Managing EPS Through Accelerated Share Repurchases: Compensation Versus Capital Market Incentives

Transcription

1 Managng EPS Through Accelerated Share Repurchases: Compensaton Versus Captal Market Incentves Carol Marquardt Assocate Professor Baruch College CUNY Chrstne Tan Assstant Professor Baruch College CUNY and Susan M.Young Assocate Professor Baruch College CUNY October 2007 We thank Bo Zhang for research assstance and Monca Bany, Mary Ellen Carter, Vctora Dcknson, Valentn Dmtrov, Peter Easton, Ncole Thorne Jenkns, Joshua Lvnat, Joshua Ronen, Tom Stober, and partcpants at the Unversty of Ednburgh and Unversty of Notre Dame accountng workshops, the Columba-NYU-Baruch-Rutgers Jont Accountng Conference, and the 2007 AAA Annual Meetng for helpful suggestons.

2 Managng EPS Through Accelerated Share Repurchases: Compensaton Versus Captal Market Incentves Abstract: Ths paper emprcally examnes the determnants of frms decsons to undertake accelerated share repurchases (ASRs). In an ASR, the frm repurchases ts own shares of stock through an nvestment bank rather than on the open market, allowng the company to acqure a targeted number of shares and record ts effects on earnngs per share (EPS) mmedately. Consstent wth our predctons, we fnd that ASR frms are more lkely to compensate ther managers explctly on reported EPS fgures and are less lkely to be concerned wth benchmark-beatng than are OMR frms. These results are robust to controllng for sgnalng effects, as well as other known determnants of stock repurchase decsons. Addtonal analyss suggests that annual cash compensaton s not adjusted for the reportng effects of the ASR. Our fndngs contrbute to the lteratures on stock repurchases, earnngs management, and executve compensaton. Key Words: share repurchases; earnngs management; manageral compensaton; captal market ncentves. Data avalablty: Data s avalable from publcly-avalable sources dentfed n the manuscrpt.

3 I. INTRODUCTION Ths paper emprcally examnes the recent phenomenon of accelerated share repurchases (ASRs). ASRs dffer from open market repurchases (OMRs) of stock n two mportant respects. Frst, n an ASR, the frm does not repurchase shares on the open market but rather borrows ts own shares of stock from an nvestment bank. Ths allows the company to both acqure a targeted number of shares mmedately and to recognze the full effects of the transacton on reported earnngs per share (EPS) n the current accountng perod. Second, the frm enters nto a forward contract wth the nvestment bank and s thus oblgated to repurchase a pre-specfed number of shares at a purchase prce determned by an average market prce over the contract perod; there s no smlar oblgaton n OMRs. 1 Our study s motvated by recent artcles n the fnancal press that have crtczed the ncreasng use of ASRs (see Maremont and Ng 2006). The man concern s whether frms are usng ASR arrangements to obtan short-term EPS ncreases but potentally damage shareholder value n the long run due to the guaranteed nature of the repurchase agreement. We shed lght on ths queston by examnng whether earnngs management ncentves are a sgnfcant determnant of frms decsons to undertake ASRs versus other methods of stock repurchase. Pror research shows that frms use OMRs as an earnngs management devce. For example, Bens et al. (2003) fnd that frms ncrease the level of ther frms stock repurchases when earnngs are below the level requred to acheve the desred rate of EPS growth, and Hrbar et al. (2006) fnd that frms use stock repurchases to meet or beat analysts forecasts of EPS. Essentally, frms explot the flexblty nherent n OMR 1

4 plans to tme repurchases so that ther fnancal reportng objectves.e., meetng or beatng an earnngs benchmark are acheved. In an ASR, however, there s no fnancal reportng flexblty that can be exploted. The full amount of shares targeted for repurchase are acqured mmedately through an nvestment bank, common shares outstandng decreases, and ths effect s mmedately reflected n reported EPS for the current accountng perod. It thus seems unlkely that an ASR would be an approprate vehcle for managng EPS f the am s to meet or beat earnngs benchmarks. We consequently do not expect that captal market ncentves to meet or beat earnngs benchmarks wll play an mportant role n the decson to undertake an ASR. The mmedate recognton of a decrease n shares outstandng that occurs wth ASRs does suggest an alternatve motvaton, however that the managers of ASR frms are more lkely to be compensated on reported EPS fgures than are the managers of OMR frms. Pror research shows that the use of earnng-based bonuses affects frm s fnancal reportng choces. For example, Beatty and Weber (2006) fnd that the lkelhood of managers recevng earnngs-based bonuses affects goodwll mparment decsons, and Marquardt and Wedman (2005) fnd that frms are more lkely to structure convertble bond transactons to ncrease EPS when manager bonuses are based on reported EPS fgures. Gven the relatvely large magntude of ASRs on reported EPS and the lack of fnancal reportng flexblty assocated wth ASRs, we beleve that compensaton ncentves are a lkely determnant n the decson to engage n ASRs. Usng probt analyss, we emprcally test these predctons usng a sample of 675 repurchase announcements from Our dependent varable, ASR, equals one f 2

5 frms repurchase stock through an ASR and zero f frms repurchase stock on the open market. We proxy for compensaton ncentves by creatng an ndcator varable, BONUS, that equals one f EPS s explctly mentoned as a determnant of annual bonuses n the frms proxy statement, and zero otherwse. We proxy for captal market ncentves usng a number of varables dentfed n pror research as related to managers ncentves to meet or beat earnngs benchmarks. The frst of these varables captures the number of consecutve quarterly ncreases n EPS n the fve-year perod pror to the repurchase announcement. If captal market ncentves drve the decson to undertake OMRs rather than ASRs, we expect longer (shorter) strngs of earnngs ncreases for OMRs (ASRs). Because Matsumoto (2002) fnds that frms wth large mplct clams by stakeholders have greater ncentves to meet or beat earnngs benchmarks, we nclude membershp n a durable goods ndustry, research and development expendtures, and labor ntensty as proxes for the magntude of mplct clams. We expect a postve (negatve) assocaton between these varables and the lkelhood of undertakng an OMR (ASR). We also nclude sales growth as a proxy for captal market ncentves; we expect hgher (lower) sales growth for OMR (ASR) frms. We control for the possblty that the sgnalng hypothess (Brav et al. 2005; Peyer and Vermaelen 2005; Ikenberry et al. 1995) mght explan frms decsons to undertake ASRs versus OMRs by ncludng pror stock prce performance, leverage, and dvdend yeld n our model. We also control for other varables known to be assocated wth the decson to repurchase stock, ncludng potental dluton from stock opton plans, frm sze, and free cash flows. 3

6 As predcted, we fnd that frms are sgnfcantly more lkely to choose an ASR over an OMR when managers are explctly compensated on EPS. We also fnd that ASR frms have shorter strngs of quarterly earnngs ncreases, lower labor ntensty, and lower sales growth than OMR frms, consstent wth captal market ncentves playng a stronger role n the case of OMRs. ASR frms also tend to be sgnfcantly larger and have lower stock prce volatlty than OMR frms. In addton, there s lttle evdence that sgnalng s a sgnfcant determnant n the decson to undertake an ASR. As a senstvty test, we also compare the determnants of ASRs and tender offers, as t may be argued that tender offers could also be used as an earnngs management devce n a manner smlar to ASRs. As wth ASRs, we fnd lttle evdence that managers decsons to undertake tender offers are drven by ncentves to meet or beat earnngs benchmarks. However, we fnd that ASR frms are sgnfcantly more lkely to reward ther executve on EPS performance, wth 58.3 percent of ASR frms mentonng EPS as a determnant of cash bonuses n ther proxy statements versus only 13.3 percent of tender offer frms. We also fnd that tender offer frms are sgnfcantly smaller, wth poorer stock performance pror to the repurchase announcement. These results suggest that tender offers are undertaken n an attempt to correct market undervaluaton, consstent wth sgnalng hypotheses, rather than to acheve specfc fnancal reportng objectves related to EPS. We also provde descrptve evdence on the settlement costs of ASRs and fnd that, on average, the settlement cost of the forward contract exceeds the ntal repurchase prce by an average of 5.7 percent. We further fnd that n cases where the settlement prce exceeds the ntal repurchase prce, the contract s always settled n cash, whle n 4

7 nstances where the settlement prce s less than the ntal repurchase prce the contract s twce as lkely to be settled n shares. Ths pattern suggests that frms choose the form of settlement n order to mnmze the contract s dlutve effect on EPS, thereby provdng addtonal evdence that fnancal reportng consderatons play an mportant role n ASRs. The above fndngs beg the queston of whether compensaton commttees adjust reported EPS for the effects of the ASRs when determnng cash compensaton levels. Usng a model smlar to Healy et al. (1987), we fnd no evdence that compensaton commttees adjust reported EPS n settng executve pay. Ths fndng mght be nterpreted as evdence that managers are enrchng themselves at the cost of shareholders by choosng to repurchase shares through an ASR, but an alternatve explanaton could be that corporate boards and compensaton commttees are smply encouragng approprate rsk-takng by provdng earnngs-based ncentve compensaton. To gan some nsght nto ths ssue, we examne two-day abnormal returns around the repurchase announcement. Whle abnormal returns around ASRs are slghtly postve, they are not sgnfcantly dfferent from OMR announcement returns after controllng for other determnants of the market response. However, Bany and Mathew (2007) fnd that ASR announcement returns are sgnfcantly smaller than those for tender offers, though the guaranteed nature of the ASR agreement would predct an equally strong market response. Based on these analyses, we cannot currently conclude that ASRs are detrmental to the frm, as alleged n the fnancal press. 2 Ths paper contrbutes to the accountng lterature n several ways. Frst, we extend the lterature on earnngs management by explctly lnkng bonus compensaton 5

8 to stock repurchases. Whle pror work by Bens et al. (2003) and Hrbar et al. (2006) have shown that benchmark-beatng s a sgnfcant determnant n the decson to undertake OMRs, we show that dfferent ncentves related to EPS reportng are at play n the decson to undertake ASRs. Ths fndng s mportant n that t deepens our nsght nto managers motvatons behnd basc fnancng decsons that affect the frm. Our fndngs also underscore the centralty of EPS n manageral decson makng. In ther survey of CFOs, Graham et al. (2005) document that earnngs, and EPS n partcular, are vewed as the most mportant performance measure, yet relatvely few papers examne questons of EPS management. Our study adds to ths nascent lterature by provdng new evdence that frms manage EPS when t s used as a performance metrc n compensaton contracts. Our results also have mplcatons for standard settng. Our evdence on ASR settlements shows that whle frms structure the forward contract transacton such that t allows them to avod mark-to-market accountng (.e., they retan the opton to settle the forward contract n cash or shares), they typcally settle the contracts n cash to avod ressung shares that would dlute reported EPS. The Fnancal Accountng Standards Board (FASB) n ts re-delberaton of SFAS 128 recently ssued a tentatve decson n October 2006, statng that contracts that may be settled n ether cash or shares at the entty s opton should presume that the contract wll be settled n shares f the effect s dlutve. That presumpton may not be overcome, regardless of past practce or stated polcy to the contrary 3 Currently, under SFAS 128, ths presumpton may be overcome f past experence or a stated polcy provdes a reasonable bass to beleve that the contract wll be pad partally or wholly n cash. Our results suggest that such a provson may be 6

9 necessary to prevent managers from structurng forward contract transactons n a manner that potentally enrches themselves at the expense of shareholders. The remander of the paper s organzed as follows. Secton II presents the accountng treatment for ASRs n more detal. We develop our hypotheses n Secton III and descrbe our research desgn n Secton IV. We outlne our sample selecton crtera n Secton V and present our results n Secton VI. In Secton VII, we present addtonal analyses and dscuss our conclusons n Secton VIII. II. ACCELERATED SHARE REPURCHASES The volume and magntude of share repurchases has reached record levels n the past few years, wth lttle evdence that ths trend wll soon subsde. A report by Standard and Poor s ssued n June 2006 showed that companes had spent a record $367 bllon on stock buybacks n the year ended March 31. Companes n the S&P 500 alone were expected to repurchase more than $435 bllon n shares durng 2006, a consderable ncrease from the approxmately $349 bllon repurchased by the 500-ndex frms n 2005 One method of share repurchases that has shown a correspondng ncrease s an ASR. An ASR s an arrangement n whch a company borrows a block of frm shares from an nvestment bank and mmedately recognzes a reducton n EPS (on a weghted average bass). At the tme of the arrangement, the company smultaneously enters nto a forward agreement wth the nvestment bank. The nvestment bank mmedately sells the shares to the company by borrowng the shares from other nvestors. The nvestment bank buys the company shares back n the open market over tme, generally less than one year, and replaces the borrowed shares (see Fgure 1). 7

10 Two accountng-related transactons occur when a frm enters nto the ASR agreement. Frst, equty s mmedately decreased by the number of shares to be repurchased tmes the current share prce, and cash s decreased or a lablty s ncreased by an equal amount. Second, the frm enters nto a forward contract wth the fnancal nsttuton, whch allows the nvestment bank to hedge ts short sale of shares. For most ASR agreements, the frm can choose to settle the contract n ether cash or shares for the volume-weghted-average-value of the dfference n share prce as of the begnnng of the ASR agreement to the settlement date. Under an ASR agreement wth a cash or share settlement opton, companes are not requred under current reportng standards to mark the forward contract to market on ther books. The assumpton behnd the accountng treatment of the forward contract (not requrng t to be marked to market as the underlyng value of the frm s stock changes) s that the company ntends to settle the forward contract n shares and therefore need not consder the change n the far value of the forward contract n the calculaton of net ncome. In realty, the large majorty of ASR forward contracts are settled n cash. At settlement, the accountng treatment s to decrease cash (or ncrease labltes) and to decrease equty, assumng the prce of the company s stock has ncreased. The repurchased shares may be kept n treasury or retred. The key dfference n accountng treatments between ASRs and OMRs s the tmng of the recognton of the decrease n shares outstandng. Therefore, the man advantage to a frm n choosng an ASR s the mmedate mpact on outstandng shares and perhaps a stronger sgnal to the market about frm value. The dsadvantage s that cash must be provded up front, and the frm must pay the average share value over the 8

11 lfe of the contract regardless of the ncrease n share prce. Frms do not have an opton to dscontnue repurchasng shares once the ASR has been entered nto as they would wth an OMR program. In fact, pror research has shown that almost 25 percent of frms that announce an OMR do not repurchase shares n the announcement quarter (Le 2005). We beleve that one reason for the ncreased frequency of ASR agreements s related to the ssuance of Statement of Fnancal Accountng Standard (SFAS) No. 150, Accountng for Certan Fnancal Instruments wth Characterstcs of both Labltes and Equty. SFAS 150 became effectve for nterm perods after June 15, Pror to SFAS 150, frms commonly wrote put optons on ther own shares to hedge aganst prce ncreases. SFAS 150 requres that frm use mark-to-market accountng on puts and forward optons, reducng the beneft to the frm by requrng changes n value to be recorded as ncreases or decreases to net ncome. However, as noted above, the forward contracts assocated wth ASRs are not requred to be marked-to-market when the frm has the opton of settlng the contract n cash or shares. 4 We provde a numercal example of the accountng treatment for ASRs n the Appendx. III. HYPOTHESES We consder the above dfferences n the accountng treatment of ASRs versus OMRs n developng our hypotheses about manageral ncentves behnd each type of repurchase. Specfcally, because the decrease n equty s recognzed mmedately for the full amount of shares announced as repurchase targets n an ASR, ths repurchase type does not provde the fnancal reportng flexblty that OMRs offer. For example, Hrbar 9

12 et al. (2006) emprcally show that n response to captal market pressures to meet analysts EPS forecasts, frms explot the flexblty that OMRs offer n terms of choosng when or whether to buy back stock when they are lkely to fall short of meetng analyst expectatons;.e., they do an OMR when they need a penny to make the forecast. Whle t s possble that frms also use ASRs to meet or exceed analyst forecasts, because the number of shares repurchased s known n advance wth certanty, t s a relatvely straghtforward exercse to adjust expected EPS for the effects of the repurchase. Indeed, anecdotal evdence shows that managers themselves explctly dsclose the reportng effects of the ASR on future EPS, and analyst adjust ther forecasts accordngly. For example, Rockwell Collns ssued a press release on September 29, 2006, announcng an ASR of 4.7 mllon shares at an ntal cost of $257 mllon. The press release also ncluded the followng statement: Wth the executon of ths agreement, the company now expects fscal year 2007 earnngs per share n the range of $3.10 to $3.20, a 5 cent ncrease over the prevously announced gudance range of $3.05 to $3.15. Smlarly, n a report ssued by Bear, Stearns & Co. on June 30, 2005 for Del Monte Foods, the analyst specfcally states that the company bought back 12 mllon shares through an ASR, and Bear Stearns s therefore rasng ther EPS estmate by one cent to account for the transacton. We therefore expect captal market ncentves to acheve earnngs benchmarks to be relatvely less mportant n the decson to undertake an ASR versus an OMR. Our frst hypothess s therefore as follows: H1: Captal market ncentves to meet or beat earnngs benchmarks play a less mportant role n the decson to undertake an ASR versus an OMR. 10

13 The nflexble nature of the reportng effect of ASRs on EPS fgures suggests to us to a qute dfferent motvaton we predct that the managers of ASR frms are more lkely to be compensated on reported EPS fgures than are the managers of OMR frms. Pror research shows that the use of earnng-based bonuses affects frms fnancal reportng choces. For example, Beatty and Weber (2006) fnd that the lkelhood of managers recevng earnngs-based bonuses affects goodwll mparment decsons, and Marquardt and Wedman (2005) show that frms are more lkely to structure convertble bond transactons to ncrease EPS when manager bonuses are based on reported EPS fgures. In addton, Healy et al. (1987) fnd that compensaton commttees do not appear to adjust earnngs for accountng choces related to nventory and deprecaton methods when settng manageral pay. There s also practtoner evdence relatng stock repurchases to executve compensaton. In ther recent report from The Center for Fnancal Research and Analyss and The Corporate Lbrary, Lehman and Hodgson (2006) speculate that some amount of share repurchase programs may actually damage shareholder value. They examned frms n the S&P 500 wth negatve cash flows pror to or durng share repurchase programs as possble perpetrators of non-benefcal programs and found that a greater percentage of the CEOs for these negatve cash flow frms were rewarded on per share performance metrc (43.11 percent) compared to S&P 500 frms as a whole (27.85 percent). They also found that bonuses were more lkely pad out to the CEOs of these frms 88 percent of these CEOs receved annual bonuses versus 78 percent of the S&P 500 frms and 37.5 percent receved cash bonuses versus 22.2 percent of CEOs n the S&P 500. The authors 11

14 conclude that share repurchase programs may be used generate hgher levels of EPS and EPS growth n order to ncrease payout of ncentves. Gven the above fndngs, the relatvely large magntude of ASRs on reported EPS, ther lack of fnancal reportng flexblty, and the possblty that compensaton commttees wll not adjust EPS for the repurchase, we beleve that compensaton ncentves are a lkely determnant n the decson to undertake an ASR versus an OMR. Stated formally: H2: Compensaton ncentves play a more mportant role n the decson to undertake an ASR versus an OMR. IV. RESEARCH DESIGN We use a multple probt regresson model to test Hypotheses 1 and 2, where the dependent varable, ASR, equals one f the frm chooses to undertake an ASR and zero f the frm chooses an OMR. As such, our analyss s condtonal on the decson to repurchase stock; that s, we assume that frms frst decde to repurchase stock and subsequently determne the type of repurchase to undertake. To proxy for captal market ncentves (H1), we employ a number of varables drawn from pror research. Our frst varable s based on the collectve fndngs of Barth et al. (1999) and Myers et al. (2007), who both fnd that the market rewards patterns of ncreasng earnngs, and Graham et al. (2005, 22), who report that chef fnancal offcers regard the same quarter of last year s EPS as the most mportant benchmark. Ths varable, STRING, equals the number of consecutve quarters pror to the announcement date of the repurchase that the frm has met or exceeded the benchmark of the pror year s EPS for the same fscal quarter, up to a maxmum of 20 quarters. If captal market 12

15 ncentves to mantan ths strng are stronger for OMR frms, we expect a negatve coeffcent on STRING n our probt analyss. Matsumoto (2002) emprcally examnes managers ncentves to avod negatve earnngs surprses and fnds that frms wth greater relance on mplct clams wth stakeholders are more lkely to meet or exceed earnngs expectatons. Stakeholders such as customers, employees, or supplers are lkely to react more strongly to earnngs surprses because they have lmted ablty or do not fnd t cost effectve to fully process all nformaton about the frm (Hrshlefer and Teoh 2003). We thus expect frms relance on mplct clams wth stakeholders to be negatvely assocated wth the lkelhood of undertakng an ASR. We follow Matsumoto (2002) and Bowen et al. (1995) and use three varables to proxy for mplct clams: membershp n a durable goods ndustry (DUR), defned as SIC codes , 245, , 283, 301, and , research and development expendtures (R&D) dvded by total assets, and labor ntensty (LABOR), defned as one mnus the rato of gross property, plant, and equpment to total assets. We measure DUR, R&D, and LABOR at the end of the fscal year precedng the repurchase announcement and predct that each varable wll be negatvely assocated wth the decson to undertake an ASR versus an OMR. Our fnal proxy for captal market ncentves s sales growth (SGROWTH). Sknner and Sloan (2002) fnd that asymmetry n the market response to postve versus negatve earnngs surprses s stronger for hgh growth than for low growth frms,.e. the dramatc losses n frm value that often occur after mssng an earnngs benchmark are more severe for growth stocks. We consequently expect SGROWTH, defned as annual 13

16 sales growth over the fscal year pror to the repurchase announcement, to be negatvely assocated wth the lkelhood of undertakng an ASR. To proxy for compensaton ncentves (H2), we follow Marquardt and Wedman (2005), and create an ndcator varable, BONUS, that equals one f EPS s explctly mentoned as a determnant of annual bonuses n the frms proxy statement and zero otherwse. We beleve that ths s the most drect measure of whether a manager s compensated based on reported EPS. If managers of ASR frms are motvated by compensaton concerns, we expect a postve coeffcent on BONUS. In addton to our test varables for H1 and H2, we also nclude ndependent varables to control for possble alternatve motvatons for undertakng an ASR nstead of an OMR. Bens et al. (2003) and Kahle (2002) present evdence that frms undertake OMRs to offset the dluton assocated wth employee stock opton plans. We further note that the flexblty nherent n OMR plans makes t a superor tool over ASRs n managng antcpated dluton, as managers can vary the amount of share repurchased as necessary. We therefore nclude DILUTION as a control varable n our analyss, where DILUTION s defned as the dfference between the shares used to calculate dluted and basc EPS, dvded by total shares outstandng, measured as of the end of the fscal year precedng the repurchase announcement. 5 We expect DILUTION to be negatvely assocated wth the decson to undertake an ASR. We also consder the possble sgnalng effects assocated wth stock repurchases. Sgnalng theory would suggest that the wllngness of managers to ncrease ther holdngs of a company s stock conveys new, postve nformaton to the market regardng the future cash flow of the company, and emprcal evdence documents that stock 14

17 repurchase announcements result n postve stock prce changes (e.g., Brav et al. 2005; Peyer and Vermaelen 2005; Ikenberry et al. 1995). We argue that the guaranteed nature of the repurchase n an ASR sends a stronger sgnal to nvestors than does an OMR, snce there s no oblgaton on the part of the ssuer to actually repurchase any shares n an OMR. We attempt to control for sgnalng effects by ncludng frms debt-to-equty ratos (DE) and dvdend yelds (DIVYIELD) as ndependent varables n our analyss, both measured for the fscal year pror to the repurchase announcement. When managers possess nsde nformaton, fnancal structure sgnals nformaton to the market, wth the value of the frm rsng wth ncreasng leverage; smlarly, when outsde nvestors have mperfect nformaton about frms proftablty, dvdends functon as a postve sgnal of expected cash flows (Barclay et al. 1995). If ASRs serve n a sgnalng role, t may be more lkely that frms undertakng ASRs have already exhausted ther sgnalng capactes by havng hgh debt levels and hgh dvdend yelds. We therefore expect DE and DIVYIELD to be postvely assocated wth the ASR decson. Another possble motvaton for undertakng an ASR versus an OMR s that frms are usng the ASR to hedge aganst stock prce fluctuatons that mght affect the cost of the stock repurchase. Because ASRs became more popular after frms were requred to mark ther wrtten put optons to market wth the ncepton of SFAS 150, t s possble that frms are now usng ASRs rather than wrtten puts to hedge aganst large stock prce ncreases. If a hedgng argument apples, we expect stock return volatlty (STKVOL), defned as the annualzed standard devaton of daly stock returns n the calendar year 15

18 precedng the repurchase announcement, to be postvely assocated wth the decson to undertake an ASR versus an OMR. 6 We also nclude other known determnants of the repurchase decson as addtonal control varables. Jagannathan et al. (2000) fnd that stock repurchasers typcally have lower stock returns and hgher free cash flows relatve to dvdend-payng frms. If managers undertake an ASR nstead of an OMR because they beleve the frm s more undervalued, then we expect buy-and-hold abnormal returns (BHAR), measured over the 90-day perod pror to the repurchase announcement date, wll be negatvely assocated wth the ASR decson. We further predct that free cash flows (FCF) are lkely to be hgher for ASR frms than for OMR frms because the targeted number of shares must be reacqured mmedately through an underwrter n an ASR, whch would requre a large cash outlay. We defne free cash flows (FCF) as operatng cash flows mnus captal expendtures, dvded by total assets, at the end of the fscal year precedng the repurchase announcement. 7 Lastly, we nclude frm sze, SIZE, defned as the log of total assets at the end of the fscal year pror to the repurchase announcement, as a control varable snce Jagannathan et al. (2000) fnd that repurchasers tend to be smaller frms. However, gven our prevous predctons that ASR frms wll have hgher dvdend yelds and lower growth than OMR frms, we expect SIZE to be postvely assocated wth the ASR decson. Our fnal model s as follows: Pr( ASR ) = β + β DILUTION β1 + β 2 + β 3 & STRING + β DE 8 DUR + β DIVYIELD 9 R D + β LABOR + β SGROWTH + β BONUS + β STKVOL + β BHAR + β SIZE + β FCF + ε where denotes frm. We predct postve coeffcents on BONUS, DE, DIVYIELD,

19 STKVOL, FCF, and SIZE and negatve coeffcents on STRING, DUR, R&D, LABOR, SGROWTH, DILUTION, and BHAR. V. SAMPLE SELECTION We dentfy our sample of ASR frms by conductng key word searches on Factva and the SEC s EDGAR database for the term accelerated share repurchase. Our ntal search over yelded 109 frms that had engaged n one or more ASR durng ths tme perod. Consstent wth reports n the fnancal press that state that the prevalence of ASRs has recently ncreased dramatcally, we note that we could dentfy only sx ASRs pror to 2004; we therefore lmt our focus to the perod. To obtan our control sample of frst-tme open market repurchasers, we conducted a search on the SDC Platnum database over and dentfed 1,739 repurchase transactons. We elmnated the followng observatons: non-asr or OMR transactons, duplcate repurchases by the same frm, frms wth no proxy statements, and frms wthout the requred Compustat and CRSP data. 8 The fnal sample conssts of 84 ASR and 591 OMR frms. We provde more detal on sample selecton n Table 1. VI. RESULTS Unvarate Tests Table 2 presents the results from unvarate comparsons of frm characterstcs across the ASR and OMR subsamples. In general, both mean and medan dfferences are sgnfcantly dfferent from zero n the predcted drecton. As expected, OMR frms have a longer seres of havng met or exceeded last year s quarterly reported EPS. Mean 17

20 (medan) STRING s (4) quarters for OMR frms versus (2) for ASRs, and both mean and medan dfferences are hghly sgnfcant (p= and p=0.0003, respectvely). Mean mplct clams by stakeholders, as proxed by DUR, R&D, and LABOR, are also sgnfcantly lower for ASR frms at the p=0.0599, p=0485, and p= levels, respectvely. Mean (medan) sales growth (SGROWTH) s 9.8 percent (6.9 percent) for ASR frms versus 18.2 percent (13.9 percent) for OMR frms; both mean and medan dfferences are sgnfcant at the p= level. These results are consstent wth H1, whch predcts that captal market ncentves play a more mportant role for OMR frms. Our results also provde evdence consstent wth H2, whch predcts that compensaton ncentves play a more mportant role for ASRs than OMRs. Mean BONUS s for ASRs versus for OMRs; ths dfference s sgnfcant at the p= level. The unvarate results also reveal that ASR frms have sgnfcantly lower mean and medan DILUTION (0.026 and 0.013, respectvely) than OMR frms (0.038 and 0.020, respectvely). Ths fndng s consstent wth the results reported by Bens et al. (2003) and Kahle (2002), who fnd that OMRs are used to offset dluton from employee stock optons plans. ASR frms have sgnfcantly hgher debt-to-equty ratos and dvdend yelds than OMR frms. Mean (medan) DE s (1.930) for ASR frms versus (1.181) for OMR frms, whle mean (medan) DIVYIELD s (0.013) for ASR frms versus (0.002); all dfferences are sgnfcant below the 0.01 level. These results provde some evdence for sgnalng arguments ASR frms already have 18

21 hgher debt ratos and dvdend yelds are therefore may have exhausted these choces as potental sgnals of good future performance. Contrary to our expectatons, however, we fnd that stock return volatlty (STKVOL) s sgnfcantly lower, not hgher, for ASR frms versus OMR frms. Mean (medan) STKVOL s (0.207) for ASR frms versus (0.310) for OMR frms, whch s not consstent wth frms usng ASRs to hedge stock prce fluctuatons. Pror stock prce performance (BHAR) s sgnfcantly hgher, not lower, for ASR frms. Mean (medan) BHAR s (-0.003) for ASR frms versus (-0.049) for OMR frms. Ths fndng s nconsstent wth the dea that frms use ASRs to correct market undervaluaton. In addton, we fnd no sgnfcant dfference n free cash flows (FCF) between ASR and OMR sample frms. Fnally, we fnd that ASR frms are sgnfcantly larger than OMR frms; dfferences n both mean and medan SIZE are sgnfcant at the p= level. Ths may reflect the fact that larger frms are more lkely to already have establshed relatonshps wth nvestment banks, whch would enable them to negotate the ASR contracts more quckly and easly than smaller frms. Larger frms may also be more lkely to have the avalable assets to repurchase a large block of stock n a sngle transacton. Multvarate Tests Table 3 presents Pearson and Spearman correlaton coeffcents for the ndependent varables. The strongest correlatons are between STKVOL and SIZE (Pearson ρ = and Spearman ρ = ) and between DE and DIVYIELD (Pearson ρ = and Spearman ρ = 0.546). In general, both STKVOL and SIZE tend to be 19

22 sgnfcantly assocated wth many of the other ndependent varables. We address potental multcollnearty problems assocated wth these varables n our probt analyss. Table 4 presents the results of a multvarate probt analyss n whch we examne the role of compensaton and captal market ncentves n determnng the decson to undertake ASRs versus OMRs. We present fve models. The frst uses our full sample of 675 observatons and omts the free cash flow (FCF) varable; the second s based on a slghtly smaller sample of 578 observatons but ncludes FCF. In the last three models, we alternatvely drop SIZE and STKVOL from the analyss to allevate multcollnearty concerns, as noted above. Of our captal market ncentve varables, STRING and LABOR are sgnfcantly negatvely assocated wth the decson to undertake an ASR versus an OMR, and SGROWTH s at least margnally sgnfcant n three out of the fve model specfcatons. Overall, these results are consstent wth H1, where we predct that captal market ncentves play a less mportant role n ASRs than n OMRs. We also fnd emprcal support for H2, where we predct that compensaton ncentves play a more mportant role n the decson to undertake an ASR versus an OMR. As expected, BONUS s sgnfcantly postve n all fve models, ndcatng that frms that explctly lnk managers annual bonuses to reported EPS are more lkely to accelerate ther share repurchases to mprove ths fgure than are frms that repurchase stock on the open market. In addton, we fnd no evdence that dfferences n the need to offset dluton from stock optons or other dlutve securtes play a role n the ASR versus OMR decson the estmated coeffcent on DILUTION s not sgnfcantly dfferent from 20

23 zero. We also control for sgnalng effects by ncludng DE and DIVYIELD as ndependent varables. As stated earler, we expect these varables to be postvely assocated wth the ASR decson f ASRs are meant to serve as a sgnal of good future performance. Our results ndcate that the estmated coeffcents on DE s sgnfcant n two of fve cases, and the estmated coeffcents on DIVYIELD are nsgnfcantly dfferent from zero, ndcatng that sgnalng consderatons do not play a promnent role n the share repurchase choce. Contrary to our expectatons however, both BHAR, the buy-and-hold abnormal returns pror to the share repurchase announcement, and STKVOL are sgnfcant determnants of the repurchase structurng decson, but n drectons opposte to our prors. If frms undertake ASRs to correct market undervaluaton, we would expect a negatve coeffcent on BHAR, but, as shown n Table 4, the estmated coeffcent on BHAR s at least margnally sgnfcantly postve n all fve models, whch ndcates that ASR frms have better pror stock prce performance n the perod leadng up to the repurchase announcement than do OMR frms. That s, there s no evdence that ASRs are the preferred repurchase structure to correct undervaluaton. We also fnd that STKVOL s strongly sgnfcantly negatve n Models 1-3. If frms use ASRs to hedge aganst fluctuatng prces that affect the cost of the stock repurchase, we would expect a postve coeffcent on STKVOL. The negatve coeffcents nstead suggest that STKVOL may be an addtonal proxy for captal market ncentves, as noted n Secton 4. Lastly, we predcted that ASR frms are more lkely to be larger n sze and have hgher free cash flows. As predcted, SIZE s sgnfcantly postve n all fve models, but FCF does not appear to be a sgnfcant determnant of repurchase structure. 21

24 Overall, our results support both our hypotheses: H1, that captal market ncentves play a strong role n open market repurchase decsons than they do for accelerated share repurchases; and H2, compensaton ncentves play a more mportant role for accelerated share repurchases than they do for open market repurchases. These fndngs are robust to controllng for sgnalng arguments, as well as for other known determnants of the decson to repurchase stock. 9 VI. ADDITIONAL ANALYSES Tender Offers In secton III, we argue that because the fnancal reportng effects of ASRs on dluted EPS occur mmedately, the motvatons for undertakng ASRs dffer from those for OMRs. However, t can also be argued that tender offers are another way to structure share repurchases to acheve smlar fnancal reportng effects as ASRs, as the reducton n shares outstandng also occurs over a relatvely short perod of tme (a tender offer typcally expres after one month, though the offer s sometmes extended f the desred number of shares were not tendered). We therefore provde an addtonal analyss n whch we compare the determnants of ASR and tender offer repurchases. We collected our tender offer sample from the SDC Platnum database for the perod After droppng multple tender offers, we arrved at a sample of 131 frms. We then drop frms that also engaged n an OMR or ASR and frms that do not have the requred Compustat, CRSP, and proxy statement data and are left wth 45 tender offer frms. Our results based on our comparson of tender offers and ASRs are presented n Table 5. 22

25 In Panel A, we examne dfferences n means and medans for the same set of varables we examned n Table 2. We observe that whle most of our tests were sgnfcant n Table 2, where we compared ASRs and OMRs, we report fewer sgnfcant dfferences n comparng ASRs and tender offers. These fndngs are not surprsng gven our overall focus on contrastng the motvatons of ASRs versus OMRs. Nonetheless, we document a number of notable dfferences between ASRs and tender offers. For example, we fnd a dramatc dfference n the percentage of frms that reward ther executves on EPS performance. Whle 58.3 percent of ASR frms menton EPS measures n ther bonus plans, only 13.3 percent of tender offer frms do. Ths dfference s hghly sgnfcant (p=0.0001) and suggests that t s unlkely that tender offer frms are structurng a stock repurchase to affect bonus compensaton. We also fnd that tender offer frms tend to be much smaller n sze (p=0.0001) and tend to have poorer stock prce performance n the perod pror to the repurchase announcement. Mean (medan) BHAR s (-0.028) for tender offer frms and (-0.003) for ASR frms; both dfferences are margnally sgnfcantly dfferent from zero, whch suggests that undervaluaton may be a motve behnd the tender offers. We also fnd that ASR frms are more hghly leveraged, wth a mean (medan) debt-to-equty rato of (1.930) versus (1.103) for tender offers frms (p= and p= for means and medans, respectvely). Stock prce volatlty s sgnfcantly lower for ASR frms, wth mean (medan) STKVOL of (0.207) for ASRs versus (0.262) for tender offer frms (p= and p= for means and medans, respectvely). Panel B presents the multvarate probt regresson results. The dependent varable s an ndcator varable that equals one f the repurchase s structured as an ASR and zero 23

26 f t s a tender offer. We present the full model n Model 1 and drop SIZE as an ndependent varable n Model 2 to allevate multcollnearty concerns. BONUS s sgnfcantly postve n both models, whch s consstent wth ASRs beng drven by compensaton motves whle there s no evdence of a compensaton motve behnd tender offers. Whle R&D s sgnfcantly postve n both models, we fnd lttle evdence of dfferences n captal market ncentves between ASRs and tender offers STRING, DUR, and LABOR are not sgnfcant determnants n ether model, and SGROWTH has a negatve sgn n Model 1. Model 1 shows that ASR frms are sgnfcantly larger than tender offer frms, whch s not surprsng gven that tender offers are often used n the decson to take a frm prvate. In Model 2, where SIZE s dropped as an ndependent varable, we fnd that ASR frms tend to have hgher debt-to-equty ratos, dvdend yelds, and stock prce performance pror to the repurchase announcement and lower stock prce volatlty relatve to tender offer frms. These results suggest that ASR frms may be usng the repurchase as a sgnalng devce whle tender offer frms may be attemptng to correct undervaluaton wth the repurchase. Overall, t appears that compensaton ncentves related to EPS-based cash bonuses do not play a sgnfcant role n the decson to undertake tender offers but are mportant for ASRs. We also fnd lttle evdence that captal market ncentves vary systematcally between ASRs and tender offers. Settlement Costs To further nvestgate the consequences of undertakng ASRs, we collected nformaton on the settlement costs of the ASR forward contracts from frms 10-K or 10-24

27 Q flngs. Ths nformaton was dsclosed by 45 of the ASR frms n our sample. Of these 45 ASR frms, eght frms chose to settle the forward contract usng shares and 37 chose to settle n cash. It s nterestng to note that of the 37 frms that chose to settle n cash, 33 frms had an ncrease n share prce over the contract perod and therefore had a further oblgaton to the nvestment bank whle none of the eght frms that chose to settle n stocks had an ncrease n stock and therefore were not requred to ssue addtonal stock to the nvestment bank. The remanng four cash settlement frms and eght stock settlement frms receved remuneraton from the nvestment bank. Table 7 contans detaled nformaton on settlement costs. EITF Topc No. D-72 and SFAS 128 provde gudance on determnng the mpact of contract settlement on the calculaton of dluted EPS. If the contract provdes the company wth a choce of net cash settlement or settlement n shares, settlement n shares s assumed. Ths presumpton may be overcome f past experence or a stated polcy provdes a reasonable bass to beleve that the contract wll be pad partally or wholly n cash (SFAS 128, 29). The pattern of contract settlement observed n Panel A s consstent wth companes choosng to settle n cash to avod ssung addtonal shares that further dlute EPS. Ths fndng has mplcatons for standard setters as they currently delberate on dsallowng frms to presume cash settlement f the effect s more dlutve. From Panel B, we note that frms that settled n cash engaged n larger valued share repurchases. The average dollar amount of the repurchase s $462.6 mllon for cash settlers compared to the $359.7 mllon for share settlers. Smlarly, the average number of shares repurchased by cash settlers s 12.6 mllon shares compared to 7.6 mllon shares. The mean magntude of the repurchase deals relatve to the number of 25

28 shares outstandng reflect the same relatonshp, 4.9 percent versus 3.5 percent for cash and share settlers, respectvely. These relatonshps do not appear to be sgnfcantly dfferent when examnng the medans; cash and share settlers are approxmately equal n sze and magntude. On average, frms pay an addtonal $12.9 mllon n cash to settle the forward contract whle the medan cash settlement amount s $7 mllon. Frms receve an average of an addtonal 1.7 mllon shares when they choose to settle the forward contract n shares. The mean (medan) addtonal costs ncurred s approxmately 5.7 (3.9) percent of the orgnal ASR deal for cash settlers and for share settlers, the mean (medan) addtonal shares receved s approxmately 4.2 (4.2) percent of the orgnal number of shares repurchased. Panel C shows that the share repurchase deal sze n terms of dollar amount and number of shares repurchased s greater for frms whose share prce subsequently decreased. However, frms whose prce ncreased repurchased a larger proporton of ther shares outstandng (5.5 percent versus 3.5 percent). For frms whose share prce ncreased over the forward contract perod, the average addtonal cash outlay to settle the contract s $17.8 mllon, representng approxmately 6.7 percent of the repurchase deal value. Conversely, for frms whose share prce decreased, the addtonal average cash receved s $2.3 mllon, representng 4.2 percent of the repurchase deal value. Overall, the descrptve evdence presented n Table 7 suggests that settlement of the forward contract can be a costly exercse for ASR frms. 26

29 Effect on Executve Cash Compensaton Gven the above fndngs that compensatng executves on EPS rases the lkelhood that frms wll choose an ASR over an OMR and that settlement costs of ASRs can be hgh, we explore the ssue of whether compensaton commttees make any adjustment to reported EPS n determnng executve pay. We estmate a cross-sectonal model n whch CEO cash compensaton for the fscal year of the repurchase s regressed on the change n reported EPS and the market value of equty. Followng Healy et al. (1987), we transform both cash compensaton and frm sze by takng natural logs of both varables so that ther dstrbutons more closely approach lnearty. Our models are as follows: log(cash Compensaton t ) = β 1 +β 2 EPS t + β 3 SIZE t + ε log(cash Compensaton t ) = β 1 +β 2 EPSASR t + β 3 SIZE t + ε. Cash compensaton s defned as the sum of annual salary and bonus; ΔEPS s defned as the dfference between current and pror year dluted EPS, dvded by the absolute value of pror year dluted EPS; ΔEPSASR s defned as the dfference between dluted EPS for the current year wth the denomnator adjusted for the tme-weghted number of shares n the ASR and pror year dluted EPS, dvded by pror year dluted EPS; and SIZE s the (logged) market value of equty as of the fscal year-end. We compare adjusted R 2 s across the two specfcatons usng a Vuong Z-test. If compensaton commttees adjust reported EPS for the effects of the ASR, we should observe a hgher adjusted R 2 n the second model. We obtan cash compensaton data from Execucomp when avalable; otherwse, we hand-collect the nformaton from frms proxy statements fled wth the SEC. We 27

30 examne both the CEO s cash compensaton, as well as the sum of cash compensaton for the top fve executves of the frm. Results are presented n Table 7. In Panel A, we present descrptve statstcs on cash compensaton and EPS. Mean (medan) CEO cash compensaton n the fscal year of the ASR s $1.965 ($1.343) mllon, whch s slghtly hgher than the mean (medan) CEO cash compensaton of $1.707 ($1.120) mllon for Execucomp frms durng our sample perod. The average CEO bonus comprses one-thrd of total cash compensaton, whch s comparable to the average for Execucomp frms. Average cash compensaton for the top fve executves s slghtly hgher than the Execucomp average of $4.823 mllon. The average bonus for the top fve executves n ASR frms comprses 32 percent of total cash compensaton compared to 36 percent for all Execucomp frms. The average cash compensaton of the mean (medan) dluted EPS s $2.647 ($2.425), whch ndcates that ASR frms are farly proftable. When dluted EPS s adjusted for the effects of the ASR, the mean (medan) fgure s $2.551 ($2.394). The mean (medan) percentage effect of the ASR on dluted EPS s 3.1% (1.7%), whch s relatvely large. In addton, mean (medan) annual growth n dluted EPS s 19.7 percent (11.5 percent), but would be 15.9 percent (8.6 percent) f the ASR had not taken place. Panel B presents the results from the OLS regressons. The adjusted R 2 s are slghtly hgher n Models 1 and 3, where the earnngs change s based on reported dluted EPS fgures, than n Models 2 and 4, where the change s based on a dluted EPS fgure adjusted for the ASR. Vuong Z-tests comparng adjusted R 2 s from Models 1 and 2 and from Models 3 and 4 are not sgnfcant at conventonal levels. Based on these fndngs, t 28

31 does not appear that compensaton commttees adjust reported EPS for the effects of the stock repurchase. Market Reacton to Repurchase Announcements As an addtonal test, we also examne the market reacton to ASR and OMR announcements. Pror research documents that repurchase announcements are typcally accompaned by postve stock market reactons. However, Bany and Mathew (2007) argue that because many OMR plans are never completed (Le 2005; Stephens and Wesbach 1998), we should expect a more postve reacton for ASR than for OMR announcements. We fnd mean two-day market-adjusted announcement returns of 0.9 percent for ASRs and 1.3 percent for OMRs, whch s roughly consstent wth Bany and Mathew s (2007) reported means of 0.7 percent and 1.1 percent for ASRs and OMRs, respectvely. However, we fnd no sgnfcant dfferences between ASR and OMR returns after controllng for other varables that have been shown to affect announcement reactons n pror research. As shown n Table 6, we regress announcement returns on the type of repurchase (ASR versus OMR), sze of repurchase as measured by the percentage of shares outstandng repurchased, frm sze, stock prce performance pror to the repurchase announcement, sales growth, leverage, and stock prce volatlty. We fnd that smaller, undervalued frms experence more postve announcement returns, as do frms undertakng larger repurchases; we fnd no other sgnfcant determnants of announcement returns. In summary, our fndngs suggest that there s no dfferental market response to the ASR versus OMR announcement. 29