Value of Fairness Opinions in US Mergers and Acquisitions,

Transcription

1 Value of Farness Opnons n US Mergers and Acqustons, Helen M. Bowers Unversty of Delaware, Lerner College of Busness and Economcs, Department of Fnance, Newark, DE 19716, USA and Wllam R. Latham III Unversty of Delaware, Lerner College of Busness and Economcs, Department of Economcs, Newark, DE 19716, USA Draft: November 23, 2004 ABSTRACT In the market for corporate control, a potental market falure of asymmetrc or nadequate nformaton arses f any of the market partcpants (the acqurng or target frms management, boards of drectors or shareholders) have nsuffcent knowledge about the real market value of a target frm. Ths falure may be mtgated by the market s partcpants choosng to purchase addtonal nformaton about the value of the target frm. An opnon by a thrd party regardng ths value s known as a farness opnon. Although t s often the case that at least one party to an acquston obtans a farness opnon, the ssue of whether they provde any nformatonal value s stll debated. US court rulngs have ncreased the potental costs to frms and ther boards of drectors of makng merger and acquston decsons wthout suffcent nformaton, thus potentally rasng the value of farness opnons. The paper examnes factors nfluencng the decsons of frms engaged n merger and acquston actvty durng the perod to obtan or not to obtan farness opnons. For each transacton nformaton s avalable on the prmary ndustry n whch the acqurng and target frms operate and on the numbers and types of addtonal nformaton, ncludng farness opnons, each of the partes to the transacton sought durng the progress of the transacton. Our results show that, for the acqurng frm n an acquston, the lkelhood of purchasng farness opnons s nfluenced sgnfcantly by (1) the market values of the acqurer and the target frm, (2) the volatlty of excess returns of both frms, (3) whether or not the transacton s a cash deal, (3) the degree of asymmetrc nformaton as measured by the smlarty of the acqurer and target frms, (4) the amount of monopoly power the target frm has, (5) whether the acquston s hostle, and (6) whether other fnancal advsory servces have been purchased by ether frm. Fnally, strong evdence s found ndcatng that (7) the behavor of acqurng frms, whether ncorporated n Delaware or not, has been sgnfcantly altered snce the 1985 Van Gorkom v. Smth decson by a Delaware court regardng farness opnons. Our results for target frms are not as strong as those for acqurers, nor are the results for fnancal advsory servces more broadly defned. The authors would lke to thank the partcpants n the Laboratore des Stratéges Idustrelles (LASI) semnar at the Unversté de Pars1 Panthéon-Sorbonne, Thomas Bates and Darren Latham for ther helpful comments. Contact nformaton: Helen Bowers: bowersh@lerner.udel.edu, Wllam Latham: latham@udel.edu.

2 2 1. Introducton A man s judgment cannot be better than the nformaton on whch he has based t. Arthur Hays Sulzberger, former publsher, N.Y. Tmes, n Address to the New York State Publshers Assocaton, August 30, 1948 A farness opnon, typcally presented n the form of a letter to the board, contans the ssuer s opnon regardng the farness to the shareholders of the corporaton of the fnancal terms of a proposed transacton. Although there are no legally specfed credentals qualfyng ssuers of farness opnons, the precedent settng court rulngs 1 have stressed that the provders of farness opnons should be qualfed and ndependent. Most often, farness opnons are wrtten by nvestment banks, but can also be provded by consultng frms, CPA frms, commercal banks, apprasers, or consultants specalzng n valuaton. 2 Farness opnons help meet a key crteron for protecton afforded by the busness judgment rule whch holds that corporate boards are protected from shareholder lablty from the consequences of adverse busness decsons f the decson was ratonally based, made n good fath that the acton was n the best nterest of the company and arrved at n an nformed manner. No other specfc document s as unversally recognzed as evdence of an nformed 1 The 1985 Delaware Supreme Court decson n Smth v. Van Gorkom (488 A.2d 858 (Del. 1985)) s usually cted as the precedent settng court rulng (for example, see Bebchuk, 1989) that created the oblgaton that when evaluatng a takeover proposal, the corporate boards of target frms must nform themselves of all reasonably avalable and relevant nformaton to the decson. 2 A June 11, 2004 Wall Street Journal artcle NASD Scrutnzes Conflcts n Bankers' 'Farness Opnons' summarzed the stuaton succnctly: Farness opnons became common after a 1985 rulng by the Supreme Court of Delaware, the leadng jursdcton for M&A law because of the large number of companes that ncorporate n Delaware. In that case, whch nvolved allegatons that Trans Unon Corp. sold tself too cheaply to the Prtzker famly, the court held that Trans Unon's board had volated a duty of care and sold the company for too lttle money. The court stated that gettng a farness opnon would have helped n fulfllng that duty. The decson prompted corporate boards to routnely seek farness opnons. But ther purpose was as much bullet-proofng a board's decson as explorng valuatons.

3 3 board as a farness opnon. Farness opnons therefore, are mbued wth power to reduce, even elmnate, lablty to shareholders for members of corporate boards arsng from dsagreements n valuaton. In ths paper, we examne whether the emprcal evdence supports ether of two theores offered to explan why acqurng or target frms obtan farness opnons. The frst s a theory of nformaton asymmetry where farness opnons are more lkely to be obtaned where there s less nformaton avalable n the market regardng the value of the target frm. The second theory s a one of ncreased lablty, where farness opnons are more lkely to be sought when there s a perceved hgher probablty of a successful shareholder lawsut as a result of the transacton. We fnd that, for the acqurng frm n an acquston, the lkelhood of purchasng farness opnons s nfluenced sgnfcantly by (1) the market values of the acqurer and the target frm, (2) the volatlty of excess returns of both frms, (3) whether or not the transacton s a cash deal, (3) the degree of asymmetrc nformaton as measured by the smlarty of the acqurer and target frms, (4) the amount of monopoly power the target frm has, (5) whether the acquston s hostle, and (6) whether other fnancal advsory servces have been purchased by ether frm. Fnally, strong evdence s found ndcatng that (7) the behavor of acqurng frms, whether ncorporated n Delaware or not, has been sgnfcantly altered snce the 1985 Van Gorkom v. Smth decson by a Delaware court regardng farness opnons. Our results for target frms are not as strong as those for acqurers, nor are the results for fnancal advsory servces more broadly defned. 2. Hypotheses 2.1 Informaton asymmetry and market value

4 4 Informaton about a partcular acquston can be characterzed nto two types: market nformaton and costly ncremental nformaton. We defne market nformaton as the nformaton that s avalable to all partcpants n a sem-strong-form effcent market. Informaton provded by the market ncludes ndustry analyss, securty analyss, and SEC flngs. Incremental nformaton s nformaton that s obtaned at a cost by some party to the acquston. Ths ncremental nformaton may eventually be dsclosed and become market nformaton, but ths s not necessarly so. The focus of ths study s on a partcular type of ncremental nformaton, the farness opnon. A farness opnon contans the ssuer s assessment of the farness of the offer to the frm s shareholders. Although the opnon may state that an offer s far, ths does not necessarly mean that the ssuer of the opnon regards ths offer as the best offer that the target could reasonably expect. If farness opnons, whle only statng that an offer s far, mtgate nformaton asymmetry we expect farness opnons to be more lkely to be sought when there s more value at stake. We measure the sze of the transacton n two ways. Frst we use the actual dollar value of the transacton n terms of the market value of the target measured as the average of the market values over days -25 to -5 relatve to the announcement date. We hypothesze that the larger the sze of the frm, the more lkely t s that a farness opnon wll be sought. Therefore, we hypothesze that a frm s more lkely to obtan a farness opnon when the other frm n the transacton s relatvely large. We also hypothesze that, the larger the target s relatve to the acqurer, the more lkely t s that the acqurer wll seek advce n the form of a farness opnon. We expect ths relaton for the acqurng frm because the acqurer should be more wllng to obtan costly ncremental nformaton when the target frm s relatvely larger because of the greater potental mpact on the acqurer s own equty value.

5 5 2.2 Uncertanty n the nformaton envronment Derkens (1991) fnds emprcal evdence that nformaton asymmetry between the managers of the frms and the market s sgnfcant for equty. The magntude of ths nformaton asymmetry s expected to ncrease wth the volatlty of the frm s stock because ths volatlty reflects the underlyng uncertanty regardng the frm s future performance. The varance of excess returns (Ret t - Mkt t ) has been used n several studes, such as Bzak, Brckley, and Coles (1993), Boone, Feld, Karpoff, and Raheja (2004), Gaver and Gaver (1993), Lehn, Patro and Zhao (2004), Smth and Watts (1992), and Yermack (1995) as a measure of the uncertanty n the nformaton envronment. We argue that when makng the decson whether to obtan a farness opnon, the relatve uncertanty regardng the nformaton envronment s a crtcal determnant. The frm would seek to obtan the costly ncremental nformaton n a farness opnon f the other frm n the transactons s n a more uncertan nformaton envronment than that of ther own frm. 2.3 Industry closeness and structure and the value of nformaton Informaton asymmetry can also arse n the market based on dfferental experences `of frms. Those wth smlar experences wll share an nformaton base not shared by those wthout such experence. One form of experence s smply operatng a busness n partcular markets wth all of the nsttutonal detals that dfferentate real-world markets from each other. We hypothesze that ndustry closeness wll affect the value of the knd of nformaton provded by farness opnons. Frms that are close to each other n the sense that they are n the same or smlar ndustres (whch s not the same s beng close n a geographc dstance sense) they may be more lkely to have hgher levels of pror knowledge about each other than frms n dssmlar

6 6 lnes of busness. It would follow that these frms wth hgher pror levels of knowledge would fnd the addtonal nformaton provded by a farness opnon less valuable than dssmlar frms. Thus we measure how close frms are to each other and then examne whether or not the closeness of frms to each other reduces the lkelhood of obtanng a farness opnon Measurng ndustry closeness Every target and acqurng frm s assgned to an ndustry based on ts prmary lne of busness. For frms that produce a wde range of products and servces these assgnments may not be precsely descrptve of the frm s actvtes. However, the noton of an ndustry n whch a frm partcpates s an ntutvely appealng concept that has also generally been found to be useful n emprcal nvestgatons. In fact, most frms produce a range of products from the same or closely related ndustres. The North Amercan Industry Classfcaton System (NAICS) and ts predecessor, the US Standard Industral Classfcaton system, both have numercal codes that group frms nto ndustres. Industres from related sectors have smlar codes: all manufacturng ndustres have NAICS codes begnnng wth the 2-dgts 31, 32 or 33, wholesale trade begns wth 42, etc. Increasngly detaled ndustres are ndcated by codes wth more and more dgts. For example, NAICS code 325 s chemcal manufacturng, 3255 s pant, coatng and adhesve manufacturng, and , s adhesve manufacturng. The codng system extends to as many as nne dgts, but sx dgts s more than enough to dentfy an ndustry n most cases. The COMPUSTAT data we use have sx-dgt NAICS (and four-dgt SIC codes) for both target and acqurng frms. The structure of the codng systems places frms that produce smlar products n ndustres wth codes that are numercally smlar. For example, NAICS ndustry code s assgned to pant manufacturng whch s obvously smlar to adhesve manufacturng whch s assgned the NAICS code We construct a measure of ndustry smlarty or closeness based on these codes. Because frms n ndustres that produce smlar

7 7 products have ndustry codes that are numercally close to each, the absolute value of the dfference between the numercal values of the target and acqurng frms NAICS (or SIC) codes mght be used as a measure of how smlar the ndustres are n whch the target and acqurng frms operate. The closeness measure for frms n pant manufacturng (325510) and adhesve manufacturng (325520) would be 10 whle the closeness measure for a frm n pant manufacturng and a frm n any knd of metal fabrcaton (332***) would be at least 6,490 ( ). Wthn the manufacturng sector, there does seem to be a rough sort of correspondence, at least ordnally, between the absolute dfferences between codes and the degree of dssmlarty between ndustres. However, the numercal code numbers were not desgned to ndcate closeness and do so only mperfectly, especally from one major ndustry group to another. There are obvously multple dmensons of closeness that cannot be reflected by such a one-dmensonal array of numbers. For example, all fnancal nsttutons have NAICS codes that begn wth 52 so that ther 6-dgt codes are 52****. Thus the closeness measure for a fnancal nsttuton and pant manufacturng would be at least 194,490 ( ). Professonal servces, whch may be just as closely related to a manufacturng ndustry as a fnancal nsttuton, but n a dfferent dmenson, have codes that begn wth 54 so ther mnmum measures of closeness to pant manufacturng would be 214,490. It does not make sense to say that any fnancal nsttuton s closer to pant manufacturng than any professonal servce. To avod ths problem, whch arses from the arbtrarness n the assgnment of the frst two dgts n the codes for major ndustry groups, we assgn a closeness measure of 30,000 whenever the acqurer and the target are from dfferent major ndustry groups. The value of 30,000 s chosen because the largest maxmum wthn-major-ndustry-group dfference, whch s n manufacturng, s 29,999. The fnal adjustment to the closeness measure we make s to nvert

8 8 the scale by subtractng each value from 30,000 so that the ntuton of low numbers ndcatng geater closeness and hgh numbers ts absence s establshed. The result s that we have a measure of the degree of closeness for pars of frms wthn the same general classfcaton and for all other pars we assgn a large number. An alternatve approach to ndustry closeness s to use a 0-1 dummy varable to dentfy as close transactons for whch both target and acqurng frms are from the same ndustry group. The precson of the defnton of closeness n ths case depends upon how dsaggregated the ndustry groups are. An a pror selecton must be made of the ndustry codes to place nto specfc groups that, for some characterstcs, are more homogeneous wthn the groups than across groups. We choose to mplement ths alternatve usng the 48 ndustry groups used by Fama and French ( Industry Costs of Equty, JFE (43) 1997, pp ). 3 Fama and French s ndustry groups are aggregatons of SIC codes of related ndustres from dfferent parts of the SIC system. For example, ther Busness Servces group ncludes the codes for ndustres from Commercal Prntng (SICs , NAICS and ), Sgns and Advertsng Dsplays (SIC 39930, NAICS 33995), parts of Busness Servces other than computers (SICs , NAICS 5418 and 5614), and a varety of other sectors. Table 1 shows the dstrbuton of ndustres n the Fama and French ndustry groups for all of the acquston announcements for whch there s fnancal advsor data avalable n SDC and for those announcements that were ncluded n our regressons. 3 Other schemes for aggregatng ndustry codes nto ndustry groups exst. For example, Compustat defnes a dfferent set of ndustry groups. In another context t s has been found that usng Compustat ndustry defntons rather than Fama-French ndustres does not affect analytcal results (e.g., Boone,et al. [2004]). Because we are not nterested n the ndvdual ndustry results explctly, but only need to determne whether two frms are n the same or dfferent ndustres, the partcular assgnment scheme should not matter as long as the levels of aggregaton are smlar. The partcular assgnment scheme mght be more mportant when ndustry s used as a proxy for other unmeasured factors such as frm complexty and n ths context we agan rely on fndng such as those of Boone,et al. [2004].

9 Industry structure and the value of nformaton We hypothesze that ndustry structure wll affect the value of the knd of nformaton provded by farness opnons. It s relatvely easy to fnd nformaton on frms n hghly compettve ndustres so that they can be compared wth other frms to determne the reasonableness of an offer. Monopoly frms, however, by ther nature, are unque and consequently more dffcult to value by referrng to other frms because less nformaton s avalable and there are not comparable frms. Thus we hypothesze that the more compettve the ndustry s, the less value a farness opnon would be expected to have. We measure compettveness usng the percentage of ndustry sales by the four frms wth the hghest sales n the ndustry (the 4-frm concentraton rato). Ths measure s admttedly somewhat arbtrary, but t s wdely used and does, at least, provde an ntal measure of compettveness to test. In addton t s avalable from the U.S. Bureau of the Census. 4 Whle ndustry compettveness vares over tme t does not vary greatly n many ndustres wthn relatvely short tme spans. Thus we use a compettveness measure for 1997 for all of the transactons n our sample. If we fnd that the compettveness measure has sgnfcantly more dscernable effects around 1997, we can refne the measure to more closely match the perod n whch the transacton occurs. 2.4 Medum of exchange The ntal terms of the acquston offer are ncluded n the announcement of the offer, ncludng whether the offer wll be cash, stock or some other medum of exchange. Because a 4 Concentraton measures for 1997 by detaled NAICS codes are avalable from the U.S. Census Bureau for all ndustres except Mnng (NAICS 21), Constructon (NAICS 23) and Management of Companes and Enterprses (NAICS 55). See Other measures calculated by the Census Bureau nclude the 8-frm, 20-frm and 50-frm concentraton ratos for all ndustres ncluded and the Hrshman-Herfndahl Index only for manufacturng ndustres. The last s vewed by most economsts as the best of these measures because t utlzes nformaton from the full dstrbuton of frms wthn the ndustry and weghts the larger shares more heavly. However, t s not avalable for ndustres other than manufacturng, so we choose to use a measure whch s avalable for all of the ndustres nstead.

10 10 farness opnon concerns the farness of the offer to the shareholders, the less ambguous the value of the offer, the less ncremental nformaton s provded n an opnon. Therefore, we hypothesze that a frm wll be less lkely to seek a farness opnon when the medum of exchange s all cash. 3. Increased lablty hypothess We hypothesze that f farness opnons can be used to satsfy one of the requrements for protecton under the busness judgment rule, then opnons are more lkely to be sought when there s the percepton that the potental for a successful shareholder lawsut s relatvely hgh. 3.1 Court rulngs and the value of nformaton Frst, we nclude a dummy varable for the Van Gorkom decson. Prevous studes have found an ntal ncrease n the frequency n the use of farness opnons after the Delaware Supreme Court decson n 1985 (Bowers 2002) and we expect o confrm ths n the present tudy even after contollng for many factors not consdered n the pror work Atttude of the deal and the value of nformaton The ntal reacton of the target frm s management to the announcement of the offer may be an ndcaton of potental ltgaton resultng from the acquston. Therefore, we hypothesze that a frm s more lkely to obtan a farness opnon f the acquston s hostle. A fndng of a negatve sgn mght ndcate that control, not valuaton, s the ssue n hostle takeovers and that addtonal valuaton nformaton s not sought because the value s not at ssue. 3.3 Number of advsory servces and the value of nformaton We regard other advsory servces as probably functonng as substtutes for farness opnons and thus hypothesze that, the larger the number of other advsory servces obtaned, the lower wll be the probablty of obtanng a farness opnon. A fndng of a postve sgn would

11 11 ndcate that other advsory servces are complements to farness opnons and not substtutes for them. 3.4 Tme perod of the transacton and the value of nformaton We use the year that the transacton s announced. We use the tme perod as a proxy for busness cycle and market condtons not related to target and acqurers excess returns or ther market values and other unobserved factors that may nfluence the decson to seek a farness opnon. We also use a contnuous tme varable to capture any tme trend effect not assocated wth the calendar years. Ths varable has a value of 1 on the frst day n the sample perod (January 1, 1980) and ncreases lnearly each day to the end of the perod (December 31, 2002). 4. Bnomal logt regresson models The followng basc model was estmated separately for acqurers and targets for =1 to T transactons 5 FOP = β + β ADVA + β ADVT + β CLSE + β CONA + β CONT + + β VALT + β RVAL + β MP + β VG + DELA + DELT + DELB + ATT + β FAIR g= 1 γ IND g 11 g + β VALA t= 1980 α YEAR t t + ε Where FOP = 1 f a farness opnon was sought by the frm, 0 otherwse; 5 Note that n the emprcal analyss the number of observatons vares from one estmated equaton to another because of mssng data for some varables for some observatons. We feel that the gan from the addtonal numbers of observaton made possble n many of the estmatons by ths approach (n terms of precson of estmates) more than offsets the loss of certanty that the partcular sample s not drvng the results when exactly the same observatons are not used n all estmatons. However, to ensure that selectvty has not based the results, Hausman s test have been run and the results ndcate that selectvty does not seem to be a problem.

12 12 or = the number of farness opnons reported by the frm ADVA = Number of non-farness opnon advsory servces used by the acqurng frm ADVT = Number of non-farness opnon advsory servces used by the target frm CLSE =Closeness = NAICSA NAICST f 1 st 2 dgts match, otherwse = 30,000 or = 1 f both frms are n the same Fama-French ndustry group, 0 otherwse CONA = 4-frm concentraton rato of acqurng frm s ndustry CONT = 4-frm concentraton rato of target frm s ndustry IND g = 1 f the target frm s ndustry s n Fama-French ndustry group g, 0 otherwse VALA = Total market value of the acqurng frm VALT = Total market value of the target frm RVAL = Market value of the target frm relatve to the market value of the acqurng frm MP = Merger premum VG = 1 for transactons after 1985, the year of the the Van Gorkham decson DELA = 1 for transactons n whch the acqurng frm s ncorporated n Delaware DELT = 1 for transactons n whch the target frm s ncorporated n Delaware DELB = 1 for transactons n whch both frms are ncorporated n Delaware FAIR = 1 for transactons after 1995, the year of the SEC s Far Dsclosure regulaton YEAR t = 1 f the year n whch transacton was completed s year t, 0 otherwse ATT = 1 f the atttude of the deal s Frendly; 0 f t s Hostle The theoretcal consderatons that lead to the model specfed have strong restrctons on nether the forms of the varables nor the forms of the equatons to be estmated. Thus we have consderable freedom to choose forms that lead to functons that ft the data best. Most of the varables are dchotomous dummes for whch alternatve functonal forms are not consdered. Several of the other varables have lmted ranges of values (e.g., FO whch only has nteger values between 0 and 7, or CONA or CONT whch have values n the 0-1 nterval) makng them less attractve canddates for transformaton also. Ths leaves the closeness measure, CLSE, based on NAICS codes, the market values of the frms, VALT and VALA and the merger premum, MP. If the Closeness varable has affects that are more attenuated the farther apart the two frms are, ths affect can be modeled by usng a quadratc form n closeness. Confrmaton of the hypothess of an attenuated effect of closeness wth greater dstance would be a negatve

13 13 lnear term and a postve squared term (the varable s values are between 30,000, for frms n the same ndustry to 0, for frms far apart. 5. Data Collecton and Summary Statstcs Sample transactons are drawn from a pool of 7,818 merger announcements between 1980 and 2002 obtaned from the Securtes Data Corporaton (SDC) domestc mergers and acqustons database. These observatons nclude those announcements for whch the acqurng and target frms were publc and the value of the deal was dsclosed. In addton, for an observaton to be ncluded n the sample, ts SDC classfcaton for the form of the deal had to be as a merger, acquston, acquston of stock assets or majorty nterest. Announcements were excluded from the sample f, accordng to SDC, the status of the deal was unknown, the transacton was classfed as rumored or where ether frm had only announced plans to seek out a buyer or seller for all or part of ts assets. Also, f SDC classfed the deal type as a spn-off, recaptalzaton, self-tender, mnorty stake purchase, prvatzaton, or acquston of remanng nterest, or f the proposed transacton was a cross-border deal, t was excluded from the sample. Of the 7,818 merger announcements, farness opnon data was avalable for 4,228 of the acqurers and 4,229 of the targets. Farness opnons are routnely by not unversally sought. Fgure 1 shows the number of farness opnons obtaned by acqurng and target frms over the perod Fgure 2 shows the frms that obtaned farness opnons as a percentage of all 4,228 of the acqurers and 4,229 of the targets for whch farness opnon data were avalable from the SDC dataset. The sample perod begns n 1980 to capture the any affect of the Van Gorkom court decson n The number of acqurers and targets obtanng farness opnons has trended upward snce

14 However, the frequency of frms obtanng farness opnons, although varyng from year to year s relatvely much more stable, exhbtng a slght upward trend over the long term. There are two possble reasons for the drop-off n frequency of frms seekng farness opnons after Frst, the decrease may be due to the purported escalaton n the prces of obtanng opnons and the declne n wllng provders of opnons because of the ncreased potental lablty after the recent corporate governance scandals. However, the Securtes Exchange Commsson (SEC) adopton of Regulaton Far Dsclosure n October 2000, may have reduced the value of farness opnons. Eleswarapu, Thompson, Venkataraman (2004) have found that nformaton asymmetry declned followng the SEC s adopton of Regulaton Far Dsclosure n October 2000, concludng that the SEC appears to have dmnshed the advantage of nformed nvestors. If the SEC s adopton of Far Dsclosure has reduced the potental value of farness opnons and then we would expect there wll be fewer farness opnons sought caters parbus followng October The change n percentage of frms seekng farness opnons on a yearly bass roughly corresponds wth change n the number of acqustons announcements per year as reported n Fgure 3. Over the entre sample perod 12 percent of acqurng frms announcng acqustons obtaned farness opnons compared wth 31 percent of the targets. 6. Determnants of Farness Opnons Table 3 summarzes the results of estmatng bnary logt regresson models for dependent varables that record whether or not acqurers and targets obtaned farness opnons or some form of fnancal advce. All of the coeffcents n ths and the followng table represent after The authors are currently nvestgatng whch of these s the reason for the decrease n farness opnons

15 15 changes n the log of the odds rato n favor of a frm usng a farness opnon or other fnancal advce when the ndependent varable changes by one unt. The values are, consequently, not as easly nterpretable n magntude as normal regresson coeffcents. Fortunately, the sgns and sgnfcance levels of the ndvdual coeffcents, shown as p-values n the table, and the characterstcs of the overall regresson are the more mportant results. In Tables 2 and 3 we have chosen to show only a sngle overall goodness-of-ft measure, McFadden s R-squared. We choose ths measure because of ts ease of nterpretaton as an R-squared showng the relatve explanatory power of the equaton and ts wdespread acceptance as a vald measure. 7 Model 1 n Table 3 presents the results for farness opnons obtaned by acqurng frms. Many of the sgns are as expected and most of the varables are hghly sgnfcant as ndcated by the low p-values. Thus a number of our basc hypotheses regardng the major determnants of the decsons of acqurers to obtan a farness opnon are supported by the data. As expected, the effect of the sze of the target frm, as measured by market value, s postve and sgnfcant, supportng the hypothess that nformaton s more valuable when there s more value at stake. In addton we fnd that the effect of the sze of the acqurng frm, also measured by market value, s negatve and sgnfcant. Ths may ndcate that larger acqurers are smply less concerned about the costs of acqustons of any gven sze. In consumer demand theory ths phenomenon s known as the budget effect: the smaller the proporton of expendture on an tem s n a consumer s overall budget, the lower the consumer s senstvty to the expendture. 7 Many aggregate measures of goodness of ft have been proposed for logt regressons and statstcal software packages often present a number of them. Wllam H. Green s among the econometrcans who recommend the use of McFadden s R-squared. (Wllam H. Greene, Econometrc Analyss, 4 th ed., Prentce-Hall Englewood Clffs, N.J. 2000) All of the equatons are hghly sgnfcant accordng to a varety of measures. For example, the p-values of the lkelhood rato statstcs for all of the equatons are 0.

16 16 The nformaton asymmetry varables have mxed effects n Model 1.We fnd that the relatve volatlty of excess returns for the two frms has a postve and sgnfcant effect on the probablty of obtanng a farness opnon. We nterpret ths result as showng that frms that have experenced more volatlty (and for whom the market assessment s that the market has less relatve nformaton about them) are more wllng to ncur the cost of obtanng a farness opnon because they do want to negatvely mpact ther volatlty further through an unwse acquston. Informaton asymmetry s greater when frms are more dssmlar or when they are not close accordng to our measure. As expected, the closer frms are to each other the lower s the probablty of obtanng a farness opnon. The closeness measure was also found to have a nonlnear effect: ts negatve effect on the probablty obtanng a farness opnon dmnshes as the dstance between the frms ncreases, as shown by the postve sgn on closeness squared. The closeness results may also be nterpreted as ndcatng that frms contemplatng horzontal mergers are more lkely to seek farness opnons than frms nvolved n vertcal or conglomerate mergers. Informaton asymmetry s also expected to ncrease n more concentrated ndustres n whch one of the results of market power may be less of a need to reveal nformaton. The postve sgn n Model 1 s results shows that ndustry concentraton n the target frm s ndustry does ncrease the probablty of an acqurer payng for the addtonal nformaton provded by a farness opnon. The varable s squared to reflect a nonlnear effect of ncreasng concentraton. 8 The varable s p-value s only.2171, ndcatng nsgnfcance at any of the usual sgnfcance 8 A Hrshman-Herfndahl Index (HHI) s often used as a measure of concentraton. HHI captures the nonlnear nature of the effect of rsng concentraton by usng the square of market shares. HHI values are sometmes computed usng only the data n Compustat, whch contans data only for exchange lsted frms. We felt that beng able to use concentraton ratos based on complete data for all frms n all ndustres was more mportant than not usng HHI.

17 17 levels. However, because of the theoretcal support for the varable, ts correct sgn, and t s z- value exceedng one, we choose to retan t n the equaton. As expected the varable whch desgnates that a transacton wll be completed wth a cash (or cash-lke) payment as the Medum of Exchange has a negatve and sgnfcant sgn showng that when the value of the payment component of a transacton s certan, there s a reduced probablty of a farness opnon beng obtaned. Ths effect s ndependent of the market value of the transacton as ndcated by a low correlaton between the target s market value and the use of cash. The results for the Atttude of the Deal beng Hostle accordng to the SDC data are not as hypotheszed: nstead we adopt an alternatve explanaton that our negatve and sgnfcant coeffcent ndcates that, when a transacton s hostle, the acqurer may have commtted to gan control by purchasng the assets and value s no longer an ssue to the acqurer. In ths case the acqurer does not need a farness opnon to help wth the decson. In the postve and sgnfcant coeffcent for Other Fnancal Advce Obtaned, we fnd evdence of complmentarty rather than the substtutablty among types of nformaton that we had hypotheszed. It appears that, when acqurers obtan other knds of fnancal advce, they are more lkely to obtan a farness opnon. There may be a behavoral nterpretaton of ths result as well: more rsk-averse acqurers can attempt to reduce rsk as much as possble by obtanng as much nformaton as possble ncludng not only farness opnons but also other knds of fnancal advce. The postve and sgnfcant value for the coeffcent on Farness Opnon Obtaned, Target may be evdence that, to some extent, there s symmetry n the need for nformaton n some transactons such that whatever the crcumstances are whch compel target frms to obtan farness opnons, the same crcumstances may also motvate acqurers to obtan farness opnons. For example, when general market condtons produce a hgh degree of

18 18 uncertanty regardng the future value of a target, both the target tself and the acqurer wll be more lkely to obtan a farness opnon. The fnal varable n Model 1 s the dummy varable whch dvdes the sample tme perod nto pre- and post- the Smth v. Van Gorkom decson, ndcated by the Van Gorkom Dummy varable. The fndng that t s postve and sgnfcant (at the 6.6% level) s as expected: followng the Van Gorkom decson, boards of drectors have become less wllng to make merger and acquston decsons wthout a formal farness opnon. However, the expectaton that, because Smth v. Van Gorkom was decded n a Delaware court, ths effect would be especally strong for frms ncorporated n Delaware was not supported by the data. Dummy varables for Delaware ncorporaton for the acqurer, for the target and for both were all found to be nsgnfcant n all models. We beleve that there are several reasons for ths fndng. Frst, the Van Gorkom decson has been used outsde of Delaware and thus t s of sgnfcance to frms not ncorporated n Delaware. Westlaw shows t cted n 90 non-delaware cases snce the year of the decson (1985). Second, Van Gorkom's prmary holdng nspred a Delaware statute the followng year that actually superseded t and then as many as 36 other states followed sut wth smlar statutes. Fnally, a number of states have adopted Delaware statutory or decsonal law on corporatons and, hence, would automatcally follow Van Gorkom or the Delaware statute wthout explct acton. After examnng the results of estmatng Model 1, a number of alternatve specfcatons were consdered. An obvous alternatve measure for the market values s relatve market value, measured as the sze of the acqurer relatve to the sze of the target whch also provdes a way to combne these effects. Table 3 also shows the results of several alternatve specfcatons for Model 1. In Table 3, Model 3 shows the results of usng the rato of the two market values n the

19 19 specfcaton. As can be seen, the result s that the relatve value measure s nsgnfcant and that the R-squared value falls as well. However, as seen n Model 4 when the log of the relatve market value measure s used, the coeffcent s sgnfcant and the R-squared value mproves as well. When the two market values are entered separately as logs, the varable representng the relatve volatlty of returns becomes nsgnfcant. When t s dropped from the equaton, the result s Model 2. Model 2 has the hghest R-square of any of the alternatve specfcatons tred for Model 1. We prefer Model 1 s specfcaton only because the relatve volatlty measure has such strong theoretcal justfcaton. In Model 5 we show an alternatve whch utlzes the merger premum and the log of the rato of excess returns. The latter s hghly sgnfcant (wth a p-value of.01) just as t was n level form n three of the frst four models. However, the merger premum s not sgnfcant n ths equaton, nor was t found to be sgnfcant n any of the other equatons for acqurer farness opnons. Other varables that were ntroduced nto the equatons but were found to ether have no meanngful sgnfcance or, when ntroduced as alternatve measures of concepts already n the model, were found not to produce as strong results. The level of annual aggregate merger actvty, as measured by the number of transactons among the 7818 n our sample, was found not to be sgnfcant n any of the specfcatons. The book value of assets s an alternatve measure of frm sze that was entered nto the equatons but dd not provde as much explanatory power as market value. We attempted to use ndustry dummy varables as these have often been found to be sgnfcant condtonng varables n other studes. We used Fama and French s ndustry defntons and ntroduced a full set of dummy varables nto the regressons but found them to be nsgnfcant as a group. In fact, only 2 ndvdual ndustres had even modest sgnfcance. For ths reason we omtted the ndustry dummes from the models reported heren.

20 20 We also attempted to use the Fama and Frecnch ndustres to obtan an alternatve measure of closeness. We defned a dummy varable that had a value of 1 f both the acqurer and the target were from the same Fama-French ndustry and 0 otherwse, but the results were that ths varable had less explanatory value n the models than the combnaton of our Closeness and Closeness-squared varables. We attempted to ntroduce ether of the tme varables dentfed n secton 3, the year of the transacton and a contnuous tme varable for the days wthn the perod, but nether was found to contrbute any sgnfcance to the explanatory power of the equatons. Table 4 presents results of estmatng the models for the decson of the target frm to obtan a farness opnon. The models whose estmaton results are n Table 4 parallel those for acqurers n Table 3. Agan the alternatve models represent experments wth the functonal form of the regresson equatons, especally regardng whether the two frm sze measures should enter separately or as a rato wth both forms ether lnear or logged. In contrast to the acqurer models, the target models do not provde support for many of the theoretcal hypotheses. Perhaps ths should not be surprsng gven that the dstrbuton of the asymmetrc nformaton n the market favors the target frms. However, ths consderaton should also lead target frms to less frequently obtan farness opnons whle the observed dstrbuton goes the other way. Only 12% of acqurers and over 30% of targets obtaned farness opnons. Examnng the results varable by varable n an attempt to gan an understandng of why the target models do not work well s not very revealng. None of the fve models n Table 4 has sgnfcant coeffcents for both market value and the rato of excess returns. Probably an as yet undscovered combnaton of these concepts needs to be found. Some results are very surprsng, such as the falure of the cash transacton varable to contrbute as t dd for the acqurer models. Smlarly, t appears that

21 21 targets are unconcerned about the ramfcatons of the Van Gorkom decson snce t s not sgnfcant n any of the models. The asymmetrc nformaton varables, closeness and concentraton appear to have no effect on the targets decsons to obtan farness opnons. There are only two consstently sgnfcant explanatory varables. The frst s the complmentary nput varable, ndcatng that other fnancal servces were obtaned by the target. The second s the ndcator that there are as-yet unmeasured general condtons affectng the frm value, of whch both acqurer and target are aware, as measured by a farness opnon beng obtaned by the acqurer. Gven the much less sgnfcant results t s not surprsng that the McFadden R-squared values n Table 4 are also sgnfcantly lower, the largest beng only.18 whereas n Table 3 Model 2 has an R-squared of.29. Are farness opnons really a specal knd of nformaton not contaned n other types of nformaton that frms can obtan? We have attempted to address ths queston. The SDC database records the numbers of fnancal advsors lsted by both acqurng and target frms. It s possble that ths knd of fnancal advce s a substtute for a formal farness opnon. In our sample almost half (49.5%) of all acqurng frms that had any nformaton of ths type lsted fnancal advsors whereas only 12.1% obtaned a farness opnon. Because a small number of frms obtaned farness opnons and dd not lst other fnancal advsors, the total percentage of acqurng frms wth ether a farness opnon or a fnancal advsor rses to 50.8%. The fgures for target frms are: 31.1% farness opnons, 71.8% fnancal advsors and 71.9% ether. To get an dea of whether the nclnaton of frms to obtan any knd of fnancal servces, ncludng but not lmted to farness opnons, s dfferent from ther decsons regardng farness opnons alone, the models n Table 5 were estmated. In ths case both the models for the acqurers and for the targets are better n some ways that the models for targets shown n Table 4

22 22 but nether s as good as the acqurer models n Table 3. Surprsngly, Model 12 for targets has a hgher R-squared value than Model 11 for acqurers. Models 11 and 12 also exhbt some curous sgns, such as the postve value of cash for target frms, although ths result mght ndcate that the when cash s gong to be receved, and there s no possble addtonal upsde gan followng the closng of the transacton, the target may want to obtan some addtonal assurances that full value s beng receved. Such behavor could be revealng a bas on the part of targets to beleve that noncash recepts are lkely to apprecate n value. In both Models 11 and 12 the Van Gorkom dummy varable s sgnfcant ndcatng that, whle the Van Gorkom decson addressed only farness opnons n a narrow sense, t s broader mplcatons have been to nduce a hgher level of consumpton of fnancal advsory servces of all knds. 7. Results Our results show that for the acqurng frm n an acquston, the lkelhood of purchasng farness opnons s nfluenced sgnfcantly by (1) the market values of the acqurer and the target frm, (2) the volatlty of excess returns of both frms, (3) whether or not the transacton s a cash deal, (3) the degree of asymmetrc nformaton as measured by the smlarty of the acqurer and target frms, (4) the amount of monopoly power the target frm has, (5) whether the acquston s hostle, and (6) whether other fnancal advsory servces have been purchased by ether frm. Fnally, strong evdence s found ndcatng that (7) the behavor of acqurng frms, whether ncorporated n Delaware or not, has been sgnfcantly altered snce the 1985 Van Gorkom v. Smth decson by a Delaware court regardng farness opnons. Our results for target frms are not as strong as those for acqurers, nor are the results for fnancal advsory servces more broadly defned.

23 23 8. Work n Progress The results reported above are encouragng, especally those for the acqurers. More systematc exploraton of alternatve model specfcatons and varable forms wll probably be productve even for acqurers. Consderable work s needed to refne a model for targets to the pont of havng results as relable and reasonable as those for acqurers. Frutful work can also be done to te the models n ths paper more drectly to those n the nformaton asymmetry lterature. Further exploraton of the legal and corporate governance aspects (board structure, etc.) also seems warranted. The authors are contnung to pursue each of the above ssues and welcome crtcsm and comments.

24 24 9. References Bebchuk, Lucan Arye and Marcel Kahan, 1989, Farness Opnons: How Far Are They and What Can Be Done About It?, Duke Law Revew, Berger, Phlp G. and El Ofek, 1995, Dversfcaton s Effect on Frm Value, Journal of Fnancal Economcs, Bzjak, J., J. Brckley, and J. Coles, 1993, Stock-based Incentve Compensaton and Investment Behavor, Journal of Accountng and Economcs 16, Boone, Audra, Laura Casarer Feld, Jonathan M. Karpoff and Charu G. Raheja, 2004, The Determnants of Corporate Board Sze and Composton: An Emprcal Analyss, Workng paper College of Wllam and Mary. Bowers, Helen M., 2002, Farness Opnons and The Busness Judgment Rule: An Emprcal Investgaton of Target Frms Use Of Farness Opnons, Northwestern Unversty Law Revew, Derkens Nathale, 1991, Informaton Asymmetry and Equty Prces, Journal of Fnancal and Quanttatve Analyss, Eleswarapu, Venkat R., Rex Thompson, and Kumar Venkataraman, 2004, The Impact of Regulaton Far Dsclosure: Tradng Costs and Informaton Asymmetry, Journal of Fnancal and Quanttatve Analyss, Elson, Charles M., The Duty Of Care, Compensaton, And Stock Ownershp, Unversty of Cncnnatt Law Revew, Fama, Eugene F. and Kenneth R. French, 1997, The Industry Costs of Equty, Journal of Fnancal Economcs, Fama, Eugene F. and Kenneth R. French, 1995, Sze and Book-to-Market Factors n Earnngs and Returns, Journal of Fnance, Fan, Joseph P. H. and Larry H. P. Lang, 2000, The Measurement of Relatedness: An Applcaton to Corporate Dversfcaton, Journal of Busness, Frankel, Rchard and Xu L, 2004, Characterstcs of a Frm's Informaton Envronment and the Informaton Asymmetry between Insders and Outsders, Journal of Accountng and Economcs, Gaver, J. and K. Gaver, 1993, Addtonal Evdence on the Assocaton Between the Investment Opportunty Set and Corporate Fnancng, Dvdend, and Compensaton Polces, Journal ofaccountng and Economcs 16,

25 25 Hare, M. Breen, 1999, Comment, The Fducary Responsbltes Of Investment Bankers In Change-Of-Control Transactons: In Re Dasy Systems Corp., New York Unversty Law Revew, Jensen, Mchael C.,1986, Agency Costs Of Free Cash Flow, Corporate Fnance, And Takeovers, Amercan Economc Revew, Kahle, Kathleen M. and Ralph A. Walklng, 1996, The Impact of Industry Classfcaton on Fnancal Research, Journal of Fnancal and Quanttatve Analyss, Kole, S., 1997, The complexty of compensaton contracts, Journal of Fnancal Economcs 43, pp Lang, Larry H. P. and Rene M. Stulz, 1994, Tobn's q, Corporate Dversfcaton, and Frm Performance, Journal of Poltcal Economy, Lehn, Kenneth, Sukesh Patro and Mengxn Zhao, 2003, Determnants of the Sze and Structure of Corporate Boards: , Workng Paper Unversty of Pttsburgh. Moeller, Sara B., Frederk P. Schlngemann and Rene M. Stulz, 2004, Do Acqurers Wth More Uncertan Growth Prospects Gan Less From Acqustons? Forthcomng Journal of Fnance. Moeller, Sara B., Frederk P. Schlngemann and Rene M. Stulz, 2004, Wealth Destructon on a Massve Scale? A Study of Acqurng-Frm Returns n the Recent Merger Wave, Forthcomng Journal of Fnance. Rhodes-Kropf, Matthew Davd Robnson and S. Vswanathan, 2001, Valuaton Waves and Merger Actvty: The Emprcal Evdence, Workng Paper Columba Unversty. Rhodes-Kropf, Matthew and S. Vswanathan, 2004, Market Valuaton And Merger Waves, Forthcomng Journal of Fnance. Shaw, Bll and Edward J. Gac, 1995, Farness Opnons In Leveraged Buy Outs: Should Investment Bankers Be Drectly Lable To Shareholders?, Securtes Regulaton Law Revew, Shlefer, Andre and Robert W. Vshny, 2003, Stock Market Drven Acqustons, Journal of Fnancal Economcs, Smth, C. and R. Watts, 1992, The Investment Opportunty Set and Corporate Fnancng, Dvdend, and Compensaton Polces, Journal of Fnancal Economcs 32,

26 26 Walker, M. Mark, 2000, Corporate Takeovers, Strategc Objectves, and Acqurng-Frm Shareholder Wealth, Fnancal Management, Yermark, D. 1995, Do Corporatons Award CEO Stock Optons Effectvely? Journal of Fnancal Economcs 39,

27 27 Fgure 1 Number of frms obtanng farness opnons by year Acqurers Targets Announcement year Fgure 2 Percentage of frms obtanng farness opnons by year % Acqurers Targets 60% 50% 40% 30% 20% 10% 0% Announcement year

28 28 Fgure 3 Dstrbuton of acquston announcements by year Year

29 29 Table 1 Dstrbuton of acquston announcements by Fama and French (1997) ndustry groups All acquston announcements from (Source: SDC) All acquston announcements ncluded n regresson results Acqurers Targets Acqurers Targets Industry N % N % N % N % Agrcrulture % % % % Food Products % % % % Candy and Soda % % % % Alcoholc Beverages % % % % Tobacco Products % % % % Recreatonal Products % % % % Entertanment % % % % Prntng and Publshng % % % % Consumer Goods % % % % Apparel % % % % Healthcare % % % % Medcal Equpment % % % % Pharmaceutcal Products % % % % Chemcals % % % % Rubber and Plastc Products % % % % Textles % % % % Constructon Materals % % % % Constructon % % % % Steel Works, Etc % % % % Fabrcated Products % % % % Machnery % % % % Electrcal Equpment % % % % Mscellaneous % % % % Automobles and Trucks % % % % Arcraft % % % % Shpbuldng, Ralroad Equpment % % % % Defense % % % % Precous Metals % % % % Nonmetal Mnng % % % % Coal % % % % Petroleum and Natural Gas % % % %