih a sand-alone coninuous double aucion. We furhermore analyze wo differen kinds of call aucions: a ransparen design wih an open order book and a nonransparen version where no order informaion is available during he order enry phase. We separaely evaluae he effecs of hese wo specificaions concerning marke

2 2 The Scienific World Journal qualiy for boh, marke opening and subsequen coninuous rading. Wefindhabohversionsofheopeningcallaucion significanly increase he informaional efficiency of opening prices relaive o he sand-alone coninuous double aucion. Regarding he effec of an opening call aucion on he subsequen coninuous rading phase, we discover a posiive spillover effec in erms of a higher liquidiy compared o he single rading venue wihou an opening call aucion. We also find ha he ransparen version of he call aucion does no lead o a more efficien opening price, nor generae higher liquidiy a he opening, han he nonransparen call aucion. The coninuous double aucion afer he ransparen opening aucion furhermore exhibis a parly lower marke qualiy han he coninuous rading phase afer he nonransparen call aucion. This paper is organized as follows: Secion 2 reviews he lieraure and describes he experimenal design and Secion 3 presens and discusses he resuls. The paper concludes wih some suggesions for fuure research on call aucion and marke opening issues. 2. Maerials and Mehods 2.1. Lieraure Review. Hisorically, many sock exchanges employed a coninuous rading mechanism for heir mos liquid securiies over he enire rading day. Nowadays, however, mos wesern capial markes have swiched o using a call aucion as heir opening mechanism. The mos frequenly advanced argumen in suppor of his focuses on he abiliy of call aucions o collec orders and pool liquidiy by including differen orders from differen invesors. One of he firs sudies confirming his argumen is ha by Madhavan[4],whofindshainhecaseofhighinformaion asymmery, a call aucion srucure is bes suied in erms of informaional efficiency and liquidiy. He concludes ha a call aucion should be used a difficul rading imes such as he marke opening. Along his line, Economides and Schwarz [5], Domowiz and Madhavan [6], and Pagano and Schwarz [7] poin ou ha in call aucions ransacion coss are reduced compared o coninuous rading srucures becauseofloweradverseselecioncossandbecauseofhe absence of a free rading opion (A free rading opion is creaed when a rader eners a limi order in he coninuous double aucion, hus giving oher raders he opporuniy o buy he sock a he limi price. If sock prices suddenly shif in he wrong direcion and he iniiaor is no able o change he limi price in ime, he iniiaor may realize a loss. This sourceofradingriskiseliminaedinhecallaucion,sinceall rades occur a he same ime a one marke clearing price, cf. Soll [8]). The advanages of call aucions are also suppored by Pagano and Röell [9], Snell and Tonks [10], and Aiken e al. [11] among ohers. Ellul e al. [12] documen ha price discovery a he opening on he London Sock Exchange is more efficien in he call aucion han in he dealer marke, alhough he call aucion is associaed wih higher ransacion coss, especially in he presence of high informaional asymmeries and grea order imbalance. Comeron-Forde e al. [13] analyze he inroducion of a call aucion a he Singapore Sock Exchange and find ha informaional efficiency is enhanced relaive o he mechanism previously in place, which employed a coninuous double aucion. Beside he menioned heoreical and empirical invesigaions, several sudies invesigaed he quesion of marke opening in an experimenal framework. Noussair and Tucker [14] and Palan [15] provide wo recen reviews of he lieraure on experimenal marke research. The paper which comes closes o our sudy is Theissen [16]. He compares a call aucion, a coninuous double aucion and a dealer marke. He discovers ha a call aucion does no lead o lower informaional efficiency compared o he coninuous double aucion bu generaes higher liquidiy. The wors performance inermsofmarkequaliyisexhibiedbyhedealermarke. Noe ha he analyses above do no invesigae he influences of differen designs of opening call aucions on marke efficiency. Madhavan [17] and Pagano and Röell [18] compare a nonransparen call aucion, in which marke paricipans have no informaion abou he order book, wih a ransparen version. Boh conclude ha he laer should lead o higher marke qualiy due o he greaer amoun of informaion revealed during he bidding process. Analyzing he preopening phase of ransparen call aucions, Biais e al. [19] and Davies [20] show ha marke qualiy is higher in he laer sage of he preopening phase compared o he beginning a he Paris Bourse and he Torono Sock Exchange, respecively. In conras o hese resuls, Madhavan e al. [21] repor ha higher ransparency a he Torono Sock Exchange proved derimenal o marke qualiy. From his Eom e al. [22] conclude ha he evidence on he effec of prerade ransparency on marke qualiy is mixed. Beside he impacs of differen marke opening designs on marke efficiency and liquidiy a he beginning of he rading day, some invesigaions deal wih he quesion wheher a specific marke opening design may also influence marke qualiy in he subsequen coninuous rading phase spillover effec. Since differen sock exchanges have been using differen specificaions of call aucions, quesions concerning, for example, he ransparency during he preopening phase or he shape of prioriy rules, have araced increased aenion in he lieraure. Chang e al. [23] repor posiive spillover effecs afer he inroducion of an opening call aucion a he Singapore Sock Exchange. Pagano e al. [24] show ha hese resuls exend o NASDAQ. Comeron-Forde e al. [13] on he oher hand documen ha marke qualiy declined afer he inroducion of an opening aucion a he Hong Kong Sock Exchange. This decline is accompanied by increasing spreads, which in urn lead o reduced liquidiy. The auhors aribue his problem o an unsuiable design of he opening mechanism which renders i unaracive for raders. Gerace e al. [25] analyze he spillover effec of a more ransparen opening call aucion on coninuous rading a he Shanghai Sock Exchange (SHX). The auhors repor ha he volailiy and spreads during inraday rading were reduced afer he inroducion of indicaive prices. This general resul receives furher suppor from Hoffmann and van Bommel [26], who compare socks lised a he more ransparen

3 The Scienific World Journal 3 Euronex Paris wih ones lised a he slighly less ransparen German Sock Exchange. In heir analysis, informaional efficiency and liquidiy in coninuous rading is higher a he Euronex Paris. We hus conclude our review of he relevan lieraure by idenifying he gap we inend o fill. No sudies have hihero analyzed a call aucion and a subsequen coninuous double aucion in a unified framework, invesigaing boh he direc and he spillover effecs, and complemening his invesigaion by analyzing he differenial impac of ransparency. This is wha we will address Experimenal Design. In oal, we conduced 21 sessions from Ocober 2009 o November 2010, using hree reamens wih seven sessions each. The experimen was programmed in z-tree [27]. Each session involved 12 paricipans from he suden body of he Faculy of Social and Economic Sciences a he Karl-Franzens-Universiy Graz, recruied using he ORSEE sysem [28]. Each paricipan ook par only once in his experimenal seing Descripion of Treamens. The experimenal design is based on he work of Theissen [16], who compares hree differen sand-alone rading srucures. In conras o his invesigaion, we have a five-minue sand-alone coninuous double aucion (CDA) and wo differen ypes of complemen markes, boh saring wih a wo-minue call aucion. Afer he opening price has been deermined, a five-minue coninuous double aucion begins also in hese reamens. The wo complemen markes differ wih respec o he ransparency level of he opening call aucion. In he firs complemen marke (CM) a compleely nonransparen call aucion is used, meaning ha no informaion abou he order flow or he indicaive prices is available o he paricipans. The second complemen marke (CMT) employs a fully ransparen call aucion; ha is, marke paricipans can see he whole order book and he indicaive prices during he enire duraion of he preopening phase. (Noe ha hese wo kinds of call aucions occupy he wo exremes in he coninuum of possible ransparency levels. Real sock markes are in general locaed somewhere in beween hese ransparency levels, alhough here are excepions like, e.g., he Paris Bourse, which employs a fully ransparen call aucion (cf. [29]).) Furhermore, he indicaive prices are seadily being updaed during he order enry phase. During he preopening phase, paricipans are allowed o ener and change (e.g., cancel) heir orders. To ener an order he rader hasoenerapriceandavolumeonheselecedmarkeside Marke Mechanisms. In he call aucion, he price which leads o he highes rading volume is used for boh indicaive prices and he opening price. In he case of wo or more prices wih he same rading volume, he middle of he price range leading o he maximum rading volume is used. Generally, buy (sell) orders wih prices higher (lower) han he opening price are being execued. If one marke side exceeds he oher in erms of order volume a he opening price, a raioning mechanism is applied. In conras o oher experimens (e.g., [16, 30]), his raioning mechanism uses a price-ime-prioriy rule like in Friedman [31]. If wo or more raders on one marke side have he same price limi, he order enered firs is processed firs. A he opening all paricipans receive informaion abou he opening price, he rading volume, and he order surplus. The coninuous double aucion has he same design in all reamens. The bes bid and ask prices are visible o all paricipans. A new bid (ask) is permissible if is price is higher (lower) han he curren bid (ask) price. If a new bid or ask order is enered and acceped, all orders enered earlier are deleed. A ransacion is execued if a rader acceps he curren bid or ask price. The order enry is again organized such ha paricipans have o ener price and volume on he seleced marke side. If a rader wans o accep a curren bid (ask) she only has o ener an order volume of beween 1 and hevolumeofhebid(ask)inhemarke Session Srucure and Payoffs. Each session of he sand-alone coninuous double aucion (complemen marke) lass approximaely 120 min (160 min). Firs, he paricipans ge insrucions explaining he rading rules, he srucure of he experimen, he payoff, and deails abou he raded asse and he informaion srucure. Nex hey complee a self-es in which hey have o answer quesions concerning he insrucions. If a paricipan gives an incorrec answer, he experimener is informed and helps he paricipan in order o achieve a level playing field in erms of subjec undersanding. In each experimenal session here are 8 rading periods, he firs wo of which serve as raining periods in which paricipans are able o ge acquained wih he rading inerface. These raining periods do no coun owards subjecs earnings, which are based only on subjecs performance in he following six rading periods. (Afer hese 8 rading periods we conduced a second experimenal reamen. A he beginning of he session, subjecs were made aware ha here wouldbeasecondparbudidnolearnanydeailsabou i.) This is communicaed o he sudens a he beginning of he experimen and is addiionally documened in he wrien insrucions hey receive. Each period is srucured as follows: he individual price signals are displayed on subjecs screens and subjecs have 20 seconds o conemplae hem before rading sars. In he complemen markes, subjecs hen have 120 seconds in which o submi call marke bids andasks.aferhe20secondshaveelapsedinhecda, he coninuous rading phase sars. Afer he 120 seconds have elapsed in he complemen markes, he opening price isdeerminedfromhesubmiedbidsandasksandsubjecs are informed abou he price and abou how many shares hey have raded. The informaion screen is displayed for 10 seconds, afer which he coninuous rading phase sars. Afer five minues his phase ends and subjecs endowmen of experimenal currency and cash is rese o heir iniial values, afer which rading resars for he nex period. (A he end of a rading period, subjecs do no receive any addiional informaion. In paricular, hey are no informed abou heir rading profis or he fundamenal value.)

4 4 The Scienific World Journal Table 1: Performance-dependen payou srucure of he experimen. Ranking Complemen markes CDA alone C 40.00C C 30.00C C 22.00C C 16.50C C 12.00C C 8.50C C 6.10C C 4.10C C 2.50C C 1.30C C 0.50C C 0.00C As in Theissen [16], each paricipan s iniial porfolio consiss of currency unis and 100 shares and is reiniialized a he beginning of every period. Subjecs can sell shor up o 100 shares and can overdraw heir currency accoun by up o a maximum of currency unis. No ineres and dividends are being paid in he experimen. Therefore, he profi of a paricipan resuls exclusively from he prices and quaniies of her rades and from he fundamenal value of her shares a he end of he experimen. The minimum ick size is one experimenal currency uni. The paymen for a paricipan consiss of a fixed show-up fee plus an amoun depending on he sum of her six periodend porfolio values. The paymen for he rading success is exponenially srucured depending on he rader s porfolio value rank wihin he session. The performance-dependen payouisshownintable1.thepayoffsrucurehusconveys a clear incenive o rade. A second reason for his ype of payou funcion is ha i allows for relaively large marginal incenives, while ruling ou he possibiliy of bankrupcy. When allowing for shor selling and margin buying, a payoff funcion linear in subjecs erminal wealh or profi admis he possibiliy of negaive final wealh. In order o circumven his problem, one sraegy would be o add a fixed lump-sum paymen o he oucome of he experimen and use an exchange rae from experimenal ino real currency which ensures habankrupcyisveryunlikely.however,inhiscasehe marginal incenives from performing well in he experimen are relaively small, since he difference in payous o good and bad performers is by necessiy small. Our payoff funcion allows for large marginal incenives (i.e., grea differences beween he bes and wors performers) wihou he risk of bankrupcy. Furhermore we argue ha convex incenives and incenives deermined by a rader s rank relaive o ohers are no uncommon in markes ouside he laboraory. Brown e al. [32], for example, argue ha he marke for muual funds can be considered a ournamen, since pas performance (relaive o ohers) drives new invesmen inflow and because fund managers rewards are usually ied oheassesundermanagemen.( Specifically,wesugges ha viewing he muual fund marke as a ournamen in which all funds having comparable invesmen objecives compee wih one anoher provides a useful framework for a beer undersanding of he porfolio managemen decision-making process. Similar o he payoffs for golf and ennis compeiions, he amoun of remuneraion ha a fund receives for winning his ournamen depends upon is performance relaive o he oher paricipans [32].) Similarly, Basak e al. [33] analyze his link beween fund flows and performance, noing ha i induces convexiies in managers objecives. A number of recen sudies are hus moivaed by he imporance of ournamen and convex incenives in real-world financial markes and analyze heir impac on rader behavior and marke oucomes (furher examples include [34 37]). In addiion o he variable payoff, every paricipan receives a 10 C show-up fee in he complemen marke and an 8Cshow-up fee in he coninuous double aucion seing. The differences in payoffs sem from he fac ha he complemen markes have a longer ime duraion of abou 30 minues compared o he seing wih he coninuous double aucion alone. On average, he paricipans of he differen reamens receive idenical expeced payoffs of 8.50 C per hour. Afer he experimenal periods have concluded, paricipans have o fill in a quesionnaire dealing wih he experimen in general and he behavior of he individual in paricular. Each period s fundamenal value V is independenly drawn from a discree uniform disribuion over all ineger values in he inerval [0.8 V 1,1.2 V 1 ], (Due o a programming error, in 1.85% of all cases (i.e., in he second period each of one sand-alone CDA and one nonransparen complemen marke) he fundamenal value changed by 24.1% from one period o he nex. Noe, however, ha subjecs were only informed abou he las period s fundamenal value.) The firs asse value of he six rading periods is randomly chosen from he inerval of 400 o 600 currency unis. (The asse values in he wo raining periods were 86 and 90. They were chosen far from he range of values employed laer for he six periods which couned owards payoffs in order o eliminae he possibiliy of anchoring.) The fundamenal value is no communicaed or displayed o he paricipans a any ime during or afer a rading period, excep for he very end of rading (i.e., afer period 8). Insead, each paricipan receives an individual, imperfec price signal for he fundamenal valueofheassebeforearadingperiodsars.inlinewih he calculaion of he fundamenal value, he price signals are uniformly disribued over all ineger values beween 10% below and above he fundamenal value. (During he preess of he final experimenal seing we also conduced some series wih a higher range of price signals and herefore a higher informaion asymmery. The resuls do no indicae ha his change in he price signal disribuion is associaed wihachangeinmarkequaliy.dueoaprogramming error, in 3.49% of all cases he signals deviaed by more han 10% (maximum: 12.01%) from he fundamenal value. Since subjecs did no learn he fundamenal value in periods 1 hrough 5, hey could have become aware of he problem in 0.47% of all cases (i.e., when a signal s deviaion was oo large in he las period), and only a he end of he experimen

5 The Scienific World Journal 5 afer rading had concluded. No subjec gave an indicaion of having become aware of one of hese cases.) Therefore, he price signal is of equal ex-ane qualiy for all raders, whereas he ex-pos signal qualiy differs in erms of precision beween he raders. We do however ensure ha, wihin each period, he average of he randomly calculaed price signals equals he fundamenal value of he asse. The variaion in price signals compued as he average absolue deviaion of each of he 12 price signals from he fundamenal value of he period ranged beween 3% and 6.5%. The average absolue variaion of price signals over all periods was 4.8%. All he facs concerning he informaion srucure wihin he experimenal marke were common knowledge. 3. Resuls and Discussion The resuls we presen are based only on he six experimenal rading periods. We srucure he presenaion ino resuls regarding he marke opening, regarding he preopening phase and regarding possible spillover effecs Marke Opening. We sar by analyzing he informaional efficiency of he opening prices of all reamens and obain he following resul. Resul 1. An opening call aucion prior o coninuous rading leads o a smaller opening price deviaion from fundamenal value and lower spreads han a coninuous double aucion alone. This resul holds for boh ransparen and nonransparen call aucion forms. Suppor for Resul 1. Themeanrelaiveerror(MRE)forhe opening prices P o, in a series measures he average absolue difference beween hese prices and he fundamenal value V, divided by he fundamenal value (cf. [16]). We firs define he relaive error (RE) as follows: RE o, = P o, V. (1) V The MRE for he opening prices is hen defined as he session average of he individual periods RE o, values as follows: MRE o = RE o,. (2) Here, is he rading period, running from 3 o 8 o exclude he raining periods 1 and 2. Lower MRE o values indicae higher informaional efficiency of he opening price. Using he MRE as a measure of informaional efficiency of a marke, we find ha he call aucion leads o a higher level of efficiency han he coninuous double aucion in erms of he opening price. We obain MRE o values of 2.61% (1.92%) [3.93%] in he ransparen call aucion (nonransparen call aucion) [sand-alone CDA]. To invesigae he robusness of hese differences, we conduc a number of saisical ess. Since he assumpion of a normal disribuion is rejeced for various daases in our analysis, we repor only nonparameric ess. I is worh menioning ha several hypoheses were also =3 esed by use of parameric mehods, yielding qualiaively equivalenresulsashereporednonparamericess. The differences beween he MRE o values of he hree reamens are invesigaed by a Kruskal-Wallis-ANOVA which shows ha hey differ significanly (P value: 0.035). In addiion we execue Wilcoxon rank sum ess comparing he MRE o values of he sand-alone CDA o hose from he call aucions. The comparison beween he opening prices of he sand-alone CDA and he nonransparen call aucion shows ha he laer generaes significanly more efficien opening prices (P value: 0.013). Comparing he prices from he ransparen call aucion o eiher he opening prices of he sand-alone CDA or he nonransparen call aucion prices yields no significan differences (P values: and 0.225, resp.). Finally, we compare he MRE o values of he call aucions o he MRE 120 seconds afer he sar of he sand-alone CDA. This comparison yields evidence on he compeiive performance of hese differen opening mechanisms, conrolling for he ime cos (i.e., he call aucions and he sandalone CDA in his comparison have boh been given 120 seconds o aggregae prices). The average MRE 120 seconds ino he sand-alone CDA is 3.56%, which is significanly (no significanly) differen from he MRE o yielded by he nonransparen (ransparen) call aucion (Wilcoxon rank sum es P values: and 0.225). However, hese ess suffer from a lack of power because hey mus compare session averages of he RE o, values for he six periods in each session (i.e., MRE o ), since he resuls from individual periods wihin a session are no independen of each oher. To remedy his shorcoming and make beer use of our daa while sill accouning for he possible dependence beween consecuive periods wihin a session, we regress RE o, on a consan, on dummies for he wo reamens using eiher ransparen or nonransparen call aucions, and on dummies for he period wihin a session, for he sessions hemselves, and for he ineracions of periods and sessions, adjusing for clusers in he sandard errors by session: RE o, =α+β CM CM +β CMT CMT (3) + β T + βt s s + βt s s +ε. In his regression, CM (CMT) is a dummy variable equal o 1 in he nonransparen (ransparen) call marke reamen and zero oherwise. The expressions β T, βt s,andβt s are ransposed coefficien vecors, is a vecor of dummies for periods 4 hrough 8, s isavecorofdummiesforsessions 2hrough7and s is he ineracion beween periods and sessions. The resuls are shown in regression model (1) in Table 2. We find significan reamen effecs for boh he ransparen and he nonransparen call aucions. We hen repea he analysis, again using RE o, for he opening prices in he reamens wih call aucions, bu using he relaive error afer 120 seconds in he CDA. The resuls, repored in model (2) in Table 2, are similar, albei wih slighly smaller (absolue) coefficiens. We inerpre his as evidence ha he exisence of a preceding call aucion of eiher ransparency level leads o a more efficien opening price han in he case of a sand-alone CDA.

6 6 The Scienific World Journal CM CMT Table 2: Regression analysis marke open. Model (1) Model (2) Model (3) (0.006) (0.005) (0.004) (0.007) (0.006) (0.004) Session dummies, period dummies, and ineracions Yes Yes Yes Cons (0.029) (0.007) (0.009) R Adj. R N P<0.1; P < 0.05; P < Sandard errors clusered a he session level (in parenheses). Session and period dummy ineracion coefficiens omied. The homogeneiy of he call aucions efficiency is also suppored when analyzing he liquidiy of he wo ypes of call aucions and he sand-alone CDA. As in mos empirical (e.g., [26, 38]) and experimenal sudies (e.g., [16, 31]), we usehebid-askspreadasameasureofmarkeliquidiy.we define he spread in he CDA o be he difference beween he lowesaskandhehighesbidinheorderbookaheime when a ransacion akes place. (Remember ha our order book has a deph of 1, since subjecs can only ener orders which improve he quoed spread (i.e., are higher han he previous highes bid and lower han he previous lowes ask) and ha previous bids (asks) are deleed upon enry of a new bes bid (ask). For his reason, in 17 ou of 42 cases no spread exiss a he ime of he opening ransacion in he sandalone coninuous double aucion.) In a call aucion, however, no direc bid-ask spread exiss. We herefore follow Friedman [31] and Theissen [16] in calculaing he relaive inside spread as RIS = PA,l o, P B,h o,, (4) V where P B,h o, (P A,l o, ) is he highes bid (lowes ask) price of a buy (sell) order which is no execued in he call aucion. The difference beween hese wo prices is an indirec spread, because an increase (reducion) of P B,h o, (P A,l o, )ohelevel of P A,l o, (P B,h o, ) would lead o he execuion of he order. An addiional order can only be execued a a price beween P B,h o, and P A,l o,, implying ha RIS represens he price impac of an addiional order. We find ha he average spreads of he opening call aucions are 2.28% for he nonransparen call aucion and 2.14% for he ransparen call aucion (medians: 2.62% and 3.27%, resp.), a difference which is no significan in a Wilcoxon ranksum es (P value: 0.848). (The same resul obains when analyzing he rading volume a he opening (P value: 0.678). Noe however ha rading volume is a relaively poor measure of liquidiy because no dimensions of liquidiy are impounded (e.g., [39]).) The average spread in he CDA, conversely, is 4.52% (median: 4.25%). Applying again he regression mehodology oulined in (3) bu using as he dependen variable he relaive inside spread in he case of he call aucion reamens and he direc spread in he sandalonecda,wefindandreporinmodel(3) of Table 2 highly significan reamen coefficiens. Summingupourresulsfromhemarkeopening,we find suppor for earlier findings by Madhavan [4], Economides and Schwarz [5], and Comeron-Forde e al. [40]. Theyshowhaanopeningcallaucionseemsobebessuied for difficul rading siuaions like he marke opening, when rader uncerainy abou he fundamenal value is high. The aribues of he call aucion wih regard o collecing orders, generaing liquidiy, and incorporaing heerogeneous orders and raders wih differen expecaions and informaion herefore end o lead o a more efficien opening price in such seings. Three findings srenghen he conclusion ha his effec is caused by he mechanism of rading (i.e., call versus coninuous double aucion). Firs, he effec is robus o analyzing prices afer he same amoun of ime has elapsed in he hree reamens. Second, he effec is presen in boh ransparency levels of he call aucion, even hough he ransparen call aucion and he coninuous double aucion provide raders wih a comparable level of informaion which is considerably greaer han ha in he nonransparen call aucion. Third, he resuls hold boh for price efficiency and for he spread. As a second high-level observaion we noe ha we deec no significan differences beween he ransparen and he nonransparen call markes regarding any of he measures invesigaed so far. If here is a difference, our resuls sugges ha i would favor he nonransparen version, which has higher (absolue) coefficiens and lower P values in mos analyses. This resul is surprising. In conras o mos of he lieraure, which finds efficiency increasing in ransparency (e.g., [22, 41 43]), relaively few sudies repor similar resuls regarding he issue of ransparency in call aucions (e.g., [21, 44]) Preopening Phase. In order o analyze our resuls a marke opening in more deail, we invesigae he developmen of prices and he order flow during he preopening phase. Such an analysis has he poenial o shed more ligh on he quesion of how he differen specificaions of call aucion mechanisms perform. We obain he following resul for he preopening phase. Resul 2. In boh forms of he opening call aucion (ransparen and nonransparen), price discovery is compleed wihin he firs minue of order submission. Suppor for Resul 2. Our experimenal framework enables us o compare a ransparen aucion mechanism wih a nonransparen version wih respec o he preopening behavior. Alhough he paricipans do no receive any informaion concerning he order flow and he indicaive prices in he laer,wearenonehelessableocompueheindicaive prices in his aucion mechanism. The amoun and variey

7 The Scienific World Journal MRE (%) Order volume Call-nonransparen Call-ransparen Time unil opening (s) Figure 1: Average MRE of indicaive prices in he call aucions. This figure displays he developmen of he average mean relaive error (MRE) in he ransparen and nonransparen call aucions over ime. of orders and marke deph in boh aucion forms increase wih increasing ime elapsed in he preopening phase. As Figure 1 illusraes, boh call aucion forms have compleed he price discovery process by he end of he firs minue of order submissions. Furhermore, hey achieve a similar level of informaional efficiency, measured by he MRE of he indicaive prices. To furher invesigae his opic, we compue he MRE for he firs indicaive prices and for he indicaive prices inhemiddleofhepreopeningphase(afer60seconds). (In addiion o compuing he MREs of he firs indicaive prices we analyze he ime a which he firs indicaive price becomes available in he call aucions. In he ransparen (nonransparen) call aucion he firs indicaive price is arrived a on average afer 23.5 (24.2) seconds. A wo-sample -es on a period-by-period basis shows ha his difference is no significan (P value: 0.596) and linear regression analysis shows ha he duraion unil he appearance of he firs indicaive price does no influence he MREs of he firs indicaive prices, nor he mid MREs or he opening MREs.) Aferwards we compare hese MRE values o he ones a marke opening. Again, hese hree MRE values are calculaed as averages over all rading periods, which leads o 7 observaions per reamen and ype (firs indicaive price, indicaive price in he middle of he preopening phase, and opening price). Using a Kruskal-Wallis-ANOVA of he MRE values per experimenal session we find significan differences beween hedisribuionsofhehreepriceypesforbohheransparen (P value 0.020) and he nonransparen call aucions (P value: 0.001). The resuls of Wilcoxon rank sum ess show ha he MREs of he firs indicaive prices are significanly greaer han he MREs of he indicaive prices afer 60 seconds in he ransparen (P value: 0.048) as well as he nonransparen call aucions (P value: 0.002). The same is rue for he comparison of he MREs of he firs indicaive prices o he opening price MREs (P values:0.009and0.003), Opening Time unil opening (s) Nonransparen call aucion gross order volume Transparen call aucion gross order volume Nonransparen call aucion ne order volume Transparen call aucion ne order volume Figure 2: Gross and ne average order volume during he preopening phase. This figure displays he developmen of he average order volume in he ransparen and nonransparen call aucions gross and ne of cancellaions over ime. respecively. The main resul of higher price efficiency a he endofhepreopeningphasecomparedohebeginningisin line wih [19], who find a similar resul for a ransparen call aucion preopening mechanism a he Paris Bourse. Figure 1 illusraes graphically ha he price discovery process is less pronounced (or even nonexisen) from he middle of he preopening phase o he opening. More echnically we find ha he MRE of he opening prices is differen from ha of he indicaive prices in he middle of he preopening phase for neiher he ransparen nor he nonransparen call aucions in Wilcoxon rank sum ess (P values: and 0.225). This conflics wih he resul of Biais e al. [19], who repor efficien price discovery especially inhelaerparofhepre-openingphase.onereasonfor his discrepancy may be ha in our experimenal sudy, a price-ime prioriy rule is implemened, possibly leading o a higher order volume already a he beginning of he preopeningphase(seefigure2).theparisboursehowever did no have such a rule during he sample period. This may have led o he higher rading aciviy in he laer par of he preopening phase and he more efficien price discovery process documened by Biais e al. [19]. In accordance wih he resuls from he opening prices, we observe no significan differences beween he firs or midperiod indicaive prices beween he wo reamens of he call aucion (P values: and in Wilcoxon rank sum ess), suggesing again ha a ransparen marke, alhough providing more informaion abou he marke siuaion, does noleadohighermarkequaliycomparedohenonransparen aucion. (As Figure 1 indicaes, he marginally significan difference beween he mid-period MREs suggess ha, on he conrary, he nonransparen marke is more efficien (average MRE: 2.29%) han he ransparen marke Opening

8 8 The Scienific World Journal (average MRE: 3.09%).) A closer look a he preopening order flow allows us o invesigae if and under which circumsances he marke qualiy in he ransparen call aucion differs from ha in he nonransparen seing. Figure 2 displays an overview of he order flow during he preopening phases of boh call aucions. I shows he gross and ne average order volumes separaely, wih ne order volume in his conex defined as he gross order volume minus he volume of cancelled orders. Figure 2 highlighs sligh reamen differences. A he beginning of he preopening phase, he (cumulaive) gross order volume is greaer in he ransparen aucion, bu he nonransparen aucion order volume caches up over ime. The ne order volumes under he wo reamens exhibi a similar developmen as he gross volumes early in he preopening phase (his is o be expeced, since order cancellaions accrue over ime and mus by definiion lag order enries), bu hen inersec in he middle of he preopening phase. Finally, cumulaive ne order volume ends up being significanly higher in he nonransparen aucion a he ime ofheopening.thisshowsha,duringhesecondhalfof he preopening phase, more orders are being cancelled in he ransparen han in he nonransparen call aucion. The higher incidence of order cancellaions in he ransparen aucion may hin a sraegic behavior in he sense ha subjecs submiing orders which are laer cancelled may no really inend o rade a he submied prices bu could be rying o move he marke hrough he revelaion of wha ohers may believe o be informaion (cf. [31]). In line wih Davies [20], we compue he fracion of cancelled orders which would have been execued a he indicaive price. For he second half of he preopening phase we find his o be 3.27% (5.94%) in he nonransparen (ransparen) aucion. (For he firs half of he preopening phase he average fracion of cancelled quoes is 2.39% (3.80%) for he nonransparen (ransparen) call aucion. This difference is no saisically significan in a Wilcoxon rank sum es (boh P values: 0.220).) A Wilcoxon rank sum es confirms ha his difference is saisically significan (P value: 0.022). However, a second inerpreaion of he greaer number of order cancellaions in he ransparen reamen would be ha here are greaer learning effecs, which cause subjecs o rehink and cancel heir orders. Since our daa are insufficien for furher pursuing his issue, we leave i for fuure research Spillover Effecs. In he majoriy of real sock markes, rading afer he opening call aucion coninues in he form of a coninuous double aucion. This raises he quesion wheher an opening call aucion no only leads o a more efficien opening price, bu also influences subsequen coninuous rading in a posiive manner in erms of marke qualiy. Our experimenal research design enables us o analyze hese spillover effecs of differen call aucion formas in a common framework by comparing measures of marke qualiy from a double aucion marke preceded by a call marke wih daa from a sand-alone double aucion marke. We sar our invesigaion by analyzing he price efficiency during he coninuous double aucion. In order o do so, we employ he following adaped RE and MRE measures: RE CDA = n i=1 ( P i, V V MRE CDA = S i, n i=1 S ), (5) i, RE CDA =3. (6) In(5)wefirscalculaeheperiodaverageofherelaive error (RE CDA ) of each ransacion in he coninuous double aucion in period by aking he relaive absolue differences beween he ransacion prices and he fundamenal value and in conras o he call aucion weighing hem by he proporion of he period s oal rading volume represened by he ransacion in quesion (cf. [16]). The weighing facor serves o limi he influence ransacions small in volume bu large in price deviaion have on he overall MRE of a coninuous rading period (MRE CDA ). In a second sep we derive he MRE CDA as he average of he sum of he RE CDA values of all ransacions in a period, over all periods for a series, as presened in (6). Welishemean,median,andsandarddeviaionof each reamen s MRE CDA values in Panel A of Table 3. The noaion we use is as follows: CDA-CM sands for he coninuous double aucion afer he nonransparen call aucion; CDA-CMT is he coninuous double aucion afer he ransparen call aucion; and CDA alone is he sandalone coninuous double aucion wihou an opening aucion. WeuseWilcoxonranksumessocomparehesandalone coninuous double aucion oucomes o daa from he markes preceded by eiher of he wo ypes of call aucions. The resuls are repored in Panel B of Table 3. Despie our smallsamplesizeofonlysevenobservaionseachofhe resuls indicaes ha prices in he coninuous double aucion are more efficien subsequen o a nonransparen opening call aucion han wihou such an opening insiuion. The difference beween he sand-alone CDA and he CDA-CMT is no significan. When using only he MRE calculaed over helasransacionpriceineachsessionofhecda,wefind no significan resuls. Weobainasrongerresulwhenweagainesimaehe regression specificaion from (3), using RE CDA as he dependen variable. The reamen dummy coefficiens for he CDA following boh he nonransparen and he ransparen call markes, repored in model (4) in Table 4, are negaive and significan.thiswouldsupporheassumpionhaopening call aucions reduce uncerainy abou he fundamenal value, hus indicaing ha a good sar of rading goes hand in hand wih an overall increase in marke qualiy. While hese findings appear promising a firs sigh, we believe ha hey acually consiue a resul of learning by he marke paricipans, which leads he spillover effecs o disappear in repeaed markes. To subsaniae his claim, we plo he average value of RE CDA by period in order o give an

9 The Scienific World Journal 9 Table 3: Descripive saisics of period and las rade MRE CDA values as well as resuls of Wilcoxon rank sum ess beween he sand-alone CDA and he CDA of he complemen markes. Panel A: descripive saisics CDA alone CDA-CM CDA-CMT Mean 3.60% 2.68% 2.82% Median 3.22% 2.65% 3.15% Sandard deviaion 1.95% 1.15% 1.15% Mean (las rade) 2.93% 2.27% 2.64% Median (las rade) 2.50% 2.46% 2.65% Panel B: Wilcoxon ess CDA alone versus CDA-CM CDA alone versus CDA-CMT CDA-CM versus CDA-CMT Z P value Z (las rade) P value (las rade) Table 4: Regression analysis spillover effec. Model (4) Model (5) Model (6) Model (7) Model (8) CM (0.003) (0.003) (0.004) (0.005) (0.004) CMT (0.003) (0.003) (0.004) (0.005) (0.005) RE o, (0.057) (0.072) S Call (0.120) Session dummies, period dummies, and ineracions Yes Yes Yes Yes Yes Cons (0.021) (0.014) (0.012) (0.014) (0.010) R Adj. R N P < 0.05; P < Sandard errors clusered a he session level (in parenheses). Coefficiens for session and period dummies and heir ineracions omied. overview of he heerogeneiy and average developmen over ime. We do so in Figure 3, which displays he developmen of RE CDA in each session, in each of he hree reamens, depiced in hin grey lines. The solid black line is he mean over all sessions. Since he firs CDA session exhibis relaively high mispricing in he firs hree periods, we also calculae hemeanwihouhissession.thedashedblacklineinhe lef panel displays he mean wihou his possible oulier, documening ha prices remain relaively high early in he CDA markes. The overall paerns in Figure 3 clearly documen relaively large heerogeneiy beween sessions, bu relaively similar convergence in laer periods. While he CM and CMT reamens seem o follow similarly fla rajecories, he panel for he sand-alone CDA suggess he exisence of a learning effec from he earlier o he laer periods. We herefore es for differences in he RE CDA values beween reamens and provideheresulsintable5.(wealsorepeaheregression analysis excluding periods 1 and 2, finding ha he significan reamen effec likewise disappears.) The resuls indicae ha mispricing is indeed significanly higher in he early periods of he CDA reamen (and ha his effec is relaively robus o he removal of one possible oulier, despie he sample size of only 7 sessions each). Noneheless, our findings also show ha he posiive spillover effec from he opening call aucions disappears afer he second period of he marke. The reason is no ha prices in he CDA following a call aucion become less efficien bu raher ha prices in he sand-alone CDA become more efficien. Given ha we run wo raining periods, we inerpre his as evidencehaourmarkeparicipansrequiresubsanial experience o largely eliminae mispricing in his seing. When hey gain his experience, however, he difference beween our reamens disappears. In oher words, our spillover effec findings seem o be driven by an ineracion wih subjec experience, wih a call aucion amelioraing he pricing errors in markes wih inexperienced subjecs.

10 10 The Scienific World Journal 10 CDA CM CMT RE (%) 6 4 RE (%) 6 4 RE (%) Period Period Period Figure 3: Average RE CDA in he CDA phase, by reamen. This figure displays he developmen of he average relaive error in each session s CDA phase (RE CDA ), in each of he hree reamens (hin grey lines), he mean over all sessions (solid black line), and he mean over all sessions wihou he firs session in he case of he sand-alone CDA (dashed black line, lefmos panel). Table 5: Average RE CDA values and resuls of corresponding Wilcoxon rank sum ess. Resuls wihou he firs CDA session in parenheses. Panel A: descripive saisics CDA alone CDA-CM CDA-CMT (1) Avg. RE CDA, Period % (4.29%) 2.59% 2.51% (2) Avg. RE CDA, Period % (4.21%) 3.24% 2.99% (3) Avg. RE CDA, Period % (2.66%) 2.84% 2.70% (4) Avg. RE CDA, Period % (2.83%) 1.98% 3.06% (5) Avg. RE CDA, Period % (2.86%) 2.40% 2.90% (6) Avg. RE CDA, Period % (3.04%) 3.05% 3.12% (7) Avg. RE CDA, las ransacion, and all periods 2.93% (2.67%) 2.27% 2.64% (8) Avg. RE CDA, las ransacion, and Period % (2.69%) 2.92% 2.39% Panel B: Wilcoxon ess CDA alone versus CDA-CM CDA alone versus CDA-CMT CDA-CM versus CDA-CMT P value (1) (0.063) (0.046) P value (2) (0.087) (0.153) P value (3) (1.000) (0.475) P value (4) (0.153) (0.775) P value (5) (0.317) (0.886) P value (6) (1.000) (0.886) P value (7) (0.386) (0.931) P value (8) (0.475) (0.886) We find no such ouliers or clear ime rends when analyzing marke liquidiy. In line wih he oucomes for MRE CDA, we perform a comparison of he spreads (In conras o he call aucion, spreads are direcly observable in he coninuous double aucion by compuing he difference beween ask and bid prices when a ransacion occurs. They are paricularly relevan in his seing, since ransacions end o ake place when spreads are lower and because raders are affeced by he spreads only when a ransacion occurs (cf. [16]). Noe ha no spread can be derived when somebody acceps he sanding bid (ask) a a ime when here is no ask (bid) in he marke.) beween he coninuous double aucion alone and he coninuous double aucions of he complemen markes. Descripive saisics as well as resuls of he Wilcoxon rank sum ess are shown in Table 6 (Panels A and B, resp.). Table 6 documens ha he spreads in he coninuous double aucion afer a nonransparen opening call aucion are significanly lower han in he sand-alone coninuous double aucion. We again esimae he regression specificaion from (3), using he average spread as he dependen variable. The resuls are repored in model (6) in Table 4. Our resuls show he reamen dummy coefficien for he CDA following he nonransparen call marke o be negaive and significan, while he coefficien for he CDA afer a ransparen call marke is no significan. This suppors he assumpion ha a nonransparen opening call aucion

11 The Scienific World Journal 11 Table 6: Descripive saisics of period spreads and resuls of Wilcoxon rank sum ess on average series spreads beween he sand-alone CDA and he CDA of he complemen markes. Panel A: descripive saisics CDA alone CDA-CM CDA-CMT Mean 4.53% 2.78% 3.91% Median 4.25% 2.62% 3.27% Sandard deviaion 2.21% 1.66% 2.43% Panel B: Wilcoxon es CDA alone versus CDA-CM CDA alone versus CDA-CMT CDA-CM versus CDA-CMT Z P value reduces uncerainy abou he fundamenal value, leading o higher liquidiy in coninuous rading afer such a call aucion. Noe ha some raders in he coninuous double aucion afer a ransparen opening call aucion ener order prices which are paricularly far away from oher raders bids or asks and from he fundamenal value. Alhough a majoriy of hese orders is no acceped, hey may lead o high repored spreads when an order of he oher marke side is acceped (since he spread is defined as he difference beween he sanding bid and ask a he ime of a ransacion). This phenomenon becomes apparen when analyzing he inraday volailiy in he coninuous double aucion. The inraday price sandard deviaion was compued for every period and session. The average is 2.26% (2.47%) [3.12%] in he CDA-CM (CDA-CMT) [sand-alone CDA]. The hypohesis of he exisence of posiive spillover effecs on price volailiy is no suppored by he Wilcoxon rank sum es resuls for eiher he nonransparen opening call aucion (P value: 0.110) or he ransparen version (P value: 0.277). We also fail o deec a significan difference beween he coninuous aucions following he wo call aucion ypes (P value: 0.848). However, employing our more powerful regression mehodology, we can isolae negaive and significan reamen effecs from he nonransparen (coefficien: , P value: 0.002) and he ransparen call aucions (coefficien: , P value: 0.019). Ineresingly, we find small bu significan downward rends when regressing volailiy on ime in CDA-CM and CDA, bu no in CDA- CMT (coefficiens: , , and ; P values: 0.020, 0.002, and 0.281, resp.). Finally, we again find no significan differences in oal rading volume beween he hree coninuous double aucions using he Wilcoxon rank sum es (P values: 0.277, 0.142, and when comparing he sand-alone CDA and he CDA-CM, when comparing he sand-alone CDA and he CDA-CMT, and when comparing he CDA-CM and he CDA-CMT, resp.). Once more, using he regression mehodology, we obain posiive and significan reamen effecs from he nonransparen (coefficien: , P value: 0.007) and he ransparen call aucions (coefficien: , P value: 0.004). To furher invesigae he quesion of spillover effecs, we perform a linear regression analyzing he impac of marke efficiency a he marke opening afer a call aucion on subsequen efficiency using he following funcional form, wih he symbols defined as in (3), and again adjusing for clusers in he sandard errors by session: RE CDA,θ =α+β RE RE θ o, +β CM CM +β CMT CMT + β T + βt s s + βt s s +ε. In his regression, θ {CM, CMT} indicaes he nonransparen and ransparen call marke reamen, respecively. This regression invesigaes he associaion beween he opening price efficiencies in he call aucions and he average price efficiencies over he course of he following coninuous double aucions and wheher here is a reamen difference in he spillover effec beween he ransparen and he nonransparen versions of he opening call aucion. As shown in model (5) in Table 4, we find β RE o be highly significan, while he coefficien dummies of he nonransparen (ransparen) call marke reamens urn ou o carry he correc sign bu are no significanly differen from zero. (This resul is robus o excluding he firs wo periods.) In oher words, lower (higher) efficiencies a he opening of he ransparen aucion marke are associaed wih lower (higher) efficiency in he subsequen coninuous double aucion, and his effec subsumes any reamen effecs ha may be presen. When we exend his line of analysis o spreads, we find a significan reamen difference bu no spillover effec. (When regressing he spreads in he coninuous double aucion on he spreads in he preceding call marke in a specificaion analogous o (7), we find he resuls repored for model (7) in Table 4. Regressinghespreadinheconinuousdoubleaucionon he opening MRE from he call aucions yields he findings lised in model (8). Asishecaseformodel(5), RE o, also has significan explanaory power for he CDA spread resuls, while he spread in he call aucion preceding he CDA does no. The dummy for he nonransparen complemen marke reamen is highly significanly differen from zero in boh models.) In conras o some of he empirical lieraure (e.g., [25, 26]) our experimenal resuls provide no suppor for he conjecure ha a higher level of ransparency in he call aucionhasaposiiveinfluenceonhemarkequaliyinhe subsequen CDA. On he conrary, hey seem o be more in line wih he argumens of Domowiz and Madhavan [6], who show ha excessive ransparency can have a negaive influence on marke qualiy. I is possible ha observing (7)

12 12 The Scienific World Journal he order flow and he price discovery process during he preopening phase increases he feeling of uncerainy in he marke, which spills over ino subsequen coninuous rading. Alernaively i is possible ha some paricipans areoverwhelmedbyheamounofinformaionrevealed in he preopening phase, which would be in line wih he concep of informaion overload (cf. [45]; see also [46] for complexiy in financial markes and informaion overload). Finally, we acknowledge he possibiliy ha he lack of a beer performance of he ransparen call aucion is due o he greaer room for price manipulaion herein, as documened by Arifovic and Ledyard [30]. Unforunaely, our daa do no allowusodisinguishbeweenheserivalhypoheses. 4. Conclusions The paper deals wih he marke srucure a he opening and is influence on subsequen rading. In conras o previous papers, we compare a single coninuous double aucion and wo complemen markes wih differen call aucion designs as opening mechanisms in a unified framework. This is also he firs experimenal examinaion of complemen markes using differen designs of an opening call aucion. The resuls of our experimenal approach can be summarized as follows. (1)Acallaucioniswell-suiedasamarkeopening insiuion, generaing a higher marke qualiy a his ime of marke sress han opening direcly wih coninuous rading. (2) Opening call aucions cause a posiive spillover effec on subsequen coninuous rading in erms of spreads, volailiy, and rading volume, alhough no necessarily on mispricing. (3) Higher ransparency during he preopening phase leads o higher marke qualiy neiher a he opening nor in subsequen rading. While here are no significan differences in erms of marke qualiy a he opening, spreads in he coninuous marke are significanly higher han in he CDA-CM seing wih a nonransparen opening call aucion. We consider his as evidence supporing he decision of mos sock exchanges o insiue an opening call aucion. As prediced by he heoreical lieraure (e.g., [4, 5]), such an insiuion seems a suiable insrumen a his crucial ime of he rading day. Furhermore, he opening call aucion has a posiive spillover effec on subsequen coninuous rading, hereby supporing mos of he empirical lieraure (e.g., [23, 24, 47]). This resul also means ha an opening call aucion creaes higher price efficiency and liquidiy for he whole markeandshouldhereforeleadoamorearacivesock marke for issuers and invesors, which in urn implies an overall welfare gain for he economy (e.g., [48]). However, as Comeron-Forde e al. [13] poined ou in heir examinaion of he Hong Kong Sock Exchange, he design of call aucion markes is a complex issue. I encompasses he choice of a level of ransparency during he preopening phase as well as price deerminaion and prioriy rules. As he presen experimenal examinaion shows, he design of he call aucion can influence he behavior of raders, wih aendan implicaions for marke qualiy. In periods wih inefficien prices during he firs seconds of he preopening phase, raders seem o ry o avoid prices geing more efficien by placing orders ha suppor he curren inefficien prices or by no placing orders, so ha no efficien price discovery process akes place. In hese periods, he efficiency of he opening price is lower han in he nonransparen aucion. Regression analysis furhermore shows ha his lower efficiency is no correced in he subsequen coninuous double aucion. In fac, he higher he MRE value a he opening of he ransparen aucion compared o he nonransparen aucion, he higher is he difference beween he MRE values in he coninuous double aucion. Furher experimenal analysis concerning he ransparency of he opening call aucions would herefore be valuable. Noe ha mos sock exchanges ouside he lab use lower levels of ransparency han ha chosen in our ransparen call aucion reamen. I would herefore be worhwhile o learn wheher he higher efficiency in our nonransparen call aucion markes is a corner soluion, or wheher here is an opimal level somewhere in beween no and full ransparency. Such a U-shaped paern is, for example, suggesed by Comeron-Forde and Rydge [41]. A sudy invesigaing his quesion could, for example, only provide raders wih indicaive prices (his is, e.g., he case a he Deusche Börse) or show a limied number of bids (his is he case, e.g., a Euronex Paris and a he New York Sock Exchange). Furhermore, manipulaive behavior by some raders, as seems o be presen in a few periods of our sudy, could be analyzed by furher experimens wih low price signals for one par of raders and high price signals for he second par of raders. Therefore, i could be analyzed wheher manipulaing behavior in a ransparen aucion can generally be observed. This would be paricularly ineresing in a seing wih heerogeneously informed raders. As a final area of possible fuure research we idenify he implemenaion of a marke making seing in he coninuous rading phase (like, e.g., a NASDAQ, cf. he empirical work of [24]). In some empirical and experimenal sudies (e.g., [16, 49]), marke making wihou an opening call aucion exhibis lower marke qualiy han he coninuous double aucion. An opening call aucion could herefore possibly have an even greaer spillover effec in his kind of insiuion. Conflic of Ineress The auhors declare ha here is no conflic of ineress regarding he publicaion of his paper. Acknowledgmens The auhors hank he Faculy of Social and Economic Sciences a he Karl-Franzens Universiy Graz for access o he Max Jung laboraory and subjec pool and for financial suppor. The auhors are graeful o Jan Pieer Krahnen and Eric Theissen for valuable discussions, and seminar audiences a he Universiy of Economics in Izmir, Turkey, in June 2010,

13 The Scienific World Journal 13 and paricipans a he EURO Working Group Workshop on Experimenal Economics in Graz in November 2009 and a he Inernaional OR Conference in Zurich in Sepember 2011 for helpful remarks. References [1] R. Wood, T. McInish, and J. K. Ord, An invesigaion of ransacion daa for NYSE socks, JournalofFinance, vol. 40, no. 3, pp , [2]H.R.SollandR.E.Whaley, Sockmarkesrucureand volailiy, Review of Financial Sudies, vol. 3, no. 1, pp , [3] M.J.Barclay,T.Hendersho,andC.M.Jones, Orderconsolidaion, price efficiency, and exreme liquidiy shocks, Journal of Financial and Quaniaive Analysis,vol.43,no.1,pp , [4] A. Madhavan, Trading mechanisms in securiies markes, Finance,vol.47,no.2,pp ,1992. [5] N. Economides and R. A. Schwarz, Elecronic call marke rading, JournalofPorfolioManagemen, vol. 21, no. 3, pp , [6] I. Domowiz and A. Madhavan, Open sesame: alernaive opening algorihms in securiies markes, in The Elecronic Call Aucion. Marke Mechanism and Trading: Building a Beer Sock Marke, R. A. Schwarz, N. Beiner, and M. J. Humbach, Eds., Kluwer Academic Publishers, Boson, Mass, USA, [7]M.S.PaganoandR.A.Schwarz, Aclosingcall simpacon marke qualiy a Euronex Paris, The Financial Economics, vol. 68, no. 3, pp , [8] H.R.Soll, Markemicrosrucure, inhandbook of he Economics of Finance, G. M. Consaninides, M. Harris, and R. M. Sulz, Eds., chaper 9, pp , [9] M. Pagano and A. Röell, Aucion and dealership markes. Wha is he difference? European Economic Review, vol. 36, no. 2-3, pp ,1992. [10] A. Snell and I. Tonks, A heoreical analysis of insiuional invesors rading coss in aucion and dealer markes, Economic Journal,vol.113,no.489,pp ,2003. [11] M. Aiken, C. Comeron-Forde, and A. Frino, Closing call aucions and liquidiy, Accouning & Finance,vol.45,no.4,pp , [12] A. Ellul, H. S. Shin, and I. Tonks, Opening and closing he marke: evidence from he London Sock Exchange, Financial and Quaniaive Analysis,vol.40,no.4,pp , [13] C. Comeron-Forde, J. Rydge, and H. Burridge, No all call aucions are creaed equal: evidence from Hong Kong, Review of Quaniaive Finance and Accouning,vol.29,no.4,pp , [14] C. N. Noussair and S. Tucker, Experimenal research on asse pricing, Economic Surveys, vol. 27, no. 3, pp , [15] S. Palan, A Review of bubbles and crashes in experimenal asse markes, Economic Surveys, vol. 27, no. 3, pp , [16] E. Theissen, Marke srucure, informaional efficiency and liquidiy: an experimenal comparison of aucion and dealer markes, Financial Markes,vol.3,no.4,pp , [17] A. Madhavan, Securiy prices and marke ransparency, Journal of Financial Inermediaion,vol.5,no.3,pp ,1996. [18] M. Pagano and A. Röell, Transparency and liquidiy: a comparison of aucion and dealer markes wih informed rading, Finance, vol. 51, no. 2, pp , [19] B. Biais, P. Million, and C. Spa, Price discovery and learning during he preopening period in he Paris Bourse, Poliical Economy,vol.107,no.6,pp ,1999. [20] R. J. Davies, The Torono Sock Exchange preopening session, Financial Markes,vol.6,no.4,pp ,2003. [21] A. Madhavan, D. Porer, and D. Weaver, Should securiies markes be ransparen? JournalofFinancialMarkes, vol. 8, no.3,pp ,2005. [22] K. S. Eom, J. Ok, and J.-H. Park, Pre-rade ransparency and marke qualiy, Financial Markes, vol.10,no.4,pp , [23] R.P.Chang,S.G.Rhee,G.R.Sone,andN.Tang, Howdoes he call marke mehod affec price efficiency? Evidence from he Singapore Sock Marke, Banking and Finance, vol. 32, no. 10, pp , [24] M.S.Pagano,L.Peng,andR.A.Schwarz, Markesrucure and price formaion a marke openings and closings, working paper, Evidence from Nasdaq s Calls, [25] D.Gerace,G.G.Tian,andW.Zheng,Call Aucion Transparency and Marke Liquidiy, The Shanghai Experience, [26] P.HoffmannandJ.vanBommel,Pre-Trade Transparency in Call Aucions,SSRNeLibrary,2009. [27] U. Fischbacher, Z-Tree: zurich oolbox for ready-made economic experimens, Experimenal Economics,vol.10,no.2,pp , [28] B. Greiner, An online recruimen sysem for economic experimens, in Forschung und Wissenschafliches Rechnen,K.Kremer and V. Macho, Eds., GWDG Berich 63, Gesellschaf für Wissenschafliche Daenverarbeiung, Göingen, Germany, [29] B. Biais and I. Marinez, Price discovery across he Rhine, Review of Finance,vol.8,no.1,pp.49 74,2004. [30] J. Arifovic and J. Ledyard, Call marke book informaion and efficiency, Economic Dynamics and Conrol, vol. 31, no. 6, pp , [31] D. Friedman, How rading insiuions affec financial marke performance: some laboraory evidence, Economic Inquiry,vol. 31, pp , [32] K. C. Brown, W. V. Harlow, and L. T. Sarks, Of ournamens and empaions: an analysis of managerial incenives in he muual fund indusry, The Finance, vol. 51, no. 1, pp , [33] S. Basak, A. Pavlova, and A. Shapiro, Opimal asse allocaion and risk shifing in money managemen, Review of Financial Sudies,vol.20,no.5,pp ,2007. [34] J. Chevalier and G. Ellison, Risk aking by muual funds as a response o incenives, Poliical Economy, vol. 105, no. 6, pp , [35] D. James and R. M. Isaac, Asse markes: how hey are affeced by ournamen incenives for individuals, American Economic Review,vol.90,no.4,pp ,2000. [36] E. J. Elon, M. J. Gruber, and C. R. Blake, Incenive fees and muual funds, Finance, vol. 58, no. 2, pp , [37] D. Acker and N. W. Duck, A ournamen model of fund managemen, Business Finance and Accouning, vol. 33, no. 9-10, pp , 2006.

14 14 The Scienific World Journal [38] C.-H. Kehr, J. P. Krahnen, and E. Theissen, The anaomy of a call marke, Financial Inermediaion,vol.10,no.3-4, pp ,2001. [39] K. H. Baker, Trading locaion and liquidiy. An analysis of U.S. dealer und agency markes for common socks, Financial Markes, Insiuions and Insrumens,vol.5,pp.1 46, [40] C. Comeron-Forde, S. T. Lau, and T. McInish, Opening and closing behavior following he inroducion of call aucions in Singapore, Pacific-Basin Finance Journal, vol. 15, no. 1, pp , [41] C. Comeron-Forde and J. Rydge, The influence of call aucion algorihm rules on marke efficiency, Financial Markes,vol.9,no.2,pp ,2006. [42] K. H. Chung and C. Chuwonganan, Transparency and marke qualiy: evidence from SuperMonage, Financial Inermediaion, vol. 18, no. 1, pp , [43] S.-F. Hsieh, H.-K. Chen, and T. Ma, Marke ransparency, invesor sraegies, and rading coss: evidence from he Taiwan sock exchange, Accouning and Finance Research, vol.1,no.1, pp , [44] T. Ma, Y. Lin, and H.-K. Chen, Are invesors more aggressive in ransparen markes? Asia-Pacific Financial Sudies, vol.37,no.2,pp ,2008. [45] A. Toffler, Fuure Shock, Random House, 1s ediion, [46] M. K. Brunnermeier and M. Oehmke, Complexiy in Financial Markes,2009. [47] C. Comeron-Forde, Do rading rules impac on marke efficiency? A comparison of opening procedures on he Ausralian and Jakara Sock Exchanges, Pacific Basin Finance Journal,vol. 7, no. 5, pp , [48] Y. Amihud and H. Mendelson, Asse pricing and he bid-ask spread, Financial Economics, vol. 17, no. 2, pp , [49] J. Bacidore and M. L. Lipson, The effecs of opening and closing procedures on he NYSE and Nasdaq, in Proceedings of he New Orleans Meeings (AFA 01),January2001.

15 Advances in Operaions Research Advances in Decision Sciences Applied Mahemaics Algebra Probabiliy and Saisics The Scienific World Journal Inernaional Differenial Equaions Submi your manuscrips a Inernaional Advances in Combinaorics Mahemaical Physics Complex Analysis Inernaional Mahemaics and Mahemaical Sciences Mahemaical Problems in Engineering Mahemaics Discree Mahemaics Discree Dynamics in Naure and Sociey Funcion Spaces Absrac and Applied Analysis Inernaional Sochasic Analysis Opimizaion