Essay 3. The liquidity impact of open market share repurchases * repurchases *

Transcription

1 Essay 3 The liquidiy impac of open marke share repurchases * repurchases * * The essay is co-auhored wih Adri De Ridder. We hank William Bradford, David Burnie, Marin Holmén and Holmén and seminar paricipans a he Financial Managemen Associaion annual meeing 2008, he Midwes Finance Associaion annual meeing 2007, and he Easern Finance Associaion annual meeing 2006, for helpful commens. Financial suppor from he NASDAQ OMX Nordic Foundaion is graefully acknowledged.

2 1. Inroducion Share repurchases have become a popular means of reurning cash o shareholders for many companies around he world. 1 The predominan mehod o repurchase shares in he US, and he res of he world, is hrough an open marke share repurchase program in which he company repurchases is own shares in he marke (Vermaelen, 2005). A fundamenal issue concerning open marke share repurchases is wheher he firm s repurchase rading and is presence in he marke as an insider has an impac on sock liquidiy. The liquidiy issue is imporan since a change in sock liquidiy affecs shareholders ransacion coss and heir required rae of reurn, which in urn affecs he firm s cos of capial. 2 The liquidiy aspec of share repurchases also seems o be of concern o firms managers. In a survey of payou policies among US companies, Brav e al. (2005) find ha more han 50% of execuives feel ha he liquidiy of heir sock is an imporan or very imporan facor affecing heir repurchase decisions. The exising lieraure argues ha open marke share repurchases can boh improve and deeriorae sock liquidiy. Barclay and Smih (1988) sugges ha repurchasing firms can, by placing limi buy orders, esablish a lower bound on he bid price and in so doing reduce he bid-ask spread. Barclay and Smih (1988) also posi a counerargumen such ha open marke share repurchase programs may widen bid-ask spreads if he managers of repurchasing firm possess informaion ha ouside marke paricipans do no have, as he rading analyses of Jaffe (1974) and Seyhun (1986) sugges. This asymmeric informaion is hough o give rise o an adverse selecion cos in he form of reduced sock liquidiy (Copeland and Gala 1983; Glosen and Milgrom, 1985). In many counries, including he US, deailed sudies of repurchase rading are difficul o conduc, as firms are no required o disclose he precise daes, prices and magniudes of heir repurchase ransacions. However, in Sweden and some oher counries, firms are required o disclose full deails of heir repurchase aciviies on a daily basis, which provides an opporuniy o sudy he liquidiy impac of repurchase rading. 3 There are few prior sudies ha examine he liquidiy impac of open marke share repurchases on he acual repurchase days. Brockman and Chung (2001) and Ginglinger and 1 Prior o 1980, buyback aciviy was illegal or else discouraged by ax laws in all counries excep for in US. 2 Amihud and Mendelson (1986) develop and es a model ha shows ha wider relaive bid-ask spreads lead o higher risk-adjused sock reurns and, hence, increased cos of capial. 3 Oher counries where firms have o disclose heir repurchase aciviy on a daily basis include, for example, Ausralia, Hong-Kong, Malaysia, Mexico, Singapore and he UK (Vermaelen, 2005). 2

3 Hamon (2007) sudy he liquidiy effecs of repurchase rading on he Sock Exchange of Hong-Kong and he Paris Sock Exchange, respecively. Boh sudies find ha repurchase rading has a derimenal effec on sock liquidiy. Conversely, Cook e al. (2004) find ha repurchase rading conribues o sock liquidiy for a sample of NASDAQ and NYSE lised firms. Given he few prior sudies ha examine he liquidiy impac of acual repurchase ransacions and heir conflicing resuls, more empirical sudies are warraned. The objecive of his sudy is o furher analyze any liquidiy impac ha open marke share repurchases may have on he acual repurchase days in an elecronic limi order marke, which is he marke sysem ha mos sock exchanges around he world use. 4 Specifically, we analyze average inraday bid-ask spreads and order dephs on repurchase days and compare hem wih spreads and dephs on surrounding non-repurchase days for repurchasing firms lised on he Sockholm Sock Exchange (SSE). 5 We conribue o prior research by spliing he sample ino repurchase days condiional upon daily repurchase sizes and repurchases execued as block rades. 6 We pariion he sample in his way because any execuion-relaed liquidiy effecs of open marke share repurchases may be dependen upon repurchase volume and because repurchases execued ouside he order sysem in he form of block rades may have a differen impac on sock liquidiy compared o repurchases execued inside he order sysem. Several key resuls emerge from our analysis. Firs, we find significanly narrower bid-ask spreads and deeper marke dephs on repurchase days relaive o surrounding non-repurchase days when repurchases are execued inside he order rading sysem. Second, afer conrolling for oal rading volume, price and reurn volailiy, we sill find significanly narrower bid-ask spreads on repurchase days relaive o surrounding non-repurchase days. Third, he reducion in he bid-ask spread on repurchase days is posiively associaed wih he daily repurchase size measured as he daily repurchase volume divided wih he average daily rading volume during he 4 weeks preceding he week of he repurchase. Fourh, repurchase rades execued ouside he order rading sysem as block ransacions have a derimenal effec on he bid-ask spread, consisen wih a negaive response o he presence of informed managerial rading. Based on our resuls, we conclude ha repurchase rading is liquidiy enhancing when repurchase rades are execued inside he order rading sysem. Our resuls are consisen wih Cook e al. (2004), who find ha ha repurchase rading conribues o sock liquidiy for 4 In a limi order marke, invesors esablish bid-ask prices by placing limi orders. 5 The official name of he Sockholm Sock Exchange is NASDAQ OMX Nordic Exchange Sockholm. 6 Brockman and Chung (2001) also es for he effec of he repurchasing firm s rade size bu no for repurchases execued as block rades. 3

4 NYSE- and NASDAQ-lised firms. Ineresingly, we find no evidence ha share repurchases execued inside an order-driven rading sysem impose a cos in he form of lower sock liquidiy on he acual repurchase days, as found by Brockman and Chung (2001) and Ginglinger and Hamon (2007) order-driven markes similar o he SSE. The paper is organized as follows. In Secion 2 we summarize he prior empirical research on he marke liquidiy effecs relaed o open marke share repurchases and develop our hypoheses. Secion 3 describes he regulaory environmen for open marke share repurchases in Sweden and he rading srucure on he SSE. Secion 4 presens he daa, he sample and marke liquidiy measures. Secion 5 repors he empirical resuls and Secion 6 concludes. 2. Previous research and marke liquidiy hypoheses 2.1. Previous research The firs sudy ha examines he marke liquidiy impac of open marke share repurchases is Barclay and Smih (1988). The auhors develop wo non-muually exclusive hypoheses and examine companies lised on he NYSE. The compeing marke-maker hypohesis predics ha he bid-ask spread will narrow if firms submi buy limi orders ha esablish a lower bound on he bid price. In conras, he informaion-asymmery hypohesis predics ha he bid-ask spread will widen if managers are beer informed and willing o rade on inside informaion. 7 Barclay and Smih (1988) es he wo hypoheses empirically by examining relaive bid-ask spreads before and afer repurchase announcemens. 8 Consisen wih he informaion-asymmery hypohesis, hey find ha relaive bid-ask spreads widen following announcemens. 9 The auhors conclude ha he presence of beer informed managers in he marke increases he bid-ask spread, and hereby increases he repurchasing firm s cos of capial. Singh e al. (1994), Wiggins (1994), and Miller and McConnell (1995), all refine and exend Barclay and Smih s (1988) sudy. However, in conras o Barclay and Smih (1988), hey find no suppor for he informaion-asymmery hypohesis. 7 Insider rading analyses by Jaffe (1974) and Seyhun (1986), among ohers, show ha managers can idenify mis-pricings in heir own firms. Marke microsrucure models by Copeland and Galai (1983), and Glosen and Milgrom (1985), show ha greaer informaion asymmery beween informed and uninformed raders deerioraes marke liquidiy. 8 The relaive bid-ask spread is he dollar spread divided by he average of he bid and ask prices. 9 Since he informaion-asymmery hypohesis and he compeing marke-maker hypohesis are no muually exclusive, heir resuls imply ha he informaion-asymmery effec dominaes he compeing marke-maker effec. 4

5 Franz e al. (1995) develop anoher hypohesis in which hey hypohesize ha he bid-ask spread will narrow afer a repurchase announcemen if he announcemen signals managers privae informaion abou he fuure prospecs for he firm. They es he hypohesis by examining bid-ask spreads around repurchase announcemens by NASDAQ-lised firms. Consisen wih heir hypohesis, hey find narrower bid-ask spreads following announcemens afer conrolling for invenory-holding and order processing coss. Franz e al. (1995) aribue he narrower bid-ask spreads o a reducion in he informed rading coss associaed wih he repurchase announcemen. More recen sudies examine sock liquidiy effecs on he acual repurchase days. Brockman and Chung (2001) refine and es Barclay and Smih s (1988) hypoheses on he elecronic order-driven Sock Exchange of Hong Kong. 10 They examine bid-ask spreads and marke dephs on repurchase and disclosure days and find ha bid-ask spreads widen and marke dephs deeriorae on repurchase days afer conrolling for oal rading volume, price and reurn volailiy. Brockman and Chung (2001) conclude ha sock liquidiy deerioraes on repurchase days because marke paricipans deec he presence of informed rading and parially - or compleely - wihdraw from he marke. Cook e al. (2004) examine bid-ask spreads around repurchase days for NYSE- and NASDAQ-lised firms. Since deails of repurchase ransacions are no publicly disclosed in he US, hey use volunarily disclosed repurchase rading daa from 64 firms. Conrary o he resuls of Brockman and Chung (2001), hey find narrower bid-ask spreads on repurchase days relaive o surrounding benchmark periods. 11 Cook e al. (2004) argue ha he spread narrowing is consisen wih he noion ha, a he ime of a repurchase ransacion, he firm compees wih marke makers o provide liquidiy on he bid side of he marke. Ginglinger and Hamon (2007) analyze he sock liquidiy impac of share repurchases for a sample of 352 French firms lised on he elecronic order-driven Paris Sock Exchange. 12 Consisen wih Brockman and Chung (2001), hey find ha bid-ask spreads widen and ha heir deph decreases on repurchase days. However, in conras o Brockman and Chung (2001), hey argue ha he adverse effec on sock liquidiy is aribued o managers rade agains he rend o suppor he firm s share price in a depressed marke, wih he rades aking place a he ask price. 10 Firms lised on he Sock Exchange of Hong Kong are required o repor deails of repurchase ransacions o he Exchange no laer han 09:30 a.m. he following business day. The informaion is hen aggregaed by he Exchange and disseminaed o daa vendors (Vermaelen, 2005). 11 In conras o Brockman and Chung (2001), Cook e al. (2004) only analyze bid-ask spreads in a univariae seing and do no conrol for volume, price and reurn volailiy. 12 Firms lised on he Paris Sock Exchange are required o publicly repor, on a monhly basis, he oal number of shares purchased or sold during he previous monh (Ginglinger and Hamon, 2007). 5

6 2.2. Hypoheses of he effec of repurchase rading on sock liquidiy Barclay and Smih (1988), as well as laer sudies, have argued ha open marke share repurchases can have an incremenal as well as a derimenal effec on sock liquidiy. We es wo non-muually exclusive hypoheses of he effec of repurchase rading on sock liquidiy based upon he hypoheses developed by Barclay and Smih (1988). Our firs hypohesis is execuion-relaed, and predics ha he bid-ask spread will narrow on repurchase days when firms submi buy limi orders a he bid or wihin he bid-ask spread in he order book. The repurchase orders esablish a lower bound on he bid price, which will narrow he bid-ask spread. We predic a posiive associaion beween he bid-ask narrowing and he daily repurchase size. Conrary o our firs hypohesis, our second hypohesis implies ha he bidask spread will widen on repurchase days. This could be due o an asymmeric informaion effec in which marke paricipans deec he firm s repurchase rades and suspec informed rading. 13 Alernaively, i could be due o an execuion-relaed effec in which he firm places limi orders a he ask price o a large exen and hereby consumes liquidiy on he sell-side. We predic a posiive associaion beween he bid-ask widening and he daily repurchase size. 3. The Swedish regulaory environmen for open marke share repurchases and rading srucure on he Sockholm Sock Exchange 3.1. Regulaory environmen for open marke share repurchases In Sweden, resricions on share repurchases were removed on March 10, 2000, when changes in he Companies Ac made i possible for public limied companies o repurchase heir own shares and keep hem as reasury sock. 14 A decision o repurchase shares mus be approved by he shareholders a a general meeing. However, he shareholder meeing can auhorize he Board of Direcors o make such a decision. 15 The auhorizaion o repurchase shares may no 13 Marke microsrucure models by Copeland and Galai (1983), and Glosen and Milgrom (1985), show ha he presence of informed raders wih superior informaion abou he fuure sock price leads o wider bid-ask spreads on a dealer-based marke. Brockman and Chung (2000) find ha limi-order raders on he elecronic order-driven Sock Exchange of Hong Kong behave like designaed dealers/specialiss when confroned wih he possibiliy of rading agains informed invesors. 14 Before March 10, 2000, i was possible for firms o redeem ousanding shares in a relaively slow and cumbersome adminisraive procedure (redempion of shares). 15 The repurchase decision is valid only if approved by shareholders holding a leas wo-hirds of he voes cas and he shares represened a he meeing. In general, he shareholder meeing auhorizes he Board of Direcors o make decisions abou share repurchases. 6

7 las longer han unil he nex annual general meeing and he maximum amoun of reasury sock he company is allowed o keep a any ime is limied o 10% of he oal number of shares in he company. Companies may buy back heir own sock on an auhorized marke or anoher regulaed marke wihin he European Economic Area or hrough a ender offer direced o all shareholders. 16 Companies repurchasing heir own shares on he SSE mus comply wih he Exchange s rules regarding share repurchases. 17 The company mus repor o he Exchange all repurchase ransacions as soon as possible, bu no laer han 30 minues before he sock exchange opens on he rading day immediaely following he repurchase day. Wih he excepion of block rades, he company may no purchase more han 25% of he average daily urnover during he prior 4 calendar weeks. The company may only place orders in he order sysem or close block ransacions in he company s own shares wihin he bid-ask spread applying on he SSE (he bes bid and ask included) Trading srucure on he Sockholm Sock Exchange Trading on he SSE is conduced in an elecronic order-driven rading sysem. Traders submi orders in he rading sysem hrough Exchange members. There are no designaed marke makers, alhough some smaller firms engage liquidiy providers. Buy and sell orders enered ino he rading sysem are auomaically mached. Limi orders are firs prioriized by price and hen by ime. If a rader requires immediae execuion, he rader mus submi a buy (sell) limi order ha his he curren bes ask (bid) price. Traders can submi hidden limi orders, where only a porion of he order volume is displayed in he limi order book. However, he hidden porion has lower prioriy han he displayed limi orders wih he same price. Exchange members can choose beween rading on-exchange or ouside he sock exchange. On-exchange, he Exchange member can eiher make rades in he elecronic order-driven rading sysem or ouside he order book as a manual rade. In boh cases, he rades mus comply wih he requiremens of he exchange. Manual rades ha ake place during he rading hours of he Exchange mus be repored as close o real-ime as possible, bu no laer 16 Companies may also repurchase heir own sock on an auhorized marke or anoher regulaed marke ouside he European Economic Area afer approval from he Swedish Financial Supervisory Auhoriy (Finansinspekionen). 17 Rules regarding purchase and sale of a company s own shares in he Rule Book for Issuers. 7

8 han 5 minues from he ime of he rade. 18 Manual rades ha ake place afer rading hours mus be repored prior o he opening of he Exchange he following rading day. The SSE offers a high degree of ransparency. Traders observe, in real-ime, bid and ask prices wih corresponding deph, concluded ransacions, and he ideniies of Exchange members behind execued ransacions. 19 The ick size depends upon he price level of he securiy. During he sample period he ick size varied from SEK 0.01 o SEK Daa, sample and marke liquidiy measures 4.1. Daa and sample NASDAQ OMX Sockholm provides daa as o all of he repurchases execued on he exchange over he period 2002 o The daa includes he name of he repurchasing firm, he repurchase dae, he oal number of shares repurchased during he day disribued by class of share, he average repurchase price, and he oal price of repurchased shares. Over he sample period, 53 Swedish firms repurchased heir own shares on he SSE. From NASDAQ OMX Sockholm, we also receive inraday bid-ask quoes wih corresponding marke dephs (samped 15 minues apar) and daily rading volumes execued inside and ouside he elecronic rading sysem over he sample period May 16, 2002, o December 30, Our iniial sample consiss of 3,039 firm repurchase days. We exclude 79 firm repurchase days on he same day or on he rading day following he release of inerim or annual repors. We lack inraday daa for 4 rading days and, herefore, we have o exclude 8 firm repurchase days. As some firms have repurchased dual class shares (A and B shares), we exclude 62 firm repurchase days and only include share repurchases in he share class wih he highes urnover. Finally, due o violaion of rading and disclosure rules by 4 firms, we exclude 84 firm repurchase days. Our final sample consiss of 53 firms and 2,806 firm repurchase days. 18 New rules by he SSE afer he sample period sae ha manual rades ha ake place during he rading hours mus be repored immediaely and no laer han 3 minues from he ime of he rade. 19 The ideniy of he exchange member (broker) behind a placed order was also shown o he exchange members during he sample period, bu is no longer displayed. 8

9 4.2. Liquidiy measures In an order-driven marke, limi orders provide marke liquidiy and esablish bid-ask spreads and marke deph. The bid-ask spread represens he price dimension and he deph represens he quaniy dimension of marke liquidiy. Lee e al. (1993) argue ha boh dimensions are required o make inferences abou marke liquidiy. We use wo spread measures and hree deph measures in our sudy. The wo spread measures we use are Absolue Spread and Relaive Spread. Absolue Spread is he daily average of absolue bid-ask spreads in SEK recorded 15 minues apar over a rading day. The absolue bid-ask spread is he difference beween he buy limi order wih he highes price and he sell limi order wih he lowes price. Relaive Spread is he daily average of relaive bid-ask spreads recorded 15 minues apar over a rading day. The relaive bid-ask spread is he absolue bid-ask spread divided by he bid-ask midpoin. The hree deph measures we use are Toal Deph, Bid Deph and Ask Deph. Toal Deph is he daily average value in SEK of all shares posed a he highes bid and lowes ask price recorded 15 minues apar over a rading day. Bid Deph is he daily average value in SEK of all shares posed a he highes bid price recorded 15 minues apar over a rading day. Finally, Ask Deph is he daily average value in SEK of all shares posed a he lowes ask price recorded 15 minues apar over a rading day. Table 1 Summary saisics of open marke share repurchase aciviy The sample comprises 2806 open marke share repurchase days over he period May 16, 2002, o December 30, 2005, for 53 Swedish firms lised on he SSE. Panel A presens repurchase aciviy per firm and Panel B provides he repurchase frequency. Panel A: Repurchase aciviy per firm Mean Median Number of repurchase programs Number of repurchase days Number of shares repurchased on a repurchase day 262,738 65,587 Value of shares repurchased on a repurchase day (MSEK) Repurchase volume on a repurchase day as a percenage of he sock s oal rading volume ha day Panel B: Repurchase frequency Number Percen Firms wih 1 repurchase day Firms wih 2 o 10 repurchase days Firms wih 11 o 50 repurchase days Firms wih 51 o 100 repurchase days Firms wih over 100 repurchase days

10 5. Empirical resuls 5.1. Descripive saisics Table 1 presens summary saisics of repurchase aciviies for our sample of repurchasing firms. As can be seen in Panel A, he median repurchasing firm iniiaes 2 repurchase programs and repurchases shares on 17 rading days during he sample period. On average, he median firm repurchases 65,587 shares valued a SEK 4.98 million (approx. US$ 0.71 million) on each repurchase day. The average daily repurchase volume by repurchasing firms is subsanial. The median firm repurchases almos 41% of is sock s daily rading volume on a repurchase day. This repurchase volume is similar o he average repurchase volume repored by Brockman and Chung (2001) for firms lised on he Sock Exchange of Hong Kong, bu higher han he average repurchase volumes repored by Cook e al. (2004) and Ginglinger and Hamon (2007) for firms lised on he US markes and he Paris Sock Exchange, respecively. 20 Panel B in Table 1 repors he repurchase frequency over he sample period. Almos one hird of he repurchasing firms repurchase shares on beween 11 and 50 rading days. Approximaely 11% of he repurchasing firms repurchase shares on only 1 rading day, while 30% of he repurchasing firms repurchase shares on more han 50 rading days. Table 2 provides rading saisics for shares of repurchasing and non-repurchasing firms lised on he SSE over he sample period. 21 Repurchasing firms have a leas one rading day involving a repurchase over he sample period. Approximaely 17% of he Swedish firms lised on he SSE execued a leas one repurchase over he sample period. Repurchasing firms have significanly higher marke capializaion, higher daily sock urnover and rade sizes - bu lower urnover raes - compared o non-repurchasing firms. The percenage of rading days shares have been raded during a year and he number of daily rades do no differ significanly beween repurchasing and non-repurchasing firms. The mean (median) end-ofday relaive bid-ask spread is 1.60% (1.07%) for repurchasing firms and 3.19% (2.09%) for non-repurchasing firms. The difference ess indicae ha he mean and median differences in he relaive bid-ask spreads are saisically significan a he 1% level. 20 Brockman and Chung (2001) find ha he average buyback represens 44% of he sock s daily rading volume on he Sock Exchange of Hong Kong. Cook e al. (2004) find ha he average buyback represens 33% of he sock s daily rading volume on NASDAQ and 19% on he NYSE. Ginglinger and Hamon (2007) find ha he average repurchase accouns for 28% of he oal value of raded shares on he ransacion dae. 21 Trading saisics are only provided for he mos raded share class if firms have dual class shares. 10

11 Table 2 Trading saisics of sample companies Trading saisics for shares lised on he SSE over he sample period May 16, 2002, o December 30, Only he mos raded share class is included for firms wih dual class shares. Firms wih one or more open marke share repurchases during he sample period are classified as Repurchase firms and all oher firms are classified as Non-repurchase firms. The measures are average annual measures averaged across he sample years for each firm. Marke capializaion is he average end-of-year marke value of he firms lised shares during he sample period. The Relaive bid-ask spread is he absolue bid-ask spread a he end of he day divided by he bid-ask midpoin a he end of he day. The -saisics are for he difference in mean measures beween repurchase and non-repurchase firms. W-saisics are from he non-parameric Wilcoxon rank sum es. p-values from he difference ess are repored in brackes. Repurchase firms (n=53) Non-repurchase firms (n=267) Difference es Mean (Median) Mean (Median) -saisic [p-value] W-saisic [p-value] Marke capializaion (MSEK) 16,284) (2,725) 6,777) (698) 2.40 [0.017] 4.34 Daily urnover (TSEK) 61,659) (4,147) 32,728) (1,351) 1.09 [0.275] 2.27 [0.018] Trade size (SEK) 173,082) (117,422) 87,401) (48,227) Turnover rae per share (%) 58.66) (40.36) 80.76) (56.19) [0.069] [0.071] Percenage of rading days he shares have been raded 93.77) (99.60) 90.96) (98.41) 1.26 [0.210] 1.14 [0.254] Number of rades per rading day 151) (37) 128) (30) 0.37 [0.715] 0.31 [0.754] Relaive bid-ask spread (%) 1.60) (1.07) 3.19) (2.09) [0.010] [0.001] Table 3 Relaive bid-ask spread for shares lised on he Sockholm Sock Exchange Relaive bid-ask spreads for shares lised on he SSE over he period 2002 o Only he mos raded share class is included for firms wih dual class shares. Firms wih one or more open marke share repurchases during he sample period are classified as Repurchase firms and all oher firms are classified as Non-repurchase firms. The Relaive bid-ask spread is he average annual absolue bid-ask spread a he end of he day divided by he bid-ask midpoin a he end of he day in percen. Repurchase firms Relaive bid-ask spread Year Mean Median Non-repurchase firms Relaive bid-ask spread Number of firms Mean Median Number of firms

12 Table 3 provides yearly average end-of-day relaive bid-ask spreads over he sample period for shares of repurchasing and non-repurchasing firms lised on he SSE. The average end-ofday relaive bid-ask spread decreases monoonically each year during he sample period for boh repurchasing and non-repurchasing firms. The decreasing bid-ask spread suggess ha he overall marke liquidiy improved on he SSE over he sample period Univariae analysis of sock liquidiy around iniiaions of share repurchase programs The focus of his sudy is on he liquidiy impac of open marke share repurchases on he acual repurchase days. However, he bid-ask spread may widen a he iniiaion of a share repurchase program and remain a a wider level for he duraion of he program if marke paricipans suspec ha he firm has an informaional advanage and will rade on i (Barclay and Smih, 1988). Even if here is an execued-relaed emporary narrowing of he bid-ask spread on he repurchase days relaive o he surrounding non-repurchase days, he programlong bid-ask widening may sill dominae a liquidiy-enhancing execuion-relaed liquidiy effec and resul in a ne-widening of he bid-ask spread during a repurchase program. We herefore examine wheher here is a general change in sock liquidiy when repurchasing firms iniiae a repurchase program and ener he marke o repurchase shares. Since repurchasing firms on he SSE mus repor heir repurchase ransacions on a daily basis, marke paricipans know he exac dae when firms firs ener he marke o repurchase shares in a repurchase program. For each repurchase program in he sample, we esimae he average of he various liquidiy measures 20 rading days before and afer he firs repurchase day. The resuls are provided in Table 4. For he sample of 79 repurchase programs, we find a saisically significan narrowing of he bid-ask spread when firms ener he marke o repurchase shares. The average absolue bid-ask spread decreases approximaely 0.13 SEK and he average relaive bid-ask spread decreases from 2.08% o 1.71%. This change in he bid-ask spread is also likely o be economically significan o large invesors in he marke. In addiion, we find saisically significan increases in order dephs when repurchasing firms ener he marke. Thus, we do no find any evidence ha a liquidiy-deerioraing asymmeric informaion effec dominaes a liquidiy enhancing execuion-relaed liquidiy effec when firms ener he marke o repurchase shares. 12

13 Table 4 Univariae analysis of sock liquidiy around iniiaions of share repurchase programs The sample comprises 79 repurchase programs wih rading daa 20 rading days before and 20 rading days afer he firs repurchase in he program. The sample period is May 16, 2002, o December 30, For each repurchase program, we esimae he average of he various liquidiy measures 20 rading days before (Pre20) and 20 rading days afer (Pos20) he firs repurchase in he program. Absolue Spread is he 20 day average of he daily average of absolue bid-ask spreads in SEK recorded 15 minues apar over a rading day. Relaive Spread is he 20 day average of he daily average of relaive bid-ask spreads recorded 15 minues apar over a rading day. The relaive bid-ask spread is he absolue bid-ask spread divided by he bid-ask midpoin. Toal Deph is he 20 day average of he daily average value in SEK of all shares posed a he highes bid and lowes ask price recorded 15 minues apar over a rading day. Bid Deph is he 20 day average of he daily average value in SEK of all shares posed a he highes bid price recorded 15 minues apar over a rading day. Ask Deph is he 20 day average of he daily average value in SEK of all shares posed a he lowes ask price recorded 15 minues apar over a rading day. Paired -ess are used o deermine he significance of measure differences in he pre-even period and he pos-even period. Pre20 Pos20 Difference Difference es Variable Mean Mean Mean -saisic p-value Absolue Spread Relaive Spread Toal Deph 3,832,737 4,900,297 1,067,560 2, Bid Deph 1,869,342 2,331, , Ask Deph 1,963,395 2,568, , Univariae analysis of sock liquidiy and rading measures on repurchase days and surrounding non-repurchase days In his secion we presen he resuls of a univariae analysis of sock liquidiy and rading measures on repurchase days and surrounding non-repurchase days. For each repurchase day in he sample, we consruc a non-repurchase period consising of five consecuive rading days before and afer he repurchase day. 22 The non-repurchase period for each repurchase day begins wih he firs rading day before and afer he repurchase day ha is no a repurchase day or a disclosure day (rading day following repurchase day). In he nonrepurchase periods, we exclude repurchase days and disclosure days. Trading days on he day before, on he same day and on he day afer he announcemen of inerim repors are also removed from he non-repurchase periods. We mach he measures on he repurchase days wih he average measures during he corresponding non-repurchase periods and calculae differences in measures for all repurchase days. 22 As a robusness check, we also use benchmark periods wih hree consecuive rading days before and afer he repurchase day, as well as five consecuive rading days before he repurchase day. These alernaive benchmark periods do no produce subsanive changes in our main resuls. 13

14 Table 5 Univariae analysis of sock liquidiy and rading measures on repurchase days and surrounding nonrepurchase days The sample comprises 2,524 repurchase days wih repurchase sizes of 25% or less of he firm s average daily rading volume during he 4 calendar weeks immediaely preceding he week of he repurchase. The sample period is May 16, 2002, o December 30, For each repurchase day in he sample, we consruc a nonrepurchase period consising of five consecuive rading days before and afer he repurchase day or repurchase period if here are consecuive repurchase days. In he non-repurchase periods, we exclude repurchase days and rading days following repurchase days (disclosure days). Trading days on he day before, on he same day and on he day afer he announcemen of inerim repors are also removed from he non-repurchase periods. We mach he measures on he repurchase days wih he average measures during he corresponding non-repurchase periods and calculae differences in measures for all repurchase days. Absolue Spread is he daily average of absolue bid-ask spreads in SEK recorded 15 minues apar over a rading day. Relaive Spread is he daily average of relaive bid-ask spreads recorded 15 minues apar over a rading day. The relaive bid-ask spread is he absolue bid-ask spread divided by he bid-ask midpoin. Toal Deph is he daily average value in SEK of all shares posed a he highes bid and lowes ask price recorded 15 minues apar over a rading day. Bid Deph is he daily average value in SEK of all shares posed a he highes bid price recorded 15 minues apar over a rading day. Ask Deph is he daily average value in SEK of all shares posed a he lowes ask price recorded 15 minues apar over a rading day. Volume is he oal rading volume recorded over a rading day. Price is he daily average of bid-ask midpoins in SEK recorded 15 minues apar over a rading day. Volailiy is he variance of reurns over a rading day where reurns are calculaed by aking he naural logarihm of bid-ask midpoin relaives 15 minues apar. The -saisics are from he paired -es for he differences in mean measures beween repurchase days and benchmark periods. The W-saisics are from he non-parameric Wilcoxon mached-pairs signed-ranks es. p-values from he difference ess are repored in brackes. Variable Absolue Spread Relaive Spread Toal Deph Bid Deph Ask Deph Volume Price Volailiy ( 10-6 ) Repurchase days Mean (Median) ) (0.5303) ) (0.0071) 7,137,606) (450,750) 3,359,366) (204,513) 3,778,239) (247,264) 947,288) (53,750) ) ( ) ) (6.3394) Nonrepurchase days Difference Difference es Mean (Median) ) (0.5612) ) (0.0079) 6,068,201) (363,202) 2,812,995) (168,507) 3,255,206) (191,471) 897,384) (34,134) ) ( ) ) (8.4552) Mean (Median) ) ( ) ) ( ) 1,069,404) (17,586) 546,371) (461) 523,032) (8,652) 49,904) (1,155) ) (0.0766) ) ( ) -saisic [p-value] [0.036] 2.67 [0.008] 1.18 [0.237] W-saisic [p-value] [0.003] 2.66 [0.008] Our marke liquidiy measures - Absolue Spread, Relaive Spread, Toal Deph, Bid Deph and Ask Deph - are described in secion 4.2. The rading measures ha we analyze are daily rading volume, price and volailiy. Volume is he oal number of shares raded during a 14

15 rading day and Price is he daily average of bid-ask midpoins in SEK recorded 15 minues apar over a rading day. Volailiy is he variance of reurns over a rading day where reurns are calculaed by aking he naural logarihm of bid-ask midpoin relaives 15 minues apar. We use a paired -es and a non-parameric Wilcoxon mached-pairs signed-ranks es o es for differences in mean and median measures beween repurchase days and non-repurchase days. Table 5 presens he resuls for 2,524 repurchase days wih repurchase sizes less han 25% of he firm s average daily rading volume during he 4 weeks preceding he week of he repurchase. We cu he sample a his repurchase size level because repurchase days wih higher repurchase sizes conain repurchase ransacions execued as block rades. Block rades are execued ouside he auomaed rading sysem and he liquidiy effecs of hese rades may herefore differ from repurchase rades execued inside he auomaed rading sysem. 23 However, we consider repurchase days wih repurchases execued as block rades in he regression analysis in secion 5.4. As repored, he average absolue (relaive) bid-ask spread is SEK 1.00 (1.14%) on repurchase days and SEK 1.19 (1.41%) on non-repurchase days. The resuls from he difference es show ha he average absolue (relaive) bid-ask spread is significanly lower (1% significance level) on repurchase days while he average deph on boh he bid and ask side is significanly higher on repurchase days han on non-repurchase days. Our rading measures, average rading volume and price are boh significanly higher on repurchase days han on non-repurchase days. Overall, our resuls from he univariae analysis show ha marke liquidiy improves while average rading volumes and share prices increase on repurchase days Regression analysis of marke liquidiy measures Prior marke microsrucure lieraure documens ha volume, price and reurn volailiy are imporan deerminans of marke liquidiy (e.g., Benson and Hagerman, 1974; Tinic and Wes, 1974; Copeland and Gala 1983). The univariae analysis in secion 5.3 shows ha hese variables change on repurchase days and, herefore, we include daily rading volume, 23 Repurchase days wih repurchase sizes lower han 25% of he firm s average daily rading volume during he four weeks preceding he week of he repurchase may also conain block rades. However, we are unable o specifically idenify hese repurchase days wih he available daa. 15

16 share price and reurn volailiy as conrol variables in a muliple regression. 24 Furhermore, he resuls in Table 3 indicae ha he average relaive bid-ask spread narrows on he SSE during he sample period. To conrol for his marke-wide change in marke liquidiy, we also include year dummy variables in he regression. Since we are ineresed in examining he liquidiy impac on repurchase days wih differen repurchase sizes, we pariion he repurchase days ino four groups based on daily repurchase size. We define he repurchase size as he daily repurchase volume relaive o he average rading volume during he 4 weeks preceding he week of he repurchase. In he regression, we include a repurchase dummy variable for each size group and use surrounding nonrepurchase days as benchmark days. 25 Repurchase days wih a repurchase size greaer han 25% are defined as a repurchase days wih a block rade because firms can only repurchase more han 25% of he average rading volume during he 4 weeks preceding he week of he repurchase wih off-exchange block rades. A pooled OLS regression ha makes no allowance for fixed unobserved differences beween repurchasing firms may produce biased resuls. We herefore conduc a Breusch- Pagan Lagrange Muliplier (LM) es of firm random effecs (Breusch and Pagan, 1980). The chi-square es saisic from he Breusch-Pagan LM es rejecs he null hypohesis a he 1% level such ha he variances of firm effecs are zero in he model for each of he five marke liquidiy measures. Thus, pooled OLS regressions may produce biased esimaors. To compare he firm fixed effecs model and he firm random effecs model, we conduc a Hausman Specificaion es (Hausman, 1978). The chi-square es saisic from he Hausman es rejecs he null hypohesis a he 1% level such ha he firm effecs are uncorrelaed wih he oher regressors in he model for each of he five marke liquidiy measures. Thus, we use he following firm fixed effecs regression: Liquidiy LNVOLUME 1 SMALL i 1 LNPRICE 2 MEDIUM 2 LNVOLAT 3 LARGE 3 BLOCK 4 (1) where Liquidiy is he dependen variable and represened by eiher he naural logarihm of Absolue Spread, Relaive Spread, Toal Deph, Bid Deph or Ask Deph. 26 Absolue Spread is 24 Prior research also documens a negaive relaionship beween bid-ask spreads and firm size. Since we use a firm fixed effecs regression in our analysis, we also conrol for his effec. 25 Since he relaionship beween repurchase size and marke liquidiy may be nonlinear, we use dummy variables for differen repurchase size inervals and no a quaniaive variable for repurchase size. 26 Owing o he possibiliy of skewness in he unransformed measures, all non-dummy variables are ransformed o heir log values, as logged variables are more normally disribued. The Shapiro-Francia es for normaliy do no rejec he null hypohesis ha he log-ransformed variables are normally disribued. 16

17 he daily average of inraday absolue bid-ask spreads. Relaive Spread is he daily average of inraday relaive bid-ask spreads. The relaive bid-ask spread is he absolue bid-ask spread divided by he bid-ask midpoin. Toal Deph is he daily average value of all shares posed a he highes bid and lowes ask price. Bid Deph is he daily average value of all shares posed a he highes bid price. Ask Deph is he daily average value of all shares posed a he lowes ask price. SMALL (MEDIUM, LARGE, BLOCK) is a dummy variable coded as 1 if he firm s repurchase volume is less han 10% (10%-20%, 20%-25% and more han 25%) of he average daily rading volume during he 4 weeks preceding he week of he repurchase, and zero oherwise. The surrounding non-repurchase days used as benchmark days are coded as zero. The daa se comprises 859 repurchase days wih small repurchase sizes, 688 repurchase days wih medium repurchase sizes, 977 repurchase days wih large repurchase sizes, 282 repurchase days wih block ransacions and 2,570 surrounding non-repurchase days used as benchmark days. LNVOLUME is he naural logarihm of he oal rading volume execued in he auomaed rading sysem during a rading day. 27 LNPRICE is he naural logarihm of he daily average of bid-ask midpoins. LNVOLAT is he naural logarihm of he variance of reurns over a rading day. All non-dummy variables excep LNVOLUME represen daily averages of inraday daa recorded a 15 minue inervals. Table 6 repors he esimaed coefficiens from he firm fixed effecs regression for each of he five marke liquidiy measures. We adjus he sandard errors of he esimaed coefficiens for heeroscedasiciy using Whie s (1980) robus esimaes of sandard errors and repor he p-values wihin parenheses. The goodness of fi is relaively high for all five regressions wih adjused R-squares beween 0.84 and The esimaed coefficiens for he conrol variables LNVOLUME, LNPRICE and LNVOLAT are all saisically significan a he 1% level and heir signs are consisen wih prior sudies. The resuls show ha larger rading volumes are associaed wih narrower bidask spreads and deeper deph. Higher share prices are associaed wih wider absolue bid-ask spreads, narrower relaive bid-ask spreads and higher deph. Higher inraday reurn volailiy is associaed wih wider bid-ask spreads and less deph. All he esimaed coefficiens on he year dummies are a leas saisically significan on he 10% level (no repored in Table 6). 27 The variable for volume is defined as he daily rading volume in he auomaed rading sysem because i deermines changes in marke liquidiy beer han he oal rading volume, which also includes on-exchange rades ouside he auomaed rading sysem. 17

18 Table 6 Firm fixed effecs regression of marke liquidiy measures across repurchase days and surrounding nonrepurchase days conrolling for volume, price and volailiy The sample comprises 2,806 share repurchase days and 2,570 non-repurchase surrounding days over he period May 16, 2002, o December 30, 2005, for 53 Swedish firms lised on he SSE. Non-repurchase days consiue five consecuive non-repurchase rading days before and afer he repurchase day or he repurchase period if here are consecuive repurchase days. In he non-repurchase periods, we exclude repurchase days, disclosure days and rading days wihou ransacions. Trading days he day before, he same day and he day afer he announcemen of inerim repors are also removed from he non-repurchase periods. The able repors he parameer esimaes from he following firm fixed effecs regression: Liquidiy LNVOLUME 1 Liquidiy is he dependen variable and is represened by eiher he naural logarihm of Absolue Spread, Relaive Spread, Toal Deph, Bid Deph or Ask Deph. Absolue Spread is he daily average of absolue bid-ask spreads in SEK recorded 15 minues apar over day for firm i. Relaive Spread is he daily average of relaive bid-ask spreads recorded 15 minues apar over day for firm i. The relaive bid-ask spread is he absolue bid-ask spread divided by he bid-ask midpoin. Toal Deph is he daily average value in SEK of all shares posed a he highes bid and lowes ask price recorded 15 minues apar over day for firm i. Bid Deph is he daily average value in SEK of all shares posed a he highes bid price recorded 15 minues apar over day for firm i. Ask Deph is he daily average value in SEK of all shares posed a he lowes ask price recorded 15 minues apar over day for firm i. SMALL (MEDIUM, LARGE, BLOCK ) is a dummy variable coded as 1 if firm i has repurchased on day, a repurchase size less han 10% (10%-20%, 20%-25%, more han 25%) of he average daily rading volume during he four weeks preceding he week of he repurchase, and zero oherwise. LNVOLUME is he naural logarihm of he oal rading volume execued in he auomaed rading sysem on rading day for firm i. LNPRICE is he naural logarihm of he daily average of bid-ask midpoins in SEK recorded 15 minues apar over day for firm i. LNVOLAT is he naural logarihm of he variance of reurns over day for firm i where reurns are calculaed by aking he naural logarihm of bid-ask midpoin relaives 15 minues apar. Regression parameers are esimaed wih year dummies. The coefficiens on firm and year dummies are no repored. p-values in parenheses are adjused for heeroscedasiciy using Whie s (1980) robus esimaes of sandard errors. *, ** and *** indicae wo-ailed significance levels of 10%, 5% and 1%, respecively. Independen variable No. of dum=1 SMALL 859 MEDIUM 688 LARGE 977 BLOCK 282 LNVOLUME LNPRICE LNVOLAT SMALL i 1 LNPRICE 2 Absolue Spread coefficien (p-value) Relaive Spread coefficien (p-value) * * (0.092) * (0.083) ** (0.044) *** (0.010) * (0.059) *** *** *** MEDIUM ** (0.044) ** (0.011) * (0.058) *** *** *** Toal Deph coefficien (p-value) *** (0.691) (0.442) *** (0.002) *** *** *** Bid Deph coefficien (p-value) *** (0.146) *** (0.003) * (0.094) *** *** *** Ask Deph coefficien (p-value) *** (0.010) (0.940) (0.244) ** (0.002) *** *** *** Year dummies Yes Yes Yes Yes Yes Number of obs. 5,376 5,376 5,376 5,376 5,376 Number of firms Adjused R LNVOLAT 3 LARGE 3 BLOCK 4 18

19 Nex, we urn o he esimaed coefficiens on he four repurchase dummy variables SMALL, MEDIUM, LARGE and BLOCK. Since we conrol for volume, price and volailiy, he esimaed coefficiens on he repurchase dummy variables deermine he marginal impac of repurchase rading on marke liquidiy. The bid-ask spread narrows significanly on repurchase days wih small, medium and large repurchase sizes (SMALL, MEDIUM and LARGE). The narrowing is also posiively associaed wih he repurchase size. These resuls are consisen wih our firs execuionrelaed hypohesis, which predics ha he bid-ask spread will narrow on repurchase days when firms submi buy limi orders a he bid or wihin he bid-ask spread in he order book. The marke deph on he bid-side increases significanly on repurchase days wih small repurchase sizes. However, on repurchase days wih large repurchase sizes he marke deph on he bid-side decreases significanly. These resuls indicae ha firms place limi orders a he bid o a large exen on repurchase days wih small repurchase sizes and limi orders wihin quoes or a he ask on repurchase days wih large repurchase sizes. On repurchase days wih a leas par of he repurchase rades execued as block rades (BLOCK), bid-ask spreads widen (saisically significan a he 10% level). This wider bidask spread is consisen wih he informaion-asymmery hypohesis, which predics ha marke paricipans deec and suspec informed rading Robusness es of rading behavior In secion 5.4, our resuls sugges ha repurchasing firms are likely o submi buy limi orders a he bid on repurchase days wih small repurchase sizes and submi buy limi orders wihin he curren bid-ask spread or a he ask on repurchase days wih large repurchase sizes. As a robusness es of hese rading behaviors, we analyze he average repurchase price relaive o he average of inraday bid-ask midpoins on repurchase days wih differen repurchase sizes. We expec o find a low average repurchase price relaive o he inraday average of bid-ask midpoins on repurchase days wih relaively small repurchase sizes and a high average repurchase price relaive o he average of inraday midpoin bid-ask on repurchase days wih relaively large repurchase sizes. As can be seen in Table 7, he percenage of repurchase days wih an average repurchase price above he average of inraday bid-ask midpoins is higher on repurchase days wih large 28 Marke paricipans can observe execued block rades during rading hours since brokers mus repor manually execued block rades wihin five minues from he ime of he rade. 19

20 repurchase sizes. On repurchase days wih large (small) repurchase sizes, we find ha 67.2% (49.1%) of he repurchase days have an average repurchase price above he average of inraday bid-ask midpoins. We also analyze he relaive price difference beween he average repurchase price and he average of inraday bid-ask midpoins over a rading day on repurchase days wih various repurchase sizes. The relaive price difference is defined as: Relaive Price difference Repurchase Price BidAsk Midpoin BidAsk Midpoin (2) where Repurchase Price is he average daily repurchase price for firm i on repurchase day and BidAsk Midpoin is he average of inraday bid-ask midpoins for firm i on repurchase day. The resuls in Table 7 show ha boh he mean and median of he Relaive Price difference is lower on repurchase days wih small repurchase sizes han on repurchase days wih large repurchase sizes. On repurchase days wih small repurchase sizes, he relaive price difference is on average -0.1%, while i is 0.3% on repurchase days wih large repurchase sizes. The difference is highly significan a he 1% level, boh in a -es and a non-parameric Wilcoxon rank sum es (no repored). The resuls are hus consisen wih he prediced rading behavior by he repurchasing firms on repurchase days wih various repurchase sizes. Table 7 Relaionship beween repurchase price and repurchase size The sample comprises 2,806 open marke share repurchase days over he period May 16, 2002, o December 30, 2005, for 53 Swedish firms lised on he SSE. SMALL (MEDIUM, LARGE, BLOCK) is a repurchase day wih a repurchase size less han 10% (10%-20%, 20%-25%, more han 25%) of he average daily rading volume during he four weeks preceding he week of he repurchase. Relaive Price difference is defined as: Relaive Price difference Repurchase Price BidAsk Midpoin BidAsk Midpoin where Repurchase Price is he average daily repurchase price for firm i on repurchase day and BidAsk Midpoin is he average of inraday bid-ask midpoins for firm i on repurchase day. SMALL MEDIUM LARGE BLOCK Number of repurchase days Percenage of repurchase days wih a daily average repurchase price above he average of inraday bid-ask midpoins Mean Relaive Price difference (%) Median Relaive Price difference (%)