Mergers as Auctions. Marc IVALDI University of Toulouse (IDEI), EHESS and CEPR, France

Transcription

1 Mergers as Auctons Marc IVALDI Unversty of Toulouse (IDEI), EHESS and CEPR, France Jrssy MOTIS GREMAQ-EHESS, France March 2005 Abstract Every day M&A are arranged brngng together separate companes to make larger ones. M&A can worth bllons of dollars and dctate the gans/loses of the nvolved companes shareholders, managers, employers, compettors and consumers for years to come. Most emprcal studes that evaluate the motves and gans of M&A conclude that, n average, the target has postve gans, whle the acquror, at best, does not lose from such deal. Wth a database rch n bds proposals, fnal bds and the number of compettors; a rcher approach may be used to estmate the acqurors gans from mergng. Snce a bd stuates a corporaton nto the game of an aucton, ths paper proposes to nterpret mergers as frst prce auctons n order to provde a powerful analytcal tool for evaluatng gans n M&A. It bulds on some dssatsfacton wth event studes and wth operatng performance studes. By estmatng acqurors true valuaton we are able to compute the true gans of the bdder n a frst prce aucton usng nonparametrc methods. The gans of bddng n a frst prce ndependent prvate value aucton are postve on average contrary to what s found n the event study. Keywords: mergers, auctons, event studes, corporate fnance. JEL codes: L10, L20, G14, G34, C14.

2 1. Introducton Ths paper proposes to nterpret a merger as an aucton n order to provde a powerful analytcal tool for evaluatng gans from mergng. It bulds on some dssatsfacton wth event studes and wth operatng performance studes that estmate the gans of a group of mergers wth fnancal and accountng data respectvely, but wthout any structural economc approach behnd those market models. Most of these emprcal studes that evaluate the motves and the gans n mergers and acqustons conclude that, even when the jont gans n stock prces are postve n average, the dstrbuton of gans s not symmetrc, that s, whle the target has postve gans, the acquror, at best, does not lose from such deal 1. A major proporton of those studes employ the already standard fnancal technque called event study whch conssts on computng the abnormal returns due to the merger announcement. Event studes attempt to determne the effects of mergers on the mergng frms and sometmes, on the market as a whole. However, stock market event measurements of the net returns provde a predcton of gans (losses) of the mergng frms rather than evdence that such gans (losses) actually occurred. Furthermore, the gans of mergers may not necessarly be mmedately reflected n the change of stock prces because stock market reactors may have nether the same nformaton, nor the same long-run perspectve than the mergng frms do when jonng ther sets of assets. Only nsders can antcpate when/how ther partcular bundle of assets operated separately can be combned n new ways to generate addtonal value. Thus, as market reactors may not posses the same nformaton to estmate the true 1 Andrade, Mtchell and Stafford 2001, fndngs reveal that the value-weghted average of the two frms return s postve, wth most of the gans accrung to the target company. 1

3 gans of the game as nsders do; outsders computatons may dffer from those of the players of the merger, for nstance, because of underestmaton of true synerges, whch mght be the reason for fnancal emprcal studes concludng that acqurors do not gan from mergng. In a prevous work, based on the event-study methodology, we have found these classcal results,.e., postve jont abnormal returns, postve abnormal returns for targets and negatve abnormal returns for acqurors none of them sgnfcant. Bearng n mnd that stock prce studes may be unable to provde evdence on the gans of mergng and on the source of any merger-related gans n the short run 2, an alternatve assessment of the mergng gans based on an accountng approach has also been largely performed, t s called operatng performance study. An operatng performance study analyzes merger performance by measurng the (accountng) profts of the mergng partes before and after the ntegraton. These studes estmate returns and the effect of mergers usng accountng data to measure changes n profts and n market shares. They are less homogeneous between them because dfferent measures of proftablty are adopted: cash flows, gross profts, profts net of nterest and taxes, proft ratos (returns on equty, on total assets, or on sales). Dfferent alternatves are also used to control for external shocks,.e., comparng the mergng frms wth ther base ndustry or wth matchng frms (frms smlar to the merged ones n ndustry and sze). However, these studes are no more perfect that event studes because accountng data are mperfect measures of economc performance and they can be affected by manageral decsons. In fact, these outcome studes show a smaller varance of results 2 Some of the event studes collect stock prces for the long-run after announcement, but do not converge n general results. Furthermore long-run movements n stock prces may not be very merger nformatve because nevtably, too much nose after the merger wll be preset e.g., external ndustry shocks; other operatons of the merged entty, etc. 2

4 due to dfferent methods (sample composton wth respect to tme horzons, control groups, merger motves, frms characterstcs, etc.) than the varance of the results n general. Ther fndngs do not dffer very much from those of event studes: n most cases post-merger profts of the mergng frms are weaker and sales perform worse wth respect to the mergng-control group. If we rely on event studes and on outcome studes n order to have an opnon about mergng gans, we end askng ourselves why mergers contnue to happen f acqurors do not show evdence of sgnfcant gans from such transacton? We consder that the ptfalls n the evaluaton of mergers mght precsely be n the employed methodology to estmate gans. Lmtng the analyss to fnancal and/or accountng technques rsks a lack of the economc ratonal behnd bdders decsons. Introducng a ratonal strategy n the decson to merge mght be useful to understand the motves to merge and to compute the true gans of mergng. Molnar (2004) models and tests the pre-empton hypothess specfyng that mergng s a ratonal strategy even when acqurors lose from mergng. Usng aucton theory Molnar models how even ratonal, shareholders value-maxmzng managers could pursue value-decreasng mergers, and then usng the event study methodology he proofs hs hypothess. Frdolsson and Stennek n a model of endogenous mergers develop a smlar dea to explan why valuedecreasng mergers occur. Our study concentrates on the acquror sde and not on the dstrbuton of the gans. It tres to show that f mergers contnue to happen and even to growth n tme and at the rhythm of economc waves, t s smply because bdders wn from mergng. It 3

5 provdes evdence of the postve gans of acqurors contrary to the general results of fnancal and accountng emprcal studes. In ths paper we hypothesze that the process of horzontal mergers parallels the process of a frst-prce aucton because a bd stuates a corporaton nto the game of aucton. Once a tender offer s open, any other potental bdder s free to propose a prce for the target; and the wnner s the bdder wth the hghest bd. Wth a database rch n bdders' and targets' characterstcs, bds proposals as well as the fnal bd pad for the deal and the number of compettors; a rcher approach may be used to estmate the acqurors ''true value'' for a target and the true gans of mergng. Furthermore, wth ths more economc approach of the merger process, the exploraton of the motves for mergng may be more accurate. As far as we know, analyzng whether an actual takeover aucton (a merger or an acquston) performs as an aucton has receved lttle attenton n the economc lterature. Mergers and acqustons have not been evaluated by an aucton process even when the course of mergers and acqustons clearly behaves as an aucton. Klemperer and Bulow (1996) support the dea that a takeover s an aucton and that targets managers get hgher expected profts by auctonng the frm than by negotatng wth fewer bdders. The value of negotatng s small relatve to the value of addtonal competton between bdders (runnng an aucton s more proftable than tryng to extract one buyer s surplus by negotaton). 4

6 On the other hand, we are nether aware of any emprcal study that estmates the gans of mergers and acqustons on the bass of a more economcally structural model as would be the aucton model when avalable data s cross sectonal 3. In ths study, we nvestgate the aucton process n mergers. By computng the economc gans of acqurers n mergers wthn a frst-prce sealed bd aucton, ths study shows that mergng s a proftable actvty for bdders. We apply the Guerre, Perrgne and Vuong (2000) methodology to estmate an ndependent prvate value aucton game. The procedure conssts n a two-step nonparametrc estmaton to recover the bdders prvate valuaton dstrbuton wthout makng any a pror assumpton of the latter. The advantages of ths method are that, t s easly mplemented and that t does not requre tocompute or to nverse the Bayesan Nash equlbrum strateges of the aucton model and that the estmated latent dstrbuton s not subject to msspecfcaton. Secton 2 presents the data. Secton 3 nterprets mergers as auctons. Secton 4 presents the nonparametrc estmaton methodology and ts dentfcaton concerns. Secton 5 presents the emprcal results. Secton 6 compares the estmated gans obtaned from wth aucton model wth the ones obtaned wth the event study. Secton 7 summarzes and concludes. 3 Structural economc models are used n case by case econometrc studes or "clncal research" studes. In those studes, data contans detaled nformaton about sales and prces of the product where the merger has taken place that allows them to measure merger effects on productvty, market shares and market power. 5

7 2. Data Our data are drawn from two versons of the SDC databases receved n dfferent perods and therefore not treated at the same tme. Those SDC databases consst on the record of the world mergers and acqustons for the perods. The database ncludes the date of the announcement of the merger as well as the year the merger s effectve. Both frms characterstcs at the announcement day are also avalable, that s, ther balance sheets elements, as well as ther ndustry and country. Data on competton ssues are also ncluded, we can observe f the merger has beng challenged and, more nterestngly for our purposes we have been able to detect f the bdder has faced compettors. For estmatons of the ndependent prvate value aucton gans we kept a subsample for whch we have also data on stock prces n order to make a comparson wth the event study gans. Then, due to the fact that we have complemented our SDC database wth the stock market nformaton from DataStream our fnal sample contans only US frms. Ths subset of mergers has regstered a non zero bd and an effectve year of the deal (to ensure that the merger has been completed). The fnal sample contans data of 150 horzontal mergers drawn over the perod Feb 2000 Feb % of the deals belong to the manufacturng ndustral sector, 37% to the Servces sector, 14% to the Fnance- Insurance-Real Estate ndustral sector, 4% to the Retal Trade sector and the remanng 5% s dstrbuted n the Mnng, Constructon and the Wholesale sector. 3. Mergers and Aucton Models Usng aucton theory for analysng M&A s a natural step because n ether arrangement, a merger or an acquston, the acquror has the self-determnaton of choosng the best 6

8 target for hm knowng that other potental acqurors wll be also nterested n submttng a bd whle the target maxmzes ts gans not by choosng tself a buyer but by lettng the dfferent buyers compete for t, that s, by runnng an aucton. The aucton aspect of corporate takeovers s explctly recognzed by regulators. Under Delaware law (the predomnant corporate law n the US), when a potental acqurer makes a serous bd for a target, the target s board of drectors s requred to act as would auctoneers charged wth gettng the best prce for the stockholders at a sake of the company 4. Furthermore, the Wllams Act requres takeover bds to reman open for at least 20 busness days on the grounds that the delay facltates auctons 5. Moreover, as well as bds n acqustons, bds n mergers are subject to delay and dsclosure provsons whch facltate the entry of compettors to a merger. Corporate regulators mght have ths preference for auctons because they know that auctons maxmze shareholders returns and that they promote effcency by shftng corporate assets nto the hands of those that value them most hghly. In ths context, matchng theory of auctons wth practce of takeover auctons, mergers and acqustons, seems reasonable. In a takeover process, the aucton starts when the frst bdder has arrved, the startng tme depends on the frst potental bdder, and once ths bdder has made publc hs nterest to obtan the target; other bdders came along and partcpate n the aucton. The man characterstcs of targets lke ther market 4 New Palgrave Dctonary of Law and Economcs, Peter Newman (ed.), London: MacMllan Press, 1, , In the 1960s, a large number of takeovers occurred unannounced. Ths created dffcultes for managers and stockholders who were forced to make crucal decsons wth very lttle preparaton. The Wllams Act was created n 1968 n order to protect nvestors from these occurrences. Ths federal act defnes the rules n regards to acqustons and tender offers. The bdders must nclude all detals of ther tender offer n ther flng to the SEC (Securtes and Exchange Commsson) and the target company. Ther fle must nclude the terms, cash source, and ther plans for the company after takeover, etc The Securtes and Exchange Commsson Rule 14e-1 7

9 value and the assets composton are publcly avalable n the market, and there s no rule that lmts the number of bds an acqurer can propose for a target. Accordngly, the aucton process s a natural mean to analyse M&A snce targets managers obtan the hghest prce by runnng an aucton nstead of by negotatng wth potental bdders. The mportant aspect s that the target has only ncomplete nformaton about the potental buyer s valuatons. If ths were not the case and targets knew the potental acquror and ts prvate valuaton for t, ts prcng problem wll be resolved and t wll ask the prce that matches the acqurer s prvate valuaton. Even when a takeover aucton can take the form of a merger or of an acquston; we beleve that there s a clear dfference between the two types of arrangements. Once passng by the process of the merger announcement and the hghest bd s proposed, the merger takes place f the two frms agree gong forward as a new sngle company rather than remanng separately owned and operated; and new company stock s ssued n ts place. On the other hand, there s no exchange of stock or consoldaton as a new company n acqustons. Several acqustons even happen n a hostle/unfrendly envronment, that s, the target company does not want to be purchased but stll, the aucton process s open. In ths sense, there s a clear dfference between a merger and an acquston. Ther common aspect s that they are both a mean to control a publcly held frm by an aucton process but; whle a merger contract s n search of a unque and specfc combnaton of target and bdder assets to become a new more valuable entty, an acquston s nterested n a porton of the target shares wthout necessarly re-combnng 8

10 ther respectve bunch of assets nether formng a new company but nstead becomng the parent/subsdary of the other. Comparably, the aucton lterature dstngushes two aucton envronments: the prvate and the common value auctons that we brefly descrbe below. For the aucton descrpton we wll denote random varables n upper case, ther realzatons n lower case and vectors n bold letters. The utlty each bdder =1,,N would receve from ownng the good s gven by U wth the same support F (). for all. U s referred to as s valuaton. Each bdder s U prvate nformaton conssts of a sgnal X whch s nformatve n the sense that E[U X =x, X - =x - ] strctly ncreases n x for all realzatons x - of s opponents sgnals. Sgnals play a purely nformatonal role and any monotonc transformaton θ ( X ) contans the same nformaton as X tself. The margnal dstrbuton of X s rrelevant, that s, wthout normalzaton on X the aucton theoretcal model s over-parameterzed. Wthout loss of generalty the followng normalzaton can be mposed X =E[U X ]. An aucton s a common value aucton f each bdder updates hs belefs about hs valuaton U when he learns X j n addton to hs own sgnal X. That s, bdders have common values f E[U X =x,..., X N =x N ] strctly ncreases n x j for all x j, j. In the common value paradgm, knowledge of opponents sgnals would alter the expectaton of the own valuaton of. The way nformaton about the value of the good s dspersed among bdders may vary, for nstance, there s the pure common value aucton n whch the value of the object s dentcal for all bdders, wth U =U 0. 9

11 In prvate values auctons, bdders do not have prvate nformaton about the valuatons of ther opponents, then E[U X =x,..., X N =x N ] = E[U X = x ] for all x,,x n, ths s equvalent to assumng that bdders know ther own valuatons X =U. We hypothesze then that whle mergers parallel a prvate value aucton mechansm, acqustons parallel a common value aucton mechansm. The reasonng s that when acqurng a porton of a target s share at a publcly known prce, each of the potental bdders pursues essentally the same objectve: obtanng a techncal effcency by acqurng control of the target and not necessarly obtanng merger-specfc gans or synerges. Thus we beleve than n the case of an acquston every bdder has essentally the same value for the shares of the target, and ths value s nfluenced by hs compettors sgnals. That s, as no merger-specfc gans are expected n the case of an acquston any bdder wll obtan the same addtonal value when acqurng shares of the target. On the other hand, for merger-specfc gans to be possble t must be the case that ths specfc combnaton of acquror-target assets gves the hghest gans to bdder, and gven that the deal s processed as an aucton the target also maxmzes gans. Synerges and merger-specfc gans are our partcular concern, for that reason ths study concentrates on horzontal mergers leavng asde acqustons deals and testng for the prvate value aucton paradgm n the a frst-prce aucton. We do not observe any data on the product market but we can stll evaluate for potental synerges drectly by computng the gans of mergng wth the aucton model. We assume that horzontal mergers are run n the context of a frst-prce sealed bd aucton and that the true gans of the acquror 10

12 should be computed not only wth the change n the stock prce but also addng the nformatonal rent obtaned when wnnng the target by the aucton process. The ndependent prvate value aucton approach s a reasonable approxmaton of the merger process because n mergers dfferent potental bdders have ther own and partcular perspectves of recombnaton of acqurng-acqured assets that are ndependent of the others bdders valuatons. The frst prce sealed bd rule s chosen for smplcty of estmatons; the ascendng aucton could also apply to the context of merger. Nevertheless, we consder ths assumpton s not very strong because n the frst prce sealed bd aucton the wnner pays the expected second hghest valuaton, E[v 2 ], whle n the ascendng aucton the wnnng bdder wll pay the second hghest valuaton, v 2. Gven the effcency of the markets t should not exsts a bg dfference, that s, we could expect [E[v 2 ] - v 2 ]=ε. The frst-prce sealed bd aucton wth ndependent prvate value s then brefly descrbed below. 3.1 Merges as Frst-Prce Sealed-Bd Prvate Value Auctons Data contanng bds and the number of actual bdders allow us to compute the equlbrum bddng strategy of the aucton game because t s a functon of the bdders prvate value and ther underlyng dstrbuton. Assume there s no reservaton prce n these takeover auctons, so that the number of potental bdders s equal to the number of actual bdders. Although, we could assume that the reserve prce of a target s ts market value, t s clear that for a bdder, that expects synergy gans from mergng and whch nternalzes the addtonal value the acquston of the target s gong to brng hm, the market value of the target; stated by outsders at the moment of the bd, s far from beng a reserve prce to hm. Assumng that bdders are symmetrc and that avalable data consst of bds from ndependent auctons of dentcal and ndvsble goods, the rules of the frst-prce sealed bd aucton are that bdders submt bds smultaneously and the 11

13 target s awarded to the hghest bdder at a prce equal to hs bd. Then characterstcs of the bdders are supposed to be drawn by nature from the same probablty dstrbuton whch s common knowledge to all bdders 6. Each bdder s supposed to know hs own valuaton as well as the number of partcpants to the aucton but does not have prvate nformaton about the valuaton of hs opponents. Then bdder s equlbrum bd s the expectaton of hs valuaton condtonal on hs own sgnal and the hghest competng bd, whch n the symmetrc case s expressed, = E U X = X = x max j. j as v ( x x, N ) = v( x, x, N ) Bdders prvate value are drawn ndependently from a common dstrbuton absolutely contnuous, F(.) wth densty f(.) and wth support [ ] + v, v R. Assumng bdders are rsk neutral, and establshng ther utlty for the target as U(v )= v, bdder s expected proft π condtonal on ts own sgnal and on t s the hghest competng bd can be expressed as: where ( ) b ( ) = ( v b ) Pr[ b b, j ] E π (3.1) v s the proft from the aucton, and ( b b, j ) j Pr s the probablty of wnnng the aucton. The bdder maxmzes hs expected gan wth respect to hs own bd and wll bd ndependently of the prvate value of ts compettors, that s, n the Bayesan Nash equlbrum, the bd s only n functon of ts own prvate value b = s( ) bdders are assumed to be symmetrc ( b ( x ) b ( x ), ) j j j j v. Because N Pr can be expressed as F 1 ( v ) 6 The bdders are assumed to be symmetrc n the sense that they draw ther characterstcs from the same dstrbuton. 12

14 and, the nverse functon of the equlbrum strategy as 1 (.) = s 1 (.) gan of the bdder can be expressed as: N 1 1 ( ) = ( v b ) F s ( b ) s ( ) E π (3.2). Then, the expected Maxmzng (3.2) wth respect to b and requrng that b =s(v ) solves the frst order condton for s(.). The symmetrc Bayesan Nash Equlbrum strategy gves then the followng dfferental equaton for N 2 : [ v b ]( N ) f ( v ) ( v ) s' ( v ) 1 = 1 (3.3) F where s (.) s the dervatve of s(.). The soluton of (3.3) s the equlbrum strategy s(.) whch s subject to the boundary condton ( ) s v = v then solvng for s(.) one obtans b = s ( v ) v 1 ( F( v )) N 1 v v ( F( z) ) N 1 dz (3.4) the equlbrum strategy, s(v(x), F(v(x)), N), s n functon of the number of bdders, the bdder s prvate value and the dstrbuton of prvate values and; t s strctly ncreasng n v on [ v, v]. Ths last equaton wll be the bass of the estmatons. 4. Estmaton and dentfcaton 4.1 Identfcaton The prmtve of nterest n the structural prvate value aucton analyss s the jont dstrbuton F(.) of bdder valuatons. Ths jont dstrbuton characterzes acqurors demand and nformaton. Equlbrum s defned by (3.4) whch lead to a closely related structural econometrc model snce bds are functons of prvate values, whch are random and dstrbuted as F(.) mplyng that the observed bds are also random wth a 13

15 dstrbuton noted G(.).G(.) conssts on a sngle mappng from the true dstrbuton of valuatons to a dstrbuton of bds mpled by the assumpton of Bayesan Nash Equlbrum. Gven that equlbrum s attaned when each player s actng optmally aganst the dstrbuton of behavor of compettors, both the dstrbuton of opponent behavor and the optmal (equlbrum) choce of each bdder are observable enablng dentfcaton of the latent jont dstrbuton of bdders valuatons. Then f the bdders prvate valuatons dstrbuton s unquely determned from observed bds, the dentfcaton problem s solved. However, gven that G(.), the dstrbuton of observed bds, depends on F(.), the dstrbutons of bdders prvate values drectly through v and ndrectly through the equlbrum strategy s(.) whch depends on F(.), the dentfcaton problem s not trval. To solve for ths, t must be that each prvate value v can be expressed as a functon of the correspondng bd b, the dstrbuton G(.) and ther densty g(.). As proposed by Guerre, Perrgne and Vuong (2000), let g(.) be the densty of observed bds n the dfferental equaton (3.3), ntroducton of g(.) and G(.) n equaton (3.3) smplfes ts expresson by elmnatng the frst dervatve s (.), the dstrbuton F(.) and ts densty f(.). Note that for every b [ b, b] = [ v, s( v) ], we have ~ ( b) = ( b b) = Pr( v~ s 1 ( b) ) = F( s 1 ( b) ) F( v) G Pr = (4.1) Then, the dstrbuton of observed bds s absolutely contnuous wth support [ s( v) ] f densty ( ) () v 1 g b = where v s ( b) s' () v ( ) ( ) g b =, the rato wll then be G g dfferental equaton (3.3) can be expressed as: f () v () v F() v v, and =. Fnally, the s' 14

16 v ( ) ( b ) 1 G b = ϕ ( b, G, N ) b + (4.2) N 1 g The bdder s latent prvate value n equaton (4.2) can be expressed as a functon of hs equlbrum bd b, the jont dstrbuton of the competng equlbrum bd he faces G(.), ts densty g(.) and the number of bdders N. Equaton (4.2) s the nverse of bdder s equlbrum bd functon, the mappng needed to nfer valuatons from bds. Snce the jont dstrbuton of bds s observable, dentfcaton of each prvate value v and therefore of the jont dstrbutons F(.) follows drectly from (4.2). That s, f b s the equlbrum bd, then the bdder s prvate value, v, correspondng to b must satsfy (4.2). Then the gans of mergng to the acquror are the change and ts market value due to the merger announcement plus the nformatonal rent obtaned from the aucton process G A t MV MV IR A A t t 1 t = + (4.3) A MVt 1 where A MV t s the acqurer s market value at the day of merger announcement, the acqurer s market value pror to the merger announcement and MV 1 s A t IR t are the gans from the merger process, that s, ( vt b t ) the acqurer s true valuaton for the target mnus the actual bd pad for t. 4.2 Estmaton method If one knows G(.) and g(.), expresson (3.4) can be estmated so as to recover every bdder s prvate valuaton v. There exst many dfferent ways to estmate such 15

17 dstrbuton. Parametrc methods would requre to specfy an apror parametrc famly dstrbuton for G(.), whch may brng some correspondence problems between G(.) and F(.) and moreover, ths choce could be subject to msspecfcaton. Wthout assumng any specfc functonal form nether on the bdders dstrbuton of ther prvate values nor on the observed bds dstrbuton, but nstead lettng the data reveal the shape of the dstrbuton, we frst estmate the emprcal cumulatve bd dstrbuton G(.) and then ts correspondng densty g(.) usng nonparametrc kernel densty estmators. Second, ntroducng the estmated G(.) and g(.) n (4.2), a sample of pseudo prvate values s obtaned. Thrd, usng the estmated prvate values of the second step, the pseudo prvate values vˆ we are able to compute the nformatonal rents of the bdder usng estmated vˆ and observed bds b. Fnally, equaton (4.3) the true gans of the merger can be computed by addng the nformatonal rents accrue to the bdder from the merger process to the change n ts market value due to the merger announcement. Hereafter, consder L auctons, each one of them wth a number of bdders N l, l=1,,l. To smplfy the presentaton assume ths number does not vary across mergers, and that no heterogenety across aucton objects s present. We observed all the bds b nl n=1,,n, l=1,,l. Let v nl n=1,,n, l=1,,l denote the prvate values. The frst step conssts n obtanng the pseudo prvate values v nl ( nl ) ( b ) 1 G b = ϕ ( bnl ) bnl + (4.4) N 1 g nl Let ~ G 1 NL L ( b) = 1( bnl < b) l = 1 N n= 1 (4.5) be the emprcal dstrbuton of observed bds; and 16

18 L N l ( ) b bnl g~ 1 b = K (4.6) NLh g l = 1 n= 1 hg be the kernel densty estmator of observed bds, where K s the trweght kernel, hg s the bandwdth and t s defned as 1/(2R+ 3) h g = cg ( NL) wth cg a *1.06σ b =, a s the factor of the kernel, σ b s the emprcal standard devaton of observed bds and R s the number of dervatves of the densty. Now, (4.4) can estmated by vˆ nl ~ 1 G b = bnl + N 1 g~ (4.7) ( nl ) ( b ) nl Equaton (4.7) defnes the pseudo prvate values. As these estmates are based at the boundares of the support the followng trmmng rule must also be appled: ~ 1 G( bnl ) bbl + f bmn + hg bnl bmax hg vˆ nl = N 1 g~ ( bnl ) (4.8) otherwse where b mn and b max are the mnmum and the maxmum bds, respectvely. Once (4.8) obtaned the estmated true gans of the merger aucton game are obtaned by Gˆ A t A A MVt MVt 1 + IRt = (4.10) A MVt 1 ˆ 5. Emprcal Results Recall that our fnal sample conssts on 150 horzontal mergers n the US. Comparng the wnnng bd b1 to the market value of the target T MVt 1 one day pror to the merger announcement and to the estmated prvate valuaton of the bdder v 1, we note that the wnner has proposed much more than the actual market value of the target at the tme of announcement. See table

19 Table 5.1: Summary Statstcs of Wnnng Bds, Target s Market Values and Estmated prvate valuatons (In mllons of dollars) N Mean S.D. Mn Max T MVt b ˆv As prvate values are recovered, we can compute the nformatonal rents to the acquror n the aucton model whch are denoted by IR and IRp where, IR=( vˆ -b) and IRp=( vˆ -b)/ vˆ respectvely. Equaton (4.10) of the gans of the bdders can also be computed. See table 5.2 Table 5.2: Summary Statstcs of Informatonal Rents of the Bdder n Mergers N Mean S.D. Mn Max IR IRp G A t In average bdders obtan a 5.7% of ther valuaton of the target whch mght appear a small number but takng nto account that targets value mllons of dollars those gans are not neglgble. The return to the bdder from mergng as ndcated by equaton (4.10) s postve, n average the value of the acqurors ncreases n 13% due to the aucton process mpled by the merger announcement. Comparson wth the Event Study Event studes rely on the presumpton that stock prces reflect the present value of the expected profts created by the mergng frms, under the assumpton that the stock market s effcent. An event study s based on the analyss of the stock return of bdders and targets relatve to a portfolo of stock representng the market where they operate. The 18

20 event s the merger announcement. Dfferences n returns of the target and the acqurer relatve to market returns are usually calculated over a perod of tme (before and after announcement). The objectve s to determne whether the announcement of the merger causes the stock return of the bdder and of the target to perform dfferently than the general market return. That s, the abnormal return s the return that s observed n excess of what t would have been f the stock had behaved relatve to market n the same way as n a benchmark perod. The behavour of the stock relatve to the market s estmated before announcement from a market model: R = a + β R + e, where R t are the actual t mt t returns to frm stock at tme t, R are the actual returns to a market portfolo for frm mt stocks at tme t, for example, the value-weghted ndex of the sector stock. The market model s estmated n a perod of tme called estmaton wndow Usng predcted values from the market model, one obtans the abnormal returns to frm a tme t by: AR t = R aˆ + ˆ β R ) for the event wndow, that s a selected perod of tme after the t ( mt estmaton wndow. The market response can be tested by computng the cumulatve abnormal returns of the mergng frms over the span of the event wndow T, that s, CAR t = T AR. Fnally to test for ther sgnfcance, a t-test s constructed as: CART t( CAR) = σˆ T CAR T Expected gans of acqurng estmated wth the aucton model are much hgher that the ones regstered n our prevous study based on the event-study methodology, were the acquror has abnormal returns for acqurors that are negatve n average and not 19

21 sgnfcant; the t-test for the sample wth stock nformaton s of See table 4 and fgure A.6. Table 5.3: Abnormal Returns, Informatonal Rents (Absolute and Relatve) and Aucton Gans N Mean S.D. Mn Max AR IR IRp G A t As t can be seen n Table 5.3 the computed true gans of the merger wth the aucton process dffer to the ones computed by the event study methodology. Whle the movements n stock prces show a decrease n the acqurer s standng alone value, the aucton gans are postve n average. 7. Concluson Ths paper proposes to nterpret a merger as an aucton n order to provde a powerful analytcal tool for evaluatng gans from mergng. It bulds on some dssatsfacton wth event studes and wth operatng performance studes that estmate the gans of a group of mergers wth fnancal and accountng data respectvely, but wthout any structural economc approach behnd those market models. We hypothesze then that horzontal mergers searchng for synerges parallel a prvate value aucton mechansm. The reasonng s that a bd stuates a corporaton nto the game of aucton. Once a tender offer s open, any other potental bdder s free to propose a prce for the target; and the wnner s the bdder wth the hghest bd. Data contanng bds and the number of actual bdders allow us to compute the equlbrum bddng strategy of the aucton game and therefore the nformatonal rents accrue to the bdder from the aucton 20

22 process. Then, by computng the economc gans of acqurers n mergers wthn a frstprce sealed bd aucton, ths study shows that mergng s a proftable actvty for bdders. The return to the bdder from mergng s postve, n average the value of the acqurors ncreases n 13% due to the aucton process mpled by the merger announcement. A parallel event study has been performed for purposes of comparson; the abnormal returns to acqurors n the event study are negatve and not sgnfcant. References Andrade, G., Mtchell, M., and Stafford, E., New Evdence and Perspectves on Mergers, Journal of Economc Perspectves 15, pp , Athey, S. and Ahle, P., A., Nonparametrc Approaches to Auctons, Workng Paper, June Bulow, J. and Klemperer, P., Auctons versus Negotatons, Amercan Economc Revew, March Cramton, P. and Schwartz, A., Usng Aucton Theory to Inform Takeover Regulaton, JLEO V7 N1, pp Frdolsson, S. and Stennek, J., Why Mergers Reduce Profts and Rase Share Prces: A Theory of Preemptve Mergers, Workng Paper, Guerre, E., Perrgne, I. and Vuong, Q., Optmal nonparametrc Estmaton of Frst-Prce Auctons, Econometrca 68, Guther T., "What do we know about success and falure of mergers", European Network of Industral Polcy, Workng Paper Healy, P.; K. Palepu; and R. Ruback, Does Corporate Performance Improve After Mergers? Journal of Fnancal Economcs, 31, Aprl 1992, Molnar, J., Pre-emptve Horzontal Mergers: Theory and Evdence, Workng Paper,

23 APPENDIX Fgure A.1 7 Estmated Prvate Values and Observed Bds n Mergers sold lne: Estmated prvate values Bds Fgure A Denstes of bd and of estmated prvate values observed bds;+: estmated prvate values :

24 Fgure A Informatonal Rents of Bdders n Mergers IR Fgure A Informatonal Rents n Percentage of Bdders n Mergers IRp 23

25 Fgure A.5 Fgure A.6 Abonormal Returns of Bdders n MErgers Densty moyen 24

26 Kernel Estmates of Bds n Mergers & Acqustons Proportonal to Target's Market Values Kernel estmates of relatve bds Bd Mergers Acqustons Kernel denstes of the true gans from mergng evaluated wth the merger process cmv and the abnormal returns ar of the event study x kdensty cmv kdensty ar 25