7/28/2019 Rest at 1994

1/10

http://www.jstor.org

Measuring the Effect of Restructuring on Corporate Performance: The Case of Management

Buyouts

Author(s): Scott B. Smart and Joel Waldfogel

Source: The Review of Economics and Statistics, Vol. 76, No. 3 (Aug., 1994), pp. 503-511

Published by: The MIT Press

Stable URL: http://www.jstor.org/stable/2109975

Accessed: 31/08/2008 21:55

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=mitpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

http://www.jstor.org/stable/2109975?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=mitpresshttp://www.jstor.org/action/showPublisher?publisherCode=mitpresshttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2109975?origin=JSTOR-pdf

7/28/2019 Rest at 1994

2/10

MEASURING THE EFFECT OF RESTRUCTURINGON CORPORATE PERFORMANCE: THE CASEOF MANAGEMENT BUYOUTSScottB. SmartandJoelWaldfogel*

Abstract-Recent researchhas attempted o documentthatthe financialgains associatedwithtakeovers,LBOs and othertypesof restructuringre attributableo subsequent mprove-ments in operatingperformance. n this paper we develop amoregeneral rameworkor measuring he effect of corporaterestructuring n performance nd apply the framework o asample of firms taken private by their management.Wedemonstrate hat the estimationapproaches mployed n theliterature mbody estrictions n thegeneral ramework hichthe data can reject. However,our best estimatesprovideevidence hatMBOs mprove orporateperformance,ndthemagnitudesof these improvementsare similar to existingestimates.I. IntroductionA GGROWINGodyof research ndicates hatcorporate restructuring enerates value forstockholders, and recent empirical evidencepointsto improvementsn operatingperformanceas a primary ource of these gains.Kaplan 1989)and Lichtenberg and Siegel (1989) study firmstaken private in managementbuyouts (MBOs)and find that both financial(sales, income, etc.)and real (factor productivity)performancemea-

sures improve after the buyout. Healy, Palepuand Ruback(1990) also discoversigns of assetproductivitymprovementsn a sampleof mergedfirms.The authorsof these studies concludethatMBOs, mergers and other types of corporaterestructuringgenerate operating efficiencies byaltering managerial incentives, reducing agencycosts and improving actor productivity hroughother means. In contrast, a number of otherstudies reportresultsat odds with the view thatcorporate restructuringgenerates long-run im-provementsin value. For example, Ravenscraftand Scherer (1987a and 1987b) detect no evi-dence of improvementsn post-takeoveroperat-ing performance n a sample of merged firms.

Mueller (1985) reaches similarconclusions n hisstudyof marketshare data.Thoughtheir conclusionsdiverge, hese studiesshare the commongoal of measuring he effect ofcorporate restructuringon the performance offirms.Answering his question is complicatedbe-cause the relevant benchmark against whichpost-reorganization erformance hould be mea-sured is not preorganizationperformance.Thepreferredbenchmark ncorporates hanges n ex-pected performance for the restructuring irmsprior to their decision to reorganize as well asunanticipated hocksto performance nrelated othe restructuring ecision.It is difficult o find acontrol measure that convincinglyanswers thequestion,"whatwould have happenedto perfor-manceat the restructuringirms n the absence ofrestructuring?"In most cases, researchersattemptingto ad-dress this issue have measured he expost perfor-mance of restructuringirms relative to the re-sults achievedby companiesoperating n the sameindustry.1The effectof restructurings thus mea-sured as a "difference n differences,"or as thechange in performanceat the restructuring irm,minusthe simultaneous hange n performance tthe controlfirm(s).The chief shortcoming f thisapproach ies in its assumption hat, if not for therestructuring, he reorganizing irm would haveexperiencedthe same change in performanceasits competitors. f individual irmshave their owndynamicperformancepatterns, t maybe unrea-sonable to assume that except when organiza-tional structurechanges, the performancemea-sures of firms n the sameindustrymovetogether.An importantreasonto doubtthat the changein performanceat non-restructuring irms pro-vides an appropriatecontrol measure is that re-Received for publicationSeptember19, 1992. Revision ac-cepted for publicationOctober18, 1993.* IndianaUniversity; ndYale Universityand NationalBu-reau of EconomicResearch,respectively.We would like to thank Steve Kaplanfor providingpost-buyout data on MBO firms. We are grateful to MichaelJensen, John Shoven,ChrisJones and two anonymous efer-ees for helpfulcomments.Anyerrorsare our responsibility.

1In addition o matching irmsaccording o industry, omestudies have attempted to match firms according to size,leverage and other financial characteristics.In contrast,Jarrell (1991) employs an expectations-basedmethodologysimilar o the approachwe propose.Copyright ? 1994 [ 503 ]

7/28/2019 Rest at 1994

3/10

504 THE REVIEW OF ECONOMICS AND STATISTICSstructuring irms are a non-random sample offirms.The reasons for restructuringmay depend,at least in part, on past and (expected) futureperformance.For example, suppose that firms'performance patterns are positively autocorre-lated and that restructurings ccur when perfor-mance is low. Suppose, further, hat at any pointin time, the performance evels of different irmsin an industryare at different levels relative totheir own historicalaverages.Then, independentof the restructuring,we expect performance m-provementat the restructuring irm. Equally m-portant,we do not necessarily expect the sameimprovement at the "control" firms. Conse-quently, he difference n differencesmay providea misleadingmeasure of the effect of restructur-ing on performance.An accurate measure of the effect of restruc-turingon firmperformances importantbecauseit sheds light on organizational fficiency. n addi-tion, the influenceof corporatecontrol events onprivate sector productivity s directly inked witha numberof public policyconcerns.For example,corporate restructuringaffects the tax revenuecollected from corporations.When companiesprivatize, hey typicallyreplace equitywith debt.While the cash flow previously paid to equityholdersfaced corporate ncome taxation,the in-terest paymentspaid to post-buyoutdebtholdersescape corporate-level axation.Jensen, Kaplan,andStiglin 1989) estimatethat the U.S. Treasuryhas actually gained from management buyouts(MBOs)because the tax revenuesgenerated romimproved corporatecash flows have been largeenough to offset corporate tax losses associa-ted with the increased usage of debt.2 Bulow,Summers,and Summers(1990) show that thesegains to the Treasurydisappearwithout the im-provements n operatingperformance.Thus, it isnecessary o determinewhetherincreasedcorpo-rate cash flows are attributable o restructuringrto other factors.Thispaper presentsa framework or estimatingthe effect of restructuring n performance.Thisframeworkembraces as special cases the differ-

ence in differencesmeasure, as well as others.Each specialcase entails a set of testable restric-tions on the general framework.Using data onperformanceat firms undertakingmanagementbuyouts MBOs),we present tests that distinguishbetweenvariousestimators orthe effect of MBOson performance,as well as our best estimatesofthe MBOeffect.The paper is organized as follows. Section IIdescribeshowthe circumstancesurrounding or-poraterestructuring ffect the measurement f itseffect on firmperformance.Section III developsa general empirical framework for measuringthe effect of reorganization n firmperformanceand discusses,thespecial cases of this frameworkthat have been employed n existingstudies. Sec-tion IV describesdata and empiricalestimationof the componentsof the framework.Section Vpresents results.A conclusion ollows.

II. The ReorganizationDecisionThe numberandsize of corporations ndertak-ing significant estructuring ctivities n the formof mergers, share repurchases, and leveragedbuyouts grew duringthe last decade, motivatingeconomists o studythe effectsof corporatereor-ganization on various aspects of firm perfor-

mance. Jensen (1986) and others advance thetheory that some types of restructuringcreatevalue by alteringthe incentivesof managersandownersin a way that enhancesefficiencyand byreducing agency costs. Shleifer and Summers(1988) suggest that corporate takeovers merelytransfer value from employees and other stake-holdersof firms to shareholdersby breaking m-plicit contracts.Similarly,Lowenstein(1985)dis-cusses the possibility that managerswho taketheircompaniesprivate n leveragedbuyoutshaveprivate information about their firm's futureprospects.This informationallows managerstobuy the companyfrom public shareholdersat apricebelowthatwhichan informed nvestorwouldpay. Thoughthis hypothesis mpliesthat financialperformance improves after the buyout, thesource of the value increaseis not relatedto thebuyout tself.Whatever he motivation or corporatereorga-nization,theory providesreasonsto suspect thatthese firmsaresystematically ifferent rom othercompanies,and empiricalevidence supportsthetheory.Neither the types of firmsthat choose to

2 There are other tax effects of these transactionsncludingextra capital gains taxes paid by selling shareholders, n-creasedrevenues rom more efficientuse of capital, axesondebtor'sncomesand asset sales and so on. Some of these taxeffects (e.g., the added capitalgains taxes) are not contingentupon future operatingperformance.However, he net effecton the Treasury's evenues s negativewithoutthe operatingperformancemprovements.

7/28/2019 Rest at 1994

4/10

RESTRUCTURING AND CORPORATE PERFORMANCE 505go private,nor the timing of the decision is ran-dom. Palepu (1986), Smart 1992),and others findthat firms involvedin corporate control transac-tions have characteristics hat distinguish themfrom other firms.3 If these characteristicsarecorrelated with future returns, then one mustaccount for these factors in order to obtain anunconditionalestimate of the effect of the re-structuringdecision on returns.Firms that are experiencingunusually ow cur-rent performancemay be more likelythan otherfirms to restructure. If this is the case, thensimply calculating pre- and post-buyoutperfor-mance changes may overstate the effect of thereorganization.n addition,part of the changeinperformanceat the restructuring irms is alsoexperiencedat other firms. For example, part ofthe improvement t a firmthat restructuresdur-ing an industry lump is experiencedby all firmsin the industryand is not attributable to thereorganization.4Thus, there are two adjustments hat must bemade to raw changes in performance o isolatethe effect of restructuring.Both the firm's ex-pected change in performanceand contempora-neous shocks felt at the firm and elsewheremustbe removed. Adjusting the change in perfor-manceat a restructuringirm with the changeinperformanceat non-restructuringirm will notyield a correct estimate unless both the controland restructuring irm have the same expectedperformance mprovement.The followingsectionlaysout a frameworkor measuring he effect ofrestructuring n performancehat is valid even ifcontrol and restructuring irms have divergentperformance xpectations.

III. An Empirical FrameworkThis section introducesa series of approachesto estimatingthe effect of restructuring n per-formance, includingmeasures which have beenadvancedn the literature.Eachof the estimation

approaches examined is a special case of ourtheoretically preferable general approach, andeach embodies testable restrictionson the gen-eral framework.The simplest possiblemeasureof the effect ofrestructuring n firm performance s simply theraw change in performanceafter the buyoutoc-curs:XtRk -XtR' (1)

whereX is some measureof firmperformance,Rdenotes a restructuringirm,and subscripts andt + k refer to pre- and post-restructuringimeperiods. We will refer to the raw changes asestimator 1). The problemwith this estimator sthat it attributes the entire change in perfor-mance to the reorganizationeven though someportion of the change at the firm would haveoccurredin the absence of any reorganization.For example,observablemeasuresof firmperfor-mance maybe autocorrelated. f a firm'sperfor-mance pattern s cyclicalandrestructuringccurswhen the performancemeasureis below its his-toricalmean, then some improvements expectedand is not properlyattributable o the reorganiza-tion.5 We can test whetherperformance hangesare expectedat restructuringirms.HYPOTHESIS A: E(XtR kIt) = XtRPrior to the restructuring, performance is ex-pected to remain at current levels.Rejectionof the hypothesis hat no changesareexpected ndicates hat the rawchangesestimatoris an inadequatemeasureof the effect of restruc-turing because significantperformance changesare expectedabsentrestructuring.A natural way to deal with the problem ofanticipatedperformancechangesis to adjusttheactual change in performance hat occurredfor(an estimateof) the expectedchange that wouldhave occurredwithout the buyout. Suppose wehave a measure of expectedperformanceat therestructuringfirm in period t + k, E(Xt/kIt).(Note that this expectationis taken at time t,See alsoShivdasani 1993)and Morck,Shleiferand Vishny(1989).4Consider the analogyof an auto workerwho decides toreturn to school to get an M.B.A. If the worker choosesgraduate chool in part becauseof temporarilyow earnings,then measuringhe effect of a graduatedegree on income bytaking he differencebetweenpre- and post-M.B.A. arningswill yield an upwardbiasedestimate.Besides this individual-specific"lowearnings hock,"wages of all autoworkersmayhave experiencedeither a positive or negative shock duringthe trainingperiod.

5Survivorship bias can generate the same type of effect.Consider he populationof firmsexperiencing elow averageperformance t time t. Some of these firmswill survive n oneformor another,but otherswill cease to exist.Becauseanysample of firmswith continuouslyobservableperformancedata over some time intervalwill exclude those that gobankrupt,we would expectto observeperformancemprove-mentsamong hose firms n the sample nitiallydoing poorly.

7/28/2019 Rest at 1994

5/10

506 THE REVIEW OF ECONOMICS AND STATISTICSprior to the restructuring ecision, and it presup-poses that no organizationalchange will takeplace). Then the effect of reorganization s onlythe part of the change in performance hat wasnot expected prior to the buyout, or the restruc-turing surprise:

Xtk - E(XtR+klt). (2)Interpreting his as a measure of the effect ofrestructuring elaxesthe assumptionof the firstapproachthat the expected change in perfor-mance in the absence of restructurings zero.While the restructuring urpriseoffers an im-provement over the raw change (1), it suffersfrom the potential shortcoming hat it attributesto the buyout any unforecastable change ex-periencedat the restructuringirm even thoughnonrestructuringirms may experience the sameperformanceshock. That is, the surprisein ex-pression 2)willtypically ncludebothfirm-specific'andmore general shocks.6 If the restructuringfirm's industry does surprisinglywell followingthe firm'sreorganization, hen it would be incor-rect to attributeallof expression 2) to restructur-ing. We can test whether industry hocks under-mine the interpretation f estimator 2).HYPOTHESISB: E(Xt+ k| t) = Xt+k (for controls)Thereare noperformance shocks at controlfirms.

Rejection of hypothesisB indicates that esti-mator (2), while better than (1), still gives amisleadingestimate of the effect of restructuring.Onlythe componentof the restructuring urprisenot experiencedby similarnonrestructuringirmsis properlyattributed o the reorganization.The possibilityof a performance hockexperi-enced by both restructuringand control firmssuggests hat the effect of restructuringhould bemeasured as the performancesurprise experi-encedat restructuringirms, ess the performancesurpriseat comparablenon-restructuringirms.7Our most general measure of the effect of arestructuring s then the restructuring urpriseminus the "controlsurprise,"or the differencensurprises:[Xt+k - E(Xt+klt)] - [Xt+k - E(Xt+klt)]. (3)

The difference in surprisesmeasure has thefollowing intuitive interpretation.Suppose thatwe have accountedfor every statistically ontrol-lable aspect of the change in performanceatrestructuring and non-restructuring irms pre-dictable prior to the reorganization, xcept therestructuringtself. What we have left is a setof change-in-performanceurprises for the re-structuring and non-restructuringfirms. Thedifference in surprises estimator is the averagedifferencein performancesurprisesbetween re-structuringand nonrestructuringirms and mayproperly be interpreted as an estimate of theeffect of restructuring n firmperformance.Separate from the progressionof estimatorsabove is a measureof the restructuring ffectthathas been used in the literature(Kaplan, 1989;Lichtenbergand Siegel, 1989, and others), thechangein performanceat restructuringirmsmi-nus the change in performancefor a controlsample of firms,

(Xt+k - XtR) - (Xt+k - Xt), (4)or the differencein differences.A shortcoming ofthis estimationapproachis that it neglects thepossibility,discussed above, that some parts ofthe changes in performanceat the restructuringand non-restructuringirmsmaybe forecastable.Our general estimator(3) reduces to expression(4) when the expected changes in performanceare equalat restructuring nd control firms.8Wecan test whether hisconditionobtains,andhencewhether the difference n differencesmeasureisappropriate,by examining he followinghypothe-sis:HYPOTHESIS C: [E(Xt+ kI) -X/R] =[E(Xt+klt) - Xt]Expectedimprovementsat restructuringand con-trolfirms are identical.An examplewill help illustrate the differenceamongthe estimators.Supposethat performanceat XYZ Corp. is expected (in the absence of arestructuring)o improveby 10%over the nexttwo years.9Supposea buyoutoccursand perfor-mance improves15%.The restructuringurprise

6 For example,a petroleum irm that went private n 1990would have experienceda performancemprovementn 1991due to the Gulf Crisis,as did firmsremainingpublic.7Of course,findinga non-MBOcontrol irm s not a trivialmatter,but we leavethat concern or the empirical ection ofthe paper.

8An intuitivespecial case of this arisesif no performanceimprovements re expectedat either type of firm.9In the empirical ectionwe use the change n the ratio ofoperating ncometo sales as our performancemeasure.Thus,if the current income to sales ratio for XYZ is 0.10, anexpected mprovement f 10% mpliesan expected ncometosales ratio of 0.11 in two years.

7/28/2019 Rest at 1994

6/10

RESTRUCTURING AND CORPORATE PERFORMANCE 507TABLE l.-FOUR ESTIMATES OF THE RESTRUCTURING EFFEcr

Raw Change = 15Restructuring Surprise = 15 - 10 = 5Difference in Differences = 15 - (1) = 14Difference in Surprises = (15 - 10) - (1 - 5) = 9

is 5%. Suppose that the performanceat Refer-ence Corp.is a relevantcontrol.PerformanceatReference Corp.was expected to improveby 5%but actuallyincreasedby 1%. The control sur-prise is thus - 4%. What is the restructuringeffect? Table 1 shows restructuring ffect esti-mates arisingfrom each of the four estimators.The raw change is 15%, but the difference insurprisess 9%.In principle,all estimatorsexceptthe difference in surprisesgive misleadingesti-mates of the true effect of restructuring.Theappropriateness f each estimatorin practice isan empirical ssuewhich we examinebelow.IV. Data

Estimation of the effect of restructuringonperformanceusingthe framework boverequiresus to create empiricalanaloguesto each of thecomponentsof expression(3). These are com-prised of two types of information,performanceandexpectedperformance,or twodifferent ypesof firms, restructuring irms and control firms.Expected performancemeasures are obtainedfrombothtime seriesmethods andanalysts' ore-casts.10The performancemeasure (X) used in thisstudy is the ratio of operatingincome to sales.Performance hanges are reported n percentageterms.We use a scaledmeasureto allowcompa-rabilityacrossfirmsand to control(at least par-tially) for post-buyoutdivestitures.In addition,the income to sales ratio may be loosely inter-preted as a measure of the efficiencywith whichfirmsutilize a givenamountof sales. We scalebysales rather than assets because the accountingvalue of assets typicallychangesat the time of abuyout,makingpre- vs. post-buyoutcomparisonsdifficult.Finally,ourmeasureof performancehasbeen used elsewhere in the literature(Kaplan,1989;Jarrell,1991),so we are able to makerough

comparisonsbetween our resultsand the resultsof previousstudies.The data on actualperformance f MBO firms(X/?) come from two sources. Pre-buyoutdataare primarilyfrom COMPUSTAT's Research File,althoughdataunavailable n COMPUSTATwereobtained from 10-k's directly.Post-buyoutdatafor 48 MBOs are obtained from the data setdeveloped by Kaplan (1989).11 These firms weresubject to public disclosure requirementsaftertheir buyouts either because they continued tohave public securities(debt or preferredstock),or because they sought additionalpublic fund-ing.12Up to three yearsof post-buyoutdata, notincluding he fiscalyearof the reorganization, reavailable or each firm. In what follows,the first,second, and thirdyears after the buyoutare re-ferred to as t + 1, t + 2, and t + 3.The data on actual performance at controlfirms Xe) comefromtwobasicsources,COMPU-STAT and ValueLine. We use three differentcontrol samples. From COMPUSTAT,we con-structan annual imeseries of the aggregate atioof operatingincome to sales at the 3-digit SIClevel.These3-digit ndustriesarechosento matchthe 3-digit industriesof the MBO firms.In anyyear,we includeall firmsthat have datafor bothvariables(COMPUSTAT ata items 12 and 13).From ValueLine, we obtaintwo differentcontrolmeasuresof performance.First, for each MBOfirm,we choose up to five individualValueLinefirmsof similarsize in the sameSICindustry.Weaggregate hese firms'operating ncome andsalesfiguresanduse the aggregateratio as our secondcontrol measure of performance.Alternatively,we associateeach MBO firmwith a ValueLineindustrygrouping,andwe use ValueLine'saggre-gate industry-groupingata as a controlmeasureof performance or that firm.13For both control and restructuring irms,wemustalso derivemeasuresof the expectedchangein performanceat the time of the restructuring

10We employbothtypes of forecastsbecause of conflictingevidenceon the qualityof timeseriesand analysts'orecasts.See, for example,Brownet al. (1987) andO'Brien 1986).

See Kaplan 1989) for a detaileddescriptionof his data.12A few firmsdisclosed heirfinancialdatawhenthey weresold to existingpubliccompanies.13 ValueLine industry ategoriescorrespond nlyloosely toSIC codes and group togetherfirmsthat operate in vastlydifferentmarketson dramatically ifferentscales. It is notuncommonfor Value Line to group two firmswhose SICcodes aredissimilar ven at the two-digitevel.We preferona priori groundsusing firm-specificdata to construct ourcontrolmeasuresbecause we can matchcontroland restruc-turing irmsmorecloselyon the basis of firmsize andindus-try.

7/28/2019 Rest at 1994

7/10

508 THE REVIEW OF ECONOMICS AND STATISTICSfirm's buyout. We derive expected changes inperformance or the MBO firms (E(XtRklt) -X/R)in two ways. First, we estimate a dynamicperformanceregressionon the MBO firms' an-nual performance istoryup to the lastpre-buyoutyear usingthe ratioof operating ncometo salesas the dependentvariable,andwe use this regres-sion to generateforecastsfor post-buyoutperfor-mance. This regressionincludes one lagged de-pendent variableand allows each firm a differentintercept. Our second measure of expectedchangein performanceat the restructuringirmsis the ratioof operating ncome to salespredictedby Value Line. Value Line publishes a multi-yearforecastcovering he second through ourthyearsfollowing the buyout (years t + 2 to t + 4). Weobtainthe expectedchange n performance s thelast forecast of performance before the an-nouncementof the buyout. Thus, these expectedchanges are the expectations of performancechangesin the absenceof a buyout.We derive expected changes in performancefor control measures(E(Xf+klt) - XI) in threeways: (1) using autoregressionson SIC 3-digitaggregate COMPUSTAThistory (in a fashionanalogous o the regressionson MBO history), 2)by aggregatingValue Line's performancefore-castsfor the 5 individual irms hatwe matchwitheach MBO firm,and (3) using ValueLine's fore-casts of industryaggregate performance.In allthree cases, the control forecasts are made giveninformationavailablepriorto the buyoutat theirmatchedMBOfirms.

V. EmpiricalResultsIn this section we estimatethe effect of restruc-turing on performanceusing data on MBOs.We test between the measuresdiscussedin sec-tion III, andwe presentourbest estimatesof the

MBO effect.A. Choosing an Estimator

In principle,the most generalmeasureof theeffectof restructurings the best measure,since itrelaxesas many assumptionsas possible.Yet, inchoosing empirically among estimators for theeffect of restructuring,what matters is whetherthe restrictionsembodiedby a particular stima-tor indeed hold. In what follows, we test therestrictions mpliedby choices of various estima-tors over others.In particular,we test hypotheses

that allow the data to distinguishbetween rawchanges and restructuring urprises 1 vs. 2), be-tween restructuring urprisesand differences insurprises (2 vs. 3), and between differences indifferencesand differences n surprises 4 vs. 3).Table 2 presents Wilcoxonstatistics in paren-theses) for these three hypotheses,estimatedoverthe three different control groups'data. Figuresin columnA representthe medianexpected per-formance improvements at restructuring irmsmeasured as a percentagechangein the operat-ing income to sales ratio.14For example, usingCOMPUSTAT atawe estimate that the medianexpected changein the ratioof operating ncometo sales is 4.0% by the firstyearafter the buyout,5.3% by the second year, and 7.0% in the third.ValueLine forecasts show even larger expectedperformance mprovementsor these firms,7.5%by year two and 17.0% by year three, and theyare significantly ifferent romzero.15Becauseweconsistentlyreject hypothesisA, it is inappropri-ate to use raw changes to measure the MBOeffect as they will give an upwardlybiased esti-mate of the true effect.ColumnB containsmedianestimatesof controlfirm surpriseswhich should theoreticallybe sub-tractedfrom restructuring urprisesto yield thetrue MBO effect. Note that accordingto all of

ourfirm-specific ata,control irmsperformworsethan expected,and significantlyo by year three.In other words, both COMPUSTAT nd ValueLine forecasts of the performanceof individualcontrol firms prove to be overly optimistic,al-thoughthe ValueLine data on aggregate ndustryperformancefail to reject the hypothesis thatcontrol surprises are zero. To the extent thatcontrolsurprisesare significantly elowzero, therestructuringurprise stimatesunderestimateheMBO effect because they neglect simultaneousnegative performance urprisesat controlfirms.Column C presents median estimates of thedifferencebetween restructuringirms'expectedperformance mprovementsand that of controlfirms.A positiveestimateindicatesthat expectedimprovements t restructuringirmsexceedthoseat control firms, and therefore that the differ-ences in differencesestimateswill exceedthe true14 These changesare measured elative o the year immedi-ately prior o the buyout.15The bottompanel of columnA is blank becauseto testthis hypothesiswe need to measure expected performance

improvements nlyat restructuringirms,not control irms.

7/28/2019 Rest at 1994

8/10

RESTRUCTURING AND CORPORATE PERFORMANCE 509TABLE 2.-MEASURES OF THE DIFFERENCES BETWEEN ESTIMATORS

AND HYPOTHESIS TESTSaControl N Ab Bc Cd

1. COMPUSTAT data used to forecast expected performancet + 1 35 4.0 -1.5 6.1(2.19)' (-1.03) (1.92)9t + 2 31 5.3 -4.7 6.0(2.19)' (-1.29) (1.96)'t + 3 14 7.0 - 28.7 -6.75.8 - 27.8 -11.7(1.6) (-2.54)' ( -0.41)

2. Value Line data used to forecast expected performance1. Individual firmst + 2 25 7.5 -7.9 1.0(3.43)f ( - 2.52)' (0.96)t + 3 11 17.0 -18.3 4.6(2.22)f ( 1.96)' (1.69)9b. Industry aggregate dataet + 2 24 2.3 1.6

1.9 1.5(0.77) (0.44)t + 3 10 2.3 4.01.9 1.6(0.87) (0.76)a Wilcoxon statistics are in parentheses. Critical values for two-tailed tests are 1.96 and 1.65 for 95% and 90%significance levels, respectively.Ho: E(XR+klt) - = 0, or that the expected performance improvement at restructuring firms is zero.Rejection indicates that estimator 2 provides an improvement over estimator 1.Ho: XI+k - E(XI+klt) = 0, or that control surprises are zero. Rejection indicates that estimator 3 is moreappropriate than estimator 2.d Ho: [E(X/R+klt) - XR- [E(X,+klt) - XI] = 0, or that the control and restructuring performance measureshave the same expected improvements prior to the restructuring. Rejection indicates that estimator 3 provides animprovement over estimator 4.e Note that since there are an even number of firms in this panel and in year t + 3 of panel 1, there is no uniquemedian, so we report both medians. Since hypothesis B does not require industry adjustment, we do not repeat theresults from panel 2.f Significant at 95%.g Significant at 90%.

MBO effect. Our evidence on this hypothesis smixed. COMPUSTAT ata suggestthat there aresignificantdifferences between expected perfor-mance at restructuring nd control firms in theshortrun(the firstandsecond post-buyout ears),but that these differencesdisappearby year three.The Value Line firm-specificdata indicate thatMBO firms have greater expected performanceimprovements, ut they are significantat conven-tional levels only in the thirdyear. The industryaggregatedataalso indicatehigherexpected per-formance or MBO firms,but the differencesarenot significant.The erraticestimates or the latteryears probablyreflect our small samples (fewerthan 15 firmsby the final year). In summary,allbut one of our estimates indicates that MBOfirmshave higherexpected performancemprove-mentsthan controlfirms.Becausethese estimatesare not consistentlysignificant,we conclude thatthere is only weak evidence that the difference ndifferences estimator overstates the true MBO

effect and that the difference n surprisesestima-tor offers an improvementover the difference ndifferences.B. Choosing an Estimate

The tests reported aboveindicate that conven-tional measures of the effect of restructuring nperformance embody restrictionsthat the datasometimes reject. The pattern of test resultsguides our choice of estimates, presentedbelow.Table 3 reports median MBO effect estimatesaswell as Wilcoxon tests of the hypotheses thatthese effects are zero. Our finding of positiveexpected improvementsat restructuringirms-and concomitantrejectionof hypothesisA-sug-gests that the raw changes (1) gives larger esti-mates than the restructuring urprises (2). Re-sults in table 3 confirm his.Withthe exceptionofthe COMPUSTATestimate for the first post-buyoutyear, all of the restructuring urpriseesti-

7/28/2019 Rest at 1994

9/10

510 THE REVIEW OF ECONOMICS AND STATISTICSTABLE 3.-MEDIAN ESTIMATES OF THE MBO EFFECr

USING ALTERNATIVE ESTIMATORSaControl N lb 2C 3d 4e

1. COMPUSTAT data used to forecast expected performancet + 1 35 4.1 5.7 6.8 6.7(1.97)9 (0.95) (1.39) (1.97)9t + 2 31 11.9 7.4 11.5 11.8(2.62)9 (1.87)h (2.00)9 (2.36)9t + 3 14 22.8 18.1 33.6 26.919.4 10.2 28.7 22.7(2.42)9 (1.66)h (2.86)9 (2.17)9

2. ValueLine data used to forecast expected performancea. Individual firmst + 2 25 18.0 7.7 21.4 15.7(2.65)9 (0.98) (1.98)g (2.19)gt + 3 11 23.9 1.4 33.9 33.1(2.31)9 (0.80) (2.13)h (2.84)9b. Industry aggregate dataft + 2 24 13.1 12.0

12.4 7.5(0.71) (1.31)t + 3 10 21.0 28.74.6 23.1(1.38) (1.17)a Wilcoxon statistics are in parentheses. Critical values for two-tailed tests are 1.96 and 1.65 for 95% and90% significance levels, respectively.

bRaw changes: X~t+k -t. All measures are reported as percentage change in the ratio of operating income tosales.c Restructuring surprise: Xt+k - E(XR+k It).d Difference in surprises: [XjR+k E(X!Rklt)] - [Xt+k - E(Xt+klt)].Difference in differences: [XI k -Xt ]-[Xt+k - X].Note that since there are an even number of firms in this panel and in year t + 3 of panel 1, there is no uniquemedian, so we report both medians. Columns 1 and 2 are blank for this panel because these estimators are basedonly on restructuring firm data.g Significant at 95%.h Significant at 90%.

mates are smaller and less significantthan themedianrawchanges.Forexample, he rawchangein the operating income to sales ratio for themedian firm is 11.9%for the sampleof 31 firmswith COMPUSTATdata in the second post-buyout year. Adjusting hese figuresfor time se-ries forecasts of expected performanceyields anincreasein the ratio of just 7.4%. While perfor-mance improvements ersist even afterremovingexpected improvements t restructuringirms,werejectthe hypothesis hat these improvements rezero with less confidence hanfor raw changes.The result of hypothesis B, that control firmsexperienced negative performanceshocks, indi-cates that the difference in surprisesestimatesshould be larger than the restructuring urpriseestimateswhich ignore these shocks.Again table3 is consistentwith our earlierfindings.In everyinstance,median differences n surprisesexceedrestructuring urprises and are more significant.For example,the median restructuring urprisesin the second post-buyout year are 7.4% and7.7% (using COMPUSTAT and Value Line fore-

casts, respectively),while the associatedmediandifference n surprisesare 11.5%and 21.4%.Finally,we compareour theoreticallypreferredset of estimates, he difference n surprises, o thedifference n differencesapproach ommonlyusedin the literature.The estimates differ when ex-pectationsof future performanceat restructuringand control firmsdiverge.The positive and sig-nificant performance mprovementsexpected atMBO firms (documented above) are somewhatoffset by the expectedchanges n performanceatcontrol firms. As a result, the medians of thedifference in differencesand difference in sur-prises are in most cases quitesimilar,eventhoughwe sometimes reject hypothesisC, that the ex-pected changesat restructuring nd control firmsare identical.The readermay find it puzzlingthat althoughin some caseswe reject hypothesisC, the medianestimates of (3) and (4) are virtuallyidentical.The answerrelates to the distributionof differ-ences in expected performancechangesfor bothtypes of firms in this sample. The adjustments

7/28/2019 Rest at 1994

10/10

RESTRUCTURING AND CORPORATE PERFORMANCE 511(which distinguishbetween estimators 3 and (4))are close to zero for firms near the median forestimator (3) and more substantial n the tails.Hence, the mediansare virtuallyunaffectedwhilethe significance evels drop somewhat.In summary,our best estimate of the MBOeffect leads us to conclude that there are largepositive ncreases n the ratio of operating ncometo sales after the buyout. This measure of per-formance improves by 6.8% between the lastpre-buyoutyear and the first post-buyoutyear.Performance mprovementsby the second post-buyout year range from 11.5% to 21.4%. Ourestimatesof performancen the thirdpost-buyoutyear indicate mprovements f up to 20%to 30%.The median ratio of operating ncometo sales inthe last pre-buyoutyear is 10.6%. Therefore a10% mprovement n performance orresponds oa post-buyoutoperating ncome to sales ratio of11.7%.Given a medianpre-buyout ales figureof$570 million, this improvement ranslatesto anincrease n operating ncome of about $6 million.If MBOsimproveperformanceby, say, 30%overlonger horizons,and if these improvementsper-sist, then the presentvalue of the firm's ncreasedoperating income would be $180 million (dis-counted at 10%).

VI. ConclusionRecent developmentshave focused attentionon the difficultquestion of whether corporaterestructuring ffects performance.The principalobstacle to measuring he effect of restructuringon performanceis accountingfor what wouldhavehappenedto performancef the reorganiza-tion had not occurred.Existingstudieshave gen-erallymeasured the effect as the change in per-formanceat the restructuringirm, ess the changein performanceat a controlfirm.We have shownin this paperthat this conven-tional measurementapproachis one of variousspecial cases of a more general measurementframework. Moreover, the restrictions on thegeneralframework equiredto justifythe use ofeach special case are testable. Using data onMBOs, we test and sometimes reject these re-strictions, uggesting hat more flexibleestimatorsof the MBO effect may be preferableto morecommonlyused estimators.

The results of our hypothesistests guide ourselectionof MBO effectestimatorsandestimates.Even though we find some evidence that ourestimatorprovidesan improvement ver the con-ventional approach, our best estimates of theMBO effect lend added supportto the view thatMBOsenhanceoperatingperformance.

REFERENCESBrown, L., et al., "SecurityAnalyst SuperiorityRelative toUnivariateTime-SeriesModels in ForecastingQuar-terly Earnings," ournal of Accounting and Economics9 (Apr. 1987),61-87.Bulow, J., L. Summers, and V. Summers, "DistinguishingDebt from Equity n the JunkBond Era," n J. Shovenand J. Waldfogel(eds.), Debt, Taxes, and CorporateRestructuring (Washington,D.C.: BrookingsInstitu-

tion, 1990).Healy, P., K. Palepu, and R. Ruback,"Does CorporatePer-formanceImproveAfter Mergers?"NBER WorkingPaper #3348, May 1990.Jarrell,S., "Do TakeoversGenerateValue? Evidenceon theCapitalMarket'sAbility o Assess Takeovers,"mimeo,SouthernMethodistUniversity,Feb. 1991.Jensen, M., "Agency Costs of Free Cash Flow, CorporateFinance,and Takeovers,"American Economic Review76 (May 1986),323-329.Jensen, M., S. Kaplan,and L. Stiglin, "Effects of LBOs onTax Revenues of the U.S. Treasury,"Tax Notes (Feb.6, 1989),727-733.Kaplan,S., "TheEffectsof ManagementBuyoutson Operat-ing Performance nd Value,"Journal of Financial Eco-nomics 24 (1989).Lichtenberg,F., and D. Siegel, "The Effects of LeveragedBuyoutson Productivity nd Related Aspects of FirmBehavior,"NBER WorkingPaperNo. 3022,June 1989.Lowenstein,L., "ManagementBuyouts,"Columbia Law Re-view 85 (1985), 730-784.Morck, R., A. Shleiferand R. Vishny,"AlternativeMecha-nisms or CorporateControl,"American Economic Re-view 79 (4) (Sept. 1989).Mueller, D., "Mergersand MarketShare,"this REVIEW 67(May 1985).O'Brien,P., "Analyst'sForecastsas EarningsExpectations,"Journalof Accounting and Economics 10(1988),53-83.Palepu, K., "PredictingTakeoverTargets:A Methodologicaland EmpiricalAnalysis,"Journal of Accounting andEconomics 8 (1986),3-37.Ravenscraft,D., and F. Scherer, "Life after Takeovers,"Journal of Industrial Economics (Dec. 1987a)., Mergers, Selloffs and Economic Efficiency (Washing-ton, D.C.: Brookings nstitution,1987b).Shivdasani,Anil, "BoardComposition,OwnershipStructure,and Hostile Takeovers,"Journal of Accounting andEconomics 16(1993),167-198.Shleifer, A., and L. Summers,"Breachof Trust in HostileTakeovers,"n A. Auerbach ed.),CorporateTakeovers:Causes and Consequences (Chicago: University ofChicagoPress, 1988).Smart,S., "LargeShareholders nd Contests for CorporateControl,"mimeo,IndianaUniversity, 992.