Information Asymmetry and Liquidity Risk

Transcription

1 Inernaional Review of Business Research Papers Vol. 8. No.1. January Pp Informaion Asymmery and diy Risk Yi-Mien Lin *, Shwu-Jen You ** and Min-Shen Huang *** This sudy firs examines he deerminans of informaion asymmery by considering boh he firm-specific variables and he marke-wide facors. We also invesigae if he firs and second momens of liquidiy, a he individual as well as he aggregae level, provide explanaory power for expeced reurns. The findings are ha informaion asymmery relaes negaively o rading volume and he number of ransacions and posiively o a firm s liquidiy risk. Marke liquidiy and marke liquidiy risk, however, have no significan associaion wih informaion asymmery. Afer adusing for he risk facors of marke, size, book-o-marke raio, and momenum, individual firm s liquidiy and marke liquidiy are posiively relaed o excess sock reurns, whereas an individual firm s liquidiy risk relaes negaively o excess reurns. Marke liquidiy risk does no have a significan effec on excess sock reurns. JEL Codes: G30, G10 and M40 1. Inroducion This sudy has a dual focus. Firs, he sudy provides an inegraive analysis of he deerminans of informaion asymmery by considering boh he firm-specific variables and he marke-wide facors. Second, he sudy invesigaes he effecs of liquidiy and liquidiy risk, a he individual as well as he aggregae level, on expeced sock reurns. Informaion asymmery has been he issue of an imporan and ineresing among academics. Richardson (2000) suggess a sysemaic relaionship beween he magniudes of informaion asymmery as measured by bid-ask spread and he level of earnings managemen. A high level of informaion asymmery will induce managers o manipulae earnings. Easley and O Hara (2004) analyze he role of informaion asymmery among invesors in he deerminaion of cos of capial. i Lamber, Leuz and * Dr. Yi-Mien Lin, Naional Chung Hsing Universiy, Taiwan. ymlin@dragon.nchu.edu.w ** Dr. Shwu-Jen You, Naional Taichung Universiy of Science and Technology, Taiwan. syou@ni.edu.w *** Specialis, Min-Shen Huang, Wisron Corporaion, Taiwan. p @gmail.com

2 Verrecchia (2009) argue ha, in perfec compeiion, he precision of informaion is he key deerminan of informaion risk which affecs cos of capial. They use informaion precision as he average qualiy of informaion ha invesors have on he expeced cash flows of he firm, and informaion asymmery as he difference in precision across invesors. Moreover, in case of imperfec compeiion, informaion asymmery will affec he cos of capial by adversely affecing he willingness of liquidiy providers o supply liquidiy. To measure informaion asymmery, Akins, Ng and Verdi (2009) use he informaion asymmery componen of bid-ask spreads and he probabiliy of informed rading. ii Informaion asymmery may influence he sock rading of a firm. diy denoes he speed and ease a which one can sock rade large quaniies. Among he numerous measures of liquidiy, bid-ask spread is he mos popular, and someimes he only measure used by researchers. iii The asymmeric-informaion models formulaed by Glosen and Milgrom (1985) and Kyle (1985) focus on he asymmeric informaion faced by he marke makers. iv Marke makers rade wih boh informed and uninformed raders and, on average, lose on rades wih he former and profi on rades wih he laer. The aforemenioned models focus on rading of individual socks and provide lile guidance abou sysemaic variaions in liquidiy. The recen sudies of Chordia, Roll and Subrahmanyam (2000), Huberman and Halka (2001), and Brockman and Chung (2002) shif emphasis oward analyzing he common deerminans of liquidiy. The inuiive appeal ha he individual firm s liquidiy is a leas parly deermined by marke wide facors comes from our knowledge ha sysemaic facors affec oher firm-specific aribues (e.g., risk and reurn). Therefore, in examining he deerminans of bid-ask spreads, his sudy considers firm-specific variables as well as marke-wide facors. Specifically, his sudy analyzes he effecs of rading volume, number of rades, individual firm s liquidiy risk, marke liquidiy and marke liquidiy risk on he magniude of spreads. Accordingly, he bid-ask spread is posiively associaed wih he degree of informaion asymmery, and herefore, his paper uses he bid ask spread as a proxy of he informaion asymmery. The relaed sudies abou he relaionship beween he level of spread and liquidiy, Soll (1978) find boh sysemaic and unsysemaic reurn risks have a significan associaion wih bid-ask spread. Since an increase in liquidiy risk, a he individual as well as he aggregae level, would lead o larger uncerainy in a dealer s opimal invenory level, his sudy conecures a significan associaion beween bid-ask spread and liquidiy risk. Hence, we also analyzes if liquidiy and liquidiy risk have explanaory power for expeced sock reurns. The concep ha liquidiy can influence asse reurns was firs 113

3 proposed by Amihud and Mendelson (1986), who argue ha invesors require a higher reurn o compensae for he higher rading cos associaed wih a less liquid asse. Following Amihud and Mendelson, numerous sudies have invesigaed he effecs of liquidiy on expeced reurns and generally find ha less liquid socks have higher expeced reurns. v In conras o he numerous sudies invesigaing he associaion beween liquidiy and expeced reurns, relaively few sudies have examined he effecs of liquidiy risk on expeced reurns. Given ha invesors are risk averse and dislike variabiliy in liquidiy, i is plausible for he second momen of liquidiy o be priced. In examining he effecs of liquidiy and liquidiy risk on sock reurns, his sudy follows Pasor and Sambaugh (2003) and Chordia, Subrahmanyam and Anshuman (2001) and aduss reurns for he risk facors of marke, book-o marke raio and size proposed by Fama and French (1993) and he momenum risk facor evidenced in Jegadeesh and Timan (1993). This sudy examines if individual as well as aggregae liquidiy and liquidiy risk are priced in addiion o he four risk facors. The empirical analysis finds ha bid-ask spread relaes negaively o rading volume and he number of ransacions and posiively o a firm s liquidiy risk, which is consisen wih our expecaion. Marke liquidiy and marke liquidiy risk, however, have no significan associaion wih spread. The evidence implies ha in seing he bid-ask spreads, he marke makers only consider firm-specific variables and disregard he marke-wide facors. As for he associaion beween liquidiy risk and excess reurns, he evidence indicaes ha, afer adusing for he well-sudied effecs of marke, size, book-o-marke raio and momenum, individual firm s liquidiy and marke liquidiy relae posiively o excess reurns, whereas individual firm s liquidiy risk relaes negaively o excess reurns. Marke liquidiy risk does no have a significan effec on excess reurns. Alhough he direcion of he effecs of liquidiy and liquidiy risk on excess reurns is conrary o our expecaion, he resuls pinpoin he imporance of hese facors in asse pricing. This sudy makes a leas wo conribuions o he lieraure. Firs, while prior sudies examining he deerminans of liquidiy focus on firm-specific variables, his sudy provides an inegraive analysis of he deerminans of bid-ask spread by considering boh he firm-specific variables and he marke-wide facors. Secondly, given ha prior research has documened evidence of he firs momen of an individual firm s liquidiy being a significan facor in asse pricing, his sudy furher invesigaes if he second momen of liquidiy, a he individual as well as he aggregae level, relaes o expeced reurns. 114

4 This paper is organized as follows. Secion 2 develops hypoheses and Secion 3 describes empirical mehods, daa, and variable measuremen. Secion 4 shows he empirical resuls and Secion 5 concludes. 2. Lieraure Review and Hypohesis A number of researchers using rading cos models have repored an inverse relaionship beween rading volume and spreads. Tinic and Wes (1972) argue ha when rading volume is high, he dispariies and disconinuiies in he inflow of buy and sell orders decline, hus reducing he need for he dealer s invenory posiioning. Inacive socks, on he conrary, require raher exensive dealer paricipaion. They invesigae he daily spreads of NASDAQ socks and documen smaller spreads in socks wih larger rading volume. Soll (1978) also uses daa from NASDAQ socks o examine he deerminans of he bid-ask spreads. Consisen wih his argumens ha rading volume deermines how long he dealer is exposed o price risk, he find a significan and negaive associaion beween rading volume and bid-ask spreads. Mcinish and Wood (1992) argue ha greaer rading aciviy can lead o lower spreads due o economies of scale in rading coss. They analyze he inraday paerns in spreads for NYSE socks o es a negaive relaionship beween spreads and rading aciviy.. Using boh he number of ransacions per inerval and he number of shares per rade o represen rading aciviy, hey find hese wo variables relae negaively o spreads. Researchers examining rading coss find ha a dealer s risk of holding a securiy is a significan deerminan of he bid-ask spread. Exisen sudies ypically use measures of oal risk and sysemaic risk. For example, Tinic and Wes (1972) and Hamilon (1978) find spreads are posiively associaed wih oal risk; Soll (1978) find boh sysemaic and unsysemaic risks have a significan associaion wih bid-ask spread. As for he relaionship beween individual liquidiy risk and he level of spread, relaively few sudies have examined i. Since an increase in an individual firm s liquidiy risk would lead o larger uncerainy in a dealer s opimal invenory level, which in urn would resul in higher invenory holding coss, his sudy expecs a posiive correlaion beween individual liquidiy risk and bid-ask spread. The popular press has long menioned sysemaic liquidiy and some researchers analyze he common underlying deerminans of liquidiy. Chordia e al. (2000) esimae a marke model for liquidiy by regressing daily percenage changes in liquidiy variables for an individual sock on marke measures of liquidiy. They find ha he quoed spread and effecive spreads co-move wih marke-wide liquidiy. The common 115

5 influences on variaions in liquidiy remain significan afer conrolling for he well-known individual liquidiy deerminans such as volailiy, volume and price. Using differen saisical echniques, Huberman and Halka (2001) examine he auoregressive srucures of he ime series of various liquidiy proxies. They find a posiive correlaion beween he innovaions of he ime series for each liquidiy proxy, indicaing he presence of a common liquidiy facor. Brockman and Chung (2002) examine commonaliy of liquidiy using inraday daa from he Sock Exchange of Hong Kong, which is an acive order-driven marke. They claim ha order-driven sysems are more suscepible o commonaliy because here is no obligaion on he par of any marke paricipan o mainain a fair and orderly marke. They show ha individual firm liquidiy is significanly influenced by a commonaliy componen. The co-movemen in individual securiy liquidiy leads o anoher quesion: wha is he relaionship beween aggregae liquidiy risk and bid-ask spread? Following he argumen of invenory holding cos, his sudy expecs a posiive associaion beween aggregae liquidiy risk and bid-ask spread since larger aggregae liquidiy risk would resul ino more uncerainy in a dealer s holdings of any individual securiy. Given hese consideraions, we formulae he following hypohesis: Hypohesis 1: Informaion asymmery negaively o rading volume, he number of ransacions and marke liquidiy bu posiively o an individual firm s liquidiy risk as well as marke liquidiy risk. Models of price formaion in securiies markes sugges ha privaely informed invesors creae significan illiquidiy cos for uninformed invesors, implying ha he required rae of reurn should be higher for securiies ha are relaively illiquid. Using a variey of liquidiy measures, empirical sudies generally find a negaive relaion beween liquidiy and required raes of reurn. Amihud and Mendelson (1986) was he firs o empirically es he relaion beween bid-ask spread and expeced reurn. They use he capial asse pricing model o adus reurns for risk and find evidence ha asse reurns include a significan premium for he quoed spread. Brennan and Subrahmanyam (1996) invesigae he relaionship beween monhly sock reurns and measures of illiquidiy obained from inraday daa. They perform generalized leas squares regression and find a significan relaionship beween required raes of reurn and heir measures of illiquidiy afer incorporaing he hree-facor model developed by Fama and French (1993), and also afer accouning for he effecs of he sock price level. Brennan e al. (1998) proxy a firm s liquidiy wih rading volume and find a significan and negaive relaionship beween reurns and rading volume afer eiher adusing for he Connor and Koraczyk 116

6 (1988) risk facors or he Fama and French (1993) risk facors. Following he resuls of hese sudies and he recen finding of a posiive associaion beween individual sock s liquidiy and marke liquidiy, his sudy expecs a negaive associaion beween risk-adused excess sock reurns and he individual firm s liquidiy as well as marke liquidiy. Given he evidence ha he level of liquidiy affecs asse reurns, i is reasonable o conecure ha he second momen of liquidiy should be posiively associaed wih asse reurns. As long as invesors are risk averse and dislike variabiliy in liquidiy, sock wih greaer liquidiy risk should command higher expeced reurns. In invesigaing he effec of liquidiy risk on sock reurns, some sudies use oal liquidiy risk and some use sysemaic liquidiy risk. For example, Chordia e al. (2001) use rading volume and share urnover as proxies for liquidiy and measure a sock s liquidiy risk wih he sandard deviaions of hese measures. Conrary o heir expecaion, hey documen a negaive and srong cross-secional relaionship beween sock reurns and liquidiy risk, afer conrolling for size, book-o-marke raio, momenum, and he level of liquidiy. Eckbo and Norli (2005) consruc a low-minus-high sock urnover porfolio as a liquidiy risk facor and find ha liquidiy is significanly priced. Marshall (2004) examine he deerminans of he level of IPO underpricing and documen evidence ha firms wih greaer liquidiy concerns a he IPO experience greaer underpricing. On he oher hand, Pasor and Sambaugh (2003) examine he relaionship beween sock reurns and sysemaic liquidiy risk and find ha a sock s liquidiy bea, measured by is sensiiviy o innovaions in aggregae liquidiy, plays a significan role in asse pricing. Socks wih higher liquidiy beas command higher expeced reurns afer adusing for he hree risk facors of Fama and French (1993) and a momenum facor. Marinez, Nieo, Rubio and Tapia (2005) use hree sysemaic liquidiy risk measures and find only one measure of sysemaic liquidiy risk is priced in he Spanish marke. Gibson and Mougeo (2004) use a bivariae GARCH (1,1)-in-mean specificaion for he marke porfolio excess reurn and find ha aggregae marke liquidiy risk is priced. Hence, he following hypohesis is formulaed: Hypohesis 2: Risk-adused excess reurn relaes negaively o individual firm s liquidiy as well as marke liquidiy, bu posiively o boh individual firm s liquidiy risk and marke liquidiy risk. 117

7 3. The Mehodology and Model 3.1 Daa This sudy rerieves daa for NYSE socks from he daabase of COMPUSTAT for he period of To be included in he sample, his sudy requires ha a sock be coninually lised on he NYSE hroughou he research period and have all he daa required for esimaing he variables involved. We exclude firms in financial disress and newly lised firms during he research period. We also exclude financial and insurance firms due o heir characerisics of operaions and financial srucures, which are differen from firms in oher indusries. Observaions wih missing informaion are also excluded. The final sample consiss of 73,950 firm-monh observaions for 1900 firms in en indusries. Table 1 displays he indusry classificaion of he sample based on an SIC code. Indusry Table 1 Indusry Classificaion of Sample Firms No. of Firms Agriculure, Foresry, and Fishing 10 Mining 136 Consrucion 38 Manufacuring 879 Transporaion, Communicaions, Elecric, Gas, and Saniary Services 393 Wholesale Trade 59 Reail Trade 149 Services 218 Public Adminisraion 18 Noe: The firs wo digis of four-digi SIC code of Agriculure, Foresry, and Fishing are 01, 02, 07, 08 and 09; Mining are 10, 12, 13 and 14; Consrucion are 5, 16 and 17; Manufacuring are 20~39; Transporaion, Communicaions, Elecric, Gas, and Saniary Services are 40~49; Wholesale Trade are 50 and 51; Reail Trade are 52~59, Services are 70, 72~73, 75~76, 78~84 and 86~89, Public Adminisraion are 91~97 and Mehods In Hypohesis 1, we expec ha informaion asymmery negaively o rading volume, he number of ransacions and marke liquidiy bu posiively o an individual firm s liquidiy risk as well as marke liquidiy risk. This sudy uses he following model o invesigae he deerminans of informaion asymmery and ries o find wheher empirical daa suppors his hypohesis: 118

8 Model 1 BAS i, = b 0 + b 1 Volume i, + b 2 TRADES i, + b 3 i, + b 4 + b 5 + b 6 SIZE i, + b 7 PRICE i, + 1 d D + (1) i, where BAS i, : he percenage of bid-ask spread in company i in monh ; rading volume of sock i in monh ; monh ; i, : firm i s liquidiy risk in monh ; : marke liquidiy risk in monh ; PRICE i, : he sock price of firm i a he end of monh ; =1,2,,8. Volume i, : TRADES i, :he number of ransacions for sock i in :marke liquidiy in monh ; SIZE i, : he size of firm i a he end of monh ; D : he indusry indicaor, This sudy includes he indusry indicaor o conrol for indusry effec. In addiion, his sudy also conrols for size and price effecs. As small firms end o be inheren wih larger degrees of informaion asymmery, his sudy expecs a negaive associaion beween firm size (SIZE) and spreads. As for he price effec, several researchers (Demsez, 1968; Tinic and Wes, 1972; Mcnish and Wood, 1992) have shown ha here is an inverse relaionship beween a sock s price and is spread. In Hypohesis 2, we expec ha Risk-adused excess reurn relaes negaively o individual firm s liquidiy as well as marke liquidiy, bu posiively o boh individual firm s liquidiy risk and marke liquidiy risk. This sudy uses he following Model 2 o examine he incremenal effecs of liquidiy and liquidiy risk, a he individual and he aggregae level, on excess reurns afer conrolling for he risk facors of marke, size, book-o-marke raio, and momenum. Model 2 AR i, = 0 b + b 1 i, + b 2 i, + b 3 + b 4 + b 5 MKT + b 6 SMB + b 7 HML + b 8 MOM + 1 b D + (2) i, where AR, :excess sock reurn of firm i in monh ; i i, :firm i s liquidiy risk in monh ; i, : firm i s liquidiy in monh ; :he marke liquidiy in monh ; : he marke liquidiy risk in monh ; MKT :he excess reurn of marke porfolio in monh ; SMB :he difference beween he reurns on he small-sock porfolio and he reurns on he big-sock porfolio in monh ; HML :he difference beween he reurns on he high-book-o-marke-raio porfolio and he reurns on he low-book-o-marke-raio porfolio in monh ; MOM :he difference beween he reurns on he winner porfolio and he reurns on he loser porfolio in monh ; D :he indusry dummy variable, 119

9 =1,2,, Variable measuremen The measuremen of he variables used in he models is described as follows: Percenage bid-ask spread(bas i, ): calculaed by (ask-bid) / [(ask+bid) / 2]*100%, where ask and bid are he averages of daily ask and bid prices during monh. Trading volume ( Volume, ): firm i s oal share volume during monh divided by firm i i s shares ousanding a he end of monh. Number of Transacion(TRADES i, ): measured by he square roo of number of ransacions in monh. Individual firm s liquidiy risk(, ):, =( V, -m(v i, ))/s(v i, ), where i i i V i, denoes he average of he daily bid and ask for sock i in monh and m(v i, ) and s(v i, ) are he mean and sandard deviaion of V i, during he research period, respecively. Marke liquidiy( ): Marke liquidiy is measured as per Pasor and Sambaugh (2003), who focus on an aspec of liquidiy associaed wih emporary price flucuaions induced by order flow. Their marke liquidiy measure is a cross-secional average of individual-sock liquidiy measures. Specifically, his sudy uses he following equaion o esimae individual sock s liquidiy in monh : e r i d, 1,, = a+b i d r, r,, + i, Sign( r e i d, Volume,, ) i, d + i, (5) where i, d :sock i s reurns on day d in monh ; i, d :abnormal reurns of sock i on day d in monh, calculaed by subracing he CRSP value-weighed marke reurns on day d in monh from r, r, r i d, r,, ; Sign( e i, d ): sign( r, e r i d, e r,, )=-1 when e r i d,, <0, sign( e r i d,, )=0 e e e when i, d =0, sign( i, d )=1 when i, d >0; Volume i, d, :he rading volume of sock i on day d in monh divided by firm s shares ousanding a he end of monh. The esimae i, represens sock i s liquidiy in monh and marke liquidiy ( ) is esimaed by he average of individual-sock liquidiy measures; ha is =(1/N) N i 1. i, Individual sock s liquidiy(, ): measured by he reciprocal of sock i s bid-ask spread. Marke liquidiy risk( is marke liquidiy and m( i ): ) and s( =( -m( ))/s( ), where ) are he average and sandard 120

10 deviaion of Excess reurn( f Lin, You & Huang over he research period, respecively. AR i, ): AR, = R, - R,, where i i f R, is sock i s reurns in monh and R, is he risk free rae in monh. Marke excess reurn( MKT ): MKT = where R m, is he CRSP value-weighed marke reurns in monh and free rae. Size-relaed risk facor( SMB ): i f R, - R,, m f R, is he risk SMB is he difference beween he reurns on small-sock porfolio and he reurns on big-sock porfolio. This sudy divides he sample firms ino hree groups, wih an equal number of firms in each group, based on each firm s monh-end marke value. The small-sock porfolio conains socks in he lowes-marke-value group, whereas he big-sock porfolio conains socks in he larges-marke-value group. Book-o-marke-raio risk facor( HML ): HML denoes he difference beween he reurns on he high-book-o-marke-raio porfolio and he reurns on he low-book-o-marke-raio porfolio. Again, his sudy divides he sample firms ino hree groups, wih an equal number of firms in each group, based on each firm s monh-end book-o-marke raio. The high-book-o-marke-raio porfolio conains socks in he highes-book-o-marke-raio group, whereas he low-book-o-marke-raio porfolio conains socks in he lowes-book-o-marke-raio group. Momenum risk facor( MOM ): MOM represens he difference beween he reurns on he winner porfolio and he reurns on he loser porfolio. This sudy divides he sample socks ino hree groups, based on he cumulaive reurns over he pas hree monhs. The winner porfolio is formed wih socks in he highes reurn group, where as he loser porfolio is formed wih socks in he lowes reurn group. Firm size(size i, ):SIZE i, = ln MV i, md ln MVm,, where MV i, denoes he naural logarihm of he marke value of firm i s equiy a he end of monh and ln represens he median of he sample firms marke value. md, Sock price(price i, ): aking naural log of he sock price a he end of monh. MV m Indusry dummy variable( D ): As shown in Table 1, he sample firms come from en indusries, ha is, = The findings 4.1 The deerminans of bid-ask spread Table 2 presens he descripive saisics of he variables. As shown, he sample firms 121

11 yield on average a monhly excess reurn (AR) of 2.62%. Among he hree risk facors proposed by Fama and French (1993), he marke risk facor (MKT) has he smalles average premium of 0.18% per monh, followed by he 0.30% of he size-relaed risk facor (SMB), and he book-o-marke facor (HML) has he larges average monhly premium of 1.34%. The momenum facor (MOM) has an even larger average premium, abou 1.49% per monh. Table 3 shows he Pearson correlaion coefficiens. As shown, he univariae correlaion reveals ha bid-ask spread (BAS) relaes significanly and negaively o rading volume (VOLUME) and number of ransacions (TRADES) and posiively o an individual firm s liquidiy risk (, ). The oher wo research variables, marke liquidiy ( ) and i marke liquidiy risk ( ), do no have a significan associaion wih he spread. Wih regard o he correlaion among he independen variables, rading volume has a posiive associaion wih he number of ransacions (TRADE), sock price (PRICE) and marke liquidiy, and a negaive associaion wih an individual firm s liquidiy risk and marke liquidiy risk. Table 4 presens he regression resuls of Model 1. As shown, he D-W saisics are very close o wo, indicaing he problems of auocorrelaion are no presen in he models. The VIF saisics furher reveals no mulicollineariy among he independen variables. Finally, he adused R 2 and he F saisics reveal ha he model is well specified. 122

12 Table 2 Descripive Saisics for Variables Minimum Maximum Mean Sandard Dev. BAS i, Volume, i TRADES i, SIZE i, PRICE i, AR i, , i , i MKT SMB HML MOM Noe: Number of Observaions=73,950; BAS represens he percenage bid-ask spread; TRADES is he number of ransacion; SIZE denoes firm size; VOLUME represens rading volume; PRICE denoes sock price; AR represens a firms monhly excess reurns; i, and i, and denoe individual and aggregae liquidiy, respecively; represens individual and aggregae liquidiy risk, respecively; MKT denoes monhly excess reurn for he marke porfolio; SMB represens monhly premium for he size-relaed risk facor; HML represens monhly premium for he book-o-marke-raio- relaed risk facor; MOM is he monhly premium for he momenum risk facor. 123

13 Table 3 Pearson Correlaion Marix on he Variables in Model 1 BAS i, TRADES i, i, Volume i, SIZE i, PRICE i, BAS i, TRADES i, (0.05) i, (0.000) (0.96) Volume i, (0.00) (0.04) (0.04) SIZE i, (0.16) (0.85) (0.67) (0.25) PRICE i, (0.07) (0.63) (0.36) (0.05) (0.56) (0.55) (0.69) (0.24) (0.01) (0.45) (0.36) (0.76) (0.25) (0.55) (0.01) (0.28) (0.21) (0.49) Noe: Number of Observaions=73,950 Figures in parenheses represen p-values; BAS represens he percenage bid-ask spread; TRADES is he number of ransacion; SIZE denoes firm size;, represens individual firm s liquidiy risk; VOLUME represens rading volume; PRICE i denoes sock price; respecively., and represen aggregae liquidiy and liquidiy risk, We conecure a negaive associaion beween rading volume, number of ransacions and bid-ask spread since larger rading volumes and more ransacions lead o smaller invenory cos for he marke makers. Consisen wih our conecure, he esimaes on rading volume and number of ransacions are significan and negaive, implying an increase in a firm s rading aciviy would improve he firm s liquidiy and hence lower he bid-ask spread. This finding is in accordance wih exisen lieraure (see, for example, Tinic and Wes, 1972; Mcinish and Wood, 1992). The significan and posiive esimae on an individual firm s liquidiy risk is also consisen wih our hypohesis. This finding highlighs ha a sock wih more liquidiy risk would resul ino larger uncerainy in a marke maker s opimal invenory level, and hence he marke maker would charge a higher spread o compensae for he invenory risk he assumes. Buil upon he recen sudies on he co-movemens in bid-ask spreads (see, Chordia e al., 2000; Huberman and Halka, 2001; Brockman and Chung, 2002), his paper also expecs a negaive associaion beween spreads and marke liquidiy. The esimaed coefficien on marke liquidiy has he expeced sign bu is no significan. Furhermore, we hypohesize a posiive effec of marke liquidiy risk on spread because larger aggregae liquidiy risk would resul ino more uncerainy in a marke maker s holdings of any individual securiy. The esimaed coefficien on marke liquidiy risk again has he expeced posiive sign bu is no significan. 124

14 Wih regard o he conrol variables, we expec firm size o have a negaive associaion wih bid-ask spreads because smaller firms are inheren wih a larger degree of informaion asymmery, implying more risk for he marke maker. The esimae has he expeced sign, bu is no significan. Finally, he posiive bu insignifican esimae on sock price runs couner o he evidence of Soll (1978) and Mcinish and Wood (1992), who show ha here is an inverse relaionship beween a sock s price and is spread. Overall, here is srong evidence ha socks wih smaller rading volume, a smaller number of ransacions and larger liquidiy risk have larger spreads, which is supporive of our hypohesis. On he oher hand, we do no find marke liquidiy and marke liquidiy risk o be significanly relaed o spreads. Our findings sugges ha marke makers consider only firm-specific variables in seing a firm s bid-ask spread and disregard he marke-wide facors. BAS i, = b 0 + b 1 Volume i, + b 2 + b 6 SIZE i, + b 7 Table 4 The Deerminans of Bid-ask Spread TRADES i, + b 3 PRICE i, + d 1 i, + b 4 D + i, + b 5 Independen Variable Coefficien -saisics P-Value VIF Consan Volume, i TRADES i, , i SIZE, i PRICE i, R 0.31 Adused R D-W saisics 2.69 F saisics Noe: Number of Observaions=73,950. BAS represens he percenage bid-ask spread; TRADES is he number of ransacion;, represens individual firm s liquidiy risk; VOLUME represens rading volume; PRICE denoes sock price; i and represen aggregae liquidiy and liquidiy risk, respecively., VOLUME represens rading volume; PRICE denoes sock price; SIZE denoes firm size. 4.2 diy, liquidiy risk and excess reurns The univariae correlaions presened in Table 5 reveal ha an excess reurn relaes 125

15 significanly and posiively o a firm s liquidiy and negaively o marke liquidiy risk and individual firm s liquidiy risk. Marke liquidiy, on he oher hand, does no have a significan correlaion wih excess reurns. The hree risk facors and he momenum facor demonsrae posiive associaions wih excess reurns. This paper uses Model 2 o examine if he level of liquidiy per se and liquidiy risk have incremenal effecs on excess reurns in addiion o he hree risk facors proposed by Fama and French (1993) and he momenum facor evidenced in Jegadeesh and Timan (1993). Table 6 presens he regression resuls of Model 2. This sudy conecures invesors would require a higher reurn for invesing in a less liquid securiy. The esimae on liquidiy ( i, ), however, is significanly posiive, indicaing securiy s excess reurn is increasing in is liquidiy. This finding is in conras o he hypohesized negaive associaion beween excess reurns and liquidiy, bu is consisen wih he resuls in Brennan and Subrahmanyam (1996) and Eleswarapu and Reinganum (1993). As individual firm s liquidiy co-moves wih marke-wide facors, his sudy also expecs a negaive relaionship beween excess reurns and marke liquidiy. Again, conrary o he expecaion, he esimae on marke liquidiy ( ) is significanly posiive. Table 5 Pearson Correlaion Marix on he Variables in Model 2 AR i, AR i, i, MKT SMB HML MOM i, 0.00 (0.26) i, 1.90 (0.05) 0.84 (0.00) MKT 0.02 (0.00) 0.00 (0.49) 0.63 (0.47) SMB 0.00 (0.02) 0.91 (0.84) 0.75 (0.62) 0.00 (0.52) HML 0.89 (0.00) 0.00 (0.66) 0.90 (0.03) 0.00 (0.57) 0.00 (0.49) MOM 0.81 (0.02) 0.03 (0.24) 0.44 (0.39) 0.92 (0.69) 0.48 (0.53) 0.65 (0.86) (0.01) (0.49) (0.10) (0.55) (0.06) (0.98) (-0.71) i, (0.05) (0.24) (0.03) (0.02) (0.74) (0.52) (0.45) 0.10 (0.55) 126

16 Noe: Number of Observaions=73,950. AR represens a firms monhly excess reurns; denoe individual and aggregae liquidiy, respecively; i, and i, and represens individual and aggregae liquidiy risk, respecively; MKT denoes monhly excess reurn for he marke porfolio; SMB represens monhly premium for he size-relaed risk facor; HML represens monhly premium for he book-o-marke-raio-relaed risk facor; MOM is he monhly premium for he momenum risk facor. Given ha invesors are risk averse and dislike flucuaions in liquidiy, his sudy conecures ha asse reurns include a significan premium for liquidiy risk. The significan and negaive esimae on individual firm s liquidiy risk (, ) reveals, however, ha a sock s excess reurn is decreasing in is liquidiy risk. Alhough inconsisen wih he expecaion, he finding accords wih he resuls in Chordia e al (2001). In addiion, as marke liquidiy risk affecs every individual sock s liquidiy risk, his sudy also expecs a posiive relaionship beween marke liquidiy risk and excess reurns. The evidence, however, is no in suppor of he hypohesis; he esimae on marke liquidiy risk ( ) is negaive bu insignifican. Consisen wih prior sudies, he esimaes on he marke risk facor and book-o-marke risk facor are significanly posiive, demonsraing ha hese wo risk facors have reliable power in explaining he cross-secion of excess reurns. The size-relaed risk facor, on he oher hand, does no have a significan esimae. While inconsisen wih Fama and French (1993), his evidence goes in he same direcion as ha repored by Brennan, e al. (1998), who find an aenuaed size effec when hey include rading volume, a proxy for liquidiy, in he model. Finally, he resuls presened in Table 6 furher show ha he momenum risk facor, when considered oinly wih oher risk facors, does no have explanaory power for excess reurns. Alhough he direcion of he associaion beween liquidiy, liquidiy risk and excess reurns is unexpeced, he resuls demonsrae srong effecs of liquidiy and liquidiy risk on excess reurns. To sum up, he evidence reveals ha variables relaed o liquidiy per se and liquidiy risk, a he individual as well as he aggregae level, play an imporan role in he cross-secion of excess reurns in addiion o he well-sudied effecs of marke, size, book-o-marke raio, and momenum. i 127

17 Table 6 The Relaion beween diy, diy Risk and Excess Reurns AR i, = 0 b + b 1 + b 6 i, + b 2 SMB + b 7 i, + b 3 HML + b 8 + b 4 MOM + 1 b + b 5 MKT D + (2) Independen Variable Coefficien -saisics P-Value VIF Consan , i i, MKT SMB HML MOM R 0.32 Adused R D-W saisics 2.86 F saisics Noe: Number of Observaions=73,950; AR represens a firms monhly excess reurns;, and denoe individual and aggregae liquidiy, respecively;, and i represens individual and aggregae liquidiy risk, respecively; MKT denoes monhly excess reurn for he marke porfolio; SMB represens monhly premium for he size-relaed risk facor; HML represens monhly premium for he book-o-marke-raio-relaed risk facor ; MOM is he monhly premium for he momenum risk facor. 5. Summary and Conclusions Using he bid-ask spread as he main measure of informaion asymmery, his sudy provides an inegraive analysis of he deerminans of bid-ask spread by considering boh he firm-specific variables and he marke-wide facors. In addiion, given ha an individual firm s liquidiy has been shown o be priced, his sudy also invesigaes if marke liquidiy, an individual firm s liquidiy risk and aggregae liquidiy risk relae o expeced reurns. The empirical analysis wih daa for he NYSE socks reveals ha bid-ask spread relaes negaively o rading volume and number of ransacions and posiively o a firm s i, i 128

18 liquidiy risk, which is consisen wih our expecaion. Marke liquidiy and marke liquidiy risk, however, do no have a significan associaion wih spread. As for he associaion beween liquidiy, liquidiy risk and excess reurns, he evidence indicaes ha, afer adusing for he risk facors of marke, size, book-o-marke raio and momenum, an individual firm s liquidiy and marke liquidiy relae significanly and posiively o excess reurns, whereas individual firm s liquidiy risk relaes negaively o excess reurns. Marke liquidiy risk does no have a significan effec on excess reurns. Alhough he direcion of he associaion beween liquidiy, liquidiy risk and excess reurns is unexpeced, he resuls highligh he role of hese facors in he cross-secion of excess reurns. The unexpeced associaion beween liquidiy, liquidiy risk and excess reurns do no lend hemselves o an obvious explanaion. Fuure research can furher invesigae hese resuls. In addiion, exending he analysis o oher financial markes, such as foreign exchange markes or inernaional equiy markes, would also be useful. Moreover, informaion asymmery variable can furher be divided ino order-processing componen and adverse-selecion componen based on he decomposiion of he bid-ask spread. Endnoes i Their model indicaes ha hey demand a reurn premium for firms wih a higher degree of informaion asymmery (i.e., a higher level of informaion risk). ii Akins, Ng and Verdi (2009) measure he informaion asymmery componen of he bid-ask spread for NYSE, AMEX, and NASDAQ firms for he period from 1983 o 2004 using he model of price formaion. iii Exising explanaions of he presence and he magniude of a spread are based on marke makers consideraions of invenory coss and informaion asymmery. The invenory-based models proposed by Amihud and Mendelson (1980) and Ho and Soll (1981) focus on marke makers exposure o risk hrough he invenories of securiies. The marke makers invenories flucuae over ime o accommodae ransiory excess demand or supply disurbances. The spread is a source of profi o compensae marke makers for exposure o risk and adminisraive coss. Accordingly, he magniude of spread is an increasing funcion of he invenory holding cos. iv v Kyle (1985) indicaes ha liquidiy providers increase bid-ask spread o guard agains adverse selecion risk. For example, Brennan and Subrahmanyam, 1996; Brennan, Chordia and Subrahmanyam,1998; Daar, Naik and Radcliffe,

19 References Amihud, Y and Mendelson, H 1986, Asse pricing and he bid-ask spread, Journal of Financial Economics, Vol.17, pp Akins, B, Ng, J and Verdi R 2009, Invesor compeiion and he pricing of informaion asymmery, Working paper. Brennan, M and Subrahmanyam, A 1996, Marke microsrucure and asse pricing: on he compensaion for illiquidiy in sock reurns, Journal of Financial Economics, Vol.41, pp Brennan, M, Chordia, T and Subrahmanyam, A 1998, Alernaive facor specificaions, securiy characerisics, and he cross-secion of expeced sock reurns, Journal of Financial Economics, Vol.49, pp Brockman, P and Chung, D 2002, Commonaliy in liquidiy: evidence from an order-driven marke srucure, Journal of Financial Research, Vol.25, pp Chordia, T, Roll, R and Subrahmanyam, A 2000, Commonaliy in liquidiy, Journal of Financial Economics, Vol.56, pp Chordia, T, Subrahmanyam, A and Anshuman, VR 2001, Trading aciviy and expeced sock reurns, Journal of Financial Economics, Vol.59, pp Connor, G and Koraczyk, R 1988, Risk and reurn in an equilibrium APT: applicaion of a new es mehodology, Journal of Financial Economics, Vol.21, pp Daar, V, Naik, N and Radcliffe, R 1998, diy and asse reurns: An alernaive es, Journal of Financial Markes, Vol.1, pp Easley, D and O Hara, M 2004, Informaion and he cos of capial, Journal of Finance, Vol.59, pp Eckbo, BE and Norli, O 2005, diy risk, leverage and long-run IPO reurns, Journal of Corporae Finance, Vol.11, pp Eleswarapu, V and Reinganum, M 1993, The seasonal behavior of he liquidiy premium in asses pricing, Journal of Financial Economics, Vol.34, pp Fama, EF and French, KR 1993, Common risk facors in he reurns on socks and bonds, Journal of Financial Economics, Vol.33, pp Gibson, R and Mougeo, N 2004, The pricing of sysemaic liquidiy risk: Empirical evidence from he US sock marke, Journal of Banking and Finance, Vol.28, pp Glosen, L and Milgrom, P 1985, Bid, ask, and ransacion prices in a specialis marke wih heerogeneously informed raders, Journal of Financial Economics, Vol.14, pp

20 Hamilon, J 1978, Markeplace organizaion and markeabiliy: NASDAQ, he sock exchange and he naional marke sysem, Journal of Finance, Vol.33, pp Hasbrouck, J 1998, Securiy bid/ask dynamics wih discreeness and clusering: simple sraegies for modeling and esimaion, Journal of Economic Lieraure, Vol.2, pp Ho, T and Soll, H 1981, Opimal dealer pricing under ransacions and reurn uncerainy, Journal of Financial Economics, Vol.9, pp Huberman, G and Halka, D 2001, Sysemaic liquidiy, Journal of Financial Research, Vol.24, pp Jegadeesh, N and Timan, S 1993, Reurns o buying winners and selling losers: implicaions for sock marke efficiency, Journal of Finance, Vol.48, pp Kyle, AS 1985, Coninuous aucions and insider rading, Economerica, Vol.53, pp Lamber, RA, Leuz C and Verrecchia, R 2009, Informaion asymmery, informaion precision, and he cos of capial, Working Paper. Marshall, BB 2004, The effec of firm financial characerisics and he availabiliy of alernae finance on IPO underpricing, Journal of Economics and Finance, Vol.28, Marinez, M, Nieo, B, Rubio, G and Tapia, M 2005, Asse pricing and sysemaic liquidiy risk: An empirical invesigaion of he Spanish sock marke, Inernaional Review of Economics and Finance, Vol.14, pp Mcinish, TH and Wood, RA 1992, An analysis of inraday paerns in bid/ask spreads for NYSE socks, Journal of Finance, Vol.47, pp Pasor, L and Sambaugh, RF 2003, diy risk and expeced sock reurns, Journal of Poliical Economy, Vol.111, pp Richardson, VJ 2000, Informaion Asymmery and Earnings Managemen: Some evidence, Review of Quaniaive Finance and Accouning, Vol.15, pp Soll, HR 1978, The pricing of securiy dealer services: an empirical sudy of NASDAQ socks. Journal of Finance, Vol.33, pp Tinic, S and Wes, R 1972, Compeiion and he pricing of dealer services in he over-he-couner marke, Journal of Financial and Quaniaive Analysis, Vol.8, pp