The effects of tender offers on target firms market value: case Sweden

Transcription

1 Master Thess n Fnance The effects of tender offers on target frms market value: case Sweden Supervsor: Hossen Asgharan Authors: Danel Ekholm Johan Wolkesson Persson

2 Ttle: The effects of tender offers on target frms market value: case Sweden Semnar date: Course: Authors: Supervsor: Key words: Purpose: Methodology: Theoretcal perspectves: Emprcal foundaton: Conclusons: Master Essay I, NEKM01, Master level course (15 ECTS) and Master Essay II, NEKM02, Master level course (15 ECTS) Danel Ekholm Johan Wolkesson Persson Professor Hossen Asgharan Tender Offers, Cumulatve Average Abnormal Returns (CAAR), Cumulatve Abnormal Return (CAR), Event Study, Regresson Analyss, Market Value and Target Frms. Investgate to what extent tender offers have created value for shareholders of target frms on the Swedsh stock market and f there are any relatonshp between specfc varables and the targets abnormal returns. The event study methodology and cross sectonal regresson analyss. The effcent market hypothess s the startng pont n our thess followed by theory concernng determnants of abnormal returns for target frms. Tender offers durng on Swedsh publc companes. We fnd that target shareholders gan approxmately 15 percent abnormal returns around the days of the announcement. Further, we fnd that the two of the varables, Tobn s Q and stock prce run-up, sgnfcantly affect the target announcement return. 2

3 1. Introducton Background Problem dscusson Purpose Thess outlne Theoretcal background A descrpton of Tender Offers The realzaton of shareholder value market effcency Revew of target performance Determnants of Target Frm Returns Method of Payment Tobn`s Q Synergy Hypothess vs. Informaton Hypothess Successful vs. Unsuccessful Tender Offers Idosyncratc volatlty Stock Prce Run-Up Strategc buyers vs. Prvate Equty Frms Control Varables Hypotheses Target Performance n Tender Offers on the Swedsh market Determnants of target performance Method of payment Tobn s Q Successful vs. Unsuccessful Tender Offer Idosyncratc Volatlty Stock prce run-up Strategc buyers vs. Investment company buyers Method The Event Study Research approach Relablty Valdty Explanatory Regressons Descrpton of varables Dependent varable Cumulatve Abnormal Return (CAR) Explanatory varables The Regresson Model Emprcal fndngs CAAR Cumulatve average abnormal return Explanatory regressons Descrptve statstcs Regressons Analyss of the hypotheses Analyss Other fndngs Concludng remarks Conclusons Further Research

4 References Artcles Books Other references Databases Appendx

5 1. Introducton 1.1 Background Global merger and acqustons actvty n 2010 s off to ts worst start n sx years (Howley, 25/6 2010). Snce 2007, M&A s peak year, the economc crss has depressed the M&A actvty around the world. Addtonally, fears over soveregn debt and a pro-longed recesson n Europe have made M&A actvty n the regon the lowest n a decade (Howley, 2010). However, European prvate equty actvty has rebound, nvestment ncreased by 18 percent compared to the last sx months. The Nordc regon was especally strong wth a 23 percent ncrease durng ths perod (Ward, 2010), whch mght mply that there s some postve momentum n the market. The low M&A actvty complcates thngs for fund managers and nvestors who nvest n companes nvolved n a merger or an acquston. Merger arbtrage hedge fund managers face dffcultes fndng enough deals to keep ther portfolos dversfed (Yang and Branch, 2001). As deal actvty wll ncrease, more and more bds wll be made on possble targets. Thus, t s relevant to conduct further research on the behavour of the target stock around the announcement of a takeover and what mght nfluence the movement of the stock. After the announcement of a merger or acquston the target s stock prce generally adjusts upward. Ths s usually because the bd for the frm s above the stock s market value. Value-creaton to targets owner s a well documented phenomenon, however the gans for the target shareholders vares consderable. Further, there are some possble target or deal characterstcs that could explan the abnormaltes n target returns around a bd announcement (Croc, Petmezas and Travlos, 2009). As an example, Huang and Walklng (1987) propose that the type of acquston, the method of payment and manageral resstance sgnfcantly shape the target returns. Consderng that a new merger wave mght start after the recent fnancal crss, we fnd t nterestng to nvestgate target stock around a tender offer. 5

6 1.2 Problem dscusson Merger and acqustons s a subject often debated and also a topc subject to extensve research wthn the feld of fnance. A common settng n the feld of research has been to nvestgate whether or not a deal s value creatng. The hypothess tested s f the shareholders of the target earn abnormal returns from mergers and acqustons. For a revew of the research see for example Jensen and Ruback (1983), Datta, Pnches and Narayanan (1992), Bruner (2002) and Campa and Hernando (2004). All conclude that target shareholders earn abnormal returns, when proposed a bd. Even though these studes vary regardng tme perod, type of deal and observaton perod, they all show that target shareholders on average earn abnormal returns. These surveys have shown that target shareholders gan an average abnormal return n the percent range. Further, most of the studes fnd that abnormal returns occur n the days followng the announcement. However, a recent study by Wong and Cheung (2009) over the perod on target frms n Asa fnds no abnormal return to target frms around the announcement perod. As our thess wll focus on tender offers we present the results of three dfferent artcles, below, that only nvestgates tender offers whch s n contrary to the artcles above who nvestgates all types of mergers and acqustons. Jarrell and Poulsen (1989a) and Lang, Stulz and Walklng (1989), ncludng only successful tender offers, found that target companys shareholders earn sgnfcant cumulatve abnormal returns n the percent regon usng dfferent event wndows. Further, Smth and Km (1994) fnd the same results usng a sample of both successful and unsuccessful tender offers. Further, research wthn ths feld has also focused on what could explan the dfferent return patterns. The varables ncluded n the dfferent studes could be sorted nto the followng four categores n accordance wth Haleblan, Devers, McNamara, Carpenter and Davson (2009); deal characterstcs, manageral effects, frm characterstcs and envronmental factors. An example of deal characterstcs s the method of payment, and how t affects target companes returns n conjuncton wth a takeover. Draper and Paudyal (1999) study mples that target companes approached wth the opton of cash or shares, earn hghest excess return followed closed by pure cash bds. The lowest excess return occurs f the method of payment 6

7 by the bddng frm s shares. Huang and Walklng (1987) fnds that cash offers gve sgnfcantly and substantally hgher returns to target shareholders. However, Suk and Sung (1997) fnds that there s no dfference n target abnormal returns between a cash offers and a share offers. Another deal characterstc s the deal type; t has been shown and argued by Jensen and Ruback (1983) that tender offers outperform mergers. Datta et al. (1992) also draws the same concluson that tender offer s to prefer over mergers, whenever possble. Further, the researchers fnd that target company managers can maxmze gans for ther owners by avodng beng acqured by a bddng company n a dfferent ndustry. The varable that has been tested s f the deal s conglomerate or non-conglomerate. In the category manageral effects, there has been research testng the relatonshp between ownershp structure and target returns. Bauguess, Moeller, Schlngemann and Zutter (2009) fnd that manageral ownershp has a postve and sgnfcant relatonshp wth target returns and outsde ownershp has a negatve and sgnfcant relatonshp wth target returns. Further, Song and Walklng (1993) and Stulz, Walklng and Song (1990) fnd that target returns are ncreasng n manageral ownershp and decreasng n nsttutonal ownershp for a sample of multple bdder tender offer contests. Both Stulz et al. (1990) and Bauguess et al. (2009) fnd that target returns seem to be best explaned by the takeover antcpaton hypothess. A further subject that has been explored s for example the relatonshp between target returns and manageral experence and compensaton. Investgatng frm characterstcs, Croc et al. (2009) fnd that hgh dosyncratc volatlty targets receve substantally hgher announcement returns compared to targets wth low dosyncratc volatlty. Other varables that have been researched upon are for example the relatonshp between hstorcal performance and frm sze wth target announcement returns. Bauguess et al. (2009) fnd that target book-to-market and targets market adjusted stock performance has a sgnfcantly postve respectvely negatve relatonshp wth target returns. Schwert (2000) fnds that there s a negatve relatonshp target frm sze and bd premums. Envronmental factors nclude varables such as dfferent merger waves, tme perod and regulatons. Croc et al. (2009) controls for the dot com effect and fnd that target announcement returns are hgher for bds announced durng the perod than other 7

8 bds. The above presented factors have all been nvestgated n order to explan target returns around an announcement of a bd. Ths s a short overvew and there are of course more varables that have been tested n conjuncton wth target returns. To our knowledge, there are no research papers publshed nvestgatng these ssues on the Swedsh market. Even though the frst ssue mentoned above has been thoroughly examned, we fnd t nterestng to see f we can draw the same concluson on the Swedsh market. Further, the research on dfferent varables s not so comprehensve and some studes show dfferent results. Thus, t wll be mportant to see what a study on the Swedsh market wll reveal. 1.3 Purpose The purpose of our thess s two ted. Frst we nvestgate to what extent tender offers have created value for shareholders of target frms on the Swedsh stock market. Secondly, we nvestgate f there are any relatonshp between specfc varables and the targets abnormal returns. 1.4 Thess outlne Ths thess ncludes seven chapters startng wth chapter 2, the theoretcal background for our thess. Chapter 3 descrbes our hypotheses followng from chapter 2. Chapter 4 outlnes the method used and the methodologcal problems we have encountered. In chapter 5 we dsplay our emprcal fndngs. Chapter 6 ncludes analyss of our fndngs. In the last chapter, chapter 7, we conclude our fndngs and present some subjects that could be of nterest for future research. 8

9 2. Theoretcal background In ths chapter we descrbe a tender offer and revew prevous research for studes examnng returns around the announcement of a tender offer as well as studes of determnants for the target announcement returns. 2.1 A descrpton of Tender Offers A tender offer s a way to acqure a target company s shares and s usually mplemented when negotatons beng held wth the target company s management has faled. Snce a tender offers provdes the possblty of ganng control over a company even though the management opposes the offer. If so, the offer s consdered as a hostle offer. The possblty of a company becomng a target n a hostle takeover can be seen as somethng postve for the shareholders. Snce ths possblty s somethng that target management s reluctant to and thereby they try to avod t by maxmzng the wealth of ther shareholders, hence, agency costs are mnmzed (Gaughan, 2007, pp ) Tender offers usually endure hgher costs for the acqurer due to publcaton costs and legal costs. Further, once a tender offer s presented the target company mght attract the attenton of other companes that are also lookng for a target company, whch can lead to an aucton, and thereby make the premum pad hgher than ntally expected and the company that orgnally ntated the deal mght not end up as the acqurer (Gaughan, 2007, pp ). Apart from the target and acqurng company there are several other players that have a role n the process of a tender offer. These players nclude an nvestment bank that helps the bdder wth fnancng and other advsory servces n the process, legal advsors, nformaton agent, depostory bank and a forwardng agent (Gaughan, 2007, pp ). Sometmes the bddng frm uses a two-tered offer n order to acheve a faster acceptance from target shareholder. Ths type of offer s usually front-end loaded whch means that there s greater payoff to target shareholders that tenders n the frst step and a lower payoff f they choose to tender durng the second step of the offer. The ratonale behnd ths type of offer s 9

10 that f a large enough number of shares s tendered n the frst step then the remanng shareholder wll be excluded and nstead they wll receve a compensaton that s consdered to be nferor, such as debentures nstead of cash (Gaughan, 2007, pp ). Although there s a framework for countres n the European Unon most countres stll have ther own natonal rules. The framework controllng tender offers n Sweden s extensve and the man ssues wll be summarzed below. An acqurng company s only allowed to ntate an offer f they have the ablty and purpose of closng the deal; ths s so because the announcement of an offer wll most lkely affect the share prces of the partes nvolved. Durng the due-dlgence process the bddng company s not allowed to trade n target stocks snce they mght obtan prvate nformaton durng ths process (NASDAQ OMX Stockholm, 2009). Once the bddng company has decded to place a bd on a target company they should communcate ths through a publc announcement, also, f the bddng company has reason to beleve that there has been some knd of nformaton leakage they should also communcate ths to the publc so that everyone has the same nformaton. The bd beng communcated must be equal to shareholders holdng the same type of shares and the bd should also be predetermned. Often, the offer s contngent on the tender of a fx mnmum number of shares. Target shareholders have no more than 12 weeks to decde f they want to tender the offer or not. Shareholder wth the same type of shares must be gven the same offer and once they have decded to tender they can t wthdraw ther tenderng. Target company board of drectors should gve ther opnon about the offer no later than two weeks before the accept perod ends (NASDAQ OMX Stockholm, 2009). 2.2 The realzaton of shareholder value market effcency Target performance has n prevous research studes typcally been measured usng an event study methodology. Based on theoretcal arguments a pre-specfed event-wndow s developed. Both short-and long run event wndows are used frequently see for example Bruner (2002) and Campa and Hernando (2004). When tryng to conclude when all of the effects of a deal have been fully ncorporated n the share prce the level of market of 10

11 effcency that one assumes s mportant. We wll therefore revew the effcent market hypothess and the emprcal evdence of t. The effcent market hypothess states that t s not possble to obtan abnormal returns for an nvestor, snce the stock prce fully reflects all avalable nformaton (Fama, 1965). In a later artcle (Fama, 1970) ntroduces a new defnton of market effcency. The author defnes a market as beng effcent f stock prces reflect all avalable nformaton. Further, he presents three levels of market effcency; strong, sem-strong and weak. Under the weak form, today s securty prces reflect all nformaton contaned n hstorcal prces. Under the sem-strong form of market effcency, securty prces mmedately adapts to new publcly avalable nformaton. Thus, stock prces wll adjust for an announced tender offer. Under the strong form, all nformaton (both prvate and publc) s reflected n the stock prce. Thus, nsders cannot expect to earn an abnormal return. Further, a tender offer announcement should not affect the stock prces, snce the tender offer s already expected and ncorporated n the stock prce. Prevous studes on abnormal returns for example see Bruner (2002) assumes sem-strong market effcency and thereby assumng that stock prces react accordngly and unbased when new nformaton arrves. 2.3 Revew of target performance As mentoned earler there has been a wde range of event studes nvestgatng the relatonshp between target company returns around the announcement of a merger or acquston. When lookng at prevous studes there s strong consensus among researchers that target frms earn statstcally sgnfcant abnormal returns around the announcement of an M&A transacton. Ths consensus s strong across markets, markets beng covered n the artcles are North Amerca, Europe and Asa, across dfferent event wndows, ncludng both short and long run event wndows, and across dfferent tme perods. For a revew of ths evdence see for example Bruner (2002), Campa and Hernando (2004) and Haleblan et al. (2009). However one artcle that have not been able to observe postve abnormal returns to 11

12 target shareholder s the artcle by Wong and Cheung (2009), coverng the perod lookng at target frms n Asa, who fnds no abnormal return to target frms around the announcement perod. Although, there s consensus concernng the sgnfcance n the returns for target companes the level of return vares substantally rangng from roughly percent wth an average of rangng from percent. 2.4 Determnants of Target Frm Returns Method of Payment The method of payment n a deal s only lmted to the magnaton of the partes beng nvolved n the transacton. The most commonly used methods are, stock, cash or a mxed cash and stock offer. Stock deals nvolve the decson of determnng the prce of the stock on whch the valuaton s based on, ether one chooses a fxed exchange rate or a floatng exchange rate. When applyng a floatng exchange rate one normally uses an average prce durng a pre-determned tme perod (Gaughan, 2007, pp ). If one assumes symmetrc nformaton, no taxes and no transacton cost the method of payment shouldn t matter. However, ths s not the envronment n whch companes compete and therefore these thngs do n fact matter, (Kargn, 2001). Prevous studes have shown that the method of payment matter when determnng reasons for abnormal returns to target shareholders. Many of these studes have found that the returns are hgher when the target company are faced wth a cash bd than wth a stock bd, see for example Wansley, Lane and Yang (1983), Huang and Walklng (1987) and Asquth, Bruner and Mullns (1990). These studes explans ths dfference by usng the tax hypothess, whch assumes that shareholders who are faced wth a cash bd wll have to pay tax on the captal gan drectly whle n a stock offer they won t have to pay taxes untl they sell ther share. Hence, ths mmedate payment of taxes has to be compensated for n a cash offer. Further, Berkovtch and Narayanan (1990) argues that the hgher returns n cash offers are due to the fact that ths payment conveys postve nformaton about the synerges that can be obtaned, 12

13 nformaton effect hypothess. However, both these hypothess yelds the same concluson, cash offers should yeld a hgher return to target shareholders. In contrast to these artcles Suk and Sung (1997) fnds no dfference n target abnormal returns dependng on the method of payment. Fnally, Draper and Paudyal (1999) fnd that target frms realze the greatest returns when the shareholders are gven the opton to choose between stock and cash. To summarze prevous research there s some consensus that cash bds yeld hgher return to target shareholders Tobn`s Q When decdng f a target company should be classfed as a value creatng company wth good growth opportuntes an often used measurement s Tobns Q see for example Dong, Hrshlefer and Rchardson (2006), Lang et al. (1989) and Servaes (1991). Tobns Q s defned as the rato between the market value of a frm s equty n relaton to the replacement cost (Lndenberg and Ross, 1981). Dong, Hrshlefer, Rchardson and Teoh (2006) argues that Prce/Book Value s a good proxy for Tobn s Q, because of the fact that the book value s a reflecton of the hstorcal costs whereas the prce or market value of equty s a forward lookng measurement. Further, n support of the Prce/Book rato they argue that the predctng power of Book/Prce ratos has been strong when analyzng the cross secton of subsequent one month returns. Also Prce/Book rato has been proven to be a good approxmaton of msprcng when one puts t n the content of theoretcal models based on psychology. Ther study conssts of approxmately 3700 M&A deals n the U.S durng the years The authors found that the abnormal returns for target shareholder are hghest when the target has a low Prce/Book rato, Tobn s Q. Ths fndng s consstent wth the nformaton hypothess that companes that are undervalued, msprced, wll see a correcton due to the takeover bd. However, f one adopts the framework behnd Tobn s Q, that takeover gans are due to assets beng reallocated, one can state that the abnormal returns to target frms wth low Prce/Book rato s due to the fact that there s a larger room for mprovement and value creaton n these companes. Ths greater mprovement s somethng that can be shared between the shareholder of target frm and the acqurng frm. Servaes (1991) uses a sample of 704 M&A 13

14 deals on the US market and measures abnormal returns to the partes n a transacton and he fnds that target frm abnormal returns are greatest when the frm has a low Tobn s Q, 32.7 percent, and s 13 percent less f the target company has a hgh Tobn s Q, ths s also consstent wth the fndngs of Lang et al. (1989) who uses a sample of 209 successful tender offer n the US between and fnds that target companes wth a hgh Tobn s Q on average earn 14 percent less than low Q targets. The few prevous studes that nvestgate the effects of Tobn s Q ponts n the same drecton, target companes wth a low Tobn s Q earn hgher returns than companes wth hgh Tobn s Q Synergy Hypothess vs. Informaton Hypothess Successful vs. Unsuccessful Tender Offers When two companes, beng combned, are able to create more wealth than the companes would together on a standalone bass they have been able to create synerges. Usually one dscusses two knds of synerges, operatonal and fnancal. Operatonal synerges nvolve mproved operatonal performance ether through cuttng costs or ncreasng revenues. Fnancal synerges on the other hand nvolve the step of mproved fnancal stuaton of the company through lowerng the cost of captal. The synergy hypothess requres that the two enttes are combned, mplyng a successful offer. The nformaton hypothess on the other hand states that the rght share prce of a target company s beng re-evaluated durng the bd process (Halpern, 1982). There are two sdes of the nformaton hypothess. Frst, one way of lookng at t, s that spreadng of new nformaton enables the market to prce prevously undervalued shares farly. Secondly, one could look at t as f the new nformaton enables the management of the target frm to pursue a more or hgher valued strategy of ther operatons. The concluson of the nformaton hypothess s that there s no need for the offer to be successful (Bradley, Desa and Km, 1983). Schwert (1996) fnds that successful target experence a slghtly hgher returns than unsuccessful targets, however ths dfference s not statstcally sgnfcant. For a revew of studes yeldng smlar results see Jensen and Ruback (1983). 14

15 In lne wth, Bradley, Desa and Km (1988) we don t use an arbtrary cut-off pont, nstead we defne a successful tender offer as an offer where any number of target shares are beng bought by the bddng frm. Because even a small purchase of shares can alter the votng power and thereby affect the operatons of the frm Idosyncratc volatlty Croc et al. (2009) nvestgate the relatonshp between dosyncratc volatlty (sgma) and target announcements returns. The authors argue that target companes wth greater pre-event dosyncratc volatlty receve consderably larger announcements returns relatve to low dosyncratc volatlty companes. Targets wth hgh sgma should have greater uncertanty and thus be more dffcult to value. Croc et al. (2009) argues that an nvestment n a hgh rsk project s expected to lead to larger returns for the bddng company, dsregardng behavoural and agency motvatons behnd the deal, and subsequently a hgher bd. Ths hgher bd wll subsequently gve hgher announcements returns for target shareholders'. The authors fnd that hgh sgma targets sgnfcantly outperform low sgma counterparts by a 9.34 percent return margn. Also, by further dvdng the sample by method of payment, the authors fnd that hgh sgma targets sgnfcantly outperform low sgma targets for all forms of fnancng. However, gven the nformaton asymmetry theory we would expect that hgh sgma targets would earn less than low sgma targets when stock s used as a method of payment. As mentoned by Croc et al. (2009) ther study s the frst that nvestgates the relatonshp between dosyncratc volatlty and target returns, hence, the theoretcal and prevous research beng avalable to us s lmted Stock Prce Run-Up The fact that stock prces of a target shows a sgnfcant ncrease n the weeks pror to a publc announcement s somethng that has been observed n a number of studes. Researchers have usually explaned ths run-up ether as a sgn of nsder tradng or as confrmaton of the effectve market hypothess Jarrell and Poulsen (1989b) and Keown and Pnkerton (1981). 15

16 However, t s beyond the scope of ths thess to test whether run-ups are due to nsder tradng or s a confrmaton of the effectve market hypothess. Instead we are nterested n testng how market antcpaton or takeover antcpaton affects target shareholder returns around the announcement of a deal. Croc et al. (2009) examne, among other thngs, how target abnormal returns are related to the market antcpaton of a takeover. Usng a sample of 2110 successful M&A deals n the US from The authors fnd that the run-up perod return has a statstcally sgnfcant negatve relatonshp wth target returns. Bauguess et al. (2009) also use the market adjusted run-up return for target companes as an approxmaton of takeover antcpaton. Usng a sample of 1668 takeovers n the US from 1996 to 2005 the authors also fnd that the run-up return s statstcally negatvely related to target returns around the announcement. The fndngs summarzed above are consstent wth the market antcpaton hypothess. Because a company that s consdered as beng more lkely to be a target should have showed hgher returns n the perod before the announcement and therefore the announcement effect should be smaller Strategc buyers vs. Prvate Equty Frms Durng the past decade prvate equty frms has become an mportant part of the M&A market. Durng the years of corporate takeovers by prvate equty frms rose from 6 percent to 30 percent, wth deal values rsng from $30 bllon n 2001 to $450 bllon n 2007 n the US (Boone, and Mulhern, 2008; Offcer, Ozbas and Sensoy, 2008). Wth the ncreased presence of prvate equty frms n M&A transactons, researchers have started to nvestgate the dfferences n deals that are performed by strategc and prvate equty nvestors. Studes have shown that target shareholders receve a lower premum when beng approached by a prvate equty nvestor compared to a strategc nvestor (Bargeron, Schlngemann, Stulz and Zutter, 2009; Roosenboom, Fdrmuc and Teunssen, 2009). Bargeron et al. (2009) argues that ths dfference s due to the fact that prvate equty bdders are more selectve n the prce they are wllng to pay than strategc buyers. Further, the authors argue that the management of strategc buyers s wllng to pay more for the target frm because they don't bear the full costs of ther decson. Also, management has an empre- 16

17 buldng mentalty. Studes examnng the relatonshp between the type of acqurer and target company returns show that shareholders of companes beng bought by prvate equty frms experence a lower return on ther stock Control Varables Hmmelberg, Hubbard and Darus (1999) argue that leverage can be used as a proxy for measurng moral hazard wthn a frm. Croc et al. (2009) and Bauguess et al. (2009) nclude leverage as a determnant for premums of target companes. The authors fnd a weak negatve relatonshp between target premums and leverage, beng statstcally sgnfcant at the 10 percent level. Followng the work of Schwert (2000), Offcer (2003) and Croc et al. (2009) we also nclude sze of the target company as a determnant of ther abnormal returns, sze s measured as the log of the market captalzaton for a frm. Schwert (2000) fnds a weak negatve relatonshp between premum and sze whle Croc et al. (2009) and Offcer (2003) fnds ths relatonshp as beng statstcally sgnfcant at the 10 percent level. Also, Hmmelberg et al. (1999) argue that larger frms mght be subject to problem of moral hazard and manageral dscreton. Bauguess et al. (2009) use the natural logarthm of sales as a determnant for target returns and manageral dscreton. The authors fnds a statstcally, at the 1 percent level, negatve relatonshp between target returns, measured as the three day cumulatve abnormal return around the announcement date. In lne wth Croc et al. (2009) our sample also ncludes the years when the dotcom bubble was present and to control for any abnormaltes that mght be due to the M&A wave n conjuncton wth ths bubble. To control for ths we nclude a dummy varable that takes the value 1 f the deal was posted after the year of 2001 and the value of 0 f the deal was posted pror to

18 3. Hypotheses In ths secton we present the varous hypotheses beng tested n our paper. The hypotheses are based on the theoretcal dscusson n the prevous chapter. Frst, the hypothess regardng the performance of Swedsh target companes beng approached wth a tender offer s presented. Secondly, we present our hypotheses concernng the determnants of the target companes returns. 3.1 Target Performance n Tender Offers on the Swedsh market As mentoned earler assumng sem-strong effcency n the market mples that an announcement and ts effects should be mmedately ncorporated n the prce of the company s stock. Prevous research on target performance shows consstent results, that on average a tender offer ncreases the value of the target company. Results are smlar both over tme, across markets and methods beng appled. However, from prevous research the level of the postve announcement returns vares across studes. We formulate our frst hypothess that target return s equal to zero. Followng the dscusson above we expect to reject our frst hypothess. Hypothess 1: Target abnormal returns are equal to zero. 3.2 Determnants of target performance Method of payment Prevous research has shown some consensus regardng the effects of method of payment on target returns, where cash bds has yelded the hghest returns target shareholders. Ths s due to the fact that the mmedate payment of tax when beng faced wth cash bds and because a cash bd conveys postve nformaton about synerges that can obtaned by combnng the target company wth the bdder. We now formulate our second hypothess, that the method of payment shouldn t affect the returns for target companes. Followng the theory and prevous research we expect to reject the null hypothess. 18

19 Hypothess 2: Target returns are not affected by the method of payment Tobn s Q Prevous studes have shown a negatve relatonshp between the level of Tobn s Q and the announcement returns to target shareholders. Ths relatonshp can be explaned by nterpretng Tobn s Q as a measurement of msprcng, and when a company s faced by a bd ther share prce wll correct tself. Another way of nterpretng Tobn s Q s as a measurement of possble mprovement n a company, where companes wth a low Tobn s Q has a greater room for mprovement whch can be dvded between the bdder and the target company yeldng a hgher return for target shareholders. We formulate our null hypothess: that Tobn s Q shouldn t have any effects on target returns. Accordng to the dscusson above we expect to reject ths hypothess. Hypothess 3: Tobn s Q does not have any effects on target returns Successful vs. Unsuccessful Tender Offer Studes nvestgatng the relatonshp between the successfulness of a tender offer and target returns haven t shown any statstcally sgnfcant relatonshp and we therefore expect to not reject the null hypothess Hypothess 4: The successfulness of a tender offer does not affect target returns Idosyncratc Volatlty Researchers have prevously suggested that companes experencng a hgh sgma could be consdered as beng more uncertan and more dffcult to value, thus, one would expect a hgher return for these companes. As mentoned earler there has only been one artcle that nvestgates ths subject, Croc et al. (2009), and they fnd that there s a postve relatonshp between sgma and target announcement returns. Therefore we expect to reject our null hypothess. 19

20 Hypothess 5: Idosyncratc volatlty has no effect on target returns Stock prce run-up Companes that are antcpated to be takeover targets experence hgh return n the perod precedng the announcement of a deal and lower returns around the days of the announcement, accordng to the market antcpaton hypothess. Prevous studes have confrmed ths negatve relatonshp. Therefore we expect to reject our null hypothess. Hypothess 6: There s no relatonshp between target returns and the level of stock prce runup Strategc buyers vs. Investment company buyers Several prevous studes have shown that target receves a lower premum when the bd s made by a prvate equty frm. Ths s explaned by the fact that these frms are more selectve because they bear the full cost of ther decson and doesn t have an empre buldng mentalty, whch could be the case for management of strategc buyers. Followng the dscusson above we expect to reject our null hypothess Hypothess 7: Returns of target companes should not be affected by buyer type 20

21 4. Method In ths chapter we wll provde the reader wth the methodologcal approach to our study and ts valdty and relablty. 4.1 The Event Study The research methodology most commonly used when measurng M&A proftablty has been the event-study framework (Bruner, 2002). Ths methodology has been extensvely used to test the mpact of some economc event on stock prces. Further, t s assumed that the mpact of the economc event s nstantly reflected n the stock prce. We wll apply MacKnlay s (1997) procedure when conductng our method. Event Defnton Snce the purpose of ths study s to examne the behavour of the stock prces around a tender offer, the prmary ssue s to select the event day. A majorty of prevous studes wthn ths feld have been usng the day of announcement as the event day, see for example Jensen and Ruback (1983) and Brown and Warner (1985), therefore we also to apply the announcement day as the event day. Below, n fgure 1, we present the outlne of an event study. Fgure 1. (Estmaton wndow) (Event wndow) (Post-event wndow) T 0 T 1 0 T 2 T 3 The dstance between T 0 and T 1 s the length of the estmaton wndow, denoted as L 1. The event wndow stretches from T 1 to T 2, denoted as L 2. Further, 1 to 2 s the event wndow beng nvestgated where T 1 < 1 2 T 2. Thereafter we defne our event wndow. A majorty of prevous research has chosen an event wndow that s rather short, however, the number of days ncluded vares over studes see for 21

22 example Bruner (2002) and Campa and Hernando (2004). Campell, Lo and MacKnlay (1997, p. 149) suggests that assumng sem strong effcency, whch mples that news of a tender offer wll nstantly be ncorporated n the securty s prce; hence the estmaton wndow should be estmated over a rather short perod. Another motvaton for usng a short event wndow s the possblty of contamnaton that can occur when usng a longer event wndow, meanng that the event that s beng observed s affected by other events and any conclusons beng drawn from the data are nadequate (Armtage, 1995). Based on the dscusson above and prevous research, see for example Croc et al. (2009) and Bauguess et al. (2009), we have chosen to use a three day perod when measurng abnormaltes n the stock prce behavour [-1, 1]. The motvaton for ncludng the day after the announcement s due to the fact that one wants to capture the prce effects that occur after the stock exchange closes on the announcement day (Campell et al., 1997, p. 151; MacKnlay, 1997). As prevously mentoned there s a wde varety of event wndows used. Therefore, we thnk t would be reasonable to apply dfferent wndows snce we thnk that ths would add to the valdty of our thess. The event wndows used are [-1,1], [-3,3] and [-5,5], not only wll ths enable us to capture a slower reacton from the market, due to the announcement, but we wll also be able to capture any nformaton leakage to the market pror to the announcement. Lastly, an extenson of the event wndow also means that any over reactons n the market wll be subsequently corrected. It s also common for studes to perform ther tests usng dfferent event wndows see for example Danbolt (2004), Bruner (2002) and Campa and Hernando (2004). Selecton crtera Set forth below s a descrpton of the selecton crtera beng appled when decdng f a deal should be ncluded n the sample or not. We have chosen to study tender offer announcements for Swedsh companes that are lsted on the Nasdaq OMX Stockholm stock exchange between and , snce Nasdaq OMX Stockholm doesn t provde data pror to ths perod. In order to retreve our sample of tender offers on Swedsh companes from 1999 untl present we have used the webste of Nasdaq OMX Stockholm stock exchange. 22

23 Further, snce we want to examne how shareholders wealth changes n conjuncton wth a tender offer the target frm has to be lsted on the Nasdaq OMX exchange. There are cases where the target frms are no longer traded however they were tradng durng the offer, these companes are ncluded n the sample, hence, we deal wth potental survval bas problems. From the Nasdaq OMX Stockholm we retreve an ntal sample of 219 successful and unsuccessful tender offers. From ths lst we do further reductons due to nsuffcent data. Frst, we only nclude the frst bd on a company, due to the fact that multple bds wll affect the estmaton perod beng appled. Secondly, we remove stocks that have not been lsted durng the whole tme perod, on whch we base our estmatons; because we can t estmate abnormal returns for these stocks usng the market model. Further, we also remove companes that don t provde data for the determnants beng used and we end up wth a sample of 172 successful and unsuccessful tender offers. Also, we modfed our data so that t doesn t nclude weekend days for the Swedsh market and f the announcement was on a non-tradng day the followng tradng day was used as the announcement day. Normal and abnormal returns In ths part of the study the normal returns for the companes, ncluded n the sample, are calculated. The normal return s assumed to be the return that the securty would yeld wthout the event. In order to make nferences, f there are any statstcally sgnfcant abnormal returns present, the normal or expected return s compared to the return on the securty durng the event wndow. Frst, the actual return s calculated for our sample over the event-wndow. The contnuously compounded returns, usng the last transacton prce, are calculated accordng to equaton (1) below. Due to bd-ask bounce t could be more approprate to use the average of bd-ask. However, we are not able to retreve ths data for the whole perod. The problem, that mght be present, when usng closng prce and not average of bd-ask prce, s that the closng prce s affected by the last trade. Further, there could be some varance present n the stock even though ts ntrnsc value remans unchanged. Ths n turn could yeld an upward bas when we quantfy abnormal returns (Blume and Stambaugh, 1983). The presence of ths problem s more common for smaller frms and snce we have some small frms n our sample ths 23

24 problem could be present. However, we try to mtgate ths ssue by usng dfferent event wndows as well as dfferent models when calculatng the normal return. R, t ln( P, t1 / P, t0 ) (1) Where P, t1 today s closng prce and P, t0 s yesterday s closng prce. Contnuously compounded returns are used n order to handle ssues concernng non-statonarty n the data. Prce data are gathered from the Thomson Reuters Datastream database and they are corrected for splts, dvdends and new ssues. In some cases we were not able to obtan quotes for A shares, nstead we use only the prce on B shares and do not create value weghted portfolos of the two types of shares. Ths problem arses because the founder sometmes keep A shares and only lets B shares float (Doukas, Holmén and Travlos, 2002). Thereafter the estmaton of abnormal returns has to be defned. MacKnlay (1997) suggests several dfferent methods to calculate ths; however, the two most commonly used are the market model and the market adjusted return model. Armtage (1995) summarzes the performance of dfferent estmatons models and concludes that most models yeld smlar results. The market model however always performs, accordng to the best alternatve and can therefore be seen as the most trustworthy model. Further, Brown and Warner (1985) also found that the two models mentoned above often gves the same results as more complcated models. Followng the dscusson above we thnk that the use of the two models wll ncrease the valdty of our results. If both models yeld smlar results t would mply that our estmates are relable and stable. The dfference between the market model and the market adjusted return model s that the latter model constrans alpha and beta to zero respectvely one. The market model assumes a lnear dependence between a stock s return and the market return over tme. A postve feature of the market model compared to the constant mean return model s that a part of the rsk adjusted return s beng removed from the market return; ths s so because the rsk of a stock should be captured n the stock s beta. Ths n turn leads to a decrease n the varance for abnormal returns (MacKnlay, 1997). 24

25 Below, both the expected return and the abnormal are defned for the market adjusted return model, equatons (2) and (3), and the market model, equatons (4) and (5). Where R,, s the expected return on asset at tme t, expressed as a functon of the expected market return at tme t. Where AR,, R, and t t R m, t s the abnormal return, actual return on securty at tme t and the benchmark return at tme t. and are OLS estmators beng estmated n the estmaton wndow precedng the event perod. t R R, t m, t t (2) AR, t R, t Rm, t (3) R, t Rm, t, t (4) AR, t R, t Rm, t (5) Estmaton procedure In order to estmate the normal returns we have to estmate the coeffcents beng used n the models for normal returns. In order to ths we need a proxy for the market return. We have chosen Affärsvärldens General Index (AFGX) as a proxy for the market; ths s a value weghted ndex that s adjusted for new ssues, splts and dvdends. Ths approach s n lne wth the approach of Doukas et al. (2002), Sanders and Zdanowcz (1992), Lakonshok and Vermaelen (1990). When decdng what tme perod that should be used as estmaton perod there are no clear evdence on how to choose ths perod. There are arguments both n favour for usng a hgh frequency data as well as data wth lower frequency see for example Merton (1980) and Scholes and Wllams (1977). Scholes and Wllams (1977) clams that betas measured frequently leads to based estmates due to non-synchronous tradng. Whch yeld an upward bas for stocks beng traded frequently and a downward bas for stocks beng traded 25

26 nfrequently. Ths mples that data wth lower frequency should be used such as weekly or monthly. Followng the argumentaton of Brown and Warner (1985) one s wrong to assume that the results from Scholes and Wllams (1977), who suggests that the use of OLS as estmaton model when estmatng beta and alpha n an event study, wll result n msspecfcatons. The ratonale behnd ths s that when ncludng an alpha, ntercept, one mposes the restrctons that the sum of the resduals, from the model beng used, s equal to zero. Ths modfcaton wll lead to an accurate specfcaton of the event study, whch s n lne wth the betas beng based, snce ths bas wll compensated by the bas n the alpha. Snce we are usng log returns t s far to assume statonarty, whch n turn means that the excess returns for a securty durng the event perod wll have a zero mean uncondtonal on the return of the market. It mght be suggested that a msspecfcaton of the event study has occurred due to the bas n the excess return of an ndvdual stock. However ths s wrong because the average bas should be equal to zero (Brown and Warner, 1985). Usng the theoretcal arguments dscussed above we feel that the methods beng most approprate to use are Ahern (2008) and Brown and Warner (1985). Daly returns for 238 days wll be calculated and used as the wndow on whch we base our alpha and beta estmates for the market model, [-244, -6]. Peterson (1980) states that the choce of estmaton perod s a choce between the parameter estmates beng more out of date but havng a greater accuracy, when the estmaton perod s expanded. Testng procedure In order to be able to draw concluson whether or not there are abnormal returns present n the event-wndow, calculatons of the average abnormal return are performed for every tme perod n the wndow. Aggregaton of the average abnormal return s performed for every securty, whch s the same as the securtes cumulatve average abnormal return as can be seen below n equaton (6) and (7). Ths s the same as computng the average of each securtes cumulatve abnormal return, equaton (8). N 1 AR AR (6) N 1 26

27 CAR, ) AR (7) ( N 1 CAR ( 1, 2 ) CAR ( 1, 2 ) (8) N 1 The null and the alternatve beng tested s the followng: H 0 : CAR( 1, 2 ) 0 H 1 : CAR( 1, 2 ) 0 A t-test s performed n order to make nferences about our hypothess. To make our nferences we calculate the varance accordng to equaton (9) through (11), where equaton (9) s a cross-sectonal approach whle equatons (10) and (11) are the tradtonal approach where the varance are relyng on past returns as oppose to the cross-sectonal approach. The cross-sectonal approach s used because the tradtonal approach to calculate the varance uses the estmaton perod returns when calculatng the varance. Whch mply that the null hypothess should be nterpreted as that there are no effects on the varance or the mean from the event. However, one mght argue that t s ratonale to assume that a tender offer mght ncrease the varance but not affect the mean. The cross-sectonal approach, presented n equaton (9) follows the method suggested by Campbell et al. (1997, pp ): Var N 1 2 CAR, ) CAR (, ) CAR(, ( (9) ) N 1 When usng the tradtonal approach, equaton (10) for varance calculatons we use the method suggested by MacKnlay (1997). When calculatng σ 2 there are two components n equaton (10), the dsturbance varance, and the other component s the varance that s due to the samplng error n β and α. However, f the estmaton wndow, L 1, s suffcently long ths component can be assumed to be zero, equaton (11). We choose to also calculate the condtonal varance dsplayed n equaton (10) consstng of the two components Rm m ( AR ) 1 e (10) 2 L 1 m 27

28 2 2 ( AR ) e (11) The dsturbance varance s estmated as n equaton (12). 2 (1/( L 1 2)) T 1 T0 1 ( R R ) 2 m (12) Where alpha and beta, n equaton (12), has been estmated durng the estmaton wndow, 244 days pror to the announcement untl 6 days pror to the announcement, and L 1 s the length of the estmaton perod, 238 days. Now we can use both estmaton of varance of the ndvdual securtes to calculate the varance for the average abnormal return and subsequently the varance for the cumulatve average abnormal return, see equatons below. var( AR N 1 2 ) ( AR ) 2 (13) N var( CAR( )) var( AR ) (14) When usng the cross-sectonal approach t s necessary to assume that the abnormal returns are uncorrelated. Brown and Warner (1985) confrm that ths s the fact when the event date s not same for all the securtes ncluded n the sample, as n our case. Usng equaton (9) and (14) we can test our hypothess usng the t-statstc obtaned from equaton (15): J 1 CAR(, ) Var CAR(, ) (15) An alternatve way to measure the level of sgnfcance for abnormal returns s to use a measurement called standardsed cumulatve abnormal returns, equatons (16), (17) and (18). Ths method s suggested to be used when the true abnormal return s stable over securtes, whle the use of the ordnary defnton s more approprate when the true return s hgher for 28

29 securtes experencng a hgher varance (Kolar and Pynnonen, 2008). Where SCAR s the standard devaton of cumulatve abnormal returns, usng the market model, adjusted for the forecast error. SCAR (, ) 1 2 CAR S CAR (, ) 1 2 ( 1, 2 ) (16) Where SCAR s the standard devaton of the cumulatve abnormal return. N 1 SCAR ( 1, 2) SCAR ( 1, 2) (17) N 1 1/ 2 N L1 4 J 2 * SCAR( 1, 2 ) L1 2 (18) 4.2 Research approach In our thess we have chosen to use a deductve approach, meanng that we wll formulate our hypotheses on the bass of the exstng lterature. After formulatng our hypotheses we collect the data n order to be able to test our hypotheses. Thereafter we analyse our fndngs and conclude f our hypotheses should be rejected or not. We then put our fndngs n a theoretcal framework and analyse them, here we can ether confrm the exstng theory or new theores can be formulated, hence, a more nductve approach s appled (Bryman and Bell, 2005, pp. 9-12). In ths thess a quanttatve research approach s appled n order to examne the behavour of target returns and possble determnants of ths behavour. 4.3 Relablty A research paper s consdered to have a hgh degree of relablty f t can be replcated easly. Therefore t s mportant that the procedures and methods beng used are descrbed n detal. Set forth below s a descrpton of our data collecton and methods beng appled. 29