Stock Price Synchronicity and Liquidity. Kalok Chan. Allaudeen Hameed. Wenjin Kang. June (Preliminary. Please do not quote)

Transcription

1 Soc Price Synchroniciy and Liquidiy Kalo Chan Allaudeen Hameed Wenjin Kang June 007 (Preliminary. Please do no quoe) * Chan is from he Deparmen of Finance, Hong Kong Universiy of Science and Technology, Clear Waer Bay, Hong Kong, achan@us.h. Hameed is from he Deparmen of Finance and Accouning, Naional Universiy of Singapore, Singapore 11759, Allaudeen@nus.edu.sg. Kang is from he Deparmen of Finance and Accouning, Naional Universiy of Singapore, Singapore 11759, bizwj@nus.edu.sg.

2 Soc Price Synchroniciy and Liquidiy Absrac We provide a simple heoreical analysis based on Kyle s (1985) framewor, and demonsrae how soc price synchroniciy can affec he adverse informaion ris ha mare maers face and herefore he liquidiy of he soc. Our empirical evidence is consisen wih our heoreical conjecure. We find ha socs which co-move more wih he mare index have higher liquidiy, compued based on effecive spread, price impac or Amihud illiquidiy measures. The resuls are obained afer conrolling for cross-secional differences in firm size, price levels, oal and idiosyncraic volailiies, and are robus o boh S&P and non S&P index socs. Besides mare co-movemen, indusry wide componen in reurns also reduces he adverse selecion ris and improves he liquidiy. We also find resuls relaed o indexing effec. Following he addiion o he S&P 500, a firm ha experiences an increase in co-movemen wih he mare is more liely o be accompanied by an improvemen in liquidiy.

3 There has been a recen surge of research wor ha invesigaes he relaionship beween he soc price synchroniciy he degree o which individual socs co-move wih he mare and he efficiency of capial mares. For example, Morc, Yeung and Yu (000) find a higher degree of soc price synchroniciy in emerging mares han in developed mares. They inerpre heir evidence as suggesing ha wea propery righs in emerging mares discourage informed ris arbirage based on firm-specific informaion. As a resul, here is more misallocaion of resources in emerging mares where less firm-specific informaion is produced. Based on a sample of 65 counries, Wurgler (000) shows ha he efficiency of capial allocaion across counries is negaively correlaed wih synchroniciy in domesically raded soc reurns. Based on a sample of U.S. firms, Durnev, Morc and Yeung (000) also demonsrae ha companies ha exhibi less synchroniciy end o use more exernal financing and allocae capial more efficienly. In his paper we examine anoher dimension of he capial mare efficiency ha he soc price synchroniciy is relaed o, namely he liquidiy of he mare. We provide a simple heoreical analysis based on Kyle s (1985) framewor, and demonsrae how soc price synchroniciy can affec he adverse informaion ris ha he mare maers face and how hey adjus he soc prices in response o incoming order flows. Our reasoning is as follows. The soc price synchroniciy measures he amoun of mare-wide informaion relaive o he firmspecific informaion. While mare maers can observe he mare-wide informaion easily, say from he mare movemen, i is more difficul for hem o observe firm-specific informaion. As a resul, he adverse informaion ris is posiively relaed o he proporion of mare-wide informaion and negaively relaed o he proporion of firm-specific informaion. When an 1

4 individual soc is highly correlaed wih he mare, mare maers can rely more on he informaion he observes from he mare price movemen, so ha heir price adjusmens are less sensiive o he order flows. In he exreme case, if an individual soc does no have any firm specific informaion (soc price synchroniciy reaches he maximum), he mare maers will no respond o he order flows for ha individual soc, as all relevan informaion abou he soc can be inferred from he mare movemen. On he oher hand, if an individual soc is less correlaed wih he mare, here is less informaion ha mare maers can base on he mare movemen, so ha hey will adjus more in response o he order flows. As a resul, we conjecure ha when a soc co-moves more wih he mare, he mare deph (or mare liquidiy) is higher. Our heoreical reasoning is relaed o a couple of recen sudies ha explain how he soc price synchroniciy can affec he lising choice of firms. Baruch and Saar (006) show ha a soc will arac more rading volume when i is lised on a mare where similar securiies are raded, or he mare ha he soc is more correlaed wih. Their empirical resuls provide supporing evidence. They documen ha he reurn paerns of securiies lised on he NYSE loo differen from he reurn paerns of Nasdaq securiies. Socs ha are eligible o lis on he oher mare bu do no swich have reurn paerns ha are similar o hose of oher securiies on heir own mare bu differen from he reurn paerns of securiies lised on he oher mare. Baruch, Karolyi and Lemmon (005) also have a similar model o show he disribuion of he rading volume across exchanges compeing for order flow is relaed o he correlaion of he cross-lised asse reurns ha arise in he respecive mares. Based on daa on 75 non-us socs cross-lised on major U.S. exchange, hey find ha volume migraes o he exchange in which he cross-lised asse reurns have greaer correlaion wih reurns of oher asses raded on ha mare.

5 We provide empirical evidence consisen wih our heoreical conjecure. We find ha soc which co-moves more wih he mare index have higher liquidiy. We compue reurn comovemen wih he mare using regression r-square from a mare model or mare beas and find similar resuls. Our resuls are robus o hree alernaive measures of liquidiy: effecive spread, Kyle s price impac measure and Amihud (00) illiquidiy measure. Our resuls canno be explained by cross-secional differences in firm size, price levels, oal and idiosyncraic volailiies, which we use as conrol variables. Furhermore, our resuls hold for boh he S&P 500 index socs and non-s&p 500 index socs We show ha he commonaliy in reurns a he indusry level is posiively relaed o liquidiy: firms wih greaer indusry-wide componen in reurns exhibi higher liquidiy. These resuls suppor our conenion ha larger mare or indusry wide componen in reurns reduces he adverse selecion ris faced by he mare maer and hence, improves he supply of liquidiy. Our resuls are consisen wih an indexing effec, where firms in he major mare index are liely o co-move more wih he mare and experience beer liquidiy. In fac, we show ha ha following he addiion o he S&P 500, a firm ha experiences an increase in co-movemen wih he mare is more liely o be accompanied by an improvemen in liquidiy. Our paper is an exension of previous sudies ha show how adverse selecion affecs he mare liquidiy (Kyle (1985) and Glosen and Milgrom (1985)). Our paper shows he adverse selecion could be miigaed if he mare paricipans could infer he mare-wide informaion componen from he oher asses. This is relaed o Admai (1985) and Callabe and Krishnan (1994) who examine how mare maers learn from addiional informaion of oher socs. Our paper also provides a conribuion o he lieraure on deerminans of mare liquidiy. Earlier 3

6 sudies lie Soll (1978) and Soll and Ho (1981) sugges ha liquidiy is affeced by he invenory riss. Breen, Hodric and Korajczy (00) have showed a se of proxies for he adverse-selecion coss, informaion-relaed coss of mare maing, and he exen of shareholder heerogeneiy can explain cross-secional variaions of mare liquidiy. Chordia, Roll and Subrahmanyam (001) documen ha liquidiy is no consan, bu varies over ime. Our paper provides indirec evidence ha he adverse-selecion coss are affeced by he relaive proporion of mare-wide versus firmspecific informaion, which hen have an impac on he mare liquidiy. The paper is organized as follows. Secion I presens he heoreical formulaion of our main hypohesis. In Secion II we describe he daa and mehodology used in he research. Secion III documens he empirical resuls, and we conclude in Secion IV. I. Theoreical Model This secion develops a simple model ha shows how he soc price synchroniciy or he correlaion of a soc wih he mare affecs he mare deph. We assume ha each soc is raded in a mare organized as in Kyle (1985), where prices are se by se by compeiive and risneural mare maers. We assume ha he payoff of an individual soc (V ~ ) is given by: V ~ = V bf ~ S ~ (1) where V 0 = a consan; F ~ = a common componen; b = he bea sensiiviy of he soc o he common componen 4

7 S ~ = a firm-specific componen We follow he sandard assumpions ha he common componen ( F ~ ) and he firm-specific componen ( S ~ ) are normally disribued, wih means zero and variances F and s, and are uncorrelaed wih each oher. As a resul, he variance of he soc ( v ) equals F S b +. We assume ha mare paricipans are able o observe a noisy signal abou he common facor ( F ~ ). One source of informaion is from he index fuures mare whereby mare paricipans observe he soc index fuures price ( f ~ ) which conains he common facor informaion wih noise: ~ f = F ~ + e~ () In equaion (), he index fuures price ( f ~ ) does no fully reflec he common facor due o he exisence of e ~, which can be inerpreed as he noise rading or liquidiy rading in he fuures mare. The variable ( ~ e ) is assumed o be normally disribued, wih mean 0 and variance. We follow Kyle (1985) and assume here are hree ypes of players in he soc mare. e The firs is a single ris-neural informed rader, who has perfec informaion abou he payoff of he individual soc. In oher words, he informed rader observes V ~ perfecly. The second ype is liquidiy raders who rade for liquidiy purpose. They do no possess any informaion, and he orders hey subm denoed u ~, are normally disribued wih mean 0 and variance u. The hird ype is he mare maer. As in Kyle (1985), he mare maer does no observe he orders submied by he informed invesors and liquidiy raders separaely. Insead, hey observe he aggregae orders submied by boh, and deermine an equilibrium price based on he aggregae order flows. 5

8 We firs derive he order submission sraegy of he informed rader. As in Kyle (1985), he informed rader is ris neural, and his objecive is o maximize expeced profis from he rading. Since he informed rader is aware ha his order submission will have an effec on he price of he aucion, he sees o choose he opimal amoun of order quaniies ( x ) o buy or sell, based on he informaion ha he receives. The informed rader assumes he mare maer se he price based on he following rule: ~ P ~ = V f ( x~ u ~ 0 + λ 1 + λ + ) (4) In oher words, he informed rader conjecures ha he price is adjused in response o he aggregae orders submied by he informed rader and noise raders in he soc mare ( x~ + u ~ ), as well as he index fuures price ( f ~ ). The pricing rule is an exension of he one in Kyle (1985) as we allows he individual soc price o be se as a funcion of also he index fuures price. As saed previously, he informed rader sees o maximize he expeced profi (π ) based on his informaion abou V ~. Alhough he index fuures price ( f ~ ) is readily observable, i is irrelevan o he informed rader since he already obains perfec informaion abou V ~. As a resul, he can maximize he expeced profi based on his own privae informaion. In oher words, ~ E [ π f = f,v ~ = V ] = E[ x(v ~ P ~ ) V ~ = V ] (5) Subsiuing Equaion (4) ino Equaion (5) and aing he expecaion: E 0 λ [ x(v ~ P ~ ) V ~ = V ] = x[v P x ] (6) To maximize he profi (π ), he informed invesor should choose he opimal order as follows: 6

9 1 x = (V V λ 0 1 ) = ( bf + S ) λ (7) Since noise raders are assumed o submi random quaniies ( u ~ ), he aggregae orders submied by informed invesors and noise raders are: 1 x~ + u ~ = ( bf ~ + S ~ ) + u ~ λ (8) As mare maers are assumed o operae in a perfecly compeiive seing, zero expeced profis imply he price will se equal o he expeced value of he soc. The linear projecion rule gives he price as follows: ~ P ~ = E[V ~ f, x~ + u ~ ] ~ = V + λ1 f + λ( x~ 0 + u ~ ) where λ = 1 Cov( v, f )Var( x Var( f )Var( x + u ) Cov( v,x + u )Cov( f,x + u ) + u ) [ Cov( f,x + u )] Cov( v,x + u )Var( f ) Cov( v, f )Cov( f,x + u ) λ = Var( f )Var( x + u ) [ Cov( f,x + u )] (9) The parameer λ is he variable of ineres, since i measures he price impac of order flow, or he inverse of mare deph. We can solve for λ based on he formula in equaion (9): λ v b F ( )( F + e ) ( b F )( ) λ λ = (10) v b F ( F + e )( + u ) ( ) 4λ λ Rearranging he erms, 7

10 8 ) ( 4 b e F u s F e s e F λ = (11) In he special case whereby he variance of he noise in he fuures mare ( e ) equals zero, equaion (11) will be reduced o u s F u s F 4 λ = = (1) The soluion for he price impac is similar o ha in Kyle (1985) whereby here is only one piece of informaion being possessed by he informed rader. In our special case whereby 0 e =, here is no uncerainy abou he common facor and ha he mare maer could infer perfecly from he fuures price. Therefore, he only unnown informaion is he soc-specific componen, which is he same as in Kyle (1985). Based on he relaionship ha s F v b + =, Equaion (11) could be simplified as: ) ( 4 )) R (1 ( ) ( 4 e F u F e v e F u s F v e λ + + = + + = (13) ) / ( 1 R where v s =, which is he R-square from he mare model and be used as a measure of he degree of soc price synchroniciy in our empirical analysis. Equaion (13) indicaes ha oher hings being equal, an increase in he soc price synchroniciy ( R ) will decrease he price impac ( λ ), which is equivalen o a higher mare deph. The inuiion is obvious. Wih an increase in he soc price synchroniciy, mare maers could infer more of he informaion from he index fuures mare, reducing he adverse selecion ris hey face in rading wih informed raders in he individual soc. Consequenly, hey will

11 adjus he price less in accordance wih order flows, enhancing he mare deph or liquidiy. II. Daa and Mehodology The sample comprises of all NYSE raded ordinary common socs, idenified by he Cener for Research in Securiy Prices (CRSP) share code 10 and 11, over he period January 1989 o December 003. We do no include closed-end funds, ADRs, ec as heir liquidiy characerisics may differ from common socs ha will mae i difficul o inerpre our findings. We also exclude socs raded on NASDAQ o avoid he influence of differences in rading proocols. The CRSP daase conains daily soc reurns, daily rading volume, number of shares ousanding, and yearly mare capializaion. We filer ou socs wih exreme price levels by discarding socs wih prices below $3 and above $999. For ransacion-level daa, we rerieve all rades and quoaions from he New Yor Soc Exchange Trades and Auomaed Quoaions (TAQ) and he Insiue for he Sudy of Securiies Mares (ISSM). 1 While our heoreical framewor models liquidiy based on he (inverse of) mare deph, here are alernaive empirical measures of liquidiy. We experimen wih several differen measures of firm level liquidiy used in he prior lieraure. Each liquidiy proxy we use can be viewed as measuring differen dimensions of liquidiy. Our firs measure of liquidiy is λ, moivaed by he price-impac measure inroduced in Kyle (1985), and derived in our heoreical framewor. For each firm, we use he price and quoe informaion from TAQ o classify every rade as buyer (seller) iniiaed, based on wheher he ransacion price is greaer (lower) han he 1 Anomalous ransacion records are deleed according o he following filer rules: (i) Negaive bid-as spread; (ii) Quoed spread > $5; (iii) Proporional quoed spread > 0%; (iv) Effecive spread / Quoed spread >

12 prevailing average of bid and as quoes. Specifically, we follow he algorihm presened in Lee and Ready (1991) in signing he ransacions as buy and seller orders, maching rading records o he mos recen quoe preceding he rade by a leas five seconds. We aggregae he buy and sell orders a he daily level and compue firm i s order imbalance, OIB i, defined as he difference beween he value of daily buyer and seller iniiaed rades. Our esimae of Kyle s λ for firm i is he coefficien from he following univariae regression, r i = α i + λioibi + εi, where r i is he daily reurn on soc i. For each year, we esimae he daily price impac coefficien λ i by regressing daily reurns on is corresponding order imbalance. Using a similar measure, Brennan and Subrahmanyam (1996) find ha he price impac measure of liquidiy is posiively relaed o average soc reurns, suggesing ha illiquid socs are compensaed wih liquidiy premium. Breen, Hodric and Korajczy (00) show ha he price impac varies across socs, and depends on a number of firm characerisics variables, including he severiy of he adverse selecion problem. Our second liquidiy measure is effecive bid-as spread. Soll (1978), Glosen and Harris (1988) and ohers show ha bid-as spreads include an adverse selecion coss of he mare maer rading wih invesors wih superior informaion. We calculae he proporional effecive spread based on wo imes he absolue difference beween he rade execuion price and he midquoe, We also use a ransacion-based measure of Kyle s λ as an alernaive. Here, we follow he approach of Brennan and Subrahmanyam (1996): le p j and q j denoe he price and he signed quaniy of he order fulfilled for ransacion j, and D j denoe he sign of he order. The regression Δ p + j = α + λt q j + ΨΔD j ε j provides he esimaes for he Kyle s λ T. Is common pracice o use Δ p j = (p j p j-1 )/p j-1 as he regressor o ensure ha our comparison of he price impac esimae across differen socs is no affeced by he price level. Our resuls are qualiaively similar when we use his ransacionbased measure. 10

13 divided by he midquoe. The daily average proporional effecive spreads for firm i is averaged each calendar year o generae our annual spread measure (ESPR i ). The hird liquidiy measure is based on Amihud (00) and does no rely on inraday ransacions daa. I is calculaed as he absolue daily reurn on soc i divided by he firm s daily dollar volume. Using his measure of he relaive price change associaed wih rading volume, Amihud finds ha firms wih greaer expeced illiquidiy earn higher expeced reurns, consisen wih a illiquidiy premium in reurns. Acharya and Pedersen (005) also use his illiquidiy measure o invesigae he effec of liquidiy on securiy reurns. Hasbrouc (005) also shows ha he Amihud illiquidiy measure is a robus measure of price impac as posied in Kyle (1985). A commonaliy among he hree measures is ha hey are in fac illiquidiy measures. If a soc is less illiquid, i will have a higher bid-as spread, is order flow will have a larger impac on soc prices, and he absolue price change per uni of volume is greaer. Therefore, an increase of any of hese measures is an indicaion of lower liquidiy. To invesigae he relaion beween liquidiy and he amoun of mare-wide informaion, we rely on he sandard mare model regressions o exrac he mare-wide componen in reurns. We sar wih he regression of weely soc reurns on soc i a wee (R ) on he CRSP valueweighed weely mare reurns (R m, ): R ε = a + β irmj, + We use wo relaed measures of he relaive amoun of mare-wide informaion esimaed from he regression. The firs is he soc s bea ( β ), which measures he responsiveness of he soc s reurn o mare reurns. The second measure is he R from he mare model regression, which reflecs he proporion of variaion in soc reurn explained by mare reurns. Since he R saisics is bounded beween zero and one, we also consruc an alernaive measure of price synchroniciy (SYNCH) by aing 11

14 he logi-ransformaion of R : ln(r /(1-R )). A higher β or SYNCH indicaes a larger amoun of mare-wide informaion in soc i. As expeced, β and SYNCH are highly correlaed measures of mare wide informaion. Table 1 conains summary saisics on our liquidiy and price synchroniciy measures for each year from 1989 o 003. Panel A repors he mean and sandard deviaion of R, soc s bea, as well as hree liquidiy measures (effecive bid-as spread, price impac, and Amihud illiquidiy measure). As expeced, he beas average o abou one wih a significan spread across securiies. The R from he mare model regression averages o abou 17.8 percen over he whole sample period, wih yearly averages in he range of 10 percen o 35 percen. All hree liquidiy measures display a downward rend over ime. For example, he effecive bid-as spread was 0.731%, 0.799%, and 0.841% in he firs hree years, and declined seadily during he sample period, and was 0.337%, 0.49%, and 0.178% in he las hree years. The oher wo liquidiy measures also have similar paern, alhough he declines are less pronounced. Over he sample period, he price impac measure declined from o 1.311, while he Amihud illiquidiy measure declined from 14.9 o Panel A also presens he cross-secional sandard deviaions. The price impac and Amihud illiquidiy measures exhibi subsanial cross-secional dispersion, as heir sandard deviaions are more han 3 imes larger han he mean values. The paern is less dramaic when we examine he percenage effecive spreads, as he sandard deviaion is slighly smaller han he mean value. Panel B of Table 1 repors he uncondiional correlaion across firm-years among he R, bea, he liquidiy proxies, and some oher conrol variables. In general, all he liquidiy measures are highly correlaed wih each oher, a resul consisen wih Korajczy and Sada (005) and 1

15 Hasbrouc (006). We find ha he daily price impac coefficien λ i has he highes correlaion wih Amihud illiquidiy measure (see Hasbrouc (005)), and is also highly correlaed wih effecive spreads. The bea and R is also highly correlaed, wih a correlaion coefficien of 68%.. III. Regression Analysis A. Cross-secional resuls Several sudies show ha cross-securiy variaion in liquidiy can be explained by firm characerisics. Soll (1978, 000), Ho and Soll (1981) and Harris (1994) show ha proporional bid-as spreads are lower for bigger firms and high volume socs as hey are associaed wih higher probabiliy of finding a counerpary o rade and hence lowers he invenory and order processing coss faced by he mare maer. They also show ha spreads are higher for socs wih high reurn variance due o compensaion for invenory riss as well as he ris of rading wih an informed rader. Habrouc (1991) repors ha greaer price impac (and informaion asymmery) is higher for smaller firms. Breen, Hodric and Korajczy (00) show ha price impac of rades are relaed o firm s relaive size, volume, absolue reurns, price synchroniciy, insiuional ownership, and oher predeermined firm specific variables. In his secion, we invesigae he cross-secional relaion beween liquidiy and he amoun of mare-wide informaion relaive o firm-specific informaion. The amoun of sysemaic informaion in soc reurns is measured by soc reurn synchroniciy (SYNCH) and bea ( β ). We inroduce a se of conrol variables ha oher sudies have shown o affec firm level liquidiy, independen of he amoun of mare-wide informaion. For each firm, we compue he oal reurn volailiy (ToVol), measured by he sandard deviaion of weely reurns wihin he 13

16 calendar year. Alernaively, we also consider idiosyncraic volailiy measured by he sandard deviaion of residual reurns from he mare model regressions, (IdioVol). A number of conrol variables are included o explain he liquidiy. The firs is firm size, which is equal o he log of mare capializaion a he beginning of he year. Firm size is a proxy for informaion asymmery, which migh affec he liquidiy of he soc, The second variable is he percenage of he firm s equiy held by insiuional invesors. For a given monh, he insiuional holdings are measured as of he end of he previous quarer. Since a higher insiuional ownership is accompanied by more informaion disclosure and lower informaion asymmery, his is accompanied by higher firm liquidiy. The hird variable is he inverse of price (1/P), where P is he beginning of year price for he firm. When liquidiy is measured by proporional effecive spreads, i is possible ha he cross-secional variaion in liquidiy is affeced by differences in he price levels. Hence, we include he inverse of price as a conrol variable. Finally, we also include urnover, which is he raio of rading volume in he previous monh o he mare capializaion. By consrucion, since urnover is endogenously deermined, i migh capure he influences of oher variables ha explain our liquidiy measures. Therefore, he effec of our synchroniciy measures on he liquidiy measures migh be undersaed. To examine he relaion beween liquidiy and mare-wide informaion, we perform he following panel regressions from 1989 o 003: Liquidiy 1 = a + b SYNCH + c ToVol + d CONTROL (14) d 1 Liquidiy = a + b Bea + c IdioVol + CONTROL (15) 14

17 where CONTROL 1 is he value of conrol variable for firm i a end of year -1, wih he conrol variables including firm size, insiuional ownership, urnover and inverse of price level. Since SYNCH capures he sysemaic porion of a soc s oal volailiy and IdioVol is he volailiy of he residual reurns, hese wo variables are expeced o be negaively correlaed. Consequenly, we do no include boh variables in he same regression. Our main findings are presened in Table. Panel A conains he resuls using effecive bid-as spread as he dependen variable. Regression resuls indicae a firm s effecive spreads are negaively relaed o he securiy s sysemaic informaion. Holding a firm s oal volailiy consan, an increase in soc reurn synchroniciy would cause a decrease in bid-as spread. Holding a firm s idiosyncraic volailiy consan, an increase in bea is also associaed wih a decrease in bid-as spread. These resuls indicae a firm s liquidiy will be improved wih a higher proporion of mare-wide informaion. Resuls for he conrol variables are consisen wih our conjecure. Bid-as spreads are negaively relaed o firm size, insiuional ownership, urnover, and posiively relaed o inverse of price level. We have also pariioned he sample period ino hree sub-periods, and he resuls are robus. Panel B conains he resuls using price impac coefficien as he dependen variable. Resuls are similar o hose based on he effecive bid-as spread. There is a saisically significan and negaive relaionship beween price impac and price synchroniciy. Consisen wih our heoreical predicion, price impac of rade is smaller for socs wih higher co-movemen wih he mare. Socs wih higher mare model R or higher mare beas are associaed wih lower price impac, indicaing ha hese socs have lower asymmeric informaion. Also, firms wih higher idiosyncraic (or oal) reurn volailiy are exposed o greaer adverse selecion ris 15

18 and, hus, experience larger price impac. For he sub-period analysis, he resuls are generally robus, excep ha he relaionship beween price impac and soc reurn synchroniciy is no saisically significan in he sub-period. Similar findings are obained when we use Amihud illiquidiy as he dependen variable, which are repored in Panel C of Table. Amihud illiquidiy is lower for firms wih higher soc reurn synchroniciy or bea, suggesing ha he liquidiy of a securiy will be improved when here he proporion of idiosyncraic volailiy is smaller. We also obain robus resuls in he subperiod analysis, excep in he firs period whereby he negaive relaionship beween Amihud illiquidiy and soc reurn synchroniciy is no saisically significan. B. Time-series resuls I is possible ha he cross-secional relaionship beween he liquidiy measures and soc reurn synchroniciy could be due o omied firm characerisic ha is correlaed wih hese variables. To circumven his problem, we also explore he ime-series paern by invesigaing he relaionship using changes of liquidiy and changes of he explanaory variables. Each year, we compue changes in R, bea, oal volailiy, idiosyncraic volailiy, as well as changes in conrol variables, and esimae he following regression specificaions:. Δ Liquidiy = a + b ΔR + c ΔToVol + d ΔCONTROL 1 (16) Δ Liquidiy = a + b ΔBea + cδidiovol + d ΔCONTROL 1 (17) Resuls are repored in Table 3. Consisen wih cross-secional regression resuls, we find 16

19 ha afer conrolling for he change of oal volailiy ( Δ ToVol ), he change of R-square ( Δ R ) has a negaive effec on change of liquidiy variables, regardless of wheher effecive spreads, price impac and Amihud illiquidiy measures are used. In oher words, afer a firm experiences an increase of soc price synchroniciy, here is a decline of adverse informaion ris so ha he securiy will see an improvemen of liquidiy. Liewise, afer conrolling for he change of idiosyncraic volailiy ( Δ IdioVol ), an increase of bea ( Δ Bea ) will have a negaive effec on he liquidiy variables. I is, however, noed ha while he effecs on bid-as spreads and Amihud illiquidiy measures are saisically significan, he effec on price impac measure is no. C. Co-movemen due o Indusry Effecs Since our focus is on he impac of sysemaic informaion on liquidiy, we also examine he influence of indusry-wide informaion, separae from mare reurns. We as he quesion of wheher greaer indusry-wide informaion reduces adverse selecion ris and, hence, increases liquidiy. We do his by compuing he bea and synchroniciy measures based on a wo-facor model wih mare and indusry facors, and compare wih hose based on a one-facor mare model. If we denoe model S as he one-facor model, and model T as he wo-facor model: Model S: R = a + Bea R + ε (18) m,i m, j, Model T: R = a + Bea R + Bea R + ε (19) m,i m, j, ind,i ind, j, where R ind,j, refers o weely reurns on indusry j porfolio corresponding o firm i a wee, (j=1,..,17 consruced using he Fama and French 17-indusry classificaion) and R m,j, is he reurn on he mare porfolio excluding indusry j. In he wo-facor model, Beam, i represens he 17

20 mare bea, so ha we could examine heir separae effecs on he liquidiy measures. We also derive he incremenal effec on soc price synchroniciy due o indusry-wide co-movemen. If we denoe R from Model S and Model T as R S and R T, respecively, we can ae he logdifference in regression R-square from he wo-facor and single-facor mare models: ln(r diff) = ln(r T) ln(r S). We can hen esimae he effec of ln(r diff) on soc liquidiy afer he marelevel reurn synchroniciy (R S )is conrolled. The following are he wo regression specificaions based on incremenal synchroniciy and indusry bea: Liquidiy Liquidiy = a + b m ln R S, + bids ln Rdiff, + c ToVol + d CONTROL 1 + ε = a + bm Beam, + bids BeaIND, + c IdioVol + d CONTROL 1 + ε (0) (1) Resuls are presened in Table 4. Evidence shows here is addiional effec on liquidiy from indusry-level co-movemen. The incremenal synchroniciy due o indusry effec (ln(r dif)) has a negaive and saisically significan effec on all hree liquidiy measures. This suggess ha afer conrolling for mare reurns, a securiy wih a higher indusry co-movemen can reduce he adverse informaion, inducing an improvemen in he liquidiy. Liewise, he indusry bea BeaIND, are also negaive and saisically significan effec in all differen specificaions. In fac, he coefficiens associaed wih associaed wih BeaIND, are almos as large in magniude as he coefficiens Bea m,. For example, in he specificaion based on effecive bid-as spread, he coefficien of BeaIND, is -5.36, which is only sligh lower in magniude han he coefficien of Bea m, (-6.1). This resul indicaes ha he indusry co-movemen is as imporan as mare co-movemen, in he sense ha boh of hem can help o improve he soc liquidiy. Even a soc 18

21 does no co-move wih he mare much, bu if invesors sill can infer he informaion of he securiy from he indusry movemen, he adverse informaion ris can remain small while he soc liquidiy remains high. D. Addiional Analysis based on index socs Our evidence ha higher mare-wide componen in reurns increases liquidiy is also consisen wih an alernae explanaion: socs reurn commonaliy and liquidiy are boh higher for index componen socs. Vijh (1994) and Barberis, Shliefer and Wurgler (004) show ha socs added o he index have higher beas or sysemaic co-movemen. Harris and Gural (1986) and Harford ad Kaul (1998) find evidence suggesing an increase in rading volume and liquidiy for socs which are added o he Index. Hence, we explore if our finding of higher liquidiy when prices are synchronous can be aribued o index socs. We obain daa on he lis of firms which belong o he S&P 500 index from Compusa. We use he yearly variable S&P Primary Index Marer (CPSPIN) o idenify socs which are included in he S&P 500 index a he beginning of each year and exclude hese socs from our analysis. Our argumen is ha if indexing is he primary cause of he relaion beween liquidiy and reurn synchroniciy, we expec he relaion o be subsanially weaer when we repea he crosssecional ess on non-s&p consiuen socs. In Table 5, we coninue o find he socs wih higher soc price synchroniciy, higher bea or higher sysemaic volailiy have lower liquidiy. Again, he resuls are very srong for hree liquidiy measures, wih he effecs of soc reurn synchroniciy, bea, and sysemaic volailiy all saisically significan. Therefore, resuls are robus even among socs ha are no included in he S&P index. 19

22 We also conduc addiional analysis by examining he liquidiy impac of he socs added o S&P 500 index. According o Vijh (1994) and Barberis, Shleifer and Wurgler (005), he socs ha are added o he S&P 500 index experience a significan increase in bea afer inclusion. Barberis, Shleifer and Wurgler (005) also provide evidence ha he increase in bea is no oally aribued o fundamenal effec, bu can also be aribued o mare-fricion and invesor senimen. Table 6 presens addiional analysis based on socs added o S&P 500 index. Panel A repors he change of bea and R afer he inclusion. As we limi our analysis o only NYSE socs, our sample is only a subse of socs used in Barberis, Shleifer and Wurgler (005) who examine boh he NYSE and NASDAQ socs. Panel A reveals a difference beween our sample and he full sample. We confirm he resuls in Barberis, Shleifer and Wurger (005) ha here is a significan increase in R and bea for he full sample of socs added o S&P 500 index. In conras, for our sample of socs, while here is a saisically significan increase in R, he increase in bea is no significan. Panel B repors he regression analysis of how he change of co-movemen afer inclusion in he S&P index affecs he effecive spreads. We do no repor he resuls based on oher illiquidiy measures o save space. Panel B shows ha here is a srong and negaive relaionship beween he increase in R and he decline in effecive bid-as spreads for hose socs added o S&P index. In conras, we do no find he change of bid-as spreads o be relaed o he increase in bea. This migh be relaed o Panel A resuls ha here is only a small increase in bea, so ha he cross-secional variaion of bea is oo small o explain he change in bid-as spread. Overall, 0

23 our resuls confirm previous evidence ha socs will exhibi an increase in co-movemen and liquidiy afer hey are added o S&P index. IV. Conclusion Recen lieraure has discussed soc price synchroniciy in he conex of efficiency of capial mares. If a mare or a company exhibi less synchroniciy, his is an indicaion of producion of more firm-specific informaion (Morc, Yeung and Yu (000)) and a more efficien allocaion of capial (Durnev, Morc and Yeung (000)). This paper proposes ha soc price synchroniciy could also affec he liquidiy of he individual firm. Since mare-wide informaion is much easier o observe han he firm-specific informaion, a firm wih a larger mare-wide componen in reurns will have less adverse informaion. Therefore, a firm wih a higher co-movemen will exhibi a higher mare deph (or liquidiy). We provide srong empirical evidence o suppor our proposiion. Regardless of wha illiquidiy measures (bid-as spreads, price impac and Amihud illiquidiy) we use, an increase in soc price synchroniciy (an increase in R-square or an increase in bea) will resul in a decline in hese illiquidiy measures. Furhermore, he effec on liquidiy is no confined o co-movemen wih he mare. Afer conrolling for he mare reurns, he indusry co-movemen also has significan effecs on liquidiy. Our resuls show ha he relaionship does no hold only for index socs, bu non-index socs as well. Our paper also sheds ligh on he change of co-movemen and liquidiy afer socs are added o S&P index. Previous lieraure ends o rea hem as separae issues. While one 1

24 srand of research (for example, Vijh (1994) and Barberis, Shleifer and Wurgler (005)) examines he increase in co-movemen for socs included in he index, anoher srand examines he effec on liquidiy (Harris and Gural (1986) and Harford ad Kaul (1998)). Our paper shows ha he wo effecs migh be indeed relaed, as he increase of R-square is negaive relaed o he decline of bidas spread for hose socs added o he S&P 500 index. While we are unable o ell while he increase in co-movemen is base on fundamenal or senimen facor, i noneheless affecs he adverse informaion ris in he mare and has an indirec impac on mare liquidiy.

25 References Adma A., 1985, A Noisy Raional Expecaions Equilibrium for Muli-Asse Securiies Mares, Economerica 53, Acharya, Viral V., and Lasse Heje Pedersen, 005, Asse pricing wih liquidiy ris, Journal of Financial Economics, 77, Amihud, Yaov, 00, Illiquidiy and soc reurns: cross-secion and ime-series effecs, Journal of Financial Mares 5, Barberis, Nicholas, Andrei Shleifer and Jeffrey Wurgler, 005, Comovemen, Journal of Financial Economics 75, Baruch, S., G. A. Karoly and M. L. Lemmon, 006, Muli-Mare Trading and Liquidiy: Theory and Evidence, forhcoming, Journal of Finance Baruch, S., and G. Saar, 006, Asse Reurns and he Lising Choice of Firms woring paper Breen, William, Laurie Simon Hodric, and Rober Korajczy, 00, Predicing equiy liquidiy, Managemen Science 48, Brennan, Michael J., and Avanidhar Subrahmanyam, 1996, Mare microsrucure and asse pricing: On he compensaion for illiquidiy in soc reurns, Journal of Financial Economics 41, Caballe, J., and M. Krishnan, 1994, Imperfec Compeiion in a Muli-Securiy Mare wih Ris Neuraliy, Economerica, 6, Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam, 001, Mare liquidiy and rading aciviy, Journal of Finance 56, Daar T, Nai N Y, Radcliffe R., 1998, Liquidiy and soc reurns: An alernaive es. Journal of Financial Mares 1, Durnev, Aryom, Randall Morc, and Bernard Yeung, 003, Value enhancing capial budgeing and firm-specific soc reurns variaion. Journal of Finance 59, Glosen, Lawrence R., and Lawrence E. Harris, 1988, Esimaing he componens of he bid/as spread, Journal of Financial Economics 1, Glosen L. R. and L. Harris, 1988, Esimaing he Componens of he Bid/As Spread, Journal of Financial Economics 1,

26 Glosen, L. and P. Milgrom, 1985, Bid, as and ransacion prices in a specialis mare wih heergeneously informed agens, Journal of Financial Economics 14, Korajczy, R. and R. Sada, 005, Pricing he Commonaliy Across Alernaive Measures of Liquidiy, woring paper Kyle, A. S., 1985, Coninuous Aucions and Insider Trading, Economerica, 53, Harris, Lawrence, 1994, "Minimum Price Variaions, Discree Bid/As Spreads and Quoaion Sizes," Review of Financial Sudies 7, Harris, L. and Gurel, E., Price and volume effecs associaed wih changes in he S&P 500: new evidence for he exisence of price pressure. Journal of Finance 41, Harford, J., A.Kaul Do concenraed rading equilibria exis? The migraion of informed rading following index addiion, Woring paper, Universiy of Oregon Hasbrouc, Joel, 005, Trading coss and reurns for US equiies: The evidence from daily daa, Woring paper, New Yor Universiy. Ho, Thomas, and Hans R. Soll, 1981, "Opimal Dealer Pricing under Transacions and Reurn Uncerainy." Journal of Financial Economics 9, Lee, Charles M. C., and Mar J. Ready, 1991, Inferring rade direcion from inraday daa, Journal of Finance 46, Morc, R., Yeung, B., Yu, W., 000. The informaion conen of soc mares: why do emerging mares have synchronous soc price movemens? Journal of Financial Economics 58, Soll, H., 1978, The supply of dealer services in securiies mares. Journal of Finance 33, Ho and Soll, 1981, Opimal dealer pricing under ransacions and reurn uncerainy, Journal of Financial Economics 9, Vijh, A., S&P 500 rading sraegies and soc beas. Review of Financial Sudies 7, Wurgler, J., 000, Financial mares and he allocaion of capial, Journal of Financial Economics 58,

27 Table 1: Descripive Saisics: Soc Reurn Comovemen and Liquidiy The proporional effecive spread for firm ESPR% i, is defined as * ransacion price midquoe / midquoe, where he midquoe is equal o he average of he as quoe and he bid quoe. The daily ESPR% j is generaed by averaging he effecive spread of all he ransacions wihin he day j, and he annual ESPR% is he average of he daily effecive spread for all he rading days in he given calendar year. For he firm i in year, we esimae he price impac by regressing he firm daily reurns on is corresponding order imbalance. The Price Impac esimae is he coefficien (λ i ) from he following univariae regression, r j =λ OIB j + e j, where r j is he daily reurn on soc and OIB j is he difference beween he dollar value of he daily buyerand seller- iniiaed rades. The Amihud Illiquidiy is defined as he absolue daily reurn on soc i divided by he firm s daily dollar volume. The annual Amihud Illiquidiy is he average of he daily Amihud Illiquidiy for all he rading days in calendar year. To measure he soc reurn comovemen, we regress weely soc reurns on soc i (R w ) on he CRSP value-weighed weely mare reurns (R m,w ): R w = a +Bea R m,w + e j in calendar year. We use wo relaed comovemen measures derived from he reurn regression above: he firs is he soc s bea, Bea, which measures he responsiveness of he soc s reurn o mare reurns; he second measure is based on he regression R, which reflecs he proporion of variaion in soc i s reurn explained by mare reurns. Since he R saisics is bounded beween zero and one, we ae he logi-ransformaion of R, ha is, Synch = ln(r /(1-R )) as he measure of price synchroniciy for soc i in year. Panel A repors he cross-secional mean and sandard deviaion of he reurn bea and R, ESPR%, Price Impac, and Amihud Illiquidiy for every calendar year from 1989 o 003. Panel B repors he Pearson correlaion among he reurn comovemen measures and liquidiy measures, as well as he log of firm capializaion (size), urnover, and percenage of insiuional ownership. 3

28 Panel A: Summary Saisics: Reurn Bea, R-square and Liquidiy measures Year R Bea ESPR% Price Impac Amihud Illiquidiy 1989 Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion Mean Sd. Deviaion ~003 Overall Average

29 Panel B: Correlaions R-square Bea Price Impac Amihud Illiquidiy ESPR% Toal Volailiy Idiosyncraic Volailiy Size Turnover Insiuional Ownership R-square Bea Price Impac Amihud Illiquidiy ESPR% Toal Volailiy Idiosyn. Volailiy Size Turnover Insiuional Ownership

30 Table : Liquidiy and Reurn Comovemen: Cross-secional Tess We regress annual soc liquidiy on is reurn comovemen measures, synch and bea, using he fixed-effec panel regression mehod. We employ hree regression specificaions: Model A: Liquidiy Model B: Liquidiy Model C: Liquidiy = a + b Synch + d Size + d Size = a + b Bea + d Size + c Toal _ Volailiy + e Insiuional _ Ownership + e Insiuional _ Ownership + c Idiosyncraic _ Volailiy = a + b Sysemaic_ Volailiy + e Insiuional _ Ownership + f Turnover ( + g (1/ P )) + ε + f Turnover ( + g (1/ P )) + ε + c Idiosyncraic _ Volailiy + f Turnover ( + g (1/ P )) + ε The hree liquidiy measures for firm i in year, which are used as dependen variables are proporional effecive spread (ESPR%), Price Impac, and Amihud Illiquidiy. The regression resuls based on he hree liquidiy measures are repored in Panels A, B, and C respecively. In each panel we provide he regression resul based on he full sample period (1989 o 003), and hree sub-periods, ha is, , , and The -saisics are repored in he parenhesis. For he independen variables, Synch is he logi-ransformaion of R from he maremodel regression for soc i in year. Bea is he bea coefficien from he mare-model regression. Toal_Volailiy represens he oal volailiy, measured by he sandard deviaion of soc i s weely reurn in year. Idiosyncraic_Volailiy represens he sandard deviaion of soc i s weely idiosyncraic reurn in year, aen from he residual of he mare-model regression. Sysemaic_ Volailiy is measured by square roo of he explained variaion of he mare-model regression. Oher conrol variables include Size, he log value of he firm capializaion measured a he beginning of year. Insiuional_Ownership sands for he percenage insiuional ownership measured a he beginning of year. Turnover represens he daily number of shares raded divided by oal shares ousanding, in year. If he dependen variable is he proporional effecive spread, we also include he inverse of he price level of he soc a he beginning of year, P. 6

31 Panel A: Dependen Variable = ESPR% Sample Period: 1989~003 Sub-Sample Period: 1989~1993 Sub-Sample Period: 1994~1998 Sub-Sample Period: 1999~003 Synch Bea Toal Idiosyncraic Sysemaic Insiuional Size Volailiy Volailiy Volailiy Ownership Turnover 1/P (-15.60) (48.9) (-67.0) (-14.3) (-30.66) (78.87) (-1.08) (51.90) (-65.71) (-14.4) (-8.0) (78.40) (50.61) (-8.98) (-65.68) (-14.69) (-8.51) (78.5) (-11.3) (40.83) (-34.8) (-6.6) (-7.3) (43.91) (-7.79) (43.84) (-33.40) (-5.69) (-6.16) (43.75) (-11.34) (30.75) (-40.9) (-8.0) (-0.98) (73.65) (-9.13) (33.46) (-39.01) (-8.07) (-19.7) (73.3) (-5.17) (6.08) (-43.0) (-7.88) (-19.67) (30.93) (-5.83) (6.81) (-41.5) (-8.08) (-17.88) (30.67) 7

32 Panel B: Dependen Variable = Price Impac Sample Period: 1989~003 Sub-Sample Period: 1989~1993 Sub-Sample Period: 1994~1998 Sub-Sample Period: 1999~003 Synch Bea Toal Idiosyncraic Sysemaic Insiuional Size Volailiy Volailiy Volailiy Ownership Turnover (-4.93) (3.73) (-33.35) (-11.4) (-19.93) (-4.99) (4.46) (-3.04) (-11.19) (-18.53) (3.67) (-3.08) (-3.14) (-11.43) (-18.78) (-1.5) (16.09) (-14.43) (-4.69) (-15.51) (-.08) (16.3) (-13.4) (-4.01) (-14.68) (-.85) (19.0) (-.35) (-8.5) (-15.56) (-.3) (19.9) (-1.1) (-8.33) (-14.61) (-4.71) (11.) (-5.37) (-8.18) (-11.46) (-3.88) (11.97) (-4.76) (-8.34) (-10.5) Panel C: Dependen Variable = Amihud Illiquidiy Sample Period: 1989~003 Sub-Sample Period: 1989~1993 Sub-Sample Period: 1994~1998 Sub-Sample Period: 1999~003 Synch Bea Toal Idiosyncraic Sysemaic Insiuional Size Volailiy Volailiy Volailiy Ownership Turnover (-4.93) (3.73) (-33.35) (-11.4) (-19.93) (-4.99) (4.46) (-3.04) (-11.19) (-18.53) (3.67) (-3.08) (-3.14) (-11.43) (-18.78) (-1.5) (16.09) (-14.43) (-4.69) (-15.51) (-.08) (16.3) (-13.4) (-4.01) (-14.68) (-.85) (19.0) (-.35) (-8.5) (-15.56) (-.3) (19.9) (-1.1) (-8.33) (-14.61) (-4.71) (11.) (-5.37) (-8.18) (-11.46) (-3.88) (11.97) (-4.76) (-8.34) (-10.5) 8

33 Table 3: Changes of Liquidiy and Reurn Comovemen We regress changes in annual soc liquidiy on he changes in is reurn comovemen measures, synch and bea, using he fixed-effec panel regression mehod. All he dependen and independen variables are measured as he log-change from year -1 o year. The liquidiy measures used as he dependen variables include he proporional effecive spread (ESPR%), Price Impac, and Amihud Illiquidiy. The regression is based on he sample period from 1989 o 003. The -saisics are repored in he parenhesis. Dependen ΔToal ΔIdiosyncraic ΔInsiuional ΔR ΔBea ΔSize Variables Volailiy Volailiy Ownership ΔTurnover Δ1/P ΔESPR% (-sa) (-8.5) (41.36) (-.7) (1.6) (-39.67) (8.63) ΔESPR% (-sa) (-3.84) (41.51) (-.57) (0.41) (-38.46) (6.79) ΔPrice Impac (-sa) (-.64) (7.6) (-13.61) (.6) (-39.6) ΔPrice Impac (-sa) (1.1) (3.40) (-13.31) (.45) (-36.66) ΔAmihud (-sa) (-6.99) (44.04) (-60.7) (-.61) (-93.16) ΔAmihud (-sa) (-1.99) (43.0) (-57.77) (-.63) (-90.64) 9

34 Table 4: Liquidiy and Reurn Comovemen: Effec of indusry reurn comovemen We esimae reurn comovemen from a single-facor model and a wo-facor model as follows: Model S: = a +ζ R + R w m, mj, w ε w Model T: R w = a + Beamj, Rmj, w + BeaIND, Rindj, w + ε w where R indj,w refers o weely reurns on indusry j porfolio corresponding o firm (j=1,..,17 consruced using he Fama and French 17-indusry classificaion) and R mj,w is he reurn on he mare porfolio excluding indusry j. The R from Model S and Model T are denoed as R S and R T respecively. To examine he incremenal effec on indusry-wide co-movemen, we ae he log-difference in regression R-square from he wo-facor and single-facor mare models, denoed as ln(r diff,) = ln(r T) ln(r S). The indusry- and mare-level reurn synchroniciy measures are hen used as independen variables in he regression below. Liquidiy = a + bm ln RS, + bids ln Rdiff, + c Toal _ Volailiy + d Sizei, + e Insiuional _ Ownership + f Turnoveri, ( + g (1/ Pi, )) Liquidiy = a + bm Beam, + bids BeaIND, + c Idiosyncraic _ Volailiy + d Size + e Insiuional _ Ownership + f Turnoveri, ( + g (1/ Pi, )) + ε IdiosyncraicVolailiy is measured by he sandard deviaion of soc i s weely idiosyncraic reurn in year, aen from he residual of he wo-facor reurn model regression in Model T. The dependen variables and he res of he independen variables are defined he same way as Table. The -saisics are repored in he parenhesis. + ε Dependen Variables ln (R_S) ln (R_diff) Bea_m Bea_IND Toal Volailiy Idiosyncraic Volailiy Size Insiuional Ownership Turnover 1/P ESPR% (-sa) (-19.9) (-1.3) (47.7) (-59.5) (-14.93) (-9.73) (79.4) ESPR% (-sa) (-11.05) (-11.05) (48.50) (-6.6) (-16.4) (-4.40) (76.61) Price Impac (-sa) (-5.58) (-.63) (3.41) (-30.17) (-11.46) (-19.69) Price Impac (-sa) (-6.33) (-4.81) (3.94) (-30.30) (-1.01) (-17.16) Amihud (-sa) (-8.45) (-3.67) (14.96) (-8.56) (-7.07) (-18.18) Amihud (-sa) (-7.73) (-8.70) (19.46) (-8.36) (-7.64) (-17.14) 10