# J-O1480-

Strategic Management Journal, Vol. 18:7, 535-556 (1997)

~ EFFECTIVE INTERFIRM COLLABORATION: HOWFIRMS MINIMIZE TRANSACTION COSTS ANDMAXIMIZE TRANSACTION VALUE

r

JEFFREY H. DYERThe Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.

This study of automotive transaction relationships in the U.S.A. and Japan offers abta whichindicate that transaction costs do not necessarily increase with an increase in reiation-specijkinvestments. We empirically examine the conditions under which transactors can simultaneouslyachieve the twin benejits of high asset specificity and low transaction costs. This is possiblebecause the dtfferent safeguards which can be employed to control opportunism have differentset-up costs and result in different transaction costs over dl~erent time horizons. We examinein detail the ~ractices of Javanese firms which result in effective inteq%n collaboration. @..1997 by John- Wiley & Sons: Ltd. “

One of the objectives of transactors seeking jointmaximization of profits should be to create con-ditions which allow them to achieve the jointmaximization result of the zero transaction costmodel (North, 1990: 107)

Economists have long recognized that ‘resourceowners increase productivity through cooperativespecialization’ (Alchian and Demsetz, 1972: 777).Indeed, the value chain in modem economies ischaracterized by interfirrn specialization such thatindividual firms engage in a narrow range ofactivities that are embedded in a complex chainof input–output relations with other firms. Pro-ductivity gains in the value chain are possiblewhen firms are willing to make transaction orrelation-specific investments (Williamson, 1985;Perry, 1989). 1 Recent empirical work confirmsthat investments in relation-specific assets are

Key words: suppliers; transaction costs; asset speci-ficity; collaborative advantage‘ I use the terms transaction and relation-specificinvestmentsinterchangeably, though I typicatly use the term “relation-s~ific” assets to suggest a shift in attention from the ttms-

actton to the economic relationship as the unit of anatysis(see Kogut, 1989).

CCC 0143-2095/97/070535-22 $17.5001997 by John Wiley & Sons, Ltd.

often correlated with superior(Parkhe, 1993; Dyer, 1996a).

However, increased specialization

performance

within a pro-duction network cannot be achieved without acost. When transactors make investments in spe-cialization, transaction costs arise because of thefear of opportunism. A central premise of trans-action cost theory is that transaction costs increaseas transactors make greater asset-specific invest-ments. The standard reasoning is that as assetspecificity increases, more complex governancestructures (i.e., more complex contracts) arerequired to eliminate or attenuate costly bar-gaining over profits from specialized assets(Williamson, 1985 ). Thus, transaction costs arepresumed to increase with an increase in assetspecificity.

The standard empirical test of the markets andhierarchies paradigm has been to predict whethertransactions are conducted within firm boundaries,or across firm boundaries, by the extent to whichthe transactors’ assets are specialized. Mostresearch to date has produced results that areconsistent with the logic that asset specificity isgreater for transactions within hierarchies thanacross markets (Monteverde and Teece, 1982;Masten, 1984; Walker and Weber, 1984). Thus,

Received 16 June 1995Final revision received 23 August 1996

536 J. H. Dyer

we have generally accepted that transaction costsincrease with increases in asset specificity. How-ever, few studies have attempted to explicitlymeasure how market transaction costs changewith_a change in asset specificity. As Williamson( 1985: 105) observed: ‘A common characteristicof these studies is that direct measures of trans-action costs are rarely attempted.’

In a recent study, Dyer ( 1996b) found thatJapanese transactors ( suppliers–automakers) madegreater asset-specific investments than their U.S.counterparts and that these investments were cor-related with superior performance. These resultsare not particularly surprising since various stud-ies have suggested that Japanese suppliers andfinal assemblers have close relationships and areoften part of a ‘keiretsu’ group (Asanuma, 1985;Nishiguchi, 1994). But what was particularlyintriguing was that Japanese transactors incurredsignificantly lower transactions costs2 than U.S.transactors, even though they had made greaterasset-specific investments. Moreover, even withinJapan (i.e., controlling for the institutionalenvironment), the Japanese automaker with themore specialized supplier group (Toyota) hadlower transaction costs than the Japanese auto-maker with the less specialized supplier group(Nissan). These findings appear to be inconsistentwith transaction cost theory which proposes thattransaction costs increase with art increase intransaction-specific investments (Williamson,1985). Moreover, these findings are importantbecause they suggest that firms (productionnetworks) can simultaneously achieve the twinbenejts of high asset specificity and low trans-action costs—a condition that could be animportant source of competitive advantage.Understanding how such a condition is achievedmay provide important insights into effectiveinterfirm collaboration.

This paper has two primary objectives. Thefirst objective is to provide evidence that trans-action costs do not necessady increase with anincrease in asset specificity. This is accomplishedby reporting supplier group specialization andtransaction costs for Toyota, Nissan, Chrysler,

2 An automaker’s transaction costs were operationalized asthe totat number of individuals employed in the procurementfunction divided by the dollar value of goods they procured.Transaction costs involve search, contracting, monitoring, andenforcement costs. Indkiduals involved in procurement areprimasily representative of these costs.

GM, and Ford. Interestingly, the automaker withthe least specialized supplier group (GM) had thehighest transaction costs while the automaker withthe most specialized supplier group (Toyota) hadthe lowest transaction costs. The second, andprimary, objective is to explore why those resultsmight have been obtained. To explain these find-ings, the paper offers five propositions whichare developed from an exploratory study of 50supplier–automaker transaction relationships inthe U.S.A. and Japan and a mail survey of 156suppliers. The exploratory study centered aroundthe question: What increases (or decreases) thecosts of transacting with suppliers (automakers)?The objective was to examine what actions takenby suppliers and automakers serve to increase ordecrease transaction costs.

THEORETICAL BACKGROUND

Extant transaction cost economics (TCE) theorysuggests that as transactors increase their invest-ments in specialized assets, transaction costsincrease because transactors must safeguardagainst the hazards of opportunism (Klein, Craw-ford, and Alchian, 1978; Williamson, 1985). AsWilliamson (199 la 282) has argued, ‘asset speci-ficity increases the transaction costs of all formsof governance.’ Transaction costs can be decom-posed into four separate costs related to trans-acting: ( 1) search costs, (2) contracting costs,(3) monitoring costs, and (4) enforcement costs(Williamson, 1985; Hennart, 1993; North, 1990).Search costs include the costs of gathering infor-mation to identify and evaluate potential tradingpartners. Contracting costs refer to the costs asso-ciated with negotiating and writing an agreement.Monitoring costs refer to the costs associatedwith monitoring the agreement to ensure that eachparty fulfills the predetermined set of obligations.Enforcement costs refer to the costs associatedwith ex post bargaining and sanctioning a tradingpartner that does not perform according to theagreement.

Transaction costs increase when asset speci-ficity increases due to opportunism (defined byWilliamson, 1985, as ‘self-interest seeking withguile’ ). Although investments in specializationboost productivity, the incentive to make trans-action-specific investments is tempered by thefact that the more specialized a resource becomes,

E#ective Interjirm Collaboration 537

the lower its value in alternative uses. The contin- Although contracts are viewed as the primarygent value of a specialized resource exposes its means for safeguarding transactions in Westernowner to a greater risk of opportunism than the economies, alternative means have been offeredowner of a generalized resource (Klein, Crawford, by scholars from various fields. These are typi-and Alchian, 1978). tally referred to as ‘self-enforcing’ agreements,

‘private ordering’, or ‘trust’ (Telser, 1980; Willi-

Types of safeguards or governance structures

To protect against the hazards of opportunism,transactors may employ a variety of safeguardsor governance structures. The term ‘safeguard’(or alternatively ‘governance structure’ ) as usedhere cart be defined as a control mechanism whichhas the objective of bringing about the perceptionof fairness or equity among transactors. The pur-pose of safeguards is to provide, at minimumcost, the control and ‘trust’ that is necessary for

amson, 1985; Sake, 1991). These self-enforcingagreements include informal safeguards such as:relational or goodwill trust (Dore, 1983; Bradachand Eccles, 1989; Sake, 1991) and reputation

(Kreps and Wilson, 1982; Weigelt and Carnerer,1988), as well as formal safeguards such as

jinancial hostages (Klein, 1980) and specializedinvestment hostages (Klein, 1980; Williamson,1983 ),4 Thus, transactors have a variety of waysto protect against the hazards of opportunism.

transactors to believe that engaging in theoff The choice of safeguard influences

exchange will make them better(Williamson, 1985).

transaction costs and value

The most prominent safeguard employed inWestern economies is the legal contract. A legalcontract specifies the obligations of each partyand allows a transactor to go to a third party(i.e., courts/state) to sanction an opportunistictrading partner. For simple transactions, whenasset specificity is low, a classical contract istypically employed. The costs of writing a classi-cal contract are relatively low since the entireobligations of each party are explicitly written‘within the four comers of the document’(Macneil, 1978). As asset specificity increases,transactors will attempt to write a more complexcontract (i.e., a neoclassical contract) with contin-gency clauses which allow for equitable adjust-ment as market conditions chartge.3 In theory,and practice, writing such a contract is morecostly than writing a classical contract (Macneil,1978; Williamson, 1985). Thus, received theorypredicts that as asset specificity increases, so doesthe full array of transaction costs (i.e., con-tracting, monitoring, and enforcement costs). Asasset specificity increases above some threshold,the costs of contacting become prohibitive andtransactors move to unified governance/hierarchy(Williamson, 1985).

There are costs associated with constructing agovernance structure which we will refer to asgovernance ‘set-up’ costs. These are the costsincurred to create the safeguard which governsthe ongoing relationship and typically include thecosts associated with writing contracts, buildingpersonal trust, creating financial hostages, etc.One may consider governance set-up costs as asubset of the total transaction costs incurred in atrading relationship over time. For example, thecosts of writing a contract may be viewed as agovernance set-up cost, but also as a transactioncost (i.e., the contracting/bargaining portion oftransaction costs). The important distinction thatwe want to make here is that governance set-upcosts involve an initial, up-front investment whichcreates the safeguard, which in turn influencesthe ongoing transaction costs (i.e., bargaining,monitoring costs ) in the exchange relationship.

To date, scholars have paid little attention tothe relative costs of different safeguards. How-ever, different safeguards are likely to have di#er-ent set-up costs and result in different transactioncosts over different time horizons. A longitudinalperspective towards transaction costs is required.Moreover, previous research has shown that the

3The transactional dllemmaa aascciated with specific assets 4Of course, asymmetric investments in specific assets do notincrease when environmental uncertainty increases. The pres- reduce the probability of opportunism but in fact raise theence of some degree of uncertainty is assumed throughout potential for opportunistic behavior. Only symmetric invest-thk paper. Without uncertainty, even highly specialized assets ments in specialized assets will reduce the probability ofmay be protected contractually (Mahoney, 1992). opportunism.

538 J. H. Dyer

choice and cost of a particular safeguard will varydepending upon the identity of the transactors andthe transaction characteristics.s Further, beyondcost minimization, Zajac and Olsen (1993) arguethat transactors should be concerned with maxim-izing transaction value through value creationinitiatives. The transactor’s choice of governancestructure influences the incentives of the transac-tors to engage in value creation behavior for‘noncontractibles’ such as innovation, quality, andresponsiveness. For example, in the auto industrythe governance structure (i.e., degree of suppIier–automaker t.mst) will influence whether or not thesupplier willingly engages in value engineering orvalue anaIysis —initiatives designed to continu-ously lower the cost (and price) of the componentbut which occur after the vehicle model is intro-duced. Such initiatives require resources andeffort on the part of the supplier that are notspecifically called for in the contractual agree-ment. Further, the governance structure influenceswhether or not suppliers bring new ideas or newtechnology to a particular automaker. Suppliersare less willing to bring a new technology ordesign to an automaker employing arm’slength/market governance if they believe theautomaker will share it with their supplier com-petitors (in order to maintain competitivebidding).

To illustrate how the choice of safeguardinfluences transaction costs and incentives forvalue creation behavior, consider a hypotheticalcase in which I am a supplier of automotivecomponents and a particular customer requeststhat I produce a customized part that requiresinvestments in specialized tools. Let us considerfour different transaction situations which resultin four different governance strategies. In the firstcase, assume that the owner of the other firm ismy brother. In this case, I agree to make theinvestment due to previously developed goodwilltrust/social knowledge. I may trust him eitherbecause I know his ‘type’ (e.g., he is ‘honest’)

s Recent empirical studies by game theorists using the pris-oner’s dilemma game (PDG) have demonstrated that oppor-tunism varies depending upon the identity of the exchangepartner (see Axehmd, 1984, for a description of the PDG).For example, one study found that husbands and wives made100% cooperative (C) choices, compared with 65% forfriends, and 22% for strangers (Argyle, 1991). Clearly, theidentity of the exchange partner has sm influence on whetherone chooses to be opportunistic or cooperative.

and/or because I can impose social sanctions onhim through our kinship network.e This safeguardis available to me because of considerable priorinvestments in the relationship. Moreover, theincremental transaction costs associated withusing this safeguard are minimal because thegovernance set-up costs have already beenincurred. Thus, I choose to make the investmentand we transact while incurnng very low trans-action costs. The safeguard employed in this case(goodwill trust) may be effective for an indefiniteduration. Further, I am willing to make effortsat value creation beyond the original agreementbecause I am confident that the resulting gainswill be fairly shared.

When contemplating the same transaction witha graduate school classmate, I may not haveenough ‘goodwill trust’ (enough social knowledgeto assess his ‘type’ or to know if he will respondto social sanctions) to protect my investment.Thus, I may choose to incur additional gover-nanceset-up costs in the form of a legal contract.In this situation I rely on both social as well aslegal sanctions to control opportunism. Of course,the contract is of finite duration and will need tobe rewritten when it expires, thereby resulting inadditional transaction costs.

When contemplating the same transaction witha complete stranger, I may be unwilling to relyon a contract to protect my investment. Conse-quently, I may require that the buying firm pur-chase the specialized assets and allow me to usethem. This solution is also likely to involve acontract which specifies ownership of the assetsas weIl as how [ am to use the assets. Moreover,the buyer must incur additional monitoring coststo ensure that I am not abusing the assets. Thus,governance set-up costs and ongoing transactioncosts again increase relative to the previous cases.Further, I am relatively unwilling to make effortsat value creation ‘beyond the contract’ due touncertainty regarding whether or not I will befairly rewarded. Of course, a final solution wouldbe for one of the transactors to vertically inte-grate, thereby bringing the specialized assets

6 Of course, I may refuse to make the investment because Iknow my brother’s ‘type’ (i.e., he is not honest) and/or Iknow that he would not respond to social sanctions. The keyis that my long history of dkct interactions with him (i.e.,my social knowledge) assists me in choosing anappropriatelefficient safeguatd (Sohn, 1994).

Effective Intetj%-rn Collaboration 539

under common ownership to mitigate the bar-gaining problem.

In each of these hypothetical cases the degreeof asset specificity did not change, but the choiceand cost of the safeguard, as well as the ensuingtransaction costs, did change. Thus, we observethat even in situations where the level of assetspecificity between transactors is identical, thetransaction costs may differ. The point is thatwhile transaction costs may increase with assetspecificity, they will also vary independently ofasset specijci~. Further, the transactors’ choiceof governance will influence not only transactioncosts but also the incentives to engage in valuecreation initiatives. Williamson ( 1985: xiii) hasargued that the central problem of economicorganization is to ‘devise contract and governancestructures that have the purpose and effect ofeconomizing on bounded rationality while simul-taneously safeguarding transactions against thehazards of opportunism.’ Due to differences inhistory, preferences, institutional environment,etc., transactors will employ different strategieswith regard to both the level of asset specificityas well as the choice of safeguard.

THE PUZZLE: HIGH ASSETSPECIFICITY AND LOWTRANSACTION COSTS

To examine the relationship between asset speci-ficity and transaction costs, we studied transactionrelationships in the auto industry. The objectiveswere ( 1) to assess the extent to which an auto-maker’s supplier group was specialized to thatparticular automaker, and (2) to measure eachautomaker’s transaction costs associated with pro-curing parts from those suppliers. Following is abrief description of the research methods andresults of this study (see Dyer, 1996a, 1996b, fora more complete description of the researchmethods and measures ).

Sample and data collection

The sample consisted of two Japanese automakers(Nissan and Toyota), all three U.S. automakers,and a sample of their suppliers. The unit ofanalysis was the supplier–automaker relationship.Each automaker purchasing department generalmanager selected a sample of 50 domestic sup-

plier relationships which were representative oftheir supply base.

A survey was then mailed to the supplierexecutive identified by the automaker purchasingdepartment as most responsible for managing theday-to-day relationship. Usable responses werereceived from 36 Nissan suppliers (72% responserate), 38 Toyota suppliers (76Y0 response rate),31 Ford suppliers (62% response rate), 24 Gen-eral Motors suppliers (48% response rate) and 23Chrysler suppliers (46% response rate). Inaddition, the purchasing agent at the automakermost responsible for the supplier relationship wasasked to complete a survey to provide the auto-maker perspective. Usable surveys were receivedfrom purchasing agents for 192 of the 250 sup-pliers.’ On objective questions (e.g., distancebetween the supplier and automaker plant ), sup-pliers and purchasing agents were asked the samequestion. In these instances, purchasing agentresponses were used only if the supplier did notrespond. Consequently, for some questions theresponse rate was as high as 48 out of 50 for aparticular automaker.

Operational measures

Williamson (1985 ) identified site, physical, andhuman asset specificity as distinct types of trans-action-specific investments. Operation measuresof each were included in this study. In addition,we attempted to measure the ‘transaction costs’incurred by the automaker in managing trans-actions with outside suppliers (see Table 1 for adescription of the asset specificity and transactioncost operational measures). Automaker transactioncosts were measured as the total number of indi-viduals employed in procurement for productionparts (including management, purchasingagents/buyers, lawyers, and support staff ) dividedby the total value of goods they procured in1991. This is expressed as the dollw value ofgoods (parts) purchased per procurementemployee. We believe this is a reasonably accu-rate measure of the relative transaction costsincurred by automakers because the procurementstaff (including lawyers) (a) is completelyresponsible for searching for new suppliers, (b)

7 One U.S. automaker limited its participation in this part ofthe study and dld not have purchasing agents fill out a surveyfor its 50 suppliers.

540 J. H. Dyer

Table 1. Summary of asset specificity and transaction costs operational measures

Variables Description

Asset specificity1. Site specificity

2. Physical assetspecificity

3. Human specificity:face-to-face contact

Transaction costsProcurement costs

The distance in miles between the supplier plant producing the highest dollarvolume component and the automaker’s small and mid-size model plants towhich it deliversPercent of supplier’s total capital equipment investment which is notredeployable to other customers (estimated by supplier respondents)Total annual ‘man days’ of face-to-face contact between supplier salesmen andengineers and automaker purchasing personnel and engineers.

The total number of individuals employed in procurement for production parts(including management, purchasing ‘agents/buyers, lawyers, and support stiff)divided by the total dollar value of goods they procured. This is expressed asthe dollar value of goods (parts) purchased per procurement employee

is completely responsible for contracting withsuppliers, (c) is primarily responsible for gather-ing information from the other operational unitsto create an overall evaluation (monitoring) ofperformance (though procurement does not actu-ally do all of the monitoring), and (d) is primarilyresponsible for enforcing performance (i.e., thelawyers will take the legal actions necessaty toensure contract fulfillment ). Thus. we believe ourmeasure is a reasonable proxy for the relativetransaction costs incurred by automakers.

Data analysis

The unit of analysis was the supplier–automakerrelationship. One-tailed t-tests were used to testfor differences in supplier-automaker asset speci-ficity. For example, the site specificity (plantdistance) supplier group mean for Toyota’s sup-pliers was compared to the site specificity suppliergroup mean for each of the other automakers.

Results

The results presented in Table 2 indicate that notonly were the Japanese transaction relationshipscharacterized by higher asset specificity than theirU.S. counterparts, but they also had lower trans-action costs. Toyota and Nissan’s suppliers weremore specialized on all three measures of assetspecialization. Their plants were significantlycloser (greater site specificity), they had a higherpercentage of capital investments which were notredeployable to other customers (greater absolute

physical asset specificity), and they had moreface-to-face contact and more guest engineers(greater human specificity). Interestingly, Toy-ota’s suppliers were significantly more specializedthan were Nissan’s on both the site and humanspecificityy measures. Furthermore, we found thatU.S. automakers, and in particular GeneralMotors, had higher transaction costs than didJapanese automakers. The purchased volume ofgoods per person was only $1.6 million for GM,while it was $5.3 million for Ford, $5.7 millionfor Chrysler, and $9.7 and $12.6 million forNissan and Toyota respectively. This analysissuggests that Chrysler and Ford incur roughlytwice the transaction costs procuring parts asJapanese automakers, while GM’s transactioncosts are six to eight times higher. These differ-ences exist despite that fact that transaction-spe-cific investments are considerably lower for theU.S. automotive transactors. Furthermore, eventhough Toyota’s supplier group was more special-ized than Nissan’s, Toyota’s transaction costswere roughly 20% lower than Nissan’s.

These findings can be interpreted as beingeither consistent or inconsistent with extant trans-action cost theory, depending on one’s point ofview. On the one hand, according to receivedtheory, a pair of transactors who employ general-purpose assets in the exchange relationship shouldincur lower transaction costs than competingtransactors who choose to make transaction-spe-cific investments. Recall that Williamson ( 1991a:282) has argued that ‘asset specificity increasesthe transaction costs of all forms of governance.’

Eflective Interjirm Collaboration 541

Table 2. Sample means on asset specificity and transaction cost variables (by automaker)

Variable Chrysler Ford GM Nissan Toyota

Site specljiciky: 543.9 508.8 427.0 113.97 59.2**Distance between manufacturingplants (miles)

Physical speci$city: 18.1~0 19.5% 13.6%** 21.4%t 21.2%tPercent of capital equipmentwhich is not redeployable

Human specijcip: 756.9 1206.2 1106.9 3344.2-/ 7235.8**Annual ‘man days’ of face-to-facecontact

Automaker transaction costs: $5.7m*** $5.3m $1.6m $9.6m $12.6mDollar value of goods procuredper procurement employee

Tests of group differences are one-tailed r-tests assuming unequal variances.**Significantly lower/higher than all other automakers (p < 0.01).tSignificarrtly lower/higher than all U.S. automakers only (p <0.01 ).*Significantly lower/higher than other U.S. automakers only (p < 0.01).***-Min&&s million;of dollars.

Our data show that in the auto industry thisrelationship does not seem to hold. However,one can argue that the relationship between assetspecificity and transaction costs will only hold ifenvironmental and other factors are held constant.Variables such as the legal and institutionalenvironment (i.e., culture ) can influence oppor-tunism and the nature of controls and safeguardsemployed. Thus, one might simply argue that wewould expect asset specificity to be higher, andtransaction costs lower, in Japanese supplyrelations because the environment is different.The Japanese institutional environment mayreduce the prevalence of opportunism due toinformal constraints (Dore, 1983). Williamson( 1985) would view the environment as a shiftparameter which must be held constant forempirical work to be valid. Further, one canargue that Japanese transactors have simplychosen more efficient safeguards than U.S. trans-actors, and thus minimized transaction costs bymore effectively aligning transactions withgovernance structures. Thus, the findings can beinterpreted as being consistent with transactioncost theory.

However, the fact that GM (which employs anarm’s length strategy with suppliers ) and Chrysler(which is now employing a partnership approachwith suppliers) operate in the identical insti-tutional environment and yet have dramatically

different transaction costs suggests that differ-ences in the institutional environment cannotsolely explain these differences in transactioncosts. It seems possible that Williamson ( 1985)may have significantly underestimated the long-term search and contracting costs associated withmanaging arm’s length market transactions. Butregardless of whether these findings are inter-preted as consistent or inconsistent with receivedtheory, an examination of why these results wereobtained is warranted.

AN EXPLORATORY STUDY:BACKGROUND AND METHOD

The findings from this study were the catalystfor an exploratory study designed to understandwhy these results might have been obtained.However, before describing the findings of theexploratory phase of the study it is important tonote that differences in procurement costs acrossautomakers may be driven by a number of factors.In particular, two factors which we would expectto influence procurement costs are ( 1) automakersize (volume ) and (2) the number and complexityof models produced. For example, larger auto-makers should enjoy economies of scale in man-aging suppliers, thereby resulting in fewer person-nel in procurement (per dollar of goods procured).

542 J. H. Dyer

However, although GM is the largest automaker,it had the highest transaction costs. Moreover,Ford and Toyota are similar in size and yettheir procurement costs differ markedly. Althoughthere is likely to be a size effect, it does notappear to be a major explanatory variable.

A second possible reason for transaction costdifferences might be due to differences in thenumber of models produced and the complexityof the models. For example, if GM producesmore models (per dollar of sales) with moredifferent types of components (i.e., perhaps dueto a differentiation strategy) then we might expectmore individuals required in procurement to man-age that complexity. However, studies by Clarkand Fujimoto (1991) and Nishiguchi ( 1994) sug-gest that Japanese automakers have developedmore models (per million dollars of sales) thantheir U.S. counterparts. Moreover, Japanese sup-pliers are more likely than U.S. suppliers todevelop unique parts for their customers.s Thus,if we were to make a prediction based on compo-nent complexity/customization, these studies sug-gest that Japanese automaker transaction costsshould be higher than U.S. automaker trans-action costs.

A third possible reason for transaction costdifferences (i.e., between GM and otherautomakers) may be that GM has chosen a marketpower strategy and is using a large procurementstaff to pit a large number of suppliers againsteach other so that GM gets its inputs at the verylowest price. Thus, GM may be trading off highertransaction costs for lower cost inputs, therebylowering GM’s production costs. This, of course,could be a viable (and not necessarily ineffective)strategy. However, every piece of research evi-dence from the auto industry to date indicatesthat GM does not get lower input prices fromsuppliers. For example, Cusumano and Takeishi(199 1) found that the input prices for Japaneseautomakers were lower than those received byU.S. automakers. Moreover, the input prices werealso declining faster for the Japanese automakers

8Clark and Fujimoto ( 1991) found that 38% of U.S. auto-maker parts were ‘off the shelf parts, while only 18% ofJapanese automaker parts were ‘off the shelf.’ Similarly,Nishiguchi (1993) found that when a U.S. supplier developsa component for a U.S. customer, the same auto componentis fitted into 8.3 car models. In contrast, Japanese supplierssell the identicat part for only 5.7 models, indicating highercustomization of Japanese auto parta for car models.

than the U.S. automakers. This held true evenwhen the Japanese transplants used U.S. sup-pliers.

Furthermore, Dyer (1996a) found that between1982 and 1992 GM, and a sample of its suppliers,were less profitable than Toyota, Nissan, Chrysler,and Ford and their suppliers. GM’s pretax returnon assets (ROA) was 2.8% while Toyota’s was13.0%. Further, GM’s suppliers’ ROA was 4.8%while Toyota’s suppliers’ was 7.1 ~o. Again, ifGM’s strategy of obtaining lower input priceswas successful, we might expect it to have somepositive effect on GM’s profits since purchasedinputs represent a high percent of the value ofany automaker’s vehicle. Although it is possiblein theory that GM is trading off higher transactioncosts for lower input prices, in practice this doesnot seem to be the case.

In summary, many of the factors (i.e., size andcomplexity ) which we might expect to explaintransaction cost differences among automakersdid not offer compelling explanations for thesevariances. Thus, other factors within Japanesetransaction relationships must explain how theysimultaneously achieve high asset specificity andlow transaction costs.9

Data collection

To understand why Toyota and Nissan had lowertransaction costs than their U.S. counterparts (aswell as why GM had higher transaction coststhan Ford or Chrysler), we conducted exploratoryinterviews with both the suppliers and automak-ers. We conducted sernistructttred interviews witheach automaker’s general manager of procurementas well as at least two other executives/ purchas-ing agents from each automaker. In addition, weinterviewed sales and engineering vice presidentsat 50 suppliers (20 Japanese and 30 U.S.

9There may be two additional reasons, which we could notcontrol for, that U.S. firms (in particular GM) might havelower procurement productivity. One is that GM may havebeen too divisionalized. Until 1992 GM’s purchasing wasdecentralized into a number of divisions. Since procurementincurs fixed costs, separating the firm into many separated]visions may have created duplication of purchasingresources. Our data were collected shortly afier purchasingwas centralized. Second. since productivity is also a functionof the degree of effective competition, the collusive behaviorof U.S. automakers joined to erect trade barriers against theirJapanese competitors may have increased the degree of slackamong U.S. automakers, thereby influencing procurement pro-ductivity.

Effective Intetjirm Collaboration 543

suppliers ). Interviews were typically 2 hours in validity are among informants and betweenlength. The purpose of the interviews was to informants and archival sources. Only data thatexplore the actions of the transactors that resulted were consistent across informants and sources

in an increase or decrease in the costs of trans- are reported here. However, to further verify theacting, namely search, contracting, monitoring, accuracy of our analysis and interpretations, weand enforcement costs (see Appendix 1 for inter- conducted a second phase of research whichview questions). involved collecting additional data from a larger

sample of suppliers. We developed a follow-up

Data analysis

We coded the interview transcripts using constantcomparative analysis in which each comment orevent was assigned to an emergent open codingscheme (Glaser and Strauss, 1967; Strauss andCorbin, 1990). Open coding is the process ofnaming and categorizing of phenomena. Duringopen coding, the data are broken down into dis-crete concepts and compared for similarities anddifferences. This produced 21 initial codes ofactions/behaviors that influenced transactioncosts. These were subsequently reduced into fiveincreasingly abstract categories through axialcoding (Strauss and Corbin, 1990). Althoughopen coding fractures the data and allows oneto identify numerous discrete concepts and theirproperties, axial coding puts these data backtogether in new ways by making connectionsbetween subcategories and combining them intocategories. This is done by utilizing a codingpa.radlgm involving conditions, context,action/interactional strategies, and consequences(Strauss and Corbin, 1990: 60-115).

Validity

We used the processes involved in the constantcomparative method as internal checks on thevalidity of the data. The process is described wellby Browning, Beyer, and Shetler (1995: 121):

As the research proceeds and new data are col-lected, they are constantly being compared toprior data in terms of categories and hypotheses.When new data yield new or inconsistent infor-mation, conceptual categories and the emergingtheory are modhied to take them into account.This process is repeated until theoretical satu-ration is reached: until no new categories areemerging and no new information inconsistentwith the categories and tentative hypotheses isbeing generated (Glaser and Strauss, 1%7;Strauss and Corbin, 1990).

survey based on the exploratory interviews andsent it to the original 250 suppliers. Responseswere received from 118 (55 Japanese and 63U.S. ) suppliers and these data are used to validatethe propositions developed during the exploratoryinterviews. In the survey we operationalized thevarious safeguards described in the first sectionto determine which safeguards were most widelyused by suppliers–automakers in each country(see survey questions in Appendix 2). The datafrom the interviews and surveys are the basis forthe propositions regarding how Japanese automo-tive transactors simultaneously achieve higherasset specificity and lower transaction costs thantheir U.S. counterparts (see Table 3 for a sum-mary of data sources).

RESULTS: WHY JAPANESETRANSACTORS HAVE LOWTRANSACTION COSTS

Through our interviews we identified five sepa-rate, but interrelated, propositions to explain thedifferences in transaction costs among automak-ers. For simplicity in developing the propositions,the remainder of the paper focuses primarily oncomparing Japanese transactors (Toyota andNissan) and U.S. transactors (GM, Ford, andChrysler). However, the propositions also serveto explain differences between GM andFord/Chrysler. 10 we ~gue that Japanese auto-

makers have lower transaction costs than theirU.S. counterparts primarily due to:

1. repeated transactions with a small set of sup-pliers;

2. economies of scale and scope in transactingwith that small supplier group (high volumeof exchange between transactors);

In qualitative research, the primary checks on10~is is me ~ith the exception of Propaition 2, which doesnot hold for the U.S. automakers.

544 J. H. Dyer

Table 3. Summary of interview and survey participants

No. of No. of TotalInterview group companies interviewees interview hours

U.S. automakers 3 12 28 hJapanese automakers 2 8 19 hU.S. suppliers 30 61 62 hJapanese suppliers 20 41 4411

No. of No. of companiesSurvey group companies responding Response rate

U.S. suppliers 150 63 44%Japanese suppliers 100 55 55%

3.

4.

5.

extensive interfirtn information sharing whichreduces asymmetric information;the use of noncontractual, self-enforcing safe-guards (i.e., goodwill trust) which are effectivefor an indefinite time horizon (as opposed tocontracts which are effective for a finite timehorizon); andinvestments in cospecialized assets.

The theoretical, and preliminary empirical, sup-port for these propositions is offered below.

Repeated transactions with a small suppliergroup

Japanese automakers work with a much smallergroup of suppliers than U.S. automakers andengage in repeated exchange with those suppliers(see Table 4). Our interviews suggest that repeat-edly working with fewer suppliers results in lowtransaction costs for three primary reasons. First,empirically we found that Japanese suppliers re-win the business at a car model change over 90~oof the time compared to Chrysler and Ford atroughly 79Y0, and GM at 52%. By making thetransaction a repeated game, Japanese automakersincrease the cost of defection/opportunism on thepart of the supplier. 11 The cost of losing the

11me ]Oglc for how repeated games result in more cnor~dve

behavior is well documented in the game theory literature(Axekod, 19S4; Parkhe, 1993). The prospect of continuinginteraction alters the pay-off structure and increases the costof ‘defection.’ Thus, the future casts a ‘shadow’ back uponthe present, affecting current behavior patterns. Cceper-ationis promoted by the establishment of a direct connection

business is greater for suppliers with: ( 1) a highervolume of exchange with the automaker, and(2) the expectation of a long-term relationship.As one Japanese supplier executive observed, ‘Forus to try to take advantage of Toyota would bevery short-sighted. We have too much businesswith them to risk such a foolish action’ (authorinterview, 1992). Thus, the cost of opportunismis higher for the typical Japanese supplier thanfor the typical U.S. supplier. The creation of ahigh-volume, repeated game increases the ‘incen-tive compatibility’ of the transactors.

Second, a ‘repeated game’ allows for moreopportunities in the future to correct for trans-action inequities, thereby reducing bargainingcosts. More than 7570 of the Japanese suppliersinterviewed indicated that they knew that if aninequity emerged in the current set of trans-actions, it would be ‘remembered’ and correctedin the future. The majority of U.S. suppliersindicated that inequities were rarely, if ever,remembered by the automakers. Japanese transac-tors are much more likely to rely on a ‘socialmemory’ which is created through repeated inter-actions. This memory allows Japanese transactorsto achieve ‘serial equity’ (equity over a longerperiod of time) rather than requiring immediateor ‘spot equity’ (Ouchi, 1984). Thus, it reducesthe need for transactors to invest heavily in bar-gaining over profits from the current set of trans-actions.

between a transactor’s present actions and anticipated futurebenefits.

Eflective Intetjirm Collaboration 545

Table 4. Sample characteristics of supplier–automaker relationships (by automaker)

Variable Chrysler Ford GM Nissan Toyota

Average number of suppliers’ 2033 2033 2033 303 303

Percent of time the supplier re- 78% 80% 52’%0** 90%t 92%1wins the part business at amodel change

Percent of time the automaker 21.1% 20.7% 67%** 4.7%+ 4.o’%frebids the part before a modelchange

Annual dollar volume of parts $100.9m*** $111.7m $237.lm* $346.8mt $687.2 m**exchanged

Average number of different 136 226 274 1613t 2379**types of parts exchanged(different part numbers)

‘For confidentiality reasons the number of suppliers reported for each automaker is the nation (U.S./Japan) average(i.e., 2033 is the average number of suppliers for Chrysler, Ford and GM).Tests of group differences are one-tailed r-tests assuming unequal variances.**slgnificmtly knver/higher than all other automakers (p <0.01 ).tSignificantly lower/higher than all U.S. automakers only (p <0.01 ).*Significantly lower/higher than other U.S. automakers only (p < 0.01).** ~m indicates mil]ionsof dollars.

Proposition 1: The higher the probability ofrepeated exchange, the lower the transactioncosts per unit of exchange.

Economies of scale and scope in transacting

On average, Japanese suppliers sell much largervolumes of goods to automakers than do U.S.suppliers (see Table 4). The fact that Japanesesuppliers sell large volumes of product, as wellas multiple products (i.e., part numbers), to theautomaker reduces transaction costs in two ways.First, the costs of sharing information go downas a percentage of the vahte of goods exchangeddue to economies of scale in transacting. Theautomaker typically must collect much of thesame information from the supplier (on martage-ment, finances, production processes, etc. )whether the automaker is buying $5 million or$500 million of parts from the supplier. Thelarger the absolute volume of goods exchanged,the lower the transaction costs per unit ofexchange. Just as scale economies lead to lowerper unit production costs, economies of scaleassociated with increasing the volume ofexchanges with a given transactor leads to lowerper unit costs associated with completing thetransaction.

Second, ex ante and ex post bargaining costsare reduced because economies of scale and scopein transacting provide many more options forcorrecting transaction inequities. For example,when suppliers bid on the business for a pwticularcomponent, they receive a projected unit volumefrom the automaker on which they base theircosts and prices. Of course, this unit volumeis an estimate but the supplier must make theinvestments in plant, tooling, manpower, etc. inorder to be able to produce the projected unitvolume. Automakers verify this investment beforeproduction to ensure that all suppliers have theability to produce the parts they expect they willneed. Of course, in most cases the estimate isincorrect-in some cases too high and in somecases too low. However, according to both sup-pliers and automakers, the estimate is much morelikely to be too high than too low. 12 In thecase of a particularly inaccurate estimate, it isimpossible for the supplier to fully recoup itsinvestments because the assets dedicated to pro-duction are not fully utilized (the supplier’s unitprice to the automaker was based on a unit

I2 Suppliers indica~ mat U.S. automakers’ estimates weretypically 20–30% too high while Japanese automaker esti-mates were more likely to be within 5-10% of actuat volume.

546 J. H. Dyer

volume that did not materialize). The suppliercart respond to this dilemma in a number ofways. First, the supplier can simply ask for ahigher price ex ante which reflects the riskinvolved. 13 Second, the supplier can insist on a

more complex, contingent claims contract to pro-tect its interests and specify damages in the eventthat the estimated unit volume does notmaterialize. Or finally, the supplier can go backto the automaker ex posr and ask for some typeof compensation.

When a supplier is only transacting with theautomaker on a small number of components (asin the U.S.), these responses-which result inhigh transaction costs—are the only ones avail-able to the supplier. However, because Japanesesuppliers typically sell multiple components formultiple car models, ex ante precision is lessimportant, and ex post bargaining is less problem-atic. Japanese suppliers claimed that if the actualunit volume for a particular model was lowerthan expected, they typically did not worry aboutit because they were usually selling a componenton another model where the estimate was low.Thus, according to numerous Japanese supplierexecutives, ‘the high and low estimates tend tocancel each other out.’ In situations where theautomaker estimates were particularly inaccurateand the supplier had made large investments thatcould not be recouped, rather than engage inexpensive haggling on that particular component,the supplier simply asked for additional businesson existing or upcoming models, perhaps at amore favorable price. Consequently, Japanesetransactors spend less time on ex ante contractingand ex post bargaining. Economies of scale andscope in the transaction relationship lower trans-action costs by providing more options to correctfor transaction inequities.

Proposition 2: The greater the total volumeof exchange between transactors, the lower thetransaction costs per unit of exchange.

13 [n ~mctice, in ~ bidding situation it is virtually impossible

for suppliers to add in a risk premium and still win thebusiness. Most suppliers claim that they price based on theautomaker’s estimated volume and hope they will find otherways to make up the shortfatl if the estimated volume doesnot materialize.

Substantial information sharing

Japanese automotive transactors share more infor-mation than their U.S. counterparts, therebyreducing information asymmetry as well as thepotential for opporhmism (see Table 5 ). This inturn reduces transaction costs. The link betweeninformation sharing and transaction costs isstraightfotward. In neoclassical economics, trans-action costs are assumed to be zero becausetransactors have perfect information. Informationasymmetry is necessary in order for transactorsto behave opportunistically. As North ( 1990 108)observes, ‘the costs of transacting arise becauseinformation is costly’ and ‘asymmetrically heldby the parties to the exchange.’ In a transactionworld of perfect information, transaction costsare negligible.

The transaction world of Japanese automakersmore closely approximates a world of perfectinformation. Japanese suppliers and automakersshare a tremendous amount of information ontheir costs, methods of production, technology,and so forth. This reduces the ability of transac-tors to behave opportunistically by concealingrelevant information (Akerlof, 1970 North,1990). The comment of a Japanese supplierexecutive illustrates this point. When asked whyhis company does not try to build additionalprofit into the cost estimates that they provide toNissan, he answered as follows:

It would be vixtually impossible for us to getaway with inaccurate cost estimates. Nissan hasmuch data; they have very good information onour operations and they can analyze our costposition. They can visit our plants and gatherinformation. They know us and our operationsso well, they would surely discover it. Moreover,tJrey have at least one other supplier that com-petes with us and they have detailed informationto compare us. We can hide nothing (authorinterview, 8 August 1992)

Over 90910of the Japanese suppliers we inter-viewed answered this question in a similar man-ner. Indeed, the supplier selection process is suchthat Japanese automakers ‘screen’ for supplierswho are willing to share information. Supplierswho are unwilling to share information are effec-tively screened out. 14lltis point is highlighted by

M~umment managers at both Japanese automakers indi-cated that they screen for suppliers who are willing to shareinformation; suppliers who are unwilling to share information

Effective Interjirm Collaboration

Table 5. Measures of information sharing (by automaker)

Variable Chrysler Ford GM Nissan Toyota

Trust automaker with 4.3* 3.6 2+6** 5.4+ 6.1**confidential information

Share detailed information on 4.I 4.5 4.0 5.3T 4.9tcost structure

Share information to assist 2.0 2.0 2.1 3.3? 3.3-1supplier with cost reduction

Share information to assist 3.1 3.7* 2.2** 3.5* 3.7*supplier with qualityimprovement

Sham information to assist 2.5 2.2 1.9 2.8 3.8**supplier with delivery/inventorymanagement

Note: Answers are on a I-7 Likert scale; 1= Not at all, 4 = To some extent, 7 = To a very great extent.Tests of group differences are one-tailed r-tests assuming unequat variances.**Significantly lower/higher than all other automakers (p < 0.01).?.SignificantlyIowerlhigher than all U.S. automakers only (p <0.01 ).*Significantly lower/higher than other U.S. automakers only (p < 0.01).

the business strategy manager of a GM compo-nent division.

I probably shouldn’t admit this but we actuallyshare more information with Toyota than we dowith our intemat assembly divisions. Why do weshare more with them? Because they demand it.That’s just the way they do business. If we didn’tgive them the information they wanted, theywould not do business with us (author interview,September 1992)

Japanese automakers demand significant cost,quality, and production information from sup-pliers for two reasons. First, a supplier’s willing-ness to share information is viewed as a signalof the trustworthiness of the supplier. Second,they do not take for granted that the supplier canperform as promised. Thus, they want to verifythe capabilities of the supplier. This high degreeof information sharing reduces information asym-metries, thereby reducing contracting and moni-toring costs because both parties are negotiatingwith similar information.

Proposition 3: The greater the degree ofinformation shan”ng between transactors, the

are not viewed as being trustworthy and therefore are elimin-ated from consideration.

547

lower the information asymmetries and thelower the transaction costs.

Use of self-enforcing vs. contractualsafeguards

Japanese automotive transactors do not controlopportunism through legal contracts but insteadrely on self-enforcing safeguards such asrelational trust and financial hostages (stockownership). These safeguards presumably havehigh initial ‘set-up’ costs but once in place theyhave relatively low maintenance costs. 15

Our supplier survey found that Japanese auto-makers, notably Toyota, have been more effectivethan U.S. automakers at getting suppliers to trustthem (see Table 6). Japanese suppliers indicatethat they are more likely to trust Japanese auto-makers to treat them fairly. ‘b Further, Japanese

15We Cmnot empirically examine whether the set-up costsare higher, but theory and some empirical evidence suggestthey would he (Sake, 1992).16 E~@nsive f~e.t~ face interactions between SupPlier Snd

automaker tended to personalize the exchange which increasedgoodwill trust (Sake, 1991). Japanese procurement personnelindicated that long-term friendships and loyatty were factorswhich prevented them from taking advantage of suppliers.U.S. purchasing personnel were much less likely to mentionfriendships and loyalty as reasons for cooperating with suppliers.

●

548 J. H. Dyer

Table 6. Measures of self-enforcing safeguards (by automaker)

Variable Chrysler Ford GM Nissan Toyota

Trust: 5.4 5.0 3.2** 6.lt 6.4?Trust automaker to treat youfairly

If given the chance, automaker 2.9 3.6 5.4** 1.8 1.4**might try to take unfairadvantage of supplier

Reputation: 5.2 4.8 2.8** 5.7? 6.3**Automaker has a reputation forfairness among suppliercommunity

Financial hostages: 070 o% 0% 12.5%~ 8.9%TPercent of supplier stockowned by the automaker

Note: Answers are on a 1–7 Likert scale; 1 = Not at all, 4 = To some extent, 7 = To a very great extent.Tests of group differences are one-tailed r-tests assuming unequal variances.Xxsignificmtly lower/higher than all other automakers (p <0.01 ).t Significantly lower/higher than all U.S. automakers only (p <0.01 ).

suppliers were more willing to make investmentsbased upon the oral promises of the automaker,without a written contract. Interestingly, GM wassignificantly less likely to have fostered trust thanChrysler or Ford. During interviews, suppliersreported cases of patent infringement, sendingproprietary design blueprints to competitors, andbroken contracts as reasons that they did not trustGM. Stated one supplier executive:

Three years ago we were honored by GM as ahigh performance supplier. In fact, we were theonly supplier featured with a picture in theirannual report for our performance. But this yearthey came back and said we would have to dropour price 2070 in order to keep the business.They said they had a lower bid. When we refusedto drop our price that much, they forced us toship the tools to a new supplier and even hadthe nerve to ask us to help the new supplier toget up and running.

Consistent with the hypothesis that ttust is anefficient governance mechanism (Dore, 1983;Sake, 1991 ), these findings support an inverserelationship between trust and transaction costs.

Finally, Japanese transactors are much morelikely to use financial hostages, rather than con-tracts, to make a credible commitment. Forexample, Japanese automakers have minorityownership positions in many key suppliers. Nis-

san and Toyota own an average of 2370 of thestock of their affiliated suppliers and roughly 10?ZOacross all suppliers in our sample (Table 6). Stockownership in a Japanese trading relationship isrepresentative of a credible commitment that onefirm has made to another firm (Gerlach, 1992).A stock ownership position held by an automakeracts as a financial hostage which encourages thesupplier to make partner-specific investments. Theuse of hostages increases the costs of unilateraldefection by the automaker and lengthens the‘shadow of the future’ by signaling good-faithintentions and long-term commitments (Schelling,1978; Parkhe, 1993).

The short-term ‘set-up’ costs of building trustor investing in financial hostages are high relativeto simply writing a legal contract to attenuate thehazards of opportunism. However, once Japanesetransactors have made the upfront investment todevelop these self-enforcing safeguards, the trans-action costs decline over the long term. Self-enforcing safeguards can control opportunismover an irtdejinite time horizon. Conversely, con-tracts control opportunism for only a jinire timehorizon. When the finite duration of the contractis over, transactors must find other means tocontrol opportunism (i.e., write a new contract).

In contrast to Japanese practice, U.S. automrtk-ers have historically worked with a large set of

Eflective Interjirm Collaboration 549

suppliers and relied on legal contracts to controlopportunism. 17The pattern for U.S. auto compa-nies has been to establish requirements and playsuppliers off against one another in a contestfor l-year contracts (Clark and Fujimoto, 1991).Helper ( 1991) found that the average length ofa legal contract employed by a U.S. automakerwas only 2.3 years in 1989, up from 1.2 yearsin 1984. This requires annual or biannual requestsfor bids from suppliers followed by the sub-sequent costs of analyzing the bids, negotiating,rebidding, selecting a supplier, and writing a legalcontract. In short, this practice leads to highsearch and contracting costs. Two-thirds of GMsuppliers in our survey indicated that historicallyGM has rebid their component before a modelchange (Table 4). In contrast, Toyota and Nissanrebid components less than 59to of the time. Asone GM executive noted, ‘It has been necessaryfor us to maintain a large purchasing departmentto manage the large number of suppliers and thefrequent rebidding of parts’ (author interview,February 1991). U.S. automakers continually facethe costs of recontracting with suppliers becauselegal contracts are only efficacious for a finitetime period.

In summary, Japanese automotive transactorsminimize transaction costs by minimizing search,contracting, monitoring, and enforcement costsover the long term. They do this by controllingopportunism in their exchange relationshipsthrough self-enforcing safeguards rather than legalcontracts. Conversely, U.S. automakers recontractwith an ever-changing line-up of suppliers andthus continue to incur considerable search andcontracting costs. While a legal contract mayminimize transaction costs in the short run, thesafeguards used in Japanese alliances result inlower transaction costs over the long run.

Proposition 4: Selj-enforcing safeguards (i.e.,goodwill trust, jinancial hostages) result inlower transaction costs than legal contracts ifthe expected duration of exchange is long term.

Investments in relation-specific orcoqwialized assets

Helper and Levine ( 1992: 566) have observedthat ‘in much of the transaction-cost literature,

IT~o”gh this patternis changing significantly, especi~ly atChrysler (Kamath and Liker, 1994; Dyer, 1996c).

asset specificity is assumed to be exogenouslydetermined by the technology’ and ‘purchasersalways choose the socially efficient level of assetspecificity . . . Governance structure follows in astraightforward way from the degree of assetspecificity.’ 18 However, our interviews suggestthat due to uncertainty and bounded rationality,transactors do not know ex ante what level of

investment in specialized assets will be optimal.Within Japanese supplier–automaker relation-ships, the level of investment in relation-specificassets tends to deepen over time, particularly astransactors make initial investments, share infor-mation, build trust, and ‘discover’ new ways toenhance performance through relation-specificassets, For example, a Nissan seat supplierdecided to build its plant on the property adjacentto a Nissan assembly plant. This decision wasmade primarily because Nissan had a minorityequity position in the supplier and the two partieshad developed a high level of relational trust.Once this site-specific investment was made, thetwo parties discovered that rather than transportthe seats by truck (a general purpose asset), itwould be more economical to build a conveyorbelt (a highly specialized asset) to carry the seatsdirectly from the supplier plant to the automakerplant. Consequently, the supplier and automakerjointly invested in building the conveyor belt.This example of an initial specialized investment(i.e., a site-specific plant) being followed by sub-sequent specialized investments (i.e., customizedequipment) is not unusual among Japanesesuppliers–automakers. A high degree of infor-mation sharing and trust results in higher levels ofrelation-specific investments because the partiesdiscover new ways to enhance performancethrough specialized investments.

In addition, we found that the level of relation-specific investments made by suppliers wasstrongly influenced by the automaker’s strategywith regard to supplier management and gover-nance. To illustrate, when suppliers consideredmaking a particular relation-specific investmentthey asked the following question: ‘Will we makethe necessary return on investment during thepayback period or length of the transaction

18Willimson ( 1985: 34) has argued that reset specificity(technology), contractual safegumds, and price are determined‘simultaneously.’ Thus, asset specificity may be viewed aspartly endogenous in his model.

550 J. H. Dyer

agreement/contract? According to Japanese sup-pliers, a sizable portion of their automaker-spe-cific investments in customized equipment, tools,processes, and capacity had a payback period ofat least 8 years or two model cycles (most Toyotaand Nissan models have a 4-year developmentcycle). Thus, for the investment to pay off, thesupplier needed to be assured that it would beawarded the component business for at least 8years. Japanese automakers provided those assur-ances by giving credible long-term commitmentsto suppliers (as evidenced by the high re-winrates ). In contrast, U.S. suppliers considered theduration of the legal contract as the paybackperiod-which was typically 5 years or less.Thus, U.S. suppliers rationally refused to makerelation-specific investments with a paybackperiod longer than the length of the contract. Thepractical result of these differences in governancestrategy was that a larger set of durable relation-specific investments made economic sense forJapanese suppliers than for U.S. suppliers.

Thus, firm-level (buyer) strategies towardsgovernance ultimately influences both ~ransactkmand production costs. If one supplier uses specialpurpose technologies and assets for productionwhile another refuses to do so due to concernsregarding opportunism, the suppliers will havedifferent production costs. 19 Governance of sup-plier relauons has an effect on production costsby influencing the level of relation-specific invest-ments employed in the relationship. Thus, invest-ments in specific assets are at least partlyendogenously determined. These findings suggestthat asset specificity is not the cause, but to alarge degree the consequence of governance strat-egy.

Finally, because Japanese automakers and sup-pliers often made symmetrical or cospecializedinvestments it resulted in a hostage situation,thereby increasing the interdependence of thetransactors (Klein, 1980; Williamson, 1983; Dyerand Ouchi, 1993). Cospecialized investmentsincrease the transactors’ interdependence and,consistent with theory, serves as an economicrationale for cooperative, long-term relationships.

19 ~i~ finding is Consisknt with Lyons ( 1994). who found

that 60% of U.K. transactors in the engineering field claimedthat they were not utilizing the optimal level of specializedinvestments with their main customer, primarily because theywere unwilling to expose themselves to the risk of beingoppofiunisticrdly exploited.

Of course, transactors will not make the initialinvestments unless they feel sufficiently protectedagainst the hazards of opportunism. Thus, trarts-action costs increase with initial investments inspecialized assets. However, once a high level oftrust is achieved and the initial relation-specificinvestments are made, subsequent investmentsserve as a credible signal of trust and commit-ment. Thus, they are self reinforcing. Specificinvestments, by their very nature, increase the‘shadow of the future’ and increase the expec-tation of future interaction, thereby leading to apattern of cooperative behavior (Heide and Miner,1992; Parkhe, 1993). Thus, we propose a curvi-linear relationship between asset specificity andtransaction costs with trust as a moderating vari-able (see Figure 1). During the early stages of arelationship, transaction costs increase as transac-tors invest in safeguards to protect initial invest-ments in specific assets. However, once a suf-ficiently high level of trust is achieved andspecialized investments are made, the investmentsthemselves serve as a signal of trust and commit-ment. Thus, transaction costs may actuallydecrease with increased investments in specificassets.

Proposition 5: Above some minimum thresh-old level of trust, aaliitional relation-specificinvestments serve to increase commitment andthe costs of unilateral defection, therebyresulting in lower transaction costs.

A model of interfhm collaboration and valuemaximization

Figure 2 proposes a model of interfirm collabo-ration that maximizes transaction value based onthe propositions developed previously. The modelproposed here suggests that the credibility of afirm’s promise to behave cooperatively increasesas transactors: ( 1) demonstrate through behaviora commitment to fiture interaction (e.g., byincreasing the re-win rate and volume ofexchange), (2) increase the amount of informationsharing, and (3) employ self-enforcing safeguardsto govern the relationship. In turn, an increase in‘promise credibility’ (or trustworthiness) withinthe trading relationship reduces transaction costsand increases the likelihood that transactors willinvest in relation-specific assets. Furthermore,increased investments in specialized assets serve

E~ective Inter&n Collaboration 551

High

TRANSACTIONCOSTS

LowI

Low High

INVESTMENTStN RELATION-SPECIFICASSETS(Intert%mInterdependence)

Figure 1. The relationship between asset specificity and transaction costs

to reinforce

DemonstratedCommitmenttoFuture Interaction m

- High Rewin Rate- HighTransaction

Volume

Information (+) TransactionValue

Sharing(Joint Performance)

I I \ 1 /

Use of Self-Enforcinrr /T\ d(+)safeguards ‘1 /

V(+)- OoodwillTrust- Reputation- FhmncialHostage I

Figure 2.

the transactors’ promise

vRcladon-SpeciticAsseta

(+)

A model of interfirm collaboration

credibilityby increasing the cost of unilateral defection andlengthening ‘tie ‘shadow of the future.’zo Finally,lower transaction costs and greater investmentsin specialized assets maximize transaction value,or the joint performance of the transactors.

m In support of this proposition, Smith and Aldrich ( 1991:28) found that asset specificity created trust by increasinginformation sharing between suppliers and buyers.

CONCLUSION

This study suggests that transaction costs do notnecessarily increase with an increase in relation-specific investments. Empirically we found thatJapanese automakers incur lower transaction coststhan U.S. automakers even though their suppliersare more specialized to them. Through explora-tory interviews we developed a number of prop-ositions to explain these results. In particular, wefound that transaction costs differ among auto-makers due to: ( 1) differences in their commit-

552 J. H. Dyer

ment to suppliers to engage in repeated exchange,(2) differences in the scale and scope ofexchanges between the supplier and automaker,(3) differences in interfirm information sharing,(4) differences in the mix of safeguards chosento govern the exchange, notably a reliance onself-enforcing safeguards which are efficaciousover an indefinite time horizon (as opposed tocontracts which are only effective for a finitetime horizon), and (5) differences in investmentsin cospecialized assets. We have suggested thatthese propositions are interrelated, meaning thatthey are highly correlated and mutually reinforc-ing. For example, engaging in repeated trans-actions (Proposition 1) is important for achievinghigh volume/scale economies (Proposition 2),sharing information (Proposition 3), developingself-enforcing safeguards (Proposition 4), and dis-covering new ways to enhance performancethrough specialized assets (Proposition 5). Wepresume that each proposition holds true if eachof the other propositions is held constant, but itmay be that in order to significantly reduce trans-action costs these actions must be taken in combi-nation.

These findings also have implications for howfirms in a production network can maximizetransaction value. In particular, a production net-work that can simultaneously achieve the twinbenejits of asset specialization and lower trans-action costs will have efficiency advantages overa less specialized network with higher transactioncosts. Williamson ( 1985: 1) has remarked that‘the transaction cost approach maintains that theseinstitutions [governance structures ] have the mainpurpose and effect of economizing on transactioncosts.’ Indeed, the fundamental governance ques-tion, as posed by Williamson, is: How canexchange relationships be structured to economizeon transaction costs?z 1 However, as Zajac andOlsen (1993 ) have suggested, this may be thewrong question. Instead, the fimdamental questionshould be: How can exchange relations be struc-

2I It is w.ofi noting that there are a number of transactioncost scholars who do not espouse the view that optimatgovernance is the one that minimizes transaction costs, butthat the institutional arrangement chosen is the one thatmaximizes the gains from trade. For example, Hennart ( 1993)argues that the efficient institutional arrangement is the onewhich results in the highest midual, a statement equivalent tosaying that the optimal institution is the one which maximizestransaction value.

tured to maximize transaction value (which

includes both production and transaction costs )?Our study suggests that beyond minimizing trans-action costs, governance influences transactionvalue by influencing the transactors’ set ofchoices regarding the level of specialized assetsthat will be employed. For example, a firm’spromise credibility (trustworthiness) can be asource of advantage because it minimizes trans-action costs. However, at least as important isthe fact that trustworthiness will often result inhigher levels of performance-enhancing invest-ments in specialized assets because trustworthytransactors: (1) engage in greater informationsharing which increases the probability that theywill discover new ways to enhance performancethrough relation-specific investments, (2) have alonger payback period during which to eam areturn on the investments, thereby increasing thenumber of relation-specific investments that makeeconomic sense, and (3) are able to implementoptimal levels of specialized assets because thecosts of safeguarding the relation-specific invest-ments are less likely to outweigh the gains. Thus,efficient governance mechanisms (i.e., must) cansimultaneously lower transaction costs andincrease relation-specific investments, thereby cre-ating competitive advantage.

However, it is important to note that Japaneseprocurement practices will only pay off if thereare gains to be obtained from making relation-specific investments. If parts manufacture wouldnot benefit from such investments, then U.S. prac-tices may be more efficient. Furthermore, buildingtrust and engaging in repeated trades is notcostless. At some point Japanese transactorshad to incur the ‘set-up costs’ associated withestablishing noncontractual governance mech-anisms. Moreover, the cost of building trustincludes the opportunity cost of not takingadvantage of one’s suppliers and the loss ofthe opportunity to use lower-cost suppliers ifthey came along. Thus, Japanese procurementpractices are not absolutely superior. They maybe superior only in the long term, and onlyif relation-specific investments create economicvalue. Hence, Japanese procurement practicesmay be optimal given a rather specific set ofconditions. However, when these conditionsexist (as they do in the auto industry), thepractices described in this paper represent amodel for effective interfirm collaboration.

Effective Inter&n Collaboration 553

ACKNOWLEDGEMENT

The Reginald Jones Center for Policy and Organi-zation at the Wharton School is gratefullyacknowledged for its support of this research.

REFERENCES

Alchian, A. A. and H. Demsetz (1972). ‘Production,information costs, and economic organization’, Amer-ican Econom”c Association, 62(5), pp. 777-795.

Akerlof, G. A. (1970). “he market for “lemons”:Quality, uncertainty, and the market mechanism’,Quarterly Journal of Economics, 84, pp. 488-500.

Argyle, M. ( 1991). Cooperation: The Basis of Socia-bility. Routledge, New York.

Asanuma, B. ( 1985). ‘The contractual framework forparts supply in the Japanese automotive industry’and ‘The organization of parts supply in theJapanese automotive industry’, Japanese EconomicStudies, 15 (Summer 1985a,b), pp. 32-78.

Axelrod, R. ( 1984). The Evolution of Cooperation.Basic Books, New York.

Bradach, J. L. and R. Eccles ( 1989). ‘Markets versushierarchies: From ideal types to plural forms’,Annual Review of Sociology, 15, pp. 97-118.

Browning, L., J. Beyer and J. Shetler ( 1995). ‘Buildingcooperation in a competitive industry: Sematech andthe semiconductor industry’, Academy of Manage-ment Journal, 38( l), pp. 113-151.

Clark, K. B. and T. Fujimoto ( 1991). Product Develop-men~ Pe#ormance. Harvard Business School Press,Boston, MA.

Cusumano, M. A. and A. Takeishi ( 1991). ‘Supplierrelations and management: A survey of Japanese,Japanese-transplant, and U.S. auto plants’, Stra~egicManagement Journal, 12(8), pp. 563-588.

Dore, R. (1983). ‘Goodwill and the spirit of marketcapitalism’, British Journal of Sociology, 34(4),pp. 459–482.

Dyer, J. H. ( 1996a). ‘Specialized supplier networks asa source of competitive advantage: Evidence fromthe auto industry’, Strategic Management Journal,17(4), pp. 271-292.

Dyer, J. H. ( 1996b). ‘Does governance matter? Keiretsualliances and asset specificity as sources of Japanesecompetitive advantage’, Organization Science, 7(6),pp. 649-666.

Dyer, J. H. ( 1996c). ‘How Chrysler created an Amer-ican keiret.su’, Harvard Business Review, 74(4),pp. 42-56.

Dyer, J. H. and W. G. Ouchi ( 1993). ‘Japanese stylebusiness partnerships: Giving companies a competi-tive edge’, Sloan Management Review, 35( 1),pp. 51-63.

Gerlach, M. L. ( 1992). Alliance Capitalism. Universityof California Press, Berkeley, CA.

Glaser, B. and A. Strauss (1%7). The Discovery ofGrounded Theory. Aldine, Chicago, LL.

Heide, J. B. and A. Miner ( 1992). ‘The shadow of thefuture: Effects of anticipated interaction and frequencyof contact on buyer–seller cooperation’, Academy ofManagement Journal, 35(2), pp. 265-291.

Helper, S. ( 1991). ‘How much has really changedbetween U.S. automakers and their suppliers’, SloanManagement Review, Summer, pp. 15-28.

Helper, S. and D. Levine ( 1992). ‘Long term supplierrelations and product-market structure’, Journal ofLaw, Economics, and Organization, 8, pp. 561-581.

Hennart, J.-F. ( 1993). ‘Explaining the swollen middle:Why most transactions are a mix of “Market” and“Hierarchy’”, Organization Science, 4(4), pp. 529–547.

Hill, C. W. L. ( 1995). ‘National institutional structures,transaction cost economizing, and competitiveadvantage: The case of Japan’. OrganizationScience, 6(2), pp. 119–1 31.

Kamath, R. R. and J. K. Liker ( 1994). ‘A second lookat Japanese product development’. Harvard BusinessReview, 72(6), pp. 154– 170.

Klein, B. ( 1980). ‘Transaction cost determinants of“unfair” contractual arrangements’, Amen”can ECO-

nomic Review, 70(2), pp. 356-362.Klein, B., R. G. Crawford and A. A. Alchian ( 1978).

‘Vertical integration, appropriable rents, and thecompetitive contracting process’, JOUITKZ1 of LZW

and Economics, 2 [, pp. 297-326.Kogut, B. ( 1989). ‘The stability of joint ventures:

Reciprocity and competitive rivalry’, Journal ofIndustrial Economics, 38, pp. 183-197.

Kreps, D. and R. Wilson ( 1982). ‘Reputation andimperfect information’, Journal of Economic Theory,27. Q~. 253-279.. . .

Lyons, B. R. (1994). ‘Contracts and specific invest-ment: An empirical test of transaction cost theory’,Journal of Economic and Management Strategy, 3,pp. 257-278.

Macneil, I. R. ( 1978). ‘Contracts: Adjustment of long-term economic relations under classicat and neo-classical, and relational contract law’, NorthwesternUniversity Law Review, 72, pp. 854-905.

Mahoney, J. T. ( 1992). ‘The choice of organizationalform: Vertical financial ownership versus othermethods of verticat integration’, Strategic Manage-ment Journal, 13(8 ), pp. 559-584.

Masten, S. E. ( 1984). ‘The organization of production:Evidence from the aerospace industry’, Journal ofLaw and Economics, 27, pp. 403-417.

Monteverde, K. and D. J. Teece (1982). ‘Supplierswitching costs and vertical integration in the auto-mobile industry’, Bell Journal of Economics, 13,pp. 206-213.

Nishiguchi, T. ( 1993). ‘Governing competitive supplierrelations: New auto-industry evidence’, MITIIMVPAnnual Sponsors’ Briefing Meeting, Cape Cod, MA.

Nishiguchi, T. ( 1994). Strategic Industrial Sourcing.Oxford University Press, New York.

North, D. C. ( 1990). Institutions, Instirurional Changeand Economic Pe@ormance. Cambridge UniversityPress, Cambridge, UK.

Ouchi, W. G. ( 1984). The M-Form Society. AvonBooks, New York.

.

554 J. H. Dyer

Parkhe, A. ( 1993). ‘Strategic alliance structuring: Agame theoretic and transaction cost examination ofinterfirm cooperation’, Academy of ManagementJOU171Ui, 36(4), pp. 794-829.

Perry, M. K. ( 1989). ‘Vertical integration’. In R.Schmatensee and R. Willig (eds.), Handbook ofIndustrial Organization. North-Holland, Amsterdam,pp. 185-255.

Sake, M. (1991). ‘The role of ‘Trust” in Japanesebuyer-supplier relationships’, Ricerche Economiche,45(2–3), pp. 449-474.

Sake, M. (1992). Prices, Quality, and Trust. CambridgeUniversity Press, Cambridge, U.K.

Schelling, T. C. ( 1978). Micromotives and h4acro-behavior. Norton, New York.

Smith, A. D. and H. E. Aldrich ( 1991). ‘The role oftrust in the transaction cost economics framework:Transaction contexts and governance structures inU.S. manufacturer-supplier relations’. Academy ofManagement Meetings, Miami, FL.

Sohn, J. H. D. ( 1994). ‘Sociat knowledge as a controlsvstem: A tmowsition and evidence from theJ~panese FD~ khavior’, Journal of International -Business Studies, Second Quarter, pp. 295-324.

Strauss, A. and J. Corbin (1990). Basics of QualitativeResearch. Sage, Newbury Park, CA.

APPENDIX 1: EXPLORATORY 2.INTERVIEW QUESTIONS

Automaker purchasing persomel3.

Telser, L. G. ( 1980). ‘A theory of self-enforcing agree-ments’, Journal of Business, 53(1), pp. 27-44.

Walker, G. and D. Weber ( 1984). ‘A transaction costapproach to make-or-buy’, Administrative ScienceQuarterly, 29, pp. 373-391.

Weigelt, K. and C. Camerer ( 1988). ‘Reputation andcorporate strategy: A review of recent theory andapplications’, Strategic Management Journal, 9(5),pp. 443-454.

Williamson, O. E. ( 1983). ‘Credible commitments:Using hostages to support exchange’, AmericanEconomic Review, 73(4), pp. 519-535.

Williamson, O. E. ( 1985). The Economic Institutionsof Capitalism. Free Press, New York.

Williamson, O. E. ( 1991a). ‘Comparative economicorganization: The analysis of discrete structural alter-natives’, Administrative Science Quarterly, 36,pp. 269-296.

Womack, J. P., D. T. Jones and D. Roos ( 1990). TheMachine rhat Changed the World. Harper Perenniat,New York.

Zaiac, E. J. and C. P. Olsen ( 1993). ‘From transaction

1.

2.

3,

4,

5.

Given your size and relative bargaining power,what prevents you from taking advantage ofsuppliers?What, if anything, increases (or decreases) the

4.

costs associated with selecting suppliers towork with?What, if anything, increases (or decreases)your costs of bargaining and contracting withsuppliers?What, if anything, increases (or decreases)your costs with regard to monitoring suppliers’

cost to transactional value analysis: Implications forthe study of interorganizational strategies’, Journalof Management Studies, 30(1), pp. 131– 145.

What, if anything, increases (or decreases)your costs of bargaining and contracting withthis particular automaker?What, if anything, increases (or decreases)your costs with regard to monitoring this auto-maker’s performance, to ensure that the auto-maker is living up to the original agreement?How do you resolve disputes with this auto-maker? (i.e., if either you or the automakerbelieve that the other party has not lived upto the original agreement?) For example, whatif the automaker’s projected unit volume (onwhich you base your unit price) does notmaterialize?

pfot-m~ce, tO ensure that they tie living uP APPE~lX z: SUPPLIER SURVEYto the original agreement?How do you resolve disputes with this sup-

QUESTIONS FOR TABLES 4-6

plier? (i.e~, if either you or the supplier believe Table 4hat the other party- has not lived up to the 1original agreement?)

Supplier saks/engineering personnel

1. What prevents you from giving inaccurate costestimates to this particular automaker? Whatprevents you from giving any other types ofinaccurate information to this particular auto-maker?

In your experience, what percent of the timedo you win business from one model to thenext when a model change is made byChrysler (stated another way, in the past whenyou have been awarded the contract fromChrysler for a specific model, what percentageof the time do you re-win the business whena model change is made)?

percent of the time a supplierwill m-win the business.

.

E@ective Interfirm Collaboration 555

2.

3.

4.

In your experience, what percent of the timedoes Chrysler rebid your business to othersuppliers during the life of a model? (e.g.,between one model change and the next)

percent of time automaker rebidsyour business before model change.a. Approximate total 1991 automotive sales?

b. Approximate percent of total automotivesales to Chrysler?Approximately how many different types ofproduction part numbers (not including service

parts) do you currently supply to Chrysler?production part numbers

out, new machines or technologies, valueanalysis/value engineering, etc. ) which haveallowed you to lower manufacturing costs?

Not To Some Very Great-at Extent ExtentAll1 234s 6 7

5. To what extent has Chrysler provided assis-tance to help you reduce defects and increasethe overall reliability and quality of the prod-ucts you sell to Chrysler?

Not To Some Very Greatat Extent Extent

Table 5 All1 2345 6 7

Information sharing6. To what extent has Chrysler provided assis-

1. To what extent do YOU trust that tance in developing a ‘Just-In-Time’ invento~

confidential /momietarv information shared management system designed to lower inven-

with Chryslerby Chrysler’s

NotatAll

1

2.

23

~ill be ‘kept strictly confidential tory costs and/or make delivery more efficient.

buyers and engineers?Not To Some Very Great

To Some Ve~ Great at Extent Extent

Extent Extent All1 2345 6 7

4 5 6 7

To what extent have You urovided recent ‘able 6detailed cost data to Ch-&ler’ (e.g., a break-down of your cost structure which estimatesexactly what it will cost you to manufacturea specific component)?

Not To Some Very Greatat Extent ExtentAll1 2345 6 7

3. To what extent do you share information withChrysler on your long-term production plans,capital investments, and capacity utilization?

Not To Some Very Greatat Extent ExtentAll1 2345 6 7

4. To what extent has Chrysler provided techni-cal, engineering, or other assistance in the pastwhich has allowed you to make changes inyour manufacturing processes (e.g., plant lay-

.-

Self-enforcing safeguards

1. To what extent do you trust Chrysler personnelto deal with you fairly?

Not To Some Very Great

at Extent Extent

All1 2345 6 7

2. If given the chance, to what extent will thiscustomer take unfair advantage of your busi-ness unit?

Not To Some Ve~ Great

at Extent Extent

All1 2345 6 ‘7

3. To what extent has Chrysler developed a repu-tation for fairness and trustworthiness amongthe supplier community?

556 J. H. Dyer

Not To Some Ve~ Great (If yes)

at Extent Extent b, What percent of your stock is owned byAll Chrysler?1 2345 6 7

Note: Each supplier was given a questionnaire which focused

4. a. Does Chrysler own any stock in your com- on a particularautomaker(the examplequestionsaboveare from

pany? Yes No—— the questionnaireto Chrysler suppliers).

.