Divisive Faultlines and the Unplanned Dissolutions of Multipartner Alliances

This article was downloaded by: [193.51.16.194] On: 08 April 2014, At: 06:18Publisher: Institute for Operations Research and the Management Sciences (INFORMS)INFORMS is located in Maryland, USA

Organization Science

Publication details, including instructions for authors and subscription information:http://pubsonline.informs.org

Divisive Faultlines and the Unplanned Dissolutions ofMultipartner AlliancesRalph A. Heidl, H. Kevin Steensma, Corey Phelps

To cite this article:Ralph A. Heidl, H. Kevin Steensma, Corey Phelps (2014) Divisive Faultlines and the Unplanned Dissolutions of MultipartnerAlliances. Organization Science

Published online in Articles in Advance 04 Apr 2014

. http://dx.doi.org/10.1287/orsc.2014.0898

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial useor systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisherapproval. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitnessfor a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, orinclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, orsupport of claims made of that product, publication, or service.

Copyright © 2014, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, managementscience, and analytics.For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org

http://pubsonline.informs.org

http://dx.doi.org/10.1287/orsc.2014.0898

http://pubsonline.informs.org/page/terms-and-conditions

http://www.informs.org

OrganizationScienceArticles in Advance, pp. 1–21ISSN 1047-7039 (print) � ISSN 1526-5455 (online) http://dx.doi.org/10.1287/orsc.2014.0898

© 2014 INFORMS

Divisive Faultlines and the Unplanned Dissolutions ofMultipartner Alliances

Ralph A. HeidlMichigan State University, East Lansing, Michigan 48824, [email protected]

H. Kevin SteensmaUniversity of Washington, Seattle, Washington 98195, [email protected]

Corey PhelpsHEC Paris, 78351 Jouy-en-Josas, France, [email protected]

Received wisdom suggests that multipartner alliances are relatively unstable because of their complexity and theincreased potential for free riding. Nonetheless, multipartner alliances do benefit from built-in stabilizing third-party

ties that mitigate opportunism and conflict between partner pairs. Previous empirical research on multipartner alliance sta-bility has been inconclusive. We shed some light on these inconsistencies by recognizing that within multipartner alliances,schisms can occur not only between a pair of partners but also between subgroups of partners that are divided by faultlines.We suggest that divisive faultlines can form between subgroups of partners within a multipartner alliance as a functionof their prior experience with one another. When a subgroup of alliance partners has relatively strong ties to each otherand weak ties to other partners, destabilizing factions can develop that hamper reciprocity among the partners. Using alongitudinal analysis of 59 multipartner alliances, we found that, in general, faultlines (as modeled by the dispersion oftie strength within multipartner alliances) increase the hazard of unplanned dissolutions. We also found that multipartneralliances comprising a mix of centrally and peripherally positioned partners within the industry network were less apt tosuffer the effects of divisive faultlines. We suggest that this is due to the greater opportunity costs of dissolution and thepresence of relatively high-status partners who can act as peacekeepers and coordinators of their lower-status partners.

Keywords : interorganizational relations; multipartner alliance; embeddedness; faultlinesHistory : Published online in Articles in Advance.

IntroductionAlliances between firms play a critical role in achiev-ing technological and economic objectives. Whereasthe simplest alliances involve two partners, multipartneralliances can be particularly effective for completinglarge-scale development projects requiring the coordi-nation and resources of multiple firms (Beamish andKachra 2004). A multipartner alliance is a single coop-erative agreement involving three or more firms boundedby a unifying goal and governed by a single overarch-ing contract (Das and Teng 2002a, Lavie et al. 2007, Liet al. 2012). Such structures are distinct from alliancenetworks comprising relatively autonomous dyadic tiesestablished through independent contracts for a diversearray of goals. Various characteristics of alliance net-works, and firms’ positions within these networks, havebeen found to influence alliance formation and theirstability (e.g., Gulati and Gargiulo 1999, Gulati et al.2012, Polidoro et al. 2011, Rosenkopf and Padula 2008).However, individual alliances in an alliance network canform and dissolve relatively independent of each other,compared with the more tightly coupled and interdepen-dent ties that compose a multipartner alliance. Becausefirms in a multipartner alliance are bound together by

one overriding contract and goal, and they depend oneach other’s contributions to achieve alliance objectives,a schism between any two partners in a multipartneralliance jeopardizes the entire alliance and may lead tothe dissolution of all the ties within the alliance.

Notwithstanding their potentially tenuous nature, mul-tipartner alliances remain popular. Studies have reportedthat 30%–50% of all alliances have three or more part-ners (e.g., Garcia-Canal et al. 2003, Makino et al.2007). The use of multipartner alliances is particularlywidespread in high-technology industries. For example,companies including Google, T-Mobile, Qualcomm, andMotorola joined forces to develop the first truly openand comprehensive platform for mobile devices, alter-ing the competitive landscape in the telecommunicationequipment industry (Helft and Markoff 2007).

Multipartner alliances differ substantially from two-partner alliances in their social exchange processes (Dasand Teng 2002a). In contrast to the direct reciprocityfound in two-partner alliances, successful multipartneralliances often require indirect reciprocity where obliga-tions to individual partners are generalized to the entiregroup. For example, in a three-partner alliance consist-ing of Firms A, B, and C, any quid pro quo benefits

1

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Heidl, Steensma, and Phelps: Divisive Faultlines and the Unplanned Dissolutions of Multipartner Alliances2 Organization Science, Articles in Advance, pp. 1–21, © 2014 INFORMS

due to Firm A for providing resources to Firm B mayneed to come from Firm C. Because of the complex-ity associated with organizing multiple member firmstoward a common goal, and the need for indirect reci-procity, multipartner alliances have greater coordinationand monitoring challenges compared with two-partneralliances (Garcia-Canal et al. 2003, Li et al. 2012, Zengand Chen 2003). Since the effectiveness of formal con-tracts is thought to be limited, broad-based trust acrossall partners may be particularly important to keep part-ners from free riding on the others (Zeng and Chen2003). Because of their complexity and unique chal-lenges, received wisdom within the alliance literature isthat multipartner alliances are inherently less stable thantwo-partner alliances (e.g., Beamish and Kachra 2004,Garcia-Canal et al. 2003). Some empirical evidence sup-ports this notion (Dussauge et al. 2000).

However, insights from social network theory suggestthat having multiple partners in an alliance may enhancestability. A distinctive feature of multipartner alliances isthat they have built-in third-party ties linked to every pairof partners. Common third parties can monitor partnerbehavior, deter opportunistic behavior, resolve conflictsbetween other partners, promote trust, and prevent desta-bilizing schisms from occurring (Gulati 1998). Thirdparties within multipartner alliances may be particularlymotivated to resolve conflicts that jeopardize an alliancein which they have a direct stake (Rosenkopf and Padula2008). Some empirical evidence supports the notion thatmultipartner alliances are particularly stable (Park andRusso 1996).

Because there is a lack of theoretical and empiricalclarity regarding the stability of multipartner alliances—alliance structures that are prevalent and distinct fromboth alliance networks and two-partner alliances—additional insight is needed. We provide nuance onthe subject of multipartner alliance stability by recog-nizing that within multipartner alliances, schisms canoccur between not only a dyad of partners but alsosubgroups of partners that are divided by faultlines.These are divisions that form within groups when sub-groups of members share particularly cohesive ties toeach other based on common experiences and identity(Lau and Murnighan 1998). Faultlines within groups ofindividuals, modeled using such attributes as gender,ethnicity, and common outside affiliations, are shownto increase conflict and hamper group performance (Liand Hambrick 2005). Similarly, we suggest that divi-sive faultlines can form between subgroups of partnerswithin a multipartner alliance as a function of their priorexperience with each other. When some alliance part-ners have relatively strong ties to each other because ofprior relationships, and weak ties to other partners, theyhave the potential to coalesce and rely on familiar normsand routines established during their prior relationships.This reliance on old ways can come at the expense of

establishing the broad-based routines and trust across allpartners in the alliance needed for indirect reciprocityand generalized social exchange (Das and Teng 2002a).

We investigated this proposition using a sample ofmultipartner technology alliances in the global telecom-munication equipment industry. Consistent with ourargument, we found that faultlines, as modeled by thedispersion of tie strength within multipartner alliances,increased the rate of unplanned dissolution. However, wealso found that multipartner alliances comprising firmsthat varied in their network centrality within the broaderindustry alliance network were somewhat resilient tothe destabilizing influence of tie strength dispersion.We suggest this resiliency results from the greater oppor-tunity costs of dissolution as a result of relatively distinctknowledge being contributed, and the presence of rela-tively high-status partners who can act as peacekeepersand coordinators.

Our study contributes to both the alliance and net-work literatures. Multipartner alliances have been inte-grated into larger industry network studies where thedyads that comprise multipartner alliances are accountedfor when modeling alliance networks (e.g., Gulati et al.2012). However, studies where the level of analysis isthe alliance and the alliances being studied comprisemultiple partners are rare (e.g., Li et al. 2012). Researchon alliance networks has shown how relational (i.e.,strong ties), structural (i.e., third-party ties), and posi-tional (i.e., network centrality) embeddedness influencealliance dynamics (e.g., Ahuja et al. 2009, Gulati et al.2012, Polidoro et al. 2011). Despite the notion thatmultipartner alliances are dense clusters of firms thatexhibit network characteristics (Rosenkopf and Padula2008), fundamental network concepts have not beenused to explore their stability. To our knowledge, thisstudy is the first to consider how the configurationsof alliance partner embeddedness influence multipart-ner alliance stability. Such a dearth may stem from theconceptual and empirical complexities inherent in thestudy of multipartner alliances. Any fine-grained analy-sis of relational embeddedness in multipartner alliancesrequires a complete collaborative history of all thoseinvolved. Consequently, scholars have tended to focus ondyadic relationships and the networks derived from them(Ahuja 2000, Gulati 1995b, Gulati and Gargiulo 1999,Gulati et al. 2000, Polidoro et al. 2011, Stuart 2000,Xia 2011), or they have controlled for the differencesbetween two-partner and multipartner alliances using asimple dummy variable (e.g., Dussauge et al. 2000) orcount of partners (e.g., Garcia-Canal 1996).

In their recent review of faultline literature, Thatcherand Patel (2012) proposed that firm-level faultlines beexplored. We do so by extending faultline research fromgroups of individuals (Lau and Murnighan 1998, 2005;Li and Hambrick 2005) to groups of firms, integrat-ing insights from this literature with those from social

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Heidl, Steensma, and Phelps: Divisive Faultlines and the Unplanned Dissolutions of Multipartner AlliancesOrganization Science, Articles in Advance, pp. 1–21, © 2014 INFORMS 3

embeddedness research. Strategy scholars have a pro-clivity for adapting individual-level concepts such asabsorptive capacity (Cohen and Levinthal 1990), learn-ing (Argyris and Schon 1978), and relational ties (Gulati1995b) to the firm level. Our conception of faultlineswithin multipartner alliances as a function of the disper-sion of partners’ tie strength expands on the notion thatfirms are capable of developing trusting and cohesiverelationships (Gulati 1995a). To the extent that firms canforge strong ties, divisive faultlines resulting from theconfiguration of those ties are also plausible.

Extant research has emphasized the benefits of strongties between firms in terms of mitigating risks ofopportunism, reducing contracting costs, and promot-ing relational stability (e.g., Uzzi 1996, Zaheer andVenkatraman 1995). Our research suggests that there isa dark side to strong relationships between firms whensuch cohesive ties within multipartner alliances engen-der divisive faultlines that impede generalized socialexchange. Our work explores the interplay between part-ners’ centrality within the broader alliance network andtheir tie strength as it pertains to multipartner alliances.Consistent with prior work (Polidoro et al. 2011), wefind that the different types of embeddedness (i.e., strongties, network centrality) are not merely additive or mutu-ally reinforcing, but rather influence the stability of mul-tipartner alliances in more complex ways than whatwould be expected based on dyadic relationship studies.

Our unfolding story and empirical results areconsistent with a recent framework suggesting howinertia, exogenous pressures, agency, and opportunityinfluence network evolution (Ahuja et al. 2012). First,inertial forces drove firms in our sample to becomehighly embedded in a subset of partners with whomthey feel comfortable and familiar. Exogenous pres-sures from technological and regulatory changes withinthe telecommunications equipment industry in the early1990s rendered existing collaborative networks inade-quate for addressing new demands. Agency and oppor-tunity cajoled incumbent firms to form multipartneralliances with sometimes unfamiliar partners, creatinglinks between central and peripheral firms. However, thelong-term influence of multipartner alliances on industrynetworks depend on their stability.

Finally, our research responds to calls for the develop-ment of methodologies and measures to more effectivelyassess the influence of clusters on larger sparsely con-nected network structures (Rosenkopf and Padula 2008).The concepts and measures we have used can be appliedto other mesolevel network structures such as clustersand multilateral agreements.

We begin by reviewing the literature on the forma-tion and stability of multipartner alliances. We thendevelop the notion of divisive faultlines within multi-partner alliances, based on the dispersion in relational tiestrength and how dispersion in partner positional embed-dedness can alter the influence of such faultlines.

Review of Multipartner AllianceFormation and Their StabilityMultipartner technology alliances are formed to bringtogether the diverse resources needed for the novelrecombination of knowledge. Such alliances have thepotential to earn greater returns than two-partneralliances, because larger pools of diverse resources canlead to the development of particularly unique prod-ucts and services that are difficult to imitate (Beamishand Kachra 2004). However, fully leveraging these largepools of resources requires the development of gener-alized social exchange and indirect reciprocity amongthe partners. With generalized social exchange, there isoften not a one-to-one correspondence between whatany two partners in a multipartner alliance give to andreceive from each other (Das and Teng 2002a). Part-ner A can contribute a resource to partner B withoutnecessarily obliging partner B to directly reciprocate.Instead, partner B fulfills its obligation by contributingin roughly equivalent value to some other partner in thefuture (Bearman 1997). In other words, partner A’s con-tribution to partner B may eventually be reciprocated bypartner C. Thus, exchange is generalized to the allianceand not individual partners. However, just when sucha reciprocal payment will occur may be uncertain andoften comes with no explicit guarantee of reciprocation(Takahashi 2000).

Partners involved in generalized exchange have anincentive to free ride to reduce their costs of coop-eration or avoid wasting their contribution if othersdo not cooperate. Because of the particularly complexexchange processes in multipartner alliances, free rid-ing can be difficult to detect (Takahashi 2000, Zeng andChen 2003). When such a practice is detected, estab-lishing sanctions against those identified is challengingbecause once the alliance is established, no partner maybe excluded from realizing some of the benefits regard-less of their contribution (Zeng and Chen 2003). Thus,for generalized exchange to function properly withoutfree riding, there needs to be broad-based trust amongall the partners. Not only must partner A trust partner Bto adequately contribute to partner C, it must trust part-ner C to reciprocate benefits to itself. Without trust, indi-rect reciprocity breaks down, making it difficult to fullyexploit larger pools of resources that typify multipartneralliances. When this occurs, the value of multipartneralliances decrease, and the likelihood of their dissolutionis enhanced.

Despite the inherent incentives to free ride, studiesexploring the influence of having multiple partners onalliance stability and performance have been mixed.Consistent with the notion that multipartner alliances areintrinsically unstable, Dussauge et al. (2000) found intheir sample of more than 200 manufacturing alliancesthat those involving more than two partners were morelikely to dissolve. However, Beamish and Kachra (2004)

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


found in a large sample of Japanese alliances that thenumber of partners generally had little influence on out-comes. Contrary to their expectations and the belief thatmultipartner alliances are relatively complex and unsta-ble, Park and Russo (1996) found that the duration ofelectronics alliances in their sample actually increasedwith the number of partners.

Three or more partners in an alliance may indeedenhance stability. Multipartner alliances provide a con-text where trust and cooperation may be promotedbecause multiple members can mutually monitor eachother’s behavior and mediate conflict that occursbetween members (Rosenkopf and Padula 2008). A keyattribute of multipartner alliances is the stabilizing third-party ties built into their structure. Such alliances areby definition structurally saturated (all partners areconnected to all other partners) and comprise triads.A three-party alliance comprises one triad, a four-partyalliance comprises four triads, and so forth. Thus, eachdyad within a multipartner alliance is embedded withinat least one triad and stabilized by the conflict resolu-tion capabilities of third parties (Simmel 1950). Third-party ties within sparsely connected industry networkshave been found to enhance the stability of two-partneralliances (Polidoro et al. 2011). Third parties within mul-tipartner alliances are likely to be even more effectiveat resolving internal conflict and stabilizing relationshipsfor two reasons. First, a conflict between two partnerswill be relatively transparent to other members of thesame alliance. Second, third parties within a multipartneralliance are likely to be highly motivated to resolve anyconflict that could jeopardize the alliance in which theyhave a direct and sizable stake. As third parties monitorand stabilize the various dyadic relationships within mul-tipartner alliances, free riding is held in check and normsof indirect reciprocity develop. Resources will flow morefreely between partners when they can trust that indirectreciprocity will occur, enabling the alliance to reach itspotential for creating value.

Such informal governance within multipartner alli-ances can make them particularly well suited for cen-trally positioned firms within an industry network whowish to safely partner with peripherally positionedand less familiar firms that possess valuable resources(Rosenkopf and Padula 2008). As a rule, firms thatare highly central in the industry network as the resultof maintaining a broad array of alliances tend to allywith other firms that are centrally positioned becausetheir quality and reliability are well established (Chunget al. 2000, Gulati and Gargiulo 1999, Podolny 1994).Although the benefits of partnering with other cen-tral firms are genuine, over time, knowledge amonghighly interdependent and centrally positioned part-ners can become redundant and hamper firm perfor-mance (Ahuja et al. 2009, Uzzi 1997). Thus, despitetheir general propensities to play it safe (Gulati 1995b,

Podolny 1994), centrally positioned firms must oftenpartner with peripheral firms to break out of the redun-dancy trap and renew their capabilities.

The respective capabilities and desires of central andperipheral firms can drive them together. Firms that arecentral to the overall industry network can use theirconnections to identify peripheral firms with valuableinformation (Ahuja et al. 2009). Those firms that areperipheral may be interested in allying with central firmsto enhance their own status (Rosenkopf and Padula2008). Nonetheless, forming alliances with peripherallypositioned and untested firms remains a relatively riskyproposition for centrally positioned firms. Rosenkopfand Padula (2008) observed that many of the multi-partner alliances in their study comprised a mixture ofpartners in terms of their positional embeddedness inthe industry network. Some were centrally positioned,whereas others were peripherally positioned. Althoughthey did not observe the outcomes of these mixedalliances, they surmised that there may be safety in num-bers. Because multipartner alliances are structurally sat-urated with built-in stabilizing third-party ties, they canbe used by centrally positioned firms to buffer the riskof reaching out to unfamiliar partners who have valuableknowledge.

In summary, multipartner alliances have the potentialto create substantial value by bringing together largepools of diverse resources. Fully exploiting such poten-tial depends on generalized social exchange processesthat require broad-based trust and norms of indirectreciprocity among all partners. Although multipartneralliances are thought to be inherently unstable because oftheir complexity and incentives to free ride, their built-inthird-party ties can also mitigate schisms from occur-ring, thereby enabling trust and norms of reciprocity toevolve. All multipartner alliances are structurally satu-rated and enjoy the potential benefits of third-party ties.Within a particular multipartner alliance, the number ofinternal third-party ties does not vary across the firmpairs. What does vary is the strength of ties between thevarious partners within an alliance, as well as the con-figuration of tie strength across multipartner alliances.

Tie strength variability within an alliance has impli-cations for its stability. Destabilizing schisms withinmultipartner alliances can occur not only betweendyads of firms, but between subgroups as well. Thesehigher-level schisms are more likely when particularlycohesive subgroups resulting from strong ties createdivisive faultlines, which can hamper the developmentof the broad-based trust necessary for generalized socialexchange.

Tie Strength Dispersion WithinMultipartner AlliancesIndividuals generally prefer to interact with those withwhom they have trusting and cohesive relationships,

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


without fear of being taken advantage of (Granovetter1985). However, within larger groups, such cohesivenesscan also lead to divisive faultlines. Members may havevarying levels of trust and kinship toward each another.Faultlines occur when particularly cohesive subgroupsexist within a larger group (Lau and Murnighan 1998),and in-group/out-group factionalism creates mistrust andanimosity within the larger group (Pearce 1997). Thosewho coalesce as an in-group tend to identify more withtheir subgroup than with the larger group (Polzer et al.2006). Factions that operate outside the formal structureof the group can advance their interests by threateningto withdraw pooled resources and manipulating gover-nance decisions (Ariño and de la Torre 1998, Brass andBurkhardt 1993). The mere potential for adverse influ-ence by a faction can cause a loss of trust by thosewho are not part of the faction. Whether perceived orreal, such factions diminish the overall sense of fairnessand equity among the members (Li and Hambrick 2005,Polzer et al. 2006).

Past research has proxied potentially divisivefaultlines by observing the distribution of sharedattributes among individual members of a group (Lauand Murnighan 2005, Li and Hambrick 2005). Thoseof the same gender, ethnicity, or age group may forma cohesive subgroup because, as a function of theircommon attributes, they are likely to have had similarexperiences in their lives and can more closely identifywith each other (Lau and Murnighan 1998). Similari-ties in background tend to elicit trust and create cohe-sive bonds (Brewer and Brown 1998). If every individual(or none) within a group shares the same attributes inquestion (e.g., similar age), faultlines and subgroup fac-tions are unlikely to form. Instead, it is moderate diver-sity within a group of individuals that provide a basisfor divisions that may adversely affect group dynam-ics (Lau and Murnighan 1998). Lau and Murnighan(2005) found that members of subgroups based on gen-der and ethnicity tended to be biased when assessingthe contributions, behavior, and trustworthiness withintheir subgroups compared with those outside of them.Li and Hambrick (2005) explored faultlines based onage, tenure, gender, and ethnicity within joint venturemanagement groups and found that strong faultlinesincreased conflict between subgroups, leading to poorperformance.

Thus differences in the extent to which members haveshared experiences are the basis for faultline forma-tion. These experiences can be shared indirectly (e.g.,those in similar age or ethnic cohorts are likely to havesimilar life experiences) such that demographic group-ings can be used to proxy differences in shared experi-ences within the group. Experiences may also be shareddirectly through a history of interaction. As partnersprove themselves to be trustworthy and honest overthe course of previous direct interactions, they develop

strong relational ties such that they identify with oneanother’s needs, preferences, and priorities (Lewicki andBunker 1996). However, similar to the effects of sub-groups based on demographic attributes, particularlystrong ties among individuals based on a history of priorinteractions can divide a larger group into subgroups andlead to counterproductive dynamics.

Strong ties based on past interactions can formbetween firms as well. Sociologists have long pointed tonumerous examples of preferential and stable bilateraltrading relationships to show that firms bond with otherfirms when they repeatedly interact (e.g., Piore and Sabel1984). Close personal ties may develop between individ-uals at collaborating firms, putting pressure on each firmto meet the expectations of the other (Macaulay 1963).As firms engage over time, and interpersonal ties growstronger, trust develops between firms (Gulati 1995a,Ring and Van de Ven 1994). Over the course of multiplerelationships, expectations of reciprocity develop, reduc-ing concerns over equity and fairness, and encouragingpartners to share information, further strengthening theirbond. Communication patterns between firms becomeroutine and norms develop for resolving conflicts (Dyerand Singh 1998, Zollo et al. 2002). The cohesivenessthat results from prior relationships between two firmstranscends the specific managers involved in the priorrelationships. Although there may be turnover in man-agement, new managers are socialized in terms of whatto expect from firm partners (Gulati and Nickerson2008). Thus, even if individual managers at partnerfirms have not interacted directly before, the prior tiesbetween their respective firms provide a basis for good-will between them (Gulati 1998). In essence, being affil-iated with firms that have interacted in the past affordsa salient common attribute for otherwise disparate indi-viduals that can lead to a sense of solidarity.

To the extent that pairs of partner firms develop strongties from prior interactions, faultlines can develop withinmultipartner alliances as a function of the differencesin tie strength across partner pairs. Potentially divi-sive faultlines form when some firms in a multipart-ner alliance share strong ties with each other but notwith others. Under such conditions, pockets of highlycohesive partners (i.e., in-groups) form, as well as part-ners relegated to out-groups. In contrast, if a multipart-ner alliance comprises firms where tie strength is eitheruniformly high (i.e., every firm has worked extensivelywith each other in the past) or uniformly low (i.e., allfirms are complete strangers), there is little potential fordivisive faultlines based on differences in tie strength.The potential for divisive faultlines occurs only when tiestrength is nonuniform across the multiple dyads.

Similar to individuals who make up cohesive sub-groups within a larger group, the managers from part-ner firms who make up particularly cohesive subgroupswithin a multipartner alliance may identify more with

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


the subgroup than the alliance. Norms of exchange andcommunication routines established during prior rela-tionships may persist at the expense of establishinginclusive norms that involve all partner firms and man-agers. Multipartner alliances are particularly susceptibleto faultlines and factions in the early stages of theirexistence when members are trying to gauge each otherand the task at hand (Lau and Murnighan 1998). Duringtimes of uncertainty, managers from member firms mayseek security with other managers whose firms share ahistory of collaboration and preexisting norms. Cohe-siveness beyond the subgroup may become constrained,leading to mistrust and a sense of inequity between thosein and outside of the subgroup.

Such sentiments among the managers and personnelrepresenting their respective firms can have dire con-sequences for alliance stability. When faultlines reducesolidarity and impede the development of broad-basedtrust in the alliance, they decrease the viability ofgeneralized exchange. The failure to develop normsof generalized social exchange and indirect reciprocityin a multipartner alliance compromises its potentialvalue. Unmet performance expectations and mistrust canfuel conflict among partners, influencing their decisionsregarding the future of the alliance (Li and Hambrick2005). When partners perceive the collaborative poten-tial to go unfulfilled and that collaborative activities arenot fair, they reduce their commitment to the alliance ordissolve it altogether (Ariño and de la Torre 1998).

The potential for faultlines within a multipartneralliance is greater when the strength of the variousdyadic ties within the alliance is particularly dispersed.Thus, we expect increasing tie strength dispersion tohamper the stability of multipartner alliances and pro-mote their unplanned dissolution.

Hypothesis 1. The greater the dispersion of tiestrength within a multipartner alliance, the greater thehazard of unplanned dissolution.

Positional Embeddedness Dispersion WithinMultipartner AlliancesAlthough faultlines within multipartner alliances pro-vide a basis for subgroup factions and internal conflict,whether these faultlines lead to schisms and dissolutionwill depend in part on the opportunity costs of dissolv-ing and the presence of relatively high-status partnerswho can act as peacekeepers and strong coordinators.

In general, conflicts between parties, whether indi-viduals, groups, or nations, are likely to be resolvedwhen there are substantial costs to all parties withescalating them (Deutsch 1994, Greig 2001). Das andKumar (2009) suggested that alliance partners areparticularly resilient to internal conflict when potentialeconomic benefits from maintaining cooperation are sub-stantial, or if alliance dissolution will adversely affect

the reputation of those involved. Under such conditions,the indiscretions of partner firms are often overlooked.

Accordingly, when there are particularly high oppor-tunity costs for the various partner firms in dissolvinga multipartner alliance, managers are likely to tolerateor cope with potentially divisive faultlines. The eco-nomic opportunity costs from dissolving technologydevelopment multipartner alliances may vary depend-ing on the composition of partners. Creative and eco-nomically valuable recombinatorial opportunities are ata peak when there is a diverse array of knowledgeto draw on (Rosenkopf and Almeida 2003). Centrallypositioned firms in the industry network often createalliances with those on the periphery to gain accessto new knowledge and combat knowledge redundancythat occurs when centrally positioned firms interact onlywith each other (Gulati and Gargiulo 1999). Multipart-ner alliances that include a mixture of centrally andperipherally positioned firms within the industry net-work can create pools of particularly diverse knowl-edge with high recombinatorial potential. All else beingequal, the value creation opportunities from multipart-ner alliances comprising a mix of partners exhibit-ing highly dispersed positional embeddedness withinthe industry network will be substantial. There arealso reputation-based opportunity costs of dissolution.The reputations of firms peripherally positioned withinthe industry network may be enhanced by affiliating withhigh-status firms that are centrally positioned (Podolny1994). Managers of peripherally positioned firms withinmultipartner alliances that also include centrally posi-tioned partners will be reluctant to destabilize an alliancethat benefits them through connections to high-statusfirms.

In contrast, recombinatorial opportunities from alli-ances comprising primarily centrally positioned firmsare relatively limited because information and resourceswithin these alliances are less diverse than those foundwithin alliances involving both centrally and peripherallypositioned firms. When word of failure spreads to thosethey are connected with, the reputations of centrallypositioned partners may be harmed by alliance disso-lution. However, their relative prominence also enablesthem to attract other partner firms to form alternativealliances to the one in question. Thus the opportunitycosts of dissolving multipartner alliances comprising pri-marily centrally positioned firms is somewhat limited.The opportunity costs of dissolving alliances comprisingonly peripheral partners will also be limited. There islittle opportunity for the peripheral partners to enhancetheir prestige by maintaining ties to high-status firmswhen there are no centrally positioned partners involved.In addition, the sparse connectivity among peripherallypositioned firms reduces reputation-based opportunitycosts, since information regarding uncooperative part-ner behavior is not diffused so readily among potentialpartners.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Conflict within a group of partnering firms can alsobe resolved through the efforts of a strong lead partner.1

When a set of firms has a common purpose, and onefirm dominates in terms of resources and status, thatfirm often becomes the hub firm, providing the gover-nance and discipline necessary for the group to achieveits goals (Provan et al. 2007). Such lead partners arelikely to emerge in multipartner alliances with membersthat are highly dispersed in terms of positional embed-dedness. Under these conditions, some partners will beof relatively higher status than the remaining partnersbecause of their centrality within the broader industrynetwork. Such partners can reduce conflict in two ways.First, they can be particularly effective third-party peace-keepers in conflicts involving deferential lower-statuspartners. Second, they can assume the role of centralcoordinators among lower-status partners, solving prob-lems as they arise before conflict festers.

In sum, the high opportunity costs of dissolution andlead partner governance associated with multipartneralliances with members that are highly dispersed in theirpositional embeddedness may be able to mitigate desta-bilizing internal conflict. Such mitigation will be partic-ularly valuable when there are faultlines. When conflictfrom faultlines occurs, partners examine the costs andbenefits of dissolution. In such make-or-break situations,a stabilizing “reset” moment is more likely to occurwhen the opportunity cost of dissolution is high andthere is a strong lead partner that can foster peace. Whenmultipartner alliance memberships are concentrated intheir positional embeddedness, neither a high oppor-tunity cost for dissolution nor a relatively high-statuspeacekeeper will exist to dampen instability because offaultlines. Thus, the destabilizing effects of faultlines asa result of tie strength dispersion will be particularly pro-nounced when there is little dispersion in the positionalembeddedness of the alliance membership and less sowhen members are highly dispersed in their positionalembeddedness.

Hypothesis 2. The greater the dispersion in posi-tional embeddedness across partners in a multipartneralliance, the weaker the influence of tie strength disper-sion on the hazard of an unplanned dissolution.

MethodsSample and DataOur research setting was the global telecommunicationsequipment industry. Firms in this industry produce andmarket hardware and software that enable the transmis-sion, switching, and reception of voice, images, and dataover short and long distances using digital, analog, wireline, and wireless technology. This industry was chosenbased on the significant changes in technology and com-petition it had experienced over the past three decades,

resulting in a growing number of technology alliancesinvolving incumbent firms (Amesse et al. 2004), partic-ularly those with more than two partners.

Many practical considerations guided the constructionof our sample. Since we employ a longitudinal design,collaboration data from multiple data sources were com-piled for the 1982–1997 time period. Our sample periodwas then limited to 1987–1997 to minimize left cen-soring. The sample frame of potential partners coveredonly public companies to ensure the availability and reli-ability of archival data. We further limited the sampleframe to the industry’s top-selling firms, which offermore complete and accurate alliance data than smallerfirms (Gulati 1995b). To minimize survivor bias, the topfirms in the industry at the beginning of the study periodwere used rather than those who dominated the indus-try at the end, since mergers, restructurings, and failureshad taken place throughout the study period (Amesseet al. 2004). To compute our measures of tie strength, wefollowed prescriptions for establishing network bound-aries in empirical research (Laumann et al. 1983) andrestricted the network to firms and alliances that cen-tered on the telecom industry, similar to criteria used inrecent alliance network research (Phelps 2010, Schillingand Phelps 2007). These restrictions produced a sampleof 104 potential partner firms involved.

We used several sources for the alliance data.Specifically, data on our sample firms were collectedusing the Securities Data Company (SDC) joint ven-ture and alliance database along with systematic archivalresearch on annual reports, 10-K and 20-F filings,Moody’s Industrial Manual and Moody’s InternationalManual, and three electronic databases: Factiva, Lexis-Nexis, and Dialog. We recorded only collaborationsthat could be confirmed by multiple sources. Duringour period of study, the 104 sample firms initiated7,978 alliances, including two-partner and multipartneralliances. Of the 7,987 alliances, 1,089 were multipart-ner alliances (13.63%). Of these, 95 involved three ormore of the 104 sample telecommunications firms wehad initially identified. However, we had to further limitour sample of multipartner alliances to those exclusivelycomprising our panel firms because we needed the com-plete alliance history for every participant in each of themultipartner alliances in our final sample to computetie strength measures. Fifty-nine multipartner alliancesinvolving 46 sample firms that operated during 1987–1997 matched this stringent criterion.

Using a panel design, we observed our sample mul-tipartner alliances and their partner firms over time.The unit of observation was the multipartner alliance-year. Data were collected on an annual basis so thatrecords for every technology alliance maintained bythe sample firms existed at the end of each sampleyear. All network embeddedness variables in this studywere calculated based on a panel of networks anchored

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


by 104 panel firms, which formed (two-partner andmultipartner) 7,978 alliances between 1982 and 1996.To assess prior ties and compute our measures associ-ated with tie strength and positional embeddedness fora given year, we considered all alliances (two-partnerand multipartner) formed in the previous five years.Consequently, we collected alliance data for each firmbeginning in 1982 and researched each sample multi-partner alliance to identify its date of dissolution or con-tinuation through 1997.

The final panel comprised 59 multipartner alliancesand 251 alliance-year observations. We counted 25 ter-minations during the observation period. Since five ofthese were planned, 20 multipartner alliances (or 34%of our sample) experienced a failure event. On average,each alliance was at risk of dissolution for 4.25 years.Sample alliances ranged from 3 to 10 member firms,with an average size of 4.13.

Measurement: Dependent Variable

Multipartner Alliance Dissolutionkt . The dependentvariable in our study was the hazard of an unplannedalliance dissolution. Such dissolutions often reflect prob-lematic collaborations (Polidoro et al. 2011). The oper-ationalization of alliance dissolution required a two-stepprocess. First, we needed to systematically identify thedate of dissolution, or continuance for each alliance, bythe end of the sample period (1997), and second, we hadto determine whether a dissolution event was plannedor unplanned. To determine the dissolution date, wefollowed the procedure implemented by Ahuja (2000)and contacted company personnel to identify dissolu-tion dates, which proved useful in identifying the status(termination or ongoing) of joint ventures. We obtainedinformation for nearly all the joint ventures regardingthe month of a venture’s termination and whether itstill existed at the end of the sample period. Wherethe termination date of a joint venture was unavailable,we assumed dissolution to have occurred the year afterthe last documentation or the year after announcement,whichever was later. For non-joint venture alliances, dis-solution was based on specified tenure mentioned inarchival sources or an announcement of dissolution ineither archival sources or by a company contact. In theabsence of specific dissolution data, an alliance wasassumed to exist through the end of the final year forwhich there was documentation or end of year it wasfounded, whichever came later. Statistical tests for differ-ences in mean duration between alliances with dissolu-tion announcements and those with assumed dissolutiondates yielded no significant differences, supporting thevalidity of this approach.

Since our theoretical arguments linked alliance dis-solution events to collaborative difficulty, the circum-stances prompting the end of a collaborative agreement

were of interest (Makino et al. 2007, Polidoro et al.2011). To distinguish unplanned from planned disso-lution events, we conducted in-depth media searchesand cross-referenced evidence from (a) SDC deal-textstatements, (b) press releases taken from company web-sites, and (c) news stories published in such elec-tronic databases as Factiva and LexisNexis. Wheredetailed information about a dissolution event couldnot be found, and no tenure was specified in thealliance announcement, we assumed the termination wasunplanned. This procedure yielded 20 unplanned dis-solutions, 5 planned dissolution events where collab-orations ended as a result of changes in the externalenvironment (e.g., product markets and technologies) orproject completion (e.g., a successful product launch),and 34 multipartner alliances that existed beyond ourwindow of study. Examples of the archival evidenceused to determine whether dissolution was plannedor unplanned are provided in Appendix A. The aver-age duration of multipartner alliances that experiencedunplanned dissolution was 2.45 years, with a standarddeviation of 2.33 years.

We measured multipartner alliance dissolution usinga time-varying dummy variable set to 0 for every timeperiod t in which alliance k existed at the end of the yearand 1 otherwise. An alliance with a planned breakup wasdesignated in our data set as censored.

Measurement: Explanatory Variables

Tie Strength Dispersionkt−1. Li and Hambrick (2005,p. 797) noted that the “measurement of faultlines isdaunting.” Previous research has dealt with this diffi-culty by relying on contrived contexts. For example,Lau and Murnighan (2005) used a randomized blockexperimental design to create clearly defined faultlineswithin groups of individuals. Perhaps most analogous toour context is the work done by Gibson and Vermeulen(2003) on subgroup strength within teams of individ-uals. They first assessed the overlap between pairs ofindividuals in terms of various demographic dimensionsand then computed the dispersion in the overlap acrossthe different pairs on a team. In a similar fashion, weconsider the tie strength between pairs of firms within amultipartner alliance and compute the dispersion in tiestrength.

To assess tie strength dispersion within each multi-partner alliance k in year t − 1, we counted the numberof prior ties formed by each dyad in the multipartneralliance for the year t−1. A prior tie is defined as directcontact between a pair of firms through either a two-partner or multipartner alliance (Gulati 1995b). Follow-ing prior alliance research, we used a five-year movingwindow (i.e., t− 6 to t− 1) to identify prior ties (Gulatiand Gargiulo 1999). This practice implies that only rela-tionships formed in the previous five years affect cur-rent behavior. We weighted each prior tie based on the

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


scope of activities that occurred in the prior alliance.Weights ranged from 1 to 4 depending on the count ofthe following activities: technology codevelopment, pro-duction, marketing, and technology licensing. We thencomputed tie strength variance across multiple dyadswithin each sample multipartner alliance for each year.Variance is essentially a measure of dispersion. A valueof 0 indicates that tie strength is equal across all partnerpairs. Higher values of variance indicate that tie strengthwithin a multipartner alliance is concentrated among asubset of partner pairs. A logarithmic transformation wasused to enhance the normality of this measure.2 Otherpotential measures of dispersion such as the Herfindahlindex are essentially functions of the measure of vari-ance but may be highly biased in relatively small sam-ples such as our alliance partner pairs.3

Positional Embeddedness Dispersionkt−1. Networkresearchers have used a variety of centrality measuresto assess a firm’s position in an industry networkstructure. Whereas many studies have used Bonacich’s(1987) eigenvector centrality (e.g., Polidoro et al. 2011,Rosenkopf and Padula 2008), we use degree centrality tomeasure a firm’s positional embeddedness (e.g., Gulatiet al. 2012). Because eigenvector centrality quantifies theextent to which central nodes are connected to other cen-tral nodes, the measure becomes relatively unstable andhighly vulnerable to sampling effects (Costenbader andValente 2003). This suggests that eigenvector centralityis best suited to networks with clearly defined bound-aries and membership. Despite our best data collectionefforts, we acknowledge the likelihood of gaps in ouralliance network and fuzzy boundaries with respect toneighboring industries. Under these conditions, degreecentrality represents a robust measure for distinguish-ing between central and peripheral firms in the network(Wasserman and Faust 1994).

We measured the degree centrality for all sample firmsbased on collaborative activities in the five years preced-ing the formation of each sample multipartner alliance.To capture the extent to which a firm enjoys informa-tion benefits related to alternative partnering opportuni-ties outside the focal multipartner alliance, to computeits degree centrality, we exclude any partners connectedthrough the focal multipartner alliance. We used thepanel of alliance networks constructed for each yearfrom 1982 to 1996 to calculate the number of col-laborative ties over the last five years each partnerhad with firms outside the focal multipartner alliance.We then calculated the variance in positional embed-dedness among the partner firms within each multipart-ner alliance. A logarithmic transformation was used toenhance the normality of this measure.

Control VariablesTo limit alternative explanations and isolate the marginaleffects of explanatory variables, we controlled for sev-eral alliance-level characteristics.

Average Tie Strengthkt−1. Prior research suggests thatstrong cohesive ties between partners enhance the sta-bility of two-partner alliances by reducing opportunismand coordination costs (Gulati and Singh 1998). Basedon received wisdom (Greve et al. 2010, Kogut 1989), weassume that cohesive and trusting relationships betweenany two firms in a multipartner alliance can also have astabilizing effect on the entire alliance. Dense networksof generally strong ties between alliance members arethought to generate a collective cohesiveness that leadsto inclusive reciprocity norms, as well as limiting self-serving behaviors (Oh et al. 2004). Consistent with thework of Reagans and Zuckerman (2001), we conceptual-ize the density and resulting cohesiveness of a multipart-ner alliance in terms of the average tie strength acrossall possible partner pair relationships within the alliance.To measure the average tie strength within a multi-partner alliance, we computed the average tie strengthof multipartner alliance k in year t − 1 as the scope-weighted count of alliances formed between all dyads ofpartners in the alliance in the previous five years dividedby the number of possible dyads. Since each dyad couldhave collaborated in several alliances over the past fiveyears, values exceeding unity were common.

Average Positional Embeddednesskt−1. Firms thatoccupy central positions in an industry network arebelieved to benefit from superior access to valuableinformation (Gulati and Gargiulo 1999). When poten-tial partner firms are collectively more central, they gainmore reliable information about one another at the timeof alliance formation, which may lead to better part-ner selection (Gulati 1995b) and greater alliance stability(Polidoro et al. 2011). As before, we based our mea-sure on the number of collaborative ties each partnerhad with firms outside the focal multipartner allianceover the last five years. We then calculated the meanof partner centralities to obtain an alliance-level averagepositional embeddedness (Polidoro et al. 2011).

Knowledge Dispersionkt−1. Disparity in the level ofinnovative or technological capability among partnerscould affect the stability of an alliance. Collaborativerelationships may become strained by asymmetricresource dependency when partners cannot equitablycontribute technological resources (Das and Teng2002b, Park and Ungson 1997). Several managementresearchers have used patent counts as an indicator ofcorporate technological capabilities (Jaffe 1986, Moweryet al. 1996, Silverman 1999). Researchers argue thatorganizational knowledge depreciates over time because(1) products or processes can change, rendering someknowledge obsolete; (2) organizational records maybecome lost or difficult to access; and (3) when organi-zational turnover occurs, employees leave and take theirknowledge with them (Argote 1999). In response to the

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


potential decay of technological competencies and con-sistent with past alliance research, we considered onlypatents received in the last five years to reflect a firm’scurrent innovative capability (Stuart 1998). For everyperiod we calculated the inequality of firm innovative-ness within each multipartner alliance using the log-arithm of the variance. Higher values indicate greaterinequality.

Number of Partnersk. The number of partners inan alliance can increase coordination costs, monitor-ing costs, and managerial complexity (Gulati and Singh1998). Alternatively, the greater the number of partners,the greater the number of potentially stabilizing third-party ties associated with each partner pair. Thus wecontrol for the count of partners in each alliance.

Joint Venture Governancek. Research has shown thatequity joint ventures can provide superior coordinationand conflict resolution mechanisms to address the com-plexity of multipartner alliances (Garcia-Canal et al.2003), suggesting that the equity joint venture structureincreases multipartner alliance stability. The choice ofjoint venture governance is correlated with tie strengthbetween partners (Gulati and Singh 1998) whose effecton stability may be confounded with that of using jointventure governance. We controlled for joint venture gov-ernance using a dummy variable set to 1 if the alliancewas governed by an equity joint venture and 0 otherwise.

Geographic Diversityk. Diversity in nationality acrosspartners in an alliance can increase communication andcoordination costs (Parkhe 1993), decreasing stability.We control for the number of distinct nations in whichthe partners in alliance k are headquartered.

Market Overlapk. Alliances are more likely to failwhen partners are also competitors (Park and Russo1996). We control for the influence of competition inmultipartner alliance k by counting the number of dyadsin which both firms share the same four-digit StandardIndustrial Classification (SIC) code.

Prior Partnership Dissolutionsk. The dissolution ofprior alliances involving members of a focal multi-partner alliance may influence its stability. For everyalliance-year, we counted the number of dyadic disso-lution events partners in the focal multipartner alliancehad experienced with each other in the prior five years.Of the 59 multipartner alliances in our sample, 54 hadmembers that had between one and four prior dissolu-tions events with each other.

Task Complexityk. To control for differences in taskcomplexity on alliance stability, we added dummy vari-ables to account for the fixed effects of the presenceof the following activities: technology codevelopment,manufacturing, marketing, and technology licensing.

Industry Classificationk. Although all firms in oursample were leading firms in the telecom industry, acloser inspection of alliance deal texts suggested thatsample alliances operated across five related industrysegments—communication equipment, computers andoffice equipment, consumer electronics, electronic com-ponents and accessories, and data processing and infor-mation services. We controlled for these differences atthe three-digit SIC level using dummy variables.

Potential EndogeneityOur study of unplanned dissolution events among mul-tipartner alliances is necessarily limited to sets of firmswho chose to participate in a multipartner alliance. How-ever, firms do not choose to form multipartner alliancesat random. Thus our coefficient estimates could bepotentially biased as a result of systematic managerialselection processes that are related to both our core the-oretical variables (i.e., tie strength dispersion and tiestrength dispersion × positional embeddedness disper-sion). If it appears that managers do not systematicallytake into account our variables of interest when deter-mining whether or not to form a multipartner alliance,confidence that self-selection is not biasing our coeffi-cients of interest in our survival models increases.

With the exception of one recent study focusing onthe new venture participation in multilateral researchand development alliances (Li 2013), there has been nosystematic study of multiparty alliance formation, letalone one that takes into account faultlines between sub-groups of partners. Thus, we used our sample to developa model of multipartner alliance formation to com-pare those multipartner alliances that actually formed topotential multipartner alliances that did not. Since gen-erating all possible permutations of potential multipart-ner alliances of varying sizes involving the 46 firms thatwere involved in our sample of multipartner allianceswould have been prohibitively complex, we used amatched pair design matching on formation year andalliance size (number of partners). To construct the con-trol sample of multipartner alliances that could haveformed but did not, we needed to construct a risk setof firms that could have formed multipartner alliancesbut did not. For a particular year in our sample, weconsidered sample firms that were involved in at leastone alliance to be eligible for participation in a mul-tipartner alliance. From this risk set of firms, we ran-domly selected n firms into an unrealized multipartneralliance to match the same number of firms in a real-ized multipartner alliance. For example, if a multipartneralliance was formed in year t by three firms, we ran-domly selected three firms from the risk set of firmsfor the same year to form a matching unrealized multi-partner alliance. Consistent with King and Zeng (2001)and Hansen and Løvås (2004), for each multipartneralliance formed in each year t, it was matched with two

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Table 1 Descriptive Statistics and Pearson Correlations for Multipartner Alliance Formation

Variable Mean S.D. Min Max 1 2 3 4 5 6 7 8

1 Alliance formation 0033 0047 0000 1000 —2 Tie strength dispersion 1028 1029 0000 4096 0043 —3 Average tie strength 2033 3039 0000 15023 0039 0041 —4 Average positional embeddedness 62095 37010 6033 202025 0045 0053 0070 —5 Positional embeddedness dispersion 7001 1052 2017 9066 0014 0032 0012 0054 —6 Knowledge dispersion 13074 2078 −9021 16071 0027 0030 0023 0039 0041 —7 Number of partners 4017 1072 3000 10000 0000 0012 0001 0000 0008 0015 —8 Geographic diversity 2063 0094 1000 6000 −0041 −0020 −0029 −0012 0000 −0002 0051 —9 Market overlap 0080 1014 0000 7000 −0010 0010 −0007 0000 0008 0004 0057 004

Notes. N = 177. Correlations with absolute value greater than 0.15 are significant at the 0.05 level.

randomly constructed unrealized multipartner alliances.These unrealized multipartner alliances constitute thecontrol sample. Thus, the overall sample used to assessthe likelihood of alliance formation was 177: 59 formedalliances plus 118 unformed.

Table 1 provides correlations and descriptive statisticsfor all variables used to assess the likelihood of multi-partner alliance formation. Model 3 in Table 2 providesour full rare events logistic regression model predict-ing the likelihood of alliance formation. In our analysis,

Table 2 Rare Event Logit Estimates on the Likelihood ofMultipartner Alliance Formation

Alliance formation

Variable Model 1 Model 2 Model 3

Tie strength dispersion −0037 −0044400285 400305

Tie strength dispersion×Positional 0012embeddedness dispersion 400195

Average tie strength 0015 0029 0027−0013 400195 400195

Average positional embeddedness 0004∗ 0004∗ 0004∗∗

400025 400025 400015Positional embeddedness dispersion −0008 −0007 −0015

400225 400215 400225Knowledge dispersion 0011 001 0012

400135 400125 400145Number of partners 0030 0032 0032

400205 400205 400205Geographic diversity −1075∗∗∗ −1078∗∗ −1078∗∗

400445 400445 400435Market overlap 0000 0006 0005

400265 400265 400265Fixed year effect Yes Yes YesFixed industry effectConstant −2009∗ −1085 −2018

Log pseudolikelihood −54047 53024 −53003Observations 177 177 177Degree of freedom 16 17 18Likelihood ratio test chi2(1) 2046 0042Prob> chi2 0012 0051Akaike information criterion 140094 140048 142006

Notes. Two-tailed tests were used for all variables. Robust,heteroskedasticity-adjusted standard errors are in parentheses.

∗p < 0005; ∗∗p < 0001; ∗∗∗p < 00001.

neither tie strength dispersion nor its interaction withpositional embeddedness dispersion was found to signif-icantly influence the formation of multipartner alliances.Thus our variables of interest appear to be randomly dis-tributed across the treatment group (formed multipartneralliances) and control group (potential alliances that didnot form). Although insignificant findings should not beoveremphasized, they provide an initial indication thatthe coefficients of interest in a survival model are notbiased as a result of self-selection bias.

There are reasons as to why firms in our samplewould not take these core theoretical variables into con-sideration when forming multipartner alliances. First,managers may simply not have been cognizant of theperformance implications of the variables we are explor-ing. The use of multipartner alliances was relativelynovel at the time of the study (1987–1997). The dif-fusion of administrative innovations such as multipart-ner alliances and the understanding of when to usethem take time (Teece 1980). Early in a diffusion cycle,when decision makers are not fully aware of the perfor-mance consequences of their choices, characteristics ofinterest are likely to be randomly distributed (Armourand Teece 1978). Under such conditions, empirical set-tings are more akin to a natural experiment, and sampleselection bias will be minimal (e.g., Park and Steensma2012). Managers were likely unaware of tie strengthand positional embeddedness dispersion and its potentialinfluence on survival. To the extent that they may havebeen aware, the opportunity to access diverse resourcesfrom unfamiliar partners using multipartner alliancesmay have overridden any concerns.

Although we found little evidence that our primaryresults of interest are biased by sample self-selection,we took additional steps to account for the effects ofunobserved heterogeneity on our dissolution models asdescribed in our tests of robustness.

Model Specifications, Estimation, and ResultsTo test our core hypothesis, we employed survival anal-ysis techniques. We chose a parametric approach to esti-mate the baseline hazard function (i.e., the rate at whichalliances dissolve when all covariates equal zero). Para-metric survival functions allow for the baseline hazard to

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


vary as a function of alliance duration (Dussauge et al.2000, Park and Russo 1996). The Weibull distributionwas used to accommodate a monotonic effect of time(Polidoro et al. 2011). We then estimated the effectsof covariates as exponentially multiplicative increases ordecreases in the baseline hazard rate.

To account for potential autocorrelation caused byunobserved alliance effects that are stable over time,we clustered observations by alliance and reported clus-tered standard errors that were robust in terms of botharbitrary heteroskedasticity and intragroup correlation(Rogers 1993). To control for such unobserved system-atic period effects such as industry conditions, whichmay influence alliance stability, year dummies wereincluded in all models. We lagged the dependent variableby one year, facilitating causal inferences by establishingtemporal precedence, which reduced concerns of reversecausality as well as avoiding simultaneity.

Table 3 provides correlations and descriptive statis-tics for all variables used to assess the likelihood ofunplanned dissolution. Table 4 provides the results ofour survival analysis. All models included year andindustry segment dummy variables. Conservative two-tailed tests were used to assess the significance of allcoefficients. Model 1 includes our control variables,Model 2 introduces tie strength dispersion, and Model 3adds the interaction between tie strength dispersion andpositional embeddedness dispersion.

Hypothesis 1 proposes a positive relationship betweentie strength dispersion and the likelihood of unplanneddissolution. In Model 2, the effect of tie strength dis-persion was positive and highly significant (� = 1041,p < 0005). A positive coefficient suggests an increasedprobability of multipartner alliance dissolution. Thus, wefind support for the general destabilizing effect of tiestrength dispersion. Hypothesis 1 is supported. Based onthese empirical results with all other covariates set to 0,a one-standard-deviation increase of tie strength disper-sion (Model 2) resulted in a more than fivefold increasein the risk of an unplanned multipartner alliance disso-lution occurring four years after formation.

Hypothesis 2 suggests that the relationship betweentie strength dispersion and the likelihood of unplanneddissolution will be relatively weaker when multipart-ner alliances comprise partners that are highly dispersedin terms of positional embeddedness. In Model 3, theeffect of tie strength dispersion × positional embedded-ness dispersion was negative and significant (p < 0005).To gain further insight into the nature of this moderat-ing effect, we plotted the survival functions for multi-partner alliances characterized by high and low levelsof tie strength dispersion and positional embeddednessdispersion, while holding all other covariates constant attheir mean values. Figure 1, panels (a) and (b), revealsthat there is a stronger positive relationship betweentie strength dispersion and unplanned dissolution when Table

3Des

criptive

Statisticsan

dPea

rsonCorrelationsforMultipartner

Allian

ceDisso

lution

Varia

ble

Mea

nS.

D.

Min

Max

12

34

56

78

910

1112

1314

1Unp

lann

eddiss

olution

0008

0027

0000

1000

—2

Tiestreng

thdisp

ersion

2022

1018

0000

4094

−00

02—

3Ave

rage

tiestreng

th70

5740

8700

0020

088

−00

1300

38—

4Ave

rage

positio

nale

mbe

dded

ness

1000

7742

089

1200

021

4050

−00

1000

4400

56—

5Po

sitio

nale

mbe

dded

ness

disp

ersion

7091

1036

2063

1000

200

0100

3100

1300

37—

6Kno

wledg

edisp

ersion

1500

510

0710

014

1607

400

0300

2600

1800

2200

30—

7Num

berof

partne

rs40

1310

7030

0010

000

0008

0010

−00

06−

0008

−00

0900

18—

8Jo

intv

enture

gove

rnan

ce00

1600

3000

0010

00−

0001

0007

−00

1000

0400

2800

1900

2—

9Geo

grap

hicdive

rsity

1071

0077

1000

3000

0003

0030

−00

0400

1900

35−

0001

−00

13−

0009

—10

Marke

tove

rlap

0066

1003

0000

4000

−00

0300

1400

00−

0006

0022

−00

2800

06−

0011

0026

11Priorpa

rtne

rshipdiss

olutions

1041

1027

04

0001

0011

0019

0028

−00

0400

3600

400

04−

0032

−00

251

12Lice

nsingag

reem

ent

0031

0047

01

−00

0700

3900

300

3300

3300

13−

0013

−00

1100

3300

07−

0028

113

Man

ufac

turin

gag

reem

ent

0008

0028

01

0002

−00

07−

0011

−00

1500

1300

05−

0018

0018

0024

0034

−00

19−

0005

114

Marke

tingag

reem

ent

0013

0034

01

−00

0300

1800

0800

0400

21−

0005

−00

1100

2200

1800

19−

0008

0029

001

1

Notes

.N=

251

allia

nce-

year

s.C

orre

latio

nsw

ithan

abso

lute

valu

egr

eate

rth

an0.

1ar

esi

gnifi

cant

atth

e0.

05le

vel.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Table 4 Weibull Regression Estimates of the Hazard of Unplanned Multipartner AllianceDissolution

Variable Model 1 Model 2 Model 3 Model 4

Tie strength dispersion 1041∗ 2028∗ 2028∗∗

400585 400965 400745Tie strength dispersion× −1012∗ −1012∗

Positional embeddedness dispersion 400575 400505Average tie strength −0014 −0045 −0036 −0036

400165 400335 400275 400275Average positional embeddedness 0002 0002 0002 0002

400025 400025 400035 400025Positional embeddedness dispersion −0018 −0028 −0095 −0095

400765 400975 410005 400705Knowledge dispersion 0044 0015 −0030 −0030

400485 400515 400605 400495Number of partners −0020 −0043 −0076∗ −0076∗

400195 400335 400415 400345Joint venture governance 2072† 3016∗ 3093∗ 3093∗∗

410405 410605 410735 410395Geographic diversity −1006 −1053 −1079 −1079∗

400945 410115 410315 400835Market overlap 0089 0073 0063 0063

400645 400635 400535 400655Prior partnership dissolutions −0068 −0062 −0054 −0054

400525 400595 400535 400495Licensing agreement 0008 0059 1026 1026

410215 410475 410415 410075Manufacturing agreement −0040 0007 −0047 −0047

400685 400645 400525 410155Marketing agreement 3043 3048 4053 4053∗∗

420435 420595 430515 410745Fixed year effect Yes Yes Yes YesFixed industry effect Yes Yes Yes YesConstant −18062∗∗∗ −22020∗∗∗ −24035∗∗∗ −24035∗∗∗

ln(Weibull shape parameter �) 1062∗ 1077∗ 1093∗ 1093∗

400395 400395 400425 400225ln(Overdispersion parameter �) −16038Log pseudolikelihood −16029 −13001 −10049 −10049Observations 251 251 251 251Degree of freedom 25 26 27 27Likelihood ratio test chi2(1) 6057 5004Prob.> chi2 0001 0002Akaike information criterion 82059 78002 74098

Notes. Two-tailed tests were used for all variables. Robust, heteroskedasticity-adjusted standarderrors are in parentheses.

†p < 001; ∗p < 0005; ∗∗p < 0001; ∗∗∗p < 00001.

positional embeddedness dispersion is low (−1 S.D.)as opposed to when it is high (+1 S.D.). For example,when positional embeddedness dispersion is low, a one-standard-deviation increase from the mean in tie strengthdispersion reduced from 97.05% to 7.02% the probabil-ity of a multipartner alliance surviving to its eighth yearwithout experiencing an unplanned dissolution. How-ever, when positional embeddedness dispersion is high,a one-standard-deviation increase from the mean in tiestrength dispersion only reduced the probability of amultipartner alliance surviving to its ninth year from91.93% to 81.60%. Together, panels (a) and (b) ofFigure 1 show multipartner alliances are particularly

vulnerable to faultlines when they comprise partners thatoccupy similar network positions.

Our results with regard to our control variables werealso notable. After controlling for all other factors includ-ing positional embeddedness and its interaction withpositional embeddedness dispersion, we find that num-ber of partners increased stability (Table 4, Model 3).This result is consistent with the notion that, despitethe added complexity, having more partners may pro-vide an alliance structural stability through additionalthird-party ties. Our results also suggest that joint ven-ture governance promotes the unplanned dissolution ofmultipartner alliances. A possible reason for this finding

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Figure 1 Effects of Faultline Strength on MultipartnerAlliance Survival Under Conditions of Low (a) andHigh (b) Positional Embeddedness Dispersion

0

0.2

0.4

0.6

0.8

1.0

0 5 10 15 20 25

Surv

ival

Age (in years)

0

0.2

0.4

0.6

0.8

1.0(b)

(a)

0 5 10 15 20 25

Surv

ival

Age (in years)

Faultline strength high

Faultline strength mean

Faultline strength low

could be that strict formal controls associated withjoint venture governance negatively affect the flexibil-ity needed by multiple partners to create a stable evolu-tionary path for the alliance through mutual adaptation(Ariño and de la Torre 1998).

Robustness and Alternative SpecificationsWe examined the robustness of our results in vari-ous ways. Although our explanatory variables did notaffect alliance formation (Table 2, Model 3), endogene-ity resulting from selection processes remains a possiblethreat. The standard solution for selection bias is to esti-mate a selection parameter from the full set of all possi-ble alliances (actual and hypothetical) and then includethis parameter in the restricted sample of those that actu-ally formed and were at risk of an unplanned dissolution(Heckman 1976). Unfortunately, we found this techniquea poor fit for addressing selection issues in our specificcontext because of (1) the intractable number of possiblemultipartner alliances that could have formed and (2) thenormality assumptions associated with a Heckman cor-rection in a two-stage model, which are problematic innonlinear models such as survival analyses (Boehmkeet al. 2006, Prieger 2002).

One means of addressing the problem of unobservedheterogeneity is to expand the survival model to includean unobserved random proportionality factor, creatingso-called frailty models. Frailty models are used to iden-tify an excessive degree of unexplained variability asa result of misspecification or omitted covariates, alsoknown as overdispersion. In Table 4, Model 4, we mea-sured the presence of overdispersion by adding a gammadistributed4 latent multiplicative effect to the hazardfunction specified in Model 3. The frailty parameter � isnot statistically significant, nor does the likelihood-ratio

test comparing the frailty models to the standard suggestthat unobserved heterogeneity is present. In addition,there were no substantial changes in our coefficients.

Another empirical strategy for contending with endo-geneity in nonlinear models is the use of flexible paramet-ric selection (FPS) models (Prieger 2002). FPS modelsare not constrained to a particular functional form, and theuse of a bivariate exponential distribution to bind selec-tion and duration equations together allows for durationdependence to assume the form of the Weibull distribu-tion (Boehmke et al. 2006). We estimated simultaneousduration-selection models using the DURSEL program inSTATA. The results show that faultline strength and itsmultiplicative effect with positional embeddedness dis-persion were not significant in the selection model, butthey significantly influence multipartner alliance dura-tion, further reducing endogeneity concerns with respectto our theoretical variables. We provide the detailedmodel estimates in Appendix B and Table B.1, along withan explanation of the FPS model.

Our results were robust using a nonparametric spec-ification employed in previous studies (Dhanaraj andBeamish 2004, Hennart et al. 1998, Kogut 1989, Parkand Russo 1996, Xia 2011). When we recomputed ouroriginal tie strength measures using a four-year win-dow instead of a five-year window, our results remainedrobust. They were also robust when we computed tiestrength based on unweighted counts of prior ties.Finally, we considered whether the influence of tiestrength dispersion on instability depends on the numberof partners. To assess whether the deleterious effects oftie strength dispersion are stronger in larger alliances,we tested for the interaction effect of tie strengthdispersion × number of partners. We found a positive,albeit marginally significant (p < 001), effect. Futureresearch may want to explore further this possibility.

Discussion and ConclusionReceived wisdom within the alliance literature is thatmultipartner alliances are inherently unstable because oftheir complexity and the increased potential for free rid-ing, yet empirical evidence regarding their stability hasbeen mixed. Insights from social network theory sug-gest that multipartner alliances benefit from built-in sta-bilizing third-party ties, which can mitigate opportunismand conflict between partners. Nonetheless, within mul-tipartner alliances, schisms may also occur between sub-groups of partners. Despite the notion that multipartneralliances are dense clusters of firms that exhibit networkcharacteristics (Rosenkopf and Padula 2008), fundamen-tal network concepts have not been used to exploretheir stability. This study considers how configurationsof partner embeddedness influence multipartner stability.

Within our sample, those multipartner alliances thathad highly dispersed tie strength across partner pairs

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


were generally more likely to dissolve than thoseexhibiting less variability in tie strength. We suggestthat when subgroups of alliance partners share relativelystrong ties to each other, but weak ties with other part-ners, in-group and out-group factionalism may occur,creating mistrust and conflict within the alliance.

Those who coalesce as an in-group identify more withtheir subgroup than with the larger group. Such factionscan hamper the development of the broad-based trustand indirect reciprocity norms that are necessary if mul-tipartner alliances are to meet partner expectations andfully exploit their potential to create value. Unmet per-formance expectations will generate further mistrust andconflict, and they could potentially lead to the dissolu-tion of the alliance.

However, we also found that when multipartneralliances involved a mix of centrally and peripherallypositioned partners within the industry network, theywere more resilient to the effects of divisive faultlines.This resiliency may be due to the high opportunitycosts of dissolution for the partners in these alliances.The potential value from combining relatively distinctiveknowledge provided by partners positioned in differentparts of the alliance network, and the opportunity forperipherally positioned firms to gain prestige by part-nering with central firms, provides strong incentive totolerate internal factions. Moreover, strong lead partnersare likely to emerge in multipartner alliances with mem-bers that are highly dispersed in their positions withinthe broader alliance network. Relatively high-status part-ners can be particularly effective third-party peacekeep-ers and coordinators of deferential lower-status partners.

The results of this study have implications for anumber of research threads. Alliance studies havedemonstrated that prior ties may influence the forma-tion, evolution, governance, and outcomes of two-partneralliances (Ahuja 2000, Gulati 1995b, Gulati and Gargiulo1999, Gulati et al. 2000, Stuart 2000). However, mul-tipartner alliances differ from two-partner alliances interms of their internal structure, exchange processes, andreciprocity expectations. To our knowledge, this is thefirst study to specifically examine the factors that influ-ence multipartner alliance stability. Moreover, we employconcepts (e.g., faultlines) and measures that are uniquelyapplicable to multipartner alliances where a distributionin tie strength across partner pairs is possible.

Nonetheless, these findings from our study comple-ment those from studies on two-partner alliances andalliance networks. In their study of alliances in theglobal liner shipping industry, Greve et al. (2010) foundthat, contrary to their expectations, the presence of par-ticularly cohesive members in a multipartner alliancewould increase the likelihood of their withdrawal fromthe alliance. They pointed to a potential dark side toembeddedness, an idea they argued needs to be devel-oped further. Similarly, we also find that particularly

cohesive subgroups within multipartner alliances hamperstability. We elaborate more fully on the notion of a darkside to embeddedness by considering the occurrence offaultlines within multipartner alliances. Polidoro et al.(2011) found that having more common third partiesconnected to the two alliance partners in a two-partneralliance reduced the likelihood of the alliance dissolving.After accounting for their configurations in tie strengthand positional embeddedness, we found that multipart-ner alliances that had a greater number of partners werein fact more stable. One explanation is that because ofthe number of members, larger alliances offer more sta-bilizing third-party ties internal to the alliance than dosmaller alliances.

Contrary to their expectations, Polidoro et al. (2011)also found that two-partner alliances where the part-ners had higher levels of combined centrality were morelikely to dissolve when these firms had prior direct ties.Greve et al. (2013) suggested that given viable alterna-tive partnering options outside of their current alliances,firms are more likely to withdraw from their currentalliances. Similarly, we found that multipartner alliancescomprising primarily centrally positioned partners weremore likely to dissolve when there were faultlines result-ing from particularly strong ties among a subgroup ofpartners. We argue that multipartner alliances compris-ing partners with an extensive number of alternativeoptions as a result of their network positions are partic-ularly susceptible to falling victim to destabilizing fault-lines. Overall, different types of embeddedness appearto exert substantially different types of influence onalliance stability.

Our theory and results highlight an additional counter-productive aspect of overembeddedness to that describedby Uzzi (1996) in his study on alliance networks.Firms within sparsely connected industry networks maybecome overembedded in a subset of partners withwhom they feel comfortable and familiar, forgoing pro-ductive exchange opportunities with firms with whomthey are less embedded. We extend this notion tosmaller, fully saturated networks such as multipartneralliances. When a subset of partners within a multi-partner alliance is highly embedded (i.e., strong ties)compared with other partners, factions can form to thedetriment of the entire alliance. In a paradoxical fash-ion, firms may try to prevent becoming overembeddedin their industry network by engaging unfamiliar anduntested partners within multipartner alliances, whichprovide informal control and safety in numbers. How-ever, the extent to which multipartner alliances endureoverembeddedness depends on their configurations ofstrong ties. In essence, the same factors that may encour-age multipartner alliance formation (i.e., embedded firmssafely partnering with unfamiliar firms) may also lead toits demise (via faultlines).

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


In contrast to the unambiguously stabilizing influencecommonly associated with strong ties in dyadic allianceresearch, our study suggests that in multilateral collab-orative contexts, strong ties may have a destabilizingeffect. When cohesive ties within multipartner alliancesresult in divisive faultlines, they can impede the develop-ment of generalized social exchange norms. Thus, strongties can impede collective action in situations wherenorms of generalized exchange are critical to collabora-tive success. Therefore, the influence of tie strength onthe collaborative stability in general depends on the rel-ative importance of preferential (dyadic) exchange andgeneralized (collective) exchange processes in specificnetworks and alliances. Gaining a better understanding ofthe dark side to strong ties could increase the explanatorypower of the social embeddedness framework for com-plex collaborations involving more than two partners.Our study provides a theoretical foundation for futureresearch on multipartner collaboration at various levelsof analysis (Rowley et al. 2000, Uzzi 1996).

Our study introduces faultlines as a mechanism thatcan influence the structural trajectory of the industry net-work by jeopardizing the stability of ties within mul-tipartner alliances. When stable multipartner alliancescontain bridging relationships linking partners from pre-viously disconnected regions of the network, they dis-proportionately influence the overall network structureby increasing the overall connectivity and clustering inthe network. However, when they unexpectedly dissolve,the connectivity and clustering in the network is sig-nificantly reduced. Our study informs network dynam-ics research by demonstrating how inertia, opportunity,exogenous pressures, and agency as fundamental driversof tie formation and dissolution affect mesolevel struc-tures in interorganizational networks.

Finally, this study has implications for managersinvolved in or contemplating multipartner alliances. Ourtheory and results suggest that when establishing a mul-tipartner alliance, the collaborative history of all poten-tial partners should be taken into account. Managersshould not rely solely on the stabilizing effect of com-mon third parties in multipartner alliances but insteadclosely examine the disparity in tie strength betweenthe various partners and the potential for faultlines.Any alliance where extensive collaborative histories cre-ate strong ties between certain partners and not oth-ers face limited long-term viability compared with thosethat offer a more uniform level of tie strength. Firmsinterested in forming multipartner alliances that featurehighly disparate tie strengths may wish to mandate team-building mechanisms, which can offset the tendencyfor cohesive partners to coalesce or rely on preexistingnorms that do not include all members.

Limitations and Future ResearchAlthough promising, the results and contributions offeredby this study are not without limitation. Rather than using

a finer-grained measure, we relied on unplanned disso-lution to measure stability. Although fine-grained dataon changes in multipartner alliance governance such asshifts in division of equity (Blodgett 1992) and alter-ation of contractual agreements (Reuer and Arino 2002,Young-Ybarra and Wiersma 1999) may allow for a betterrepresentation of alliance stability, the lack of availablearchival data on our sample’s changes in governance pre-cluded their use. Despite such limitations, we believe ourmeasure of alliance stability is valid, since alliance disso-lution is a recognized indicator of collaborative stability(Kogut 1989) and because any misclassification of termi-nation events led to conservative tests of our hypotheses.Such a measure of alliance stability is consistent withthe research on the stability of two-partner joint ventures(Polidoro et al. 2011).

Other limitations concern the nature of our sample.Because our sample frame consisted of large firms and weused a nonrandom sampling process, our results may notgeneralize beyond the population of multipartner telecomalliances. And because our sample was based on horizon-tal technology alliances where coordination and gover-nance hazards were exacerbated (Gulati and Singh 1998),alliances employing vertical relationships as in manufac-turing or marketing may not reflect the same dynamics.Additional evidence based on larger random samples anddata on different types of multipartner collaborations areneeded to externally validate these results. Despite ourbest efforts to address potential sample selection con-cerns, the numerous selection steps (focus on realizedmultipartner alliances, publicly announced, encompass-ing only publicly traded telecom firms) make it diffi-cult to fully rule out the possibility of selection biases.Although our empirical results are consistent with thenotion of destabilizing faultlines, in-depth case studieswould enrich our understanding of such processes andtheir effects on multipartner alliance stability.

Despite these concerns, we believe the results willlead to interesting avenues for future research. Little isknown about the consequences of multipartner allianceson firm-level outcomes. Although our study focuseson alliance stability, we did not focus on which part-ner(s) chose to exit a multipartner alliance. A study byLavie et al. (2007) used a competitive dynamics frame-work to examine the link between firm performanceand entry/exit decisions. In contrast, our theoreticalargument suggests that multipartner alliance outcomesdepend partly on the distribution of tie strength amongpartners in that alliance. Following this logic, when rel-ative strangers enter multipartner alliances with a sub-group of cohesive partners, they will exit earlier thaninitially planned, decreasing derivable benefits of thealliance. Testing this theoretical logic with detailed datashould allow for a better understanding of the relation-ship between multipartner alliance and individual partnerperformance.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


An overall understanding of the benefits and detri-ments of strong ties within multipartner alliances re-mains incomplete. Future studies should investigatethose conditions where tie strength dispersion offers astabilizing effect, perhaps buffering the consequences ofan imbalance in the distribution of power. Multipartnertechnology alliances frequently emerge when relevanttechnological knowledge about a new product or serviceis dispersed among many firms, leading to a power struc-ture based on partners’ technological resource positions,embedded in a network of social relations. We arguethat subgroups of highly embedded partners are morelikely to form coalitions operating outside formal legit-imated structures to advance their agenda through con-certed actions (Brass and Burkhardt 1993). Investigatinghow tie strength dispersion within a multipartner alliancecan moderate the influence of formal power distributionon stability and other outcomes would make for an inter-esting extension of our findings. The theoretical perspec-tive advanced in this study may inform future researchon the dynamics within larger strategic alliance struc-tures in high-technology settings. Prominent and eco-nomically consequential examples include patent poolsand standard setting consortia.

Appendix A. Examples of Archival Evidence Related to Assessing the Dissolution of Multipartner Alliances

Dissolution event Participants Dissolution context

Unplanned Motorola Inc., Apple ComputerInc., International BusinessMachines (IBM) Corp.

IBM Corp. entered into a strategic alliance with Apple Computer Inc. and MotorolaInc. to jointly develop a new computer hardware system. The hardwareprototypes were expected to be manufactured during 1995 and on the market in1996. The collaboration was dissolved when IBM and Apple Computerannounced on November 17, 1995, plans to scrap their Kaleida Labs softwarejoint venture. Almost from the start, engineers had haggled over jobresponsibilities, and board members were distracted by more important jobs.Conceived in 1991 as part of a grand alliance meant to put giant Microsoft in itsplace, the venture gained only a handful of customers for its ScriptX multimediaprogramming language. (Source: Hammonds 1995)

Hitachi Ltd., NTT Corp., NECCorp., Oki Electric IndustryCo. Ltd., Matsushita ElectricIndustrial Co. Ltd., Fujitsu Ltd.,Toshiba Corp., Sharp Corp.,Mitsubishi Electric Corp.

Japan’s fifth-generation computer, a highly anticipated project that took 11 years todevelop, failed to work properly, thus causing plans for a logical computer tocome to a halt. Consequently, the project came to a quiet close. After thedissolution, the Institute for New Generation Computer Technology (ICOT)software was offered free of charge to all comers, but no other institute offeredto take the technology further. “ICOT’s programs will run only on custom-builtcomputers, and none of the companies that built the parallel interface machinemodules plans to put such machines into production.” (Source: Cross 1992)

Oki Electric Industry Co. Ltd.,Toshiba Corp., Fujitsu Ltd.,NEC Corp., Hitachi Ltd.,Mitsubishi Electric Corp.

The collapse of the Open Systems Interconnection (OSI) project in 1996 severelydamaged the reputation and legitimacy of the organizations involved, especiallyISO. The worst part was that OSI’s backers took too long to recognize andaccommodate the dominance of the TCP/IP protocol suite. The financial damagedone to Japan and Europe (where Internet deployment was delayed by years) isdifficult to estimate. (Source: Walrand and Varaiya 2000)

Planned Compression Labs Inc., RelianceElectric Co., Philips

“Reliance Comm/Tec and three partners have been jointly developingvideo-on-demand solutions for nearly a year. In the latest demonstration, theblockbuster film The Firm was delivered in compressed digital video format byReliance Comm/Tec, Compression Labs, On-Demand Technologies, and PhilipsConsumer Electronics. It was brought to the small screen by transporting MotionPicture Experts Group signals through a coaxial network by RelianceComm/Tec’s integrated multimedia access platform, from On-DemandTechnologies’ RAIS digital video service to a Philips 31-inch TV set.” (Source:Philips Business Information 1994)

ConclusionWe integrate network and faultline concepts and suggestthat particularly cohesive subgroups within a multipart-ner alliance may be detrimental to its stability. Whenfaultlines impede the development of broad-based trustand norms of indirect reciprocity across all partners, thepotential for multipartner alliances to create value is ham-pered. To the extent that tie strength dispersion leads todivisive faultlines (applicable only to multiparty struc-tures), the evolutionary path of the entire collaborativenetwork is affected. Multipartner alliances are partic-ularly popular and prevalent in high-growth industriessuch as telecommunications, aerospace, and software.Thus there is good reason to develop a more granularunderstanding of what determines their performance andsuccess.

AcknowledgmentsThe authors are grateful for thoughtful comments from WarrenBoeker and Jeffrey Barden, as well as senior editor GautamAhuja and two anonymous reviewers throughout the reviewprocess.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Appendix B. Robustness Test with ConcurrentEstimation of Selection and Duration ModelsWe used an FPS model to address sample selection (Prieger2002). FPS models are not constrained to a particular func-tional form, and the use of a bivariate exponential distributionto bind selection and duration equations together allows forduration dependence to assume the form of the Weibull distri-bution (Boehmke et al. 2006). At first, this analytical techniqueappeared uniquely suited to our empirical context becauseit parallels the Heckman approach and would be consistentwith our finding that the baseline hazard increases over time.However, the results of the simultaneous duration-selectiontechnique calculated using the DURSEL program in STATAin Table B.1 should be interpreted with caution because theduration-selection technique only allows for time-invariantcovariates. This raises theoretical and analytical issues.

First, our study is based on the premise that multipartneralliances are at risk of experiencing an unplanned dissolutionas a function of factors that change over time. This is con-sistent with our theoretical focus on the collaborative dynam-ics affecting multipartner alliance stability as a function ofvarying tie strength within the alliance. Second, by shiftingfrom a dynamic to a static interpretation, we move analyticallyfrom a panel design describing year-to-year variation in ourtheoretical variables to an exclusive focus on the formation

Table B.1 Full Information Maximum Likelihood Duration withSelection Model

Alliance Allianceformation duration

Variables Model 1 Model 2

Tie strength dispersion −0018 −1064∗∗∗

400235 400315Tie strength dispersion× −0001 0016∗∗∗

Positional embeddedness dispersion 400025 400045Average tie strength 1032∗ −0010∗∗∗

400605 400025Average positional embeddedness 0000 −0000

400015 400005Positional embeddedness dispersion −0018 0010∗

400155 400055Knowledge dispersion 0025 0021∗∗∗

400115 400045Number of partners 0008 −0009∗

400125 400045Joint venture governance −0038

400745Geographic diversity −0063∗ −0044

400245 400825

Fixed year effect No NoFixed industry effect NoConstant −2062∗ −0022

ln(Weibull shape parameter �) 1.81∗

Rho (error correlation) 0.24∗∗

Log pseudolikelihood −43088Observations (uncensored) 84 (25)

Note. Two-tailed tests were used for hypothesized variables andcontrols.

∗p < 0005; ∗∗p < 0001; ∗∗∗p < 00001.

and termination events. Whereas the estimates in the selec-tion model derive from the conditions preceding the formationof the 59 observed multipartner alliances, the correspondingduration model is estimated based only on the conditions pre-ceding the 25 observed termination events. In other words,this model did not take into account observations that werecensored because of a termination event not occurring. Thisresults in a dramatic reduction of our data support for modelestimation and hypothesis testing (i.e., 84 relevant observationsfor both selection and duration models).

This reduced sample size did not support the complexity ofour original model specification, and we could only computeestimations for simplified models. Because of the sample size,the results for some controls appeared rather sensitive to spec-ification changes. Despite this rather tenuous analytical envi-ronment, model results related to our theoretical propositionwere consistent with our findings in alternative approaches.Positive coefficients in the selection model indicate factorsthat increase the probability of multipartner alliance formation,whereas positive coefficients in the duration model increase theduration and thus reduce the likelihood of an unplanned disso-lution event. Despite the caution that is warranted in the inter-pretation of these results, three observations merit attention.

First, tie strength dispersion and its multiplicative effectwith positional embeddedness dispersion do not seem to affectthe formation but increase the likelihood of an unplanned dis-solution event, further reducing concerns that the observedeffects of our theoretical variables are artifacts of selec-tion bias.

Second, average tie strength has a positive effect on multi-partner alliance formation but a negative effect on multipartneralliance duration. This result is consistent with the notion thathigh levels of relational embeddedness among partners repre-sent an attractive force during partner selection that can turninto a collaborative liability as the alliance evolves.

Finally, the correlation between error terms in the selectionand duration models is significant. This suggests that, in thissimplified specification, the selection model indeed removedbias from the duration model estimates.

Endnotes1We thank an anonymous reviewer for suggesting this line ofthinking.2Prior to the logarithmic transformation, we added 1 to allvalues of tie strength variance such that the natural log ofmultipartner alliances with zero variance was defined.3Herfindahl index = 41/N5+N × (variance), where N is equalto the size of the sample. Adjustments for sample size areparticularly appropriate for measuring concentration of marketshare within an industry, where N is equal to the number ofindustry participants.4Results were also robust to frailty models using an inverseGaussian distribution.

ReferencesAhuja G (2000) Collaboration networks, structural holes, and innova-

tion: A longitudinal study. Admin. Sci. Quart. 45(3):425–455.

Ahuja G, Polidoro F, Mitchell W (2009) Structural homophily orsocial asymmetry? The formation of alliances by poorly embed-ded firms. Strategic Management J. 30(9):941–958.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Ahuja G, Soda G, Zaheer A (2012) The genesis and dynamics oforganizational networks. Organ. Sci. 23(2):434–448.

Amesse F, Latour R, Rebolledo C, Seguin-Dulude L (2004)The telecommunications equipment industry in the 1990s:From alliances to mergers and acquisitions. Technovation24(11):885–897.

Argote L (1999) Organizational Learning: Creating, Retaining, andTransferring Knowledge (Kluwer Academic Publishers, Boston).

Argyris C, Schon DA (1978) Organizational Learning (Addison-Wesley, Reading, MA).

Ariño A, de la Torre J (1998) Learning from failure: Towardsan evolutionary model of collaborative ventures. Organ. Sci.9(3):306–325.

Armour HO, Teece DJ (1978) Organizational structure and economicperformance: A test of the multidivisional hypothesis. Bell J.Econom. 9(1):106–122.

Beamish PW, Kachra A (2004) Number of partners and JV perfor-mance. J. World Bus. 39(2):107–127.

Bearman P (1997) Generalized exchange. Amer. J. Sociol. 102(5):1383–1415.

Blodgett LL (1992) Factors in the instability of international jointventures—An event history analysis. Strategic Management J.13(6):475–481.

Boehmke FJ, Morey DS, Shannon M (2006) Selection bias andcontinuous-time duration models: Consequences and a proposedsolution. Amer. J. Political Sci. 50(1):192–207.

Brass DJ, Burkhardt ME (1993) Potential power and power use—Aninvestigation of structure and behavior. Acad. Management J.36(3):441–470.

Brewer MB, Brown RJ (1998) Intragroup relations. Gilbert DT, FiskeST, Lindzay G, eds. Handbook of Social Psychology (McGraw-Hill, Boston), 554–594.

Chung S, Singh H, Lee K (2000) Complementarity, status similar-ity and social capital as drivers of alliance formation. StrategicManagement J. 21(1):1–22.

Cohen WM, Levinthal DA (1990) Absorptive capacity—A newperspective on learning and innovation. Admin. Sci. Quart.35(1):128–152.

Costenbader E, Valente TV (2003) The stability of centrality measureswhen networks are sampled. Soc. Networks 25(4):283–307.

Cross M (1992) Japan’s fifth generation fails the artificial intelligencetest. Independent (London) (June 8). Retrieved from Factiva.

Das TK, Kumar R (2009) Interpartner harmony in strategic alliances:Managing commitment and forbearance. Internat. J. Bus.Alliances 1(1):24–52.

Das TK, Teng BS (2002a) Alliance constellations: A social exchangeperspective. Acad. Management Rev. 27(3):445–456.

Das TK, Teng BS (2002b) The dynamics of alliance conditionsin the alliance development process. J. Management Stud.39(5):725–746.

Deutsch M (1994) Constructive conflict resolution: Principle, trainingand research. J. Soc. Issues 50(1):13–32.

Dhanaraj C, Beamish PW (2004) Effect of equity ownership on thesurvival of international joint ventures. Strategic Management J.25(3):295–305.

Dussauge P, Garrette B, Mitchell W (2000) Learning from competingpartners: Outcomes and durations of scale and link alliancesin Europe, North America and Asia. Strategic Management J.21(2):99–126.

Dyer JH, Singh H (1998) The relational view: Cooperative strategyand sources of interorganizational competitive advantage. Acad.Management Rev. 23(4):660–679.

Garcia-Canal E (1996) Contractual form in domestic and internationalstrategic alliances. Organ. Stud. 17(5):773–794.

Garcia-Canal E, Valdes-Llaneza A, Arino A (2003) Effective-ness of dyadic and multi-party joint ventures. Organ. Stud.24(5):743–770.

Gibson C, Vermeulen F (2003) A healthy divide: Subgroups as astimulus for team learning behavior. Admin. Sci. Quart. 48(2):202–239.

Granovetter M (1985) Economic action and social structure: Theproblem of embeddedness. Amer. J. Sociol. 91(3):481–510.

Greig M (2001) Moments of opportunity: Recognizing conditionsof ripeness for international mediation between enduring rivals.J. Conflict Resolution 45(6):691–718.

Greve HR, Mitsuhashi H, Baum JA (2013) Greener pastures: Out-side options and strategic alliance withdrawal. Organ. Sci.24(1):79–98.

Greve HR, Baum JAC, Mitsuhashi H, Rowley T (2010) Built to lastbut falling apart: Cohesion, friction, and withdrawal from inter-firm alliances. Acad. Management J. 53(4):846–864.

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

Gulati R (1995b) Social structure and alliance formation patterns:A longitudinal analysis. Admin. Sci. Quart. 40(4):619–652.

Gulati R (1998) Alliances and networks. Strategic Management J.19(4):293–317.

Gulati R, Gargiulo M (1999) Where do interorganizational networkscome from? Amer. J. Sociol. 104(5):1439–1493.

Gulati R, Nickerson JA (2008) Interorganizational trust, governancechoice, and exchange performance. Organ. Sci. 19(5):688–708.

Gulati R, Singh H (1998) The architecture of cooperation: Manag-ing coordination costs and appropriation concerns in strategicalliances. Admin. Sci. Quart. 43(4):781–814.

Gulati R, Nohria N, Zaheer A (2000) Strategic networks. StrategicManagement J. 21(3):203–215.

Gulati R, Sytch M, Tartarynowicz A (2012) The rise and fall of smallworlds: Exploring the dynamics of social structure. Organ. Sci.23(2):449–471.

Hammonds K (2005) Kaleida Labs calls it quits. Bloomberg Business-week (December 3) http://www.businessweek.com/stories/1995-12-03/kaleida-labs-calls-it-quits.

Hansen MT, Løvås B (2004) How do multinational companiesleverage technological competencies? Moving from single tointerdependent explanations. Strategic Management J. 25(8/9):801–822.

Heckman JJ (1976) Sample selection as a specification error.Econometrica 47(1):153–161.

Helft M, Markoff J (2007) Google enters the wireless world. NewYork Times (November 6) http://www.nytimes.com/2007/11/06/technology/06google.html.

Hennart JF, Kim DJ, Zeng M (1998) The impact of joint venturestatus on the longevity of Japanese stakes in U.S. manufacturingaffiliates. Organ. Sci. 9(3):382–395.

Jaffe AB (1986) Technological opportunity and spillovers of researchand development—Evidence from firms’ patents, profits, andmarket value. Amer. Econom. Rev. 76(5):984–1001.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


King G, Zeng L (2001) Logistic regression in rare events data. Polit-ical Anal. 9(2):137–163.

Kogut B (1989) The stability of joint ventures—Reciprocity and com-petitive rivalry. J. Indust. Econom. 38(2):183–198.

Lau DC, Murnighan JK (1998) Demographic diversity and faultlines:The compositional dynamics of organizational groups. Acad.Management Rev. 23(2):325–340.

Lau DC, Murnighan JK (2005) Interactions within groups and sub-groups: The effects of demographic faultlines. Acad. Manage-ment J. 48(4):645–659.

Laumann EO, Marsden PV, Prensky D (1983) The boundary speci-fication problem in network analysis. Burt RS, Minor MJ, eds.Applied Network Analysis (Sage, Beverly Hills, CA), 18–34.

Lavie D, Lechner C, Singh H (2007) The performance implications oftiming of entry and involvement in multipartner alliances. Acad.Management J. 50(3):578–604.

Lewicki R, Bunker BB (1996) Developing and maintaining trust inwork relationship. Kramer RM, Tyler TR, eds. Trust in Orga-nizations: Frontiers of Theory and Research (Sage, ThousandOaks, CA), 114–139.

Li D (2013) Multilateral R&D alliances by new ventures. J. Bus.Venturing 28(2):241–260.

Li D, Eden L, Hitt MA, Ireland RD, Garrett RP (2012) Governancein multilateral R&D alliances. Organ. Sci. 23(4):1191–1210.

Li JT, Hambrick DC (2005) Factional groups: A new vantage ondemographic faultlines, conflict, and disintegration in workteams. Acad. Management J. 48(5):794–813.

Macaulay S (1963) Noncontractual relations in business: A prelimi-nary study. Amer. Sociol. Rev. 28:55–67.

Makino S, Chan CM, Isobe T, Beamish PW (2007) Intended andunintended termination of international joint ventures. StrategicManagement J. 28(11):1113–1132.

Mowery DC, Oxley JE, Silverman BS (1996) Strategic alliances andinterfirm knowledge transfer. Strategic Management J. 17(Spe-cial issue):77–91.

Oh H, Chung M, Labianca G (2004) Group social capital and groupeffectiveness: The role of informal socializing ties. Acad. Man-agement J. 47(6):860–875.

Park HD, Steensma H (2012) When does corporate venture cap-ital add value for new ventures? Strategic Management J.33(1):1–22.

Park SH, Russo MV (1996) When competition eclipses cooperation:An event history analysis of joint venture failure. ManagementSci. 42(6):875–890.

Park SH, Ungson GR (1997) The effect of national culture, orga-nizational complementarity, and economic motivation on jointventure dissolution. Acad. Management J. 40(2):279–307.

Parkhe A (1993) Partner nationality and the structure-performancerelationship in strategic alliances. Organ. Sci. 4(2):301–324.

Pearce RJ (1997) Toward understanding joint venture performanceand survival: A bargaining and influence approach to transactioncost theory. Acad. Management Rev. 22(1):203–225.

Phelps CC (2010) A longitudinal study of the influence of alliancenetwork structure and composition on firm exploratory innova-tion. Acad. Management J. 53(4):890–913.

Philips Business Information (1994) Reliance Comm/Tec, partnersdevelop video on demand. Fiber Optics News 14(8). Retrievedfrom Factiva.

Piore M, Sabel C (1984) The Second Industrial Divide (Basic Books,New York).

Podolny JM (1994) Market uncertainty and the social character ofeconomic exchange. Admin. Sci. Quart. 39(3):458–483.

Polidoro F, Ahuja G, Mitchell W (2011) When the social struc-ture overshadows competitive incentives: The effects of networkembeddedness. Acad. Management J. 54(1):203–223.

Polzer JT, Crisp CB, Jarvenpaa SL, Kim JW (2006) Extending thefaultline model to geographically dispersed teams: How colo-cated subgroups can impair group functioning. Acad. Manage-ment J. 49(4):679–692.

Prieger JE (2002) A flexible parametric selection model for non-normal data with application to healthcare usage. J. Appl.Econometrics 17(4):367–392.

Provan KG, Fish A, Sydow J (2007) Interorganizational networks atthe network level: A review of the empirical literature on wholenetworks. J. Management 33(3):479–516.

Reagans R, Zuckerman EW (2001) Networks, diversity, and produc-tivity: The social capital of corporate R&D teams. Organ. Sci.12(4):502–517.

Reuer JJ, Arino A (2002) Contractual renegotiations in strategicalliances. J. Management 28(1):47–68.

Ring PS, Van de Ven AH (1994) Developmental processes of cooper-ative interorganizational relationships. Acad. Management Rev.19(1):90–118.

Rogers WH (1993) Regression standard errors in clustered samples.Stata Tech. Bull. 3(13):19–23.

Rosenkopf L, Almeida P (2003) Overcoming local search throughalliances and mobility. Management Sci. 49(6):751–766.

Rosenkopf L, Padula G (2008) Investigating the microstructure ofnetwork evolution: Alliance formation in the mobile communi-cations industry. Organ. Sci. 19(5):669–687.

Rowley T, Behrens D, Krackhardt D (2000) Redundant governancestructures: An analysis of structural and relational embeddednessin the steel and semiconductor industries. Strategic ManagementJ. 21(3):369–386.

Schilling MA, Phelps CC (2007) Interfirm collaboration networks:The impact of large-scale network structure on firm innovation.Management Sci. 53(7):1113–1126.

Silverman BS (1999) Technological resources and the direction ofcorporate diversification: Toward an integration of the resource-based view and transaction cost economics. Management Sci.45(8):1109–1124.

Simmel G (1950) The Sociology of Georg Simmel (Free Press,Glencoe, IL).

Stuart TE (1998) Network positions and propensities to collabo-rate: An investigation of strategic alliance formation in a high-technology industry. Admin. Sci. Quart. 43(3):668–698.

Stuart TE (2000) Interorganizational alliances and the performanceof firms: A study of growth and innovation rates in a high-technology industry. Strategic Management J. 21(8):791–811.

Takahashi N (2000) The emergence of generalized exchange. Amer.J. Sociol. 10(4):1105–1134.

Teece DJ (1980) The diffusion of an administrative innovation. Man-agement Sci. 26(5):464–470.

Thatcher MB, Patel PC (2012) Group faultlines: A review, integration,and guide to future research. J. Management 38(4):969–1009.

Uzzi B (1996) The sources and consequences of embeddedness forthe economic performance of organizations: The network effect.Amer. Sociol. Rev. 61(4):674–698.

Uzzi B (1997) Social structure and competition in interfirm networks:The paradox of embeddedness. Admin. Sci. Quart. 42(1):35–67.

Walrand J, Varaiya P (2000) High Performance Communication Net-works, 2nd ed. (Morgan Kaufmann, New York).

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Wasserman S, Faust K (1994) Social Network Analysis (CambridgeUniversity Press, New York).

Xia J (2011) Mutual dependence, partner substitutability, and repeatedpartnership: The survival of cross-border alliances. StrategicManagement J. 32(3):229–253.

Young-Ybarra C, Wiersma M (1999) Strategic flexibility in informa-tion technology alliances: The influence of transaction cost eco-nomics and social exchange theory. Organ. Sci. 10(4):439–459.

Zaheer A, Venkatraman N (1995) Relational governance as interor-ganizational strategy: An empirical test of the role of trust ineconomic exchange. Strategic Management J. 16(5):373–392.

Zeng M, Chen XP (2003) Achieving cooperation in multipartyalliances: A social dilemma approach to partnership manage-ment. Acad. Management Rev. 28(4):587–607.

Zollo M, Reuer JJ, Singh H (2002) Interorganizational routines andperformance in strategic alliances. Organ. Sci. 13(6):701–713.

Ralph A. Heidl is an assistant professor of managementat the Eli Broad College of Business, Michigan State Univer-sity. He received his Ph.D. from the Foster School of Busi-ness, University of Washington. His research interests includecollaborative networks, knowledge transfer, and technologyentrepreneurship.

H. Kevin Steensma is a professor of management andorganization at the Foster School of Business, University ofWashington, Seattle. He received his Ph.D. from Indiana Uni-versity Bloomington. His research interests include alliances,knowledge spillover, and intellectual property strategy.

Corey Phelps is an associate professor of strategy and busi-ness policy at HEC Paris. He received his Ph.D. from the SternSchool of Business, New York University. His research inter-ests include external corporate venturing, interorganizationalnetworks, and innovation.

Dow

nloa

ded

from

info

rms.

org

by [

193.

51.1

6.19

4] o

n 08

Apr

il 20

14, a

t 06:

18 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Documents

Divisive Faultlines and the Unplanned Dissolutions of Multipartner Alliances