Does Stock Price Synchronicity Represent Firm-Specific Information? The International Evidence

Transcription

1 Does Sock Price Synchroniciy Represen Firm-Specific Informaion? The Inernaional Evidence Hollis Ashbaugh-Skaife Universiy of Wisconsin Madison 975 Universiy Avenue Madison, WI Joachim Gassen Ruhr-Universiä Bochum Universiässraße Bochum, Germany Ryan LaFond Sloan School of Managemen Massachuses Insiue of Technology 50 Memorial Drive Cambridge, MA March 005 Revised Ocober 005 We hank S.P. Kohari, seminar paricipans a he 005 European Accouning Associaion Annual Congress and he 005 Columbia Inernaional Symposium for helpful commens and suggesions.

2 Does Sock Price Synchroniciy Represen Firm-Specific Informaion? The Inernaional Evidence ABSTRACT Much of prior inernaional accouning research implicily assumes ha sock prices impound informaion wih similar precision across counries. Recen research assers ha sock price synchroniciy, defined as he R from asse pricing regressions, is a useful measure of he amoun of informaion refleced in sock prices. This paper invesigaes he validiy of he informaion-based inerpreaion of sock price synchroniciy in inernaional markes. The resuls of our analyses provide lile suppor for using sock price synchroniciy as a measure of informaion inernaionally. We develop an alernaive measure of informaion based on he percenage of zero reurn days, i.e., zero-reurn meric, and repea he analyses. Overall, our resuls sugges ha he zero-reurn meric is a beer measure of he relaive amoun of informaion impounded ino share prices han he synchroniciy measure in inernaional markes.

3 Does Sock Price Synchroniciy Represen Firm-Specific Informaion? The Inernaional Evidence 1. Inroducion Prior research documens differences in he value relevance, imeliness, and conservaism of accouning informaion across counries (see e.g., Alford, Jones, Lefwich, and Zmijewski 1993; Ball, Kohari, and Robin 000; Ball, Robin, and Wu 003). An underlying assumpion of his line of research is ha sock prices impound informaion wih similar precision across counries. Morck, Yeung, and Yu (000) propose ha sock price synchroniciy, defined as he R from asse pricing regressions, can be used as a measure of he relaive amoun of firm-specific informaion refleced in reurns inernaionally. There are significan differences in mandaed and volunary informaion flows, ownership srucures, rading aciviy, and marke fricions across counries, ha affec he price formaion process in inernaional markes. If sock reurns do no capure informaion wih he same precision across counries, hen sock price synchroniciy is limied in is use as a measure of he informaion refleced in sock prices inernaionally. In his paper, we invesigae he exen o which he sock price synchroniciy measure, i.e., he R measure, is associaed wih firm-specific informaion in he inernaional seing. Morck e al. (000) inerpre higher R values, i.e., greaer sock price synchroniciy, as reurns ha reflec more marke-wide informaion and lower R values as reurns ha reflec more firm-specific informaion. We evaluae his inerpreaion of he synchroniciy measure by conducing five analyses using daa from six of he larges equiy markes; Ausralia, France, Germany, Japan, he U.K., and he U.S. 1 Our firs analysis builds on he work of Durnev, Morck, Yeung, and Zarowin (003) who assess wheher he R measure is associaed wih he informaiveness of U.S. firms sock prices. If he R measure consisenly reflecs firm-specific informaion in reurns inernaionally, we expec lower R values o be associaed wih more informaive prices wih respec o fuure earnings. Overall, he resuls are no consisen wih his 1 The Appendix displays he summary findings in 15 addiional counries ha have sufficien daa o conduc our five analyses.

4 expecaion. We find lower R values are associaed wih more informaive prices in Ausralia. However, we find higher R values are associaed wih more informaive prices in Germany, Japan and he U.S., and find no saisically significan associaion beween fuure earnings and reurns condiional on R values in France or he U.K. Thus, in conras o Durnev e al. (003) who sugges ha greaer firm-specific reurn variaion is associaed wih more informaive sock prices in he U.S., our resuls sugges ha here is no consisen relaion beween he R measure and he pricing of fuure earnings informaion inernaionally. Our second analysis invesigaes wheher sock price synchroniciy is associaed wih analys forecas errors. Prior research documens a negaive relaion beween firm disclosures and analyss forecas errors (Lang and Lundholm 1996, Hope 003). The work of Ashbaugh and Pincus (001), and Lang, Lins, and Miller (003) indicaes ha analyss forecass errors decline when a non-u.s. firm s public informaion se expands due o providing addiional disclosures as a resul of adoping Inernaional Financial Reporing Sandards or U.S. Generally Acceped Accouning Principles, respecively. Therefore, if low R values reflec greaer amouns of firm fundamenals being capialized ino share prices following he release of firm-specific informaion, we expec a posiive relaion beween analyss forecas errors and he R measure. The resuls are consisen wih his expecaion in Japan, where we documen a posiive associaion beween he R measure and analys forecas errors. In Ausralia, France, Germany, he U.K., and he U.S., however, we find ha firms wih higher R values have lower analys forecas errors. These findings conradic he noion ha he synchroniciy measure consisenly capures he relaive informaion refleced in reurns in inernaional markes. Durnev e al. (003) use a mached pair design and addiional conrols in he U.S. o asses wheher firms wih lower R are associaed wih more fuure earnings informaion refleced in reurns. While he use of he mached paired design has advanages, he resuls poenially do no address he average cross-secional relaion beween he R and fuure earnings informaion in he marke. Since we are ineresed in assessing he associaion beween he R measure and he pricing of fuure earnings wihin he respecive marke, we use a cross-secional design. Thus our resuls speak o he sensiiviy of he Durnev a al. (003) findings o alernaive design choices. Morck e al. (000) view he R as a summary measure of he amoun of informaion refleced in reurns, under his inerpreaion he inclusion of conrol variables, such as size, which influence boh he R measure and he pricing of fuure earnings informaion is no appropriae.

5 In our hird analysis, we invesigae wheher here is a change in sock price synchroniciy surrounding firms cross lising in he U.S. Cross lising in he U.S. represens a major informaion even in ha U.S. foreign regisrans are required o provide more disclosures relaive o he disclosures required in home markes (Ashbaugh 001, Lang e al. 003), and, in general, are more visible o new invesors, which increases invesors informaion search (Karolyi 004). If lower R measures represen relaively more firm-specific informaion in reurns, we expec a decline in R values following firms cross lising in he U.S. We find no evidence ha Ausralian, French, German, Japanese, or U.K. firms R values declined following cross lising in he U.S. Raher, we find ha he R values of firms domiciled in France and he U.K. increased afer a U.S. lising. The resuls of his analysis raise furher doub on wheher he synchroniciy measure can be used o capure differences in informaion impounded in sock price inernaionally. Our fourh analysis ess he associaion beween he synchroniciy measure and variables ha prior research uses o proxy for firm-specific informaion flows and firm fundamenals. Specifically, we use he reporing of a loss, he disclosure of research and developmen coss, analys following, and he percenage of closely held shares o proxy for he quaniy and qualiy of firms informaion flows, and he sandard deviaion of sales and reurn-on-asses o proxy for firm fundamenals. If sock price synchroniciy measures he relaive amoun of firm-specific informaion refleced in prices inernaionally, we expec consisen relaions beween R values and he variables ha proxy for firms informaion flows (fundamenals) across our sample counries. Afer conrolling for firm size, urnover, and indusry regulaion, we find significan associaions beween he R measure and he informaion proxies (firm fundamenals) across counries. However, he associaions are inconsisen in ha we find posiive relaions beween he R measure and informaion proxies (firm fundamenals) in some counries and negaive relaions in oher counries. These findings provide addiional evidence ha he informaionbased inerpreaion of he R measure is no valid on a consisen basis in inernaional markes. 3

6 The work of Andrade e al. (005), Barberis e al. (005), Kumar and Lee (005), and Greenwood (005), suggess ha non-fundamenal facors affec firms sock price synchroniciy. Building on Barberis e al. (005), we conduc one more analysis ha examines he associaion beween German firms membership in a German-marke index and heir R values. The German marke provides a unique seing o assess he usefulness of he synchroniciy measure as a measure of informaion because membership in cerain German indices during our analysis period requires firms o provide addiional disclosures ha are inended o increase firm-specific informaion flows. 3 If he synchroniciy measure reliably reflecs he amoun of firm-specific informaion, we expec he R values of German firms ha are members of an index o be lower han oher German firms. Conrary o he informaion-based explanaion for he synchroniciy measure, we find ha membership in a German index is associaed wih significanly higher R values. These resuls sugges ha non-fundamenal facors significanly influence sock price synchroniciy in he German marke. Collecively, he resuls of our se of analyses sugges ha he variaion in firms synchroniciy measures inernaionally does no consisenly capure differences in firm-specific informaion flows. Having provided evidence ha he informaion-based inerpreaion of he synchroniciy measure is no reliable in inernaional markes, we develop and invesigae wheher an alernaive marke measure beer capures he relaive amoun of informaion in reurns inernaionally. Building on he work of Bekaer, Harvey, and Lundblad (003) and Lesmond, Ogden, and Trzcinka (1999), we use he percenage of zero-reurn days as a simplified measure of firmspecific-informaion. Lesmond e al. (1999) noe if he value of an informaion signal is insufficien o exceed he coss of rading, hen he marginal invesor will no rade. If he marginal invesor does no rade, hen here is no change in price, and a zero reurn resuls. We apply his noion o he inernaional seing assuming ha when sufficien informaion arrives in 3 We limi his analysis o he German marke because of he difficuly in idenifying non-u.s. firms index memberships over ime. 4

7 he marke, invesors rade, and a reurn is generaed. Therefore, we conjecure ha he proporion of zero reurns days (hereafer referred o as he zero-reurn meric) represens he frequency of informaion arrival, where a lower zero-reurn meric (i.e., smaller proporion of zero-reurn weeks) reflecs more informaionally efficien share prices. We repea our five analyses using he zero-reurn meric in place of he synchroniciy measure and find he following. In assessing wheher he zero-reurn meric is associaed wih more informaive prices wih respec o fuure earnings, we find, as expeced, ha he zero-reurn meric is negaively associaed wih he amoun of earnings relaed informaion refleced in reurns in France, Germany, Japan, he U.K., and he U.S. Consisen wih expecaions, he resuls of he analys forecas error analysis indicae ha he zero-reurn meric is posiively associaed wih analyss forecas errors in Ausralia, France, he U.K., and he U.S. In addiion in France and he U.K. we find a significan decline in he zero-reurn meric following firms cross lising in he U.S. When we regress he zero-reurn meric on informaion proxies and firm fundamenals, we find he relaions beween he zero-reurn meric and informaion proxies and fundamenals o be consisen wih expecaions. Based on he resuls of hese analyses, we conclude ha he zero-reurn meric is a beer measure of he relaive amoun of informaion impounded in share prices han he R measure inernaionally. Our sudy makes several conribuions o he lieraure. Firs, he resuls of our sudy sugges ha lower sock price synchroniciy does no capure he amoun of firm-specific informaion refleced in sock prices in inernaional markes. Morck e al. (000) documen ha sock prices move ogeher more in poor counries relaive o rich counries, and sae ha crosscounry differences in propery righs explain he cross-counry variaion in sock price synchroniciy. Morck e al. (000) conclude ha srong propery righs promoes informed arbirage, which capializes deailed firm-specific informaion ino prices leading o lower sock price synchroniciy. The analysis and inferences drawn from Morck e al. (000) are based on aggregae counry-level measures, i.e. he counry averages. As discussed by Freeman (004), 5

8 Freedman, Pisani, and Purves (1998) and Greenland and Robbins (1994) using aggregae daa as opposed o firm-specific daa can resul in differen and poenially incorrec inferences being drawn. While he Morck e al. (000) resuls speak o aggregae differences across counries, heir analysis does no allow hem o disenangle wheher he synchroniciy measure reflecs firmspecific informaion wihin a counry and, consequenly, wheher he synchroniciy measure capures common informaion across counries. Our sudy measures and assesses R values a he firm level wihin a counry. By examining he properies of he synchroniciy measure across firms wihin a counry we hold marke micro-srucure and insiuional feaures, which poenially affec securiy pricing, consan hereby allowing us o es wheher differences in he synchroniciy measure wihin a counry reflec more informaion-laden sock prices. The resuls of our firm-level analyses sugges ha differences in synchroniciy across firms are no driven by differences in informaion. Thus, we conclude ha he informaion-based explanaion for he synchroniciy measure is no valid in inernaionally, and quesion wheher across-marke comparisons can be made using he synchroniciy measure. 4 Second, our research conribues o prior and concurren research ha invesigaes measures inended o capure he degree o which sock prices are informaionally efficien (see e.g., Gelb and Zarowin 00, Kelly 005). We develop he percenage of zero reurn days, i.e., zero-reurn meric, as an alernaive measure of informaion in reurns ha capures he degree o which prices are informaionally efficien. Unlike oher measures of informaion efficiency (e.g., he breadh of insiuional ownership), he zero-reurn meric can be consruced for every firm lised in a public equiy marke. We provide evidence ha he zero-reurn meric is a more reliable measure of informaion-laden sock prices han he synchroniciy measure, and link he zero-reurn meric o facors ha capure informaion flows and firm fundamenals. Having a 4 Ohers have begun o assess he robusness of prior inernaional research ha draws inferences based on counry-wide measures. For example, Holderness (005) demonsraes he conclusions regarding he influence of weak legal insiuions on ownership srucure appear o be due o he use of aggregae (i.e., counry-level versus firm-level) measures of ownership srucure. 6

9 simple measure ha capures differences in he degree of firms price informaiveness is imporan o researchers and regulaors ineresed in he inegraion of capial markes as well as invesors whose opimal resource allocaion depends upon informaionally efficien prices. The paper proceeds as follows. Secion provides an overview of he prior lieraure on he R measure. Secion 3 exends he findings of Morck e al. (000) o our sample period. Secion 4 repors he resuls of he analysis examining he R measure a he firm level. Secion 5 describes he zero-reurn meric ha we develop o capure he relaive amoun informaion refleced in sock prices and repors he resuls of he analysis conduced using his new measure of firm-specific informaion. Secion 6 repors he resul of addiional sensiiviy analysis and Secion 7 concludes he sudy.. Overview of he R Measure and Relaed Lieraure Roll (1988) noes ha ypical asse pricing regressions yield relaively low explanaory power and proposes ha one poenial explanaion for his is he incorporaion of privae, firmspecific informaion ino prices. 5 Asse pricing models ypically regress firms reurns on a se of common facors. For example, he capial asse pricing model (CAPM): (1) RETi = β 0, i + β1, i RETMKT + ε i where RET i is he firms i reurn for period and RETMKT is he reurn on he marke for period. For his model o yield a high explanaory power he firm in quesion mus rade wih he marke, meaning is share price mus align wih he share prices of oher firms in he marke, i.e., i mus exhibi synchronous sock price movemens. The convenional inerpreaion of he residual from equaion (1) is ha afer removing he reurn effecs due o sysemaic facors, he remaining reurn volailiy is due o idiosyncraic, firm-specific evens. A low R from equaion (1) is poenially due o firms reurns capuring 5 The decline he R values overime was documened by research prior o Roll (1988). Blume (1968), King (1966) and Officer (1971) all find a decline in he explanaory power of he marke model overime. Roll (1988), however, was he firs o propose and invesigae differences in R values across firms due o differences in he amoun of firms-specific informaion refleced in prices informaion. 7

10 unique firm-specific informaion or reflecing greaer idiosyncraic noise in reurns. Roll (1988) noes ha he incorporaion of firm-specific informaion ino prices generally increases he volailiy of an individual firm s sock price, which resuls in lower explanaory power from asse pricing regressions such as equaion (1). He finds ha he low R from asse pricing models are primarily due o high firm-specific reurns volailiy and ha his volailiy is no associaed wih public news announcemens. Based on his finding Roll conens ha privae informaion or else occasional frenzy unrelaed o concree informaion (p. 566) is driving he high firm-specific reurn volailiy. 6 Morck e al. (000) is he firs in a series of papers o use measures capuring he degree o which sock prices move ogeher as a measure of he relaive amoun of firm-specific informaion refleced in sock prices. Using counry level measures of sock price synchroniciy, Morck e al. (000) find ha sock prices in poorer counries wih less developed equiy markes, weaker proecion of invesor righs, and weaker legal regimes end o move ogeher more. Morck e al. (000) conclude ha sronger proecion of invesor righs promoes informed rading resuling in more informaive sock prices as evidenced by less synchronous rading in hese counries. Jin and Myers (005) confirm he findings of Morck e al. (000) and documen a decline in counry level R values overime across a sample of 40 counries. In addiion, Jin and Myers (005) find ha counries wih higher R values experience more frequen marke crashes and ha he higher R values and more frequen crashes boh resul from more opaque informaion environmens. Li, Morck, Yang, and Yeung (004) invesigae he behavior of counry level measures of sock price synchroniciy in emerging markes finding ha R values are generally declining overime and lower R values are associaed wih greaer capial marke openness, more 6 One poenial way in which firm-specific informaion is impounded ino prices is hrough he acions of raders wih privae informaion abou firm fundamenals. Acions underaken by informed marke paricipans such as analyss or insiders will resul in firm-specific sock price movemens, hus providing a poenial explanaion for high firm-specific reurns volailiy. 8

11 efficien legal sysems and less corrup economies. While he counry level resuls of Morck e al. (000), Jin and Meyers (005) and Li e al. (004) are consisen wih an informaion-based inerpreaion of he R measure, he resuls do no address wheher he synchroniciy measure reflecs informaionally efficien prices a he firm level. 7 Research examining wheher he synchroniciy measure reflecs informaion efficien prices a he firm level has focused on he U.S. seing. One line of research builds on he fac ha firm-specific informaion is impounded ino prices hrough he public disclosure of informaion or hrough he acions of informed marke paricipans. Durnev e al. (003) examine wheher firms wih low synchroniciy are associaed wih more fuure earnings informaion being refleced in reurns. They find ha U.S. firms wih lower R values have more fuure earnings informaion refleced in sock prices, consisen wih differences in synchroniciy across firms being due o differences in he amoun of informaion refleced in prices. Pioroski and Roulsone (004) es he associaion beween synchroniciy and acions of informed marke paricipans (i.e., analyss, insiuional invesors, and insiders). 8 They find ha acions underaken by informed marke paricipans are associaed wih firms sock price synchroniciy. Finally Durnev e al. (004) invesigae wheher firms wih lower R values in he U.S. make beer capial allocaion decisions. They find ha firms wih lower R values end o make more efficien invesmens (less over or under invesmen). Their finding is consisen wih he synchroniciy measure represening firm-specific informaion, in ha firms wih lower R values suffer from fewer problems wih asymmeric informaion improving he coordinaion beween capial suppliers and he firm, resuling in more efficien invesmens by he firm. 7 As noed by Pioroski and Roulsone (004, p. 116) radiional differences a he counry level are no likely he cause of observed differences in synchroniciy. Insead, differences in R s are a resul of he economics underlying each firm and he relaive flow of informaion ino prices. 8 Chan and Hameed (005) invesigae he associaion beween he R measure and analys following in emerging markes, and find ha higher analys following is associaed wih higher R values. They inerpre heir findings as being consisen wih analys impounding marke wide (no firm-specific) informaion ino reurns. 9

12 Anoher line of U.S. research explores wheher low R values are a resul of excess noisein-reurns resuling from oher facors unrelaed o firm fundamenals. Shiller (1981) and Wes (1988) find ha he level of sock price volailiy is oo high o be explained by he volailiy in he underlying fundamenals, e.g. dividends. Wes (1988) provides a heoreical model where increased firm-specific reurn volailiy is associaed wih less firm-specific informaion and more noise-in-reurns. In Wes s (1998) model relaively more informaion resuls in prices being closer o fundamenal values, and he release of new informaion resuls in smaller price movemens and lower firm-specific reurn volailiy. Wes (1988) empirically ess his model and repors resuls ha indicae ha firm-specific reurn volailiy is posiively associaed wih bubbles, fad, and oher non-fundamenal facors. Oher sudies sugges ha behavioral facors, bubbles, herding and oher non-fundamenal facors affec sock price volailiy (see Shleifer 000 for a review), and ulimaely he usefulness of he synchroniciy measure as gauge of firm-specific informaion. Barberis e al. (005) find significan changes in firms R values surrounding addiions and deleions o he S&P 500 index in he U.S., consisen wih marke fricions influencing synchroniciy. 9 In addiion, Greenwood and Sosner (00), and Greenwood (005) find similar resuls in Japan using addiions and deleions from he Nikkei 5 index. Since addiions and deleions o indices do no signal new informaion o he marke regarding firms fundamenals, he changes in firm s R values surrounding changes in he composiion of indices is inconsisen wih an informaion-based explanaion of he R measure. Kumar and Lee (005) find ha noise raders (uninformed reail invesors) influence sock price synchroniciy. Their findings are consisen wih he noise in reurns explanaion for he R measure. Andrade e al. (005) develop a model in which rading imbalances combined 9 Barberis e al. (005) develop a model o explain he changes in R values based on marke fricions and senimen. Boh heir empirical and heoreical work provides evidence inconsisen wih he informaionbased explanaion of he R measure. 10

13 wih limied risk bearing capaciy of arbirageurs resuls in correlaed price movemens across socks. An imporan feaure of heir model is ha synchronous price movemens resul from cross sock price pressure, no informaion. Andrade e al. (005) es heir model in Taiwan and heir resuls indicae ha limied risk bearing capaciy on he par of arbirageurs can explain a significan porion (more han 50%) of observed sock price synchroniciy, which is inconsisen wih he informaion-based inerpreaion of synchroniciy. Thus, he findings of Andrade e al. (005), Barberis e al. (005), Greenwood and Sosner (00), Greenwood (005), and Kumar and Lee (005) indicae ha marke fricions, i.e., facors unrelaed o informaion, have a significan influence on sock price synchroniciy. Campbell, Marin, Malkiel, and Xu (001) documen he rise in firm-specific reurn volailiy in he U.S. overime and he resuling decrease in R values. They inerpre heir findings in he spiri of Wes s (1988) model conending ha he decrease in R values is no likely a resul of increased firm-specific informaion. Wei and Zhang (004) invesigae he poenial causes for increased firm-specific volailiy overime in he U.S., and find ha he variance of firm fundamenals (reurn on equiy) has increased overime hereby providing a parial explanaion for he findings of Campbell e al. (001). However, Wei and Zhang furher documen ha he increase in volailiy of firm fundamenals and he associaion beween fundamenal volailiy and reurn volailiy is driven, for he mos par, by newly lised firms, casing doub on an informaion-based explanaion for declining R values. Rajgopal and Venkaachalam (005) documen a posiive associaion beween informaion risk, as measured by accrual qualiy and analys forecas dispersion, and firm-specific reurns volailiy. Their findings are consisen wih he heoreical work of Pasor and Veronesi (003), who demonsrae ha uncerainy abou firms fundamenals (informaion risk) influences reurns volailiy. These sudies provide furher evidence agains he informaion-based inerpreaion of he synchroniciy measure. If greaer firm-specific reurns volailiy is associaed 11

14 wih poorer qualiy informaion (greaer uncerainy) hen how can higher firm-specific reurn volailiy also be associaed wih more firm-specific informaion being refleced in reurns? Overall heoreical and empirical sudies provide lile suppor for he informaion-based inerpreaion of he synchroniciy measure. In addiion, argumens relaed o he limis and risk of arbirage indicae ha firm-specific reurn volailiy may hinder informed rading raher han be a consequence of informed rading as claimed by Roll (1988) and Morck e al. (000). Subsequen inernaional research ends o assume ha Morck e al. s (000) counry-wide measure of sock price synchroniciy is a measure of he relaive amoun of firm-specific informaion refleced in firms sock prices. 10 To dae, however, we know of no evidence ha validaes he informaion-based explanaion for he synchroniciy measure inernaionally. 3. Replicaion We measure a firm s sock price synchroniciy following Morck e al. (000) who define synchroniciy as he percen of he variaion in a firm s sock reurns explained by variaions in he firm s domesic marke reurn and he U.S. marke reurn. Specifically, he synchroniciy measure is he R from esimaing he following firm-specific regression: () RETi = β 0, i + β1, i RETMKTc + β, i RETMKTUS + ε i where RET i is he reurn for firm i for he wo week period, RETMKT c is he reurn on he marke for counry c for period, and RETMKTUS is he reurn on he US marke over period. 11 Like Morck e al. (000), we use bi-weekly reurns o deal wih infrequen rading in inernaional markes. We use value weighed marke reurns, where all reurns, including he reurn on he U.S. marke, are calculaed in he local currency and are colleced from Daasream. We esimae equaion () by firm over he 5 week period encompassing he firm s fiscal year. We require 10 For example, Wurgler (000) examines he associaion beween counry-level measures of sock price synchroniciy and counry-level measures of he efficiency of capial allocaions, and DeFond and Hung (004) invesigae he associaion beween counry-level synchroniciy measures and CEO urnover inernaionally. 11 For he U.S. sample we include only he reurn on he U.S. marke in equaion (). 1

15 firms o have a minimum of 30 weeks of non-zero reurns o esimae equaion (). To deal wih exreme reurns, we exclude all reurn observaions wih absolue values greaer han 0.5 (Morck e al. 000). 1 The counry-level synchroniciy measures are defined as: (3) R c, R i i, c, SSTi = SST i i, c,, c, where SST i,c, is he oal sum of squared variaions from he firm-specific esimaes of equaion () wihin each counry. Panel A of Table 1 repors he descripive saisics for he counry-wide R measures of 1 developed equiy markes, where all firm-year observaions relae o firms ha have sufficien daa o esimae equaion (). We selec hese 1 counries because hey are a subse of he counries sudied by Morck e al. (000) ha have firms wih sufficien accouning and marke daa o be included in our empirical ess. 13 For breviy, however, we able and discuss only he resuls of our empirical ess for Ausralia, France, Germany, Japan, he U.K., and he U.S. The Appendix summarizes our findings in he 15 oher counries. Panel A of Table 1 repors he mean (median) values of he R measure presened in order of average counry rank. To calculae he average counry rank, each year we rank he 1 sample counries by heir R value and repor he average counry rank across he 13 sample years. The U.S. repors he lowes mean and median R values (mean value of and median value of 0.097) as well as he lowes mean counry rank of 1.615, followed by he Canada, Ausralia and 1 Including hese reurn observaions does no change any inferences drawn from he resuls. 13 To be included in he analysis presened in Table 1 we only require firm-year observaions o have sufficien weekly reurns daa o calculae he R measure and o be on Worldscope. The requiremen ha firm-year observaions are on Worldscope reduces our sample sizes compared o Morck e al. (000). However his requiremen ensures ha firm-year observaions included in Panel A of Table 1 have he necessary financial informaion available o conduc our oher empirical ess. 13

16 France. The highes mean counry ranks are found in Spain (16.69), Ialy (17.077), and Singapore (19.31). Morck e al. (000) repor ha wealhier counries (as measured by gross domesic produc) wih common law legal regimes, and wih greaer proecion of invesor righs have lower sock price synchroniciy. To replicae Morck e al. (000), we esimae he following OLS regression: 14 00, 1,, 3, ε fye= 1990 (4) R c = β LEGALc + β RIGHTSc + β GDPc + α fyeyear + i, where LEGAL is equal o one if he counry is classified as having a code law legal origin (La Pora e al. 1998). RIGHTS is equal o he invesor righs index developed by La Pora e al. (1998), where counries receive one poin for each of he following, allowing voing by mail, he requiremen of invesors o deposi heir shares prior o shareholder meeings, if cumulaive voing or proporional represenaion of minoriy shareholder on he board is allowed, if here are mechanisms in place o for oppressed minoriy shareholders, he minimum ownership required o call an exraordinary shareholder meeing, and if shareholders have preempive righs. GDP is equal o he log of he per capia gross domesic produc for he counry year. YEAR is equal o a series of fiscal year fixed effecs. The firs hree columns in Panel B of Table 1 display he resuls of esimaing parial forms of equaion (4) where only one insiuional variable and YEAR are included in he model due o he high correlaion across he insiuional variables. Considered in isolaion, we find ha counries having code law legal regimes and lower levels of invesor righs have higher synchroniciy values. However, when we esimae equaion (4) wih all hree insiuional feaures (he resuls of which are repored in column 4 of Panel B), we find only a significan posiive coefficien on LEGAL. While, he counry level resuls are consisen wih he resuls presened in Morck e al. (000), a counry-level analysis does no differeniae beween differences in R values across firms being due o firms sock prices reflecing relaively more informaion abou firm fundamenals or differences in R values across firms being due o non-fundamenal facors resuling in greaer noise in reurns. In he nex secion, we explore he exen o which he 14 All regressions are esimaed including fixed-year effecs, resuling in differen inercep values for each fiscal year. We do no able he fiscal year inerceps, which in general are significan a convenional levels. 14

17 synchroniciy measure is associaed wih facors ha represen firm informaion flows and fundamenals in five of he larges inernaional equiy markes. 4. Wihin-counry Analysis of he Synchroniciy Measure 4.1 SAMPLE AND DESCRIPTIVE STATISTICS Panel A of Table presens he descripive saisics on he R measure esimaed using all firm-year observaions over for Ausralia, France, Germany, Japan, he U.K., and he U.S. ha have he necessary daa o conduc our empirical ess. Our empirical ess require reurns, colleced from Daasream, and accouning daa, colleced from Worldscope. Firm-year observaions meeing he daa requiremens resul in sample sizes of,895, 5,368, 3,515, 3,58, 14,48, and 56,95 for Ausralia, France, Germany, Japan, he U.K., and he U.S., respecively. The sample sizes repored in Table (e.g., Ausralia n=,895) are smaller han hose repored in Table 1 (e.g., Ausralia n=,895) due o he addiional daa required for our ess. Japanese firms have he highes R values (mean=0.319, median=0.98) whereas he U.S. has he lowes R values (mean=0.118, median=0.067). 4. EMPIRICAL TESTS Our firs analysis examines he associaion beween he synchroniciy measure and accouning measures of sock price informaiveness. Earnings are one of he primary sources of firm-specific informaion, and differences in he amoun of earnings informaion refleced in sock prices is one poenial reason for differences in R values across firms. Collins, Kohari, Shanken, and Sloan (1994), Gelb and Zarowin (00), and Lundholm and Myers (00) use he amoun of informaion abou curren and fuure changes in earnings refleced in reurns as a measure of price informaiveness. Durnev e al. (003) use his definiion of price informaiveness o draw inferences on wheher firm-specific sock price movemens in he U.S. marke reflec firm-specific informaion or increased noise in reurns. They find lower R measures are associaed wih more price informaiveness. We es he associaion beween he R 15

18 measure and he amoun of informaion relaed o changes in curren and fuure earnings refleced in sock prices by esimaing he following regression equaion: (5) ABRET = β E 1 * * + β RR 6 + β E α fye fye= 1990 RR YEAR + β E + ε + β E RR + β ABRET where ABRET is he marke adjused buy and hold reurn over fiscal year. RR is equal o he decile rank of he R value, deermined by ranking observaions each year based on he R value wihin each of he five counries. E is equal o he change in ne income before exraordinary iems scaled by beginning of period marke value of equiy over fiscal year. YEAR is equal o a series of fiscal year fixed effecs. 15 The ABRET +1 erm is included in he model o correc for he errors in variables problem idenified by Collins e al. (1994). Collins e al. (1994) noe ha he correc specificaion of equaion (5) would include he expeced change in fuure periods earnings. Since expecaions are unobservable he acual changes in fuure periods earnings is used, inroducing an errors in variables problem which hey demonsrae can be correced by including nex period s reurn in he model. Given he resuls of Durnev e al. (003) in he U.S. we expec β and β 4 o be negaive and significan if lower R values are associaed wih more informaion abou curren and fuure changes in earnings being refleced in reurns. Table 3 displays he resuls of esimaing equaion (5). In all counries, we find a posiive and significan coefficien on he curren change in earnings. The resuls are more mixed wih respec o he change in fuure earnings. In Ausralia, Japan and he U.K., he coefficien on he change in fuure earnings is posiive and saisically significan whereas in France and Germany, however, he coefficien on he change in fuure earnings is negaive and significan. Turning o he primary variables of ineres, we find lile evidence ha he R measure consisenly reflecs firm-specific informaion being impounded ino share prices. France and he U.S. are he only counries where he R measure is associaed wih he curren change in earnings and reurns. 15 All analyses are esimaed including year fixed effecs. For breviy, we do no able he year inercep values. 16

19 However, inconsisen wih expecaions, higher R values in France and he U.S. are associaed wih more informaion abou he curren change in earnings being refleced in sock prices. When examining he resuls of ineracing fuure earnings changes wih he R measure, we find he ineracion o be saisically significan in Ausralia (β 4 =-0.00, p-value=0.01) indicaing ha lower R values are associaed wih more informaion abou fuure earnings changes being refleced in reurns. In Germany, Japan, and he U.S., however, he ineracion is posiive and significan a he 0.06 level or beer indicaing ha higher R values are associaed wih more informaion abou fuure earnings changes being priced. 16 Overall he resuls presened in Table 3 indicae ha lower R values are no associaed wih more earnings informaion being refleced in reurns. Raher, we find higher R values are associaed wih more earnings informaion being priced in France, Germany, Japan, and he U.S., which is opposie of wha is expeced if differences in he R values reflec differences in he amoun of firm-specific informaion refleced in reurns. 17 Our second analysis examines he associaion beween he synchroniciy measure and analys forecas errors. Our inquiry is moivaed by prior inernaional and U.S. research examining he properies of analys forecas errors. In general, his lieraure finds ha beer informaion in he form of addiional firm disclosures is associaed wih lower forecas errors (Lang and Lundholm 1996, Ashbaugh and Pincus 001, Hope 003, Lang e al. 003). Following his line of lieraure, we expec a posiive relaion beween analyss forecas errors and he synchroniciy measure if he synchroniciy measure reflecs informaion. We es he associaion beween firms R values and forecas errors using he following equaion: 16 We repea he analysis presened in Table 3 using he unranked R values in each of he five counries, he resuls of his analysis are similar o hose presened in Table Our findings are consisen, in par, wih Wes s (1988) model. Wes (1988) claims ha lower firmspecific reurn volailiy, higher R s, is associaed wih more informaion abou firm fundamenals being refleced in sock prices. 17

20 00 (6) F _ ERROR = β RR + α fyeyear + ε 1 fye= 1990 where RR is equal o he decile rank of he R value for fiscal year. F_ERROR is equal o he decile rank of EPS ac EPS forecas / EPS forecas for fiscal year, where EPS ac is he firm s acual earnings per share and EPS forecas is he mean consensus earnings per share forecas. Analys forecass are provided by IBES. YEAR is equal o a series of fiscal year fixed effecs. 18 We esimae equaions (6) wihou conrols due o he fac ha Morck e al. (000) argue ha he R measure is a summary measure of informaion and hus he inclusion of conrols is no appropriae as he poenial effecs of he conrol variables are capured by he R measure. Table 4 presens he resuls from esimaing equaion (6). In Japan, he coefficien on he RR erm is posiive and significan a he 0.00 level. This indicaes ha in Japan, consisen wih he informaion-based explanaion for he synchroniciy measure, lower R values are associaed wih lower analys forecas errors. In Ausralia, France, Germany, he U.K., and he U.S., however, we find he coefficien on he RR erm o be negaive and significan a he 0.01 level or beer. These resuls are, in general, inconsisen wih he R measure reflecing differences in he relaive amoun of firm-specific informaion. Thus, in he majoriy of our sample counries, we find ha higher R values are associaed wih lower analys forecas errors, which is opposie of wha is expeced if lower synchroniciy measures reflec relaively more firm-specific informaion. Our hird analysis invesigaes wheher here is a change in firms synchroniciy measures afer cross lising in he U.S. Cross lising in he U.S. represens a significan informaion even as a U.S. lising subjecs firms o increased regulaion and disclosure requiremens ha resul in more informaion abou he firm being made available o invesors (Ashbaugh 001, Lang e al. 003). Furhermore, cross lising in he U.S. increases firm visibiliy, increasing he invesor base and subsequen informaion search by invesors (see Karolyi (004) for an overview). If 18 The sample sizes in he analys forecas errors es are furher reduced due o he requiremen ha firms be followed by an analys. 18

21 synchroniciy is a funcion of firm-specific informaion, i follows ha firms R values are expeced o decrease afer hey cross lis in he U.S. Table 5 presens he resuls of he cross lising analysis where we define he change in R values as he R value in he 1 monhs following he cross lising monh minus he R value in he 1 monhs preceding he cross lising monh. 19 Panel A of Table 5 presens he mean and median change in R measure ( R ) for all cross lisings over he ime-period. The R means in Ausralia (-0.00) and Japan (-0.003) are direcionally consisen wih he predicions, bu are no saisically differen from zero. In France and he U.K., however, he mean change in he R measures (0.080 and 0.055, respecively) are posiive and significanly differen from zero (a he 0.0 level or beer). We find similar resuls when examining he median R, where saisical significance is assessed using he Wilcoxon rank sum es. Panel B of Table 5 examines only he Level and 3 ADR s since hese ypes of cross lising are associaed wih he greaes informaion disclosures. The resuls presened in Panel B of Table 5 are similar o he resuls for cross lisings as a whole. We find ha in France and he U.K. he mean and median R are posiive and saisically differen from zero a he 0.0 level or beer. None of he oher R measures are significanly differen from zero. Overall, he resuls presened in Table 5 indicae ha cross lising in he U.S. is no associaed wih a decline in he R values as one would expec under an informaion-based inerpreaion of he synchroniciy measure. 0 Our las analysis uses he framework of Pioroski and Roulsone (004) o es he exen o which firm-specific informaion and firm fundamenals are relaed o he R measure in 19 Cross lising daes and he ype of cross lising (e.g. Level 1,, 3 or Rule 144A) are provided by J.P. Morgan Chase & Co. 0 We conduc wo addiional sensiiviy ess. Firs, we repea he cross lising analysis esimaing equaion () wihou he U.S. reurn due o he poenial mechanical effec ha cross lising may have on he coefficien on U.S. reurn. Second, we esimae an OLS fixed effecs model for each counry using a dummy variable o capure he pos period. We draw similar inferences from he resuls of hese wo sensiiviy ess. 19

22 inernaional markes. Pioroski and Roulsone (004) use U.S. firms R values as a benchmark of firm-specific informaion incorporaed ino prices, and es he associaion beween R values and variables proxying for firms informaion environmen. Based on heir work, we esimae he following fixed effecs model: SYNCH (7) = β LOSS β STDROA α fye fye= β R & D + β REG 7 YEAR + ε i + β ANALYST 3 + β RELSIZE 8 + β % CLHLD 4 + β % MVE 9 + β STDSALES + + β % TURNOVER where SYNCH is equal o log(r /(1-R )) for fiscal year ; LOSS is equal o one if ne income before exraordinary iems is negaive, and zero oherwise; R&D is equal o one if he firm repors a value for research and developmen expense, and zero oherwise; ANALYST is equal o he log of one plus he number of analyss making a forecas for fiscal year s earnings; %CLHLD is he proporion of shares ha are closely held as of he end of he fiscal year ; STDSALES is he sandard deviaion of sales scaled by oal asses over calculaed requiring a minimum of hree and maximum of five fiscal years; STDROA is he sandard deviaion of ROA calculaed requiring a minimum of hree and maximum of five fiscal years where ROA is equal ne income before exraordinary iems divided by fiscal year end oal asses; REG is equal o one if he firm is a financial insiuion or uiliy; RELSIZE is he firm s sales divided oal sales of is primary indusry (-digi SIC code); MVE is defined as he naural log of fiscal year end marke value of equiy; TURNOVER is he average weekly urnover (number of shares raded divided by number of shares ousanding) over he fiscal year; and YEAR is equal o a series of fiscal year fixed effecs. The dependen variable in equaion (7), SYNCH, is he R measure ransformed o creae coninuous variable ha yields a dependen variable wih a more normal disribuion han he disribuion of R values ha are bounded by zero and one (Morck e al. 000, Pioroski and Roulsone 004). We use four variables o proxy for firms informaion environmen (LOSS, R&D, ANALYST, %CLHLD) and wo variables o proxy for firm fundamenals (STDSALES, STDROA). LOSS is included in he model as he reporing of losses is a news even expeced o be refleced in reurns (Joos and Plesko 005, Hayn 1995). Likewise, he reporing of research and developmen expendiures is also considered o be a news even refleced in reurns (Aboody and Lev 000). R&D is an indicaor variable idenifying wheher he firm discloses research and developmen expendiures. Research and developmen is a highly firm-specific invesmen and 0

23 he disclosure of research and developmen expendiures in many counries is a volunary disclosure over our analysis period. The number of analyss following he firm, ANALYST, is included in he model as a proxy for he firms informaion environmen because higher analys following is associaed wih richer informaion environmens (Lang and Lundholm 1996, Bushman, Pioroski, and Smih 005). If firms R values serve as a measure of firm-specific informaion incorporaed ino prices, we expec negaive coefficiens on LOSS, R&D, and ANALYST. We use he percen of shares ha are closely shares, %CLHLD, o proxy for insider ownership (Himmelberg, Hubbard, and Love 00, Lins and Warnock 004). Greaer insider ownership will resul in lower R values when insiders are able o gaher and rade on privae informaion abou firm fundamenals (Roll 1988). Alernaively, greaer insider ownership may resul in higher R values if insiders manipulae financial informaion ransparency for he purpose of hiding heir wealh exracion or if insiders are par of group ownership where group ownership and he coordinaed acions wihin he group, such as financing oher firms in he group, resuls in a common componen o firm s fundamenals. Given he compeing explanaions, we make no predicion on he relaion beween %CLHLD and SYNCH. The sandard deviaion of sales (STDSALES) and he sandard deviaion of reurn-onasses (STDROA) are included in he model o capure he volailiy of firm fundamenals. One poenial reason for high firm-specific reurn volailiy is he volailiy of underlying fundamenals. Wei and Zhang (004) find ha wihin he U.S., greaer volailiy in firms reurn on equiy is associaed wih increased reurn volailiy. We include boh he volailiy of reurnon-asses and sales due o differences in income smoohing inernaionally and he poenial influence of income smoohing on reurn-on-asses (Leuz, Nanda, and Wysocki 003). If firms R values reflec firm fundamenals being incorporaed ino prices, we expec negaive coefficiens on STDSALES and STDROA. 1

24 The remaining variables in equaion (7) (REG, RELSIZE, MVE, and TURNOVER) serve as conrol variables. REG is used o conrol for he fac ha all firms operaing in a regulaed indusry face similar consrains due o regulaion, and hus, heir prices are expeced o have high sock price synchroniciy (Pioroski and Roulsone 004). RELSIZE is used o conrol for a firm s indusry presence. The larger marke share ha a firm has in is indusry he more likely he firm s sock price drives indusry reurns. Therefore we expec a posiive relaion beween RELSIZE and SYNCH. We include MVE o conrol for firm size. Larger firms are generally associaed wih richer informaion environmens indicaing in a negaive associaion beween firm size and R values. However, larger firms also poenially have more diversified operaions resuling in hese firms rading more in line wih he marke resuling in a posiive associaion beween firm size and he R measure. We include TURNOVER in he model o capure he level of rading in a firm shares. Under he informaion based inerpreaion of he R measure, he associaion beween he R measure and TURNOVER would be negaive as more rading represens increased informaion being impounded ino firms share prices. However if one assumes he R measure proxies for noise rading, rading unrelaed o fundamenals, hen he associaion beween he R values and TURNOVER is expeced o be posiive. Given he uncerainy, we make no predicion of he sign of he coefficiens on MVE and TURNOVER. Panel A of Table 6 displays he Pearson correlaions beween synchroniciy and he independen variables of equaion (7). In general, here is quie a bi of variaion across counries in he sign and significance of he correlaions beween firms R values and he variables proxying for firms informaion flows and fundamenals. In conras, he correlaions beween he synchroniciy measure and he conrol variables drawn from prior research are more consisen across counries. Panel B of Table 6 presens he resul of esimaing equaion (7) wihin each of our six counries. In general, he explanaory power of he model is similar across counries ranging

25 from a low of 14% in he U.S. o a high of 5% in Germany. The las wo columns in Table 6 summarize he signs and significance of he esimaed coefficiens on he independen variables. The resuls indicae ha in Japan, he U.K. and he U.S. LOSS is significanly associaed wih SYNCH. However, in Japan and he U.K. he coefficien on LOSS is significanly posiive, which is conrary o wha is prediced if he R measure reflecs firm-specific informaion. Likewise, in four counries he coefficien on R&D is significanly posiive, which is opposie of wha is prediced. The resuls are mixed on ANALYST. As prediced he coefficien on ANALYST is negaive and significan in he U.K. In conras, he coefficien on ANALYST is posiive and significan in he oher five counries. These resuls indicae ha he R measure is no consisenly associaed wih variables ha prior research has used o proxy for firms public informaion environmen. Focusing on he impac of insider ownership on sock price synchroniciy, he resuls indicae a significan negaive associaion beween %CLHLD and SYNCH in all five counries in which we have closely held ownership daa. This finding suppors he claim ha insiders in hese markes are able o gaher privae informaion abou firm fundamenals (Roll 1988). Turning o he specific variables represening firm fundamenals, we find inconsisen relaions beween STDSALES and SYNCH across our sample counries. Specifically, he coefficien on STDSALES is significanly posiive in Ausralia, France, Germany, and he U.S., and marginally significan and negaive in he U.K. The coefficien on STDROA is only significan in wo counries, Germany and Japan, and he sign is posiive raher han negaive as prediced if he R measure represens firm-specific informaion being impounded in share prices. The inconsisen resuls relaed o informaion environmen variables and he volailiy of firm fundamenals is paricularly roubling for an informaion-based explanaion of he R measure across counries. Summarizing he findings on he conrol variables, we find mixed resuls on REG and RELSIZE is posiively associaed wih SYNCH in only Ausralia. We find a posiive and significan associaion MVE and SYNCH in all counries, which is consisen wih he resuls of 3