bridging ties: a source of firm heterogeneity in competitive capabilities

24
Strategic Management Journal Strat. Mgmt. J., 20: 1133–1156 (1999) BRIDGING TIES: A SOURCE OF FIRM HETEROGENEITY IN COMPETITIVE CAPABILITIES BILL McEVILY 1 * and AKBAR ZAHEER 2 1 Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh, Pennsylvania, U.S.A. 2 Carlson School of Management, University of Minnesota, Minneapolis, Minne- sota, U.S.A. What explains differences in firms’ abilities to acquire competitive capabilities? In this paper we propose that a firm’s embeddedness in a network of ties is an important source of variation in the acquisition of competitive capabilities. We argue that firms in geographical clusters that maintain networks rich in bridging ties and sustain ties to regional institutions are well- positioned to access new information, ideas, and opportunities. Hypotheses based on these ideas were tested on a stratified random sample of 227 job shop manufacturers located in the Midwest United States. Data were gathered using a mailed questionnaire. Results from structural equation modeling broadly support the embeddedness hypotheses and suggest a number of insights about the link between firms’ networks and the acquisition of competitive capabilities. Copyright 1999 John Wiley & Sons, Ltd. Since capabilities are critical to the pursuit of competitive advantage (Nelson, 1991; Teece, Pis- ano, and Shuen, 1997), understanding differences in firms’ competitive capabilities is a central question in the field of strategy. Among the more prominent explanations for differences in com- petitive capabilities are imperfections in factor markets or luck (Barney, 1986); path dependence (David, 1985); causal ambiguity and uncertain imitability (Lippman and Rumelt, 1982); and relatively immobile internal resources (Barney, 1991). In contrast to these, and other eco- nomically derived approaches that take an atom- istic view by assuming that firms act alone, this research takes an embeddedness perspective. Specifically, the embeddedness perspective high- lights the role of firms’ social, economic, and professional networks (Granovetter, 1985) to explain economic actions such as alliance forma- Key words: network resources; firm capabilities; geo- graphical clusters; bridging ties *Correspondence to: Bill McEvily, 323 GSIA, Carnegie Mel- lon University, Pittsburgh, PA 15213, U.S.A. CCC 0143–2095/99/121133–24 $17.50 Received 30 July 1997 Copyright 1999 John Wiley & Sons, Ltd. Final revision received 19 July 1999 tion (Gulati, 1995), interfirm exchange (Uzzi, 1997), and organizational survival (Baum and Oliver, 1992). We extend this literature by exploring how firms’ embeddedness in geographi- cal clusters relates to their competitive capabili- ties. A recently renewed interest in the phenomenon of geographical clusters, also known as industrial districts (Piore and Sabel, 1984) or ‘hot spots’ (Pouder and St. John, 1996), has spurred research in economic geography (Krugman, 1991; Scott, 1992), sociology (Lazerson, 1988), political science (Sabel, 1993), and international and stra- tegic management (Porter, 1990). This research documents the impressive economic growth and vitality of regions such as the ‘Third Italy’ (Pyke and Sengenberger, 1990) and Silicon Valley (Saxenian, 1994). A prominent feature of these and similar geographical clusters is extensive interfirm networks supporting frequent and repeated knowledge sharing and collaborative innovation. The literature on geographical clusters argues that firms within a region tend to exhibit

Upload: bill-mcevily

Post on 06-Jun-2016

216 views

Category:

Documents


2 download

TRANSCRIPT

Strategic Management JournalStrat. Mgmt. J.,20: 1133–1156 (1999)

BRIDGING TIES: A SOURCE OF FIRMHETEROGENEITY IN COMPETITIVE CAPABILITIES

BILL McEVILY1* and AKBAR ZAHEER2

1Graduate School of Industrial Administration, Carnegie Mellon University,Pittsburgh, Pennsylvania, U.S.A.2Carlson School of Management, University of Minnesota, Minneapolis, Minne-sota, U.S.A.

What explains differences in firms’ abilities to acquire competitive capabilities? In this paperwe propose that a firm’s embeddedness in a network of ties is an important source of variationin the acquisition of competitive capabilities. We argue that firms in geographical clusters thatmaintain networks rich in bridging ties and sustain ties to regional institutions are well-positioned to access new information, ideas, and opportunities. Hypotheses based on theseideas were tested on a stratified random sample of 227 job shop manufacturers located in theMidwest United States. Data were gathered using a mailed questionnaire. Results from structuralequation modeling broadly support the embeddedness hypotheses and suggest a number ofinsights about the link between firms’ networks and the acquisition of competitive capabilities.Copyright 1999 John Wiley & Sons, Ltd.

Since capabilities are critical to the pursuit ofcompetitive advantage (Nelson, 1991; Teece, Pis-ano, and Shuen, 1997), understanding differencesin firms’ competitive capabilities is a centralquestion in the field of strategy. Among the moreprominent explanations for differences in com-petitive capabilities are imperfections in factormarkets or luck (Barney, 1986); path dependence(David, 1985); causal ambiguity and uncertainimitability (Lippman and Rumelt, 1982); andrelatively immobile internal resources (Barney,1991). In contrast to these, and other eco-nomically derived approaches that take an atom-istic view by assuming that firms act alone, thisresearch takes anembeddednessperspective.Specifically, the embeddedness perspective high-lights the role of firms’ social, economic, andprofessional networks (Granovetter, 1985) toexplain economic actions such as alliance forma-

Key words: network resources; firm capabilities; geo-graphical clusters; bridging ties*Correspondence to: Bill McEvily, 323 GSIA, Carnegie Mel-lon University, Pittsburgh, PA 15213, U.S.A.

CCC 0143–2095/99/121133–24 $17.50 Received 30 July 1997Copyright 1999 John Wiley & Sons, Ltd. Final revision received 19 July 1999

tion (Gulati, 1995), interfirm exchange (Uzzi,1997), and organizational survival (Baum andOliver, 1992). We extend this literature byexploring how firms’ embeddedness in geographi-cal clusters relates to their competitive capabili-ties.

A recently renewed interest in the phenomenonof geographical clusters, also known as industrialdistricts (Piore and Sabel, 1984) or ‘hot spots’(Pouder and St. John, 1996), has spurred researchin economic geography (Krugman, 1991; Scott,1992), sociology (Lazerson, 1988), politicalscience (Sabel, 1993), and international and stra-tegic management (Porter, 1990). This researchdocuments the impressive economic growth andvitality of regions such as the ‘Third Italy’ (Pykeand Sengenberger, 1990) and Silicon Valley(Saxenian, 1994). A prominent feature of theseand similar geographical clusters is extensiveinterfirm networks supporting frequent andrepeated knowledge sharing and collaborativeinnovation.

The literature on geographical clusters arguesthat firms within a region tend to exhibit

1134 B. McEvily and A. Zaheer

‘similar resources, cost structures, mental mod-els, and competitive behavior’ (Pouder and St.John, 1996: 1195). In effect, this stream ofresearch assumes that firms in these regionsare homogeneous and achieve similar levels ofperformance. However, other research suggeststhat whereas some firms manage to acquireand maintain the capabilities to successfullycompete, others languish with obsolete skillsand routines (Saxenian, 1994). The naturalquestion, then, is what accounts for the vari-ation in competitive capabilities of firms withina geographical cluster?

We explain difference in firms’ competitivecapabilities by viewing economic action asembedded in firms’ networks of ties, includingnonmarket ties (Oliver, 1996). From this stand-point, firm actions and outcomes are substan-tially influenced by the ongoing pattern ofrelationships maintained with other firms andnonmarket organizations. Specifically, theembeddedness perspective stresses that‘net-works of social relations penetrate irregularlyand in differing degrees in different sectorsof economic life%’ (Granovetter, 1985: 491.Emphasis added). Applied to geographical clus-ters, the embeddedness perspective suggests thatfirms are embedded in highly differentiatedways that link them to different sets of playersand thereby present them with sharply distinctopportunities and constraints. Put differently,firms vary in terms of their potential to discoverand exploit competitive capabilities throughtheir networks. Consequently, it is important tounderstand how a firm’s distinctive pattern ofnetwork ties relates to its competitive cap-abilities.

Extant research on firm capabilities has focusedprimarily on the link between capabilities andperformance-related outcomes (Lieberman, Lau,and Williams, 1990; Clark and Fujimoto, 1991;Henderson and Cockburn, 1994). However, farless research attention has been paid to thesourcesof firm capabilities. What research hasbeen conducted in this area has focused onsources internal to the firm. In contrast, wemaintain that there are importantexternalsources of capabilities that firms draw uponto varying degrees (Galaskiewicz and Zaheer,1999). We propose that these ‘networkresources’ (Gulati, 1999) enable and constrainfirms’ abilities to acquire competitive capabili-

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

ties through differential exposure to informationand opportunities.

We highlight two key differentiating facetsof firms’ network resources:bridging ties andlinkages to regional institutions. The networkon which we focus in this study is the firm’sset of linkages to external sources of advice.Focusing on the diversity of a firm’s advicenetwork, we develop the concept of bridgingties. By linking the firm’s internal capabilitieswith its external network of bridging ties, wepropose to understand and explain differencesin competitive capabilities among firms in ageographical cluster.

Firms’ network resources in geographicalclusters are distinguished further by linkages toregional institutions. We argue that regionalinstitutions facilitate the development of com-petitive capabilities among local firms by actingas network intermediaries for interaction andinformation exchange among firms. Althoughthe services provided by regional institutionsare available to all the firms in a geographicalcluster, not all firms participate to the samedegree or benefit to the same extent. Participat-ing in regional institutions raises issues ofcooperation with competitors and the potentialexposure of proprietary information. Therefore,a question arises as to which firms actually douse the available services and to what extentregional institutions contribute to the develop-ment of firms’ competitive capabilities.

We propose that heterogeneity in firms’ net-works of bridging ties and variations in theirlinkages to regional institutions are importantsources of differences in firms’ competitive capa-bilities in a geographical cluster. For strategyresearch, this proposition suggests a need torevisit a basic premise of the resource-based viewof the firm—that atomistic firms generate capa-bilities internally. Our research also addressesa central debate in the network literature overalternative mechanisms underlying informationaccess: Is the causal mechanism structural holes,as Burt (1992) proposes, or tie strength, asGranovetter (1973) argues? By examining ques-tions that lie at the intersection of strategy andnetwork research—such as the influence of inter-firm network structure on intra-firm competitivecapabilities—we advance both streams ofresearch and show how they can be usefullyintegrated.

Bridging Ties 1135

KNOWLEDGE SHARING INGEOGRAPHICAL CLUSTERS

Following Porter (1990), we define a geographi-cal cluster as a spatially concentrated group offirms competing in the same or related industriesthat are linked through vertical (buyer–supplier)or horizontal (technology, information, or otherresource-sharing) relationships. Illustrative ofsuch clusters are the semiconductor and otherelectronics firms located in Silicon Valley and theRoute 128 area of Boston. Through professional,social, and exchange relationships, firms in theseclusters share advice, engineering solutions, andinformation about new technologies and practices(Nohria, 1992; Saxenian, 1994).

The transfer of situation-specific knowledgeand the creation and exchange of new ideas isan important attribute distinguishing geographicalclusters from market coordination. Marshallrecognized this point nearly a century ago, not-ing that

inventions and improvements in machinery, inprocesses and the general organization of thebusiness have their merits promptly discussed; ifone man starts a new idea, it is taken up byothers and combined with suggestions of theirown; and thus it becomes the source of newideas (1920: 271).

More recently, Best notes that geographical clus-ters have

the institutional capacity to continuously learn,adjust, and improve in economic performance.% [I]nitiatives from one firm intersect withothers and modify the production capabilities andopportunities for each firm (1990: 235).

Network-based explanations emphasize that tight-knit industrial communities are characterized byhigh levels of trust that enable such knowledgesharing among firms (Saxenian, 1994). In thesecommunities, the network of relationships amongfirms is typically characterized as a web of denseand overlapping ties. Via this web, knowledge israpidly diffused throughout the geographical clus-ter.

Yet, taken to the logical extreme, it is unclearhow new ideas and information enter an entirelyclosed social system. Firms interacting exclu-sively with each other presumably would maintaina similar pattern of ties and thereby be relatively

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

insulated from developments and advances inexternal communities. It seems unlikely, however,that all firms in a geographical cluster would infact maintain an identical pattern of linkages.Interfirm relationships are frequently formed as aresult of family ties, membership in civic groupsand social clubs, and spin-offs from the sameparent corporations or universities. Firm networksthat share such common heritage overlap moreextensively. Stinchcombe’s (1965) notion thatfirms created under similar environmental circum-stances are likely to share certain attributes,including common sets of relationships (Kogut,1993) is apropos. Correspondingly, firms’ net-works are likely to be quite distinct from thoseof other firms with whom they do not share acommon history. Viewed this way, geographicalclusters contain ‘cliques’ or subclusters of thickand multiplex networks, with fewer linkagesbetween subclusters. This perspective of geo-graphical clusters suggests that firms maintainunique and idiosyncratic patterns of network link-ages and are consequently differentially exposedto new knowledge, ideas, and opportunities.

Knowledge sharing in geographical clusters isalso influenced by the common infrastructure sup-porting economic activity within a region. A cen-tral theme of the geographical clusters literatureis the notion that firms co-located in close geo-graphical proximity can benefit from agglomer-ation effects by drawing on a common infrastruc-ture (Maarten de Vet and Scott, 1992; Porter,1990). The common infrastructure of geographi-cal clusters typically includes pools of skilledlabor, available capital, qualified suppliers, pro-fessional service firms, and R&D labs (Best,1990; Saxenian, 1994; Romo and Schwartz,1995).

Our focus is one specific element of thisinfrastructure—regional institutions. For purposesof this research, we define regional institutionsas locally-oriented organizations that provide ahost of collective support services to firms in theregion. Examples of regional institutions includetechnical assistance centers, university outreachprograms, vocational training centers, and localresearch institutes. While some trade associationsmay fulfill the criteria of regional institutions aswe define them (e.g., SEMATECH), others aregeared primarily toward lobbying activities (e.g.,the Semiconductor Industry Association) and areoutside the scope of our definition. Also excluded

1136 B. McEvily and A. Zaheer

from our concept of regional institutions aresocial and civic organizations such as the KiwanisClub or the United Way.

THEORY AND HYPOTHESES

Network Resources

We view network resources as representing theinformational advantages associated with a firm’snetwork of ties (Gulati, 1999). To sustain a com-petitive advantage firms must constantly seek outnew opportunities for upgrading and renewingtheir capabilities. However, acquiring capabilitiesentails uncertainty regarding the value of thecapability and the extent to which it can benefitthe firm. Consequently, firms may benefit fromhaving a network of knowledgeable contacts thatprovide a reliable source of information aboutoptions for enhancing competitive capabilities.

Not all firms possess comparable levels ofnetwork resources, and the variation in networkresources across firms influences their individualability to discover and exploit useful information.Specifically, firms’ networks vary in terms ofstructure, or the pattern of ties, and nodal hetero-geneity, or the variation in the mix of contactsin firms’ networks (Galaskiewicz and Zaheer,1999). We argue that network structures rich inbridging ties offer superior informational advan-tages about competitive capabilities. Further,nodal heterogeneity in the form of linkages topublic regional institutionsalso enables the acqui-sition of competitive capabilities.1

Bridging ties

We define bridging ties as those that link a focalfirm to contacts in economic, professional, andsocial circles not otherwise accessible to the firm.Specifically, ‘a bridging tie between two personsis the sole path by means of which the twopersons (and their direct contacts) are joined ina network’ (Friedkin, 1980: 411). We theorizethat bridging ties connect a focal firm to sourcesof information and opportunities that are not

1 ‘Acquisition of competitive capabilities’ refers to the processthrough which firms discover, internalize, and exploit industry-specific practices, methods, and techniques for achieving com-petitive advantage. By ‘acquisition’ we do not mean eitherthe outsourcing or the wholesale purchase of capabilities.

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

available from other network contacts. In geo-graphical clusters, firms that act as connectorsbetween subclusters maintain bridging ties.

The concept of bridging ties has its basis intheories of social networks, which argue thatnew information is obtained through ‘weak ties’(Granovetter, 1973) and ‘structural holes’ (Burt,1992). Granovetter posits that new information isobtained through casual acquaintances (weak ties)rather than through close personal friends (strongties). The ‘strength of weak ties’ thesis is basedon the premise that strong ties characterize adense cluster of actors who are all mutuallyconnected to each other. Since this subcluster ofstrongly connected actors is likely to interactfrequently, much of the information circulatingin this social system is redundant.

Conversely, weak ties are often links withactors who move in social circles other than thoseof the focal actor. Weak ties enable the discoveryof opportunities because they serve as bridges tonew and different information. In support of thistheory, Granovetter (1973) finds that a person’scasual acquaintances, with whom only sporadiccontact is maintained, are critical to discoveringnew employment opportunities.

Granovetter conceptualizes weak ties as therelative infrequency of interactionbetween thefocal actor and contacts. However, infrequencyof interaction alone may not be a sufficient con-dition for discovering opportunities if an actor’sweak tie contacts are either connected to eachother or connected to the actor’s strong tie con-tacts. Burt (1992) extends the strength of weakties argument by asserting that it is not so muchthe strength or weakness of a tie that determinesits information potential, but rather whether astructural hole exists between a focal actor’scontacts. In other words, the causal agentdetermining whether a tie will provide access tonew information and opportunities is the extentto which it is nonredundantor, in Burt’s termin-ology, whether a tie spans a structural hole. AsFigure 1 illustrates, the weak tie between Actors1 and 2 serves as a bridge by connecting thefocal actor (Actor 1) to a contact (Actor 2)disconnected from the focal actor’s other directcontacts. However, as shown from the linkagebetween Actors 1 and 3, strong ties may serveequally as a bridge, since the connection is nonre-dundant.

While both arguments offer compelling theo-

Bridging Ties 1137

Figure 1. Illustration of weak ties and structural holes

retical explanations for the discovery of newopportunities, a question remains as to whetherinfrequency of interaction (weak ties) or nonre-dundancy (structural holes) offers a more empiri-cally accurate explanation. A key objective ofthis paper is to compare the relative explanatorypower of these two theories of bridging ties.

Our conception of bridging ties is consistentwith both Granovetter (1973) and Burt (1992) butdiffers in important ways. Whereas Granovetter(1973) assumes that ‘all bridges are weak ties,’we argue that bridges are not always weak ties.We also recognize Burt’s position that a bridgingtie spans a structural hole. However in contrastto Burt, who argues that a bridge is ‘a chasmspanned, and the span itself’ (Burt, 1992: 28),we maintain that there is in fact a distinctionbetween a bridging tie and the structural hole itspans. In our view, the structural hole (chasm)presents information opportunities, but the bridg-ing tie (span) is how an actor exploits thoseopportunities to realize certain benefits (e.g.,information).

Bridging ties in firms’ advice networks

In their original formulations, the concepts ofweak ties and structural holes were used toexplain individual career mobility on the onehand and industry profitability on the other. How-ever, little research exists that has consideredthese theories in explaining firm-level outcomes.We propose that the causal mechanisms underly-

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

ing these theories are equally applicable to under-standing firm-level actions resulting from infor-mation search. We see firms’ acquisition ofcompetitive capabilities as significantly influencedby the search for novel information through net-works. From this perspective, firms use their net-works entrepreneurially by searching for ideasand opportunities to achieve competitive advan-tage. Firms’ advice networks are likely to be oneimportant channel through which a firm(particularly a small one) learns about new prac-tices and techniques and other opportunities forenhancing its competitiveness.2

We conceptualize bridging ties as embodied inthree elements, two of which (nonredundancy andinfrequency of interaction) are drawn from theliterature reviewed earlier. We definenonredun-dancy as the extent to which the contacts in afocal firm’s advice network are not linked to oneanother. Infrequency of interactionrefers to therarity with which a focal firm converses with thecontacts in its advice network. To these twoconcepts we add a third,geographic dispersion,which is particularly relevant to geographicalclusters.

Geographic dispersionin the advice networkis a proxy for face-to-face interaction amongadvisors. In our view, a dispersed advice networkis comprised of contacts located far enough apartspatially to impede routine face-to-face inter-action. With a geographically scattered advicenetwork, contacts interact less readily with thefocal firm, as well as with each other, and there-fore provide nonredundant information.

We argue that bridging ties exist when highnonredundancy, infrequency of interaction, andgeographic dispersion characterize a firm’s advicenetwork. We view these three concepts as com-plementary formulations, because each capturesin a different way the potential (or the lackthereof) for bridging ties. Nonredundancy is basedon the structural configurationof a firm’s con-tacts relative to each other (i.e., the presence or

2 A question may be raised about the applicability of individ-ual-level network theories to explain organizational actions.In our theoretical and empirical framing we have focused onsmall organizations that are typically owner-managed (91%of the firms in this study are owner-managed). The owner-CEO is invariably the only person who has the authority forall major decisions taken by the organization. In this regard,the advice network of the owner-manager is tantamount tothat of the small organization.

1138 B. McEvily and A. Zaheer

absence of links among contacts in the network),infrequency of interaction emphasizes thetiestrength between a focal firm and its contacts,and geographic dispersion focuses on thespatiallocation of contacts.

Bridging ties as a source of competitivecapabilities

Networks have the potential of conferring bothinformation and control benefits (Burt, 1992).Our focus in this paper is on information benefits,especially those occurring from access, which isdefined as ‘receiving a valuable piece of infor-mation and knowing how to use it’ (Burt, 1992:13). We maintain that firms differ in theircapacity to access useful information throughtheir networks, and that these differences are akey source of variation in firms’ competitivecapabilities.

Specifically, we propose that firms with advicenetworks rich in nonredundant ties will acquirecompetitive capabilities to a greater extent thanfirms with advice networks lacking in nonredun-dant ties. Actors with such optimized networksgain access to useful information to a greaterextent because ‘they know about% morerewarding opportunities’ (Burt, 1992: 13). How-ever, the information benefits stemming from net-works are not automatic. Rather, establishing andsustaining nonredundant contacts ‘in the placewhere useful bits of information are likely to air,and % providing a reliable flow of informationto and from those places’ (Burt, 1992: 15) isthe overriding determinant of information access.When actors seek out contacts that are tappingfundamentally different informational domains,they are likely to discover unique opportunitiesand information not available from a networkwith redundant ties. On the other hand, redundantties are likely to produce information that islargely superfluous and unoriginal. Thus, therange, novelty, and diversity of informationobtained from a nonredundant tie are predictedto be much greater than from a redundant tie.

When advisors in a firm’s advice network areall tied to each other, they will have access tosimilar information about competitive capabilities.In other words, little fresh news circulates in ahighly redundant social system. In contrast, if allof a firm’s advisors are complete strangers toone another, the advice network will be high in

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

nonredundancy. In such a case, all of the firm’sadvisors will operate in different economic, pro-fessional, and social circles and know of differentopportunities to innovate and develop competi-tive capabilities.

Accordingly, firms’ contacts ‘are ports ofaccess to clusters of people’ (Burt, 1992: 23)that serve as independent sources of information.Firms that maintain linkages to diverse infor-mation sources gain access to novel informationand learn about competitive capabilities to agreater degree than firms without such ties. Bymaintaining such a network of diverse and non-overlapping ties, a firm increases its chances ofgaining access to new ideas and information aboutcompetitive capabilities.

We define competitive capabilities, followingAmit and Schoemaker, as ‘a firm’s capacity todeploy resources% using organizational proc-esses’ (1993: 35) to achieve a strategic objective.We view acquisition of capabilities as a continu-ous process that begins with knowledge andawareness of some opportunity, continues with adecision to internalize the capability, and endseventually in firm actions to implement the capa-bility. We propose that a firm’s exposure via itsnetwork to more diverse sources of informationis commensurate with a richer set of opportuni-ties. Thus, by raising awareness about a capabilityand the potential benefit to the firm, networkdiversity initiates the process of capability acqui-sition. Following the initial learning about a capa-bility, firms engage in a process of evaluationand eventual adoption. Since nonredundantadvisors are likely to possess different infor-mation about a given capability, the focal firmcumulates superior information about the potentialadvantages and disadvantages associated with thecapability. Put differently, when a particular capa-bility is beneficial to the focal firm, a nonredun-dant network aids the firm in realizing the capa-bility’s true potential. This is akin to the argumentmade by Uzzi that through embeddedness, net-work ties improve decision making because they‘appear to reduce bounded rationality byexpanding the range of data attended to and thespeed of processing’ (1997: footnote, p. 46). Thisbrings us to our first hypothesis:

Hypothesis 1: Nonredundancy in the firm’sadvice network will positively influence itsacquisition of competitive capabilities.

Bridging Ties 1139

Similarly, an advice network composed of con-tacts with which the focal firm interacts frequentlyis less likely to contain bridging ties. Granovetter(1973) posits that the contacts with which anactor interacts frequently (i.e., strong ties) tendto interact with each other frequently as well. Inthese cases, we should observe a tightly knitclique, relatively isolated from external sourcesof information. The underlying reasoning for thisprediction is that a disproportionate amount oftime spent interacting with those in the cliqueleaves less time to interact with actors from otherparts of a social system.

Granovetter’s theory of weak ties also suggeststhat the less a firm interacts with the contacts inits advice network, the less a firm’s advisorsinteract with each other. On this point Granovetterexplicitly argues that ‘two factors—time andsimilarity—indicate why weaker A-B and A-Cties make a C-B tie less likely%: C and B areless likely to interact and less likely to be compat-ible if they do’ (1973: 1362). Further, where afirm’s advisors interact with each otherinfrequently they are more likely to operate indifferent economic, professional, and socialcircles. Hence, the advisors are more likely tobe privy to diverse information and ideas. Bymaintaining weak ties with advisors, who areeach likely to be tied into sources of informationdifferent from the focal firm’s other advisors, thefirm accesses a greater diversity of advice aboutacquiring competitive capabilities. These ideas aresummarized in the following hypothesis:

Hypothesis 2: Infrequency of interaction withthe firm’s advice network willpositively influ-ence its acquisition of competitive capabilities.

We also expect to find an advice network richin bridging ties when advisors are geographicallydispersed. Building on the preceding discussion,we argue that in a scattered network, interactionamong a focal firm and its contacts is likely tobe limited. As the geographic distance betweenthe focal firm and its advisors increases, theopportunity for face-to-face interaction is lowerand it is more difficult to maintain strong ties.In contrast, an advice network densely clusteredaround the firm provides greater opportunity fora firm to interact with its contacts and for thecontacts to interact with each other. As a result,the potential for the focal firm to access different

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

information, including learning about new capa-bilities, is more limited in a densely clusteredadvice network because each of the contacts islikely to know the same things that the othercontacts know. Hence, the diversity of infor-mation obtained by a firm is likely to increasewith the geographic dispersion of the firm’sadvice network. Diverse information about com-petitive capabilities will enhance the likelihoodthat the focal firm will learn about and acquirecompetitive capabilities. Accordingly, we hypoth-esize that:

Hypothesis 3: Geographic dispersion in thefirm’s advice network willpositively influenceits acquisition of competitive capabilities.

Participation in regional institutions andcompetitive capabilities

Beyond being embedded in a network of interfirmrelationships, firms are also embedded in abroader set of regional linkages. In geographicalclusters specifically, an important feature of thelocal infrastructure is regional institutions thatprovide collective support services to firms in theregion. The connection between firms’ forminglinkages with regional institutions (throughparticipation) and the acquisition of competitivecapabilities is based on regional institutions’specialized network role of intermediary. Asintermediaries, regional institutions facilitate theacquisition of competitive capabilities by compil-ing and disseminating knowledge and by reducingsearch costs.

Beyond providing specific support services(e.g., vocational training, market research, appliedR&D) and other resource benefits (Baum andOliver, 1992) to local firms, regional institutionsact as repositories for knowledge and opportuni-ties about competitive capabilities. Becauseregional institutions interact with a relatively largenumber of firms in the geographical cluster, theyare exposed to a wide variety of solutions toorganizational challenges typically faced by thefirms in a region. Based on broad experienceobserving others who have dealt with similarproblems, regional institutions compile and dis-seminate ready-to-use summaries about competi-tive capabilities and routines (Suchman, 1994).In effect, regional institutions facilitate managerialinnovation by providing access to information

1140 B. McEvily and A. Zaheer

and resources for acquiring new, and extendingexisting, capabilities.

Regional institutions also mitigate search costsassociated with locating external sources ofknowledge and specialized expertise critical tothe acquisition of competitive capabilities. Thesearch economies generated by intermediariesstem from their maintaining an extensive networkof ties to different parts of a social system.Individual members of a network thus can ‘relievethemselves of a heavy burden and free up theirown time and energy to engage in other activities’(Galaskiewicz, 1985: 113). Therefore, rather thanmaintaining numerous ties to different parts ofthe network, an actor can maintain a single con-nection with the intermediary that specializes inproviding access to and information about com-petitive capabilities.

A key insight developed in the social networkliterature is that intermediaries serve as go-betweens for potential exchange partners who areotherwise disconnected. In effect, intermediariesbridge the social gaps in a network by ‘linkingpersons having complementary interests, transfer-ring information, and otherwise facilitating theinterests of persons not directly connected to oneanother’ (Aldrich and Zimmer, 1986: 16). In thecontext of geographical clusters, rather than allfirms being tied to one another, each can maintaina single connection with the regional institutionthat specializes in providing access to and infor-mation about potential exchange partners.

In brief, regional institutions fulfill a networkintermediary role for participating firms by serv-ing as a repository of knowledge and by reducingsearch costs. Accordingly, we hypothesize that:

Hypothesis 4: Participation in regional insti-tutions will positively influence the firm’sacquisition of competitive capabilities.

Bridging ties and participation in regionalinstitutions

In spite of the seemingly obvious advantages ofparticipating in regional institutions, firms are notalways inclined to use their services. Whileregional institutions do provide firms with oppor-tunities to learn about new capabilities, parti-cipants are also exposed to the risk that pro-prietary information about their own competitive

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

advantage may be revealed to rival firms. Firmsmay hesitate to participate in regional institutionsbecause they are concerned that they may losemore than they gain.

Our preliminary fieldwork confirmed the reluc-tance of some firms to participate in regionalinstitutions because of concerns about leakage ofproprietary information. As one executive com-mented,

I think I find [the regional institution] moreuseful to our association as a group. I’m notsaying there’d never be anything we would workwith them on, but the problem in working withthem is whatever we do, and do develop, is nowpublic knowledge% anything they came in hereand did would be common knowledge andwouldn’t be proprietary to [our company].

Ambivalence about participating in a regionalinstitution stems from the fact that firms oftenneed to share a certain amount of informationabout their own operations with the regional insti-tution. To receive assistance with an organi-zational problem, a firm must provide some infor-mation about its current operations so that theissue can be analyzed and diagnosed. In manycases, this process involves a number of visitsto, and inspection of, the firm’s operations. Suchvisits present the potential for exposing pro-prietary and sensitive information, some of whichmay be unrelated to the specific problem at hand.

The knowledge, expertise, or skills developedin the process of resolving a specific organi-zational challenge confronted by a firm becomepart of the regional institution’s ‘stock of knowl-edge.’ As a quasi-public entity, the regional insti-tution seeks to spread the knowledge it acquiresthrough interactions with firms as widely as pos-sible. This knowledge may be shared with thefirm’s direct competitors. From this perspective,using the specialized services available from aregional institution may indirectly help a firm’scompetitors resolve similar problems. Conse-quently, participation in a regional institution maybe viewed as a threat to a firm’s strategic advan-tage.

We suggest that firms that have created net-works rich in bridging ties to access diversesources of information are more likely to beconscious of the value of information and thecompetitive risk of leakage of such informationthrough participating in regional institutions.

Bridging Ties 1141

Moreover, we expect that firms with advice net-works rich in bridging ties would want to keepproprietary whatever knowledge and skills theyhave acquired through their contacts. As before,bridging ties are conceptualized as nonredun-dancy, infrequency of interaction, and geographicdispersion. Based on these arguments we pre-dict that:

Hypothesis 5: Nonredundancy in the firm’sadvice network will negatively influence itsparticipation in regional institutions.

Hypothesis 6: Infrequency of interaction inthe firm’s advice network will negativelyinfluence its participation in regional insti-tutions.

Hypothesis 7: Geographic dispersion in thefirm’s advice network willnegatively influenceits participation in regional institutions.

RESEARCH METHODS

Research setting

The recent emergence and growth of a nationwideinfrastructure of regionally based industrial exten-sion centers provided an ideal opportunity forresearch relating acquisition of competitive capa-bilities with participation in regional institutions.The field setting for our research consisted ofgeographical clusters located in two states of theMidwest United States. These clusters are servedby industrial extension centers affiliated with anational system of such centers, which in thisstudy represents regional institutions.

Beginning in the late 1980s, a new set ofinstitutional arrangements began to appear in theU.S. with a specific focus on promoting thedeployment of new manufacturing technology,production techniques, and business practicesamong small manufacturers. This new publicinfrastructure is embodied in a system of regionalindustrial extension centers managed by theNational Institute of Standards and Technology(NIST). NIST assists small manufacturers withproductivity enhancement through upgrading theirmanufacturing technology and improving theirorganizational practices.

The cornerstone of NIST’s efforts to improvethe competitiveness of small manufacturers is

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

its Manufacturing Extension Partnership (MEP)initiative.3 The regionally based MEP centers pro-vide a common set of in-house services such asinformation ‘hot lines,’ training courses, work-shops and conferences, equipment demonstrations,supplier certification, and manufacturing assess-ments.

Each MEP center also coordinates a staff offield agents that performs a range of outreachactivities such as reconfiguration of existingmanufacturing operations, selection and instal-lation of new equipment, and identification andimplementation of new management practices. Allof the MEPs included in this study were foundedat the same time, have the same charter, aresimilar in size (in terms of number of employeesand annual budget), and offer some of thesame programs.

Because of the MEP program’s public charter,the services provided are widely available to thefirms in the regions. The manufacturing extensionprogram advertises the availability of its servicesthroughout the region by routinely mailing news-letters and other promotional material to localfirms, by attending local trade fairs, and throughother direct and indirect channels. The voluntarynature of the MEP program means that individualfirms choose to use the services available.

Research design and data collection

To test the hypotheses presented in this study,we used a stratified (by participation in a MEP)random sample of firms. A stratified random sam-pling strategy was used to ensure a sufficientdegree of variation in the participation variable.As Singlettonet al. note,

[S]tratifying by variables correlated with thedependent variable in a study increases precisionbecause it systematically introduces relevantsources of variability (or heterogeneity) in thepopulation into the sample (1988: 145)

In essence, stratifying contributes to samplingefficiency and eliminates error that would occur

3 A question may arise as to what the MEP centers’ owninterests and objectives are. MEP centers fit within the rubricof economic development programs that aim to preserve andcreate manufacturing jobs and enterprises, support innovation,enhance firms’ productivity and profitability, and strengthenregional manufacturing economies more generally.

1142 B. McEvily and A. Zaheer

if a relevant segment of the population werenot included.

The sampling frame consisted of all job shopmanufacturers located in several regions of twoMidwestern states and operating in the metal-working sector (i.e., electroplating, coating andpainting, printed circuit board manufacturing,screw machining, stamping, sheet metal fabri-cation, and machining). The two geographicalregions were selected because each operates anoffice of the state’s MEP. The metalworkingindustries were chosen in order to obtain roughlyequivalent proportions of firms participating ineach region’s MEP. In addition, these metal-working industries are all populated predomi-nantly by small and mid-sized firms and tend torely on similar production technologies.

A primary dataset was constructed using amailed questionnaire. Officials from the MEPs inboth states compiled and made available a list ofparticipating and nonparticipating firms in eachof the industries under study. From this list, allparticipating firms in each industry and an equalnumber of randomly selected nonparticipatingfirms in the same industry and state wereextracted.

A preliminary version of the survey instrumentwas pretested among a group of local executivesof job shop manufacturing companies from anindustry other than those just listed. The question-naire was administered to 22 job shop manufac-turers, including a roughly equal number of parti-cipants and nonparticipants. Feedback from theseexecutives was incorporated into a revised versionof the survey instrument, along with commentsand suggestions from industry experts, officialsfrom the MEPs, and several colleagues knowl-edgeable in survey design.

The final questionnaire was mailed to 1,000chief executive officers or presidents, who wereconsidered best able to respond to questions aboutorganizational and strategic issues relating to theirrespective firms. This approach is consistent withthe selection of key informants knowledgeableabout organizational matters by virtue of theirposition (John and Weitz, 1988). The sampleincluded approximately 500 participants and 500nonparticipants across both states and all metal-working industries. We also implementedDilman’s (1978) techniques for maximizing theresponse rate. Most important, in addition to theinitial survey mailing, extensive followup com-

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

munications were carried out. The followupincluded: (1) sending a reminder/thank you post-card two weeks after the first round of mailing,(2) sending a second round of mailings to nonre-spondents, and (3) placing a telephone call toany remaining nonrespondents.

A total of 309 executives responded to therequest for information about their company. Thisnumber is approximately 31% of the original1,000 firms surveyed. The followup telephonecalls revealed that 178 of the original 1,000 firmswere not eligible to participate in the study,4 andwe eliminated these firms from our initial sam-pling frame. The actual response rate then equaled38% (i.e., 309/822) of eligible firms. Completeresponses were obtained from 227 of these firms.All of the firms from which we obtainedresponses had 500 or fewer employees. Further,75% of the firms had 63 or fewer employees.Thus, the majority of firms in our sample weretruly small.

Testing for nonresponse bias

Although the overall response rate of 38% wasreasonable given the general reluctance of smallmanufacturers to respond to mailed surveys, thepossibility remained that the sample of respondingfirms systematically differed from the remainderof the population. We addressed the potentialfor nonresponse bias by comparing certain keyattributes of respondents (firm size in terms ofthe number of employees and annual sales) tothose of a group of 50 randomly selected nonre-spondents. We obtained size and sales data forthe 50 nonrespondents from one of the MEPs.T-tests revealed no significant differences betweenthe mean size (t= −1.29) and the mean sales(t = −1.83) of respondents and nonrespondents,although the near significance of the differencein the mean sales suggests the possibility of somebias in our sample toward firms with higher sales.To further confirm the representativeness of oursample, we conducted a Kolmogorov-SmirnovTwo Sample test.5 For the variables of both size(as total employees) and sales we found no sig-

4 Firms deemed ineligible to participate in our study includethose that were not job shop manufacturers, had gone out ofbusiness, or had zero employees.5 We are grateful to an anonymous reviewer for suggestingthis test to us.

Bridging Ties 1143

nificant differences between our sample and therandom sample of 50 nonrespondents. P-valueswere, respectively, 0.225 and 0.357, suggestingthat the two samples were drawn from the samepopulation. These tests provided adequate assur-ance that the sample of firms responding to thequestionnaire was representative of the broaderpopulation surveyed (Siegel, 1956).

Operational measures

This study used a multimeasure approach to oper-ationalize the theoretical constructs. Wheneverpossible, measurement instruments available fromextant research were used to operationalize thetheoretical constructs. Several instruments weremodified to make them more suitable for thecurrent research setting. Table 1 presents thedetails of the measurement instruments and scalesused to operationalize our theoretical constructs.The Cronbacha reliabilities for each constructare also reported in Table 1. With the exceptionof the quality management construct, which ismarginal at 0.61, constructs were at or above thevalue of 0.70 (Nunnally, 1978).

Acquisition of competitive capabilities

The acquisition of competitive capabilities con-struct is based on the ‘Assimilation of Innovation’scale developed by Meyer and Goes (1988).These authors argue that innovation adoption inthe organizational context is not simply a discreteevent that is carried out instantaneously by indi-vidual decision makers. Rather, an innovationpresents organizations with an adoption oppor-tunity and triggers a set of formal and informaldecision processes that may progress through thestages of awareness, evaluation, adoption, utili-zation, and eventually end in institutionalization.Similarly, we view competitive capabilities asbeing acquired through a multistage, organi-zational process since capability acquisition is aform of organizational innovation. A widelyaccepted definition of an innovation is any newproduct, process, practice, or idea (Rogers, 1995).This definition clearly encompasses the competi-tive capabilities studied and described in thispaper (i.e., pollution prevention, quality man-agement, and competitive scanning).

To develop a variable for statistical analysis,we created a 7-point scale that captures three

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

primary decision making stages associated withthe adoption of technology: knowledge–awareness, evaluation–choice, and adoption–implementation. Since we view the acquisition ofnew competitive capabilities as analogous to theadoption of technological innovation, we adaptedthe wording of the Meyer and Goes’ (1988)Assimilation of Innovation scale. Specifically, themodified scale refers to competitive capabilities,rather than technological innovations, and focuseson small, rather than large, organizations.

We evaluated three competitive capabilitiesusing the acquisition scale: pollution prevention,competitive scanning, and quality management.Following Amit and Schoemaker (1993), wefocused on industry-specific competitive capabili-ties that were identified during the field research.

In the initial phase of our research we conduc-ted a series of in-depth field interviews with13 individuals from different segments of themetalworking industries, including the owner–CEOs of job shop manufacturers and their cus-tomers and suppliers. We also interviewed indus-try consultants and field agents from regionalinstitutions working with the industry. All inter-views were conducted face-to-face at companylocations. A semistructured interview format wasused. The length of interviews varied from one tofour hours, and all interviews were tape recorded(approximately 30 hours) and transcribed(roughly 300 pages). A careful analysis of thewritten material and subsequent confirmation bythe interviewed industry experts suggested thathazardous materials management, minimization ofvariations in production processes, and awarenessof competitor’s strategy and behavior are centralto competing successfully in the metalworkingsector. For these reasons, we selected pollutionprevention, competitive scanning, and qualitymanagement as the industry-specific competitivecapabilities to study.

Pollution prevention

Pollution prevention capabilities have becomesalient with recent amendments to environmentalregulations, such as the Clean Water Act, whichgovern the use and emission of hazardousmaterials integral to the production process ofindustries in the metalworking sector. Pollutionprevention emphasizes the judicious use ofresources through source reduction, energy

1144 B. McEvily and A. Zaheer

Table 1. Measurement instruments

Measurement items Scales Internalconsistencyreliability

(a)

Pollution Prevention Capabilities1. Substitute less hazardous raw 1= We know little about this practice 0.72

materials for more hazardous ones 2= We know about this practice, but doNOT do it2. Offer new products/services because 3= We have considered doing this

of low waste disposal costs 4= We decidedNOT to do this after considering it3. Discontinue products/services high in 5= We do this from time to time

environmental management costs 6= We do this most of the time7 = We do this all of the time

Competitive Scanning Capabilities1. Monitor your competitors’ strategies 1= We know little about this practice 0.81

and tactics 2= We know about this practice, but doNOT do it2. Search for information about which 3= We have considered doing this

customers your competitors supply 4= We decidedNOT to do this after considering it3. Collect information about your 5= We do this from time to time

competitors’ market share 6= We do this most of the time7 = We do this all of the time

Quality Management Capabilities1. Collect data on your company’s 1= We know little about this practice 0.61

production process variations 2= We know about this practice, but doNOT do it2. Provide charts and graphs to 3= We have considered doing this

production employees reporting defect 4= We decidedNOT to do this after considering itrates 5= We do this from time to time

3. Conduct experiments to isolate causes 6= We do this most of the timeof defects 7= We do this all of the time

Participation1. Obtain on-site assistance at your 1= We know little about this service 0.75

company from [name of center] 2= We know about this service, but doNOT do it2. Select/install new equipment or 3= We have considered doing this

computer systems with [name of 4= We decidedNOT to do this after considering itcenter] 5= We did this once

3. Participate in user groups or networks 6= We did this a couple of timesorganized by [name of center] 7= We did this several times

Non-Redundancy1a. Please write the initials of the five (not applicable) (not

most important peoplenot employed applicable)by your companythat you rely on foradvice about managing your business

1b. Now, using the table providedindicate if these people know eachother. If so, circle ‘Y’ for yes.

Continued

efficiency, reuse of scrap materials during pro-duction, and reduced releases of hazardous ortoxic materials (U.S. Environmental ProtectionAgency, 1990). Product and process changes aretwo common methods used to reduce or eliminate

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

the creation of pollutants or wastes at the source.Product changes include altering the compositionor use of intermediate or end products. Processchanges are manufacturing modifications thataffect the amount of waste generated as a result

Bridging Ties 1145

Table 1. Continued

Measurement items Scales Internalconsistencyreliability

(a)

Infrequency of Interaction1. Please tell us approximately how (not applicable) (not applicable)

many conversations per month (onaverage) you have with each of youradvisors about your business

Geographic Dispersion1. What is the travel time by car to (not applicable) (not applicable)

each advisor’s office?

Firm Size*1. Roughly how manyfull-time (not applicable) (not applicable)

equivalentemployees worked for youin fiscal year 1995?

2. Roughly how manytemporary andseasonalemployees worked for youin fiscal year 1995?

*Firm size = sum of items 1 and 2

Organization Age1. How long has your company been in (not applicable) (not applicable)

business?

of the way a product is produced. Accordingly,we operationalized pollution prevention with threeitems measuring various product and processchanges (Table 1, Pollution Preventioninstrument). These items were adapted from anational survey of pollution prevention conductedin one of the job shop industries selected for thisstudy (Cushnie, 1994).

Competitive scanning

Given the need to monitor competitors’ actionsclosely in an environment of low switching costsand lack of customer loyalty, competitive scan-ning is critical to firm competitiveness in thesemetalworking industries. Central to the notion ofsustainable competitive advantage is the idea thatfirms strive to insulate themselves from the uncer-tainties of competitive interaction by establishinga unique position relative to rivals (Grant, 1995).Consequently, strategy formulation implicitlyentails scanning the external environment forinformation about competitors to identify threatsand opportunities (Aguilar, 1967; Andrews,

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

1971). The concept of competitive scanning wasoperationalized in this study using three itemsthat tap into the practices a firm uses to gatherinformation about rivals (Table 1, CompetitiveScanning instrument). These items are based onthe measurement instrument created and validatedby Miller (1987).

Quality management

Quality management has increasingly become aprerequisite for doing business in metalworkingindustries that are part of supply chains support-ing the automobile and other global industries.Although total quality management (TQM) meansdifferent things to different people, the elimi-nation of production defects through continuousimprovement of all processes is a common featureof most definitions (Hackman and Wageman,1995). A core principle guiding the managementof quality improvement is the analysis of varia-bility in processes and outcomes through system-atic data collection and statistical analysis(Deming, 1986; Juran, 1974).

1146 B. McEvily and A. Zaheer

In this study, TQM was measured with threeitems capturing the use of statistical process con-trol charts to provide operators with feedback.The feedback allowed operators to base theiractions on the variability of the manufacturingprocess (Table 1, Quality Managementinstrument). These items are based on themeasurement instrument developed by Flynn,Sakakibara, and Schroeder (1995: 1327).

Participation in regional institutions

This construct indicated the extent to which afirm used the services available from a regionalindustrial extension center. Analogous to theacquisition of competitive capabilities, partici-pation in a regional institution is not a discreteevent, but rather represents an ongoing oppor-tunity for a small manufacturer to access infor-mation and ideas.

A 7-point scale similar to the acquisition scalewas developed to measure participation and capturethree primary decision-making stages: knowledge–awareness, evaluation–choice, and utilization. Par-ticipation was measured with a three-item scalecapturing the use of services available (Table 1).

Nonredundancy

To operationalize nonredundancy, this study usedan ego-centered network measure based on aninstrument designed and developed specificallyfor use in the small firm context (Aldrich, Rosen,and Woodward, 1986). This instrument asksrespondents (ego) to identify the five mostimportant external sources of advice (alters)relied upon and to report the extent to whichthese five6 sources know each other (Table 1).Using this matrix, a nonredundancy score wascomputed as follows:

Nonredundancy = (Potential Ties− ActualTies)/Number of Advisors

where,Potential Ties = the maximum number

of ties that could existamong advisors (0 to 10),

6 The data show that there is not a bias toward listing fiveadvisors. In fact, 50% of the firms report fewer than 5advisors, with the mean number of advisors being 3.5.

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

or n(n−1)/2; where, n isthe total number ofadvisors listed

Actual Ties = the number of ties thatdo exist among advisors(0 to 10)

Number of = the total number ofAdvisors advisors listed (0 to 5).

This equation defines nonredundancy as a ratioof nonredundant ties per advisor.7 The values forthis variable ranged from zero to two, with lownumbers indicating low nonredundancy and highnumbers reflecting high nonredundancy. In thecase where all advisors knew each other, nonre-dundancy equals zero (i.e., there was no nonre-dundancy, but rather complete redundancy). Incontrast, when no advisors knew each other non-redundancy reached its maximum of two,reflecting a network of ties that was entirelynonoverlapping. In other words, a low percentageof advisors that know each other indicates that afirm’s network is rich in bridging ties.

Infrequency of interaction

This variable measures the rarity of interactionwith advisors in terms of the number of conver-sations per month (Table 1). Respondents wereasked to report the average number of conver-sation held per month with each of the advisorslisted. Using this information, an overall averageinfrequency of interaction score for each respon-dent is computed as8

Infrequency of Interaction=1

Îmean (conversations per month)

Geographic dispersion

To operationalize this variable respondents wereasked to indicate how far away each advisor was

7 We normalize our nonredundancy measure by the numberof advisors (or the size of the network) since a larger networkwould permit greater opportunity for nonredundancy amongcontacts. We also evaluated a number of alternative compu-tations of nonredundancy, including network size alone, anddetermined that the present formulation proved to be the mostrobust and consistent with our theoretical conceptualization.8 A square root transformation is used to normalize these data.

Bridging Ties 1147

located in terms of travel time by car (Table 1).These data were used to compute an aggregategeographic dispersion score for each respondentbased on the average distance to an advisor as fol-lows:9

Geographic Dispersion= Îmean (distance)

Control variables

We included as a control firm size, in terms ofthe total number of employees (as standardizedz-scores). Larger firms may be less inclined toparticipate in regional institutions and better posi-tioned to acquire competitive capabilities. Scaleeconomies, associated with spreading the costs ofimplementing capabilities over a larger base ofoperations, are greater in larger firms.

We also included organization age as a controlbecause older organizations tend to be relativelymore inert (Hannan and Freeman, 1977) andtherefore less likely to acquire new capabilitiesand participate in regional institutions that pro-mote change.

Construct validity

A measure of a construct is valid to the extentthat it actually measures what it purports to meas-ure (Carmine and Zeller, 1979). The logic ofconstruct validity suggests that multiple indicatorsof the same theoretical construct should be posi-tively and strongly related. In particular, conver-gent validity refers to ‘the degree to which multi-ple attempts to measure the same concept bydifferent methods are in agreement’ (Campbelland Fiske, 1959; Phillips, 1981: 399). The degreeto which two theoretical constructs differ fromeach other indicates discriminant validity.

The ego-centered network data described earlierrelies on a surveying strategy that asks a focalactor (ego) to identify a set of alters connectedto ego, and then report on the network of tiesamong these alters. The validity of the networkconstructs based on this ego-centered networkdata obviously hinges on ego’s ability to assessaccurately the ties between pairs of alters(Krackhardt, 1996; Marsden, 1993). Despite theimportance of accurate assessments by ego, the

9 A square root transformation is used to normalize these data.

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

validity of ego-centered network measures hasreceived little attention (Wasserman and Faust,1994). For this reason, we evaluated the validityof the bridging ties constructs using data gatheredfrom a second knowledgeable respondent whocould provide an accurate report on the networkdata.

Validity of network constructs

To validate the constructs based on the ego-centered network data gathered, this study sur-veyed the contacts (alters) of 40 randomly se-lected firms (egos), of which 21 agreed to provideaccess to their contacts. All of the 85 contactsidentified by the 21 respondents were telephoned.All contacts agreed to confirm their ties with theother alters listed by ego (i.e., the focal firm).We found that 72% of the time, ego’s assessmentof the tie between a pair of alters correspondswith both alters’ report of the tie between them-selves. Further, 86% of the time, ego’s report ofthe tie between a pair of alters is in agreementwith at least one of the alters. In other words,we can be fairly sure that, on average, thesemeasures are no less than 72% accurate andwell within the range of generally acceptablemeasurement accuracy. Based on this analysis, itis reasonable to conclude that the validity of theego-centered network measures is acceptable.

Discriminant validity

The discriminant validity of two constructs canbe assessed by demonstrating that the correlationbetween a pair of constructs is significantly differ-ent from unity. We test discriminant validity bycomparing an unconstrained with a constrainedstructural equation model. The constrained modelsets the correlation between two constructs equalto one. A significantly lowerx2 value for theunconstrained model supports the discriminant va-lidity criterion. As Table 2 indicates, the threebridging ties constructs, as well as the threecompetitive capabilities constructs, all exhibit sat-isfactory discriminant validity.

ANALYSIS AND RESULTS

To test the hypotheses developed earlier, we usedthe maximum likelihood estimation procedure in

1148 B. McEvily and A. Zaheer

Table 2. Discriminant analysis

Model d.f. x2 Dx2 p-valueof Dx2

Independent (exogenous) Latent VariablesAll fs left free 3 1.98 — —f12 fixed to 1.0 (Non-Redundancy & Infrequency of 4 70.48 68.50 0.000Interaction)f13 fixed to 1.0 (Non-Redundancy & Infrequency of Geo- 4 43.72 41.74 0.000graphic Dispersion)f23 fixed to 1.0 (Infrequency of Interaction & Geo- 4 52.77 50.79 0.000graphic Dispersion)

Dependent (endogenous) Latent VariablesAll cs left free 48 64.80 — —c23 fixed to 1.0 (Pollution Prevention & Quality 49 134.49 69.69 0.000Management)c24 fixed to 1.0 (Pollution Prevention & Competitive 49 136.70 71.90 0.000Scanning)c34 fixed to 1.0 (Quality Management & Competitive 49 130.40 65.60 0.000Scanning)

LISREL 7 (Joreskog and So¨rbom, 1993) to spec-ify a structural equation model. Rather thanassuming that constructs are measured withouterror, LISREL explicitly models the measurementerror in the indicators. This feature was relevantto this study, which relies heavily on psycho-metric measurement instruments. LISREL alsoprovides the capability of simultaneously estimat-ing a system of structural equations. This wasappropriate for the present study also, since wetested a partial-mediation model (Venkatraman,1989) that predicts direct as well as indirect(i.e., through participation in regional institutions)effects of the independent latent variables on theacquisition of competitive capabilities.

The structural equation model estimates a seriesof path coefficients reflecting the relationshipsspecified among the latent variables. The Betacoefficient (b) indicates the structural path amongdependent (endogenous) latent variables, whichin this case is the link between participation inregional institutions and acquisition of competi-tive capabilities. The Gamma coefficient (g) rep-resents the structural paths between the inde-pendent (exogenous) latent variables and thedependent (endogenous) latent variables. The signand statistical significance of the Beta andGamma path coefficients serve as a test of thehypothesized relationships.

The results for the structural equation model

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

include the path coefficients and t-values (shownin parentheses) corresponding to each hypothe-sized relationship, and a series of standard fitindices10 reflecting the degree of overall fitbetween the actual and predicted covariancesamong variables of a model. Thex2 statistic teststhe correspondence between the model and theunderlying data. A nonsignificantx2 value isdesirable and indicates that the model is notsignificantly different from the underlying data.Descriptive statistics and zero-order correlationsamong the constructs are reported in Table 3.

The structural equation model specified therelationship between the three bridging ties con-structs and the three acquisition of competitivecapabilities constructs. The model also estimatesparticipation in regional industrial extension cen-

10 Four fit indices are reported for the model estimated. Thegoodness of fit index (GFI) indicates the relative amount ofvariance and covariance jointly explained by the model. Theadjusted goodness of fit index (AGFI) is similar to GFI, butadjusts for the number of degrees of freedom in the model.The normed fit index (NFI) (Bentler and Bonett, 1980)represents the point at which the model being evaluated fallson a scale from a null model (specifying mutual independenceamong indicators) to a perfect fit. The comparative fit index(CFI) (Bentler, 1990) is the same as the NFI, but correctsfor small sample size by subtracting the degrees of freedomfrom their corresponding chi-square values. Each index rangesfrom zero to 1.00, with values closer to 1.00 indicating agood fit. A commonly accepted rule of thumb is that a fitindex should be greater than 0.90.

Bridging Ties 1149

Table 3. Descriptive statistics and zero-order correlations among constructs

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

1 Pollution Prevention 3.831 1.778 1.000Capabilities

2 Competitive Scanning 3.363 1.551 0.260** 1.000Capabilities

3 Quality Management 4.510 1.501 0.295** 0.323** 1.000Capabilities

4 Participation 2.131 1.405 0.132* 0.098 0.215** 1.0005 Non-Redundancy 0.990 0.629 0.130* 0.135* 0.032−0.121* 1.0006 Infrequency of 0.626 0.335−0.014 −0.115 0.016 −0.0583 −0.071 1.000

Interaction7 Geographic Dispersion 1.095 0.886−0.021 0.126* 0.083 0.007 0.091 0.031 1.0008 Firm Size 0.000 1.000 0.020 0.176** 0.264** 0.032−0.038 −0.064 0.026 1.0009 Organization Age 29.930 22.120 0.066 0.083 0.066−0.43 −0.020 0.004 0.030 0.170**

**p , 0.01, *p , 0.05

ters as a mediator of the relationships betweenbridging ties and capabilities constructs whilecontrolling for firm size and age (Figure 2). Thestructural equation model demonstrated ax2 (94df) value of 101.87 (p= 0.272). All fit indices

Figure 2. Bridging ties structural equation model

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

are above 0.9, indicating a satisfactory fit of themodel with the data. Parameter estimates for themodel are provided in Table 4.

Hypothesis 1. The first hypothesis predicting apositive relationship between nonredundancy and

1150 B. McEvily and A. Zaheer

Table 4. Parameter estimates for structural model(Figure 2)

Parameter Standardized t-valuesolution

b21 0.200 2.252*b31 0.130 1.593b41 0.275 3.153**g11 −0.175 −2.094*g21 0.219 2.486*g31 0.159 1.950†g41 0.068 0.803g12 −0.083 −1.010g22 −0.006 −0.076g32 −0.145 −1.824†g42 0.020 0.240g13 0.034 0.410g23 −0.065 −0.763g33 0.136 1.722†g43 0.058 0.704g14 0.056 0.667g24 −0.057 −0.654g34 0.195 2.401**g44 0.411 4.798**g15 −0.076 −0.905g25 0.143 1.642†g35 0.067 0.826g45 0.053 0.625lx1 0.900 (Fixed Parameter)lx2 0.900 (Fixed Parameter)lx3 0.900 (Fixed Parameter)lx4 0.900 (Fixed Parameter)lx5 0.900 (Fixed Parameter)ly1 0.817 (Fixed Parameter)ly2 0.763 8.536**ly3 0.593 7.665**ly4 0.776 (Fixed Parameter)ly5 0.662 7.355**ly6 0.598 7.000**ly7 0.828 (Fixed Parameter)ly8 0.716 9.961**ly9 0.755 10.320**ly10 0.801 (Fixed Parameter)ly11 0.476 5.490**ly12 0.485 5.565**

**p , 0.01, *p , 0.05, †p, 0.10

competitive capabilities islargely supported. Twoof the three competitive capabilities are signifi-cantly associated with nonredundancy in theadvice network. Specifically, while the linkbetween nonredundancy and pollution preventioncapabilities is positive and statistically significant(g21 = 0.219, t = 2.486, p , 0.05), and therelationship between nonredundancy and competi-tive scanning capabilities is positive and sta-tistically significant at the 0.05 level (g31 = 0.159,

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

t = 1.950, p = 0.05), the link to quality man-agement capabilities is not significant (g41 =0.068, t = 0.803, n.s.).

Hypothesis 2. The hypothesized positiverelationship between infrequency of interactionand competitive capabilities isnot supported.Although infrequency of interaction exhibits anassociation with one of the three competitivecapabilities studied (competitive scanning) that isstatistically significant at the 0.07 level, the signis negative rather than positive as predicted (g32

= −0.145, t= −1.824, p= 0.07). The relationshipbetween infrequency of interaction and pollutionprevention capabilities is not statistically signifi-cant (g22 = −0.006, t = −0.076, n.s.), nor is thelink with quality management capabilities (g42 =0.020, t = 0.240, n.s.).

Hypothesis 3. The predicted positive relation-ship between geographic dispersion and competi-tive capabilities ispartially supported. There isa positive relationship between geographic disper-sion and competitive scanning significant at the0.09 level (g33 = 0.136, t = 1.722, p = 0.09),however the data reveal nonsignificant statisticalrelationships for both the geographic dispersionto pollution prevention link (g23 = −0.065, t =−0.763, n.s.) and the geographic dispersion toquality management relationship (g43 = 0.058, t= 0.704, n.s.).

Hypothesis 4. The fourth hypothesis positivelyrelating participation in regional institutions to theassimilation of competitive capabilities islargelysupportedby the data. Indeed, we observed posi-tive and statistically significant relationships forthe link between participation and pollution pre-vention capabilities (b21 = 0.200, t= 2.252, p,0.05) and for the link from participation to qualitymanagement capabilities (b41 = 0.275, t= 3.153,p , 0.01). However, the relationship betweenparticipation and competitive scanning capabilitiesis not statistically significant (b31 = 0.130, t =1.593, n.s.).

Hypothesis 5. The model providedsupport forthe fifth hypothesis predicting a negative relation-ship between nonredundancy and participation inregional institutions. As predicted, the nonredun-dancy to participation link is both negative andstatistically significant (g11 = −0.175, t= –2.094,p , 0.05).

Hypothesis 6. The prediction that infrequencyof interaction and participation in regional insti-tutions would be negatively related isnot sup-

Bridging Ties 1151

ported, since the relationship is nonsignificant(g12 = −0.083, t = −1.010, n.s.).

Hypothesis 7. The hypothesized relationshippredicting a negative relationship between geo-graphic dispersion and participation isnot sup-ported; the link between these two variables isnonsignificant (g13 = 0.034, t = 0.410, n.s.).

Limitations of the research

Before discussing our findings it is important tomention some limitations. Because of the cross-sectional nature of our research design, we arecautious in inferring the direction of causalityamong the key constructs. Although we havepresented the findings in a manner that impliesa certain causal chain, it is possible that thedirection of causality may be reversed in somecases. For instance, Hypotheses 5–7, predictinga relationship between firms’ networks of bridg-ing ties and participation in regional institutions,could be interpreted as causally reversed. Thiswould suggest, for example, that participationactually lowers nonredundancy, rather than nonre-dundancy lowering participation, as we havehypothesized.

To explore this possibility, we compared theduration of the relationship between a firm andthe regional institution with the average durationof the relationship between a firm and itsadvisors. We found that 96% of the time, therelationships between a firm and its advisors werein place prior to participating in the regionalinstitution. Moreover, on average, firms knewtheir advisors for nine years and two monthsbefore their first participation in the regional insti-tution. We believe these data provide some indi-cation that firms’ network structure precedes partici-pation in regional institutions, as we hypothesize.

A question may also be raised as to the diver-sity of the advisor network. If the firm’s advisorsare predominantly from the same industry as eachother, then it is possible that the informationaccessed by the focal firm is less diverse. Toexplore this possibility we analyzed additionaldata collected that reported each advisor’s indus-try category (e.g., supplier, customer, consultant,banker, competitor, university contact). Thesedata indicated that 46% of the firms’ advisorswere in different respective categories (i.e., notwo advisors belonged to the same category).Further, 36% of the firms had exactly two of

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

their advisors in the same category. Only 18%of the firms reported more than two advisorsfrom the same category. These data are consistentwith the notion that the majority of firms’advisors represent diverse sources of information.

A further challenge we confronted was oper-ationalizing the bridging tie concept. Because welimited ourselves to five advisors and the linksamong them, we may not have captured the fulleffects of nonredundancy in the extended net-work. We justify this approach by pointing outthat our research strategy emphasized a multi-construct conceptualization of bridging ties andwas theoretically derived. Further, we oper-ationalized bridging ties with the specific aim ofcreating a parsimonious measure.

Generally, the reliance on egocentric networkmeasures introduces the potential for bias in theassessment of network structure. One issue arisingwith the use of ego-centered network data is theaccuracy of the reports from focal actors (egos)about the relationships among their contacts(alters) (Krackhardt, 1996; Marsden, 1993). Asdiscussed earlier, we addressed this issue by con-ducting a follow-up study with a sample of altersand find a substantial degree of correspondencebetween egos’ and alters’ reports of relationshipsamong alters.

A second issue associated with ego-centerednetwork data is the inability to capture thebroader network of ties beyond an actor’simmediate contacts. The implication is that anapparent structural hole between two primary con-tacts may be bridged by an indirect connectionto a common third party. While unidentified thirdparty connections are a legitimate concern, wenote that the inability to capture the broadernetwork is a challenge for the large body ofresearch relying on ego-centered network data.

In addition, our choice of small manufacturingfirms as the focus of this study limits the general-izability of the findings to this population. At thesame time, small manufacturing firms constitutea large and critical segment of the economy.

DISCUSSION

This study holds important implications for stra-tegic management and social networks research.For strategy research, the study suggests the needto revisit the implicit assumptions prevalent in

1152 B. McEvily and A. Zaheer

the resource-based view that firms are atomisticand that capabilities are internally generated.Rather, as this research shows, sources of com-petitive capabilities can be embedded externallyin firms’ network resources—their network ofbridging ties and linkages to regional institutions.

For network research, we provide insight intoa central debate over the underlying causal agentleading to the acquisition of unique informationand the discovery of productive opportunities. Inparticular, we shed light on the relative explana-tory power of structural holes versus strength ofweak ties theories.

Contributions to strategy research

Our contribution to strategy research consists ofusing a key set of ideas from the embeddednessliterature to deepen understanding of the funda-mental question of why firms differ (Nelson,1991). Existing strategy research employing theresource-based view of the firm tends to explainfirm heterogeneity and profitability differences asarising primarily from internally generated capa-bilities (Barney, 1991). Moreover, this, and othereconomics-based perspectives explaining firm het-erogeneity, implicitly suggest that firms are auton-omous and atomistic in their pursuit of competi-tive advantage.

Our research challenges both assertions bypointing to the role of network resources, andthe externally embedded nature of capabilitiesacquisition, and highlighting the central role offirms’ ties with other economic and noneconomicactors. Put differently, in contrast to firm-centricapproaches to explaining firm heterogeneity, ourresearch adopts the embeddedness notion thatidiosyncratic patterns of linkages account for dif-ferences in competitive capabilities across firms.Our central premise is that a firm’s networkresources of interfirm ties and institutional link-ages expose it to new ideas, information, andopportunities about competitive capabilities. Ineffect, the distinctiveness in firms’ networkresources, and the information access they pro-vide, have important implications for firms’capacity to compete.

Drawing on the social network literature, wedevelop three complementary explanations for thediscovery of new information and opportunities:network structure (nonredundancy), tie strength(infrequency of interaction), and spatial location

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

of contacts (geographic dispersion). The resultsdemonstrate that the structure of firms’ networksof bridging ties exerts the greatest influence onboth the acquisition of competitive capabilitiesand participation in regional institutions.

In geographical clusters, keeping on top ofdevelopments is achieved by maintaining a net-work of advisors who are nonoverlapping anddisconnected. This finding implies that a firm’sconfiguration of linkages with other actors in thecluster is an important vehicle through whichthe firm’s skills, competencies, and routines arecontinually upgraded, refreshed, and renewed.

The link between networks and capabilitiescontributes a valuable insight to the literature oncompetitive capabilities, which has yet to formu-late a comprehensive theory for thesource ofcompetitive capabilities. In particular, while thestrategic management literature has explained per-formance differences in terms of resources andcapabilities, there is scant theory explaining howfirms identify, obtain, and develop competitivecapabilities. Some previous work in this area hassuggested that capabilities derive from externallinkages such as alliances (Hamel, 1991) andnetworks of information (Zaheer and Zaheer,1997), or learning (Powell, Koput, and Smith-Doerr, 1996). We extend this stream of researchby elaborating upon how firms’ patterns of bridg-ing ties and the nature of their linkages withregional institutionsboth influence the acquisitionof competitive capabilities.

We do not mean to imply that firms’ networkresources entirely explain differences in competi-tive capabilities, which are undoubtedly also theoutcome of other factors, including immobileresources, causal ambiguity, luck, and path depen-dence. However, we do argue that firms’ embed-dedness in networks of social, economic, andprofessional ties is an important, if somewhatneglected, perspective in strategy research.

Consistent with the notion of embeddedness,we suggest that bridging ties are not necessarilyonly connections between a focal firm and itscompetitors. Rather, bridging ties also commonlyexist with symbiotic partners such as customers,suppliers, professional service providers, inves-tors, and firms in related industries, all of whichmaintain a considerable interest in the competitivesuccess of the focal firm. Moreover, such bridgingties with a firm’s main advisors are most likelyto exist with trusted actors that provide reliable

Bridging Ties 1153

and valuable information. The implication fornetwork theory is that, rather than focusing onthe strength or weakness of a tie, research shouldrecognize that new ideas, information and oppor-tunities can be sourced through contacts that aresimultaneously nonredundant and trusted. Forstrategy research, because high-trust relationspenetrate irregularly and in differing degrees,bridging ties offers an important explanation forthe differences among firms’ competitive capa-bilities.

Contributions to network research

Our research contributes to the network literatureby drawing a theoretical link between the notionof embeddedness and firm heterogeneity. Specifi-cally, we theorize and demonstrate that onesource of important firm heterogeneity is the idio-syncratic and unique manner in which firms areembedded in networks. Since firms are eachembedded in highly differentiated ways, theyoccupy unique network positions that link themto different sets of players and thereby presentthem with distinct opportunities and constraints.In this way, our research highlights a significantavenue through which embeddedness influenceseconomic actions.

Our research further contributes to the net-work literature through our conceptualization ofthe bridging ties constructs in terms of nonre-dundancy, infrequency of interaction, and geo-graphic dispersion. By relating each of thebridging ties constructs to competitive capabili-ties and to participation, we test the explanatorypower of the underlying network theories (Burt,1992; Granovetter, 1973). The results indicatethat, consistent with the structural holes thesis,nonredundancy in the firm’s advice networkexplains the acquisition of capabilities and par-ticipation in regional institutions; whileinfrequency of interaction and geographic dis-persion show virtually no effects. The impli-cation of these findings is that the diversity ofinformation sources is best reflected in the lackof overlap among contacts (nonredundancy),rather than in the intensity of interaction withadvisors (infrequency of interaction).

At the same time, our results show that, con-trary to structural hole theory, structural holes (asmeasured by nonredundancy) and weak ties (asmeasured by infrequency of interaction) are in

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

fact unrelated. Although Burt (1992) argues thatweak ties and nonredundancy overlap to a certaindegree, we found no systematic relationshipbetween these concepts as reflected in the nonsig-nificant zero order correlation between nonredun-dancy and infrequency of interaction (Table 3).Given received theory about the relationshipbetween structural holes and weak ties, this resultis surprising. It implies, as Burt also acknowl-edges, that ‘information benefits are expected totravel over all bridges, strong or weak’ (1992:30).

Before concluding that the weakness of a tieis not related to the discovery of new informationand opportunities, we encourage future researchto test other possibilities. For example, despitethe explicit use by Granovetter (1973),infrequency of interaction may not be the bestmeasure of weak ties. Future research may con-sider alternative dimensions of weak ties, such asbandwidth of communication, emotional intensity,or reciprocity.

Alternatively, a study mapping the entire net-work of advisors would allow for distinctions tobe examined between weak ties with reclusesversus weak ties with strategically-situated actors.The lack of an association between nonredun-dancy and infrequency of interaction raisesimportant questions about boundary conditions ofthe strength of weak ties theory. For example,it would be useful to understand under whatcircumstances tie weakness might act as a substi-tute for structural holes.

This research further demonstrates that partici-pation in regional institutions mediates the linkbetween bridging ties and the acquisition of com-petitive capabilities. On the one hand, the findingsindicate that firms with an advice network richin bridging ties tend to acquire and develop com-petitive capabilities independent of the regionalinstitution. Conversely, those firms with an advicenetwork limited in bridging ties participate to agreater extent in regional institutions. We interpretthis finding as suggesting that a participatingfirm’s link with a regional institution may act asa ‘surrogate tie’ by serving as a functional substi-tute for its own cluster of bridging ties. Ineffect, a regional institution acts to condensenetworks spatially, thereby allowing a firm’slink through its ‘surrogate tie’ to serve as aconduit for a wide range of information andopportunities.

1154 B. McEvily and A. Zaheer

Directions for future research

While our research has deepened understandingof the links between networks, resources, andcapabilities, it has raised a number of furtherquestions. The finer-grained process throughwhich network structure translates into the acqui-sition of competitive capabilities is an interestingand important area for future research. Mediatingvariables, such as the diversity of informationobtained, could be examined. Another fruitfularea of inquiry is the longitudinal dynamics ofhow firms’ networks evolve and change inresponse to emerging competitive challenges andopportunities. As a firm’s competitive environ-ment changes, to what extent does inertia con-strain a firm’s ability to reconfigure its pattern ofbridging ties?

Further, our understanding of the processes ofcooperative competition in geographical clusterscould greatly benefit from a more detailed analy-sis of the mix of cooperation and competition innetworks. The balance between interfirmcooperation and competition, while a popular idea(Hamel, 1991), warrants greater research atten-tion, particularly in the network context.

Yet another interesting extension of thisresearch would be to investigate the outcomes andconsequences of the two-way flow of informationamong a focal firm and its contacts, and theexchange dynamics that may accompany it.

Concluding remarks

Despite the centrality of competitive capabilitiesto the field of strategic management, research hasdevoted far less attention to exploring the sourcesof capabilities than to performance-related out-comes. Consequently, we know relatively littleabout how firms acquire and develop capabilitiesand what explains differences in the level ofcapabilities among firms. To this end, ourresearch contributes to a better understanding ofthe role of firms’ network resources of bridgingties and participation in regional institutions inthe development of competitive capabilities. Wethus take an initial step in the direction of embed-ding the economic action of capabilities acqui-sition in the broader context of firm networks ingeographical clusters.

The findings of this study demonstrate that firmdifferences exist, in terms of network structure

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

and the degree of participation in regional insti-tutions, and that these differences do matter forunderstanding firms’ acquisition of competitivecapabilities.

ACKNOWLEDGEMENTS

This research was funded in part by the EwingMarion Kauffman Foundation’s Center forEntrepreneurial Leadership, the Great Lakes Pro-tection Fund, and the University of Minnesota’sGrants-in-Aid program. We gratefully acknowl-edge the support of these organizations as wellas the thoughtful suggestions of Phil Bromiley,James Gillespie, Alfred Marcus, Peter Robertsand the participants of The European ScienceFoundation Conference ‘Inter-Firm Networks:Outcomes and Policy Implications,’ held in Mod-ena, Italy, September 1996. We also recognizethe valuable input of two anonymous reviewers.Sharon Hansen and Emily Fritzke providedadmirable administrative support. The contents ofthis paper are solely the responsibility of theauthors.

REFERENCES

Aguilar, F. (1967).Scanning the Business Environment.Macmillan, New York.

Aldrich, H., B. Rosen and W. Woodward (1986).‘Social behavior and entrepreneurial networks, sum-mary’, Frontiers of Entrepreneurship Research.Bab-son College, Wellesley, MA, pp. 239–240.

Aldrich, H. and C. Zimmer (1986). ‘Entrepreneurshipthrough social networks’. In D. L. Sexton and R.W. Smilor (eds.), The Art and Science ofEntrepreneurship.Ballinger, Cambridge, MA, pp.3–23.

Amit, R. and P. J. Schoemaker (1993). ‘Strategic assetsand organizational rent’,Strategic ManagementJournal, 14(1), pp. 33–46.

Andrews, K. (1971).The Concept of Strategy. DowJones-Irwin, Homewood, IL.

Barney, J. (1986). ‘Strategic factor markets: Expec-tations, luck, and business strategy’,ManagementScience,32, pp. 1231–1241.

Barney, J. (1991). ‘Firm resources and sustained com-petitive advantage’,Journal of Management,17, pp.99–120.

Baum, J. A. C. and C. Oliver (1992), ‘Institutionalembeddedness and the dynamics of organizationalpopulations’,American Sociological Review,57, pp.540–559.

Bentler, P. M. (1990), ‘Comparative fit indexes in

Bridging Ties 1155

structural models’,Psychological Bulletin,107, pp.238–246.

Bentler, P. M. and D. G. Bonett, (1980), ‘Significancetests and goodness of fit in the analysis of covariancestructures’,Psychological Bulletin,88, pp. 588–606.

Best, M. (1990).The New Competition: Institutions ofIndustrial Restructuring.Harvard University Press,Cambridge, MA.

Burt, R. S. (1992).Structural Holes: The Social Struc-ture of Competition.Harvard University Press, Cam-bridge, MA.

Campbell, D. T. and D. W. Fiske (1959). ‘Convergentand discriminant validation by the multitrait-multimethod matrix’,Psychological Bulletin, 56, pp.81–105.

Carmine, E. G. and R. A. Zeller (1979).Reliabilityand Validity Assessment.Sage, Newbury Park, CA.

Clark, K. and T. Fujimoto (1991).Product Develop-ment in the World Automobile Industry. HarvardBusiness School Press, Boston, MA.

Cushnie, G. (1994).Pollution Prevention and ControlTechnology for Plating Operations.National Centerfor Manufacturing Sciences, Ann Arbor, MI.

David, P. (1985) ‘Clio and the economics of QWER-TY’, American Economic Review,75, pp. 191–211.

Deming, W. E. (1986).Out of the Crisis.MIT Centerfor Advanced Engineering Study, Cambridge, MA.

Dilman, D. A. (1978). Mail and Telephone Surveys:The Total Design Method.Wiley, New York.

Flynn, B. B., S. Sakakibara and R. G. Schroeder(1995). ‘Relationship between JIT and TQM: Prac-tices and performance’,Academy of ManagementJournal, 38(5), pp. 1325–1360.

Friedkin, N. (1980). ‘A test of structural features ofGranovetter’s strength of weak ties theory’,SocialNetworks,2, pp. 411–422.

Galaskiewicz, J. (1985).Social Organization of an UrbanGrants Economy.Academic Press, Orlando, FL.

Galaskiewicz, J. and A. Zaheer (1999). ‘Networks ofcompetitive advantage’. In S. Andrews and D.Knoke (eds.),Research in the Sociology of Organi-zations. JAI Press, Stamford, CT, pp. 237–261.

Granovetter, M. (1973). ‘The strength of weak ties’,American Journal of Sociology,78(6), pp. 1360–1380.

Granovetter, M. (1985). ‘Economic action and socialstructure: The problem of embeddedness’,AmericanJournal of Sociology,91(3), pp. 481–510.

Grant, R. M. (1995).Contemporary Strategy Analysis:Concepts, Techniques, Applications.Basil Blackwell,Cambridge, MA.

Gulati, R. (1995), ‘Does familiarity breed trust? Theimplications of repeated ties for contractual choicein alliances’,Academy of Management Journal,38,pp. 85–112.

Gulati, R. (1999). ‘Network location and learning: Theinfluence of network resources and firm capabilitieson alliance formation’,Strategic Management Jour-nal, 20(5), pp. 397–420.

Hackman, J. R. and R. Wageman (1995). ‘Total qualitymanagement: Empirical, conceptual, and practicalissues’, Administrative Science Quarterly,40, pp.309–342.

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

Hamel, G. (1991). ‘Competition for competence andinter-partner learning within international strategicalliances’, Strategic Management Journal,WinterSpecial Issue,12, pp. 83–103.

Hannan, M. T. and J. Freeman (1977). ‘The populationecology of organizations’,American Journal of Soci-ology, pp. 929–964.

Henderson, R. and I. Cockburn (1994). ‘Measuringcompetence? Exploring firm effects in pharmaceu-tical research’, Strategic Management Journal,Winter Special Issue,15, pp. 63–84.

John, G. and B. A. Weitz (1988). ‘Forward integrationinto distribution: An empirical test of transactioncost analysis’, Journal of Law, Economics, andOrganization,4, pp. 337–355.

Joreskog, K. G. and D. So¨rbom (1993).LISREL 7: AGuide to the Program and Applications(2ndedition). SPSS, Chicago, IL.

Juran, J. M. (1974).The Quality Control Handbook(3rd edition). McGraw-Hill, New York.

Kogut, B. (1993). ‘Learning, or the importance of beinginert: Country imprinting and international compe-tition’. In S. Goshal and D. E. Westney (eds.),Organi-zational Theory and the Multinational Corporation.St. Martin’s Press, New York, pp. 136–154.

Krackhardt, D. (1996). ‘Comment on Burt and Knez’sthird party effects on trust’,Rationality and Society,8(1), pp. 111–120.

Krugman, P. (1991).Geography and Trade.MIT Press,Leuven University Press, Leuven and Cambridge, MA.

Lazerson, M. H. (1988). ‘Organizational growth ofsmall firms: An outcome of markets and hier-archies?’,American Sociological Review,53 (June),pp. 330–342.

Lieberman, M. B., L. J. Lau and M. D. Williams(1990). ‘Firm-level productivity and managementinfluence: A comparison of U.S. and Japanese auto-mobile producers’,Management Science,36, pp.1193–1215.

Lippman, S. and R. Rumelt (1982). ‘Uncertain imita-bility: An analysis of interfirm differences inefficiency under competition’,Bell Journal of Eco-nomics,131, pp. 418–438.

Maarten de Vet, J. and A. J. Scott (1992). ‘The South-ern California medical device industry: Innovation,new firm formation, and location’,Research Policy,21, pp. 145–161.

Marsden, P. V. (1993). ‘The reliability of networkdensity and composition measures’,Social Networks,15, pp. 399–421.

Marshall, A. (1920).Principles of Economics: An Intro-ductory Volume.Macmillan, London.

Meyer, A. D. and J. B. Goes (1988). ‘Organizationalassimilation of innovations: A multilevel contextualanalysis’, Academy of Management Journal,4, pp.897–923.

Miller, D. (1987). ‘Strategy making and structure:Analysis and implications for performance’,Acad-emy of Management Journal,30(1), pp. 7–32.

Nelson, R. R. (1991). ‘Why do firms differ and howdoes it matter?’,Strategic Management Journal,Winter Special Issue,12, pp. 61–74.

1156 B. McEvily and A. Zaheer

Nohria, N. (1992). ‘Information and search in thecreation of new business ventures: The case of the128 Venture Group’. In N. Nohria and R. G. Eccles(eds.), Networks and Organizations: Structure,Form, and Action.Harvard Business School Press,Boston, MA, pp. 240–261.

Nunnally, J. (1978).Psychometric Theory. McGrawHill, New York.

Oliver, C. (1996). ‘The institutional embeddedness ofeconomic activity’. In J. A. C. Baum and J. E.Dutton (eds.),Advances in Strategic Management.JAI Press, Greenwich, CT, pp. 163–186.

Phillips, L. W. (1981), ‘Assessing measurement errorin key informant reports: A methodological note onorganizational analysis in marketing’,Journal ofMarketing Research,18, 395–415.

Piore, M. J. and C. F. Sabel (1984).The SecondIndustrial Divide: Possibilities for Prosperity.BasicBooks, New York.

Porter, M. E. (1990).The Competitive Advantage ofNations.Free Press, New York.

Pouder, R. and C. H. St. John (1996). ‘Hot spotsand blind spots: Geographical clusters of firms andinnovation’,Academy of Management Review,21(4),pp. 1192–1225.

Powell, W. W., K. W. Koput and L. Smith-Doerr(1996). ‘Interorganizational collaboration and thelocus of innovation: Networks of learning in biotech-nology’, Administrative Science Quarterly,41, pp.116–145.

Pyke, F. G. and W. Sengenberger (1990).IndustrialDistricts and Interfirm Cooperation in Italy.Inter-national Institute for Labour Studies, Geneva.

Rogers, E. M. (1995).Diffusion of Innovations. FreePress, New York.

Romo, F. P. and M. Schwartz (1995). ‘The structuralembeddedness of business decisions to migrate’,American Sociological Review,60(6), pp. 874–907.

Sabel, C. F. (1993). ‘Studied trust: Building new formsof cooperation in a volatile economy’,HumanRelations,46(9), pp. 1133–1170.

Saxenian, A. (1994).Regional Advantage: Culture and

Copyright 1999 John Wiley & Sons, Ltd. Strat. Mgmt. J.,20: 1133–1156 (1999)

Competition in Silicon Valley and Route 128.Har-vard University Press, Cambridge, MA.

Scott, A. J. (1992). ‘The Roepke lecture in economicgeography: The collective order of flexible pro-duction agglomerations: Lessons for local economicdevelopment policy and strategic choice’,EconomicGeography,68, pp. 219–223.

Siegel, S. (1956).Nonparametric Statistics for theBehavioral Sciences.McGraw-Hill, Tokyo.

Singletton Jr., R., B. C. Straits, M. M. Straits and R. J.McAllister (1988). Approaches to Social Research.Oxford University Press, New York.

Stinchcombe, A. L. (1965). ‘Social structure andorganizations’. In J. G. March (ed.),Handbook ofOrganizations. Rand McNally, Chicago, IL, pp.142–193.

Suchman, M. C. (1994). ‘On advice of counsel: Lawfirms and venture capital funds as information inter-mediaries in the structuration of Silicon Valley’.Doctoral Dissertation, Stanford University.

Teece, D. J., G. Pisano and A. Shuen. (1997).‘Dynamic capabilities and strategic management’,Strategic Management Journal,18(7), pp. 509–533.

U.S. Environmental Protection Agency, Office ofResearch and Development (1990).Guide to Pol-lution Prevention. The Fabricated Metal ProductsIndustry. U.S. Government Printing Office, Wash-ington, DC.

Uzzi, B. (1997). ‘Social structure and competition ininterfirm networks: The paradox of embeddedness’,Administrative Science Quarterly,42, pp. 35–67.

Venkatraman, N. (1989). ‘The concept of fit in strategyresearch: Toward verbal and statistical correspon-dence’,Academy of Management Review,14(3), pp.423–444.

Wasserman, S. and K. Faust (1994).Social NetworkAnalysis: Methods and Applications.CambridgeUniversity Press, Cambridge, UK.

Zaheer, A. and S. Zaheer (1997). ‘Catching the wave:Alertness, responsiveness, and market influence inglobal electronic networks’,Management Science,43, pp. 1493–1509.