FIRMOWNERSHIPPREFERENCESANDHOSTGOVERNMENT … · 2019-03-08 · (SIC32),andprimarymetals(SIC33).Subsidiary's productfrom HarvardMNEdatabase. (mean=0.25; sd=0.43) 5. Marketing/RevSubInd

FIRMOWNERSHIPPREFERENCESAND HOST GOVERNMENTRESTRICTIONS:AN INTEGRATEDAPPROACH

BenjaminGomes-Casseres*Harvard University

Abstract.lWo approachesmay explainhow multinationalenter-prises (MNEs) select ownershipstructuresfor subsidiaries.Thefirst arguesthat MNEs preferstructuresthat minimizethe trans-action costs of doing businessabroad. The second arguesthatownershipstructuresare determinedby negotiationswith hostgovernments,whose outcomes depend on the bargainingpowerof the firm. This paper presentsa frameworkintegratingthesetwo approachesand uses statisticalmethods to separatetheireffects empirically.The statisticalanalysissupportsan importanthypothesisof thebargainingschool-that attractivedomesticmarketsincreasetherelative power of host governments.But it finds no supportfor other hypotheses of this school, such as those predictingthat firms in marketing- and R&D-intensiveindustries havemore bargainingpower than others. These latter factors wereapparentlymore importantin determiningfirm ownershippref-erences. Futhermore, the paper measures when governmentownershiprestrictionsdeter firm entry, concluding that rela-tively large firms, and those with high intra-systemsales aredeterredmorethan others.

In recent years, there has been a renewed interest in the determinants ofownership structures of foreign subsidiaries. The basic question of whenand why multinational enterprises (MNEs) form joint ventures abroad isbeing addressed with a refined set of tools and with new theories. This

*BenjaminGomes-Casseresis AssistantProfessorof BusinessAdministrationattheGraduateSchoolof BusinessAdministration,HarvardUniversity.HeholdstheB.A.degreefromBrandeisUniversity,theM.P.A.degreefromPrincetonUniversity,andthe D.B.A.degreefromHarvardUniversity.His relatedarticleson joint ventureshaveappearedintheJournalofEconomicBehaviorandOrganizationandtheSloanManagementReview.Hiscurrentresearchandconsultingfocusonthemanagementof internationaltransfersof technologyand theireffects on competitiveness.

I receivedvaluablecommentson this researchfromLouisT. Wells,Jr.,DonaldR. Lessard,RobertSchlaifer,DennisJ. Encarnation,StephenJ. Kobrinandthreeanonymousreferees.The researchwasfundedby the HarvardBusinessSchool'sDivisionof Research.Harvard'sMultinationalEnterpriseProjectandtheStrategicPlanningInstituteinCambridge,Massachusetts,gavemeaccessto theirdata.The viewspresentedherearemine.

Received:August1988;Revised:December1988& March& May 1989;Accepted:May 1989.

1

Palgrave Macmillan Journalsis collaborating with JSTOR to digitize, preserve, and extend access to

Journal of International Business Studieswww.jstor.org

®

2 JOURNAL OF INTERNATIONALBUSINESS STUDIES, FIRST QUARTER1990

renewedinterestis partlydue to two developments:(1)a newconceptionof the roleof ownershipin internationalbusiness;and (2) a betterunder-standingof the processof negotiationbetweenMNEs and host countrygovernments.Thesetwo new developments,however,havenot been inte-gratedinto one framework.This articleattemptsto do that.Inthelastdecade,numerousauthorshavesuccessfullyusedtransactioncostideas to analyze the role of ownershipin internationalbusiness (forexample,Buckley&Casson[1976];Hennart[1982];anda recentreviewinTeece[1986]).This approachhas also been appliedspecificallyto MNEs'choicebetweenwholeandjoint ownershipin foreignsubsidiaries[Gomes-Casseres 1989b;Anderson & Gatignon 1986; Hennart 1988]. Almostwithoutexception,however,this streamof researchhasignoredtheeffectsof hostgovernmentownershiprestrictionson MNEchoices.None of theseauthorswould deny,of course,that such restrictionscan makean MNEforma joint ventureevenwheretransactioncost analysiswouldpredictawholly-ownedsubsidiary.But the tendencyin this literaturehas been tofocus on only one of these two effects at a time.Anotherschoolof thought,usingthebargainingpowerapproach,hasmadethe oppositeomission.In an effortto explainthe outcomeof negotiationsbetweenMNEsandhost countrygovernments,virtuallyall the authorsinthisschoolhavedownplayedthefactors,thatinfluenceMNEownershippref-erences (for example, Fagre & Wells [1982]; Grieco [1982]; Lecraw[1984];and a recentreviewin Kobrin1987).Again, most of them haveacknowledgedthatfirmsmayprefersomethingotherthanwholeownership,and that somehowthis had to be takeninto account.But they,too, havetendedto focus on only one side of the equation.Thispatternof academicspecializationwouldnothavebeensurprisingwereit not for the complementarityof the two approaches.Thesetwo theoriesarenot competingexplanationsof the samephenomenon,butaddresstwodistinctquestions.Toputit simply,thetransactioncost modelanswersthequestion: What ownershipstructuredoes the firm want? The secondapproach,thebargainingpowermodel,answersthequestion:Whatowner-ship structurecan thefirm get?Thesetwo questionsarenot only distinctin substance,but usuallyalso intime. At the risk of oversimplifyingthe decisionmakingprocess,one canvisualizean MNE firstdecidingwhatit wants,and then seekingthe hostgovernment'sagreement.Conceptualizingthe decision as a sequentialprocessis the first step in integratingthe two approaches.Thesecondstepin integratingthetwomodelsliesin separatingtheirpredic-tions empirically.Kobrinrecentlyrecognizedthe complementarityof thetwo approachesin his studyof bargainingpower,but lamentedthat "it isdifficultto determinewhetherdifferencesin observedownershipstemfromdifferencesin preferencesor bargainingpower"[1987,p. 624]. The diffi-culty, of course,is that firm preferencesare not directlyobservable.Butwithoutseparatingthe effects, one cannothope to test eithertheory.

FIRM PREFERENCES& HOST GOVERNMENTRESTRICTIONS 3

Thispaperhas fourpurposes,First,it discussesa conceptualsynthesisofthe transactioncost and bargainingmodels.Second,it presentsa methodforseparatingtheireffectsin empiricaltests.Third,it appliesthesemethodsand discussesresultsof newtestsof the bargainingpowermodel. Fourth,it extendsexistingtestsof thebargainingapproachbymeasuringthecondi-tions underwhichhost governmentpolicies deterentryby the MNE. Adetaileddiscussionof my tests of the transactioncost model appearsinGomes-Casseres[1989b].

OWNERSHIPTRADE-OFFSAND NEGUIHATION

OwnershipPreferences

Evenwhenhost governmentsdo not imposerestrictionson foreignowner-ship,anMNEchoosingthe ownershipstructurefor a newsubsidiarymustmakesome keytrade-offs.Takethe exampleof an MNE venturinginto anewcount-rywherethereareestablishedcompetitors.A joint venturemaybe an effectivewayto acquirelocal expertise,but theremaybe significantcoststo sharedmanagement.InStopford&Wells[1972],theMNE'schoicewoulddependon a trade-offbetweenthe firm's"needfor resources,"andits "needforcontrol."It willprefera joint venturewhenthe firstis largerthan the second, and a wholly-ownedsubsidiaryotherwise.Inthetransactioncostmodel,theproblemis formulatedslightlydifferently.Herethetrade-offis betweenthecostsof usingmarketor internalchannelsfortransferringorganizationalcapabilities.Intheexampleabove,theMNEcouldget the neededlocal expertiseeitherby hiringthe servicesof a localfirm(e.g.,to supplymarketresearchandrepresenttheMNEin negotiationswithgovernment),or by forminga joint venturewitha local firm. In thefirstcase it wouldbe acquiringan organizationalcapabilityby using themarket;in the secondit wouldbe usingan internalchannel,becausethecapabilitieswouldbe transferredfroma partywithan ownershipshareinthe venture.'This distinction-between internaland marketchannelsfortransferringfirmcapabilities-is alsocriticalto thetransactioncost theoryof the multinationalenterprise[Hennart1982].Inthetransactioncostapproach,too, thebenefitsof usingownershipchan-nelswouldbe reducedbythecost of sharingequity.Foremostamongtheseare"shirking"by thejoint venturepartners-each of whichhas less thana fullstakeintheventure-and conflictsof interestbetweenthem.Notealsothatthismodelreliesnot only on transactioncosts, butalso on argumentsaboutfirmcapabilitiesandstrategies.ThelatterdeterminewhentheMNEneeds contributionsfrom local firms; transactioncosts then determinewhetherownershipchannelswill be used for these transfers.This modelis discussedfurtherin Gomes-Casseres[1989b];similarargumentsappearin Hennart[1988].Thisprocessof weighingcostsandbenefitsof variousownershipstructuresresultsin theMNE'spreference,i.e.,whatit wants. Buttwofirmsthatboth

4 JOURNAL OF INTERNATIONALBUSINESS STUDIES, FIRSTQUARTER1990

preferwholly-ownedventuresmaydo so to differentdegrees,i.e., onemaywantit moreintenselythat the other.2It is thereforemoreusefulto thinkof thefirm'spreferencesasbeinga rankingamongownershipoptionswithvaryingdistancesbetweenthem.TheMNE'scapabilitiesandthetransactioncostsof differentownershipstructurescanthenbethoughtof asdeterminingthese distances.

OwnershipConcessions

MNEsfacinghost governmentownershiprestrictions3can be expectedtogo througha similarprocessto derivetheirpreferencerankings.Inaddition,however,the relativepowerof firm and governmenthelps determineinthese cases what ownershipstructurethe firm can get. It is criticaltorealizethat the firmswill makeconcessionson ownershipbasedon theirpreferencerankings.In otherwords,a firmrankingwholeownershipsignif-icantlyhigherthanajointventurewillbe lesslikelyto givein on theowner-shipissuethan a firmthat, whilestillpreferringwhole ownership,sees ajoint ventureas a close second choice.As a result,the ownershipstructurethat the MNE actuallyendsup withis a functionof: (1)the intensitywithwhichit preferswholeownership(ifat all);and(2) itsbargainingpowerrelativeto thegovernment's.Excludingthe calculationof firmpreferencesfromthissynthesisis just asmisleadingas ignoringthe bargainingpowerof the parties.Thisframeworkrecognizesthatownershipis not an all-importantissuetofirms,or to governmentsfor that matter.Both partiesmightbe willingtoacceptsomethinglessthantheirtopownershippreferenceinreturnforgainson otherissues.Thus,the governmentmightdropits insistenceon a jointventurein returnforincreasesin theMNE'scontributionto othernationalgoals.4Or the firmmightsettle for its secondchoice in returnfor accessto a lucrativemarket.Inthiscontext,therefore,theslipperyconceptof "bargainingpower"doesnot indicatehow the total economic benefits of an investmentare dis-tributed.Rather,it referssimplyto the abilityof one partyto skewtheoutcomeof negotiationsinthedirectionof theownershipstructureit prefers[Lax&Sebenius1986].Thatabilitydependsnot only on whatthe govern-menthasto offerthe MNE, but also on howbadlytheMNE wantswholeownership.

DeadlockedNegotiations

The synthesis proposed above allows analysis of a phenomenon notadequatelydealtwithin eitherthetransactioncost orthebargainingpowermodel.Authorsusingthelatterframeworkhavepointedoutthatsometimeshost governmentownershiprestrictionsseemedto deterentryby MNEs[Fagre&Wells1982].Butwhendoes thishappen:whenthe firmhas rela-tivelymorepowerthan the government,or vice versa?Theseauthorsdo


not answerthis question,partlybecausethe conceptof bargainingpoweritself is misleadingin this context.In my framework,entrydeterrenceoccurswhen:(1)the MNE has a pref-erenceschedulewherethejoint ventureformranksbelownot investingatall;and(2) thegovernment'srankingplaceswholeownershipbelowtheno-investmentalternative.It is not a questionof bargainingpowerat all, butoneof comparingthecost-benefitcalculationsof thetwoparties.Eachpartyconsidersnot onlythe variousownershipoptions,but also the alternativeof foregoinginvestment.Thatalternativesets the "reservationvalue"foreachparty;an agreementwillnot be reachedunlessits netbenefitsexceedthis value [Lax& Sebenius1986].One party'sno-agreementalternativemay, in turn, affect its bargainingpowerin ownershipnegotiations.WhenanMNEcango elsewherewithitsinvestment,it is less likelyto give in to governmentdemandsthan other-wise.Similarly,if a restrictivegovernmentcan find otherfirmsthatwouldaccepta forcedjoint venture,it maynot give in to one MNE's insistenceon wholeownership.Hostgovernmentownershiprestrictionscanthushaveboth ownershipeffectsandentryeffects.Becausethesetwo sets of effectsareinterrelated,testsof the bargainingmodel shouldattemptto estimateboth, as is done below.

SEPARATINGEFFECTSON PREFERENCESAND BARGAINING

Thestatisticalmethodsusedhereattemptto estimatehowtheownershipandentryeffects of restrictivepoliciesvarywith characteristicsof the MNE,its subsidiary,the industry,and the host country.

EstimatingOwnershipEffects

The main problemin estimatingthe separateownershipeffects of trans-actionandMNE-governmentbargainingmodelsis thattheysharea numberof explanatoryfactors.Forexample,both modelspredictthat the R&D-intensityof a subsidiary'sbusinessshouldaffectownershipchoice,but fordifferentreasons.ThefirstmodelclaimsthatR&Dintensityleadsto wholeownershipbecauseMNEspreferinternalchannelsovercontractswhentrans-ferringtechnologicalcapabilities.Thesecondmodelpredictsthesamefinaloutcome,butcreditstheMNE'sincreasedbargainingpowerwiththeresult.Recognizingthis problem, previousresearchershave tested bargaininghypothesesby comparingownershippatternsin countrieswhose govern-mentsrestrictedforeignownershipwithpatternselsewhere[FagreandWells1982;Lecraw1984;Kobrin1987].But this introducesa differentproblem.The ownershippatternsin a restrictivehost maybe due to otherfeaturesof the country'senvironment,or to systematicdifferencesin the typesoffirms investing in each group of countries. Although most of theseresearchersrecognizedthese possibilities,theywerenot able to deal withthem in satisfactoryways.


The method used here resolvesboth these problems.5It, too, makes adistinctionbetweengroupsof "restrictive"and "open" countries,but itcontrolsfordifferencesin firmandcountrycharacteristicsbetweenthetwogroups. The method is analogous to that used in testing for structuralchangein time-seriesmodels[Johnston1984].Assumingonly one explan-atoryvariable,X, the regressionmodel is:

Y=Ao + (A1-Ao)D+BoX+ (B1-Bo)DX (1)

whereY = observedownershiplevel of subsidiary,D = dummyvariableequal to 1 if subsidiaryis in a restrictive

country,and 0 otherwise,X = independentvariable that is hypothesizedto effect both

bargainingpowerand firm preferences.In this model, the termsAo and Bo represent,respectively,the interceptandtheeffectof X on ownershiplevelin non-restrictivecountries,becauseD= 0 for theseobservations.Thesecoefficientscan thusbe interpretedasreflectinga model of firm preferencesalone, becauseMNE-governmentbargainingplays no role in these countries.The termsA1 and B1 thenrepresent,respectively,theinterceptandtheeffectof X on ownershiplevelin restrictivecountries,becauseD= 1 for theseobservations.Thesecoeffi-cients,therefore,canbe interpretedas reflectingboth firmpreferencesandbargainingpower,becauseboth processesareat workin thesecountries.OnecanthentestwhetherX affectseitherfirmpreferences,or bargainingpower,or both, by consideringwhetherthe coefficientsin the regressionmodelarestatisticallysignificantlydifferentfromzero.If Bo is zero,thenX hasno effectin opencountries;thisimpliesthatX hasno effecton firmpreferences.If (B1-BO)is zero,thenX has the sameeffect in open coun-tries as it does in restrictivecountries(i.e., B1=BO).Becausethe owner-ship levelsin restrictivecountriesappearto be affectedby X in the samewaywhetheror not bargainingtakesplace,X canbe saidto haveno effecton bargainingpower.If X affects both firm preferencesand bargainingpower,the coefficients on X as well as on DX should be significantlydifferentfrom zero.In the data set used here, Y was definedas a binaryvariable,whichwasequal to 1 if the subsidiarywas a joint venture.This does not affect theinterpretationof the model,but it requiredthe use of a binaryregressiontechnique.Inthebinomiallogitmodelused,theeffectsof theindependentvariablesare assumedto be linearin the logarithmof the odds that thedependentvariableis equalto one.As in ordinaryleast-squareregressions,theone-tailedsignificancelevelsof theestimatedeffectsrepresenttheproba-bilitiesthatthetrueeffectshavesignsoppositethoseof theestimates;thesearereportedbelow.Similarly,thestandardizedbetacoefficientsgivetheesti-matedamount,in standarddeviations,thatthe dependentvariablechanges


withonestandarddeviationchangein theindependentvariable;these,too,are reportedbelow.6

EstimatingEntryEffects

The degreeto which restrictivepolicies deterentrycan also be expectedto dependpartlyon characteristicsof theMNEor of theproposedventure.Forexample,MNEs seem to preferwhole ownershipof subsidiariesthattrade extensivelywith other membersof their global systems, becauseconflicts of interestare likelyto arisebetweenthe partnersin these situ-ations.7Consequently,an MNE can be expectednot to placesuchsubsid-iaries in restrictivecountries,unless, of course, it can gain an exceptionto the ownershiprulesthroughthe bargainingprocess.Themethodsusedhereattemptto measuretheimpactof a numberof firm,industry,andsubsidiaryvariableson the entryeffect of ownershiprestric-tions.8Onesimplewayto do thiswouldbe to comparethe meansof thesevariablesin openand restrictivecountries.The problemwiththis methodis that this differencebetweenmeans may be due to other factors. Forexample,thegroupsof openandrestrictivecountriesmayalso differsyste-maticallyin economicterms,andthismightbe the reasonbehindthevari-ancein subsidiarycharacteristics.Thefollowingregressionmodelcontrolsfor such effects:

X=A+B1D+B2Z+B3V (2)

whereX = firmorsubsidiaryvariablehypothesizedto affectthedegreeof

entrydeterrence,D =dummy variableequal to 1 if subsidiaryis in a restrictive

country,and 0 otherwise,Z = otherhost countrycharacteristic,V = other firm or subsidiarycharacteristic.

Inthismodel,thecoefficienton thedummyvariableindicatinghostcountrypolicy (B1)measuresthe differencebetweenthe level of X in open andrestrictivecountries,whilecontrollingfortheeffectsof othercountry,firm,and subsidiaryvariableson X. If the coefficientis positive,then entrybysubsidiarieswith low levelsof X aredeterredmorethan entryby subsid-iarieswithhigherlevelsof X. The reverseis truefor negativecoefficients.Therecan, of course,be severalZ and Vvariables,but theircoefficientsare of no interesthere.

TheSampleand Variables

Sources of Data. The datausedherecamefromseveralsources.The bulkof it wascollectedbytheHarvardMultinationalEnterpriseProject.The187parentfirmsin this projectwereall U.S. Fortune 500 companieswith atleastsix foreigninvestments.Amongothervariables,the databasecontains


for each subsidiarythe host country,ownershipstructure,industry(at 4-digitSIClevel),andcharacteristicssuchassizeof assetsandextentof intra-systemsales [Curhan,Davidson& Suri, 1977].A selectsampleof the datacollectedin this projectwasused here.First,thesampleincludesonlysubsidiariesthatdidat leastsomemanufacturing.Second,the sampleis limitedto subsidiariesthatwerestill activein 1975,the last year for which there are data.9These cross-sectionaldata thuscontain ownership"corrections"that the MNEs mighthavemade afterinitialentry[Gomes-Casseres1987],andarethereforemorelikelyto repre-sent a long-run,stablepattern.Third,the sampleexcludesjoint venturesbetweentwo or moreMNEs.A numberof variablesderivedfromthe ProfitImpactof MarketingStrat-egies(PIMS)projectwereaddedto theHarvarddata.Thisproject,admin-istered by the StrategicPlanning Institute, collected detailed businessinformationon over2,000domesticstrategicbusinessunits(SBUs)of some200 large U.S. firms [Schoeffler,Buzzell & Heany 1974;Clark 1984].Industryaverages(4-digitSIC)of selectedvariablesfromthesedatawereused as proxiesfor the characteristicsof the subsidiariesin the Harvarddata. As usual in this type of work, it wouldhavebeen betterto use theactual characteristicsof the subsidiaries,as was indeed done wheneverpossible.Butsubsidiary-levelinformationwasnotavailableon keyvariableslikeR&Dandmarketingintensity,so thatproxieshadto be used.Amongavailableproxies,thePIMSdataseemmoreappropriatethanothers,astheyarebasedon informationfromSBUsof firmsverysimilarto those in theHarvarddata. Furthermore,the proxiesusedherewereat the 4-digitSIClevel,and so aremorelikelyto reflectsubsidiarycharacteristicsthanmoreaggregateproxiescommonlyused in the literature.Tomeasurethe effects of the subsidiary'senvironment,the analysisuseseconomicinformationon hostcountriesgatheredbytheWorldBank[1978],anddataon hostgovernmentownershippoliciesfromtheU.S.Departmentof Commerce[1981].Finally,a measureof how familiarU.S. firmswerewith differentcountrieswas derivedfromDavidson[1980].Definitionsof Variables.FollowingFranko[1971]and Stopford& Wells[1972],a joint ventureis definedhereas any subsidiarywherethe MNEownedless than95%of the equity.10Actualownershiplevelwasnot usedbecausethe difference,say,between100%foreignownershipand80%owaslikelyto beperceiveddifferentlybyfirmsandgovernmentthanthatbetween800/ and 60Wo.Separatetestswithanotherqualitativedependentvariableusinga 50Wo-of-equitycutoff point yieldedsubstantiallythe sameresults[Gomes-Casseres1985].ThisandallothervariablesusedintheanalysesbelowaredefinedinTable1.The independentvariablesare numberedto facilitatecomparisonacrosstables.Eachdefinitionalsogivesthevariable'ssource,andits samplemeanand standarddeviation(sd).


RESULTS AND DISCUSSION

Resultsof logit analysesto estimatethe ownershipeffects of restrictivepolicies are shown in Table2. Becausethe binarydependentvariableisequalto one whenthe subsidiaryis a joint venture,a negativecoefficientinthistableimpliesthatthevariablediscouragedjointventures(andencour-agedwholeownership),whilea positivecoefficientimpliesthatthevariableencouragedjoint ventures. The first analysis (shown in column 2.1)containeda seriesof independentvariablesthoughtto influencefirmpref-erences,butwithouttheinteractiontermsthatwouldseparatetheseeffectsfromthose of bargaining.In otherwords,only the Xs fromequation(1)abovewereincluded,in additionto the restrictive-countrydummyD, buttheinteractiontermsbetweenthesetwo(theDXs) wereexcluded.Theotheranalysis(shownin column2.2)includedtheseinteractionterms(DXs above)to estimateseparatelythe effectsof bargaining.Theseresultsarediscussedin the next threesections.Becauseall theinteractiontermsin column2.2 arebasedon theRestrictiveHost Gov dummy variable, one might expect some multicollinearityproblems.But only two of the pairwisecorrelationsbetweenthese inter-actionterms(no. 13throughno. 23) and the dummyvariable(no. 12)areover .80 and four are over .60. Furthermore,the resultsin the two caseswithhighcorrelations-theFamiliarityandGDP Growthterms-are stillsta-tisticallysignificantat the .005 level(Table2, discussedbelow), implyingthatthe collinearitydid not leadto highstandarderrorson the estimates.Elsewherein the datatherewerefewcorrelationsover .50;most werewellbelow .20. In each of the caseswherethe coefficientsin Table2 weresta-tisticallyinsignificant,this did not seem due to collinearityproblems.

Firm OwnershipPreferences

ElsewhereI discussedat lengththeresultsof testsof a modelof firmowner-shippreferences[Gomes-Casseres1989b].Thebaseregressionin thatpapercontainedthefirstsevenvariablesin column2.1andtherestrictive-countrydummy."The resultswereessentiallythe same as shownhere,and wereconsistentwith the argumentsaboveon how firm strategies,capabilities,and transactioncosts influenceownershippreferences.TheysuggestthatMNEs preferwhole ownershipwhen they havea lot of experiencein anindustryor a country,whenintra-systemsalesof the subsidiaryarehigh,or whenthe subsidiarywas in a marketing-intensiveindustry.But MNEspreferredjoint ventureswhen they reliedon local inputsof rawmaterialandwhenlocalfirmscouldcontributeskillsto ajoint venture.Ratherthandiscusstheseconclusionsin moredetailhere,it will be moreusefulto testsomeof the argumentsaboutfirmpreferencesusingtheresultsof the fouradditionalvariablesnot discussedin my previouspaper(nos. 8-11).12

Stopford&Wells[1972]andFranko[1987]foundthatrelativelysmallfirmsin an industrytendedto favorjoint venturesmorethanleadingfirms.Thisis consistentwith the argumentsabove and in Gomes-Casseres[1989b].


TABLEIDefinitions of Variables

Dependent VariableMNE Owns<95% Dummy variable equal to 1 if the parent MNE owned

less than 95% of the subsidiary's equity in 1975.From Harvard MNE database. (mean=0.29; sd=0.46)

Independent Variables1. R&D/Rev Sub Ind Average percentage share that R&D expenses repre-

sented in revenues of PIMS SBUs in the subsidiary'sprincipal 4-digit SIC industry. Based on PIMS.(mean = 2.6; sd = 2.3)

2. MNE's #Subs in Ind Number of foreign manufacturing subsidiaries thatthe parent MNE had in 1975 in the same 3-digit SICindustry as the subsidiary's principal product. FromHarvard MNE database. (mean=14; sd=13)

3. >10% Sales Intrasystem Dummy variable equal to 1 if more than 10% of thesubsidiary's sales in 1975 were to other members ofthe parent MNE's system. From Harvard MNE data-base. (mean=0.18; sd=0.38)

4. Resource-Based Sub Dummy variable equal to 1 if the subsidiary's mainproduct was in one of the following 2-digit SICgroups, which can be considered "resource-based"industries: food and beverage (SIC 20), tobacco (SIC21), textile mills (SIC 22), wood except furniture (SIC24), pulp and paper (SIC 26), petroleum (SIC 29),rubber (SIC 30), leather (SIC 31), stone and glass(SIC 32), and primary metals (SIC 33). Subsidiary'sproduct from Harvard MNE database. (mean=0.25;sd = 0.43)

5. Marketing/Rev Sub Ind Average percentage share that marketing expensesrepresented in revenues of PIMS SBUs in thesubsidiary's principal 4-digit SIC industry. Based onPIMS. (mean=11.8; sd=9.9)

6. Industrial GNP of Host Size of the host country's industrial sector in 1976, inmillions of U.S. dollars. From World Bank (1978).(mean = 69,300; sd = 73,600)

7. Familiarity with Host Index (from 0 to 16) of how "familiar" foreign hostcountries were to U.S. MNEs, based on how oftenthese MNEs entered one country before anotherduring 1900-1976. From Davidson [1980]. (mean=9.5;sd = 5.5)

8. Parent Assets<lndavg Dummy variable equal to 1 if the parent's assets in1975 were less than the average for firms in thesame principal 3-digit SIC industry as the parent.Based on Harvard data. (mean=0.39; sd=0.49)

9. Sub Assets>$1OM Dummy variable equal to 1 if the subsidiary's assetsin 1975 were greater than $10 million. Based onHarvard data. (mean=0.34; sd=0.47)


TABLEI(continued)

10. #MNEs in Industry Number of parent firms with at least one foreignsubsidiary in the same 4-digit SIC industry as thesubsidiary's main product. Based on Harvarddata.(mean=14; sd =11)

11. GDP Growthof Host Average annual percentage growth rate of the hostcountry's real per capita GDP in 1960-76. FromWorldBank (1978). (mean=3.6; sd=1.5)

12. Restrictive Host Gov Dummy variable equal to 1 if the subsidiary was inone of the following host countries, which in 1975had policies restricting foreign ownership or encour-aging joint ventures: Australia, Brazil, Colombia,Ecuador, France, Japan, India, Indonesia, Iran,Malaysia, Mexico, New Zealand, Nigeria, Pakistan,Peru, Philippines, South Korea, Spain, Sri Lanka,and Venezuela. (mean=0.45; sd=0.50)

Interaction terms between Independent variables13. RSTRxR&D/Rev Product of Restrictive Host Gov and R&D/RevSub

Indvariables.14. RSTRxMNEs #Subs Product of Restrictive Host Gov and MNEs #Subs in

Ind variables.15. RSTRxlntrasys>10% Product of Restrictive Host Gov and >10% Sales

Intrasystem variables.16. RSTRxRes Based Product of Restrictive Host Gov and Resource-Based

Sub variables.17. RSTRxMktng/Rev Product of Restrictive Host Gov and Marketing/Rev

Sub Ind variables.18. RSTRxlndstr GNP Product of Restrictive Host Govand IndustrialGNP of

Host variables.19. RSTRxFamiliarity Product of Restrictive Host Gov and Familiaritywith

Host variables.20. RSTRxParAssets<lndavg Product of Restrictive Host Gov and Parent Assets

<lndavg variables.21. RSTRxSub Assets>$10M Product of Restrictive Host Gov and Sub Assets>

$10M variables.22. RSTRxMNEs in Industry Product of Restrictive Host Gov and #MNEs in

Industryvariables.23. RSTRxGDP Growth Product of Restrictive Host Gov and GDP Growth of

Host variables.

Becauseof oligopolisticcompetition[Knickerbocker1973],smallfirmscanbeexpectedto bepressedto expandabroadeventhoughthisstretchedtheirorganizationalcapabilitiesto the limit. In these situations,local partnersmightmakevaluablecontributionsto newventures.Indeed,thecoefficienton the variablemeasuringwhetherthe MNE's assets are smallerthanaverageforits industry(ParentAssets< Indavg)is positiveandstatisticallysignificantat the .05 levelin column2.1. It remainspositive,but declinesin significancein column2.2, whichincorporatesbargainingeffects (seebelow).


TABLE2Logit Analyses: Ownership Effects of Firm, Country,and Industry

Variables (standardized beta coefficients with t-statistics) (N=1,877)

Dependent Variable:MNEOwns<95%

Column 2.1 Column 2.2

Independent Variables Betas t-stats Betas t-stats

1. R&D/RevSub Ind -0.031 -0.97 -0.039 -0.752. MNE's#Subs in Ind -0.135*** -4.22 -0.132** -2.753. > 10% Sales Intrasystem -0.075*** -2.59 -0.083** -2.084. Resource-Based Sub Ind 0.116*** 4.00 0.153*** 3.835. Marketing/RevSub Ind -0.214*** -5.94 -0.229*** -3.826. Industrial GNP of Host 0.161*** 4.74 0.185*** 3.707. Familiaritywith Host -0.291 *** -8.82 -0.400*** -7.848. Parent Assets<lndavg 0.052** 1.93 0.058* 1.389. Sub Assets>$1OM 0.075*** 2.68 0.130*** 3.25

10. #MNEs in Industry 0.061** 2.18 0.064* 1.5611. GDP Growthof Host 0.114*** 4.07 -0.035 -0.6712. Restrictive Host Gov 0.266*** 9.50 -0.041 -0.3513. RSTRxR&D/Rev - 0.012 0.2014. RSTRxMNE's#Subs - -0.004 -0.0715. RSTRxlntrasys>10% - 0.015 0.4316. RSTRxRes Based - -0.062** -1.5917. RSTRxMktng/Rev - 0.021 0.3018. RSTRxlndstr GNP - -0.061 -1.0719. RSTRx Familiarity - 0.214*** 3.0620. RSTRx ParAssets < Indavg - -0.007 -0.1621. RSTRxSub Assets > $1OM - -0.081** -2.0322. RSTRxMNEs in Industry - -0.011 -0.2123. RSTRxGDP Growth - 0.315*** 3.03

R2 0.185 0.201

Note: Because the dependent variable is equal to one when the subsidiary is a jointventure, positive effects indicate variables that encourage joint ownership and negativeeffects indicate those that encourage whole ownership.

*Statistically different from zero at .10 level (one-tailed)**Statistically different from zero at .05 level (one-tailed)

***Statistically different from zero at .005 level (one-tailed)

Similar arguments would predict that MNEs investing in relatively largesubsidiaries should prefer joint ventures. In pursuit of scale economies,firms may be forced to set up larger plants than they can support andmanage by themselves. The coefficient on the variable measuring whetherthe subsidiary's assets exceed $10 million (Sub Assets > $10M) is indeed posi-tive and statistically significant at the .005 level in the regression.

Another factor that can be expected to lead MNEs to prefer joint venturesmay be a high degree of competition in the subsidiary's industry. Twoseparate arguments suggest that the costs of a joint venture may decline asthe number of competitors rises. First, the incentive for internalization maydecline as the opportunity for monopolistic pricing falls [Buckley & Casson1976]. Second, transaction costs for transferring technology through contrac-tual means may decline with a rise in the number of competitors [Stobaugh


1984],thusloweringthebenefitsof usingownershipchannelsandreducingthe costs of joint ventures.The coefficienton the variablemeasuringthenumberof MNEsin anindustry(#MNEsinIndustry)is indeedpositiveandstatisticallysignificantat the .05 levelin column2.1. It remainspositive,but declinesin significancein column2.2, whichincorporatesbargainingeffects (see below).A fourthvariablethatappearsto influencetheownershipchoiceincolumn2.1,servesasanexcellentexampleof howbargainingeffectscanbecomeentangledwith firmpreferencesin a mis-specifiedstatisticalmodel. The coefficienton the variablemeasuringthe host country'sgrowthrate(GDP GrowthofHost) is positiveand significantat the .005 level in column 2.1, but theotherregressionshowsthatthisis entirelydueto theeffectthatthisvariablehasin restrictivecountries.Incolumn2.2, thecoefficienton theinteractiontermwiththerestrictive-countrydummy(RSTRx GDP Growth)is positiveandsignificantat the .005level,butthaton themainvariableis not signif-icant.In opencountries,therefore,thisvariableseemsto haveno effectonownershippatterns.(This resultis discussedfurtherbelow.)

OwnershipEffectsof RestrictivePolicies

ComparisonwithPreviousStudies.This last exampleillustrateswhy thespecificationincolumn2.1cannotdistinguishbetweentheindependentvari-ables'effectson ownershippreferencesandon bargaining.Yetthisis essen-tially the specificationused in Kobrin[1987].The resultsdiscussednextsuggestthat, becauseof this mis-specification,many of the factorsthatKobrinconcludedaffectedthe bargainingprocess,in fact influencedfirmpreferences.The adjustmentsthat Fagre & Wells [1982] and Lecraw[1984]madeto attemptto controlfor firmpreferencesalso seemedto havebeen incomplete. Many of the results that they attributedto MNE-governmentbargainingseemalso to be due to the processby whichfirmsselect theirpreferredownershipstructures.Overall,it is strikingthatthevariablesincludedinthisstudyseemto affectfirmpreferencesmuchmorethantheydo relativebargainingpower.Sevenof the twelvemainvariables13in 2.2 arestatisticallysignificantat the .05level,comparedto onlythreeof thetwelveinteractionterms.As discussedabove,thetestof whetheranyvariableaffectsbargainingpowerdependsonthe statisticalsignificanceof its interactionwith the restrictive-countrydummy.In otherwords,column2.2 suggeststhateightof the twelveinde-pendentvariablesin this studydo not affect the relativepowerof MNEsand governments.14Whatis remarkableis thatmanyof the variablesthatheredo not seemtoaffectbargainingoverownershiparepreciselythosethatpreviousresearchersfound didjust that. Forexample,Fagre& Wells[1982]and Lecraw[1984]found that the extent of intra-systemsales, the R&D- and marketing-intensityof the subsidiary'sbusiness,andthe degreeof competitionin theindustryall affectedbargainingoverownership.None of the coefficients


on thecorrespondinginteractiontermsin column2.2 arestatisticallysignif-icant evenat the .10level(RSTRxIntrasys> 10/o, RSTRxR&D/Rev, RSTRxMktng/Rev, and RSTRx MNEs in Industry).

Ininterpretingthesedifferences,someof thelimitationsof thisstudyshouldbe kept in mind. Many of the proxiesused hereare crude,eventhoughmostareidenticalorsimilarto thoseusedin previousstudies.Inparticular,themeasureof whatconstitutesa restrictivehostgovernmentpolicy-whichis centralinthisanalysis-is binaryandbasedon aggregatestudies.Further-more,althoughmulticollinearitydoesnot seemto havebeena problem,itmightstillhaveinflatedthe standarderrors,leadingto low t-statistics.Asa result, some of the insignificantcoefficientsmay well be due to themethodsand proxiesused.Factors Affecting Ownership Outcome. A substantial body of literatureargues that technology,marketpower,and other factors affect MNE-governmentbargaining.Howcantheseviewsbe reconciledwiththeresultsin Table2? It is best to examinethis questionseparatelyfor each factor.This study did find supportfor one of Fagre& Wells'shypothesesthatfailed theirown tests. They expectedthat MNEs makingrelativelylargeinvestmentswouldhavegreaterbargainingpowerthanothers,butcouldfindno conclusiveevidenceof this.ThecoefficientonRSTRx SubAssets> $10Min Table2 is indeednegativeand statisticallysignificantat the .05 level.Thisimpliesthatwhenmakingmajorinvestmentsabroad,MNEsin restric-tivecountrieswerelesslikelyto formjointventuresthanthosein opencoun-tries, as impliedby Fagre& Wells'shypothesis.The estimateof the effect of R&Dintensityis perhapsone of the mostsurprising.In Gomes-Casseres[1989b]I describedhowR&Dcanhavetwoopposinginfluenceson ownershippreferences.15Firmsexploitingtheirowntechnologypreferwholeownershipin industrieswithhighR&Dspending,butfirmsacquiringtechnologypreferjointventures.Theseopposingeffectsseemto canceleachotherout to yielda coefficienton R&D/RevSubIndthatis not significantlydifferentfromzero(Table2). As a result,theeffecton bargaining(measuredwithRSTRx R&D/Rev)canalsobe expectedto bestatisticallyinsignificant.Furthermore,substantialcase evidencesuggeststhat bargainingpowerofR&D-intensivefirmsvarieswithintheirindustries.Grieco's[1982]studyofthecomputerindustryin India,forexample,showsthatwhileIBMrefusedto givein to the government'sdemandfor local participation,otherfirmsdid.Similarly,Franko's[1987]studyof a largegroupof countriesfoundthat"insider"firmsrefusedjoint ventureswhile "outsider"firmsgavein togovernmentdemandsin boththecomputerandpharmaceuticalindustries.Recently,IBM gained an exceptionto Mexico'sownershippolicies, butAppleandDECdidnot.16 Sucheffects,of course,cannotbemeasuredwithan industryaveragesuch as used here.A numberof variablesthat wereincludedto try to capturethe effectsofintra-industrydifferencesin bargainingpoweralsodidnotyieldstatistically


significantresults.The relativesize of theMNE affectedownershippref-erencesas discussedabove,butnotbargainingpower(RSTRx ParAssets<Indavg).The samewastruefor the internationalexperienceof the MNE(RSTRxMNE's #Subs),and the degreeof competitionin each industry(RSTRxMNEs inIndustry).17All of theseproxiesmayhavebeentoo crudeto measurethe kindsof effectsdescribedby Franko[1987].Fagre&Wells[1982]reportedsimilarproblemswiththeirmeasureof industrycompetition.Twovariablesincludedto measurethe effectof verticalintegrationyieldedsomewhatmixed results.The extentof a subsidiary'sintra-systemsalesseemednot to affectbargainingpowerin Table2 (RSTRx Intrasys> 107o),contraryto expectationsand earlierresults[Fagre& Wells1982;Lecraw1984;Reuber1973].However,as discussedbelow, this variableaffectedentrydeterrencesignificantly.Theotherindicatorof verticalintegration,thistimebetweentheMNEandlocal rawmaterialproducers,did seemto affect the bargainingpoweroftheMNEs.TWoconflictinghypothesesmightbe reasonablein thiscase.Ontheonehand,MNEsdependingon localrawmaterialsmightyieldto govern-ment pressurein orderto gain access to the inputs. On the other hand,governmentsseekingto developandexploittheircountry'snaturalresourcesmightyield to the firms' demands.In fact, investmentin resource-basedindustriesgenerallyled MNEs to prefera joint venture.18But when, dueto other circumstances,they chose whole ownership,MNEs investinginresource-basedindustrieshad greaterbargainingpowerthan others.ThecoefficientonRSTRx ResBasedis negativeandstatisticallydifferentfromzero at the .05 level, supportingthe secondhypothesisabove.19In otherwords,it appearsthat governmentrelianceon MNEsfor the developmentand exportof resourcesincreasedthe bargainingpowerof the firms.Whilemanyof the resultsaboveare surprising,it is somewhateasiertounderstandwhymarketingintensitydoesnot seemto affectthebargainingprocess.ThecoefficientonRSTRxMktng/Revis not statisticallydifferentfromzeroat the .10level.20Here,too, theremaybe differenceswithinindus-tries.But, inaddition,therearefewerreasonswhyhostgovernmentsshouldbeexpectedto givein to MNEdemands.Unlikewithhigh-technologyindus-tries,mostgovernmentsdo not setoutto developindustriesthatcreatevaluethroughadvertising.Some governmentsevenarguethat royaltiespaid toforeignfirmsforuseof trademarksis a netlossto thecountry,unlikeroyal-ties for importedtechnology.It seems morereasonableto assumethat host governmentshavegreaterbargainingpowerthanMNEsin suchindustries.21In returnfor ownershipconcessions,host governmentscan grantconsumergood firmsprotectionagainstimportsor restrictthenumberof domesticcompetitors.Thisseemsto have been the patternin India [Encarnation& Vachani1985], andperhapstoday in China. Thesecountries,in effect, use the attractionoftheirdomesticmarketsto gain concessionsfrom MNEs.


Thereis strong evidencethat the attractionof the domestic marketincreasesthegovernment'sbargainingpoweracrossa broaderrangeof countries,assuggestedby previousresearchers[Reuber1973; Lecraw1984; Kobrin1987].The moreattractivea host countrymarket,the morean MNE willbe willingto tradeawayits ownershippreferencesforaccessto themarket.Butmarketattractivenessmightbe measuredin variousways.The resultsin Table2 suggestthatthe sheersizeof a country'smarketdoes not affectbargaining,but thatits growthratedoes. TheformerwasmeasuredbyIn-dustrialGNP of Host, whichappearsto encouragejoint venturesequallyin both open and restrictivecountries.22Economicgrowth,on the otherhand, seemsto encouragejoint venturesonly in restrictivecountries:thecoefficienton the interactionterm(RSTRx GDP Growth)is positiveandsignificantat the .005 level,but that on the mainvariable(GDP Growthof Host) is not significant.Onereasonfor thisresultmaybe thatthe largenumberof competitorslikelyto be presentin largemarketsmightreducetheattractivenessof entry,whilethecapacityconstraintslikelyin a rapidlygrowingmarketwill increaseit.Anothercountryvariablealso seemsto affectbargainingoverownership.Whencountriesarearrangedaccordingto theirrelativefamiliarityto U.S.firms[Davidson1980],governmentsof the leastfamiliarcountriesappearto have less bargainingpowerthan others.Firmsgenerallypreferwholeownershipin familiarcountries,as indicatedbythenegativecoefficientonFamiliaritywithHost. One reasonfor this is that U.S. firmswould haveless to learnaboutthesehost environments[Gomes-Casseres1989b].But,MNEscanbe expectedto yieldto the ownershipdemandsof governmentsof familiarcountriesmoreoftenthanto thedemandsof governmentsof lessfamiliarcountries.Onereasonforthis is that familiaritywiththeenviron-mentis likelyto resultin betterworkingrelationshipswithlocal partners,and thus lower risks in "forced" joint ventures.This expectationisconsistentwiththepositivecoefficientonRSTRxFamiliarity,whichis statis-tically significantat the .005 level.It is striking that these two country factors-economic growth andfamiliarity-seem to havethegreatestimpacton bargainingthantheothervariablesconsideredhere. In the simple regressionin column 2.1, therestrictive-countrydummy(RestrictiveHost Gov)hada positivecoefficient,but in the regressionwith interactionterms it is effectivelyzero. Thissuggeststhatownershippoliciesdo not haveanyacross-the-boardeffectonownershipstructure.23 Rather,theireffectdependscompletelyon thechar-acteristicsof the industry,the subsidiary,and,especially,the hostcountry.This broadconclusionaffirmsthe thrustof the bargainingmodel, eventhoughotherresultsleadmeto discountspecificvariablesreportedassignif-icant by previousauthors.

EntryEffectsof RestrictivePolicies

MNEshavetwo choicesif theirownershippreferencesconflictwiththoseof the government:(1)negotiatea compromise;and (2) declineto invest.


The resultsin Table2 discussedabovesuggestthat the abilityof MNEsto gainownershipconcessionsdidnot varymuchwithfirmcharacteristics.But theirdecisionsof whento declineto investdid. Thus, some types offirms turnedawaymore easily when faced with restrictivepolicies thanothertypes.Suchdifferentialeffectsof ownershiprestrictionshavereceivedlittle attentionin the literature.Theresultsof teststo measuretheeffectof restrictivepolicieson entrydeter-renceareshownin Table3. Thefirmcharacteristicslistedin thistablewereindependentvariablesin separateOLSorLogitregressions.Theindependentvariableswerecountryandfirmfactors,aswellas thedummyvariableindi-catinga restrictivecountry(RestrictiveHost Gov).As a result,thecoefficientson this dummyvariablemeasurethe differencein the dependentvariablebetweenopen andrestrictivecountries.Thisis model (2) describedabove.Amongthe variableswith strongeffectson entrypatterns,the proxiesforvertical integrationstand out. Subsidiariesin restrictivecountriesweremuchless likelyto havehigh intra-systemsales (> 10%SalesIntrasystm)thanthose in opencountries;this differencewasstatisticallysignificantatthe .005level.On average,12%oof the subsidiariesin restrictivecountrieshadintra-systemsalesexceeding10o, comparedto 22Woin opencountries.This resultis consistentwith otherfindingson the effect of this variable.MNEsinvestingin subsidiariesthatwerehighlyintegratedintotheirglobalsystempreferredwhole ownership[Gomes-Casseres1989b].Becausetheyfailedto gain concessionsfromhost countrygovernments(Table2), theymayhavedecidedto avoidinvestingin restrictivecountries.Wheretheydidgivein to governmentdemandsforajointventureandwentaheadwiththeinvestment,they could be expectedto modify the subsidiary'sstrategytoreducethe extentof intra-systemsales.That, too, wouldleadto the resultobservedhere.Verticalintegrationbetweenthe subsidiaryand local firmsalso seemedtohavehadimportanteffectson entrydecisions.Firmsinvestingin resource-basedsubsidiaries(Resource-BasedSub)werealso morelikelyto investinopen countriesthan in restrictiveones;this effect was statisticallysignif-icantat the .05level.Onaverage,22%oof all subsidiariesin restrictivecoun-trieswereresource-based,comparedto 27Woin opencountries.It appearsthat firmsseekingresourcesabroadwereconcernedwith the governmentinterventionimpliedby the restrictivepolicies. This fear may havebeenaccentuatedby the fact that resource-basedsubsidiariesfaceda relativelygreaterriskof expropriationthanothers[Bradley1977].Thispatternmayhavecontributedto the MNEs'abilityto gainownershipconcessionsfromhostgovernments,as shownin Table2. Thegovernments'goalof exploitingnationalresources,combinedwiththe MNEs'willingnessto go elsewhereforrawmaterials,couldwellhavetippedthebalanceof powerin ownershipnegotiations.Restrictiveownershippolicies also seemed to have deterredfirms withrelativelylittle internationalexperience.TheMNEsinvestingin restrictive


TABLE 3Effects of Firm Characteristics

on Deterrent Effects of Restrictive Policies

Coefficients on Restrictive Host GovControlling for Effectsof Countryand Other Regression

FirmCharacteristics FirmVariables Method(Dependent Variable) Coefficient t-statistic Used

1. R&D/RevSub Ind 0.04 0.40 OLS2. MNE's#Subs in Ind 1.51*** 2.57 OLS3. > 10% Sales Intrasystem -0.17*** -4.83 Logit4. Resource-Based Sub Ind -0.06** -2.29 Logit5. Marketing/RevSub Ind 0.26 0.58 OLS8. Parent Assets<lndavg 0.05** 1.96 Logit9. SUB Assets>$1OM -0.03 -1.00 OLS

10. #MNEs in Industry 0.30 0.63 OLS

Notes: The regression analyses include variables 1 through 12. Natural coefficients areshown for the OLS regressions, and standardized coefficients for the Logit regressions.Because variable Restrictive Host Gov is equal to one in restrictive countries, positive co-efficients indicate that the average of the firm characteristic is higher in restrictive coun-tries than in open countries; the reverse is true for negative coefficients. For example,the results on variable 3 above indicate that subsidiaries in restrictive countries havelower intra-system sales than those in open countries.

*Statistically different from zero at .10 level (one-tailed)**Statistically different from zero at .05 level (one-tailed)

***Statistically different from zero at .005 level (one-tailed)

countrieshad more subsidiariesworldwidethan those investingin opencountries,as shownby the positivecoefficienton MNE's #Subsin Ind,which is statisticallysignificantat the .005 level. Theremight be twoexplanationsforthisresult.First,theexperienceof operatingin a varietyofenvironmentsmighthavehelpedthesefirmsdealwithrestrictivehostgovern-ments, eventhoughthey could not win ownershipconcessions(Table2).For example,they may have learnedhow to managerelationswith hostgovernments,or how to managejoint ventureseffectively.Second,whilerestrictivecountriesmaynot belocationsof choicefora firmjustbeginningexpansionabroad,theymayofferpositivemarginalreturnsto a firmthatis alreadyactivein all the choice locations.But whilethe firmsthat refusedto be deterredby restrictivepolicieshadextensiveinternationalnetworks,theyweretypicallynot thelargestin theirindustries.Thepositivecoefficienton ParentAssets<Indavg suggeststhatthelargestfirms in anindustrystayedawayfromrestrictivecountriesmorethan smallerfirms;this effect was statisticallysignificantat the .05 level.Onaverage,47%opercentof MNEsin restrictivecountrieshadassetsbelowtheirindustry'saverage,comparedto 37Woin opencountries.Thesmaller,second-tierfirmsweremorelikelyto formjoint venturesvoluntarilythanothers (Table2), and were also more likely to give in to governmentdemandsin orderto gain marketaccess [Franko1987].


Alonga numberof otherdimensions,theMNEsinvestinginrestrictivecoun-triesweresimilarto those in opencountries;in otherwords,the restrictivepoliciesherehad little deterrenteffect. In particular,this deterrenteffectdid not varywith the R&Dor marketingintensityof the subsidiary,withthe numberof firms in an industry,or with the size of the investment.Exceptingthis last factor,all of these also seemedto havelittle effect onownership outcomes (Table 2), contrary to some arguments of thebargainingschool. It is thereforenot surprisingto find that theyalso hadlittleeffecton entry.Firmsinvestingin largesubsidiaries,however,didgainownershipconcessions(Table2), and so may not have neededto avoidrestrictivecountries.

CONCLUSION

The choice of ownershipstructurefor a foreignsubsidiarydependsonfactorsin twodifferent,but interacting,processes.Onesetof factorsaffectswhatthefirm wants,i.e., itspreferredownershipstructurefora subsidiary.These factorsincludethe capabilitiesof the firm, its strategicneeds, andthetransactioncostsof differentwaysof transferringcapabilities.Anothersetof factorsdeterminewhatthefirm canget, whichmaybe differentfromwhatit wants.Inparticular,whenthefirmpreferswholeownershipbutthehostgovernment'spoliciestryto encouragejoint ventures,thentheowner-shipstructureof the subsidiarywillbe determinedin negotiationsbetweenfirmandgovernment.In thisprocess,the relativebargainingpowerof thepartiesaffect the outcome.The existenceof two distinct processesled to the developmentof twoschoolsof thoughtaboutthe ownershipdecision:the transactioncost andthebargainingpowerapproaches.Butbecausetheownershipstructuresthatweactuallyobserveall stemfroma mixtureof the two processes,previousresearchershavehad difficultytestingthe argumentsof eitherschool ofthought.As a result,therehas beenconsiderablecontroversyaboutwhichapproachbest explainsownershippatterns.Thispaperdevelopsa conceptualframeworkthatcombinesthe argumentsof both schools of thoughtand uses statisticalmethodsthat empiricallyseparatethe effectsdueto the two processes.Thesemethodsareusedhereprimarilyto test the predictionsof the bargainingschool. Overall,theresults suggest that several factors previously thought to affect thebargainingprocess (e.g., R&D intensity,marketingintensity,and intra-systemsales),in factdo so onlyto a limitedextentor not all all. Butotherfactors(e.g., marketattractivenessand subsidiarysize) do seemto affectthe outcome of ownershipnegotiations.This analysisthus constitutesapartialconfirmationof the overallbargainingpowerapproach.Thepaperalsoexplorestheconditionsunderwhichfirmsfacingrestrictivegovernmentsdecideto foregoinvestingaltogetherinsteadof yieldingto theownershipdemands.This option has been suggestedin the literatureonbargaining,but it has not been examinedempirically.The resultssuggest


thatwhetheror not firmsforegoinvestingdependson theircharacteristicsandthoseof theproposedsubsidiary,suchas the firm'ssizeandtheextentof intra-systemsales of the subsidiary.Thus, some factors previouslythoughtimportantinbargainingappearto affectnottheownershipoutcome,but the entrydecisionitself.Theseresultshaveimportantimplicationsforinternationalbusinesstheoriesandforbusinessandgovernmentpolicies.Theysuggestthata theoryaboutmultinationalenterprisebehaviorshouldconsidernot only the costs andbenefits perceivedby the firm, but also the impactof relationsbetweenfirmsand governments.Recognizingboth theseeffects,managersshouldanalyzethe two aspectsof globalstrategy,i.e., whatis ideal for the firm,andwhatthe firmcanget. And governmentpolicymakersshouldconsiderthatforeigninvestmentpoliciesmayaffectnot onlyhowfirmsorganizetheirsubsidiariesbut also whetherthey will invest at all. Takingaccount ofmultiple processes in theory and policy is never easy, but integratedapproachessuch as that in this papercan help.

NOIES

1.ThisdiscussionassumesthattheMNE'sowncapabilities,too,arebesttransferredthroughownershipchannels.Otherwise,theMNEcouldwritea licensingcontractwiththelocalfirm.TheexampleabovealsoassumesthattheMNEneedsto acquirethelocalexpertise.If it didnot, it wouldprefera wholly-ownedventure.

2. Of course,the samemaybe truefor one firmfacingtwo differentsituations,suchas subsidiariesin differentcountriesor industries.Ownershippreferencesdependnot onlyon firmfactors,but alsoon industryandcountryfactors.See Gomes-Casseres[1989b].

3. Tousetheterminologyintroducedabove,thesegovernments'preferenceis forajointventure,althoughtheirreasonsmayvary,andtheircost/benefitcalculationsonthisscoredifferfromthoseof theMNE.

4. Inotherwords,inthegovernment'spreferencerankingof ownershipstructures,100%foreignowner-shipmayrankcloseor farbehindjoint ventures,andthe MNE'scontributionsmaywellreversetheranking.5. I am indebtedto an anonymousrefereefor valuablesuggestionson statisticalmethods.

6. LogitanalysisandOLSregressionsdiffersignificantly,however,inthedefinitionandinterpretationof R2as a measureof goodnessof fit. Amemiya[1981,pp. 1504-7]considersseveraldefinitionsandfindsallareflawed;butheoffersnogoodalternative.Themeasureusedhereis thatinMaddala[1983,pp. 37-8]whichis definedanalogousto R2 in OLSregressions,i.e., the squaredcorrelationbetweenYandE(Y).BothMaddalaandAmemiyareportacontroversyintheliteratureaboutwhethertheupperlimitof this measureis one or less thanone, anda suggestionthat thisR2 maywellbe verylow inlogitmodels.At anyrate,testsof thegoodnessof fit of theregressionspresentedherearenot criticaltotheinterpretationof theresults.Thepaperdependsmuchmoreonsignificancetestsof thecoefficients,whichareanalagousto testsin OLSmodels.

7. Thishypothesiscan, of course,be testedwiththe methodsdescribedabove.See Gomes-Casseres[1989b].

8. Becausethemethodrelieson comparingcharacteristicsof investmentsinopenandrestrictivecoun-tries,it cannotbe used to estimatethe impactof countryfactorson entrydeterrence.

9. Unfortunately,therearenot morerecentdatawiththe levelof detailandthe breadthof coverageof thisdatabase.However,therearegood reasonsto believethatresultspresentedherearestillvalidtoday.Kobrin[1988]comparesownershippatternsinthe1975Harvarddatawiththoseinhismorelimited1985surveyandfindsno substantialdifferences.InGomes-Casseres[1985]I testedwhethertheeffectsof a number of independentvariables examined here changed over the years 1960-1975,


andfoundthat theydidnot. Therelationshipsexaminedherethusseemedto havebeenquitestable,eventhoughtheremighthavebeenchangeinnumbersof subsidiaries,distributionacrosscountriesandindustries,andso on. Furtheranalysisof timetrendsis in Gomes-Casseres[1988].10.Actually,the Harvarddatadidnot collectinformationon subsidiarieswithless than5% foreignownership,so that the joint venturesin this analysishavebetween95%and 5% MNEownership.11.Toreduceconfoundingeffectsof bargaining,thepaper'ssampleexcludedsomeof themostrestrictivecountriesthat areincludedhere,i.e., Japan,Spain,SriLanka,India,Mexico,andPakistan.

12.Thesevariablesareincludedherebecausetherearecompellinghypothesesabouttheireffectsonthe bargainingprocess.

13. "Mainvariables"is usedhereto referto the Xs in equation(1) above.Theyincludeall firm,industry,andcountryvariables,exceptfortherestrictive-countrydummy(RestrictiveHostGov)andtheinteractiontermswiththatdummyvariable.

14.Notethatthesevariablesmightstillaffectthedeterrenceeffectof restrictivepolicies,as discussedbelow.15.Fagre&Wells[1982],too, reportedthattherelationshipbetweentechnologyandbargainingpowerwascomplexandnon-linear.However,I foundno evidenceof suchnon-linearityin theresidualsfromthe regressionsin Table2.

16.A comparisonof IBM'sbargainingwithIndiain 1978andMexicoin 1985is in Gomes-Casseres[1989a].17.Highlyexperiencedfirmswerelikelyto preferwholeownership,becausetheydidnotneedthecontri-butionof local firms.ThecoefficientonMNE's#Subsin Ind is negativeandstatisticallysignificantcoefficientat the .05 level.See also Gomes-Casseres[1989b].18.PositivecoefficientonResource-BasedSubInd.Dependenceonlocalresourcesmightimplya bilat-eralmonopolysituationleadingto hightransactioncoststhatcanbe minimizedbysharingownershipwithproducersof the local resources.See Gomes-Casseres[1989b].

19. Becauseof the two conflictinghypotheses,somemightpreferto applya two-tailedtest here.Inthat case,the null hypothesisof no effect is rejectedonly at the 10%confidencelevel.

20.Fagre&Wells's,Kobrin's,andLecraw'sresultsto thecontraryarealmostcertainlydueto thestrongeffectof marketingintensityonfirmpreferences.ThecoefficientonMarketing/RevSubIndis negativeand statisticallysignificantat the .005 level.

21. Usingdifferentstatisticalmethods,I foundsomeevidencefor this. SeeGomes-Casseres[1985],Chapter7.

22. As discussedin Gomes-Casseres[1989b],the numberof local firmsthatcan contributevaluableskillsto the MNE'soperationis likelyto be greaterin countrieswith largeindustrialsectors.As aresult,MNEsaremorelikelyto preferjoint venturesin thesecountriesthanin smallerones.

23. Intermsof equation(1)above,theinterceptsin openandrestrictivecountriesareequal:A1=AO,so that (A1-AO)=0.

REFERENCES

Amemiya,Takeshi.1981.Qualitativeresponsemodels:A survey.Journalof EconomicLiterature,December:1483-1536.

Anderson,Erin&HubertGatignon.1986.Modesof foreignentry:A transactioncostanalysisandpropo-sitions.Journalof InternationalBusinessStudies,Fall:1-26.

Bradley,DavidG. 1977.Managingagainstexpropriation.HarvardBusinessReview,July-August1977.

Buckley,PeterJ. &MarkCasson.1976.Thefutureof themultinationalenterprise.NewYork:Homesand Meier.

Clark,KimB. 1984.Unionizationandfirmperformance:Theimpactonprofits,growth,andproduc-tivity.TheAmericanEconomicReview,December:893-919.

Curhan,JoanP., WilliamH. Davidson,&RajanSuri.1977.Rtacingthe multinationals.Cambridge,MA:Ballinger.

Davidson,WilliamH. 1980.Experienceeffectsininternationalinvestmentandtechnologytransfer.AnnArbor:U.M.I. ResearchPress.


Encarnation,DennisJ. & SushilVachani.1985.Foreignownership:Whenhosts changethe rules.HarvardBusinessReview,September-October:152-60.

Fagre,Nathan&LouisT. Wells,Jr.1982.Bargainingpowerof multinationalsandhostgovernments.Journalof InternationalBusinessStudies,Fall:9-23.

Franko,LawrenceG. 1971.Joint venturesurvivalin multinationalcorporations.NewYork:Praeger.

. 1987.Newformsof investmentin developingcountriesby U.S. companies:A five industrycomparison.ColumbiaJournalof WorldBusiness,Summer:39-56.

Gomes-Casseres,Benjamin.1985.Multinationalownershipstrategies.DBAdissertation,HarvardUniver-sity.

. 1987.Jointventureinstability:Is it a problem?ColumbiaJournalof WorldBusiness,Summer:97-102.

. 1988.Jointventurecycles:Theevolutionof ownershipstrategiesof U.S. MNEs,1945-75.InFarokJ.Contractor&PeterLorange,eds.,Cooperativestrategiesininternationalbusiness.Lexington:LexingtonBooks.

. 1989a.Jointventuresinthefaceof globalcompetition.SloanManagementReview,Spring:17-26.

. 1989b.Ownershipstructuresof foreignsubsidiaries:Theoryandevidence.Journalof EconomicBehaviorand Organization,January:1-25.

Grieco,JosephM. 1982.India'sexperiencewiththeinternationalcomputerindustry.InternationalOQgan-ization,Summer:609-32.

Hennart,Jean-Francois.1982.A theoryof multinationalenterprise.AnnArbor:Universityof Michigan.

. 1988.A transactioncosttheoryof equityjointventures.StrategicManagementJournal,July-August:36-47.

Johnston,John. 1984.Econometricmethods(thirdedition).NewYork:McGraw-Hill.

Knickerbocker,F.T.1973.Oligopolisticreactionandthemultinationalenterprise.Boston:HarvardBusi-ness School.

Kobrin,StephenJ. 1987.Testingthebargaininghypothesisin the manufacturingsectorin developingcountries.InternationalOrganization,Autumn:609-38.

. 1988.Trendsinownershipof U.S.manufacturingsubsidiariesindevelopingcountries:Aninter-industryanalysis.InFarokJ.Contractor&PeterLorange,eds.,Cooperativestrategiesininternationalbusiness.Lexington:LexingtonBooks.

Lax, DavidA. & JamesK. Sebenius.1986. Themanageras negotiator.Boston:HarvardBusinessSchoolPress.

Lecraw,DonaldJ. 1984.Bargainingpower,ownership,andprofitabilityof transnationalcorporationsindevelopingcountries.Journalof InternationalBusinessStudies,Spring-Summer:27-43.

Maddala,G. S. 1983.Limited-dependentandqualitativevariablesineconometrics.Cambridge,U.K.:CambridgeUniversity.

Reuber,GrantL. 1973.Privateforeigninvestmentin development.Oxford:ClarendonPress.

Schoeffler,Sidney,RobertD. Buzzell&DonaldF.Heany.1974.Impactof strategicplanningon profitperformance.HarvardBusinessReview,March-April:137-45.

Stopford,JohnM. &LouisT.Wells,Jr.1972.Managingthemultinationalenterprise.NewYork:BasicBooks.

Teece,DavidJ. 1986.Transactioncosteconomicsandthemultinationalenterprise.Journalof EconomicBehaviorand Organization,7: 21-45.

U.S.Departmentof Commerce.1981.Theuseof investmentincentivesandperformancerequirementsbyforeigngovernments.A reportpreparedbytheInvestmentPolicyDivision,Offlceof InternationalInvestment,Internationalliade Administration,October,mimeographed.

WorldBank. 1978. Worlddevelopmentreport:1978.Washington:The WorldBank.

Documents

FIRMOWNERSHIPPREFERENCESANDHOSTGOVERNMENT … · 2019-03-08 · (SIC32),andprimarymetals(SIC33).Subsidiary's productfrom HarvardMNEdatabase. (mean=0.25; sd=0.43) 5. Marketing/RevSubInd