Interorganizational Trust, Governance Choice, and Exchange ... · allow us to analyze...

OrganizationScienceArticles in Advance, pp. 1–21issn 1047-7039 !eissn 1526-5455

informs ®

doi 10.1287/orsc.1070.0345©2008 INFORMS

Interorganizational Trust, Governance Choice, andExchange Performance

Ranjay GulatiKellogg School of Management, Northwestern University, Evanston, Illinois 60208,

[email protected]

Jack A. NickersonJohn M. Olin School of Business, Washington University in St. Louis, St. Louis, Missouri 63130,

[email protected]

This paper looks at when and how preexisting interorganizational trust influences the choice of governance and inturn the performance of exchange relationships. We theorize that preexisting interorganizational trust complements

the choice of governance mode (make, ally, or buy) and also promotes substitution effects on governance mode choicewhile impacting exchange performance. We evaluate hypotheses using a novel three-stage switching regression model anda sample of 222 component-sourcing arrangements of two assemblers in the automobile industry. Analysis of our databroadly supports our hypotheses. High levels of preexisting interorganizational trust increased the probability that a lessformal, and thus less costly, mode of governance was chosen over a more formal one. This finding suggests a substitutioneffect of interorganizational trust on governance mode choice that in turn shapes exchange performance. We also found acomplementary effect of trust on performance: Regardless of the governance mode chosen for an exchange, trust enhancedexchange performance. Additional evidence of the complementary effect of trust on performance was that trust somewhatreduced interorganizational conflict.

Key words: interorganizational trust; governance choice; exchange performance; interorganizational relations; transactioncost economics

History : Published online in Articles in Advance.

A growing debate in the strategic management, orga-nization theory, and contracting literatures concerns therole of interorganizational trust in determining gover-nance choice and performance in exchanges. Broadlydefined, interorganizational trust is an organization’sexpectation that another firm will not act opportunis-tically (Bradach and Eccles 1989). Some scholarsstudying interorganizational trust argue that “trust cansubstitute for hierarchical contracts in many exchangesand serves as an alternative control mechanism” (Gulati1995, p. 93). The implication is that if trust existswhen firms enter an exchange relationship, they mayuse less formal modes of governance, and therefore pre-existing trust enhances exchange performance. A richbody of research suggests that trust may substitutefor formal governance if the cooperative behavior thattrust generates offers a less costly and more effectivesafeguard than complex contracts or vertical integra-tion (e.g., Bradach and Eccles 1989, Dyer 1997, Gulati2007, Lincoln and Gerlach 2004, Nooteboom et al.1997, Zaheer and Venkatraman 1995). Substitution oftrust for formal governance may also arise if the useof contracts “crowds out” the use of trust in gover-nance (e.g., Frey 1997). Indeed, the use of a contractmay signal distrust, thereby undermining the develop-ment of relational exchange and perhaps encouraging

opportunistic behavior between exchange partners (Fehrand Gachter 2000, Ghoshal and Moran 1996, Malhotraand Murnighan 2002, Sitkin and Roth 1993).Focusing on the performance implications of interor-

ganizational trust, scholars also have argued thatpreexisting trust between firms entering exchangerelationships benefits those relationships regardless ofthe chosen governance structure. This occurs becausetrust reduces transaction costs and facilitates coordina-tion (Aulakh et al. 1996, Dodgson 1993, Gulati andSingh 1998, Zaheer and Venkatraman 1995, Zaheer et al.1998). Along these lines, some researchers have pro-posed that formal governance and relational governance(via trust) can act as complements (Poppo and Zenger2002) to enhance exchange performance (Mayer 1999).Some have even argued that formal governance is neces-sary for trust, implying that high levels of exchange per-formance cannot be achieved by trust alone (Lazzariniet al. 2004). This body of research broadly echoesNorth’s (1990, p. 46) assertion that “formal rules cancomplement and increase the effectiveness of informalconstraints.”In contrast to these arguments, our view is that the

question is not whether trust is a substitute or comple-ment to formal governance, but rather when and howit may serve as both simultaneously. In this paper, we

1

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.

Published online ahead of print April 25, 2008

Gulati and Nickerson: Interorganizational Trust, Governance Choice, and Exchange Performance2 Organization Science, Articles in Advance, pp. 1–21, © 2008 INFORMS

develop a theory about (1) when and how dyadic pre-existing interorganizational trust—the level of trust thatexisted prior to a dyad’s current exchange—leads to thesubstitution of one governance mode for another (and tosubsequent performance effects); and (2) when and howinterorganizational trust complements any formal gov-ernance mode in enhancing exchange performance. Wedevelop our theory by examining the three prototypicalgovernance choices for dyadic exchanges, “buy,” “ally,”and “make,” which Williamson (1991) called “market,”“hybrid,” and “hierarchy.” In the buy mode, an organi-zation procures an input from another using simple con-tracts. In the ally mode, one organization enters a formalarrangement with another using a complex (yet incom-plete) contract to procure the input. In the make mode,one organizational unit procures from another unit that ispart of the same broader organization.We theorize that trust between organizations at the

time they enter an exchange relationship can lead to thesubstitution of a less formal governance mode (e.g., buy)for a more formal mode (e.g., ally). In other words, part-ners are likely to use less formal governance when thereis trust between them. In Williamson’s (1991) parlance,preexisting trust is a “shift parameter” whose magni-tude may affect choice of governance mode, which mayenhance exchange performance by allowing use of less-costly governance structures. A key contribution of ourpaper is to increase understanding of trust as a shiftparameter for governance modes. The substitution effectarises because trust can mitigate some of the contract-ing hazards associated with exchange, which reducesthe need for more formal controls. Such substitutionsare particularly likely to occur when the exchange haz-ard is near certain “critical values”—the level of assetspecificity without trust, when buy and ally are closesubstitutes or ally and make are close substitutes. Insuch situations, trust tips the balance toward less formalmodes.We further argue that preexisting trust can complement

any mode of governance and thus improve exchange per-formance whenever contracting hazards are present byreducing both instances of conflict and the costs of res-olution. Because contracting hazards are present in allmodes of governance—including intrafirm exchange—we propose that preexisting interorganizational trust hasa complementary effect in all modes. Thus, our theorysuggests that preexisting trust simultaneously has a sub-stitution effect on governance mode and a complemen-tary effect on exchange performance.To examine our propositions, we use a novel data set

of 222 sourcing arrangements from two U.S. auto assem-blers. Our data, drawn from a comprehensive survey,allow us to analyze interorganizational trust, governancechoice, and perceived performance in a single statisti-cal framework. We use a three-stage switching regres-sion model to account for unobserved heterogeneity

(e.g., Hamilton and Nickerson 2003). This analysis is thefirst to empirically assess endogenous governance modechoice and concomitant performance implications, usingthe tripod of alternative governance modes.

Theory and HypothesesInterorganizational TrustTrust is a complex topic that has been examined througha wide array of disciplinary lenses (for a review ofthis diverse literature, see Gambetta 1988, Kramer 1999,Rousseau et al. 1998). In recent years, the inher-ently interpersonal notion of trust has been extendedto organizations. One premise of this extension is thatinterorganizational trust is linked to the predictability ofa partner firm’s behavior toward a vulnerable focal firm.If the partner firm fulfills positive expectations, the focalfirm develops greater confidence in the partnership, andthis confidence in turn mitigates future concerns aboutopportunism (Gulati 1995, Hill 1990, Parkhe 1993,Nooteboom 1996, Nooteboom et al. 1997, Zaheer andVenkatraman 1995). Over time, even though turnover ofworkers and managers may occur at both firms, insti-tutionalizing processes crystallize expectations, and newboundary spanners are socialized to accept the firmwideexpectations for this partner’s behavior (Zaheer et al.1998, Zucker 1987).We focus on the effects of interorganizational trust

existing prior to a current exchange and its effectson governance choice and performance in the currentexchange. In particular, we look at the trust engenderedby past interactions that generate trust by shaping expec-tations of subsequent behavior (Gulati 1995, 2007). Forinstance, over time a firm may avoid partners who havedemonstrated a propensity for opportunistic behaviorand retain those who have not. In this way, recurrentinteraction with a partner leads to an expectation ofbehavior superior to that of an untested partner. Pastinteractions also engender social processes (Granovetter1985, Zucker 1987) and interaction patterns (Blau 1964)that scholars have identified as important foundations oftrust (Dore 1983; Gulati 1995, 2007; Gulati and Sytch2008; Heide and Miner 1992; MacNeil 1980; Jones et al.1997).1 Our focus on preexisting trust also allows us toconsider the role of such trust in governance structurechoice at the time an exchange is formalized, as wellas trust’s effect on the subsequent performance of theexchange.

A Baseline Theory of Governance Choice andExchange PerformanceGovernance cost is a class of transaction cost thatrefers to (1) the costs of drafting, negotiating, and safe-guarding an agreement between two or more actorsand (2) the ex post costs of contracting, generated

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.

Gulati and Nickerson: Interorganizational Trust, Governance Choice, and Exchange PerformanceOrganization Science, Articles in Advance, pp. 1–21, © 2008 INFORMS 3

by maladaptation, haggling, administration, and bond-ing (Williamson 1991).2 Lower governance costs implythat exchange partners make adjustments and adapta-tions at low cost. Governance costs have a critical impacton expected exchange performance because performancelevels increase when governance costs are lower, all elseheld equal.To examine the impact of governance costs on per-

formance, transaction cost theorists largely focus on thedichotomy of make versus buy, and practically all extantempirical work examines this dichotomy. More recenttheory adds ally as a governance mode (Bradach andEccles 1989, Williamson 1991, Gulati 1995), but fewempirical studies address this expanded set of modes.Figure 1 graphically represents the traditional transactioncost theory predictions that we use to describe our base-line argument and also use for expositional purposes.Let M!k " #$, X!k " #$, and H!k " #$ in Figure 1 repre-sent expressions of the expected governance costs asso-ciated with buy, ally, and make modes of governance, asa function of asset specificity (k) and a vector of shiftparameters (#) (Williamson 1991). We set aside theseshift parameters for the moment, but will return to themsubsequently. Williamson also identified two additionaltransaction attributes—uncertainty and frequency—thatinteract with asset specificity in determining the mag-nitude of exchange hazards. In this paper, we suppressconsideration of uncertainty and frequency to focus onWilliamson’s “main locomotive,” asset specificity. Doingso eases exposition without sacrificing generality.Williamson asserted that the shape of the gover-

nance cost curves are such that M!0$ < X!0$ < H!0$and %M/%k > %X/%k > %H/%k > 0 and, importantly,that these curves intersect so that M!k1$ = X!k1$ andX!k2$ = H!k2$. These assertions imply that procure-ment through simple contracts (buy) is the least costlymode of organization (i.e., offers superior performance)for low levels of asset specificity (i.e., k < k1); thatprocurement through internal production (make) is the

Figure 1 Governance Costs as a Function of Asset Specificity

Gov

erna

nce

cost

0 k1 k2 k

Asset specificity

M(k)

X(k)H(k)

Buy

Ally

Make

least costly mode of organization for high levels ofasset specificity (i.e., k > k2); and that procurementthrough complex contracts (ally) is the least costly modeof organization for moderate levels of asset specificity(i.e., k1 < k < k2). The intersections of the curves Mand X, denoted by k1, and the curves X and H , denotedby k2, identify the critical values of asset specificityat which the transaction-cost economizing governancechoice shifts from one mode to another. At these crit-ical values, the adjacent governance modes are nearlyequally efficient because they incur equivalent gover-nance costs.In the following sections, we argue that preexisting

interorganizational trust acts as a shift parameter forthe three governance cost curves. Investigating how pre-cisely preexisting trust affects the curves for buy, ally,and make allows us to predict how trust affects bothgovernance choice and exchange performance.

Trust and the Buy Mode of GovernanceAccording to a substantial literature in economics andsociology, trust lowers transaction costs in all kindsof exchange relationships in which a risk of oppor-tunism is present (e.g., Bromiley and Cummings 1995;Bradach and Eccles 1989, p. 104; Nooteboom 1996;Nooteboom et al. 1997). Bradach and Eccles (1989),for instance, argued that transactions are rarely governedsolely by the market. Along similar lines, Arrow (1974,p. 23) maintained that “trust is an important lubricantof the social system” that adds efficiency to many eco-nomic exchanges. Supporting these statements, empiri-cal research has shown trust to be an important elementof market exchanges in industries as varied as residentialhomebuilding (Eccles 1981), banking (Eccles and Crane1988), electrical and electronic components (Nooteboomet al. 1997), and textiles-clothing (Mariotti and Cainarca1986), and that businesspeople often rely on trust evenwhen a transaction involves exposure to exchange haz-ards (Macaulay 1963, p. 58). Prior research suggests thattrust between firms involved in an exchange is likely toincrease confidence in, and positive expectations about,each partner’s behavior, reducing the need for controlthrough formal governance mechanisms. A buy modeof governance may therefore be chosen even whensome level of exchange hazard is present, and trust willlower expected governance costs. Thus, the efficiency-enhancing, and hence cost-reducing, effects of trust canincrease exchange performance in the buy mode ofgovernance when exchange hazards are present (Gulati1995, 2007; Hill 1990; Parkhe 1993).In this context, we turn to mapping the effect of pre-

existing trust on the governance cost curve for buy (seeFigure 1). Preexisting trust lowers expected maladapta-tion and haggling costs because exchange partners aremore likely to avoid disputes or resolve them quickly.Under the buy mode of governance, inconsequential

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


setup and running costs and no bonding costs areincurred, so these costs need not be considered. Trusthas no effect on the buy governance cost curve whenexchange hazards are absent (k = 0), but otherwiseshifts governance costs downward (i.e., when k > 0). Asexchange hazards grow, trust reduces costs even morebecause avoiding or quickly resolving disputes, whichtrust facilitates, has greater economic significance forlarger contractual hazards. Thus, the greater the expectedtrust, the lower the governance costs.We identify preexisting trust as a particular shift

parameter (&) that relocates the governance cost curvefor the buy mode of governance, M!k' &$. We assumethat the presence of trust implies that & > 0 (althoughour theory may apply equally well to distrust, whichwould imply & < 0). Codifying this relationship interms of our buy governance cost curve implies thatM!0'0$ = M!0' &$ "& > 0, which indicates that trustdoes not lower governance costs when exchange haz-ards are absent (i.e., k = 0) and 0 < %M!k' &$/%& <%M!k'0$/%& "k > 0, & # 0, which indicates that thecost curve for the buy mode of governance with trust isflatter and below the cost curve for buy without trust.Figure 2 depicts this relationship between M!k'0$ andM!k' &$.

Trust and the Ally Mode of GovernanceTrust also shifts the ally governance costs curve, but toa lesser extent than it moves the buy curve. We proposethat trust influences expected governance costs in thecontext of complex contracts—the ally mode. As men-tioned earlier, in the presence of interorganizational trustamong alliance partners, disputes arise less frequentlyand, when they do, are often resolved by the partiesthemselves without recourse to third parties, bureau-cratic policies, or written contracts. Trust is also asso-ciated with sustainable mutual interdependence betweenexchange partners in an atmosphere of bilateral com-mitment, making relationships more flexible and adap-tive to unforeseen contingencies (Dwyer et al. 1987,

Figure 2 Governance Costs as a Function of Asset Specificity

Gov

erna

nce

cost

Asset specificity

M(k,0)

X(k,0)

H(k,0)M(k, !) X(k, !)

H(k, !)

0 k1 k1! k2

!k2 k

Gulati and Singh 1998, Gulati and Sytch 2007). Fur-thermore, trust complements formal governance in allymodes by attenuating the expectation of opportunisticbehavior and facilitating partner coordination (Barneyand Hansen 1994; Gulati 1995, 2007; Mayer 1999).Although trust can facilitate adjustments with a com-

plex contract, the contract nonetheless constrains theefficacy of trust for making such adjustments. Macaulay(1963) described complex contracts as frameworks foradjustment.

At times relatively contractual methods are used to makeadjustments in ongoing transactions and to settle dis-putes. Demands of one side which are deemed unreason-able by the other occasionally are blocked by reference tothe terms of the agreement between the parties. The legalposition of the parties can influence negotiations eventhough legal rights and litigation are never mentioned intheir discussions. (p. 64)

The fact that partners can appeal to the contract there-fore shapes and limits the range of feasible adjust-ments even when trust is present. Indeed, many scholars(e.g., Fehr and Gachter 2000, Ghoshal and Moran 1996,Lubell and Scholz 2001, Malhotra and Murnighan 2002,Sitkin and Roth 1993, Tenbrunsel and Messick 1999)have asserted that complex contracting may limit theeffectiveness of trust and potentially dissipate it. Thiscan happen if actors attribute cooperation to contractualconstraints rather than to a partner’s goodwill (Malhotraand Murnighan 2002) and adopt a calculative rather thana moral or ethical frame of reference (Tenbrunsel andMessik 1999), increasing actors’ proclivities for oppor-tunistic and other trust-damaging pursuits (Ghoshal andMoran 1996). Indeed, Macaulay (1963, p. 64) noted that“businessmen object that in [complex contracts] one getsperformance only to the letter of the contract. Such plan-ning indicates a lack of trust and blunts the demands offriendship, turning a cooperative venture into an antag-onistic horse trade.” Per this description, complex con-tracts truly constrain the efficacy of trust for makingadjustments.Preexisting trust may expand the scope of adjustments

that exchange partners are willing to embed in a con-tract or leave out of it, choices that in turn influencetheir expectations (Dwyer et al. 1987, Macaulay 1963,MacNeil 1983). In other words, preexisting trust mayexpand the scope of the contract or the set of adjust-ments perceived as reasonable (Larson 1992, Lorenz1988, Luhmann 1979). At the same time, the detailedand sophisticated contracts typical of the ally mode canlimit the scope of adjustment available for exchangepartners. Such contracts often either set explicit expec-tations or specify permissible adjustments, narrowingthe window for spontaneous relational adjustment thattrust can open. In contrast, simple contracts that do lit-tle to set expectations or specify adjustments offer nosuch constraint. Additionally, because parties tend to

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


develop deep-rooted patterns of interdependence in allyrelationships (Dwyer et al. 1987), a fear of betrayal thatmay trigger sizeable losses additionally constrains trust-driven adjustment (Nooteboom 1996).Thus, although trust can reduce governance costs in

the ally mode, the complex contracts associated with thisstructure constrain its benefits. Therefore, at best, trustmay have the same effect on the ally mode as it doeson the buy mode, at least for low levels of exchangehazards. However, our discussion suggests that the effectof trust, although significant, should be smaller for allythan for buy. As in the buy mode of governance, inthe ally mode trust is expected to (1) have no effecton governance costs when exchange hazards are absent(therefore, X(0'0)=X(0' &) "& # 0) and (2) lower thegovernance cost associated with the ally mode wheneverexchange hazards are present. We conclude that preexist-ing trust lowers expected governance costs more for buythan for ally. Thus, %M!k' &$/%& < %X!k' &$/%& < 0"k > 0, & > 0.

Trust and the Make Mode of GovernanceSome scholars assume that trust affects only marketmodes of organization (Chiles and McMackin 1996), yetearlier studies suggest that trust can matter in hierar-chical exchanges as well (e.g., Perrow 1986). Bradachand Eccles (1989, p. 107), for instance, argued that“trust * * *plays an important role in intrafirm interac-tions.” We expressly assume that interorganizational trustcan exist between distinct organizations within a firm,such as divisions and business units, that are engagedin interfirm exchange (e.g., component procurement).Although scholars have identified antecedents of trustin dyadic relationships between senior managers (e.g.,Becerra and Gupta 2003), very few have investigated theeffects of trust on the governance costs or performanceof intrafirm exchanges (Eccles and White 1988, Gulatiet al. 2005). More generally, a substantial literature insociology and psychology examines the effects of trustin organizations, but most of this research looks at trustwithin groups or between subordinates and superordi-nates (e.g., Dirks and Ferrin 2001).Despite the dearth of prior theoretical and empiri-

cal work, we expect preexisting trust to affect gov-ernance costs and performance in the make mode aswell. As argued above for the buy and ally modes, dis-putes should arise less frequently in the presence oftrust, and when they do, they should more often beresolved by the exchange partners themselves, withoutrecourse to authority. We therefore expect interorganiza-tional trust to have no effect when hazards are absentfrom an exchange (i.e., H (0'0)=H (0' &) "& > 0$ andto lower the governance cost associated with the makemode whenever exchange hazards are present, so that0< %H!k' &$/%k < %H!k'0$/%k "k > 0, & > 0.

However, a critical question remains: Does trust lowergovernance costs in the make mode more than in theally mode, all else being equal? In the make mode,bureaucracy, administrative controls, and authority shapeorganizational behavior and limit the ability of exchangepartners to make adaptations other than those prescribedby management. Thus, the lubricating effect of trust islikely to be more limited in this context. The constraintsimposed by governance in the make mode may also bemore limiting than those in an ally mode because the for-mer relies on authority as the ultimate means of resolv-ing disputes, so any agreement can be overruled by fiat,whereas the latter relies on the courts, which have con-tracts as guides to resolving disputes.Furthermore, anecdotal evidence suggests that man-

agement may sometimes intentionally propagate conflictsamong organizational units. By so doing, managementmay not only obtain otherwise unavailable information,but also encourage units to monitor each other’s perfor-mance. Although interunit monitoring potentially bene-fits overall organizational performance (Rotemberg andSaloner 1995), the resulting tension among units cer-tainly further constrains the role of trust in shapingthe performance of hierarchically organized exchanges.This discussion suggests that although trust lowers gov-ernance costs for intrafirm exchanges, it lowers gover-nance costs for ally even more, so that %X!k' &$/%& <%H!k' &$/%& < 0 "k > 0, & > 0.

Predicting Governance Choice andExchange PerformanceNow we use our theoretical discussion to generate pre-dictions for the effects of trust on governance choice andexchange performance. We expect preexisting interorga-nizational trust to substitute for formal governance byenabling the use of less-formal governance modes andto complement governance by enhancing performance inall modes. Our theory suggests that trust lowers gover-nance costs for all modes of governance, but more forbuy than for ally and more for ally than for make. Wehave argued that this differential exists because formalgovernance shapes behavior, constraining what trust cando. These constraints are greatest in the make mode,least in buy, and intermediate in ally.This differential lowering of governance costs has

two effects on governance choice: (1) preexisting trustextends the range of asset specificity within which thebuy mode offers efficient governance and (2) trust inconjunction with the ally mode becomes the efficientchoice for some range of asset specificity within whichmake was previously the most efficient choice. Preex-isting trust shifts the critical values of k1 and k2 to theright, toward higher levels of exchange hazards. Per ourprior arguments, these shifts also would lead to greatersubstitution effects, but the substitution of buy for ally

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


and of ally for make would occur only near the criticalvalues k1 and k2.Figure 2 shows these effects, graphing the expected

baseline governance cost curves for buy, ally, and make,along with expected cost curves when preexisting trustis present (when # = & > 0), and illustrating thatpreexisting interorganizational trust shifts the reduced-form equations from M!k'0$, X!k'0$, and H!k'0$ toM!k' &$, X!k' &$, and H!k' &$. The substitution effectof interorganizational trust on governance costs can beseen in Figure 2 because the critical value of k1—where M!k'0$ and X!k'0$ intersect—shifts to the rightto k&1—where M!k' &$ and X!k' &$ intersect. Similarly,the critical value of k2—where X!k'0$ and H!k'0$intersect—shifts to the right to k&2—where X!k' &$ andH!k' &$ intersect. This figure shows that trust enablesbuy to substitute for ally between k1 and k&1 and for allyto substitute for make between k2 and k&2 . We predictthat trust encourages the use of less-formal arrangementsfor certain ranges of exchange hazards. Our hypothesesmake clear the expected substitution effects of trust:

Hypothesis 1A. The greater the level of preexistinginterorganizational trust, the greater the exchange haz-ard that can be governed by a buy instead of an allymode of governance.

Hypothesis 1B. The greater the level of preexistinginterorganizational trust, the greater the exchange haz-ard that can be governed by an ally instead of a makemode of governance.

In addition to having these substitution effects ongovernance modes, trust complements them by lower-ing their costs, thus enhancing exchange performanceregardless of mode. Trust lowers expected governancecosts over all three modes whenever exchange hazardsare present. Because lower costs translate into higherperformance, trust enhances exchange performance in allthree modes of governance. Put differently, trust alwayscomplements buy, ally, and make with respect to per-formance because it facilitates making adjustments thatlower costs. This assertion resonates with Poppo andZenger’s (2002) conclusion that trust complements gov-ernance. Interorganizational trust thus causes two inter-related effects: It has a substitution effect with respectto governance by affecting the choice of mode for agiven level of asset specificity, while at the same time itlowers governance costs. Both of these effects improveexchange performance.Notice in Figure 2 that the minimum expected gov-

ernance cost scribed out by M!k' &$, X!k' &$, andH!k' &$ is lower at all points than the one scribed out byM!k'0$, X!k'0$, and H!k'0$, except when no hazardis present. This complementary effect lowers governancecosts when hazards are present, which translates intoa positive effect on exchange performance. The higher

the preexisting interorganizational trust, the lower theexpected governance costs, all else being equal. It is pos-sible that such lower costs could lead to increases on themargin in asset-specific investments, but such a substi-tution is not likely unless significant gains are expected.Note also, however, that the expected governance costreductions from preexisting interorganizational trust aregreatest for buy and least for make for every level ofasset specificity greater than zero, as we suggested inour earlier theoretical discussion. That is, the strength oftrust’s complementation differs systematically for buy,ally, and make:

Hypothesis 2A. The higher the level of preexistinginterorganizational trust, the lower the expected gover-nance cost for exchanges organized as buy, ally, andmake, all else being equal.

Hypothesis 2B. Preexisting interorganizational trusthas a greater effect on lowering expected governancecost for exchanges organized as buy than as ally, all elsebeing equal.

Hypothesis 2C. Preexisting interorganizational trusthas a greater effect on lowering expected governancecost for exchanges organized as ally than as make, allelse being equal.

MethodSample and Data CollectionData came from a comprehensive 1995 survey of leadcomponent buyers at the Ford Motor Company and at theChrysler Corporation. The sampling frame consisted ofall commodities that go into the assembly of an automo-bile; this methodology offers an advantage over automo-tive studies with more restricted samples of exchanges(e.g., Walker and Weber 1987). Drawing on a previ-ous study of the automobile sector (Monteverde andTeece 1982) and discussions with industry informants,we listed 120 components that go into most automo-biles. The comprehensiveness of this list was verifiedwith several industry executives and also against com-ponent lists the firms used to monitor parts’ quality. Foreach component, senior managers at the two automo-bile assemblers supplied the names of the buyers whooversaw sourcing. Additionally, the controller’s office ineach company verified the expert status of each surveyrespondent. The approach of contacting the most knowl-edgeable informant is consistent with prior research inother contexts (e.g., Heide and John 1990, Venkatramanand Grant 1986, Walker and Poppo 1991). The sampleincluded both internal and external sourcing relation-ships. The unit of analysis for this study was existingcomponent exchange ties, with each survey respondentproviding data on the component itself and on its com-pany’s two largest suppliers (or one supplier, if only

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


one existed) of that component. Our unit of analysisimproves on several prior studies of governance choicein the auto industry (Masten et al. 1989, Monteverde andTeece 1982) because of our unique focus on individualtransactions. This focus, in turn, enables us to conduct afine-grained analysis of governance choice and its con-sequences in each transaction.We complemented our survey with 37 in-depth inter-

views in two waves (16 at Chrysler, 21 at Ford)with individuals responsible for sourcing internally andexternally. The initial interviews were exploratory andopen-ended and were intended to clarify the procure-ment relationships. In later interviews, we sought infor-mation relevant to our central constructs of trust andthe performance of different governance modes. Wepretested the questionnaire with executives at the par-ticipating companies to remove ambiguities and confirmthe face validity of the measures.Survey implementation involved several standard steps

for ensuring a good response rate (Fowler 1993). Sixty-four buyers responded from Ford, and 67 from Chrysler,numbers representing response rates of 53% and 56%,respectively, and an overall response rate of 55%. Afterexclusion of 40 entries with missing data, our sampleconsisted of 222 existing sourcing arrangements, someinternal and some external.We examined nonresponse bias by comparing the

characteristics of the components for which responseswere received with the characteristics of those for whichwe received none. We looked at two key character-istics of commodities identified in prior research—type of sourcing and engineering complexity—and usedMonteverde and Teece’s (1982) ratings of these charac-teristics as a basis for this comparison. A Kolmogorov-Smirnov two-sample test to assess the probability ofdifferences in the distribution of respondents and nonre-spondents for these two variables (Siegel and Castellan1988) indicated that respondents and nonrespondentscame from the same population.To ensure that some of our self-reported measures of

constructs like performance did not suffer from seriousperceptual biases, we conducted some additional analy-ses for some of our survey measures. Given that exactperformance data were inaccessible to us for confiden-tiality reasons, we asked our contacts at each firm tocorrelate scores on four measures from a random sub-sample of 20% of valid survey responses with objective,internally tracked indicators of these measures from theyear before the survey. This was done for the past target-price ratio, defect rate (inverted), change in componentprice (inverted), and improvement in defect rate. Theobtained correlation coefficients of over 0.90 attested tothe reliability of the survey measures.

Measures

Dependent Variables. For Trust, our first dependentvariable, we used three items, listed in Table 1, that

reflect the three elements that define interorganizationaltrust (Cummings and Bromily 1996, Dyer 1997, Zaheeret al. 1998, Zaheer and Venkatraman 1995). Zaheer et al.(1998) used two additional questions to define interorga-nizational trust. We omitted these because they representantecedents of preexisting trust.3 Each question used aseven-point Likert scale ranging from strongly disagreeto strongly agree. We evaluated the reliability of ourdependent variable by estimating Cronbach’s alpha coef-ficient (Nunnally 1978) for the battery of three items;this alpha was 0.80.Mode, our second dependent variable, identified how

an exchange to procure a component was organized. Weclassified as make those exchanges with a division ofa respondent’s own company, as ally those exchangescharacterized by long-term complex contracts, and asbuy those exchanges characterized by short-term (i.e., ayear or less) contracts and competitive bidding. Alongwith providing these three choices of governance mode,a clear definition of each was provided to the respondentso they could make an informed selection. Respondentswere asked to identify which of these three categoriesgoverned a relevant exchange. Of the 222 exchanges onwhich information was complete, 24 were organized asmake, 126 were organized as ally, and 72 were orga-nized as buy. MODE was an ordered categorical vari-able, coded zero for buy, one for ally, and two for make.We assumed that lower governance costs correspond

to higher economic performance. Because firms keeparchival figures on the economic performance of ex-changes confidential, we based our Performance mea-sure on 13 questions, derived from a literature survey,fieldwork, and our pretests, concerning assembler sat-isfaction. To measure assembler satisfaction, we tappedthe survey respondents’ opinions of a component sup-plier’s attractiveness; as noted, respondents were compo-nent buyers employed by the assemblers. All questions(displayed in Table 2) employed a seven-point Likertscale. The first 10 questions reflected the respondent’sview of how attractive a supplier was compared with thebest alternative supplier for the commodity in question(1, much less attractive; 7, much more attractive). Thenext three questions pertained to the frequency of buyer-supplier disagreement and the difficulty of negotiationsover sharing the financial burden of engineering changerequests and raw material cost increases (1, fairly easy,to 7, very difficult).Unlike some prior studies that have assessed exchange

performance as subjective satisfaction only (e.g., Gulatiet al. 2005), we looked at both subjective and quasi-objective measures of performance. The latter werescaled measures of average past target ratio, average pastprice change rate, average defect rate, and improvementin average defect rate. We used factor analysis to identifyrelevant performance dimensions based on the above-mentioned comprehensive list of survey questions on

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table 1 Measures and Survey Questions

Survey questions

Dependent variables

Trust A constructed scale$ measuring the opinion of the buyer about the supplier compared with the bestalternative supplier for this commodity. The three questions concerned whether the supplier hasalways been evenhanded in its negotiation with your company, this supplier may use opportunitiesthat arise to profit at your expense, and you trust this supplier to treat you fairly (!= 0"80)

Mode 0 if Buy, 1 if Ally, 2 if Make.Performance A constructed scale$ measuring the opinion of the buyer about the supplier compared with the best

alternative supplier for this commodity. The 10 questions concerned the supplier’s pricecompetitiveness, support and services, flexibility in production, product quality, product innovations,overall performance, average past target ratio, average past price change rate, average defect rate,and improvement in average defect rate. (!= 0"93)

Conflict A constructed scale$ measuring the opinion of the buyer about the supplier compared to the bestalternative supplier for this commodity. The three questions concerned the supplier’s frequency ofsignificant disagreements, the ease of negotiation over sharing cost-engineering changes, and theease of negotiation over sharing cost-material cost increases. Higher levels of this construct equateto higher levels of conflict (!= 0"78).

Covariates

Experience: Based on past experience, you cannot with complete confidence rely on n this supply to keep promisesmade to you. (7-point Likert scale.)

Component History Logarithm of the number of years assembler has purchased component from the supplier.Organization History Logarithm of the number of years assembler has purchased any component from the supplier.Buyer Tenure Logarithm of the number of years that the assembler’s buyer has been with the company.Buyer History Logarithm of the number of years that the assembler’s buyer has personally dealt with the supplier.Supplier Asset Specificity This supplier has made significant investments in terms of equipment, facilities, and engineering

designed specifically to meet the buyer’s supply requirement for the commodity. (7-point Likert scale.)Buyer Asset Specificity Your company has made significant investments in tooling and equipment that are specific to your

relationship with the supplier. (7-point Likert scale.)Breadth Ordered categorical measure of the extent to which a commodity type is used company wide (0), more

than one platform (1), one platform (2), one model (3), or in one trim line (4).

Control variables

Firm Dummy variable to identify assembler-specific effectsInfo Extent to which you share the following 10 kinds of business information: quality information, inventory

information, schedule and delivery information, detail cost information, marketing information,long-term volume projections, manufacturing process information, proprietary technical information,design information, and production capacity. (7-point Likert scale.) (!= 0"81)

Component Revenue Categorical measure of the annual dollar value of purchases from supplier for focal exchange. Thequalitative categories are less than $10 MM (0), $50 MM (1), $100 MM (2) or more than $100 MM (4).

Overall Revenue Categorical measure of the annual dollar value of purchases from supplier for all exchanges betweenthe buyer and supplier. The qualitative categories are less than $10 MM (0), $50 MM (1),$100 MM (2) or more than $100 MM (4).

$All constructed variables were calculated by standardizing the items and calculating the average of the items.

the performance of exchange relationships. We analyzedour 13 performance measures using exploratory factoranalysis with varimax rotation, which measures unob-servable theoretical constructs using reflective indicators(Carmines and Zeller 1979, Zeller and Carmines 1980).Table 2 displays factor loadings for the three factorsthat emerged from the analysis. A single factor with aneigenvalue of 6.15 accounted for 75% of the variationin our data. Only the first two factors yielded eigenval-ues greater than 1; the second factor had an eigenvalueof 1.43, explaining no more than 18% of the variation.Using a loading coefficient greater than or equal to 0.30and eliminating items with high cross-factor loadings(Kim and Mueller 1978), we concluded that the first10 of our variables loaded onto the first factor. All ofthese variables were qualitative measures of performance

consistent with strong economic performance. The threeremaining variables loaded onto the second factor, whichwe called Conflict, with one of them exhibiting a highcross-factor loading. Zaheer et al. (1998) used thesethree items to represent conflict and negotiation coststhat mediate overall performance, although neither con-struct was statistically significant in their study. Giventheir weak empirical results, we considered the aggre-gate measure of conflict a potentially ambiguous mea-sure of performance; it cannot differentiate betweenfunctional and dysfunctional conflict, which may havevastly different implications for performance (Gulati andSytch 2007). With the factor underlying Performanceexplaining over four times more variance than the fac-tor underlying Conflict, we used Performance as ourprimary dependent variable, but we also include the

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table 2 Factor Loadings for Performance Constructs

Factor 1 Factor 2 Factor 3

Buyer’s opinion of supplier compared withthe best alternative supplier for this commodity

1 Price competitive 0"620 %0"127 0"3562 Support and services 0"764 0"096 0"0193 Flexibility in production 0"795 0"092 0"1124 Product quality 0"817 0"174 %0"2615 Product innovations 0"738 0"074 %0"1146 Overall performance1 0"904 0"134 %0"0197 Average past target ratio2 0"698 %0"064 0"3688 Average past price change rate3 0"678 %0"107 0"3999 Average defect rate4 0"793 0"167 %0"321

10 Improvement in average 0"828 0"128 %0"206defect rate5

11 Frequency of significant %0"398 0"487 0"158dissagreements6

12 Ease of negotiation over %0"260 0"731 0"111sharing cost-engineeringchanges7

13 Ease of negotiation over %0"173 0"711 0"109sharing cost-materialcost increases8

Eigenvalue 6"149 1"425 0"710Proportion of variance 0"753 0"175 0"087

eigenvector explains

1We include “Overall Performance” as an item in a battery ofquestions about the buyer’s opinion of suppliers even though it canbe viewed as a summary category. Alternatively, this item can beviewed as a different category of performance. As a robustnesscheck we reran the factor analysis as well as all analysis and veri-fied that our results persist.

2Target-price ratio = (actual part price at market introduction)/(target price your company set when it selected the supplier for thepart).

3Price change rate= average annual rate of price change afterthe market introduction (excluding the price change when the part’sdesign was changed due to engineering changes in specification).

4Defect rate = (number of defective parts)/(number of partsreceived).

5Improvement in defect rate = average annual rate of defectchange after market introduction.

6“During the past year how often were there significant disagree-ments between your business unit and this supplier?”

7“How easy are negotiations between your business unit and thesupplier over sharing the burden of cost (not exactly covered by thecontract) when your business unit requests engineering changes?”

8“How easy are negotiations between your business unit and thesupplier over sharing the burden of cost (not exactly covered bythe contract) when the supplier’s raw material costs increased?”

analyses on Conflict as an empirically derived construct,on the assumption that it may reflect lower transactioncosts.We evaluated the reliability of our Performance con-

struct by estimating Cronbach’s alpha for the battery of10 items; the estimate, 0.93, indicated a sufficient levelof scale reliability. The battery of items comprising Con-flict had Cronbach’s alpha of 0.78, also representing suf-ficient reliability.

Covariates. Our first set of covariates identifiedantecedents of interorganizational trust:Component History was the logarithm of the number

of years an assembler had purchased the focal compo-nent of a current exchange from the supplier, which typ-ically encompassed the history of transacting episodesbetween the exchange partners.Organization History, the logarithm of the number

of years an assembler had purchased any componentfrom the supplier, captured the history of all transact-ing episodes between the exchange partners. Assemblersare presumably likely to recontract with suppliers whohave performed well. Thus, good performance engendersinterorganizational trust.4 Ongoing selection of suppliersby assemblers would suggest the presence of trust, allelse being equal.Buyer Tenure was the logarithm of the number of

years that a lead component buyer had been with thepresent company. The longer a buyer has been with anassembler, the more likely it is that expectations heldby others have been institutionalized and adopted by thebuyer.Buyer History was the logarithm of the number of

years that an assembler’s buyer had personally dealt witha supplier. The longer a buyer has worked with a sup-plier, the greater the likelihood that social attachments,which engender trust, exist.Experience captured whether, on the basis of past

experience, a buyer could not rely on a supplier to keeppromises made. This was based on the response to a sin-gle seven-point Likert item rated from strongly disagreeto strongly agree. The higher the response, the lower thepreexisting trust we expected.Our second set of covariates related to the degree

to which dyad partners had made cospecialized invest-ments—the level of asset specificity in an exchange. Wedid not include contract duration in our model becauseduration is likely to be endogenous with asset specificityand governance. Our model indicates that higher assetspecificity corresponds to more formal modes of gov-ernance. Our survey provided three items intended toassess the degree of asset specificity:Supplier Asset Specificity measured a buyer’s agree-

ment (1, strongly disagree, to 7, strongly agree), withthe statement, “This supplier has made significant invest-ments in terms of equipment, facilities, and engineeringdesigned specifically to meet the buyer’s supply require-ment for the commodity.” Our auto industry intervie-wees had suggested that suppliers typically make specificinvestments, when these are called for.Assembler Asset Specificity measured the buyer’s

agreement (1, strongly disagree, to 7, strongly agree),with the statement, “Your company has made signifi-cant investments in tooling and equipment that are spe-cific to your relationship with the supplier.” We did notcombine Supplier Asset Specificity and Assembler Asset

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Specificity because the latter may signal a credible com-mitment in response to a supplier’s specific investments,which may reflect an exchange of hostages to mitigatehazards (Williamson 1993).An additional measure of asset specificity was

Breadth, the extent to which an assembler used a focalcomponent. Components used over multiple platformsare likely to be standardized, whereas components with asingle use are more likely to require cospecialization. Toconserve degrees of freedom in our analysis, we codedBreadth as an ordered categorical measure of the extentof use: companywide, in more than one platform, in oneplatform, in one model, or in one trim line; higher levelscorresponded to higher degrees of cospecialization.

Control Variables. To insure the robustness of ourresults and to identify our regression models, we in-cluded several control variables (listed in Table 1): Firmwas a dummy variable that identified assembler-specificeffects and distinguished between Ford and Chrysler.Info measured the extent to which a buyer and supplier

exchanged information. This variable, constructed from10 items (+= 0*81; specified in Table 1), accounted forhow information channels between organizations affecttrust, governance choice, and exchange performance,independent of factors that may build information chan-nels. For instance, the variables capturing length of rela-tionship defined above may affect trust and informationchannels, both of which may affect performance. Includ-ing Info as a control allowed us to distinguish amongpotential sources of performance benefits. We also per-formed a factor analysis of the 10 items (not reported).All loaded onto a single factor.Component Revenue and Overall Revenue were cate-

gorical measures of the annual dollar value of purchasesfrom a supplier for, respectively, a focal exchange andall exchanges between the assembler and supplier. Thequalitative categories were less than $10 MM, $50 MM,$100 MM, and more than $100 MM. To conservedegrees of freedom in our analysis, however, we usedwhole numbers ranging from one for the lowest categoryto four for the highest category for each variable. Weused these variables to control for the effect that the sizeof a transaction might have on either the emergence ofpreexisting trust or governance choice. Controlling forthe amount of information exchanged, we assumed thatthe size of a transaction affects governance mode choice,but not performance. The greater the transaction in termsof revenue, the more economical it is for a firm to incurthe fixed cost of setting up and running an ally or makeform of organization. Hence, we expected size to act asa shift parameter in such a way that the organizationalmode chosen would move from buy, to ally, to make asexchange size increases.Table 3 displays summary statistics and correlation

coefficients for our variables. No correlation is large

enough to pose estimation problems. Table 4 providessummary statistics for all variables by mode of exchangegovernance.

AnalysisAn important concern in evaluating the effect of trust ongovernance choice and performance was unobserved het-erogeneity, because unobserved factors not accounted forstatistically might correlate with trust, governance, andperformance. For instance, personal friendships betweenbuyers and personnel at suppliers are unobservable inour data but could have an impact on all three variables.Also, we had to identify the level of interorganizationaltrust that existed prior to a current focal exchange. Toaccount for unobserved factors and identify preexistingtrust, we implemented a three-stage switching regressionmodel (e.g., Hamilton and Nickerson 2003, Masten et al.1991). In the first stage, we modeled interorganizationaltrust as a function of exchange attributes and antecedentsof preexisting trust. In the second stage, we modeledchoice of governance mode as a function of transactionattributes and predicted preexisting trust (estimate in thefirst stage), while omitting the antecedents of preexistingtrust. Finally, in the third stage we modeled exchangeperformance as a function of governance choice, trans-action attributes, and the predicted level of preexistingtrust. We omitted from the third equation variables thataffect governance choice but not performance, to statis-tically identify the predicted governance mode choice.Admittedly complicated, this method was statisticallynecessary, and its use here represents one of the fewattempts in an empirical study on trust to account forunobserved heterogeneity. In view of our theory, it wasalso important to consider the simultaneous effects oftrust on governance and performance.

Regression Equations. Equation (1) took the form:

Trusti = +0 ++1 $Firm++2 $ Info++3 $Component Revenue++4 $Overall Revenue++5 $ Supplier Asset Specificity++6 $Assembler Asset Specificity++7 $Breadth++8 $Buyer_Tenure++9 $Buyer History++10 $Component History++11 $Organization History

++12 $Experience+ ,1i' (1)

where ,1i is a random error term. All variables in Equa-tion (1) need not be viewed in causal terms. Someof them (e.g., Assembler Asset Specificity and Sup-plier Asset Specificity) are included because they are

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table3

Sum

maryStatis

ticsan

dCor

relatio

ns(N

=21

6)

12

34

56

78

910

1112

1314

1516

1Trus

t2

Gov

erna

nce

%0"054

Mod

e3

Perfo

rman

ce0"58

9%0

"096

4Con

flict

%0"474

%0"033

%0"369

5Firm

0"08

9%0

"078

0"13

4%0

"137

6Info

0"01

9%0

"034

0"18

6%0

"048

0"40

37

Com

pone

nt%0

"102

0"17

3%0

"103

0"07

2%0

"095

0"01

9Rev

enue

8Ove

rall

%0"059

0"26

9%0

"077

0"08

7%0

"002

0"07

80"66

1Rev

enue

9Buy

erTe

nure

0"26

6%0

"059

0"22

7%0

"077

0"28

80"06

50"09

90"11

410

Buy

erHistory

0"13

60"15

70"08

3%0

"055

0"07

2%0

"205

0"01

6%0

"003

0"19

611

Com

pone

nt%0

"075

0"25

3%0

"123

0"11

4%0

"117

0"00

60"26

80"25

20"06

1%0

"004

History

12Organ

ization

%0"129

0"27

7%0

"208

0"13

8%0

"004

0"02

40"19

70"40

50"10

20"01

30"59

5History

13Bread

th0"09

40"23

1%0

"035

%0"063

%0"222

%0"185

0"05

30"08

9%0

"031

0"03

70"01

10"05

514

Supp

lierAss

et0"16

10"13

30"27

5%0

"077

0"13

10"18

10"01

20"04

60"23

30"17

4%0

"071

%0"016

0"04

9Sp

ecificity

15Buy

erAss

et%0

"062

0"07

30"04

0%0

"055

0"07

30"25

30"14

20"19

8%0

"009

0"04

40"02

80"05

60"07

00"25

0Sp

ecificity

16Ex

perie

nce

%0"634

0"12

7%0

"527

0"37

00"01

80"05

4%0

"022

0"03

3%0

"229

%0"024

0"12

20"14

6%0

"046

%0"116

0"08

1

Mea

n0"00

01"77

3%0

"003

0"00

00"50

50"00

02"55

13"19

42"37

10"53

62"48

02"94

22"49

15"70

44"69

93"13

0Std.

dev.

0"84

20"62

50"76

70"83

10"50

10"61

61"07

71"01

10"93

80"56

40"84

60"72

01"35

71"15

11"75

71"48

2Min

%2"742

1%2

"792

%1"994

0%2

"264

11

%1"100

%1"792

%1"833

01

21

1Max

1"55

733

1"92

62"10

41

1"33

34

43"71

41"15

34"31

74"31

75

77

7

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table 4 Summary Statistics for Buy, Ally, and Make

Buy (n= 72) Ally (n= 126) Make (n= 24)

Variable Mean Std. dev. Min Max Mean Std. dev. Min Max Mean Std. dev. Min Max

Trust 0"030 0"769 %2"249 1"333 0"015 0"868 %2"742 1"557 %0"170 0"933 %2"025 1"557Mode 1 0 1 1 2 0 2 2 3 0 3 3Performance %0"002 0"760 %1"791 1"701 0"068 0"750 %2"792 1"926 %0"461 0"747 %1"938 1"425Conflict %0"001 0"790 %1"320 2"104 0"032 0"843 %1"994 2"104 %0"164 0"907 %1"770 1"882Firm 0"486 0"503 0 1 0"579 0"496 0 1 0"208 0"415 0 1Info 0"027 0"686 %2"264 1"333 %0"008 0"524 %1"218 0"979 %0"028 0"775 %1"133 1"333Component Revenue 2"417 1"045 1 4 2"548 1"070 1 4 3"208 1"021 1 4Overall Revenue 2"875 1"061 1 4 3"294 0"972 1 4 3"750 0"608 2 4Buyer History 0"374 0"591 %1"099 1"070 0"632 0"525 %1"792 1"179 0"485 0"695 %0"875 1"070Component History 2"274 0"763 0 4"317 2"480 0"753 0 4"317 3"109 1"193 %1"833 4"248Organization History 2"757 0"676 1"099 4"317 2"942 0"729 0 4"317 3"534 0"350 2"996 4"248Supplier Asset Specificity 2"069 1"271 1 5 2"627 1"384 1 5 2"917 1"248 1 5Buyer Asset Specificity 5"500 1"199 2 7 5"762 1"148 2 7 5"958 0"806 4 7Breadth 4"819 1"673 1 7 4"429 1"813 1 7 5"833 1"050 2 7Experience 2"889 1"240 1 7 3"183 1"577 1 7 3"458 1"587 1 6

explanatory variables in the second or third stages. Also,because responses from a given survey respondent couldbe correlated over suppliers/exchanges, we controlledfor such correlation by using the clustering option inSTATA. Clustering affects the estimated standard errorsand the variance-covariance matrix of estimators, typi-cally leading to larger standard errors, but not affectingestimated coefficients.In the second stage of our model, we used an

ordered probit model (e.g., Greene 2002) to examine theeffect of preexisting interorganizational trust, accountingfor the level of asset specificity, on governance modechoice. This procedure allowed us to assess the substi-tution effect. To address the potential for endogeneitybetween our construct for trust and governance choice(i.e., governance choice might engender trust during anexchange), we constructed the predicted level of preex-isting trust, P_Trust, and omitted antecedents of trustfound in the first stage from the second stage. We usedthis predicted level of trust in Equation (2) to investigateHypotheses 1A and 2A. Our second-stage equation tookthe form:

Mode$i = -0 +-1 $Firm+-2 $ Info+-3 $Component Revenue+-4 $Overall Revenue+-5 $ Supplier Asset Specificity+-6 $Assembler Asset Specificity+-7 $Breadth+-8 $P_Trust+ ,2i' (2)

where ,2i is a random error term and Mode$i repre-sents the index of an ordered probit estimation. Withrespect to actual governance mode choices, Modei = 0if Mode$i &.1, Modei = 1 if .1 < Mode$i & .2, andModei = 2 if Mode$i >.2, where .1 and .2 are referred

to as break points in the ordered probit. We usedP_Trust, the level of interorganizational trust predictedby Equation (1), instead of Trust. Doing so allowed usto account for possible endogeneity between the levelof trust and the governance mode choice. Again, weallowed clustering on respondents to control for correla-tions among a buyer’s assessments of different suppliers.Finally, we examined to what extent governance mode

choice and preexisting trust affect exchange perfor-mance, which allowed us to examine complementaryeffect while simultaneously accounting for possible sub-stitution. Our third stage employed a switching regres-sion analysis in which we estimated three models—oneeach for buy, ally, and make—and included the appro-priate inverse Mills ratios, calculated from the orderedprobit analysis to assess the performance delivered byeach organizing mode (Idson and Feaster 1990, Maddala1983). The inverse Mills ratio corrects for sample selec-tion bias that may arise from self-selection of organi-zational modes (Hamilton and Nickerson 2003, Mastenet al. 1991). Without such a correction, our coefficientestimates could be biased by unobservable factors affect-ing both governance mode and performance. Our third-stage model had three equations:

Performanceij = &j0 +&j1 $Firm+&j2 $ Info+&j3 $ Supplier Asset Specificity+&j4 $Assembler Asset Specificity+&j5 $Breadth+&j6 $P_Trust+&j7 $Mills_Ratioj + ,ji' (3)

where j is [buy, ally, make], Mills_Ratioj is the inverseMills ratio for organizing mode j , and ,ji is a ran-dom error term. The switching regression model allowedus not only to evaluate the performance delivered byeach governance mode, but also to quantify the effect of

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


inappropriate matches of exchange conditions to mode(Maddala 1983). We omitted the controls ComponentRevenue and Overall Revenue from Equation (3) to iden-tify our ordered probit model econometrically becausewe assumed that the size of a transaction had no directeffect on performance. Without omitting these variables,our estimate of predicted mode choice would be iden-tified only by the nonlinearity of the ordered robit. Weagain allowed for clustering to control for correlationsamong a buyer’s assessments of different suppliers. Wealso reestimated Equation (3) with Conflict as our depen-dent variable:

Conflictij = &j0 +&j1 $Firm+&j2 $ Info+&j3 $ Supplier Asset Specificity+&j4 $Assembler Asset Specificity+&j5 $Breadth+&j6 $P_Trust+&j7 $Mills_Ratioj + ,ji* (3a)

ResultsTable 5 displays the set of nested models used to assessTrust. Model 1 includes the dummy variable control-ling for the assembler and our two controls for revenue.No coefficient is significant, and the model has littleexplanatory power. Model 2 adds our three measures ofasset specificity, but explains little variance (R2 = 0*06).Only the coefficient for Supplier Asset Specificity (p <0*10) is statistically significant, which offers a signifi-cant but limited improvement over Model 1 (p < 0*05).

Table 5 Interorganizational Trust and Governance Mode

Interorganizational trust Governance mode

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Firm 0"140 (0.157) 0"153 (0.157) 0"076 (0.124) %0"179 (0.181) %0"137 (0.185) %0"090 (0.188)Info %0"011 (0.138) 0"022 (0.127) 0"097 (0.100) %0"044 (0.169) %0"036 (0.179) %0"032 (0.183)Component Revenue %0"072 (0.089) %0"063 (0.086) %0"145$ (0.065) %0"010 (0.095) 0"002 (0.093) %0"016 (0.095)Overall Revenue 0"011 (0.089) 0"007 (0.087) 0"074 (0.060) 0"334$$ (0.098) 0"311$$ (0.099) 0"317$$ (0.099)Supplier Asset Specificity 0"119† (0.067) 0"030 (0.054) 0"137 (0.077) 0"176$ (0.076)Buyer Asset Specificity %0"060 (0.041) %0"020 (0.033) 0"000 (0.054) %0"019 (0.054)Breadth 0"084 (0.057) 0"061 (0.039) 0"156$ (0.064) 0"182$$ (0.068)P _Trust %0"302$ (0.142)Buyer Tenure 0"085 (0.073)Buyer 0"147 (0.096)Component History 0"112 (0.090)Organization History %0"140 (0.093)Experience %0"348$$ (0.041)

Constant 0"080 (0.225) %0"547 (0.354) 0"814$ (0.383)#1 0"473 (0.329) 1"601 (0.549) 1"790 (0.538)#2 2"263 (0.348) 3"452 (0.574) 3"667 (0.560)

R2 0.01 0.06 0.48F -test 3"08$ 3"60$$

Psuedo R2 0.05 0.07 0.08$2 2"94$ 18"23$$

Model 3 shows an R2 of 0.48 and a substantialimprovement over Model 2 (p < 0*01). Only the coef-ficients for Experience, Component Revenue, and theconstant are significant in Model 3. The coefficient forExperience is negative (p < 0*01), which indicates thattrust is lower when past experience suggests a supplierwill not keep promises. The coefficient for ComponentRevenue (p < 0*05) is also negative, which suggests thatthe higher the expectations for a specific component, thelower the level of trust in a supplier.It is important to note that none of our parame-

ter estimates for asset specificity are related to Trust.These estimates suggest that contracting hazards in anexchange had little to do with the level of trust in ourcontext, a finding consistent with our model capturingpreexisting trust.Models 4, 5, and 6 in Table 5 present results from our

ordered probit analysis of governance choice. Model 4is our baseline model, with an R2 of 0.05. Model 5,adding our asset specificity proxies, yields a pseudo R2

of 0.07 and is a significant improvement (p < 0*05) overModel 4. Model 6, adding the predicted level of trust,shows a pseudo R2 of 0.08 and significant improvementover Model 5 (p < 0*01).The coefficients are generally consistent over these

three nested models, so we focus our attention onModel 6. The coefficient for Overall Revenue is sig-nificant (p < 0*01) and positive, indicating that theprobability of choosing a more hierarchical governancemode increases with the sum of revenue over all theexchanges between an assembler and a supplier. Thisfinding suggests that interdependencies that influence

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


organizational choice may exist over dyadic transactions(Argyres and Liebeskind 1999). The coefficient for Sup-plier Asset Specificity was nonsignificant in Model 5, butpositive and significant (p < 0*05) in Model 6, indicat-ing that governance choice is more likely to shift frombuy, to ally, to make as supplier asset-specific investmentgrows. This change in significance might have arisenbecause P_Trust, which was omitted in Models 4 and 5,is correlated with both Supplier Asset Specificity andMode. The coefficient for Breadth is significant (p <0*01) and positive, which implies that the more narrowlya component is used, the more likely governance choiceis to shift from buy, to ally, to make. Both of thesefindings support Williamson’s prediction that transactionattributes are matched in a transaction-cost economizingway to governance structure. The coefficient for Assem-bler Asset Specificity is nonsignificant, suggesting thatbuyer-specific investments have little influence on gov-ernance choice.Our estimated level for preexisting interorganizational

trust is significant (p < 0*05) and negative. The negativesign indicates that the greater the trust, the less likely anexchange is to be organized hierarchically. SupportingHypotheses 1A and 1B (predicting a substitution effect),higher levels of trust shift the governance choice frommake, to ally, to buy.Before discussing the third stage of our model, we

focus our attention on assessing the marginal effectsof Supplier Asset Specificity, Breadth, and P_Trust, tocompare the effects of asset specificity and preexist-ing trust on governance choice.5 Table 6 reports themarginal probabilities at the mean value, 20th percentile,and 80th percentile for each covariate, with all othercovariates held at their means. We are mainly inter-ested in comparing the magnitude of the marginal effectfor P_Trust at the 20th percentile with the magni-tude of the marginal effect for asset specificity prox-ies like Supplier Asset Specificity at that percentile. Asexpected, higher Supplier Asset Specificity or Breadthincreases the marginal probability of an ally or makechoice and decreases the marginal probability of eithera buy or an ally choice. Conversely, higher levels ofP_Trust decrease the probability that make will be cho-sen and increase the probability that either ally or buywill be chosen. Perhaps the most interesting finding evi-dent from Table 6 is that marginal effects are greaterfor P_Trust than for either measure of asset specificity.Indeed, this finding suggests that managing preexistingtrust deserves much attention if governance choice hassubstantial cost implications, as we assert in our the-ory. We conclude that trust may have a greater marginaleffect than asset specificity on governance choice.Table 7 reports results for our switching regres-

sion model of performance. Models 7, 8, and 9 reportnested regressions analyzing exchange performance in

Table 6 Marginal Effect of Regressors!

Marginal effects on probability

Buy Ally Make

Supplier Asset Specificity %0"065 0"042 0"023at 20th percentile

Supplier Asset Specificity %0"062 0"034 0"027at mean

Supplier Asset Specificity %0"054 0"017 0"037at 80th percentile

Breadth at 20th percentile %0"071 0"052 0"019Breadth at mean %0"064 0"036 0"028Breadth at 80th percentile %0"054 0"014 0"040

P _Trust at 20th percentile 0"098 %0"041 %0"057P _Trust at mean 0"106 %0"059 %0"047P _Trust at 80th percentile 0"113 %0"075 %0"038

$All covariates except focal one set to mean.

the 72 exchanges organized as buy. We focus our atten-tion on Model 9 because its coefficients are consistentwith previous models, and it offers not only the largest R2

but also a significant improvement over Model 8. Onlytwo coefficients are statistically significant (p < 0*05)in Model 9. The coefficient for Info is positive, indi-cating that higher levels of information exchange cor-respond to higher levels of performance within the buymode. The coefficient for P_Trust is also positive, indi-cating that higher levels of trust correspond to higherperformance within the buy mode of governance. Thisfinding is consistent with Hypothesis 2A (predicting acomplementary effect). The coefficients for our Millsratios are nonsignificant, indicating that unobserved het-erogeneity is not problematic in estimating coefficientsfor this mode of governance.Models 10, 11, and 12 (see Table 7) report nested

regressions for analyzing exchange performance in the126 exchanges organized as ally. Only the coefficientfor the Mills ratio is significant (p < 0*05) in Model 10.It is negative, which indicates that selection bias is anappropriate concern. Model 11 (R2 = 0*32), which incor-porates our three asset-specificity measures, providesa significant and substantial improvement in explana-tory power over Model 10 (R2 = 0*08). The coefficientfor Supplier Asset Specificity is positive and significant(p < 0*01), and the coefficient for Breadth is positiveand weakly significant, indicating that performance ishigher when specific supplier investments are higheror when a component is used more narrowly by anassembler. Our inverse Mills ratio for ally again is sig-nificant (p < 0*01) and negative. Also, our constantterm is highly significant and negative. Model 12 incor-porates P_Trust, which increases our R2 to 0.49 andoffers a significant improvement over Model 11. InModel 12, the coefficients for Supplier Asset Specificity(p < 0*01) and Assembler Asset Specificity (p < 0*10)are significant and positive. These coefficients indicate

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table 7 Switching Regression Model of Performance (N = 222)

Buy Ally Make

Model 7 Model 8 Model 9 Model 10 Model 11 Model 12 Model 13 Model 14 Model 15

Firm %0"137 %0"088 %0"128 0"111 0"009 %0"057 0"625 0"358 0"288%0"227& %0"231& %0"209& %0"208& %0"159& %0"131& %0"660& %0"508& %0"531&

Info 0"349$$ 0"325$ 0"297$$ 0"161 %0"017 %0"005 %0"272 %0"150 %0"0950"118 %0"124& %0"100& 0"217 %0"158& %0"138& %0"231& %0"285& %0"311&

Supplier Asset Specificity 0"056 %0"114 0"367$$ 0"220$$ 0"646$ 0"544$

%0"109& %0"088& %0"081& %0"070& %0"241& %0"227&Buyer Asset Specificity 0"032 0"004 0"007 0"055† %0"076 %0"096

%0"067& %0"052& %0"040& %0"032& %0"117& %0"138&Breadth 0"122 %0"065 0"107† %0"046 0"440$ 0"390$

%0"098& %0"081& %0"055& %0"051& %0"173& %0"150&P _Trust 0"989$$ 0"637$$ 0"245

%0"187& 0"109& %0"231&Mills ratio-buy 0"117 0"443 %0"419

%0"383& %0"448& %0"401&Mills ratio-ally %0"466$ %1"136$$ %0"432$

%0"227& %0"260& %0"206&Mills ratio-make 0"479 1"667$ 1"335†

%0"537& %0"789& %0"668&

Constant 0"173 %0"237 0"338 %0"126 %2"684$$ %1"423$$ %1"338 %7"799$$ %6"351$$

%0"437& %0"460& %0"449& %0"152& %0"588& %0"535& %0"787& %2"070& %1"678&

R2 0"09 0"13 0"40 0"08 0"32 0"49 0"09 0"55 0"56F -stat 0"87 28"05$$ 7"22$$ 34"04$$ 4"41$ 1"12

†p < 0"10; $p < 0"05; $$p < 0"01.

that performance is higher when specific investments byeither supplier or buyer are higher. The coefficient forP_Trust is positive and significant (p < 0*01). Support-ing Hypothesis 2B, preexisting trust complements allymode governance by improving exchange performance.Models 13, 14, and 15 (see Table 7) report nested

regressions for analyzing the level of performance in the24 exchanges organized as make. No coefficient is sig-nificant in Model 13. Model 14, incorporating our asset-specificity variables, is a significant improvement (p <0*01) over Model 13 (R2 0.55 versus 0.09). The coef-ficients for Supplier Asset Specificity and Breadth aresignificant (p < 0*05) and positive. These estimates indi-cate that higher performance is correlated with higherlevels of specific investments by the supply divisions andmore narrow use of a component by the buying divi-sions. The inverse Mills ratio (p < 0*05) and the constant(p < 0*01) are significant, although the former is positiveand the latter negative. The coefficient for the inverseMills ratio indicates that assemblers would experiencelower performance if the exchange were organized underally or buy.Model 15, which incorporates P_Trust, offers no sig-

nificant improvement in fit. The coefficient for P_Trustis nonsignificant, which suggests that exchange perfor-mance is not significantly enhanced when preexistingtrust is present (a complementary effect). Therefore, wereject Hypothesis 2C. Estimates for other coefficients are

similar to those found in Model 14, except that the Millsratio is now weakly significant.We reestimated our switching regression model with

Conflict in place of Performance as the dependent vari-able. Table 8 reports our results. Only 216 observationswere available for Models 16–24. Because the indepen-dent variables for our model of Conflict are the samefor our model of Performance and because most coef-ficient estimates are nonsignificant, we focus only onconsistently significant findings. First, the negative andsignificant coefficient estimates for Firm in Models 19through 21 indicate that one of the firm’s ally exchangesexperienced less conflict than did the other one. Second,only the Mills ratios for make exchanges are statisticallysignificant, albeit weakly, which indicates that exchangesorganized internally experience less conflict than theywould if organized otherwise. Finally, the coefficientsfor P_Trust in Models 18 and 21 indicate that higherlevels of trust correlate with lower conflict in the buyand ally modes.

Discussion and ConclusionIs interorganizational trust a substitute for or com-plement to formal governance? The resolution ofthis debate is of more than academic interest. Thedisparate views described herein offer fundamentallydifferent implications for how firms manage and sup-port exchanges, and competing prescriptions for strong

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Table 8 Switching Regression Model of Conflict (N = 216)

Buy Ally Make

Model 16 Model 17 Model 18 Model 19 Model 20 Model 21 Model 22 Model 23 Model 24

Firm 0"090 %0"036 0"081 %0"399$$ %0"389† %0"282† %1"256† %1"246 %1"202%0"290& %0"286& %0"248& %0"199& %0"201& %0"161& %0"729& %0"789& %0"787&

Info 0"038 0"077 0"082 %0"084 %0"077 %0"064 0"624 0"653 0"623%0"159& %0"135& %0"128& %0"193& %0"190& %0"166& %0"409& %0"490& %0"504&

Supplier Asset Specificity %0"019 0"061 %0"059 0"025 %0"241 %0"217%0"085& %0"078& %0"100& %0"094& %0"280& %0"268&

Buyer Asset Specificity %0"063 %0"030 0"018 %0"051 0"066 0"081%0"060& %0"058& %0"052& %0"054& %0"176& %0"222&

Breadth %0"256$$ %0"161 %0"030 0"031 %0"097 %0"109%0"096& %0"107& %0"082& %0"081& %0"179& %0"216&

P _Trust %0"783$$ %0"602$$ %0"081%0"187& %0"132& %0"497&

Mills ratio-buy %0"087 %0"693 %0"546%0"352& %0"418& %0"359&

Mills ratio-ally 0"029 0"154 0"108%0"309& %0"425& %0"421&

Mills ratio-make %1"050 %1"381† %1"472†

%0"806& %0"732& %0"785&Constant %0"134 0"248 %0"428 0"267 0"635 0"232 1"766 3"614 3"540

%0"424& %0"367& %0"424& %0"153& %0"710& %0"680& %1"363& %2"203& %2"262&

R2 0"01 0"12 0"32 0"07 0"08 0"23 0"23 0"25 0"26F -stat 4"01$ 17"49$$ 0"14 20"88$$ 0"30 0"03

†p < 0"10; $p < 0"05; $$p < 0"01.

exchange performance. The substitution view suggeststhat exchange performance is superior when trust ispresent, allowing for less-formal governance to safe-guard an exchange, whereas the complementation viewsuggests that exchange performance is superior whentrust operates together with formal governance and isthus independent of the governance structure used. Per-haps one reason for the disconnect between alternativeperspectives on the role of trust and formal governancein exchange performance arises from the emergence oftwo parallel research streams: one on governance choiceand another on exchange performance. Very few studieshave considered both.Seeking to resolve this debate, we focused on the

impact of interorganizational trust existing prior to agiven exchange, arguing that such trust complementsthe use of appropriate formal governance in shap-ing exchange performance. Trust lowers costs for allmodes of governance whenever exchange hazards arepresent, and thus enhances performance regardless ofmode. Governance cost decreases because trust facili-tates adaptation: Exchange partners are more likely toavoid disputes or resolve them quickly (Gulati and Singh1998, Gulati et al. 2005). Reduced governance coststranslate directly into better exchange performance.However, our theory also suggests that trust can lead

to a substitution of less formal for more formal modes ofgovernance. We argued that the governance-cost reduc-ing benefits of trust are greater for buy than for ally and

greater for ally than for make. This differential impact oftrust on governance costs over organizing modes arisesbecause formal governance limits the range of adapta-tions that trust can facilitate. Make imposes greater con-straints than ally, thus limiting the benefits of trust morein the former mode. The result of this differential impactis that the buy mode of governance, with the additionof trust, can be used over a broader range of exchangehazards than can buy without trust, which in turn offerslower governance costs and enhances exchange perfor-mance. Also, an ally exchange with trust can substi-tute over some range of exchange hazards for make,which enhances exchange performance. These substi-tutions over various types of governance of exchangerelationships take place near the critical values of assetspecificity, where buy and ally or ally and make havesimilar governance costs without trust. Thus, preexistinginterorganizational trust is both a complement to and asubstitute for formal governance modes and can greatlyenhance exchange performance.Our empirical analysis of the performance of ex-

change relationships for the sourcing of components inthe U.S. auto industry broadly supports our theory. Wesaw a substitution effect: Higher preexisting interorga-nizational trust corresponded to less hierarchical gover-nance. The higher the level of trust, the more likely buywas, rather than ally, and ally rather than make. We alsosaw a complementary effect of trust on exchange perfor-mance. Trust enhanced performance for buy and ally and

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


lowered conflict for both. It had no effect on make. Thus,trust complements buy and ally governance choices bylowering conflict and enhancing performance.In sum, this study makes several distinct contribu-

tions to the study of interorganizational trust, gover-nance, and exchange performance. First, it contributes tothe existing body of research by explicating not only therelationship between interorganizational trust and for-mal governance, but also the performance implicationsof each for exchange relationships. Although prior stud-ies have focused on the relationship between formaland informal governance mechanisms (e.g., Gulati 1995,Nooteboom et al. 1997), few studies have explicitlyaddressed the ramifications of such factors for exchangeperformance (e.g., Gulati et al. 2005).Second, this study adds to the small number of empir-

ical investigations that highlight the intricate interplayof trust and formal governance arrangements using fielddata. Prior studies on the interplay between governanceand trust have been either conceptual (e.g., Dwyer et al.1987, Hill 1990, Nooteboom 1996), experimental (e.g.,Malhotra and Murnighan 2002), or ethnographic (e.g.,Larson 1992). Finally, although much prior researchon the interplay of trust and governance has beenrestricted to a particular governance mode—studies ofthe heterogeneity within ally arrangements (e.g., Gulati1995, Larson 1992, Parkhe 1993), for example—thisinvestigation spans the whole continuum of governancearrangements, from hierarchical exchanges to simpleopen market transactions. In doing so, this study revealshow trust and formal governance influence exchangeperformance in all major governance modes.This study also has important managerial implications.

Our findings on substitute and complementary effectsbetween trust and governance imply substantial costsand performance loss within exchanges if managers usean inappropriate mode of governance to structure anexchange. Managers who choose unwisely will eitherincur added costs by adopting and maintaining unneededformal governance structures or expose themselves topotentially costly adaptation losses by relying on inad-equate formal safeguards. Both outcomes can gener-ate substantial costs (Mayer and Nickerson 2005). Ourfindings suggest that the performance of exchange rela-tionships benefits from preexisting interorganizationaltrust. However, simply developing interorganizationaltrust may not be enough to fully realize available per-formance benefits; the level of benefit depends not onlyon generating trust, but also on relying on it to supportless-costly governance. Thus, ongoing and central ques-tions for managers should concern not only how to cre-ate trust, but also how to obtain its full benefits.Several areas of potential improvement are worth

noting to aid future research efforts. First, our studysuggests that it is empirically important to distinguishbetween preexisting trust and trust emerging during

an exchange through ongoing interaction. Our cross-sectional data forced us to identify preexisting trusteconometrically. Panel data on trust before, during, andafter exchanges would greatly enhance understanding ofthe timing and complex interplay between trust and gov-ernance choice. Relatedly, recent work has suggestedthe need for a more dynamic approach to understand-ing exchange partner selection and the emergence oftrust. For instance, Bercovitz et al. (2006) found thatexchange partners that exceed expectations are morelikely to enter future exchanges. Gulati and Sytch (2008)uncovered the intricate relationship between a history ofinterfirm interaction and formation of interorganizationaltrust. This relationship, positive in later stages, is charac-terized by initial ambivalence and virtually nonexistenttrust, because opportunities for demonstrating trustwor-thiness and evidence for judging the trustworthiness of apartner are insufficient. Further, history ceases to matteronce partners are far along into transacting with eachother, because processes of learning about, and identifi-cation with, the partner subside. These dynamics, whichare related to the emergence of trust and to its possiblydiminishing marginal returns, require further scholarlyinvestigations.A more specific caveat is that we measured Perfor-

mance and Conflict in reference to the best alternativesupplier for a commodity. We of course would have pre-ferred an archival and objective measure of performancefor each exchange, but we had no access to such data,owing to issues of confidentiality. An absolute qualita-tive measure (i.e., How well does this supplier performcompared with all suppliers?) would also be problematicbecause of the great variety of components and the reli-ability issues this diversity creates. That is, performancein one component area may not be equivalent to perfor-mance in another. Thus, in view of feedback from man-agers and buyers, we asked respondents to compare eachfocal supplier of each component to the best alternativesupplier for the same component so that the performancemeasure would be comparable within a class of com-ponents. This approach may have led to certain biases.For instance, our discussion with buyers suggested thata next-best supplier might be willing in some contextsto make idiosyncratic investments similar to those of theactual supplier, but the alternate supplier could still beviewed as inferior, owing to lack of trust. In this case,we would expect the coefficients for the effect of assetspecificity on performance to be biased downward. Italso could be the case that the next-best supplier couldbe equally trusted, which would also bias our coeffi-cients downward. In sum, we think that our measurementapproach would bias our coefficients toward zero, whichsuggests that our estimates are conservative.One more-related caveat with our measure of per-

formance arises from our measuring exchange perfor-mance only from the standpoint of the assembler and at

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


one point in time. While an ideal approach would havebeen to obtain dyadic data with performance measuresfrom both partners conducted over a period of time, thiswas not feasible with our sample. Consistent with mostprior studies (e.g., Gulati and Sytch 2007, Parkhe 1993,Zaheer and Venkatraman 1995), we relied upon a one-sided cross-sectional measure of performance.Another important caveat to this research is that we

chose not to consider uncertainty. Technological, behav-ioral, and demand uncertainty all could play importantroles in determining governance choice and performancefor exchanges. We did not include uncertainty in ourstudy for several reasons. First, we think that given thecentral focus of this work on asset specificity, formaliz-ing our theory around this factor is consistent with theway much of the literature has developed (e.g., Mastenet al. 1991, Shelanski and Klein 1995). Including uncer-tainty would offer only incremental value and wouldreduce parsimony and prevent a focus on the centralissue of the interplay among trust, governance, and per-formance. Also, we are unaware of any work examin-ing asset specificity with uncertainty and rejecting assetspecificity as the driver of Williamson’s theory.In sum, our approach was consistent with how much

of the conceptual and empirical research in this bur-geoning domain has approached these issues. Ultimately,making a trade-off between conceptual and empiricalcomplexity and parsimony and focus, we favored thelatter, especially given long precedent in the vast liter-ature on transaction costs. Nonetheless, we did conductsome analyses with two measures of uncertainty (vol-ume uncertainty and technical uncertainty), which wecollapsed into one using an alpha score, and includedeight additional variables (two measures of uncertaintyplus six interaction terms between uncertainty and assetspecificity measures) into our models. In results notreported here, our main effects did not change signifi-cantly. The incorporation of various types of uncertaintyinto the model remains an open and important questionworthy of further research.Notwithstanding these caveats, our paper advances the

literature on trust and interorganizational exchange inseveral ways. First and most importantly, it developsand empirically tests our theory that interorganizationaltrust always complements governance choice and canalso facilitate the substitution of a less hierarchical gov-ernance mode for a more hierarchical one. This theorybrings together two seemingly disparate views of preex-isting interorganizational trust as either a substitute for,or complement to, governance choice. We suggest thattrust can be both. The implication is that both trust andthe use of appropriate governance arrangements shapeexchange performance.Our findings support and further develop many of the

claims of transaction cost economists (e.g., Williamson1991). As asset specificity deepened, we found that the

governance mode shifted from buy to ally to make. Inparticular, we found that specific investments by sup-pliers and breadth of product use were critical pre-dictors of governance choice. Such investments alsoclearly affect exchange performance for ally and makemodes of governance, as performance increases withsupplier-specific investment. Unfortunately, lack of adirect measure of economic or transaction costs made itdifficult to assess the effects of these costs on financialperformance. Although we distinguish between preex-isting trust and trust arising during the exchange, thismerits further investigation. Furthermore, scholars havealso distinguished between calculative and noncalcula-tive trust. More research is needed to more fully identifyand evaluate impact of each distinct facet of trust onorganizational choice and performance.This paper utilized methodological elements not typi-

cally employed in research on trust. For example, endo-geneity has not been much of a concern in empiricalinvestigations on trust. However, it is likely that trust isindeed endogenous to a variety of antecedents, so that,absent controls for endogeneity, an analysis of the per-formance effects of trust will likely yield biased coeffi-cient estimates. Indeed, our results not only support ourtheory, but also demonstrate that correcting for endo-geneity is warranted.Our paper is also important to the growing body of

research on hybrid or alliance forms. Our analysis pro-vides one of the first empirical assessments involvingthe trichotomy of buy, ally, and make. This trichoto-mous choice model is particularly important given thegrowing interest in ally modes of organization overthe past decade (Dyer 1997, Gulati 2007, Zaheer andVenkatraman 1994). Ours is one of the first empiricalexaminations to include ally along with buy and make(e.g., Gulati et al. 2005). It is also of note that althoughthe automobile industry has been the context for manyempirical transaction cost economics studies, none ofthese has had the breadth of observations and micro-analytic detail of our study. For instance, even thoughtransaction cost economics is predicated on transactionas the unit of analysis, most studies use aggregatedunits of analysis, such as all the transactions involvinga particular component (e.g., Masten et al. 1989), andrarely evaluate performance. Alternatively, automotiveindustry studies that have used the transaction as the unitof analysis have had far less data than ours (e.g., Walkerand Weber 1987). Thus, our analysis is a valuable exten-sion of both studies of trust and governance and thosebased in the automobile industry.

AcknowledgmentsThe authors are grateful to Paul Lawrence, who was a col-laborator in the field research, survey design, and data col-lection. They also thank Lyda Bigelow, Kurt Dirks, PhanishPuranam, Maxim Sytch, Jim Wade, Aks Zaheer, and ToddZenger for their comments on drafts. All errors remain thoseof the authors.

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Endnotes1“Institutional context” offers another source of interorgani-zational trust (Fukuyama 1995, Zucker 1987) and can asserta powerful influence on the behavior of exchange part-ners through positive and negative reinforcement (Zukin andDiMaggio 1990). In contexts rich with institutional constraints,opportunistic behavior may be less likely (for examples fromthe Italian districts, see Lazerson 1988, Weiss 1984). Thus, theinstitutional context may also affect general expectations aboutanticipated behavior and in turn engender interorganizationaltrust.2Williamson (1985, p. 21) describes expected ex post gov-ernance costs that arise once a governance structure is cho-sen as including (1) the “maladaptation costs” incurred whenbilateral parties are not able to respond quickly and easilyto problems stemming from disagreements and self-interestedbargaining (see also Williamson 1996, p. 107); (2) the “hag-gling costs” incurred when partners attempt to realign theirexchange should it have drifted out of alignment due to unan-ticipated events; (3) the setup and running costs associatedwith the governance structures to which disputes that arise dur-ing the exchange are referred (e.g., management costs to mon-itor and resolve disputes, administrative controls that establishrules and procedures for resolving disputes, etc.); and (4) the“bonding costs” that come from investing in commitments,such as mutual investments to signal financial commitmentto the exchange, which thereby mutes incentives for oppor-tunism. Our theoretical arguments, which are provided below,will feature how interorganizational trust affects these sourcesof governance costs.3While it is difficult to separate antecedents of preexistingtrust from trust itself in cross-sectional data (Gulati and Sytch2007), we omitted these because they may potentially influ-ence preexisting trust.4Zaheer et al. (1998) argue that expectations held by employ-ees of an organization can be transmitted to and from bound-ary spanners through institutionalization. These expectationsdevelop over time (Ring and Van de Ven 1994) at an organiza-tional level. Thus, the history of exchange interactions betweenorganizations is a relevant antecedent.5We note that the marginal effects of the regressors on theprobabilities in an ordered probit do not equal regressor coef-ficients and depend on .1 and .2. The marginal effects forany particular regressor, Xi, are calculated by:

%PBuy

%Xi

=%/!.1 %-X$-i0

%PAlly

%Xi

= (/!.1 %-X$%/!.2 %-X$)-i0

%PMake

%Xi

=/!.2 %-X$-i*

The marginal effect thus varies with the value of Xi.

ReferencesArgyres, N. S., J. P. Liebeskind. 1999. Contractual commitments,

bargaining power, and governance inseparability: Incorporatinghistory into transaction cost theory. Acad. Management Rev. 2449–63.

Arrow, K. 1974. The Limits of Organization. Norton, New York.

Aulakh, P. S., M. Kotabe, A. Sahay. 1996. Trust and performancein cross-border marketing partnerships: A behavioral approach.J. Internat. Bus. Stud. 27(5) 1005–1032.

Barney, J. B., M. H. Hansen. 1994. Trustworthiness as a source ofcompetitive advantage. Strategic Management J. 15 175–190.

Becerra, M., A. K. Gupta. 2003. Perceived trustworthiness withinthe organization: The moderating impact of communication fre-quency on trustor and trustee effects. Organ. Sci. 14(1) 32–44.

Bercovitz, J., J. A. Nickerson, S. Jap. 2006. The antecedents and per-formance implications of cooperative exchange norms. Organ.Sci. 17(6) 724–740.

Blau, P. 1964. Exchange and Power in Social Life. Wiley, New York.

Bradach, J., R. Eccles. 1989. Rice, authority, and trust: From idealtypes to plural forms. Annual Rev. Sociol. 15 97–118.

Bromiley, P., L. L. Cummings. 1995. Transaction costs in organiza-tions with trust. Res. Negotiation Organ. 5 219–247.

Carmines, E. G., R. A. Zeller. 1979. Reliability and Validity Assess-ment. Sage, Beverly Hills, CA.

Chiles, T. H., J. F. McMackin. 1996. Integrating variable risk prefer-ences, trust, and transaction cost economics. Acad. ManagementRev. 21(1) 73–99.

Cummings, L. L., P. Bromily. 1996. The organizational trust inventory(OTI): Development and validation. R. Kramer, T. Tyler, eds.Trust in Organizations. Sage, Thousand Oaks, CA.

Dirks, K. T., D. L. Ferrin. 2001. The role of trust in organizationalsettings. Organ. Sci. 12 450–467.

Dodgson, M. 1993. Learning, trust, and technological collaboration.Human Relations 46(1) 77–94.

Dore, R. 1983. Goodwill and the spirit of market capitalism. BritishJ. Sociol. 34 459–482.

Dwyer, F. R., P. H. Schurr, S. Oh. 1987. Developing buyer-sellerrelationships. J. Marketing 51 11–27.

Dyer, J. H. 1997. Effective interfirm collaboration: How firms mini-mize transaction costs and maximize transaction value. StrategicManagement J. 18(7) 535–556.

Eccles, R. G. 1981. Bureaucratic versus craft administration: The rela-tionship of market structure to the construction firm. Admin. Sci.Quart. 26(3) 449–469.

Eccles, R. G., D. B. Crane. 1988. Doing Deals" Investment Banks atWork. Harvard University Press, Boston.

Eccles, R. G., H. C. White. 1988. Price and authority in inter-profitcenter transactions. Amer. J. Sociol. 94 S17–S51.

Fehr, E., S. Gachter. 2000. Do incentive contracts crowd out voluntarycooperation? Working paper, Institute for Empirical Research inEconomics, University of Zurich, Switzerland.

Fowler, F. J. 1993. Survey Research Methods. Sage, Beverly Hills,CA.

Frey, B. S. 1997. A constitution for knaves crowds out civic Virtues.Econom. J. 107 1043–53.

Fukuyama, F. 1995. Trust" The Social Virtues and the Creation ofProsperity. The Free Press, New York.

Gambetta, D. 1988. Can we trust trust? D. Gambetta, ed. Trust" Mak-ing and Breaking Cooperative Relations, Vol. 13. Electronic edi-tion, Department of Sociology, University of Oxford, 213–237.

Ghoshal, S., P. Moran. 1996. Bad for practice: A critique of the trans-action cost theory. Acad. Management Rev. 21(1) 13–47.

Granovetter, M. 1985. Economic action and social structure: Theproblem of embeddedness. Amer. J. Sociol. 91(Nov.) 481–510.

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Greene, W. H. 2002. Econometric Analysis. Prentice-Hall, UpperSaddle River, NJ.

Gulati, R. 1995. Does familiarity breed trust? The implications ofrepeated ties for contractual choice in alliances. Acad. Manage-ment J. 38(1) 85–112.

Gulati, R. 2007. Managing Network Resources" Alliances, Affilia-tions, and Other Relational Assets. Oxford University Press,New York.

Gulati, R., H. Singh. 1998. The architecture of cooperation: Manag-ing coordination costs and appropriation concerns in strategicalliances. Admin. Sci. Quart. 43 781–814.

Gulati, R., M. Sytch. 2007. Dependence asymmetry and joint depen-dence in interorganizational relationships: Effects of embed-dedness on a manufacturer’s performance in procurementrelationships. Admin. Sci. Quart. 52 32–69.

Gulati, R., M. Sytch. 2008. Does familiarity breed trust? Revisiting theantecedents of trust. Managerial Decision Econom. 29 169–190.

Gulati, R., P. Lawrence, P. Puranam. 2005. Adaptation in vertical rela-tionships: Beyond incentive conflict. Strategic Management J.26 415–440.

Hamilton, B., J. Nickerson. 2003. Correcting for endogeneity in strate-gic management research. Strategic Organ. 1(1) 53–80.

Heide, J. B., G. John. 1990. Alliances in industrial purchasing:The determinants of joint action in buyer-supplier relationships.J. Marketing Res. 27(1) 24–36.

Heide, J. B., A. S. Miner. 1992. The shadow of the future: Effects ofanticipated interaction and frequency of contact on buyer-sellercooperation. Acad. Management J. 35(2) 265–291.

Hill, C. W. L. 1990. Cooperation, opportunism, and the invisible hand:Implications for transaction cost theory. Acad. Management Rev.15(3) 500–513.

Idson, T. L., D. J. Feaster. 1990. A selectivity model of employer-sizewage differentials. J. Labor Econom. 8(1) 99–122.

Jones, C., W. S. Hesterly, S. P. Borgatti. 1997. A general theory of net-work governance: Exchange conditions and social mechanisms.Acad. Management Rev. 22(4) 911–945.

Kim, J. O., C. W. Mueller. 1978. Factor Analysis" Statistical Methodsand Practical Issues. Sage, Newbury Park, CA.

Kramer, R. 1999. Trust and distrust in organizations: Emerging per-spectives, enduring questions. Annual Rev. Psych. 50 569–598.

Larson, A. 1992. Network dyads in entrepreneurial settings: A studyof the governance of exchange relationships. Admin. Sci. Quart.37 76–104.

Lazerson, M. 1988. Small firm growth: An outcome of markets andhierarchies. Amer. Sociol. Rev. 53(3) 330–342.

Lazzarini, S. G., G. J. Miller, T. R. Zenger. 2004. Order with somelaw: Complementarity vs. substitution of formal and informalarrangements. J. Law, Econom. Organ. 20(2) 261–298.

Lincoln, J. M., M. L. Gerlach. 2004. Japan’s Network Economy.Cambridge University Press, New York.

Lorenz, E. H. 1988. Neither friends nor strangers: Informal networksof subcontracting in French industry. D. Gambetta, ed. Trust"Making and Breaking Cooperative Relations. Basil BlackwellLtd., Cambridge, MA, 194–210.

Lubell, M., J. T. Scholz. 2001. Cooperation, reciprocity, and thecollective-action heuristic. Amer. J. Political Sci. 45(1) 160–178.

Luhmann, N. 1979. Trust and Power. Wiley, New York.

Macaulay, S. 1963. Non-contractual relations and business: A prelim-inary study. Amer. Sociol. Rev. 28(1) 55–70.

MacNeil, I. R. 1980. Power, contract, and the economic model.J. Econom. Issues 14(4) 909–923.

MacNeil, I. R. 1983. Values in contract: Internal and external. North-western Univ. Law Rev. 78 340–418.

Maddala, G. S. 1983. Limited-Dependent and Qualitative Variablesin Econometrics. Cambridge University Press, Cambridge, UK.

Malhotra, D., J. K. Murnighan. 2002. The effects contracts on inter-personal trust. Admin. Sci. Quart. 47(3/4) 534–559.

Mariotti, S., G. C. Cainarca. 1986. The evolution of transaction gover-nance in the textile-clothing industry. J. Econom. Behav. Organ.7(4) 351–374.

Masten, S. E., J. Meehan, E. Snyder. 1989. Vertical integration inthe U.S. auto industry: A note on the influence of transactionspecific assets. J. Econom. Behav. Organ. 12 265–273.

Masten, S., J. Meehan, E. Snyder. 1991. The costs of organization.J. Law, Econom. Organ. 7 1–25.

Mayer, K. J. 1999. Buyer-supplier relationships in high technol-ogy industries. Ph.D. dissertation, University of California atBerkeley, Berkeley.

Mayer, K. J., J. A. Nickerson. 2005. Antecedents and performanceimplications of contracting for knowledge workers: Evidencefrom information technology services. Organ. Sci. 1(3) 225–242.

Monteverde, K., D. J. Teece. 1982. Supplier switching costs and ver-tical integration in the automobile industry. Bell J. Econom. 13206–213.

Nooteboom, B. 1996. Trust, opportunism and governance: A processand control model. Organ. Stud. 17 985–1010.

Nooteboom, B., H. Berger, N. Noorderhaven. 1997. Effects of trustand governance on relational risk. Acad. Management J. 40(2)308–338.

North, D. C. 1990. Institutions, Institutional Change and EconomicPerformance. Cambridge University Press, Cambridge, UK.

Nunnally, J. C. 1978. Psychometric Theory. McGraw-Hill, New York.

Parkhe, A. 1993. Strategic alliance structuring: A game theoreticand transaction cost examination of interfirm cooperation. Acad.Management J. 36 794–829.

Perrow, C. 1986. Complex Organizations" A Critical Essay, 3rd ed.Random House, New York.

Poppo, L., T. R. Zenger. 2002. Do formal contracts and relationalgovernance function as substitutes or complements? StrategicManagement J. 23(8) 707–725.

Ring, P. S., A. H. Van de Ven. 1994. Developmental processes ofcooperative interorganizational relationships. Acad. ManagementRev. 19(1) 90–118.

Rotemberg, J. J., G. Saloner. 1995. Overt interfunctional conflict (andits reduction through business strategy). RAND J. Econom. 26630–653.

Rousseau, D. M., S. B. Sitkin, R. S. Burt, C. Camerer. 1998. Notso different after all: A cross-discipline view of trust. Acad.Management Rev. 23 393–404.

Shelanski, H. A., P. G. Klein. 1995. Empirical research in transac-tion cost economics: A review and assessment. J. Law, Econom.Organ. 11(2) 335–361.

Siegel, S., N. J. Castellan, Jr. 1988. Nonparametric Statistics for theBehavioral Sciences. McGraw-Hill, New York.

Sitkin, S. B., N. L. Roth. 1993. Explaining the limited effectiveness oflegalistic “remedies” for trust/distrust. Organ. Sci. 4(3) 367–394.

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.


Tenbrunsel, A. E., D. M. Messick. 1999. Sanctioning systems,decision frames, and cooperation. Admin. Sci. Quart. 44684–707.

Venkatraman, N., J. H. Grant. 1986. Construct measurement in organi-zational strategy research: A critique and proposal. Acad. Man-agement Rev. 11(1) 71–87.

Walker, G., L. Poppo. 1991. Profit centers, single-source suppliers,and transaction costs. Admin. Sci. Quart. 36 66–87.

Walker, G., D. Weber. 1987. Supplier competition, uncertainty, andmake-or-buy decisions. Acad. Management J. 30(3) 589–596.

Weiss, L. 1984. The Italian state and small business. Eur. J. Sociol.25(2) 214–241.

Williamson, O. E. 1985. The Economic Institutions of Capitalism"Firms, Markets, Relational Contracting. The Free Press, NewYork.

Williamson, O. E. 1991. Comparative economic organization: Theanalysis of discrete structural alternatives. Admin. Sci. Quart. 36269–296.

Williamson, O. E. 1993. Calculativeness, trust and economic organi-zation. J. Law Econom. 36(April) 453–486.

Williamson, O. E. 1996. The Mechanisms of Governance. OxfordUniversity Press, New York.

Zaheer, A., N. Venkatraman. 1995. Relational governance as aninterorganizational strategy: An empirical test of the role of trustin economic exchange. Strategic Management J. 16 373–392.

Zaheer, A., B. McEvily, V. Perrone. 1998. Does trust matter? Explor-ing the effects of interorganizational and interpersonal trust onperformance. Organ. Sci. 9 141–159.

Zeller, R. A., E. G. Carmines. 1980. Measurement in the SocialSciences" The Link Between Theory and Data. Cambridge Uni-versity Press, New York.

Zucker, L. G. 1987. Institutional theories of organization. Annual Rev.Sociol. 13 443–464.

Zukin, S., P. DiMaggio. 1990. Structures of Capital" The SocialOrganisation of the Economy. Cambridge University Press,Cambridge, UK.

Cop

yright:INFORMSho

ldsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

institu

tiona

lsub

scrib

ers.

The

filemay

notb

epo

sted

onan

yothe

rweb

site,inc

luding

theau

thor’ssite.Pleas

ese

ndan

yqu

estio

nsrega

rdingthispo

licyto

perm

ission

s@inform

s.org.

Interorganizational Trust, Governance Choice, and Exchange ... · allow us to analyze...

Documents

Transcript of Interorganizational Trust, Governance Choice, and Exchange ... · allow us to analyze...