Risk and the Prediction of Successful Corporate Takeovers

Keith C. Brown and Michael V. Raymond

Keith C. Brown is Assistant Professor at the GraduateSchool of Business, Universityof Texas at Austin. Michael V. Raymondis a Financial Analyst at Electronic Data Systems, Dallas.

I. Introduction longed lack of recognition was the perception in the financial that on and Risk arbitrage, in its most common usage, refers to press gambling mergers acquisi- tions the skill and acumen of a Wall Street the purchase of the securities of a firm targeted for requires "insider" for Much and Laws acquisition. By purchasing shares immediately after a (see, example, [12] As of definite cash tender or exchange offer is established, [10]). Ivan Boesky, one its leading proponents, an investor is able to lock in the fixed spread that flatly stated, "the practice of risk arbitrage . .. is a craft that borders on an art" it is typically exists between the offer price and the post- [1, p. 15]. Ironically, himself who is credited with tech- announcement market price for the target firm. Of Boesky bringing the into the one of the first risk course, this price differential will only be guaranteedif nique open by forming the merger ultimately occurs. Thus, the primary func- arbitrage limited partnerships in 1975. Since then, in the to tion of the risk arbitrageuris to determine the likeli- competition market has steadily increased the where there are now a multitude of investment hood that a proposed takeover will succeed, thereby point and mutual funds in and substituting event risk for market risk in the process. pools specializing merger Until recently, risk arbitragehas been a game played acquisition arbitrage, offering smaller investors access to the field. The of this outside is almost exclusively by the institutional specialist. extent participation such that the as Welles [17] noted that although the strategy has been by middle of 1981 it was estimated that much as billion employed since the 1930s it was not until the merger $1.5 had been invested by nonprofes- wave of the last decade that the "shroud of mystery" sionals, with the professional arbitragecommunity in- a like amount. surroundingtakeover arbitrage was lifted for the gen- vesting Coincident with the in influ- eral public. Perhaps the biggest reason for this pro- rise outside investor ence has been an increase in the amount of information that becomes available about a Most The authors would to John proposed merger. like thank Howe, Scott Lummer, Robert has observed that investment Taggartand two anonymousreferees for theirhelpful suggestionson an notably, Rosenberg [13] earlier draft of this paper. bankerswill often consult with an arbitrageprofession-

54 BROWN AND RAYMOND/ARBITRAGE 55

al while the deal is still being structured.Thus, an ble on the terms of the merger, arbitrageursprovide assessmentof the market'spredisposition toward the liquidityto those stockholdersin the targetfirm who mergercan be embeddedin the exchange or tender are satisfiedtaking the smaller,but certain,profit that offer before any public announcementis made. This resultsfrom the announcementof the deal. In essence, notion is fully consistentwith the finding of Ruback then,the takeoverarbitrage market is one in whichless [14], who reportedthat successfuloffer prices are in- speculativeinvestors can transferany undesiredmerg- deed competitivelyset. Further,once the termsof the er-specificrisk to a risk professional.This is a point deal are announced,the arbitrageurs,many of whom that is missed by Hetherington[6], who apparently representlarge investment interests, must compete assumedthat any investorselling sharesof an acquisi- with one anotherin the open marketto securethe best tion candidateprior to the completionof a takeover price spreadpossible. One result of this competition was incapableof assessing the proposalefficiently. for informationis that risk arbitrageis no longer a Thebasic mechanics of the riskarbitrage investment strategyyielding abnormally high returns,as indicated are fairly simple. Supposethat a firm targetedfor ac- by Govoni [5]. Equallyimportant, with an increasein quisitionreceives an offerequivalent to $30 foreach of thenumber of informed"outside" participants, the pre- its outstandingshares. For simplicity,suppose further vailing prices of the firms involved will reflect the thatprior to the announcementof the deal the stock of consensusview of whetheror not the deal will eventu- boththe target and acquiring firms sold for $20 a share. ally be consummated.Of course, this is exactly the Afterthe mergerplans are publiclyrevealed, assume samereasoning used to suggestthat the stockmarket in thatthe investorsreact to the announcementby bidding generalsets prices so as to efficientlyreflect all avail- the price of the targetfirm's stock up to $28. At this able information. time, the arbitrageurmust judge whetherthe deal ap- Thepurpose of this articleis to arguethat, while risk pearssufficiently likely to go throughthat purchasing arbitragemay remainan arcaneart on the individual the targetfirm's stock in orderto capturethe $2 price level, competitionamong investors is such that the spreadis justified.If it does appearso, the exactstrate- marketsfor mergingcompanies are quiteefficient. As gy employeddepends upon the natureof the acquiring a result,it is shownthat an ongoingprediction as to the company'soffer. Witha cash tenderoffer, the arbitra- eventualsuccess of the mergercan be inferredfrom the geurhas only to purchasethe stockof the targetfirm in prices set in the post-announcementperiod. This pro- orderto lock in the pricedifferential. When shares are cess, whichis basedon the investmentstrategies of the exchanged,however, the post-announcementvalue of arbitrageur,complements the work of Lewellen and the exchangeoffer will vary with the price of the ac- Ferri[11 ] who firstconsidered the estimationof merg- quiringfirm's stock. Thus, in orderto lock in a particu- er probabilitiesfor acquisitionsinvolving share ex- lar spread,the stock of the acquiringfirm will have to change offers only. To demonstratethe technique,a be sold shortin appropriatequantities at the sametime sampleof announcedtakeover attempts utilizing both thatthe targetfirm's stock is purchased.In the present cash tenderoffers and stock swaps is gathered,and example,as $28 is spentfor a shareof the targetfirm, empiricalpredictions of a successfulmerger are then one and a half sharesof the suitorcompany will be generated.The majorfinding indicates that the market shortedat the assumedprice of $20. If the mergeris is able to discriminate,in a statisticallymeaningful then consummated,the investorcan cover the way, betweenmergers that will eventuallyfail or suc- sale with the exchangedvalue of the targetshare. ceed as faras threemonths in advanceof the respective The risk involved with such is, of events. This evidenceis interpretedas consistentwith course, thatthe attemptedmerger may neverbe com- the notion that the prices in the post-announcement pleted.If the dealproposed heretofore does indeedfall stage of a corporatetakeover attemptare set in an through,it is likely thatthe targetstock will returnto efficient and competitivemanner. its pre-announcementlevel and$8 couldbe lost. Con-

II. The Mechanics of Risk Arbitrage 'The price to which the target firm's stock will fall in the event of an un- unsuccessfultakeover attempt is a matterof considerableuncertainty. If The risk arbitrageuris a speculator.However, the pre-announcementprice of $20 already incorporateda degree of like a currencyspeculator concerned with overalleco- anticipationabout the merger proposal, then the subsequent decline nomic trends, the investor in an attemptedtakeover could be even more severe. However, if the takeover attempt, albeit the outcomeof flawed, has revealednew andpositive informationabout the targetfirm, cares only aboutthe factorsaffecting $20 may be the lower end of the post-failureprice range. This issue will the proposeddeal. Throughtheir willingness to gam- be considered in more detail in the empirical analysis that follows. 56 FINANCIALMANAGEMENT/AUTUMN 1986

sequently, the time and money spent researching the has concentrated on the reaction of the takeover on the front end is most important. However, either around the time of the initial announcement or while considerable resources may be expended, the after the resolution of the deal. Thus, while these stud- evaluation process is hardly a scientific endeavor. As ies do offer excellent insights into the ex post returns Boesky [1] noted, assessment of the probability of a realized by both the acquiring and target firms, they successful merger is essentially subjective, depending fail to give an adequate ex ante indication of the prob- not only on whether the deal makes economic sense ability that a merger will be completed. This knowl- but also on the personalities involved in the negotia- edge would seem to be particularly important given tions. Once the arbitrageurmakes this prediction, the that Dodd and Ruback [4], as well as Bradley, Desai, decision to invest rests on two remaining questions. and Kim [3], have shown that the shareholders of the First, is the existing differential between market and target firm tend to earn positive (zero) abnormal re- offer prices sufficient compensation for the perceived turns if the merger succeeds (fails). Second, while risk? Second, will the merger occur soon enough for Hoffmeister and Dyl [7] and Walkling [16] have adapt- the available price spread to translate into an adequate ed sophisticated statistical techniques for the purpose returnon the investment? To answer the first question, of predicting tender offer success, both models are notice that if $20 is thought to be a reasonable estimate designed to use only information available to a poten- of the post-failure price of the target firm, then the tial bidding firm before a formal proposal is made. arbitrageuris risking eight dollars to earn two. Thus, Thus, neither of these methodologies provides an ex- for the investment to make sense, it must be thought plicit means of evaluating the progress of an offer in that the merger has at least a four-out-of-five chance of the post-announcement period. Finally, the procedure succeeding. On average, then, 80% can be viewed as developed hereafter exploits the market's ability to set the "breakeven" probability inferred from the price prices for the target company in an efficient manner. In spread. The obvious implication here is that since indi- this sense, our methodology is in keeping with Lewel- vidual investors are helping to set the same post-an- len and Ferri's [ 11] argumentthat the securities market nouncement prices on which they base their decisions, should be considered the ultimate authority on the de- those prices must reflect the aggregate opinion of sirability of any corporate decision. However, while whether and when the merger will ultimately take Lewellen and Ferri focused on measuring the synergis- place. This will be true even though the requiredrate of tic benefits of two firms merging via an exchange of return is difficult to assess precisely since the risk in shares, the probability measure suggested hereafter al- acquisition arbitrage is almost wholly unsystematic. lows for a greater variety of offers while concentrating The significance of the preceding discussion is that on the price fluctuations of the target firm. In the case as a result of the competition among arbitrageursit is of cash tender offers, this segmentation is possible possible to infer an aggregate opinion about the likeli- because the risk arbitrage investment occurs on only hood that a merger will occur from observable market one side of the merger transaction. prices. Thus, it should be of considerable interest to Returning to the earlier example, consider the peri- see how accurate these forecasts are. In particular, is it od t decision of the arbitrageurwho can acquire at the possible for the market to consistently discriminate prevailing market price P,t a share in a firm for which between proposed takeovers that will ultimately suc- there is an outstanding tender offer at the price PTt.2 ceed and those that will fail well in advance of the This purchase will therefore yield a return of [(PTt- actual event? It is to just this topic that the empirical Pmt) Pmt]if the merger is completed. However, since experiment of the following sections is addressed. an announced tender offer is not necessarily an accept- ed one, there is also the chance that the merger will III. The Prediction of Successful ultimately fall through. Assuming that the price to Corporate Takeovers which the target's stock falls in the event of an unsuc- The purpose of this section is to use the information cessful merger can be written PF, the arbitrageuralso implicit in the arbitrageur's decision-making process faces the possibility of realizing a potentially negative to estimate the likelihood that a cash tender or ex- change offer will eventually be accepted. Before this is 2Theperiod t subscripton the offer price reflects the fact thatacquisition accomplished, three things should be mentioned. proposalsinvolving stock swaps, while maintaininga fixed exchange ratio, can in dollaramount over time. In this First, as noted Jensen and Ruback most of the change drastically context, by [8], the term "tenderoffer" is used to connote any definite form of merger literature concerning itself with corporate takeovers proposal. BROWN AND RAYMOND/ARBITRAGE 57

returnof [(PF - P) Pmt]. While this latter expres- absenceof competingbids the marketdecides that a sion cannotbe set in advance(since PFis not known mergeris certainwhen the period t marketprice is with certaintyat time t), the use of any of severalpre- equal to or greaterthan the prevailingtender offer. announcementprices as a proxy can still give the in- Conversely,the takeoverwill be labeleda certainfail- vestoran indicationof the overallrisk-return tradeoff. urewhenever the marketallows the stockto fall below Onceestablished, these parameters can thenbe usedto the estimated"failure" price.4 One potentialproblem set the market'sprediction of the likelihood(or, as will thatarises from this latterdefinition is thatif the mar- be discussed,the desirability)of the merger,which can ket has anticipatedthe public announcementof the in turnbe comparedwith the arbitrageur'sown subjec- merger,it is possible that using a pre-announcement tive assessment. price as a proxy for PFwill bias x, downwardsince To formalize this process, compute the period t (6xt/6PF) < 0 whenever Pm < PTt, which is typically merger probability (i.e., xt) such that the expected risk the case. Therefore,one consequenceof such a bias arbitragepayoff is zero. mightbe an overpredictionof mergerfailure. This is an empiricalquestion and one thatwill be considered - - - E(t) = Xt(PTt Pmt) Pmt] + (1 x)[(PF Pmt) Pmt] in subsequentsections.5 - - - = x[(Tt/Pmt) 1] + (1 xt)[(PF/Pmt) 1] =0. IV. Data and Methodology In orderto test the market'sability to predictthe Thus: eventualoutcome of an announcedtakeover attempt, a of and failed - sampleconsisting both successful merg- 1 (PF/Pmt) ers was collected from the 1980-1984 With Xt ,(1) period. - (PTt/Pmt) (PF/Pmt) several hundred acquisition proposals per year to choose from, it was possibleto refinethe database by where (PTt/Pmt)and (PF/Pmt) are defined as the tender imposing the following a priori constraints: (i) the andfailure premiums, respectively. To see moreclear- proposedtransaction must involve eithera mergeror ly the intuitionbehind this probabilitymeasure, re- an acquisition,i.e., divestituresor otherforms of reor- write Equation(1) in the following form: ganizationwere excluded;(ii) the announcementand completionor terminationdate of the deal, as well as - + X = (Pmt PF) (PTt- PF) (2) theterms of the offer, mustbe publiclyavailable infor- and the stock of both firmsinvolved must From it is to see thatthe mation; (iii) Equation(2) straightforward be listed. As an additionalmeans of estab- marketas a whole sets the likelihoodthat the takeover exchange the of all the will be successfulas the of the incremental lishing predictivepower Equation(3), percentage selectedwere classified to tenderoffer thathas been assimilatedinto the mergerproposals according already whetheror not the firms were involved in a market as of t.3 To ensurethat this mea- target price period These distinctionsleft surecan be as a rewrite competitivebidding process. strictlyinterpreted probability, three non- so thatit is restrictedto remainwithin the separatesubsamples: completed mergers, Equation(2) failed andcom- [0,1] interval: competing; attempts,non-competing; pletedmergers where at least one othercompeting bid failed. xt = Min{[Max{(Pmt- PF), } + (PTt- PF)] 1} (3) Two primarysources of informationwere used in By way of interpretingEquation (3), notice thatin the 4Fromthe perspectiveof a shareholderof the targetfirm, an outstanding tender offer can be judged successful if it is accepted or, if it fails, it new informationabout the to lead to a subse- 3Noticethat the probabilitymeasure is calculatedby takinga percentage conveys enough company offer. See Jensen and Ruback for more on this of the difference between the failure and tender prices, rather than quent [8, pp. 14-16] simply inverting the tender premium (i.e., Pmt/PTt). To see why it is point. necessaryto also account for the failure price, consider two targetfirms each receiving tenderoffers for $10 more than their initial shareprices 5Itshould be noted that a similarprediction method has been developed of $20 and $60. Assuming that the marketsubsequently reacts to these independentlyby Samuelsonand Rosenthal[15]. However, the empiri- announcementsby bidding the respective prices to $28 and $68, Equa- cal applicationoffered in [15] differs from that presentedsubsequently tion (2) would correctlypredict that each proposedmerger has the same in threeprimary ways: (i) the sample period runs from 1976-1981, (ii) (0.8) chance of success. However, if the reciprocals of the tender no non-cash exchange offers are considered, and (iii) all competitive premiums are used, the predictions would be (28/30) = 0.93 and tenderoffers are excluded from the sample. Nevertheless, the results of (68/70) = 0.97, respectively. the two studies seem to be robust with respect to these differences. 58 FINANCIALMANAGEMENT/AUTUMN 1986

Exhibit 1. Average Probabilities of Completed (Non-Competing and Competing) and Failed Merger Proposals, PF = Four Weeks Prior to Announcement Weeks Prior to Resolution 0 1 2 3 4 5 6 7

Completed (Non-Competing): Average 0.984 0.964 0.951 0.932 0.913 0.882 0.849 0.842 Std. Deviation 0.023 0.044 0.050 0.047 0.066 0.111 0.105 0.106 Sample Size 36 36 36 36 36 36 36 36 t-statistic 257.7 131.5 114.1 119.0 83.0 47.7 48.5 47.7 Completed (Competing): Average 0.934 0.946 0.923 0.884 0.867 0.864 0.880 0.925 Std. Deviation 0.098 0.070 0.099 0.222 0.238 0.246 0.227 0.081 Sample Size 18 17 17 17 17 16 15 13 t-statistic 40.4 55.7 38.4 16.4 15.0 14.0 15.0 41.2 Failed: Average 0.209 0.413 0.421 0.454 0.489 0.435 0.436 0.420 Std. Deviation 0.284 0.303 0.323 0.311 0.310 0.282 0.265 0.244 Sample Size 35 34 33 29 26 23 20 20 t-statistic 4.35 7.95 7.49 7.86 8.04 7.40 7.36 7.70 All average probabilitiesare significant at the 0.05 level. the selection process. For the non-competitive propos- quires three pieces of information: the prevailing mar- als, both the "Roster" and "Out the Window" sections ket price of the target firm, the offer price, and the in various issues of Mergers and Acquisitions were price to which the target firm's stock will fall if the deal used to locate a total of 71 representative takeover is cancelled. All but the latter are directly observable attempts. Announcement dates as well as the initial and so an estimate must be used for the value of PF. For and, if applicable, revised offer terms were gathered purposes of this investigation, four different proxies from The Wall Street Journal. Of these merger candi- were selected. These are the actual marketprices of the dates, 36 were successful and 35 ultimately failed. target firm one, two, three, and four weeks prior to the Further, approximately half of the proposed non-com- initial announcement of the attempted takeover. Thus, peting deals were purely for cash while the other half for each of the merger proposals in the sample, four involved stock swaps or combinations of stock and separate sets of probabilities were calculated. cash. The competitive subsample, on the other hand, As discussed earlier, the risk arbitrageprocess takes was comprised solely of acquisition attempts relying place in the period between the announcement and on cash tender offers by all the bidding firms. Using resolution of the takeover attempt. Using these two the Tender Offer Statistics data base provided by dates as endpoints, weekly price information was ob- Douglas Austin & Associates, a set of 18 target firms tained from The Wall Street Journal and probabilities satisfying the preceding requirements was isolated. were computed over the entire investment interval. In This data base, which collects information directly other words, starting with the day on which the pro- from tender offer forms (14D-1) filed with the SEC, posed deal was either completed or terminated, a pre- provided all the necessary announcement and offer diction was made according to Equation (3) for each terms for the competing tender offers. Interestingly, priorweek all the way back to the announcementdate.7 the 18 target firms selected represent the full set of 6It also should be mentionedthat none of the mergerproposals sampled companies receiving multiple registered bids during involved "two-tiered"bids, wherein the bidding firm makes its initial the sample period. While this would seem to be a offer for only a portionof the targetcompany's shares. To evaluate this small number of cases, two should kind of deal, the risk arbitrageurmust estimate a "blended"tender offer surprisingly points as the of all received the shareholdersof the be in mind. as noted Knoeber tender weighted average prices by kept First, by [9], targetfirm. For more on both the practicaland technical aspects of this offers accounted for only about 5-10% of all merger process, see Boesky [1] and Bradley [2]. announcementsduring these years. Second, since only bids were considered, all rumorsor 7For notational convenience in the work that follows, a "week" is formally registered defined to be a unit of number to five business were consisting any up days. about proposed takeovers ignored.6 For instance, a proposed merger announcedon Tuesday, May 1 and The probability measure given by Equation (3) re- terminatedon Friday, May 25 would be considered four weeks long. BROWN AND RAYMOND/ARBITRAGE 59

Exhibit 1. (Continued) maximum number of observations as of the last day in Weeks Prior to Resolution the investment interval. It will be the difference be- that indicates the stock 8 9 10 11 12 tween these weekly averages market's ability to accurately anticipate the outcome of 0.830 0.798 0.785 0.767 0.743 a proposed takeover. 0.121 0.190 0.172 0.171 0.158 36 35 32 30 28 V. Empirical Results 41.1 24.8 25.8 24.6 24.9 As a starting point, the prediction measure of Equa- tion was calculated on a basis for all 89 0.774 0.750 0.705 (3) weekly 0.914 0.846 firm four 0.095 0.273 0.360 0.377 0.295 proposed deals using the price of the target 11 7 6 6 4 weeks prior to the announcement date as a proxy for 31.9 8.20 5.27 4.87 4.78 the "failure" price, PF. The probabilities for a given week relative to the resolution date were then averaged 0.438 0.423 0.375 0.391 0.442 over both of the These cross-sectional 0.274 0.256 0.265 subsamples. 0.235 0.290 devi- 18 17 13 11 10 averages, along with the associated standard 7.91 6.01 4.93 5.07 5.27 ations, are reported for the three months preceding the completion or failure of the merger in Exhibit 1. The results are presented in a graphical form in Exhibit 2.8 The first fact to from an examination of Since this period of time lasted anywhere from three to emerge Exhibit 1 is that for all the weeks, some method of standardizing the every subsample weekly forty-six are different from results was necessary. Given the stated purpose of the average probabilities significantly zero. If else, this serves as an indication that empirical analysis, the probability estimates were nothing the used produces results of con- averaged on a weekly basis for all three subsamples. prediction technique Each set of average probabilities was initially calculat- 8As all of the tests summarizedbelow were con- ed from the at which the uncertainty was finally mentioned, empirical point ductedwith four differentproxies for PF. However, given that identical resolved (i.e., the ultimate failure or completion date). inferenceswere generatedby each of the various"failure" prices, only a Consequently, each partition of the sample contains its representativeset of results will be listed.

Exhibit 2. Average Probabilities of Completed (Non-Competing and Competing) and Failed Merger Proposals, PF = Four Weeks Prior to Announcement

~1. 0 VCompleted-C 0.9 -_ 0.8 - 0.7 - Completed- NC 0.6 - PROBABILITY 0.5 OF TAKEOVER --- 0.4 *- --*·'^ a--* ·--- ·^ 0.3 - .~/ ~Failed 0.2 1 0.1

0.0 I I I I I I I I I I I 0 1 2 3 4 5 6 7 8 9 10 11

NUMBEROF WEEKSTO RESOLUTION 60 FINANCIALMANAGEMENT/AUTUMN 1986

Exhibit 3. Difference in Average Probabilities between (i) Completed (Non-Competing) vs. Failed and (ii) Completed (Competing) vs. Failed Merger Proposals, PF = Four Weeks Prior to Announcement Weeks Prior to Resolution 0 1 2 3 4 5 6 7

Completed-NC vs. Failed: Avg. Difference (Dt) 0.775 0.551 0.530 0.478 0.424 0.447 0.413 0.422 Pooled Std. Dev. (spt) 0.200 0.213 0.226 0.210 0.206 0.196 0.178 0.168 t-statistic 16.3 10.8 9.73 9.11 7.98 8.56 8.30 9.01 Completed-Cvs. Failed: Avg. Difference (D,) 0.725 0.553 0.502 0.430 0.378 0.429 0.444 0.460 Pooled Std. Dev. (Spt) 0.239 0.252 0.270 0.282 0.284 0.268 0.250 0.237 t-statistic 10.5 7.13 6.23 4.99 4.27 4.92 5.21 5.68 All differences in average probabilitiesare significant at the 0.05 level (except where denoted by an asterisk). sistent quality. Beyond this point, however, the prob- Another difference in the weekly mean probabilities abilities generated by the failed and completed merger is the manner in which the market revises its predic- cases are apparently quite different. For instance, no- tions as the resolution date draws closer. As seen most tice that over the 12-week period leading up to the clearly from Exhibit 2, the average probability for the completion of either a competitive (C) or a non-com- successful merger subsamples steadily increases up to petitive (NC) merger, the market never forecasts less the time of completion. Conversely, while it was noted than a 70% chance that the proposed target will be heretofore that the unsuccessful merger attempts are acquired. This is particularly impressive in the case of never considered to be very promising, there appears the competitive takeovers since the market was able to to be no substantial downward reassessment as the correctly evaluate the target firm's acquisition, despite failure date approaches. To formalize this, consider the fact that at least one of the competing bids failed. regressing the average probability for the completed Conversely, at no time during the reported sample (CMPRB-NC and CMPRB-C) and failed (FLPRB) period does the average probability for the failed merg- takeovers against the number of weeks to resolution er candidates ever exceed 0.5. From the outset, then, (WTR). The results of these equations, with the t- investors in the arbitragemarket set prices that indicate statistics listed parenthetically, are as follows: that it is far more likely that these takeover attempts - - will eventually be terminated. This fact is notable (CMPRB-NC)t= 0.9867 0.0203 (WTR)t;R2=0.99 since, for the 35 failed studied, the average (284.4) (-41.5) mergers = 0.9584 - 0.0160 R2=0.68 ...,12 of time between the announcement and resolu- (CMPRB-C)t (WTR)t; t=0,1 length (40.7) (-4.80) tion dates was only 9.20 weeks. Therefore, from the (FLPRB)t = 0.3818 + 0.0049 (WTR)t;R2=0.08 inception of the typical failed takeover, the market (10.9) (0.99) provided ample indication of the tenuous nature of the deal. On the other hand, the mean length of time it took As the disparity between the coefficients of determina- to complete the 36 non-competitive and 18 competitive tion ratherclearly reveals, only the predictions for the successful mergers was 17.31 and 9.83 weeks, respec- completed mergers are revised with any consistency. tively. Thus, as with the failed subsample, the average By directly comparing the t-statistics for the respective length of time that it takes to acquire a target firm with parameterestimates on the independent variable it is multiple bidders is shorter than the three months of seen that only the failing acquisitions yield an insig- data reported. Further, although not shown on the dis- nificant temporal relationship. Consequently, it can be plays, even 18 weeks prior to resolution the average argued that the two completed subsamples are also probability of the 13 non-competing mergers that took differentiatedfrom the failed grouping according to the this long to complete is 0.671 with a standarddeviation process by which the weekly probabilities are of 0.142. The t-statistic of 17.04 implied by these generated. figures is also significant at the 5% level. These find- A closer inspection of Exhibit 1 reveals that even ings might therefore be interpretedas an initial demon- after the termination of a takeover attempt (i.e., week strationof the market's ability to discriminate between zero) the market prediction of success is still about the various types of acquisition proposals. 20%. Further, the week zero standarddeviations indi- BROWN AND RAYMOND/ARBITRAGE 61

Exhibit 3. (Continued) Here,NFt and Nct represent the week t numberof failed Weeks Prior to Resolution andcompleted takeovers, respectively. The resultsof 8 9 10 11 12 comparingeach of the successfulmerger subsamples with the unsuccessfulmerger group are in Exhibit3. 0.392 0.375 0.410 0.376 0.301 To be conservative,specify the degreesof freedom 0.167 0.227 0.206 0.196 0.190 for a particularvalue of Dt to be the minimumof 8.13 5.59 6.06 5.43 4.29 {NFt- 1, Nt- 1}. Also, the significance tests will be conducted a two-sidedalternative even 0.487 0.491 0.471 0.383 0.308 against though 0.187 0.235 0.274 0.295 0.310 a one-sided alternativewould be appropriatesince 7.14 5.40 3.67 2.57 1.93* thereis no a priorireason to believethat D, will everbe negative.The t-statisticsfor each of twelve weeks pre- ceding the resolutionof the attemptedtakeover are displayedby comparisongrouping. Finally, it should again be noted that all of the tests describedin this cate that the markethas more trouble coming to a section were replicatedusing three other proxies for definitiveconclusion about the failed mergersubsam- PF. In every case, inferenceswere identicalto those ple. Bothof theseresults are caused by the fact thatthe reportedsubsequently and thereforewere omitted. actualpost-failure price for the targetfirm is above its Perhapsthe most strikingresult found in Exhibit3 is estimatedlevel. The most plausible explanationfor theremarkable similarity in the probabilitiesgenerated this occurrenceis that, althougha particularmerger by both competingand non-competingacquisition at- may fall through,the marketrecognizes that the target tempts.With a single exception,the averageprobabil- firm is still an attractiveacquisition, even if no other ities for each of the completedsample partitionsare offer is currentlyavailable. In the presentsample this significantlydifferent from those of the failed merger is exemplifiedby FlagshipBanks, whichwas success- proposalsas far as three months in advance of the fully mergedwith Sun Banksin 1983 afteran effortby respective events. This is particularlynoteworthy Royal Trustcotwo years prior had failed. If this is when it is recalledthat the averagelength of both an indeedthe case, the probabilitymeasure summarized unsuccessfulacquisition and a competitivetender offer by Equation(3) must be consideredto be not only a is less thanten weeks. Since this findingis consistent predictionabout a currenttakeover attempt but also an over a wide rangeof estimatedvalues it mustbe con- expressionof the perceivedlikelihood that the target cludedthat prices set in the risk arbitragestage of a firm will be involved in a futuredeal. In theory, this mergercan significantlydiscriminate between eventu- interpretationclearly makes it more difficultto sepa- al outcomes.Thus, thereis strongevidence to suggest rate failed from completed mergers, particularlyin thatthe averageprobabilities generated by failingtake- situationsinvolving competingtender offers. From a overattempts are significantly lower than those gener- practicalstandpoint, however, it merelyraises an em- ated by successfulones. pirical issue which requiresan alterationin the test Before concluding,two final points meritsome at- methodology. Specifically, a modification will be tention. First, it may be of some interestto consider madeso as to directlytest the abilityof the prediction how accuratelythe estimatedvalues of PFforecast the measureto differentiatethe two completedgroupings actualpost-failure prices. To do this, for eachof the 35 fromthe failed subsample.This can be accomplished failedmergers compute the cross-sectionalaverage ra- by definingDt as the week t differencein the valuesof tio of the actualprice of the targetfirm j weeks afterthe the for averageprobabilities one of the completedand terminationdate (PN+j)and the pricej weeks priorto the failed A mergersamples. t-statisticfor the signifi- the announcement date (PA_j).That is, for each of the cance of this measureis the given by ratio: failed takeovers, calculate (PN+j) - (PA-j) and com- putethe averagefor all for proxyweeks j. The result- T = * Dt {pt \/[(l/NFt) + (l/N,)]}, (4) ing meanratios for j = 1,2,3,4 are 1.04, 1.09, 1.10, and 1.11, respectively,none of which is significantly where representsthe cross-sectionalstan- sp, pooled, differentfrom one at the 0.05 level.9 Thus, it appears darddeviation for the respectivesubgroups, i.e.:

- + + Spt = {[(SFt2 NFt) (NFt Nt)] 9Therespective cross-sectional standard deviations for the fourcases are [(sc,2 Nc,) (NFt + N,)]}. (5) 0.28, 0.35, 0.35 and 0.46 witht-statisticsof 0.85, 1.60, 1.67 and 1.49. 62 FINANCIALMANAGEMENT/AUTUMN 1986

Exhibit 4. Average Probabilities of Failed (Competing and Non-Competing) Merger Proposals, PF = Four Weeks Prior to Announcement WeeksPrior to Resolution 0 1 2 3 4 Failed (Competing): Average 0.887 0.951 0.904 0.819 0.641 Std. Deviation 0.112 0.096 0.286 0.337 0.447 Sample Size 18 17 12 8 6 t-statistic 33.60 40.79 10.94 6.87 3.51 Failed (Non-Competing): Average 0.209 0.413 0.421 0.454 0.489 Std. Deviation 0.284 0.303 0.323 0.311 0.310 Sample Size 35 34 33 29 26 t-statistic 4.35 7.95 7.49 7.86 8.04 All averageprobabilities are significantat the 0.05 level. that all of the estimated "failure" prices used are rea- however, the price of the target firm in the week before sonable forecasts of the actual values. the successful bid is issued was already an average of The second point worth considering is how the mar- 0.4% higher than the level of the initial offer. Thus, ket evaluates tender offers which will eventually be even when a particularmerger proposal fails to achieve outbid. For the competing tender offers examined, the the desired result, participants in the risk arbitrage average time between the initial and second registered marketare able to accurately evaluate the nature of the offers was 2.39 weeks. Since the mean length from the failure and arrive at the correct conclusion about the first offer to the eventual acquisition was about two ultimate fate of the target firm. and half months, it is apparent that competitors for a By way of summarizing the results of this section, it target firm formally reveal themselves relatively early should be recalled that the question to be addressed in the bidding process. The more important issue, was whether or not the prices set in the post-announce- however, is how the price of the target firm reacts ment stage of a corporate takeover reflect enough in- during the time between when a proposal is initiated formation to differentiate between eventual outcomes. and when it is rejected. Specifically, are the probabil- According to findings contained in Exhibit 3, the an- ities set by the market similar for all unsuccessful of- swer to this question is clearly yes. As indicated by fers, regardless of whether they failed for competitive both the competitive and non-competitive subsamples, or non-competitive reasons? To consider this question successful and unsuccessful takeover attempts do im- more carefully, Exhibit 4 lists the average probabilities ply significantly different estimated probabilities. for both of the relevant subsamples in the four weeks prior to the respective failure dates. For the competi- VI. Conclusion tive grouping, these four weeks correspond to the time Speculating in merger and acquisition arbitrage, immediately prior to the registration of the eventually once the domain of a few well connected insiders, has successful bid while the non-competitive probabilities recently become an increasingly competitive endeav- are replicated from Exhibit 1. Not surprisingly, the or. The purpose of this paper has been to demonstrate results indicate that the probabilities are substantially that an importantimplication of this speculative activ- higher in the competitive subsample. Since the only ity is that the market prices of the firms involved in a apparentdifference between the two groups is the pres- takeover attempt will necessarily reflect the attitudes ence or absence of alternative bidders, these findings of the investing public. In particular, it was argued that undoubtedly reflect the market's ability to anticipate through the actions of the arbitrageurit is possible to larger (or more attractive) subsequent offers. Interest- use the post-announcement price of the target firm to ingly, the results of Exhibit 4 may actually understate infer the probability that an acquisition will ultimately this disparity due to the fact that Equation (3) restricts take place. To demonstrate this point, it has been the probability measure to equal one whenever the shown that successful and unsuccessful merger pro- current market price of the target firm exceeds the posals do in fact imply significantly different probabil- existing tender offer. For the competitive sample, ities well in advance of the eventual outcome. Further, BROWN AND RAYMOND/ARBITRAGE 63

it was indicatedthat this predictionprocess is robust 4. P. Dodd and R. Ruback, "TenderOffers and Stockholder withrespect to the only piece of informationthat is not Returns: An Empirical Analysis," Journal of Financial 351-374. directlyobservable; that is, the post-failureprice of the Economics (December 1977), pp. firm. it was concludedthat the 5. S. J. Govoni, "Tough Stretch for Arbitragers,"Financial target Thus, arbitrage 96-97. marketdoes differentiatebetween the two World (February6-19, 1985), pp. effectively 6. N. S. the Risk Out of Risk Arbi- of takeover Hetherington,"Taking types corporate attempts. Journal of Portfolio Management(Summer 1983), The direct of the resultsin this is trage," suggestion study pp. 24-25. thateven thoughthe risk associatedwith the arbitrage 7. J. R. Hoffmeisterand E. A. Dyl, "PredictingOutcomes of processis largelyunsystematic, the marketcan none- Cash Tender Offers," Financial Management (Winter thelessbe useful in helpingto assess the resolutionof 1981), pp. 50-58. an uncertainacquisition. It shouldbe quite reassuring 8. M. C. Jensen and R. S. Ruback, "The MarketFor Corpo- to both the private investor and managementof the rate Control:The Scientific Evidence," Journal of Finan- mergingfirms that the pricesset in the post-announce- cial Economics (April 1983), pp. 5-50. R. Shark and mentperiod reflect prevailingattitudes and thatthese 9. C. Knoeber, "GoldenParachutes, Repellants Hostile Tender American Economic Review attitudesare discriminating predictors of futureevents. Offers," While still be "artists"at (March 1986), pp. 155-167. professionalarbitrageurs may 10. M. "Risk and Reward: The Game Pro- their work Laws, Arbitrage forming subjective merger probabilities, vides Plenty of Both," Barron's (November30, 1981), pp. may appropriatelybe likenedto thatof the fundamen- 9-20. tal analystin an efficient stock market.At the very 11. W. G. Lewellen and M. G. Ferri, "Strategiesfor the Merg- least, it now appearsthat the aggregateproduct of the er Game: Managementand the Market,"Financial Man- risk arbitragecommunity is less mysteriousthan the agement (Winter 1983), pp. 25-35. methoditself. 12. M. Much, "Arbitragers:Wall Street's Mystery Men," In- dustry Week (October 1, 1979), pp. 68-72. 13. H. Rosenberg, "Lowering the Risk in Risk Arbitrage," Financial World(February 22-March 6, 1984), pp. 26-28. References 14. R. S. Ruback, "Assessing Competitionin the Marketfor 1. I. F. Boesky, MergerMania-Arbitrage: Wall Street's Best CorporateAcquisitions," Journal of Financial Economics KeptMoney Making Secret, New York, Holt, Rinehartand (April 1983), pp. 141-153. Winston, 1985. 15. W. Samuelson and L. Rosenthal, "Price Movements as 2. M. Bradley, "InterfirmTender Offers and the Marketfor Indicatorsof Tender Offer Success," Journal of Finance CorporateControl," Journal of Business (October 1980), (June 1986), pp. 481-499. pp. 345-376. 16. R. A. Walkling, "PredictingTender Offer Success: A Lo- 3. M. Bradley, A. Desai, and E. H. Kim, "The Rationale gistic Analysis," Journal of Financial and Quantitative BehindInterfirm Tender Offers: Information or Synergy?," Analysis (December 1985), pp. 461-478. Journal of Financial Economics (April 1983), pp. 183- 17. C. Welles, "Inside the ArbitrageGame," InstitutionalIn- 206. vestor (August 1981), pp. 41-58.