Tilburg University How Organizational Field Networks shape ...

Tilburg University

How Organizational Field Networks shape Interorganizational Tie-formation Rates

Kenis, P.N.; Knoke, D.

Published in:Academy of Management Review

Publication date:2002

Link to publication in Tilburg University Research Portal

Citation for published version (APA):Kenis, P. N., & Knoke, D. (2002). How Organizational Field Networks shape Interorganizational Tie-formationRates. Academy of Management Review, 27(2), 275-293.

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 06. Oct. 2021

https://research.tilburguniversity.edu/en/publications/9d982ea0-4293-4bb9-9652-487dcf1a3b9d

f Academy o/ Monagement fleview2002. Vol. 27, No. 2. 275-293,

HOW ORGANIZATIONAL FIELD NETWORKSSHAPE INTERORGANIZATIONAL

TIE-FORMATION RATES

PATRICK KENISTilburg University

DAVID KNOKEUniversity of Minnesota

We investigate the impact oi communication in field-level networks on rates offormation ol interorganizational collaborative ties, such as strategic alliances andjoint ventures. After developing the concept of an organizational field network ("field-net"), we derive a set of testable propositions and corollaries that relate field-netproperties, such as density, reciprocity, centralization, multiplexity, and hierarchy, tosubsequent nonlinear changes in interorganizational tie-formation rates. We con-clude by discussing aspects of empirical research ior testing the empirical validity ofthese propositions.

In attempting to explain organizational behav-ior, many analysts recently have paid greaterattention to resource exchange networks and col-laborative ventures in which organizations areembedded. However, although networks are fre-quently mentioned in both academic circles andthe business press as influencing organizationalbehavior, organizational theory still lacks de-tailed explanations of how and why networks ofinterorganizational relationships influence organ-izational characteristics and actions, such as theirstructural features, behavioral processes, and lifechances. Our objective is to contribute to a betterunderstanding of these complex relationships.

Three primary problems motivate our efforts;(1) the general absence of explicit network ele-ments in most neoinstitutional theory, espe-cially from analyses of organizational fields(e.g., Powell & DiMaggio, 1991; Scott, 2001); (2) anapparent shift in the locus of many organization-al actions from internal to external origins, par-

We thank P. Devareaux Jennings and the anonymousAMR reviewers for their many useful suggestions for improv-ing the quality of this article. David Knoke's work on thispaper was supported by a grant from the National ScienceFoundation (SBR-9507914) and a Scholar of the Collegeaward from the University of Minnesota College oi LiberalArts, Earlier versions of this paper were presented at theEGOS Colloquium in Maastricht. Netherlands. July 1998, andthe American Sociological Association meetings in Chicago,Illinois, August 1999. We thank the audiences and commen-tators for their suggestions.

ticularly in interorganizational relationships(Auster, 1994; Burns & Wholey, 1993; Nohria &Eccles, 1992); and (3) unresolved general ques-tions about causal relationships among struc-ture and action in diverse social systems {Emir-bayer & Goodwin, 1994; Pescosolido, 1992). Toadvance our understanding of these problems,we develop propositions that link changes overtime in properties of interorganizational fieldnetworks ("field-nets") focused around commu-nication to changes in rates of formation of col-laborative ties among organizations. We hopethat explicating the details of these macrolevelprocesses will ultimately contribute to greatertheoretical and empirical knowledge about or-ganizational change.

We define an organizational field-net as theconfiguration of interorganizational relationsamong all the organizations that are membersof an organizational field. Thus, a field-net con-sists of a particular pattern of both present andabsent links among the entire set of organiza-tional dyads occurring in a specified organiza-tional field. Our field-net concept connects anexplicit network component to the antecedentconcept of an organizational field, whichDiMaggio and Powell define as ". . . those or-ganizations that, in the aggregate, constitute arecognized area of institutional life: key suppli-ers, producers, regulatory agencies, and otherorganizations that produce similar services orproducts" (1983: 148). The organizational field

275

276 Academy of Management Review April

concept itself differs from the more familiar eco-nomic term industry, which, in its most narrowconstrual, refers to a set of equivalent firms thatall produce the same product or service (Alter SfHage, 1993: 44-68). The pop music sector is anorganizational field, consisting of bands, talentagencies, recording studios, radio stations, pub-lishers and distributors, concert halls, and tourpromoters. Other well-known fields include col-legiate athletics, fine arts, commercial banking,medicine, national defense, and internationaltourism.

The organizational field concept is insuffi-ciently attuned to the interorganizational rela-tions among member organizations. That is, inspecifying a particular field, an analyst onlyidentifies the set of organizational actors that heor she believes may be relevant to an empiricalinvestigation. In contrast, the organizationalfield-net concept explicitly focuses analyticattention on the dyadic relations—or ties—between every pair of organizations in a field.Field-net properties at the meso and macro levelare built up from dyadic relations among fieldmembers. Combinations of present and absentdyadic relations aggregate into various networksubstructures—for example, the occurrence ofsuch components as cliques, groups, positions,action sets, and structural holes—^as well as intostructural attributes of the entire field, such asdensity, connectivity, and centralization {Was-serman & Faust, 1994). With our propositions weseek to relate variations in organizational field-net properties to the member organizations' ac-tions.

No single type of relationship constitutes"the" network of an organizational field. Multi-ple types of ties may be relevant in constructingan explanation of structure and action. Whichtypes of networks an analyst should take intoaccount and which ones he or she can safelyignore ultimately depend on the substantive is-sues driving a specific empirical inquiry. Al-though the variety of interorganizational rela-tions is potentially inexhaustible, we contendthat most of these diverse substantive contentsmay be classified under five broad substantiveheadings: information transmissions, resourceexchanges, power relations, boundary penetra-tions, and sentimental attachments (see Knoke,2001: 65, for formal definitions of these terms).

Although we stress that multiple network tiesare often essential for understanding diverse

behaviors, in the formulation of propositionswe give a privileged analytical position tointerorganizational information transmission orcommunicafion. The primary reason is that com-municating preferences, intentions, values, nor-mative expectations, and other varieties of data isa necessary prelude to all other forms of inter-organizational interaction. A second reason is thatbecause our propositions concentrate on the rateof tie formation as the organizational behaviorsto be explained, communication provides afoundation on which organizational agentsbuild trust relations among partners. Thus, in-formation transmission serves as an indispens-able prerequisite before proceeding towardmore complex and riskier interactions.

Information transmission takes many forms,ranging from such relatively low-cost interac-tions as verbal and written messages to moreintense commitments of time and resources,such as conferences and clearinghouse opera-tions. Too often researchers measure communi-cation only as a generalized exchange pro-cess—for example, by asking informants "Withwhom do you regularly discuss (business/political/scientific) affairs?" (e.g., Knoke, Pappi,Broadbent, & Tsujinaka, 1996). A distortedpicture can emerge from aggregating all re-spondents' reports to reconstruct the image of asocial system's communication network. De-pending on the length of time spanned by themeasurement instructions, perhaps only a fewlatent communication links may be operationalat any moment.

We reconceptualize this static image as a dy-namic sequence of specific communicationevents or "episodes" (Pescosolido, 1992: 1104).That is, an organizational field's communicationnetwork is socially constructed through a seriesof temporally dated information exchanges, oi-ten involving relatively precise and narrow sub-stantive contents. The important thread tyingtogether such dynamic sequences is the the-matic consistency of the messages being ex-changed. For example, negotiations to acquire acorporate spinoff may involve prolonged discus-sions, proposals, and counterproposals amongthe buyers and sellers, banks, unions, govern-ment regulators, and other field members. Bytaking a longitudinal perspective, we can betterexamine the impact of an antecedent communi-cation network on subsequent relations be-tween organizations.

2002 JCenis and Knoke 277

Our central analytic interest in this article isto explain specifically how properties of thecommunication network in an organizationalfield-net influence subsequent rates of collabo-rative ties among the field member organiza-tions. In the next section, drawing key conceptsfrom quantitative network analysis, we derive aset of testable propositions and corollaries,based on four theoretical assumptions, that relatefield-net properties, such as density, reciprocity,centralization, multiplexity, and hierarchy, to sub-sequent nonlinear changes in interorganizationaltie-formation rates. In the concluding section wediscuss the underlying assumptions and consideraspects of empirical research for testing the em-pirical validity of the propositions.

RESEARCH PROPOSITIONS

The domain of our propositions is delineatedby the relationship between the structural at-tributes of organizational field-nets and subse-quent rates of interorganizational tie forma-tion^specifically, the creation of strategicalliances. A central question in any theory con-struction is why one would expect a relationshipbetween the different factors included; in otherwords, what is the underlying logic of the prop-ositions (Whetten, 1989: 491)? Theory develop-ment requires more than just specifying the re-lationship between concepts A and B. Everytheory is based on several underlying assump-tions, which, at best, are explicitly stated or, atworse, only implicitly alluded to. Hence, westate simply here the assumptions on which ourpropositions are based, to provide a theoreticalfoundation for expecting those relationships.Our claim that rates of interorganizational tieformation are related to the structure of thefield-net is based on four theoretical assump-tions: (1) organizations are generally acknowl-edged by analysts to be influenced by the socialcontexts in which they are embedded; (2) organ-izational fields serve as significant environ-ments for their member organizations; (3) therelational properties of organizational fields ex-ert a strong influence on organizational actions:and (4) relational structures per se, and not justthe positional attributes of organizations, arecritical sources of organizational behavior. Afterpresentating the propositions, we return tothese underlying assumptions in the discus-sion section.

To develop concrete propositions relating or-ganizational field-net dimensions to tie-forma-tion rates that have potentially testable impli-cations, we focus on a somewhat narrow rangeof relevant phenomena. In particular, we giveprimacy to interorganizational informationtransmission or communication as the key macro-level network that shapes the more microlevelstrategies and actions chosen by individual or-ganizations. An information network involves ex-changes of substantively important messagesamong the members of an organizational field.The specific contents flowing through these com-munication channels may comprise economic, sci-entific, social, or political data that are relevantboth to performing individual organizationaltasks and to achieving a field's collective objec-tives. Variations across several dimensions of afield's communication network structures consti-tute the propositions' antecedents or independentvariables.

The microlevel behavior that we explicitly at-tempt to explain in the propositions—in effect,the consequent or dependent variable—is therate of interorganizational tie formation. Interor-ganizational ties involve the initiation of collab-orative activities, where two or more organiza-tions pool resources and coordinate theiractions to achieve some joint outcome thatmight otherwise prove either too difficult or pro-hibitively costly to achieve by acting alone.

The tie-formation process goes beyond arm's-length market transactions in requiring that thepartnering organizations work closely togetherto attain a common desired objective. Examplesof interorganizational ties include relationalcontracting (Domberger, 1998), strategic alli-ances (Borys & Jemison, 1989; Dussauge & Gar-rette, 1998), joint ventures (Barkema, Shenkar,Vermeulen, & Bell, 1997; Bergquist, Betwee, &Meuel, 1995), research and development consor-tia (Dimanescu & Botkin, 1986), and governmen-tal lobbying campaigns (Knoke et al., 1996; Lau-mann & Knoke, 1987). The proposition set assertsthat the macrolevel relational properties of aninterorganizational field's communication net-work significantly shape the rates at which in-terorganizational ties form among field mem-bers. The unifying question permeating thepropositions is "How do structural dimensions ofan interorganizational communication networkat Time 1 affect the interactions among member


organizations—specifically, their formation ofties to other organizations—at Time 2?"

We assume that organizational participationin both information exchanges and collabora-tive ties involves explicitly instrumental mo-tives and rational decisions undertaken inten-tionally to obtain an organizational benefit. Bothorganizational capacities and informationalneeds are intertwined in forging and sustainingcommunication linkages. One organization mayinitiate contact with another because the agentsof the first perceive those of the second to pos-sess important information (e.g., about technol-ogies and production skills, market opportuni-ties, socioeconomic connections, politicalintelligence) that might enable the initiator toaccomplish its tasks and achieve its goals moreefficiently. Likewise, the second organizationmight respond to an overture to exchange infor-mation because its agents also believe that link-age offers organizational benefits. Perhaps theinitiator can provide complementary data capa-ble of boosting the second organization's perfor-mance, or the exchange might be prized be-cause it enhances an organization's reputationand status as a good organizational citizen.

Similarly, we presume such mutual-benefitconsiderations underlie any decision by two ormore organizations to enter into a collaborativerelation. Thus, the decisions by potential part-ners to engage in communication exchangesand to form collaborative ties involve explicitcalculations about likely gains and costs fromcreating, maintaining, and terminating interor-ganizational interactions.

From a longitudinal perspective, the changingrate of tie formation requires tracking thenumbers of new interorganizational relationsforming within successive intervals. In a cross-sectional perspective, rates of new collabora-tions could also be assessed relative to the num-bers of existing relations among field members.Both conceptual considerations place importantconstraints on propositional forms. If attention isrestricted only to the creation of new relations("virgin ties") among previously unconnected or-ganizations, many network properties, such asdensity, could soon reach a theoretical maxi-mum value, or arrive at an empirical carryingcapacity, beyond which further system growthbecomes either impossible or unlikely.

Repeated interorganizational ties among al-ready connected organizations initially face

fewer restrictive obstacles. For example, in theglobal information sector, previously partneredfirms frequently announce additional strategicalliances to continue their joint research anddevelopment activities (Knoke, 2001: 139-150).However, even recurring collaborations mayeventually reach a field's saturation point, thusslowing or blocking the further proliferation ofties. A strong implication is that it would be illadvised to use linear propositional forms to cap-ture these shifting temporal dynamics. Toachieve a broader analytic scope—making thepropositions more germane to an organizationalfield's full life span, from birth through maturityand beyond—we intend the propositions to ap-ply to both virgin and repeat ties. In discussingindividual propositions below, we considerwhether corollaries should attend to possibledifferences in the formation rates of virgin andrepeat ties.

We formulated all propositions in simple bi-variate forms that depict nonlinear relation-ships between field-net communication proper-ties and interorganizational tie-formation rates.Nonlinearity appears more plausible for ex-plaining population dynamics, as demonstratedby numerous empirical studies relating organi-zational vital rates to changing density depen-dence (Carroll & Hannan, 2000: 218-219). For ex-ample, an initial low density of communicationearly in the history of a field-net probably im-pedes virgin tie formation, because informationabout the advantages of interorganizational co-operation spreads slowly through tenuous indi-rect connections. But as the density of directcommunication ties increases, formation ratesof both virgin and repeat ties should accelerate,as laggard organizations more quickly seize theremaining opportunities to collaborate. Finally,as communication density approaches a ceilingbeyond which additional ties prove difficult tocreate and sustain, the tie-formation rate shoulddecelerate. The rate of change may cease if thesystem reaches its carrying capacity—an equi-librium between forming and breaking relation-ships.

Hence, despite the absence of empirical sup-port in organization studies, we decided to for-mulate all propositions as general nonlinear ex-pectations about the dynamic relationship offield-net communication density to interorgani-zational tie-formation processes. Of course, sub-sequent empirical studies might conclude that

2002 Kenis and Knoke 279

linear relations account for most of the relevanteffects, but our more encompasing nonlinearpropositional forms avoid placing premature re-strictions on any investigation.

In the following proposition set, we attemptto explain how changes in eight formal aspectsof information transmission networks lead to thesubsequent formation oi interorganizationalrelations among the members of an organiza-tional field. Field-nets vary in such macrolevelproperties as their density, connectivity, central-ization, and the multiplexity of ties composingtheir distinct communication networks. Each di-mension may separately affect the tie-formationpropensities of a field's members.

Although we present the eight propositions ina bivariate ceteris paribus format, the relativemagnitude of each macrolevel characteristic ina multivariate analysis can only be answeredempirically. The propositions may strike somereaders as obvious assertions, because theseformal network dimensions were derived di-rectly from network principles (our primarysource was Wasserman & Faust, 1994}. However,they have never been formally stated for organ-izational theory. Our intention is to make ex-plicit a set of conjectures that seemed latent innetwork perspectives on organizations, therebyencouraging organization researchers to exam-ine their empirical validity.

The propositions should be seen, first, as apreliminary attempt at theory constructiondrawing from institutional and social embed-dedness perspectives and, second, as buildingclosely on extant social network analyses. Net-work analysis is central to these propositions,because in that discipline precise concepts andmeasures applicable to describing the rela-tional structures of organizational field-netshave been formulated. Whether these measureswill ultimately prove the most appropriate onesfor proposition testing can only be demonstratedthrough empirical research. After presentingand discussing each proposition, we concludewith speculations about possible sequences, in-teractions or conditional effects, and feedbackprocesses among network phenomena.

Density

Network density is a macrolevel property, de-fined as the proportion of present dyadic ties toall potential ties. The speed with which informa-

tion may be transmitted among the corporatemembers of a field varies directly with the den-sity of communication ties. A very low-densitycommunication network implies that messagesare likely to propagate only slowly through thefield via lengthy chains of intermediaries, be-cause relatively few alternative routes areavailable to link particular dyads indirectly. Theaverage path length (the minimum number ofindirect steps necessary to connect a dyad) islikely to be longer in low-density networks,meaning that both the time required to transmitmessages and the potential for distorted com-munication are greater than in high-density net-works, whose path lengths are much shorter.Many members of low-density fields may beonly tenuously connected to one another, whilesome organizations may be completely discon-nected and, thus, unable to gain access to infor-mation available elsewhere in the field (seeProposition 4). In contrast, in higher-densitycommunication networks, average path lengthsbetween pairs of organizations are likely to bemuch shorter (including more numerous directties), multiple alternative routes should link therelatively fewer dyads that lack direct ties, andfew or no organizations will likely remain com-pletely out of the field's information loops.

These extreme macrolevel configurations atTime 1 carry quite different implications aboutthe capacity of field members to form both virginand repeat interorganizational relations at Time2. As noted above, in our network conceptualiza-tion we assume that information transmission isa necessary antecedent to constructing otherinterorganizational ties, such as resource ex-changes, power relations, and boundary pene-trations. Thus, an existing macrolevel communi-cation network structure enables a field'sparticipants to learn more quickly from one an-other about new opportunities for interaction,exchange, and collaboration. A high-densitycommunication net rapidly floods the systemwith information from numerous, perhaps re-dundant, sources that offer several alternativechannels for filtering and assessing data qual-ity and reliability. Trust sentiments should beeasier to establish when an organization'sagents can tap other informants capable of ver-ifying the credibility of a potential partner. Per-haps only a small proportion of all extant in-formation linkages may be subsequentlytransformed into more expensive interorganiza-


tional relations. But the more dense the commu-nication network, the greater the chance thatany particular corporation can find other or-ganizations disposed to transact further busi-ness. In contrast, the organizations in fieldscharacterized by low-density information trans-mission networks face greater structural barri-ers to obtaining timely and useful data fromwhich to construct subsequent instrumental andaffective ties.

A dynamic aspect of this process is that as acommunication network's density changes, theother types of relations correspondingly rise orfall. Even if the rate at which information trans-mission promotes other interactions were con-stant at every density level, an increase in com-munication network density would raise thesheer volume of new interorganizational rela-tions. More likely, an increasing density initiallypromotes a subsequent increase in interorgani-zational tie formation. Across the history of anorganizational field, the initial predominance ofvirgin ties yields to repeat collaborative actionsamong field members. In an organizational fieldof fixed size N. establishing additional commu-nication links eventually enlarges the meanego-net size, meaning that a modal organizationgains access to more sources of information,both directly and indirectly. With informationabout potential partners multiplying as thefield's communication web grows denser, oppor-tunities for every organization to make collabo-rative deals expand nonlinearly.

As explained earlier, we expect the effects ofcommunication network density on the rate oftie formation to exhibit a curvilinear patternover the history of an organizational field. Aninitial slow start in constructing a field's com-munication network is succeeded by a rapid in-crease in density and the initiation of virgincollaborative ties among organizations. Subse-quently, as communication density continues torise, both virgin and repeat ties should acceler-ate, with the rising rate of new ties obviouslypreceding the formation of repeat collabora-tions. However, as communication network den-sity approaches a field's carrying capacity, thisasymptote or ceiling slows (and may eventuallyhalt) the rate of formation of ensuing interorgan-izational ties. Presumably, with the rapid dwin-dling of potential new organizational partners,the virgin tie rate levels off sooner than therepeat tie rate diminishes. We expect these

changing effects of field-net density on interor-ganizational collaborative actions to produce acumulative S-shaped growth curve over the fullhistory of an organizational field. Our first prop-osition reflects these nonlinear temporal dy-namics, which we believe will apply to bothvirgin and repeat ties, considered separatelyand jointly.

Proposition 1: As the density of an or-ganizational field's communicationnetwork increases, the subsequentrate of interorganizational tie forma-tion initially accelerates and then de-celerates toward zero.

As a visual aid. Figure 1 schematically repre-sents the hypothesized nonlinear relationshipbetween network density and the rate of interor-ganizational tie formation. When a field-net'sdensity is low, the tie-formation rate is also low.As density increases, the rate accelerates in re-sponse to the increased opportunities for collab-oration. Finally, as the communication networkdensity reaches saturation, the tie-formationrate declines and eventually reaches zero.

Reciprocity

Network density refers to the ratio of actual topotential communication ties, without consider-

FIGURE 1Schematic Representation of Proposition 1:

Relationship Between Network CommunicationDensity and Changing Rate of

Interorganizational Tie Formation

0.00 0.50

Communication network density

LOO


ation of whether reciprocity (mutuality) of infor-mation exchanges occurs between both mem-bers of a dyad. For example, actor A may sendinformation to actor B, but actor B may not re-ciprocate. Two organizational fields with iden-tical densities can differ drastically in how theirtotal volume of directed communication is struc-tured at the dyadic level: at one extreme no tie isreciprocated, while at the other extreme everytie is mutual.

An organizational field that sustains a largerproportion of two-way communication channelsis more likely to generate more subsequent inter-organizational relations than a field whose com-munications rely predominantly on one-way ex-changes. A field whose members are connectedby information exchange channels that are openin both directions enables new information to cir-culate more rapidly and reliably through the sys-tem. Hence, its members enjoy a greater capacityto obtain, assess, and act quickly on opportunitiesfor forging additional interorganizational rela-tions. The contrasting situation—an organization-al field having largely unreciprocated communi-cation exchanges—suffers from insufficient flowsof information that alert members to potentialpartnership opportunities. To the extent that afield's participants recognize such structuralblockages to meeting their needs for more infor-mation, they may deliberately cultivate more re-ciprocal information exchanges over time, whichsubsequently increases rates of new interorgani-zational collaborations.

We suspect that the impact of communicationreciprocity may be more substantial for virgininterorganizational ties than for repeat ties.Fields characterized by relatively fewer recipro-cal communication linkages will suffer fromblocked information flows, retarding the subse-quent development of virgin collaborative tiesamong previously unconnected organizations.But if the volume of mutual communication tiesincreases over time, an initially rising rate ofvirgin tie formation should be followed by slow-down and stabilization as the field reaches itscarrying capacity. In contrast, once a specificdyad forms an initial collaboration, the pairmust establish a two-way communication chan-nel that endures as long as that alliance re-mains active {i.e., two organizations cannot co-ordinate their joint activities without frequentreciprocal information exchanges). Because mu-tual communication characterizes most partici-

pants in an initial collaboration, the degree ofreciprocity in an organizational field should be-come less important for the formation of re-peated interorganizational ties. Thus, we expectthe effect of reciprocal communication to have agreater impact on the rate of virgin tie formationthan on repeat collaborations. We formulate anhypothesis and corollary that reflect the nonlin-ear effects of field reciprocity on rates of tieformation.

Proposition 2: As the reciprocation ofdirected communication linkages in afield-net increases, the subsequentrate of interorganizational tie forma-tion initially accelerates and then de-celerates toward zero.

Corollary 2: The impact of recipro-cated communication linkages on thechanging rate of interorganizationaltie formation is substantially greaterfor virgin than foi repeat ties.

Confirmation of Ties

Communication ties are often measured bythe self-reports of organizational informants,with the attendant problems of subjectivity andimprecision. Whether a mutual or unrecipro-cated exchange occurs depends on whether theresearcher takes a weak or strong stand regard-ing the con/irmafion of ties reported by bothmembers of a dyad. For example, an informantfor organization A reports sending informationto organization B, and the informant for B ac-knowledges receiving such communication fromA. Thus, the A^B channel is confirmed by bothparties. However, suppose that although B re-ports sending information to organization A, thelatter denies receiving such communication. Inthis case the unidirectional B4>A channel is un-confirmed. Whether an analyst decides to iden-tify a communication link between organiza-tions A and B as a reciprocated tie thus dependson whether the presence and absence of all foursending-receiving linkages must be explicitlyacknowledged by both parties in the dyad.

Kathleen Carley and David Krackhardt (1996)have used constructural theory to disentanglecognitive inconsistencies or disagreements po-tentially arising from informants' disjointed per-ceptions of their dyadic relationship. In particu-lar, nonconfirmation should be distinguished

282 Academy ol Management Review April

from nonsymmetric and nonreciprocated ties,such as those considered in Proposition 2:

Non-confirmations reflect a lack of agreement be-tween the two parties about the existence of a tiefrom i to j (or vice versa). Non-symmetries stemfrom the perceiver's inconsistency in his or herbelief that the tie is reciprocated {resulting in anon-symmetric matrix within his or her cognitivemap of the structure). Non-reciprocities (or re-flected non-reciprocities) represent a differencebetween the two parties each sending (or receiv-ing) a relation to (or from) the other (Carley &Krackhardt, 1996: 4).

Because organizational communication de-pends heavily on fallible humans for informa-tion exchanges, a possibility arises that thoseagents will experience frequent sociocognitivefailures either to recognize or to acknowledgecommunication acts. The proposition belowabout the consequences of a field's confirmedcommunication exchanges is based on our as-sumption that ties confirmed by both partieswill exert more powerful effects than will di-rected ties reported by only one party. Our rea-soning applies both to reciprocated and nonre-ciprocated ties, as well as the symmetric andasymmetric ties examined in Proposition 8.

If both members of a dyad agree that a spe-cific communication link exists^whether it bea uni- or bidirectional channel—informationflowing through that channel will have ahigher probability of producing interorganiza-tional collaborative relations, compared to in-formation spreading more tenuously throughunconfirmed links. Therefore, we speculatethat organizational fields composed of higherproportions of confirmed communication tiesare more likely to promote nonlinearly in-creasing interorganizational ties than are net-works with more unconfirmed ties. In the twocorollaries we explicitly assert that the non-linear confirmed tie effect occurs equivalentlyfor reciprocated and nonreciprocated commu-nications. However, we offer no speculationsabout whether these effects differ for virginand repeat ties.

Proposition 3; As the proportion of con-finned communication ties in a field-net increases, the subsequent rate ofinterorganizational tie formation ini-tially accelerates and then deceler-ates toward zero.

Corollary 3a: As (he proportion of con-firmed reciprocated communicationties incieases, rates of interorganiza-tional tie formation increase nonlin-early. "

Corollary 3b: As the proportion of con-firmed nonrecipiocated communicationties increases, rates of in terorganization-al tie formation increase nonlinearly.

Connectivity

Another macrolevel network property derivedfrom elementary graph theory concepts is con-nectivity (Wasserman & Faust, 1994: 92-150). Anetwork's connectivity is a function of the num-ber of actors or relations that must be removedin order to leave some members unconnected.Basic to the connectivity concept is a path be-tween a pair of organizations. Formally, a pathis a sequence of distinct actors linked by di-rected ties, starting from one member of a dyadand ending at the second member. The pathlength between these two organizations is thenumber of indirect ties linking them. A networkis connected if at least one path exists betweeneach of its {N'^-N) dyads; otherwise, the networkis disconnected. If no path exists between adyad, its two members are mutually "unreach-able."

In our discussion about connectivity in an or-ganizational field's communication network, wepresume directional relations—that is, a tie isdirected from one actor in a dyad to the secondactor, but a reciprocated relation or indirect pathmay not occur. (Whether a directed tie is alsoconfirmed or unconfirmed is a separate issue, asdiscussed above). Researchers may be temptedto treat every tie in an information transmissionnetwork as evidence of undirected connections(e.g., by symmetrizing all reported ties). However,assuming that any communication channel per-mits messages to pass in both directions betweenindirectly linked pairs may distort a field-net'sactual structure. Unreciprocated infonnation ex-changes provide evidence of unavailable pathsthat may have important consequences for net-work activity. For example, the directed pathA=>B=>C enables actor A to pass a message toactor C via intermediary B, but the absence of thereverse path A<=B<=:C prevents a reply. In con-

2002 is and Knoke 283

meantrast, the reciprocated ties in paththat A and C are mutually reachable.

Analysts should avoid symmetrizing their net-work data and should retain directed tie infor-mation about the connectivity of communicationpaths between every pair of organizations. Di-rected ties permit classification of dyads andentire field-nets according to four increasinglystrict forms of connectivity: weakly, unilaterally,strongly, and recursively connected networks(see Wasserman & Faust, 1994: 132-133, for for-mal definitions of connectedness at both dyadicand network levels of analysis). Over time, afield-net may strengthen its connectivity as itsmember organizations forge increasingly recip-rocal communication paths.

The importance of connectivity for an organ-izational field's communication network liesin its capacity for more speedy and reliableinformation transmission. In recursively andstrongly connected networks, all pairs of par-ticipants are mutually reachable throughdirect communications or via reciprocated in-direct paths. Hence, information and data cantravel faster with greater accuracy from oneactor to another. But in weakly or unilaterallyconnected networks, or in field-nets that arevulnerable to easy disconnection by uncoup-ling a few crucial ties, some organizations areunable to send or receive information fromothers. And, by the definition of a discon-nected network, no members of one componentcan reach the members of other components.The absence of two-way channels or go-betweens (brokers) spanning subsectors of afield at Time 1 means that some dyads remainunreachable. Thus, an organization may beunable to locate and signal potential newpartners about its interests in forming virgininterorganizational relations at Time 2. In con-trast, fields that are characterized by stronglyand recursively connected communicationnetworks and that are not vulnerable to dis-ruption through the disappearance of a hand-ful of ties should experience robust rates ofinterorganizational collaboration. Organiza-tions should more readily gain low-cost accessto candidates willing to exchange resourcesand form high-risk collaborative relations. Be-cause the creation of an initial interorganiza-tional tie undoubtedly reinforces reciprocalcommunication paths, we expect that connec-

tivity effects will be stronger for virgin alli-ances than for repeat collaborations.

Proposition 4: As the strength of commu-nication connectivity in a tield-net in-creases, the subsequent rate of inter-organizational tie formation initiallyaccelerates and then decelerates to-ward zero.

CoroUary 4: The impact of communi-cation connectivity on the changingrate of interorganizational tie forma-tion is substantially greater for virginthan ior repeat ties.

Centralization

The network literature distinguishes betweencentralization as a macrolevel property and ego-centric concepts of "actor centrality" that char-acterize a specific ego's power relative to othernetwork alters. Analysts conventionally con-sider three basic types of graph-theoretic cen-trality: degree, closeness, and betweenness(Freeman, 1977). (A fourth basic form, informa-tion, is an extension of closeness [see Stephen-son & Zelen, 1989, and Wasserman & Faust, 1994:169-219].) As an analytic class, centrality con-cepts capture aspects of an ego-actor's visibilityor popularity, as indicated by the actor's in-volvement in many direct and indirect relations.Variations among the basic centrality measurestake into account differences in the directional-ity of ties (sending or receiving) and the "quali-ty" of the other actors (in terms of their owncentralities) to which an ego is connected.

Briefly, degree simply counts the number ofdirect ties to or from an organization; ciosenesstakes into account both direct and indirect links,representing efficiency or independence fromall other network actors; and betweenness cal-culates the extent to which actors fall betweenother pairs on the shortest paths connectingthem (Brass & Burkhardt, 1992: 194-195). Eachconcept and associated measurement reflectstheoretically distinct ideas about important di-mensions of network structures, such as the nec-essary conditions for acquiring systemic powerand avoiding dependence, controlling and ac-cessing alternative information sources, or ob-taining political support and participating incoordinated collective actions. In the absence ofcompelling theoretical arguments favoring one

284 Academy of Management Beview April

centrality dimension above the other two, weapply all three concepts in formulating the prop-ositions below.

To derive corresponding macrolevel central-ization measures, the ego-actor centrality mea-sures can be aggregated, thus revealing the ex-tent to which the information transmission tiesin a field-net tend to concentrate around a sin-gle organization, with the other members sub-stantially more peripheral. For example, themaximally centralized "star" network concen-trates all relations on one central organizationthat communicates directly with the others. Nodirect connections link the N-l noncentral ac-tors. In contrast, a "circle" network is completelydecentralized: each organization communicateswith just two partners, each of which also ex-changes information with another unique actor,thus forming a closed chain with no central or-ganization.

Freeman (1979) proposed a mathematical def-inition of a normed group-level centralizationindex for a network of N actors. It ranges be-tween 0 and 1, with the lowest score occurringwhen all actors have the same centrality valueand higher scores reflecting the tendency of oneactor to predominate in relations with the oth-ers. Thus, network centralization reflects the ex-tent of relational inequality in a network (vari-ation or dispersion among the ego-levelcentralities) and permits comparison of changesover time in one network or differences amongseveral field-nets.

Each oi the three basic types of centralizationat the field-net level may independently affectthe rates of interorganizational tie formation. Ingeneral, we expect an inverse relationship, withincreasing centralized communication networksreducing the rate of collaborative tie formation.However, the different centralization measuresrepresent divergent processes. Higher levels ofdegree centralization indicate a growing con-centration of information exchanges among arelatively smaller number of organizations. Al-though these fewer popular egos gain greateraccess to information held by their numerousalters, the system as a whole is starved for in-formation.

Similarly, higher closeness centralization in afield-net means that fewer organizations have theshorter direct and indirect ties that provide themwith greater reachability and capacity to avoidcontrol by others. But the flip side is that most

organizations experience relatively greater de-pendence on and control by those central actors,which reduces efficiency in finding potential part-ners with whom to cut deals.

Finally, higher betweenness centralizationsignifies a growing control over information ex-changes and ability to manipulate these pat-terns for the central organizations' advantages.Such severely impacted information also hin-ders the less central actors' efforts to developcollaborative relations. Hence, for all threetypes of systemic centralization, we expect in-creasing field-net centralization to reduce ratesof interorganizational collaboration; however,we have no expectations about differential ef-fects on virgin and repeat ties.

Proposition 5a; As the degree central-ization of a field-net increases, thesubsequent late of interorganizationaltie formation decelerates nonlinearlytoward zero.

Proposition 5b: As the closeness cen-tralization of a field-net increases, thesubsequent rate of interorganizationaltie formation decelerates nonlinearlytoward zero.

Proposition 5c: As the betweennesscentralization of a field-net increases,the subsequent rate of interorganiza-tional tie formation decelerates non-linearly toward zero.

Multiplexity

Two or more different types of relationshipsoccurring together is network multiplexity. Mul-tiplex ties may flow in the same direction or inopposite routes. For example, an organizationthat transmits scientific-technical data to an-other organization may also lend personnel tothe same company. Typical dyadic market pur-chases illustrate the other pattern of multiplexexchange relations: goods going to one organi-zation and money returning by the oppositepath. Multistranded relations reinforce the tiesamong a field's members, making them moreresistant to complete dissolution than are ties ina single-stranded network. In an importantsense, multiplexity indicates the strength oi tiesbinding a field's members together. Frequentand intense encounters spanning a diversity of


interactions should expose the actors' mutualdependencies and obligations, leading them toresist lapses in any of their relations. In con-trast, the disappearance oi a sole connectionbetween a pair of organizations leaves no re-maining links to serve as a reminder that theymight consider reestablishing their broken con-nection.

We are specifically concerned with the effectsof communication network multiplexity on theformation of interorganizational collaborations.Up to this point, we have treated the communi-cation network as an undifferentiated type of tie,but here we consider how information transmis-sion may actually encompass various forms andcontents, frequencies, and intensities. For exam-ple, ties may join organizational agents operat-ing at diverse levels within their organizations,ranging from sales and technical personnel toexecutive leaders. The information exchangemechanisms they use may vary from superficialpublic relations announcements to regular com-mittee meetings. The communication contentsmay involve social, political, economic, and/orscientific data exchanges. Our basic point isthat information transmission can encompass amultiplexity of linkages among field members.

An organizational field pervaded by multiplexinformation ties offers more opportunities for itsindividual members to develop subsequent re-lations than a field with only tenuous communi-cation ties. When communication is confined tonarrow channels, organizations may acquireonly limited knowledge about potential partnerscapable of providing solutions to organizationaldilemmas. For example, if only marketing per-sonnel interact about supplier-customer rela-tions, other key agents may fail to becomeaware of the full implications of technical prob-lems confronting a pair. But when engineers,accountants, lawyers, and other specialists si-multaneously engage in multisided conversa-tions over diverse field issues, then chances in-crease that important data could reach the mostappropriate decision makers.

At the policy-making level, the more tightlycomplementary communication ties amongchief executive officers and board directorsweave a field together, the more easily its or-ganizations can socially construct commonali-ties among their problems, interests, and expla-nations. Bringing the more disinterestedperspectives of third parties, such as banks and

government regulatory agencies, into the infor-mation mix may facilitate deals betweenweakly connected organizations. Third-party in-terpretations may serve to check tendencies forunvetted information to stampede apprehensiveagents into risky ventures.

In effect, a field integrated through a multi-plex information network resembles a classic"garbage can" system, whose participants ben-efit from a continually innovative churningamong its components. In contrast, the absenceof a strongly reinforced information network of-fers an impoverished foundation on which toconstruct secure interorganizational relations.

Multiplex ties are particularly important forexplaining the virgin tie formation process. Asmany-voiced conversations among organization-al agents steadily weave multistranded webs ofinformation, potential partners increasinglylearn enough about one another's interests andexpertise to take a calculated chance on a firstalliance. Although an initial collaboration likelycontributes to further broadening of the interor-ganizational communication channels, addi-tional multiplex ties probably yield smaller in-crements to each partner's familiarity with theother. Hence, we expect that multiplexity shouldbe less important for repeat ties than for initialcollaborations.

Proposition 6: As multiplex communi-cation relations in a field-net in-crease, fhe subsequent rate of interor-ganizational tie formation initiallyaccelerates and then decelerates to-ward zero.

Corollary 6: The impact of multiplexrelations on the changing late of inter-organizational tie formation is sub-stantially greater for virgin than forrepeat ties.

Cohesion

A subset of actors "among whom there arerelatively strong, direct, intense, frequent, orpositive ties" makes up a cohesive subgroup{Wasserman & Faust, 1994: 249). A group that iscompletely connected by strong mutual ties is a"clique" (i.e., a maximal complete subgraph). Aclique imposes a very demanding definition ofcohesion: the absence of a single direct tie be-tween any members prevents it from being a

286 Academy of Management Review AprU

clique. Less strict forms (e.g., "n-cliques" and"k-plexes"} relax the clique criterion by requir-ing relatively short indirect paths to reach allgroup members.

At the macro level, subgroup cohesion is cap-tured by the extent to which information trans-mission is concentrated within a subgroup, rel-ative to communication ties between subgroups.Key cohesion indicators characterizing a field-net's structure include the number of subgroups,the number of actors in each subgroup, and theextent of overlap among these subgroups' mem-bers. Unless the entire network comprises a sin-gle clique, two or more cohesive subgroups mayexist. In some networks multiple subgroupsoccur with substantially overlapping member-ships, while in sparser networks less numeroussubgroups share relatively few members incommon.

At the field-net level, we are primarily con-cerned about subgroups emerging from a de-veloping communication network, rather thanexogenously determined subgroups. (In intraor-ganizational analyses, formally designatedsubgroups such as work teams, departments,and functional divisions impose very signifi-cant constraints on informal communicationpatterns.) We also assume that such sub-groups maintain relatively high levels of in-ternal communication linkages, approachingif not actually achieving clique status. (Theconcept of a group whose members are uncon-nected seems vacuous; see the discussion ofbetween-group hierarchy in the next section.)Two dimensions are important for ascertain-ing the impact of cohesion at the organization-al field level: (1) the number of distinct sub-groups and (2} the extent to which communicationrelations link members of those subgroups.

One extreme configuration is a fragmentedfield whose many subgroups have no or veryfew connections to one another. In the absenceof bridges and go-betweens that facilitate infor-mation exchanges among subgroups, organiza-tions cannot easily learn about their mutualinterests, impairing subsequent rates of interor-ganizational collaboration. If communicationsare largely confined within numerous small, co-hesive subgroups, information becomes se-verely impacted. That is, organizations in eachsubgroup lack access to information held byother subgroups. As a result, the field-net suffersfrom the classic predicament of strong-tie rela-

tions: an incapacity to acquire and apply poten-tial useful information squirreled away in inac-cessible locations (Granovetter, 1973).

At the other extreme is a solidary field con-sisting of a few subgroups with large, cohesivememberships that also maintain substantial in-tergroup communication relations. Through fre-quent contacts, subgroups avoid an impactedinformation situation by offering better opportu-nities for members from different subgroups tolearn about possibilities for interorganizationalcollaboration. High rates of intergroup contact,especially when they lead to strong ties such asalliances, typically reduce disruptive conflictsand maintain the permeability of group bound-aries (Nelson, 1989). Stronger outgroup interac-tions also may attenuate such negative conse-quences of within-subgroup intimacy asconformity, radicalism, and hostility. The othertwo logical combinations (many subgroups withhigh intergroup information relations, few sub-groups with low intergroup channels) probablyprovide intermediate information about poten-tial partners for different subgroups.

Although organization studies have shown lit-tle evidence about the evolution of cohesivesubgroups within organizational fields, wespeculate that field-nets most likely changefrom fragmented toward solidary patterns. Oneconsequence would be enhanced interorganiza-tional ties, especially for virgin collaborations.

Proposition 7: As the numbers of cohe-sive subgroups in a field-net decreasewhile intergroup communication rela-tions increase, the subsequent rate ofinterorganizational tie formation ini-tially accelerates and then deceler-ates toward zero.

Coroiiaiy 7: The impact of numbers ofcohesive subgroups with high inter-group communication relations on thechanging rate of interorganizationaltie formation is substantially greaterfor virgin than for repeat ties.

Hierarchy

Hierarchy in a network of directed relationsrequires that, for every dyad, one actor canreach a second actor, but the second cannotreach the first (Krackhardt, 1994: 97). Militarychains-of-command and bureaucratic organiza-


tional charts are classic examples of strict hier-archies in which each position directly issuescommands only to those positions immediately"below" it. Lower-level subordinates cannotsimultaneously exercise authority over theirhigher-level superiors, but their reporting pat-terns are a mirror image of the hierarchicalcommand sequence. Thus, any reciprocated orsymmetrical relation violates the strict orderingin a hierarchical structure. A somewhat morerelaxed hierarchical pattern involves tie transi-tivity. That is, direct asymmetric ties occur be-tween every "higher" position and all the lowerpositions, but no commands flow from lower tohigher positions.

The network structures oi most organizationsare very unlikely to conform strictly to eitherideal hierarchical pattern. For example, in astudy oi five types of networks in thirty-six so-cial service organizations, researchers mea-sured hierarchy as the ratio of symmetric (tie-reciprocated) and asymmetric (unreciprocated)dyads in each organization (Shrader, Lincoln, &Hoffman, 1989). Consistent with classical hierar-chy models, the status-ridden mechanistic or-ganizations' networks exhibited greater asym-metry, but the networks oi more egalitarianorganic organizations contained more symmet-ric dyads.

Organizational fields also seem likely to vary 'in the extent of the hierarchy within their com-munication networks. Many industries are moreheavily regulated by antitrust restrictions on thetypes oi economic information they can ex-change, and trade or proiessional associationsmay exert strong influence over communicationpatterns in some fields. Although most fieldsseem unlikely to develop the hierarchy oi a mil-itary command, greater or lesser degrees oi hi-erarchical pecking orders in information ex-changes can occur when higher-statusorganizations disdain to reply to communica-tions from lower-status organizations. Organiza-tional reputations probably display such verti-cal diiierentiation, given that well-known andwell-regarded celebrities receive attention andacclaim from the less renowned, to whom theydare not reply without degrading their ownprestige (Fombrun, 1996).

To distinguish hierarchical ties from the recip-rocal ties in Proposition 2, we formulate propo-sitions specifically about intergroup ties be-tween the cohesive subgroups discussed in the

preceding section. That is, whereas reciprocityrefers to dyadic information exchanges, hierar-chy applies only to the extent oi symmetry orasymmetry in intergroup ties, without regard tocommunication structures occurring within sub-groups. We speculate that the more closely aiield-net communication structure conforms to ahierarchy among subgroups, the less likely or-ganizations are to participate in interorganiza-tional collaborations. Information hierarchyerects substantial barriers to members of onesubgroup developing sufficient familiarityabout potential partners irom other subgroupsto risk launching new alliances. The impact ofhierarchical constraints is probably muchgreater for virgin than for repeat ties, on thepresumption that well-established subgroupswould be less responsive to collaborative over-tures from their social inferiors.

Proposition 8: As hierarchy increasesamong the subgroups in an organiza-tional field's communication network,the subsequent rate of interorganiza-tional tie formation decelerates non-linearly toward zero.

Corollary 8: The impact of hierarchicalcommunication among subgroups onthe changing rate of interorganiza-tional tie formation is substantiallygreater for virgin than for repeat ties.

RELATIONS AMONG NETWORK DIMENSIONS

We do not specify a causal structure or inter-locking order among the propositions and corol-laries, primarily because we feel that such astep would push iield-net ideas prematurely to-ward more complex processes. Thus, we havesimply stated several plausible bivariate rela-tions about the effects of communication net-work properties on the subsequent nonlinearrates oi interorganizational collaboration. Webelieve these basic propositions together consti-tute a credible foundation upon which later the-orists could construct more elaborate theoriesabout the joint impacts oi field-net properties inshaping a field's interorganizational relations.Such an undertaking would benefit from empir-ical research that tests the basic propositions, inconjunction with logical deductions about inter-relations among various network dimensions.

288 Academy of Management Heview April

Efforts to disentangle the interplay of networkproperties would best proceed from a historicalperspective regarding the development of organ-izational fields from their nativity through in-fancy and adolescence into maturity and ripeold age. The earliest directed communicationsnetworks in an organizational field should typ-ically exhibit low levels of density, reciprocity,confirmation, connectivity, centralization, multi-plexity, cohesion, and hierarchy. Hence, the sub-sequent initial rates of interorganizational col-laborations, primarily consisting of such virginties as R&D alliances and joint ventures, wouldalso likely be low. As a field's information ex-changes increase over time, several dimensionsof communication network properties shouldshift upward, thence accelerating the subse-quent rates of both virgin and repeat interorgan-izational ties. For example, as a field-net ma-tures, the growing density of communicationlinks among its members should also raise thefield's connectivity, owing to the intimate rela-tionship among these two concepts.

Krackhardt (1994) demonstrated for simulatedhierarchical organizations (with "outtree" inter-nal structures) that a closely related concept(connectedness) approached unity at higher lev-els of network density. Whether similarly robustcovariation among these network properties oc-curs in the information exchanges of less hier-archically structured real-world field-nets is anissue needing examination. Unfortunately, or-ganization studies presently lack empirical evi-dence and well-honed theoretical argumentsabout precisely how and when relationshipsamong various network properties would affectone another. We raise three issues to considerfor broadening the future theoretical reach.

First, we have not specified the scope condi-tions of the field-net propositions—that is, thesituations when and where they apply. Instead,we formulated the proposition set in fairly gen-eral terms, without clearly indicating whether itmight be restricted to particular times, places,and types of fields. In other words, it now hasuniversal applicability. Of course, if empiricalresearch demonstrated that the field-net rela-tionships applied only under some conditionsand not in other circumstances (as other analystsoften suggest), then their explanatory powerwould have a more restricted scope. We primarilyhad in mind explanations of processes within for-profit business fields, but analysts of nonprofit

social services and governmental sectors mayalso find these ideas relevant to their researchefforts. We consider scope conditions to ultimatelybe a matter best resolved through empirical inves-tigation than a priori.

Second, the propositions do not include feed-back processes from interorganizational rela-tions that could restructure a field's communica-tion processes—that is, how does the formationof virgin and repeat collaborative ties amongpartners subsequently alter the dimensions ofinformation networks? As we noted at the begin-ning of the proposition section, both organiza-tions' capacities and needs may drive them toseek information exchanges with other fieldmembers. But, as briefly mentioned in our dis-cussion of reciprocity (Proposition 2), when apair of organizations forms an alliance, thoseorganizations most likely engage in increasedreciprocal communication. As a field's membersgain more experience in forming and imple-menting alliances, especially those involvingrepeat ties among previously partnered organi-zations, the field-net undoubtedly increases indensity, reciprocity, and presumably severalother dimensions.

Prominent examples of institutionalizedmechanisms deliberately developed by organi-zational fields to foster interorganizational com-munications around issues of common interestinclude business trade associations, researchconsortia, standards committees, and certifica-tion boards. The feedback arising from such inter-organizational activities to a more elaborate infor-mation exchange network suggests a "virtuouscycle" of reinforcing effects over time. Of course,just as in our nonlinear propositions we indicatethat all good things never proliferate indefinitely,in feedback loops the constraints on a field-net'scarrying capacity should be taken into account.

Third, another intriguing possibility is thattwo or more field-net properties interact in theireffects on tie-formation rates—that is, the effectof one network dimension may depend on levelsor rates of a second dimension. For example,Propositions 5a through 5c state that as commu-nication centralization in a field-net increases,the subsequent rate of tie formation deceleratesnonlinearly. But those effects may depend onconcurrent changes in the field-net density—that is, the rate at which rising centralizationreduces tie formation may be altered as thefield's communication density changes. In that

2002 Jfenis and Knoke 289

case, a more complex proposition might, for ex-ample, state: "At lower network densities, morecentralized communication networks lead toslowly decelerating rates of tie formation, whileat higher network densities, more centralizednetworks result in more rapidly deceleratingrates."

Although our propositions all specify nonlin-ear relations, the possibility of complex interac-tion effects transcends the current sophistica-tion of most network theories. Hence, we believethat developing formal propositions about inter-actions among the network dimensions is alsotheoretically premature and must await a moreextensive accumulation of empirical evidence.One further implication arising from potentialfeedback and interaction effects is that empiri-cal researchers should use dynamic methods ofdata gathering and analysis, as discussed in thenext section.

DISCUSSION

We have asserted that explanations of interor-ganizational action must take into account thetemporal dynamics of an organizational field-net—a configuration of interorganizational tiesamong all members of an organizational field.We consider the field-net concept to be a promis-ing analytic tool, because it combines elementsfrom two well-established theoretic concepts: itgives structure to the somewhat amorphous con-cept of "organizational field," and it provides sub-stantive content to the somewhat abstract conceptof "network."

We have indicated the significance of thefield-net concept for organization studies, argu-ing that interorganizational relations can be ex-pected to influence organizations embedded ina field. We also have attempted to demonstratethe fertility of the field-net concept by formulat-ing several testable propositions about the non-linear relationships among various dimensionsof network structure in organizational fields andsubsequent rates of interorganizational tie for-mation. Of course, our efforts are only the firststeps toward a theory construction process, andseveral further inquiries lie ahead. Here we dis-cuss some theoretical, empirical, and practicalimplications of the proposition set relating net-work properties to interorganizational collabo-rations.

Theoretical Assumptions

At the beginning of this article, we simplyasserted that four theoretical assumptions un-derlie our proposition set. We now providegreater detail and illustrate some contrasts be-tween our structural approach and other macroor-ganizational theories in order to highlight thecontributions of our theory.

The first general theoretical assumption un-derlying our approach—^that organizations areinfluenced by the social contexts within whichthey are embedded—was identified by JeffreyPfeffer (1997) as the social behavior model. Theidea that organizations are influenced by thefield-nets in which they are embedded is basedon the assumption that action is shaped by thesocial environment: "social models of behavioremphasize the context of behavior more gener-ally and networks and social actors' positions inthem and their social relations more specificallyas causal explanations" (Pfeffer, 1997: 55). Al-though the embeddedness perspective providesone fundamental basis for our propositions, it isnot a sufficient foundation. According to somecritics, embeddedness says little about how so-cial ties shape social behavior. Consequently,additional theoretical distinctions and assump-tions were necessary in order for us to arrive atour field-net propositions.

A second assumption is that organizationalfields serve as significant environments for theirmember organizations. An organizational fieldapproach, however, is narrower than most neoin-stitutional conceptions of environment, becauseit is assumed that the immediate ties are morerelevant than broader societal and institutionalpatterns and that organizational actions aremore strongly shaped by concrete communica-tion structures than by diffuse beliefs, norms,and rules. In contrast to neoinstitutionalists, weassume that organizational fields often embodyconcrete structures and localized cultures thatare only loosely coupled to broader national so-cietal contexts. Thus, the field-net propositionsshould be seen as a specific attempt to articu-late a process-oriented approach that treats so-cial structure (field-nets) as both a product ofand a constraint on organizational action. Actorbehavior is not passively determined solely bythe environment; rather, organizations are si-multaneously autonomous subjects and depen-dent objects. Field-nets are produced by human

290 Academy of Management fleview April

actions, but they also subsequently assumestructural properties. That is, they are sociallyconstructed by organizations acting in particu-lar social contexts. However, once developedand deployed, field-nets tend to become reified,thus losing their connections with and depen-dence on the human agents who originally con-structed them. Hence, field-nets appear to theirparticipants as objective structural properties ofan organizational field.

Our third assumption is that the relationalcharacteristics of the organizational field exertprimary influence on organizational action.Whereas in resource dependence theory schol-ars conceptualize environments primarily innonrelational terms, in the field-net perspectivewe explicitly assume that what matters most arenot a field's characteristic attributes but its in-terorganizational structures. Structural analystsassume that structured social relationships aremore powerful sources of theoretical explana-tion than are the individual attributes of systemmembers. We view the field-net approach asclearly falling within a structural analysis per-spective because both start from a basic as-sumption that interorganizational relations aremore important than organizational attributesfor explaining organizational behavior.

The final assumption underlying our struc-tural approach is that the critical sources oforganizational behavior are not the positionalattributes of organizations but, rather, the rela-tional structures of the field-net per se. In ourfield-net approach, we assume that the macro-level configuration of interorganizational tiesamong field members serves as an opportu-nity structure that both constrains and facili-tates its member organizations' potential ac-tions. Our propositions emphasize howaggregate relations within a field-net canerect substantial barriers to collective ac-tion—for example, more centralized networksprovide fewer opportunities for peripheralparticipants to locate potential partners withwhom to forge new collaborative ties.

Research Implications

Our main task has been to specify and justifythe expected relationships between the basicdimensions of communication networks and thesubsequent rates of interorganizational tie for-mation. However, we would be remiss in not

offering some suggestions about research de-signs that could be used to test these proposi-tions and, hence, contribute to theory buildingby uncovering evidence that either strengthensor alters our initial claims. Our arguments referto changing structures and rates that obviouslyrequire some form of longitudinal data collec-tion and analysis. Such projects seem to placemany daunting demands on researchers' pa-tience in identifying, selecting, accessing, mea-suring, and analyzing temporal data.

Studying the structures of interorganizationalnetworks is not an impossible undertaking, asexemplified by research on U.S., German, andJapanese national policy domains (Knoke et al.,1996; Laumann & Knoke, 1987); Midwest urbanyouth services agencies (Shrader et al., 1989);Scottish knitwear producers (Porac, Thomas,Wilson, Paton, & Kanfer, 1995); and New Yorkwomen's better dress manufacturers (Uzzi, 1996,1997). Joseph Galaskiewicz's study of Twin Cit-ies, Minnesota, philanthropy provides a para-digmatic research design for a panel of corpo-rations and nonprofit associations whosecommunication links, grant making, and dona-tive activities were persistently tracked andobserved periodically across two decades(Galaskiewicz, 1985, 1997; Galaskiewicz &Bielefeld, 1998). These successful investigationsof interorganizational relations were conductedby one or two principal investigators and a fewresearch assistants, with relatively modest re-search funds.

Researchers should attempt to observe the en-tire membership of an organization field for aprolonged period as its organizations communi-cate and engage in collaborative activities.Conceivably, data collection might span severalyears, undeniably requiring a substantial timeinvestment by the investigator before he or sheobtains a payoff. Archival records from newsreports and organizational publications mightextend the time frame into the recent past—forexample, by reconstructing interlocking direc-torates and joint memberships in a field's inter-est groups. Archival methods are better for iden-tifying such key collaborative events as jointventures and strategic alliances (Knoke, 2001)than for mapping details of a historical informa-tion exchange network comprising the crucialindependent variable in the propositions. Forthis reason, we suggest that longitudinal for-


ward tracking is preferable to a retrospectivedata collection research design.

Studying a field-net during its early develop-mental stages offers a nnore suitable test of thepropositions than observing a mature fieldwhose stable relationships have become thor-oughly institutionalized. Consequently, re-searchers should begin investigating recentlyemergent organizational fields of the twenty-first century. Such sites would be equivalent tothe organization fields launched in the 1900s forautomobiles, in the 1920s for airlines, in the1970s for biotechnology, and in the 1980s for in-formation technologies. Of course, anticipatingwhich fields-in-formation might emerge as fruit-ful arenas to study is a critical task. Small fieldsrapidly constructed around highly specializedtechnologies might provide the most tractableresearch opportunities. Within nonprofit andpublic sectors, initiation of a new social serviceprogram, such as welfare reform, could provideopportunities for tracking emergent interorgan-izational structures.

Another important question for empirical re-searchers is whether to identify field-nets pri-marily in objective or in subjective terms. Thatis, the criteria for determining membership in afield and selecting specific organizations couldinvolve quantified indicators, such as industryand size, or more cognitive approachs, such asinformant judgments. In previous empiricalanalyses of national policy domains, research-ers combined several behavioral and reputa-tional indicators to identify the core organiza-tions in energy, health, and labor policy fields(Knoke et al., 1996; Laumann & Knoke, 1987). AsScott notes, "In stable and more highly institu-tionalized fields, there is high consensus on thedefinitions as to who the critical players are,what activities and interactions are appropriate,and which organizations are included, marginalto, or outside field boundaries" (1998: 129). Byinference, the newly emergent fields most ap-propriate for testing the propositions would lackconsensus about which governmental, private,and nonprofit organizations were members.Hence, researchers might develop and deploymultiple, overlapping measurements for bound-ing a field prior to data collection.

Empirical examination of the propositionsmight eventually falsify some of our proposi-tional statements. For example, Provan and Mil-ward (1995) found counterintuitive evidence that

centralized social welfare networks were moreeffective than decentralized systems in deliver-ing services. If analyses spanning diverse or-ganizational fields disproved the validity of spe-cific propositions, such outcomes would yield arevised set more consistent with the observedrelations between network structures and interor-ganizational relations.

Another issue with important implications forresearch on social change in general, and net-work analysis in particular, is the question ofontological validity. Briefly, collecting data onexisting social relations at one particular timeruns some risk of incompletely capturing thephenomenon. Because most interorganizationalcommunication exchanges occur intermittentlyrather than continuously, using a very narrowtime frame may result in underreporting of la-tent ties because of their low salience to infor-mants. Measures spanning a longer time inter-val may produce a more comprehensive tally ofrelations but, in turn, might inaccurately repre-sent the actual network structures existing at aspecific moment. That is, measuring an organi-zational field over time and verifying its chang-ing network structures depend crucially on howresearchers operationalize interorganizationalties within and across particular intervals.

This entity-measurement conundrum, with itsechoes of the quantum physics uncertainty prin-ciple, requires empirical researchers to ponderhow best to measure changing social structures.Although we do not offer a definitive solution,we recognize that this difficult issue deservesgreater attention than network analysts haveyet paid it. For now, we reiterate our contentionthat communication-event or "episode" mea-sures, which would identify specific informationexchanges occurring within a dated interval,are preferable to asking informants about theirgeneralized communication links during avague period. However, the ontological validityof our assertion awaits empirical confirmationthrough a comparative investigation of alterna-tive measurement instruments.

Implications for Practitioners

On first consideration, our macrolevel propo-sitions would seem to offer little practical guid-ance for harried administrators trying to copewith the daily contingencies of organizationalmanagement. Some network properties may

292 Academy of Management fieview April

have contradictory implications across differentlevels of analysis. For example, although in-creasingly centralized field-nets reduce the rateof interorganizational collaborations {Proposi-tions 5a through 5c), individual organizationstypically benefit by occupying the more centralpositions in their field. Despite such cross-levelconundrums, leaders who ignore interorganiza-tional relations potentially imperil their organi-zations' performance and even survival.

Although large field-net structures are proba-bly highly resistant to overt manipulations byindividual organizations, top executives shouldcontinually monitor their field for signs of sig-nificant change. Creating and maintaining ef-fective communication links to diverse fieldmembers are vital in gathering useful organiza-tional intelligence about potential opportunitiesand threats. By proactively forming and break-ing information linkages, top executives shouldtry to move their firms and agencies into moreadvantageous positions. For example, theycould reap organizational benefits by facilitat-ing exchanges of reliable information or by bro-kering virgin collaborations of otherwise uncon-nected organizations. As practitioners learnmore about how an organization field's commu-nication network shapes the rates of interorgan-izational collaborations, they may acquire a pro-active interest in developing and sustaining thisvaluable collective asset.

REFERENCES

Alter, C, & Hage, J. 1993, Organizations working together.Newbury Park, CA: Sage,

Auster, E. R. 1994. Macro and strategic perspectives on interor-ganizational linkages: A comparative analysis and re-view wilh suggestions for reorientation. Advances inStrategic Management. ME): 3-40.

Barkema, H. G., Shenkar. O., Vermeulen, F,, & Bell, I. H, I-1997, Working abroad, working with others: How firmslearn to operate international joint ventures. Academyoi Management Journal. 40: 426-442.

Eergquiat, W. H., Betwee, J,, & Meuel, D. 1995. Bui/ding stra-tegic alliances: How to extend your organization's reachthrough partnerships, alliances, and joint ventures. SanFrancisco: Jossey-Bass,

Borys, B,, & Jemison, D. B, 1989. Hybrid arrangements asstrategic alliances: Theoretical issues in organizationalcombinations. Academy of Management Review, 14:234-249.

Brass, D. J,, & Burkhardt, M, E, 1992. Centrality and power inorganizations. In N. Nohria & R. G. Eccles (Eds.), iVef-

works and organizations: Structure, form and action:191-215. Boston: Harvard Business School Press.

Burns, L. R., & Wholey, D. R. 1993, Adoplion and abandon-ment of matrix management programs: Effects oi organ-izational characteristics and interorganizational net-works. Academy of Management Journal, 36: 106-138.

Carley, K, M. & Krackhardt, D, 1996, Cognitive inconsisten-cies and non-symmetric friendship. Social Networks, 18:1-27.

Carroll, G,, & Hannan, M. T, 2000. The demography of corpo-rations and industries. Princeton, NJ: Princeton Univer-sity Press,

DiMaggio, P,, & Powell, W, 1983. The iron cage revisited:Institutional isomorphism and collective rationality inorganizational fields. American Socio/ogica/ Review, 48:147-160.

Dimanescu, D., & Botkin, J. 1986, The new alliance: America'sR and D consortia. Cambridge, MA: Ballinger.

Domberger, S. 1998, The contracting organization: A strategicguide to outsourcing. New York: Oxford University Press,

Dussauge, P., & Garrette, B. 1998. Anticipating the evolutionsand outcomes of strategic alliances between rival lirms,/n(erna(iona/ Studies of Management and Organization,27(4): 104-126.

Emirbayer, M., & Goodwin, J, 1994. Network analysis, culture,and the problem of agency. American /ournai of Sacial-ogy, 99: 1441-1454.

Fombrun, C. ]. 1996, Heputation: Reaiizing value from the cor-porate image. Boston: Harvard Business School Press.

Freeman, L, 1977. A set of measures of centrality based onbetweenness, Sociometry, 40: 35-41.

Freeman, L. 1979. Centrality in social networks: I. Concep-tual clarification. Social Networks. 1: 215-239,

Galaskiewicz, J. 1985. Social organJzafion of an uihan grantseconomy: A study of business philanthropy and non-profit organizations. Orlando, FL: Academic Press,

Galaskiewicz. J, 1997. An urban grants economy revisited: Cor-porate charitable contributions in the Twin Cities, 1979-81, 1987-89. Administrative Science Quarter])', 42: 445-471.

Galaskiewicz, J,, & Bielefeld, W. 1998. Nonprofit organiza-tions in an age of uncertainty. New York: Aldine deGruyter,

Granovetter, M. 1973, The strength oJ weak ties, American/ournai of Sociology, 78: 1360-1380,

Knoke, D, 2001, Changing organizations: Business networksin the new political economy. Boulder, CO: Westview.

Knoke, D,, Pappi, F. U., Broadbent, ].. & Tsujinaka, Y. 1996,Comparing policy networks: Labor politics in the U.S.,Germany, and Japan. New York: Cambridge UniversityPress,

Krackhardt, D. 1994, Graph theoretical dimensions of infor-mal organizations. In K. M. Carley & M. I. Prietula (Eds.),Compufationai organization theory: 89-111, Hillsdale,NJ: Lawrence Erlbaum Associates.

Laumann. E, O.. & Knoke, D, 1987, The organizational state: Aperspective on the sociai organization of national en-


ergy and health policy domains. Madison: University oiWisconsin Press.

Nelson, R, E. 1989. The strength of strong ties: Social net-works and inlergroup conflict in organizations. Acad-emy of Management Journal, 32: 377-401.

Nohria, N., & Eccles, R. G. (Eds.). 1992. Networks and organi-zations: Structure, form and action. Boston: HarvardBusiness School Press.

Pescosolido, B. 1992. Beyond rational choice: The social dy-namics oi how people seek help. American Journal olSociology. 97: 1096-1138.

Pieifer, J. 1997. New directions lor organization theory: Prob-lems and prospects. New York: Oxford University Press.

Porac, J. F.. Thomas, H., Wilson. F., Paton, D., & Kanier, A.1995. Rivalry and the industry model ot Scottish knit-wear producers. Administrative Science Quarterly, 40:203-227.

Powell, W. W.. & DiMaggio, P. J. 1991. Introduction. In W. W.Powell & P. J. DiMaggio (Eds.). Tbe new institutionalismin organizational analysis: 1-38. Chicago: University ofChicago Press.

Provan, K. G., & Milward. H. B. 1995. A preliminary theory ofinterorganizational effectiveness: A comparative study

of four community mental health systems. Administra-tive Science Quarterly. 40: 1-33.

Scott, W. R. 1998. Organizations: Rational, natural, and opensystems (4th ed.). Englewood Clifts, NJ: Prentice-Hall.

Scott, W. R. 2001. /nstifutions and organizations (2nd ed.).Thousand Oaks, CA: Sage.

Shrader. C. B., Lincoln. J. R., & Hoffman, A. N. 1989. Thenetwork structures of organizations: Effects ol task con-tingencies and distributional form. Human Relations. 42:43-66.

Stephenson, K., & Zelen, M. 1989. Rethinking centrality: Meth-ods and examples. Social Wefworics, 11: 1-37.

Uzzi, B. 1996. The sources and consequences of embeddednessfor the economic performance of organizations: The net-work effect. American SocioJogicai fleview, 61: 674-698.

Uzzi, B. 1997. Social structure and competition in interfirmnetworks: The paradox of embeddedness. Adminisfra-tive Science Quarterijr, 42: 35-67.

Wasserman. S., & Faust. K. 1994. Sociai networfe anaJysis:Methods and applications. Cambridge: Cambridge Uni-versity Press.

Whetten. D. A. 1989. What constitutes a theoretical contribu-tion? Academy of Management Review. 14: 490-495.

Patrick Kenis is a professor of policy and organization studies at Tilburg University,the Netherlands. He received his Ph.D. in political and social science from the Euro-pean University Institute in Florence, Italy. His research interests include interorgan-izational networks, policy networks, and network theory.

David Knoke is professor of sociology at the University of Minnesota. He received hisPh.D. from the University of Michigan. His research interests include organizationaltheories, strategic alliances, and social network analysis.

Tilburg University How Organizational Field Networks shape ...

Documents

Transcript of Tilburg University How Organizational Field Networks shape ...