cooperative or this study examines the influence of the...

Cooperative orControlling? The Effectsof CEO-board Relationsand the Content ofInterlocks on theFormation ofJoint Ventures

Ranjay GulatiNorthwestern UniversityJames D. WestphalUniversity of Texas at Austin

© 1999 by Cornell University.0001-B392/99/4402-0473/$1.00.

The order of authorship was determinedalphabelicaily. We thank the associateeditor Don Palmer, three anonymous re-viewers, and The following individuals fortheir helpful comments: Gautam Ahuja,Martin Gargiulo, Mark Mizruchi, Marc-David Seidel, Karen Westphai, and Ed-ward Zajac. We would like to thank LindaJohanson for editorial guidance.

This study examines the influence of the sociai networkof board interlocks on strategic alliance formation. Ourtheoretical framework suggests how board interlock tiesto other firms can increase or decrease the likelihood ofalliance fornnatlon, depending on the content of relation-ships between CEOs (chief executive officers) and out-side directors. Resuits suggest that CEO-board relation-ships characterized by independent board control reducethe likelihood of alliance formation by prompting distrustbetween corporate leaders, while CEO-board cooperationin strategic decision making appears to promote allianceformation by enhancing trust. The findings also showhow the effects of direct interlock ties are amplified fur-ther by third-party network ties.*

The board interlock network has been viewed as an idealarena in which to develop and test the embeddedness per-spective on interorganizational relations. The board of direc-tors is a unique formal mechanism linking top managers oflarge corporations; it provides an opportunity for leaders toexchange information, observe the leadership practices andstyle of their peers, and witness firsthand the consequencesof those practices. Thus, from this perspective, board ties toother firms should have a strong influence over corporatepolicy and strategy decisions. The empirical literature onboard interlocks has extended research on the diffusion ofinnovations by specifying the social networks through whicha variety of policies and practices are spread across firms(e.g,, Mizruchi, 1992; Haunschild, 1993; Palmer, Jennings,and Zhou, 1993; Westphal and Zaiac, 1997).

While interlock research has advanced our understanding ofthe consequences of interlocks for firms, significant con-cerns have also been raised that reflect more genera! con-cerns about the application of network theory to interorgani-zationai relations (Mizruchi, 1996), Several authors haveexpressed concern about the consistency and magnitude ofnetwork effects (Stinchcombe, 1990; Fligstein, 1995). Weakor inconsistent findings may result from two limitations com-mon to most prior studies. First, prior interlock research hasnot adequately specified the content of network ties (Hirsch,1982; Pettigrew. 1992; Mizruchi, 1996: 288). Content hereimplies a specification of the nature of the relationship andbehavioral processes underlying a connection between twoactors. Although recent research in the governance literaturesuggests that relationships between top managers on corpo-rate boards may be characterized by independence and dis-trust in some cases (Westphal, 1999), in the interlock litera-ture all ties are generally treated as equally positiveconnections that facilitate social cohesion and the exchangeof information between firms. This ignores heterogeneitythat may exist among interlocks in the extent to which theychannel information and engender trusting relations betweenboard members.

Another concern with interlock research is its primary focuson the effect of direct ties or relational embeddedness onfirm behavior to the exclusion of more distant network tiesor structural embeddedness (Granovetter, 1992). While thereis ample evidence in network research that both relational

473/Admmistrative Science Quarterly, 44 (1999): 473-506

and structural embeddedness can influence behavior {e.g.,Burt, 1987; Gulati and Gargiulo, 1999), this has had limitedapplication in interlocks research. Studies in the larger net-work literature have shown that indirect network ties be-tween actors can strongly condition the effects of direct tiesbetween them (Gulati, 1995b). Moreover, recent researchalso suggests that indirect network ties can amplify differ-ences in the magnitude of the effect of direct ties, such thatdistrust between actors is exacerbated in the presence ofindirect ties between them (Burt and Knez, 1995). Thus, itmay be possible to uncover stronger interlock effects bymodeling variation in the content of direct ties and examin-ing how such ties are conditioned by the larger social struc-ture.

The present study examines the influence of heterogeneoussocial processes that underlie interlock ties, and the moder-ating effects of indirect network ties, on the creation of stra-tegic alliances between firms. Some ties may promote thecreation of a new alliance, while others could actually reduceits likelihood, depending on the behavioral content of the tie.As a result, there may be both bright and dark sides to em-beddedness in interorganizational relationships. Although re-cent studies have focused on how the network of prior alli-ances provides valuable information to potential partnersabout each other's reliability, capabilities, and needs (Kogut,Shan, and Walker, 1992; Guiati, 1995b; Powell, Koput, andSmith-Doerr, 1996), this literature has not considered therole of alternative networks such as board interlocks in guid-ing the formation of new alliances or in strategic cooperationbetween firms. In this paper, we examine the role of boardinterlocks, focusing on a subset of alliances known as jointventures, which entail the creation of a separate legal entityin which the parent firms take equity, and use the term alli-ance to refer specifically to joint ventures. Such alliancestypically entail a considerable outlay of resources and createenduring and irreversible commitments between partners,which can make the influence of the board interlock networkon their formation even more important. Such a network canbe an important source of information for top managersabout the reliability and capabilities of potential venture part-ners.

CONTENT OF INTERLOCK TIES AND EFFECTSON ALLIANCES

Empirical studies examining the consequences of interlock-ing directorates for the diffusion of innovations and the likeli-hood of strategic change have typically viewed interlock tiesin broad terms as a mechanism for resolving uncertainty fortop management decision makers (Galaskiewicz, 1985a}. Indiscussing how interlock ties may facilitate the diffusion ofan innovation, scholars have emphasized the value of directcommunication between managers and directors in reducingambiguity about the implications of adoption. From this per-spective, information from fellow corporate leaders is par-ticularly influential because it comes from a trusted source(Davis, 1991; Haunschild, 1993). Research on the conse-quences of interlocking directorates would also suggest thatInterlock ties could help resolve uncertainty for top manage-

474/ASa September 1999

Cooperative or Controlling?

ment decision makers about the implications of forming stra-tegic alliances with another firm. Moreover, the questionhere relates not only to the adoption of strategic alliances ingeneral, but also to the choice of a specific partner. Whileprior research has typically described interlocks as conduitsof information about administrative innovations, it is reason-able to expect that board members also communicate infor-mation about their respective parent organizations. The so-cial embeddedness created by interlock ties should helpresolve uncertainty for top managers about the motives andmanagement capabilities of other organizations as potentialalliance partners.

Despite their explosive growth, strategic alliances are associ-ated with a variety of risks and pitfalls that result in consider-able uncertainty about the decision to enter such ties. Thisuncertainty stems from two main sources (Gulati, 1995a,1995b). First, organizations have difficulty in obtaining infor-mation about the competencies and needs of potential part-ners. Such information is often confidential and may not berevealed outside a close relationship, but organizations mustunderstand the needs and capabilities of potential partners ifboth organizations are to derive benefits from the alliance.The second source of uncertainty that affects strategic alli-ances stems from the paucity of information about the reli-ability of the potential partners, whose behavior is a key fac-tor in the success of an alliance. Organizations enteringalliances face considerable moral hazard concerns becauseof the unpredictability of the behavior of partners and thelikely costs to an organization from opportunistic behavior bya partner, if it occurs (Kogut, 1989; Doz, Hamel, and Pra-haiad, 1989; Gulati, Khanna, and Nohria, 1994; Khanna, Gu-lati, and Nohria, 1998). A partner organization may either freeride by limiting its contributions to an alliance or may simplybehave opportunistically, taking advantage of the close rela-tionship to use resources or information in ways that maydamage the partner's interests.

Recent research builds on Granovetter's notion of embed-dedness (1985) and suggests that organizations address thepotential hazards associated with building alliances by relyingon information provided through existing interorganizationalnetworks (Galati, 1998). While the focus of this research hasbeen on the role of the network of prior alliances (e.g., Gulatiand Gargiulo, 1999), board interlocks may also channel infor-mation between firms and thus serve as a catalyst for thecreation of new alliances between firms. Beyond allowingtop managers to form relationships with managers of pro-spective alliance partners, board ties may also enable direc-tors to acquire firsthand knowledge about another firm's ca-pabilities, activities, and plans through their communicationswith top management and their involvement in the decision-making process. Top managers can identify and pursue alli-ances by jointly discovering opportunities for collaboration inongoing discussions. Networks can also provide informationin a timely manrier, which can be important when a firmseeking attractive alliance partners must approach them atthe right juncture and preempt their seeking alliances else-where. This suggests an initial, baseline hypothesis on theeffect of interlock ties on alliance formation:

475/ASQ, September 1999

Hypothesis 1: An interlock tie between two firms will increase thelikelihood of subsequent alliance formation between them.

The discussion thus far has assumed that interlock ties indi-cate positive social contact between top managers and out-side directors of the focal firm, A similar rationale has beenused in much prior research on board interlocks and the dif-fusion of organizational innovations. As interlock researchershave generally recognized, however, there is considerablevariation in the nature of management-board relationships,though the consequences of this heterogeneity have yet tobe systematically examined (Herman, 1981; Johnson,Hoskisson, and Hitt, 1993; Mizruchi, 1996), The form ofmanagement-board relationships can range from a positiveand relatively cohesive relationship between top managersand outside directors to a negative and independent one,with very different consequences for the likelihood of ven-ture formation.

Independent Board Control and Alliances

According to agency perspectives, while top managers areresponsible for ongoing decision management, the board ofdirectors is responsible for decision control, which involvesmonitoring and evaluating management decision making andperformance (Fama and Jensen, 1983}. In effect, the boardis viewed as an efficient control device that can help alignmanagement decision making with shareholders' interests(Beatty and Zajac, 1994). For instance, to the extent thatmanagers' personal preferences regarding executive com-pensation, corporate diversification, or other strategy andpolicy issues conflict with the interests of shareholders,boards can intervene to ensure that shareholders' interestsare protected (Hermalin and Weisbach, 1988; Hill and Snell,1988). Moreover, from this perspective, outside directors inparticular are critical to the board's ability to exercise control,because as non-employee directors they are formally inde-pendent from management and thus better abie to evaluatemanagement decisions and actions objectively on behalf ofshareholders' interests.

In prior years, this agency model of the relationship betweenthe chief executive officer (CEO) and the board could be dis-missed as an anomaly. Organization theorists have typicallysuggested that while outside directors are in a position toexercise Independent control over management, various be-havioral factors effectively limit the social independence ofoutsiders, impairing their ability or willingness to exert con-trol. For instance, given evidence that CEOs traditionally dic-tate the selection of new directors, several authors havesuggested that CEOs can appoint personal friends or otherindividuals with whom they have preexisting social ties (e.g.,Finkelstein and Hambrick, 1988; Wade, O'Reilly, and Chan-dratat, 1990; Cannella and Lubatkin, 1993). Such ties arethought to inhibit the board's willingness to contradict man-agement's preferences on behalf of shareholders. Moreover,organization theorists have long nnaintained that generalizednorms of support among managerial elites enforce a passiverole for outside directors in strategic decision making (e.g.,Herman, 1981; Whisler, 1984). From this perspective, boardshave little potential to serve as independent agents of con-


Cooporative or Confroliing?

trol and, supporting the assumption of interlock theorists,management-board ties are characterized by sociai cohesion.

The recent literature on boards of directors, however, hasprovided some evidence that widespread norms about therole of corporate boards may be changing. Useem (1993)and Westphal and Zajac (1997) have documented the spreadof changes in board structure, composition, and executivecompensation that appear to indicate increased board controlover management among large corporations over the pastfifteen years. This trend may have originated in response toexternal criticism from institutional investors and other stake-holders and the threat of lawsuits over perceived negligencein protecting shareholders' interests (Kesner and Johnson,1990; Davis and Thompson, T994). External constituentshave demanded evidence that boards are willing to challengemanagement's decisions on their behalf. For instance,boards have been told to expand the search for new direc-tors beyond the CEO's close circle of personal friends and toalter board structure and processes in ways that diminishthe CEO's direct control over board meetings (Kaplan andHarrison, 1993; Daily, 1996). In effect, boards have beenpressured to adopt a role characterized by more independentmonitonng and control over management. Nevertheless,while there has been a general move toward more assertiveboards that assert greater control over CEOs, there remainsconsiderable variance across boards in the extent to whichthey have adopted a controlling orientation.

The consequences for alliance formation. There are sev-eral possible consequences of independent board control onthe prospects of alliance formation between the focal firmand manager-directors' home companies. On one level, aCEO-board relationship characterized by monitoring and con-trol simply entails lower cohesion between the CEO and theboard, or the absence of a strong tie, but it may go furtherthan that. Independent board control over management mayactually produce a negative relationship between the CEOand the board characterized by a lack of mutual understand-ing and distrust. When benevolence and support toward theCEO is replaced with independent control over the CEO,leaders of the firm can become effectively divided into sepa-rate groups: decision managers {i.e., the CEO and other topmanagers) and decision controllers (i.e., outside directors)(Fama and Jensen, 1983), where they were previously com-mon members of a mutually supportive, inner circle of elites(Useem, 1982). The literature on intergroup relations has pro-vided consistent evidence, in both laboratory and field set-tings, that dividing a single group of individuals into two ormore separate groups has a variety of negative effects onrelations between members of the different groups (Millerand Brewer, 1996). Empirical studies have demonstratedthat when individuals are divided into separate groups, atti-tudes about the out-group members become significantlymore negative (Gaertner et al., 1989; Messick and Mackie,1989). In particular, group categorization has been shown tofoster distrust toward out-group members while also creat-ing the perception of intergroup conflict (Kramer, 1996;Miller and Brewer, 1996; Labianca, Brass, and Gray, 1998).


Out-group categorization, which in the case of interlocks oc-curs when CEOs view outside directors as controllers ratherthan supporters or fellow managers, can promote distrustboth with respect to the capabilities of the other party (task-based trust) and the risk that they might limit their contribu-tions to the relationship (relational trust) (Creed and Miles,1996), thus prompting negative evaluations of the perceivedcapability and personal reliability of the other party. This out-group bias occurs even when the basis for group categoriza-tion is arbitrary or minimal (Brewer, 1979). Moreover, Kramer(1994, 1996: 224) and others (Fenigstein and Vanable, 1992)have found evidence that when individuals are subjected to"evaluative scrutiny" or control by out-group members, "apattern of exaggerated mistrust" may develop.

Applying research on intergroup relations to the CEO-boardcontext, we expect that when outside directors assert them-selves as an independent group of controllers accountable toshareholders rather than management, distrust can arise be-tween top managers and outside directors. Whereas outsidedirectors on passive and supportive boards are effectivelyinsiders with regard to their orientation toward management,on controlling boards such directors adopt the perspective ofan independent outsider. As a result, the perception of a di-vision between insiders and outsiders can reinforce "a gen-eralized sense of distrust" across groups and lead to "esca-lating cycles of distrust" when out-group members areexercising control (Sitkin and Stickel, 1996: 199). A behav-ioral manifestation of distrust is "reduced cooperative effortsof all kinds" and enhanced competition for resources andstatus between groups (Brewer and Kramer, 1985; Gaertneret a),, 1989; Creed and Miles, 1996: 27). Thus, intergroupbias would lead each party of the management-board rela-tionship to view members of the other group as less trust-worthy in both professional and personal terms, reducinginterest in various forms of cooperation,

The literature on strategic alliances suggests that trust playsa critical role as an enabling condition of alliance formation(Ring and Van de Ven, 1992; Gulati, 1995a, 1995b; Dyer,1996; Gulati and Singh, 1998). Trust fundamentally entails awillingness to put oneself at risk (Barney and Hansen, 1994;Mayer, Davis, and Schoorman, 1995), and several authorshave emphasized the potential for opportunistic behavior toderail ioint ventures {e.g., Doz, Hamel, and Prahalad, 1989;Gulati, Khanna, and Nohria, 1994). In the presence of trust,managers will be less concerned about the incentive of apartner to cheat or free ride in cooperative relations by limit-ing its contribution to the joint enterprise. Since intergroupbias resulting from independent board control can diminishboth task-based trust and relational trust, one might expectthat when directors have asserted themselves as an inde-pendent group responsible for controlling managers ratherthan supporting them, CEOs may view them as less trust-worthy alliance partners. Board independence can also pre-vent top managers and manager-directors from becomingfamiliar with each other's management and decision-makingstyles and developing a professional rapport, and managerstend to believe that teamwork and rapport between them isa critical factor in the success of alliances (Alster, 1986;

478/ASa September 1999


Borys and Jemison, 1989). Moreover, given that distrust to-ward an independent, controlling group is a basic and power-ful human response (Kramer, 1994, 1996; Fenigstein andVanable, 1992), independent board control may have a par-ticularly strong, negative effect on alliance formation be-tween top managers and manager-directors:

Hypothesis 2: The greater the board's control over the CEO, thelov /er the likelihood of subsequent alliance formation between thefocal firm and outside directors' home companies.

CEO-board Cooperation and Alliances

While empirical research on boards has typically assumedthat board involvement in corporate affairs entails indepen-dent monitoring and control by outside directors (Johnson,Hoskisson, and Hitt, 1993), the larger literature on boardshas suggested another form of involvement. In his classicqualitative study. Mace (1971: 179) concluded that, whileboards often did not challenge management's final deci-sions, they may nevertheless provide "advice and counsel"to management on strategic issues during the decision-mak-ing process. Pfeffer and Salancik (1978: 170) also distin-guished the provision of advice and counsel from board con-trol as two different forms of board administration (see alsoMintzberg, 1983). In a recent large-sample empirical study,Westphal (1999) found support for this general classification.Factor analysis showed that CEO-board relationships couldbe classified into three categories: independent monitoringand control, close cooperation (i.e., advice and counsel), orinaction. Moreover, qualitative and survey evidence suggeststhat advice and counsel is typically provided at the CEO'srequest (Lorsch and Maclver, 1989; Demb and Neubauer,1992). Thus, rather than remain independent of top manag-ers to permit objective monitoring and evaluation of manage-rial decision making, some boards enter closer working rela-tionships with CEOs by providing advice and counsel at theCEO's request. In such cases, CEOs direct a cooperativeform of board involvement in which boards work togetherwith them to govern the firm, rather than separately in aprincipal-agent relationship.

The consequences for alliance formation. CooperativeCEO-board relationships may influence alliance formation be-tween the focal firm and manager-directors' home compa-nies in several ways. On one level, CEO-board cooperationshould enhance trust between top managers and outsidedirectors through social interaction alone. Simmei's (1964)theory of trust emphasized how the mere occurrence of so-cial interaction builds trust or the expectation of faithfulness,and other theorists have suggested that more frequent inter-action increases trust by enhancing mutual affect and famil-iarity (Laumann, Galaskiewicz. and Marsden, 1978; Gulati,1995a; Creed and Miles, 1996). Accordingly, the heightenedsocial interaction that results when there is greater CEO-board cooperation (i.e., advice seeking) should reinforce rela-tional trust between CEOs and outside directors.

The connection between cooperative interactions in CEO-board relationships and the extent of trust between manag-ers can also be understood by considering some of the evi-dence from research on intergroup relations. According to


this literature, cooperative interactions between group mem-bers make common goals more salient, which builds mutualtrust and respect (Gaertner et al., 1990, 1999). Thus, whileindependent board control may reduce trust by effectivelydividing top managers and outside directors into separategroups, the CEO's seeking advice from the board should en-hance trust by drawing outside directors into a collective de-cision-making team. In effect, just as negative affect and dis-trust toward an independent, controlling group is a basic andpowerful human response, cooperation between groupmembers can engender in-group biases that lead to positiveaffect and higher, even excessive levels of trust betweenindividuals (Fenigstein, 1979; Kramer, 1996). Given the im-portance of intermanagement trust in facilitating alliance for-mation, cooperative CEO-board relationships should promotealliances between a focal firm and those of outside directorsby enhancing confidence in each other's reliability and mana-gerial capability and lowering the perceived risk of opportun-ism:

Hypothesis 3: The greater the cooperation betv\/een the CEO andthe board, the higher the likelihood of subsequent alliance forma-tion between the focal firm and outside directors' home companies.

The Role of Indirect Ties

While little empirical research has examined how socialstructural factors such as indirect ties moderate the effectsof dyadic interlock ties between firms, qualitative evidencesuggests that managers may have access to indirect infor-mation about directors through their appointments on otherboards (Useem, 1984; O'Neal and Thomas, 1993). An indi-rect or third-party tie could provide top managers with animportant source of information about outside directors whosit on their board. For example, a top manager A is exposedto second-hand information about outside director B on hisor her board when A has a common appointment on anotherboard with a third director C, who sits on B's board. Theseindirect ties are particularly relevant to the present study, inlight of recent evidence suggesting that third-party ties canaffect the level of trust between individuals or organizations(e.g., Raub and Weesie, 1990; Burt and Knez, 1995; Gulati,1995b; Gulati and Gargiulo, 1999).

It is typically supposed that third-party ties will enhance trustbetween parties to a relationship by increasing the reputa-tional costs of noncooperative behavior (Van de Ven, 1976).For instance, if A is cheated by relationship partner B, and Ahas third-party ties to B through C, A can impose reputa-tional costs on B by spreading the word to C that B cannotbe trusted. Given this threat, A can trust B not to defectfrom cooperative exchange (Kreps, 1990}. The claim thatthird-party ties enforce cooperation through reputational ef-fects assumes that noncooperative behavior is illegitimate ornon-normative, like cheating a friend (Granovetter, 1992). Inmany cases, however, noncooperative behavior involvingcompetition or control is not normatively proscribed in thelarger social structure, or the related norms are ambiguous.As several authors have noted, norms governing CEO-boardrelationships have become uncertain: it is not clear whetherindependent board control is more or less normative or legiti-



mate than CEO-board cooperation (Lorsch and Maclver,1989; Useem, 1993). Accordingly, noncooperative behavior,such as exercising independent control, does not necessarilyhave negative reputational consequences for the partici-pants; it does not necessarily damage a director's career, asempirical research on director selection has shown (Zajacand Westphal, 1996), Thus, third-party ties between a CEOand his or her board members may not necessarily reducethe likelihood of noncooperative behavior in CEO-board rela-tionships.

While traditional perspectives on indirect network ties maynot apply to board interlocks, recent research on the effectsof third-party ties suggests a more germane perspective.Burt and Knez (1995, 1996) have extended existing theorieson how social structure affects trust by proposing that third-party ties amplify existing trust or distrust in professional re-lationships (see also Labianca, Brass, and Gray, 1998). Theyshowed empirically that when the immediate relationshipbetween managers tended to foster trust between them,third-party ties further enhanced trust in the relationship. Atthe same time, third-party ties amplified any distrust that al-ready existed in the relationship. They concluded that third-party ties influence the intensity but not the direction of trustin managerial relationships; that Is, indirect ties make manag-ers more certain of their trust (or distrust) in another. In de-veloping their theory, Burt and Knez suggested that manag-ers exchange information or gossip with third-party tiesabout other managers, and the social dynamics underlyingsuch interactions lead third parties to reaffirm whichever pre-disposition managers have toward their colleagues. This isconsistent with anthropological and social psychological re-search on network gossip, which suggests that people gos-sip with third parties in a search for affirmation of their feel-ings and beliefs about other individuals in their network; inthe process, gossip also serves to reaffirm the values thatunderlie those beliefs (Cox, 1970; Haviland, 1977; Besnier,1989), Moreover, by validating ego's trust or distrust in alter,a third party strengthens his or her relationship with ego(Byrne, Ciore, and Worchel, 1966). Such behavior can be mo-tivated by political self-interest or simply by the desire tomaintain social cohesion for its own sake (Cox, 1970; Burtand Knez, 1995).

We can extend our previous hypothesis by considering thepossibility that third-party ties between a CEO and his or herboard members resulting from appointments on otherboards may amplify the effects of these different relation-ships on trust between CEOs and outside directors. Asnoted above, qualitative research on boards suggests thatthe relationship between top manager A and outside directorB is influenced by third-party ties when A has a connmonappointment to another board with a third director, C, whosits on B's board. From the third-party gossip perspective,when A and C discuss B (or A's relationship with B), the so-cial dynamics underlying such interactions will lead C to con-firm A's predisposition by drawing on his or her prior experi-ence with B. For example, if A expresses doubt to C aboutwhether B can be trusted to support A's decisions, C willtend to affirm A's distrust, either by providing explicit infor-


mation or "replicating accounts" of B's behavior or throughmore subtle affirmations (e.g., "that doesn't surprise me") ornonverbal signals (Cox, 1970; Burt and Knez, 1995: 260).Such interactions are especially likely in that top managershave become increasingly concerned in recent years aboutwhether they can count on the loyalty and support of theiroutside directors (Lorsch and Maclver, 1989). This suggestsadditional hypotheses, predicting that Indirect ties betweenCEOs and outside directors through third-party directors inthe interlock network will tend to amplify the relationshipbetween each kind of CEO-board tie and the likelihood ofalliance formation:

Hypothesis 4: Indirect interlock ties between the CEO and outsidedirectors through thirdi:5arty directors v 'ill interact with the contentof the focal CEO-board tie to predict alliance formation between thefocal firm and outside directors' home companies.

Hypothesis 4a: The more indirect interlock ties there are betv\/eenthe CEO and outside directors through third-party directors, thestronger the negative relationship between board control over theCEO and the likelihood of subsequent alliance formation betweenthe focal firm and outside directors' home companies.

Hypothesis 4b: The more indirect interlock ties there are betweenthe CEO and outside directors through third-party directors, thestronger the positive relationship between CEO-board cooperationand the likelihood of subsequent alliance formation between thefocal firm and outside directors' home companies.

METHOD

Sample and Data Collection

The sample frame for this study included 600 firms selectedfrom the Fortune and Forbes 500 indexes of U.S. industrialand service firms. We collected both archival and survey-based information on these firms. Archival information wascollected on alliances formed, board interlocks and otherboard characteristics, strategic variables, and financial data.To measure board control and CEO-board cooperation, aquestionnaire survey was distributed in April 1995 to allCEOs of the 600 firms. In addition, to assess interrater reli-ability, another questionnaire was sent to individuals servingas outside directors at one or more companies whose CEOresponded (N = 1,312 directors).

Surveys of top managers have been plagued by low re-sponse rates. To ensure the highest possible response inthis case, the following steps were taken (Forsythe, 1977;Fowler, 1993; Groves, Cialdini, and Couper, 1992): First, anin-depth pre-test was used to refine the format and length ofthe survey. Second, the cover letter linked the present studywith prior surveys on top management issues conducted bya major business school, while noting that hundreds of theirpeers had responded to the prior surveys; the letter alsohighlighted the need for research on CEO-board relations,which also engages respondents' natural interest in the topic(see Groves, Cialdini, and Couper, 1992). Third, nonrespon-dents were sent a second letter with a new questionnaireabout 21 days after the initial mailing. As a result of theseefforts, 263 of the 600 CEOs in the sample frame re-sponded, a response rate of 44 percent. Moreover, 564 ofthe 1,312 outside directors responded, yielding a response



rate of 43 percent. These response rates are high comparedwith other top management surveys (Pettigrew, 1992).

To check for nonresponse bias, respondents and nonrespon-dents were compared across a variety of firm characteristicsusing the Kolmogorov-Smirnov two-sample test (Siegel andCastellan, 1988). This assesses whether significant differ-ences exist between the distribution of respondents andnonrespondents for a given variable. As shown in table 1,the results of this test consistently suggest that respondentsand nonrespondents come from the same population. Wealso assessed nonresponse bias according to the presenceor absence of specific board structures and practicesthought to indicate board control (cf. Hoskisson, Johnson,and Moesel, 1994; Beliiveau, O'Reilly, and Wade. 1996).These analyses provided further evidence that nonresponsebias is not present in our data. In particular, a series of differ-ence in proportions tests showed that respondents and non-respondents were not significantly different with respect tothe existence of (1) an executive committee on the board (D= ,018; p = ,260); (2) a nominating committee composed ofoutsiders (D = .011; p = .585); or (3) a management develop-ment and compensation committee (D = ,009; p - .649).Moreover, respondents were not significantly different intheir use (vs. non-use) of stock compensation for directors(D - .019; p - .212), and CEOs of responding firms wereneither more nor less likely to serve as an ex-officio non-voting director on the board (D = .016; p = .435).

We collected data on all alliances initiated by firms in thesample frame from 1970 to 1996, This sample includes allinterfirm partnerships that entail the creation of a new legalentity in which both partners hoid equity, also referred to asjoint ventures. We coded these data manually from the Funkand Scott Predicasts Index of Corporate Change and fromLexis/Nexis- We recorded only joint ventures that had actu-ally been formed and excluded reports of probable joint ven-tures that never materialized. We made an effort to ensure

Table 1

Results of Kolmogorov-Smirnov Difference Test for Nonresponse Bias*

Largest positive Largest negativeVariablest difference difference p-value

SizePerformanceSofvencyLiquidityOutsider ratioJoint tenureConstraintNumber of interlock tiesPrior alliance activityInstitutional ownership

.073

.037

.067

.030

.048

.052

.058

.002

.041

.025

-.037-.069-.023-.100-.035-.013-.029-.059-.091-.082

.499

.510

.544

.149

.911

.882

.763

.754

.216

.361

* N = 218 respondents and 274 nonrespondents. Differences indicate thelargest positive/negative deviations observed when the cumulative fre-quency distribution (i.e., step function) for nonrespondents is subtractedfrom that of respondents.

t This test could not be applied to board leadership structure, but respondentsand nonrespondents did not differ significantly on the portion of firms withseparate CEO and board chair positions {.29 versus .27).


that these data were comprehensive in covering all alliancesduring the time period. We predict alliance formation overthe two-year period following the survey date (1995-1996).and we used the remaining historical alliance data to com-pute some key control variables described below. We col-lected data on board interlocks and board structure for theperiod 1994-1995 from Standard and Poor's Register of Cor-porations, Directors, and Executives and the Dun and Brad-street Reference Book of Corporate Management. To calcu-late measures of market constraint (discussed below), weobtained input-output data from the database created by theInterindustry Economics Division of the Bureau of EconomicAnalysis (cf. Burt, 1992; Mizruchi, 1992). Data on financialcharacteristics and other firm attributes were obtained fromCOMPUSTAT.

Measures

Alliance formation was measured with a dichotomous vari-able, coded 1 if the two firms in a dyad entered into an alli-ance during the two-year period following the survey date(i.e., 1995-1996). We also conducted two separate analyseswith alliance formation measured for different time periods.In one analysis we measured alliance formation over oneyear (1996), and in the second analysis, we observed alli-ances formed over the period 1993-1994, using retrospec-tive measures of board control and CEO-board cooperation in1992. For each of these separate analyses, results for thehypothesized relationships were very similar to the resultspresented below, suggesting that our findings are robust fordifferent time periods.

We measured board interlocks as directional ties, which arecreated by individuals who are principally affiliated as officersor owners with the firms they connect {Davis, 1991; Haun-schild, 1993; Palmer, Jennings, and Zhou, 1993). Thus, twofirms, A and B, are coded as having an interlock tie when atleast one officer or owner from firm A serves as an outsidedirector at firm B, or vice versa.

A pre-test involving in-depth pilot interviews with 22 topmanagers and board members was used to refine and re-word the survey items (cf. Fowler, 1993: 102). Board controland CEO-board cooperation were measured with two multi-item scales from the CEO survey that were carefully vali-dated with responses from the outside director survey andalso with archival measures of board characteristics. Items inthe control scale assessed key behavioral elements of boardcontrol that have been theorized to entail board indepen-dence from management, including the board's tendency tomonitor and evaluate CEO decision making and performanceand the frequency with which directors challenge the CEO'sposition on strategic issues, rather than deferring to theCEO's judgment. Items in the cooperation scale were basedon prior qualitative research about how CEOs may engage inongoing collaboration with outside directors by seeking theiradvice and counsel on strategic issues, as discussed above.

We conducted a confirmatory factor analysis using LISREL.The chi square for the measurement model comprising thecontrol and cooperation constructs was not significant (pvalue = .48), suggesting that the model provides an ad-


Cooperative or Controliing?

This method differs from the ordinary re-gression method for estimating factorsby mintmizing the sum of squares of theunique factors over the range of items,rather than minimizing the discrepanciesbetween the true and estimated factors.The Bartlett method produces less biasedestimates than the regression method formoderate to large samples, although itmay also produce less accurate esti-mates (Harman, 1976) In this case, how-ever, separate analyses confirmed thatthe results were identical when factorswere estimated using the regressionmethod rather than the Bartiett method.

These measures effectively assume thatboards engage in control or cooperationas a group (fama and Jensen, 1983; Al-derfer, 1986), so that the effects on theCEO's relationship with each directorshould be similar across directors, leadingto similar effects on the Hkelihood of alli-ance formation between the focal firmand each of the outside directors' homecompanies. While this assumption is con-sistent with the prior literature on CEO-board relationships, in separate modeiswe measured the individual participationof directors in control and cooperationactivities with responses to separatequestions in the director survey. Theseadditional results are discussed furtherbelow.

equate fit to the data (Bolten, 1989). Thus, we proceeded toestimate cooperation and control factors using the Bartlettmethod."* Interitem reliability was acceptable for both scales,with alphas of .89 for the cooperation items and .88 for thecontrol items (Nunnally, 1978).^

We also assessed interrater reliability by comparing CEOs'and outside directors' responses on the board control andcooperation Items. We used the kappa correiation coeffi-cient, which corrects for the level of correlation that wouldbe expected by chance. The results of this analysis and abrief description of the questions used for each constructare provided in table 2. Values exceeding .75 are typicallythought to indicate excellent agreement, and values be-tween .40 and .75 indicate fair to good agreement (Landisand Koch, 1977; Fleiss, 1981). Kappas exceeded .75 for allitems but one, and the overall value was .82. The sample forthis analysis included companies with a responding CEO andat least one responding outside director (N = 188). As dis-cussed further below, we ran separate analyses in whichcontrol and cooperation were measured with directors' re-sponses rather than CEOs' responses.

We also conducted tests of convergent validity for our mea-sures of board cooperation and control. First, we developedarchival measures of each construct. The archival measurefor control includes multiple aspects of board structure andpractice that are thought to facilitate controlling behavior byboards: the use of stock to compensate directors, institu-tional ownership, and the presence or absence of the spe-cific committees listed above (Hoskisson, Johnson, andMoesel, 1994; Belltveau, O'Reilly, and Wade, 1996; David,Kochhar, and Levitas, 1998), Institutional ownership wasmeasured as the percentage of total common stock held bypension funds, banks and trust companies, savings andloans, mutual fund managers, and labor union funds (Hansenand Hill, 1991), The dichotomous measures were combinedinto a Guttman scale and then combined with institutionalownership using principal components (Jackson, 1991).

The archival measure of cooperation is a composite of threevariables: joint tenure of the CEO and directors, complemen-tary functional backgrounds of the CEO and directors, andCEO stock ownership. The organizational demography litera-ture has provided consistent evidence that higher joint ten-ure increases the level of task-related communication andproblem-solving behavior among group members (Zengerand Lawrence, 1989; O'Reilly, Snyder, and Boothe, 1993;Smith et al., 1994; Williams and O'Reilly, 1997). Moreover,the upper echelon perspective would suggest that directorsare more valuable to CEOs as a source of strategic adviceand counsel if they have a complementary base of functionalexpertise and experience {e.g., if the CEO has a financialbackground and directors have marketing backgrounds)(Hambrick, Cho, and Chen, 1996). Functional backgroundwas measured using Hambrick and Mason's (1984) classifi-cation, calculated as the percentage of directors who hadfunctional backgrounds complementary to the CEO. Finally,given that stock ownership aligns CEOs' interests withshareholders' interests, it may motivate CEOs to engage thecooperation of board members in the strategic decision-mak-


Table 2

Resulte of Interrater Reliability Assessment iN s 1S9)*

Itemst

Agreement

Observed

86.50

85.65

87.76

88.19

(%)

Expected

24.86

21.16

25.82

27.40

Kapp

.82

.82

.84

.84

Board control1. To what extent does the board monitor top management

strategic decision making?2. To what extent does the board formally evaluate your

performance?3. To what extent does the board defer to your judgment on final

strategic decisions?4. Over the last twelve months in how many board meetings

have one or more directors challenged your positfon on astrategic issue?

CEO-board cooperation1. To what extent do you solicit board assistance in the

formulation of corporate strategy?2. To what extent do you use outside directors as a "sounding

board" on strategic issues?3. How often have you sought the advice and counsel of outside

directors in discussions outside of board/committee meetings(by telephone or in person)?

Overall kappa

79.32

87.34

86.50

24.00

23.43

24 93

.73

.83

.82

.82

* W = 188. When multiple outside directors responded for the same company, directors' responses were averaged toensure that reliability estimates were not inflated by common perspectives derived from holding the same position.

t The phrasing of each item is taken from the CEO survey; most items were altered appropriately for the director survey(e.g., "To what extent does the CEO . . ."). For purposes of comparison across items, kappas were calculated forcontinuous-scale items by converting them into categorical variables {i.e., divided into quartilesj.

t Z-statistics for all kappas are highly significant.

ing process (Murphy, 1986; Jensen and Murphy, 1990).These three variables were combined into a composite mea-sure using principal components.

The archival measure of control was significantly correlatedwith the survey measure of control (r = .42), while the archi-val measure of cooperation was significantly correlated withthe survey measure of cooperation (r = .34). Moreover, thesurvey measure of cooperation was not significantly corre-lated with the archival measure of control, while the surveymeasure of control was not associated with the archivalmeasure of cooperation. This analysis provides further evi-dence for the construct validity of the survey measures.

We also examined the correlation between our archival mea-sures of board cooperation and control and a survey mea-sure of trust in the CEO-board relationship. This measure isa multi-item scale included in both the CEO and director sur-veys; specific items in the CEO scale assess, for instance,the degree to which the CEO feels that he or she can trustthe board and the extent to which his or her relationshipwith outside directors is characterized by distrust. This mea-sure showed high interitem reliability (alpha = .91), as wellas high interrater reliability Ikappa = .85). Moreover, the trustmeasure is positively associated with the measure of coop-eration (r = .47) and negatively correlated with the measureof board control (r =i -,41). This further supports the conver-gent validity of the survey measures.

486/ASQ. September 1999


Indirect ties between CEOs and outside directors throughthird-party directors (third-party ties) were measured for eachdyad as the number of board appointments shared by theCEO and board members of the outside director's homecompany board, excluding the focal board. We tested thehypothesized interaction effects between third-party ties andthe focal CEO-board ties using the product-term approach(Jaccard, Turrisi, and Wan, 1990).

Control variables. To ensure the robustness of our results,we included a number of control variables considered to in-fluence the formation of ventures between firms. Resourcedependence perspectives suggest that firms may use coop-erative strategies to manage their dependence on otherfirms (for a review, see Oliver, 1990). According to this view,firms that are particularly dependent on each other maychoose to form an alliance to secure future access toneeded resource flows. To capture the role of resource con-siderations in promoting joint ventures, we built on Burt's(1983, 1992) measure of market constraint to capture thedegree of resource interdependence between firms. Data onmarket constraint are available only at the industry level andhave typically been used to compute aggregate constraintscores for firms. Burt (1983) defined constraint as the pro-portion of industry A's total transactions that involve B, mul-tiplied by the four-firm concentration of B. This measure indi-cates the extent to which firms in industry A are dependenton resources provided by firms in industry B. Burt's firm-level measure of constraint was based on the dependenceof the firm's primary industry (i.e., the firm's primary two-digit SIC code). Galaskiewicz (1985b) modified this measureby calculating the median constraint score across all indus-tries in which the two firms participated. Palmer, Friedland,and Singh's (1986) measure involved summing the numberof significant constraint relations between industries inwhich the firms participated. We followed Mizruchi (1992) inreducing these three operationaiizations to a single variableusing principal components analysis. Mutual interdepen-dence between firms in a dyad is then measured as the totalamount of constraint between them. This measure is high-est when both firms are in highly concentrated industrieswith a heavy flow of transactions between them.

Since constraint is measured at the industry level, we alsocontrolled for industry overlap, measured as the percentageof total sales between the two firms that are made to thesame industry, to assess strategic complementarity at thefirm level {Mowery, Oxiey, and Silverman, 1996). This dyad-level measure allows us to assess the extent to which everyspecific pair of firms overlaps in the industries in which theyoperate. This measure has some limitations in that it treatsall industries as equivalent and does not take into account thecharacteristics of the industries in which the overlap occurs.

We included a number of measures that prior researchwould suggest may be important strategic and economicdrivers for the creation of new ventures: (1) size, measuredas the log of total sales; (2) performance, measured as acomposite of return on assets and market-to-book value,which were combined using principal components analysis;(3) solvency, measured as the total amount of long-term


In out primary analyses, we measuredthese variables at the focal firm tevel, butthese variables might 3)scj be specified althe dyad level; several authors have sug-gested that alliances form hetween firmswith complementary needs and capabili-ties {e.g., Aiken and Hage, 1968, Richard-son, 1972; Paulson, 1976; Burgers, Hill,and Kim, 1993) Thus, for instance, poorlyperforming firms may seek alliances togain access to capabilities of high-per-formmg firms, while the latter seek toleverage those capabilities. Similarly,small firms may enter into aUiances withlarge firms to gain access to economiesof scale and scope, vi'hile large firmsseek to spread the risks associated withnew ventures. To measure each attributeat the dyad level, we divided the smallervalue by the larger value. Larger values ofthis measure indicate greater similarity onthe given attribute, in separate analyses,we included these dyad-level measureswith the firm-level measures, but the re-sults were not subsrantiveiy drfferantfrom those presented below. We alsoconducted analyses in which separatevariables were included for each firm inthe dyad, and again, the results weresubstantively unchanged. Finally, the re-sults were robust to inclusion of addi-tional measures of diversification, includ-jng the percentage of sales made toglobal markets.

The effects of prior alliance activity wereweaker when thrs variable was measuredoniy for the focal firm or at the dyadlevel. This was not true for the economic/strategy variables, as noted above, per-haps because alliance activity of one firmin the dyad was not correlated with activ-ity of the other firm, v^hile the strategyvariables were significantly correlated be-tween firms in the dyad, such that addinga separate economic/strategy variable foreach firm in the dyad had little effect onthe results.

debt divided by current assets; (4) research and develop-ment (R&D) intensity, measured as research and develop-ment expense divided by total sales of the focal firm; (5) ad-vertising intensity, measured as advertising expense dividedby total sales of the focal firm; and (6) diversification, mea-sured with the entropy variable, which takes into accountthe number of segments in which a firm operates andweights each segment according to its contribution to totalsales (Palepu, 1985). Each variable was measured in theprior year.^

We also controlled for the firms' prior alliance history. Firmsmay acquire unique information about other firms throughtheir prior alliances with them, while developing routines thatare specifically adapted to cooperation with those firms.Such investments may increase the likelihood that firms willcontinue to ally with the same firms over time {Gulati,1995a, 1995b; Powell, Koput, and Smith-Doerr, 1996), Thus,we controlled for the number of prior alliance ties betweenfirms in the dyad {prior alliance ties). To control for the his-torical propensity of each firm to initiate alliances, we alsoincluded variables indicating the total number of prior alli-ances initiated by each firm in the dyad {prior alliance activ-ity] (Gulati. 1999).* To compute this measure as accuratelyas possible, we observed alliance activity from 1980 to 1994,since alliances became more common among large firms inthe early 1980s. In separate analyses, we measured allianceactivity over several different time periods, including longerperiods (e.g., from 1970), as well as shorter periods (e.g.,since 1990 or 1993), and the results for the hypothesizedrelationships were substantively unchanged from the resultspresented below. These two sets of measures serve as use-ful controls for unobserved heterogeneity that results fromunobserved propensities by the actor to engage in those ac-tivities in the future (Heokman and Borjas, 1980).

We also controlled for two other kinds of board ties thatcould influence CEO-board relationships and alliance forma-tion. First, we controlled for common board appointments{common appointments) held by the CEO and the outsidedirector on other boards. A common tie exists if CEO A andan outside director B on the focal board both serve as out-side directors on another board. Social cohesion resultingfrom common membership as directors on another boardcould enhance the potential for trust and cooperation be-tween CEOs and directors on the focal board. Second, wecontrolled for the total number of appointments {total ap-pointments) held on other boards by the CEO and the out-side director. As discussed above, our measure of third-partyties effectively includes indirect ties with a distance of twolinks: if CEO A sits on another board with C, who sits onoutside director B's board, A and B are separated by twolinks. When this measure is held constant, the control vari-able for total appointments captures the effect of indirectties of greater length (i.e., three links or more). We con-trolled for such ties because they could amplify trust or dis-trust in the focal CEO-board relationship, although we wouldexpect them to have a weaker effect than ties to third par-ties that are directly connected to both members of theCEO-board relationship.


Cooperative or Controiting?

Given that alliance formation could also be influenced by thecontent of sent interlock ties (i.e., relationships between topmanagers and manager-directors at the latter directors'home company boards, which are not included in measuresof cooperation and control), we controlled for reciprocatedappointments, coded as 1 if a top manager serves on thehome company board of an outside director who is on thefocal firm's board.

Further, we controlled for several other possible exogenousinfluences on management-board relationships. Several au-thors have speculated that personal friendship ties betweentop managers and outside directors could influence profes-sional relationships on the board. Mace (1971), Finkelsteinand Hambrick (1988), and others have suggested that friend-ship ties reduce the board's tendency to independentlymonitor management, although recent evidence suggeststhat such ties may enhance cooperation without reducingthe level of monitoring activity (Westphal, 1999). There isalso anecdotal evidence that friendship ties between topmanagers may facilitate the formation of strategic alliances(Alster. 1986). Thus, we included a survey measure of friend-ship ties, indicating the portion of the board composed ofthe CEO's personal friends. In separate analyses, describedbelow, in which cooperation and control were measured foreach CEO-director dyad, friendship ties were also measuredat the dyad level. Descriptive statistics and bivariate correla-tions are provided in table 3.

Analysis

We used maximum-likelihood logit regression analysis totest the effect of interlock ties on the likelihood of allianceformation (Aidrich and Nelson, 1984; Hosmer and Leme-show, 1989). Because the appropriate risk sets to test eachof the hypotheses differ somewhat, we conducted a numberof additional analyses to ensure consistency across our find-ings. Since hypothesis 1 examined the effect of interlockties on alliance formation, the risk set for this analysis in-cluded all possible dyadic combinations between each of thefocal firms in the final survey sample and all firms in the totalsample frame (73,510 dyads). Since hypotheses 2-4 assumethat an interlock tie exists, because board control or coopera-tion only occur when there is an interlock, the risk set nar-rows here from all possible dyads to only those dyads forwhich there was an interlock tie between the two firms.Thus, to test the effects of board control vs. cooperation onalliance formation, as well as the moderating effects of third-party ties, we conducted an initial set of analyses using iogitregression on the sample of dyadic combinations betweenthe focal firm and each of the home companies of CEO-di-rectors on the board (N = 898). Moreover, for the sample ofpossible dyadic combinations that included a responding out-side director, we also examined whether individual CEO-board-member relationships mattered by estimating separatemodels using each responding director's assessment of hisor her individual relationship with the CEO (N = 412 dyads).

In addition, we estimated Heckman selection models to en-sure that logit estimates were not biased by any unmea-sured differences between the smaller sample of CEO-direc-


Table 3

Descriptive Statistics and Pearson Correlation Coefficients for Analyses of Board Control andCEO-Board Cooperation*

Variable Mean S.D. 7 2 3 4

1. Alliance2. Interlock tie3, Board control4. CEO-board cooperation5. Third-party ties6- Prior alliance activity, firm 17. Prior alliance activity, firm 28. Prior alliance ties9, Constraint

10. Size11, Performance12. Solvency13. R&D intensity14, Advertising intensity15. Diversification16. Industry overlap17. Common appointments18. Total appointments19. Reciprocated appointments20. Friendship ties

.15

.01

.00

.002.795.986.35

.09

.007.69

.00

.38

.02

.02

.74

.091.238.21

.11

.37

.36

.11

.88

.822.035.575.94

.311.331.501.01

.31

.02

.03

.56

.251.015.31

.31

.34

—.06

-.26.31.12.29.34.23.17.02

-.23^.08-.18-.14-.05

.22

.09

.07

.03

.16

-

.16

.05

.07

.18

.17-.05

.05-.01

,06-02.08.25.15.11.29.02

-.16.05.07.08

-.17.18

-.04-.08-.05-.03

.02-.12-.10-.08-.05-.09-.18

.06

.09

.09

.21

.13-.08-.04-.01

.00

.06-.17

.20

.07

.02

.17

.21

.11,11.14

-.02.03

-.01.01.04.02.00.04.18.37.03.04

.02

.17

.12

.06-.13

,02-.11-.05-.04

.03

.04

.01

.02

.03

Descriptive statistics and correlation coefficients are calculated for the sample of interlocked firms [N = 898), exceptstatistics for interlock ties, which are calculated for the larger sample of all possible dyads IN = 73,510).

tor dyads and dyads in the larger sample frame used fortesting hypothesis 1. This approach uses the larger risk setto assess hypotheses 2-4 (i.e., N = 73,510). The Heckmanmodel is essentially a two-stage procedure that estimatesthe likelihood of interlock ties with probit regression andthen incorporates estimates of parameters from that modelin a second-stage regression model to predict alliance forma-tion among dyads with an interlock tie; the second-stagemodel is also estimated with probit regression (van de Venand van Praag, 1981).

Formally, the Heckman mode! assumes that a potential ob-servation is observed if x^Bi + u > 0, where u-, has a stan-dard normal distribution. In addition, there is another regres-sion equation, y = X2B2 + aU2, vv'here U2 also has a standardnormal distribution but is potentially correlated with Ui withcorrelation p. In our case, the latter equation represents alli-ance formation while the former represents the likelihood ofinterlock ties between CEOs in a dyad. When p is signifi-cantly different from 0, standard regression techniques ap-plied to the second equation yield biased results, to the ex-tent that error terms in both equations contain somecommon omitted variables (van de Ven and van Praag,1981). For example, if firms that form interlock ties are moreresponsive to a given level of resource dependence or moresensitive to the need for trust between top managers informing alliances than those firms not forming an interlocktie, then specification error would be present. Heckman'sprocedure generates consistent, asymptotically efficient esti-mates for such models, allowing us to generalize results tothe larger sample frame (cf. Heckman, 1979),

RESULTSTable 4 provides the results of the logistic regression analy-sis of aliiance formation, and table 5 gives the Heckman se-



Table 3 (continued)

Variable

8, Prior alliance ties9. Constraint

10. Size11. Performance12. Solvency13. R&D intensity14. Advertising intensity15. Diversification16. Industry overlap17. Common appointments18. Total appointments19. Reciprocated appointments20. Friendship ties

7

.15

.10

.03-.08

.05-.10-.08-.07

.03

.03

.03-.01-.02

8

.14

.01

.03

.04-.12-.05-.06

.14

.16-.01

.04

.17

9

-.08-.04-.11

.03

.01-.09-.32

.05-.04

.19

.10

10

-.06.36

-.31-.09

.26

.18

.08

.04

.06-.04

11

.17

.14

.10-.03-.03

.01-.02

.05

.09

12

.27

.05-.12-.06

.10

.00

.04

.06

13

-.18.16.02.07.05.06.03

14

.15-.02

.00

.45-.01

.10

15

-.09.13.08.03.04

16

.16-.11-.02

.01

17

.45

.01

.06

18 19

.00-.07 .05

lection model results. The hypothesized effects are in bold.Model 1 in table 4 tests hypothesis 1, that an interlock tiebetween two firms wiil increase the likelihood of subsequentalliance formation between them. The results in model 1 donot support this hypothesis: after controlling for the extentof market constraint (i,e,, resource interdependence) be-tween firms, as well as other financial and strategic factors,the existence of an interlock tie is not significantly related tosubsequent alliance formation.

Model 2 in table 4 tests hypotheses 2 and 3, which addressthe influence of management-board relationships on subse-quent alliance formation between the focal firm and outsidedirectors' home companies. The results for board control andCEO-board cooperation shown in model 2 provide strongsupport for these hypotheses. Consistent with hypothesis 2,board control over the CEO is negatively related to the iikeii-hood of forming an alliance between the focal firm and out-side directors' home companies. The results aiso supporthypothesis 3: CEO-board cooperation is significantly andpositively related to subsequent alliance formation. The hy-pothesized effects of board control and cooperation werealso supported in Heckman selection models of alliance for-mation, as shown in model 1 of table 5.

In summary, the first set of results indicates that the merepresence of a board interlock tie between firms does notpredict the formation of strategic alliances between firms;instead, such ties may either increase or decrease the likeli-hood of alliance formation, depending on the nature of theCEO-director relationship that underlies the tie. The greaterthe extent to which an interlock tie results in cooperationbetween top managers of different firms in strategic deci-sion making (i.e., at the focal firm), the greater the likelihoodof subsequent strategic cooperation between the focal firmand the outside director's home company. At the same time,the greater the extent to which an interlock tie results in anindependent control relationship between top managers ofdifferent firms, the lower the likelihood of subsequent strate-gic cooperation between them.

The next set of results tests hypothesis 4, that third-partyties resulting from appointments of focal-firm CEOs on otherboards amplify the effects of independent board control and


Table 4

Logistic Regression Analysis of Aliiance Formation*

Independent variable

Interlock tie

Board control

CE0^3oard cooperation

Third-party ties

Third-party ties x board control

Third-party ties x CEO-boardcooperation

Prior alliance activity, firm 1


Prior alliance ties

Constraint

Size

Performance

Solvency

R&D intensity

Advertising intensity

Diversification

Industry overJap

Common appointments

Totai appointments

Reciprocated appointments

Friendship ties

Constant

NChi square

1

0.44(.047)

.006(.005)

.002"(.0005).003*

(.0007).057"

(.018).022*

(.008).002

(.004)-.013*(.005)

-.032(.021)

--323*(.161)

-.377"(.159)

-.013(.011),047*

(.017).010

(.007).002

(.002).011

(.050)

.041(,045)

73,510104.18*

• Standard errors are in parentheses. Hypothesized

2

- .391"(.149).575*

(.183).034

(.080)

.037*(.009).041"

(.009)1.678*(,552).289*

(.109).034

(,092)-.292*(,132)

-.463(.442)

-9-490(4.837)

-10.389"(4,169)-.177(.257)1.330"

(4.92).139

(.127}.053

(.036}.018

(-428}2.839"

(1.022}

1.977(1.011)

898118.44*

effects are

3

-.589*(.203).757"

(.249).098

(.080)-.199*(.074).227"

(.073).037*

(-010).040*

(.009}1.701"(.555}.290"

(.109}.045

(.092}-,288"(.132}

-.491(.438}

-8.682(4.850}

-10.507*(4,176}-,195(.256}1.356*

(4.95}.168

(.127)-051

(.036},008

(.428}2.906*

(1.029)

1.928(1.017}

898140.36*

in bold.

CEO-board cooperation on alliance formation. The interactioneffects in model 3 of table 4 support this hypothesis. Consis-tent with hypothesis 4a, the results show that as the num-ber of third-party ties between the CEO and outside direc-tors increases, the negative relationship between boardcontrol over the CEO and the likelihood of subsequent alli-ance formation between the focal firm and outside directors'home companies becomes stronger. The results also sui>port hypothesis 4b: as the number of third-party ties be-tween the CEO and outside directors increases, the positive



Table 5

Heckman Selection Models of Alliance Formation (fV= 73,510)*


Board control

CEO-board cooperation

Third-party ties


Third-party ties x CEO-board cooperation



Prior alliance ties

Constraint

Size

Performance

Solvency

R&D intensity


Diversification

Industry overlap

Common appointments

Total appointments


Friendship ties

Constant

Chi square

• p < .01.

1

-.480*(.184}.581*

(.201).142

(.114)

.034'[.010).037'

(.010)1.627'(.564).312*

(.116).044

(.096)-.342*(.138)

-.119(.460)

-10.756(5.718)

^11.085*(4.881)

.156(.267)1.330"(.492).071

(.134).050

(.040).018

(.428)2 799*(1.101)

2.446(1.077)

132.45'

* Standard errors are in parentheses. Hypothesized effects are

2

-.678'(.2661.767'

(.270).139

(.113)-.259*(.097).268*

(.092),035'

(.010).036'

(.010)1.594'(.567).311*

(.116).041

(.096)- .333'(.138)

-.424(.460)

-9.383(5.760)

-10.691 '(4.893)

.179(2.70)1.346(.494).089

(.137).052

(.040).011

(.428)2.882'(1.113)

2.386(1.081)

151.22*

in bold.

In these models, board control and CEO-board cooperation, which we assume tobe group variables, are measured at theboard ievel (Fama and Jensen, 1983);thus, effects on alliance formation shouldbe similar across dyads on the sameboard. To check this assumption, we esti-mated separate models using measuresof the individual participation of directorsin control and cooperation activities (N =412), The results for these models wereconsistent with the results in tabies 4and 5.

relationship between CEO-board cooperation and the likeli-hood of subsequent alliance formation between the focalfirm and outside directors' home companies also becomesstronger. The hypothesized interaction effects were alsosupported in Heckman selection models of alliance forma-tion, as shown in model 2 of table 5.^

Results for several of the control variables provide furtherinsights. For instance, the degree to which firms are mutu-ally constrained by resource interdependence, as indicatedby resource flows between their respective industries, ispositively associated with subsequent alliance formation,consistent with the traditional resource dependence perspec-


tive on alliance formation. While this effect has previouslyonly been observed at the interindustry level, our resultsdemonstrate that such effects for resource dependence oc-cur at the dyad level as well. Results also show that friend-ship ties between CEOs and outside directors are positivelyand significantly related to subsequent alliance formation be-tween the focal firm and outside directors' home companiesin each of the models. In contrast, common appointments toother boards are not significantly associated with allianceformation, nor are the main effects of third-party ties signifi-cant. In general the various network variables do not haveindependent effects on alliance formation; instead, networkeffects are contingent on the content of CEO-director rela-tionships.

We conducted several additional analyses to examine therobustness of our findings to alternative independent vari-ables and different samples. First, we conducted analyses ofalliance formation using the archival measures of cooperationand control discussed above. As shown in table 6, the re-sults of these analyses are very similar to the results ofanalyses using the survey measures: model 1 shows thatthe archival measure of board control is significantly andnegatively related to subsequent alliance formation, whilethe archival measure of cooperation has a significant andpositive effect on alliance formation. Model 2 shows thatthese effects are amplified by third-party ties, as hypoth-esized,

To further assess the robustness of our findings, we con-ducted separate logistic regression analyses with the fullsample of dyads by estimating the interaction effects be-tween the presence of an interlock tie and board control orcooperation. Results are shown in table 7. As shown inmodel 2, there are significant interaction effects between aninterlock tie and both cooperation and control. These resultsindicate that, while the mere presence of an interlock tie be-tween firms does not affect the likelihood of alliance forma-tion (as shown in model 1), the effect of an interlock tie be-comes significantly more positive as the level of CEO-boardcooperation increases and significantly more negative as thelevel of control increases. These results again suggest thatthe effect of board interlock ties on alliance formation is con-tingent on the nature of the CEO-board relationship. Wewould not expect significant main effects for board controlor cooperation, since these relationships should only in-crease or decrease the likelihood of alliance formation be-tween two specific firms provided an interlock tie exists be-tween them, vi/hich is not necessarily the case with thislarger dataset. We also examined interaction effects usingthe individual, continuous-scale indicators of cooperation andcontrol, to identify the level of cooperation and control atwhich interlock ties decreased the likelihood of alliance for-mation. The results of these analyses showed, for instance,that interlock ties have a negative effect on alliance forma-tion when a minority of directors have the same functionalbackground as the CEO. In addition, the results showed that


Coop^'ative or Controlling?

Table 6

Supplementary Hedcman Selection Models


Board control


CEO-board trust

Third-party ties


Third-party ties x CEO-board cooperation



Prior alliance ties

Constraint

Size

Performance

Solvency

Research and development intensity


Diversification

Industry overlap

Common appointments

Total appointments


Eriendship ties

Board approval

Ingratiation

Constant

Chi square

" p < ,01.

of Alliance Formation I7V=73,510)»

With Archival Measuresof Cooperation/Control

1

- .322'(.120).603*

(.230)

,139(.113)

.034'(.010).036*

(.010)1.566'(.565).310*

(.1161.053

(.093)- .290'(.137)

-.427(.453)

-10.729(5,724)

-10.937*(4.845)

.164(.268)1,333*(.494).071

(.133).041

(.039).022

(.427)2.804*(1.103)

.124t208).052

(.174)

2.500(1.071)

125.16'

* Standard errors are in parentheses. Hypothesized effects are in bold

2

-.339*(.123).568*

(.233)

(.167).126

(,113)- .161*(.073).268*

(.100).035*

(.010),037-

(.010)1.590*(.569).306*

(,115).047

(.094)-.332*(.138)

-.431(.460)

-9.384(5.772)

-10-625*(4.875)

.186(.268)1.351*(.497).093

(.136).051

(.040).019

(.429)2.886*(1.114)

.116(.208).053

(.074)

2.363(1.077)

139.38*

With Measuresof CEO-board Trust

3

-.252(.1851.311

(.203).605'

1.167}.141

(.114)

.035*(.010),037*

(.010)1.589*(.565).309*

(,116).045

(.094)-.339*(.137)

-.436(.463)

-10.192(5.745)

-11.677*(4.911)

.169(.270)1.358'(.500).070

(.134).050

(.040).019

(.425)2.810'(1.103)

.118(,208).057

(.077)

2.565(1.083)

130,01'

4

-.330(.266).433

(.268).599*

.139(.114)

-.173(.097).119

(.094).036*

(.010).037*

(.010)1.601*(,571).311*

(.117),049

(.092)-.330*(.139)

-.434(.462)

-9.150(5.788)

-11.392*(4.909)

.185(.270)1.380*(.503).091

(.136).051

(.040).019

(.426)2.877*(1.113)

.114(.209).062

(.074)

2.366(1.094)

148.15*

interlock ties decrease alliance formation when the CEOowns less than 1 percent of outstanding common stock.

While our theoretical argument suggests that trust in theCEO-board relationship can explain how control and coopera-tion affect alliance formation, our primary analyses did notexplicitly model the mediating effect of trust. Thus, one


Table 7

Suppl«nentary Logistic Regression73,510)*


Interlock tie

Board control


Interlock tie x board control

Interlock tie x CEO-board cooperation

Third-party ties



Prior alliance ties

Constraint

Size

Performance

Solvency

R&D intensity


Diversification

Industry overlap

Common appointments

Total appointments


Constant

Chi square

Mod^s of Alliance

1

.043(.047)

-.002{,004).002

(.008)

.006(.005).002'

(.0005).003'

(.0007).057*

(.017),02 r

(.008).002

(.004)-.012*(.005)

-.029(.021)

-^.320(.160)

- . 37T(.158)

-.013i.OU).045"

(.016).009

(.007).002

(.002).010

(.050)

.042(.046)

102.97'

Formation (/Va

2

.039(.046)

-.002(.004).003

(,008)-.409"(.125).331*

(.111).006

1.005).002*

(.0005).002"

(.0006).051"

(.016).022'

(.008).003

(.004)-.013*(.005]

-.028(.021)

-.307(.160)

-.364"(.156)

-.014(.010).042"

(.016).008

(.006).002

(.002).010

(.049)

.033(.042)

128.20'

" p < .01.* Standard errors are in parentheses. Models are analyzed for the full sampleusing archival measures of board control and CEO-board cooperation. Hypoth-esized effects are in bold.

might question whether other, related social processes me-diate these relationships. For instance, cooperation might beassociated with political influence processes such as ingratia-tion, which could affect the likelihood of alliance formationbetween the focal firm and manager-directors' home compa-nies by enhancing directors' affect toward the CEO, withoutnecessarily enhancing trust in the relationship. Similarly, co-operation could increase the board's approval of the CEO'sperformance and thus increase the likelihood of alliance for-mation independent of CEO-board trust. To assess the rela-


CoopM-ative or Controliing?

We measured trust in our survey as de-scribed in the method section above.Prior research has measured ingtatiationas a unidimensional construct composedof multiple behaviors (Jones and Wort-man, 1973: Kipnis and Schmidt, 1988),Our jngratiation measure is based on Ku-mar and Beyerlein's 11991} scale, whichtaps key aspects of ingratiation such asflattery, favor-dotng, opinion conformity,and impression management; while pnorevidence for the reliability of this scale ismixed (e,g,, Kacmar and Valle, 1997), theeight items used in the present studywere adapted to the CEO-board context,and the scale showed acceptable inter-item reliability (alpha = .86) and interraterreliability {kappa = .81}. Factor analysisclearly indicated that CEO advice andcounsel did not load on the ingratiationfactor (loadings were less than .2 for thecooperation items), suggesting that CEO-board cooperation is distinct from ingra-tiation (i.e., advice and counsel is not anaspect of ingratiation}. Moreover, the re-sults presented in table 6 were un-changed when ingratiation was measuredusing any combination of the four behav-iors assessed with the scale (e,g,, opin-ion conformity and flattery, favor-doingand flattery, etc.). Board approval wasmeasured with two survey items thatassessed the board's apparent overailsatisfaction with the CEO's leadership.

tive importance of these different social processes in ex-plaining how control and cooperation affect allianceformation, we conducted further exploratory analyses usingsurvey measures of trust, political influence (ingratiation),and board approval of the CEO.® As shown in models 3 and4 of table 6, CEO-board trust has a strong and positive rela-tionship with alliance formation, while the effects of ingratia-tion and board approval are nonsignificant, in addition, whentrust is added to the models, the effects of cooperation andcontrol become nonsignificant, suggesting that CEO-boardtrust mediates the effects of cooperation and control on alli-ance formation {Baron and Kenny, 1986). Moreover, the ef-fects of ingratiation and board approval of the CEO are insig-nificant in all models.

The results are not consistent with the view that preexistingtrust in the CEO-board relationship led to cooperation, whichthen facilitated alliance formation through some othermechanism. The findings suggest that trust nnediates theeffects of cooperation/control, and not the reverse. We alsomeasured trust using responses to the director survey, andresults were substantively unchanged from results pre-sented in table 6. While researchers have typically viewedtrust and distrust as one bipolar construct, Lewicki, McAllis-ter, and Bies (1998) recently suggested that distrust is a dis-tinct construct from trust. Thus, we conducted further analy-ses using only items that refer to distrust. Our results werenearly identical: distrust was strongly (and negatively) associ-ated with alliance formation, and the control and cooperationvariables became nonsignificant when distrust was added tothe models, suggesting that distrust mediates the effects ofcontrol and cooperation on alliance formation. Thus, even iftrust and distrust are viewed as distinct concepts, the re-sults suggest that both predict alliance formation and medi-ate the effects of cooperation and control. Moreover, theCEO-board relationship appears to satisfy several of Lewicki,McAllister, and Bies' (1998) conditions for a high (negative)correlation between trust and distrust, including high valuecongruence and interdependence between the parties.

DISCUSSION

This study shows how board interlock ties can have qualita-tively different effects on the formation of joint ventures be-tween firms depending on the behavioral processes that un-derlie CEO-board relationships. The first set of resultssuggested that the mere presence of a board interlock tiebetween two firms does not appear to increase (or de-crease) the likelihood that they will enter into a strategic alli-ance with one another. Further results showed that theseaggregate effects of board interlock ties appear to maskmore specific effects that depend on the content of the tie.Specifically, the results suggest that there can be dark andbright sides to the presence of a board interlock tie betweentwo firms depending on the underlying relationship. Higherlevels of independent board control over management actu-ally decreased the likelihood of subsequent alliance forma-tion between them, while higher levels of CEO-board coop-eration in strategic decision making raised the likelihood thatthe two firms would enter into an alliance. These results


were confirmed with both survey-based and archival mea-sures of board cooperation and control, and they hold aftercontrolling for a variety of economic and strategic variablesthat could influence alliance formation. Thus, the first set ofresults demonstrated how the consequences of board inter-lock ties for strategic cooperation depend critically on thebehavioral content of the tie.

Our theoretical framework goes beyond looking at the influ-ence of dyadic ties and their content to address whether theindirect ties in which parties to an interlock tie are embed-ded could influence interorganizationat action. The findingsare consistent with the perspective that third-party ties pri-marily amplify whatever relational dispositions already existamong directly connected actors—they not only appear toamplify trust resulting from cooperative interaction in CEO-board relationships, but they also amplify distrust resultingfrom independent board control. At the same time, such in-direct ties did not have significant main effects on allianceformation. Thus, the results appear to support the proposi-tion developed by Burt and Knez (1995) that third-party tiestend to reaffirm or amplify whichever predisposition manag-ers have toward their colleagues (see also Labianca, Brass,and Gray, 1998).

Our finding that the effects of third-party connections arestrongly contingent on the behavioral content of immediate,professional ties between top managers is not consistentwith the view that such ties uniformly enhance trust be-tween individuals by increasing the reputational costs of non-cooperative behavior (Kreps. 1990; Raub and Weesie, 1990).Third-party ties are not effective in promoting cooperationwhen noncooperative behavior is normatively acceptable inthe larger sociai structure, such that individuals do not incurreputational costs for noncooperative behavior, in our con-text, the reputation of an outside director is not necessarilydamaged by noncooperative behavior (i.e., exercising inde-pendent control over CEOs), Instead, directors may even beincreasingly rewarded for exercising independent control inthe market for corporate directors; some evidence suggeststhat directors on controlling boards gain more subsequentappointments on other boards with relatively high levels ofcontrol over management, although they may also gainfewer appointments on cooperative boards {Zajac and West-phal, 1996). As a result, in the absence of reputational costsfrom noncooperation, third-party ties do not necessarily en-force such behavior. Future research conducted in other con-texts might examine explicitly whether the effects of third-party ties on trust between individuals are contingent on thenormative status of cooperation vs. control in the larger so-cial structure.

The results on trust offer additional support for our theoreti-cal argument by providing evidence that CEO-board coopera-tion and control influence aDiance formation through theireffects on CEO-board trust. That is, the resuits are consis-tent with the view that cooperation increases the likelihoodof alliance formation by increasing trust between top manag-ers and manager-directors, while board control towers thelikelihood of alliance formation by reducing trust betweenthem. Nevertheless, these additional analyses are merely



exploratory, and further research is clearly needed that usesalternative measures of trust and sociopolitical influence toverify more conclusively the social mechanism by whichCEO-board interaction affects alliance formation. This couldbe further supplemented with studies that use alternativemeasures of cooperation and control. While this research isperhaps unique in demonstrating support for hypothesesabout CEO-board relationships with both archival and surveymeasures of key constructs, there is a great need for re-search that uses alternative approaches to measuring coop-eration, control, and other forms of CEO-board interaction.

The results for control variables included in our analysis alsoprovide some valuable insights. Looking across the logit re-gression and Heckman selection model results, we foundthat previous ties between dyad members increase the likeli-hood of the firms entering an alliance. This is consistent withresearch highlighting the importance of the social network ofprior alliances in influencing subsequent alliance formation(Kogut, Shan, and Walker, 1992; Gulati, 1995b, 1998; Powell,Koput, and Smith-Doerr, 1996; Gulati and Gargiulo, 1999).Variables indicating prior alliance activity also capture any un-observed propensities of the firms to enter into alliances thatare not captured by the independent variables and furtherattest to the robustness of our results (Heckman and Borjas,1980). We also found that some of our measures of re-source dependence were significant, indicating that resourcedependence was indeed an important consideration for thecreation of new alliances. Our measure for dyadic constraintwas positive and significant as expected: the greater theconstraint, the greater the likelihood of alliance formation.Moreover, the significant effect of friendship ties betweenCEOs and manager-directors provides further evidence thatpositive ties between CEOs and board members encouragealliance formation.

In delineating the critical role of tie content in moderatingnetwork effects and then showing how it may be moderatedfurther by third-party ties, this study makes several relatedcontributions to research on interorganizational networks.First, as several authors have noted, very little empirical re-search has examined when network ties may lead to morenegative relations between individuals or between organiza-tions (Brass and Burkhardt. 1992; Burt and Knez, 1995; Labi-anca. Brass, and Gray, 1998; Westphal and Zajac, 1997). Byexploring how the content of network ties might diminishmutual trust between individuals, thus impeding alliance for-mation, this study investigates what Burt and Knez (1995:261) called the "dark side" of social networks. This is furtherdeveloped by showing that both negative and positive tiesbetween dyads of firms are amplified by third-party connec-tions in which firms are embedded. This of course assumesthat alliances are a beneficial strategy that are somehow pre-empted by negative connections between firms.

Second, research on social networks has typically treatednetwork ties as exogenous; relatively little empirical researchhas examined the origin of organizational networks (Gulati,1998; Gulati and Gargiulo, 1999). It is important to recognizethat both interlocks (our key independent variable) and iointventures (our dependent variable) are interorganizational rela-


tionships that accumulate into a social network. This studyconsiders how the creation of interorganizational networks ofstrategic alliances may be influenced in significant ways bythe social networks of board interlocks in which firms areplaced. As a result, this study examines the multiplexity ofties in which firms are embedded and the relationship be-tween those ties: new social networks result from a socialprocess in which preexisting ties may shape their creation.This is also distinctive because these two networks exist atdifferent levels of analysis: the relationships between corpo-rate leaders that make up board interlocks are individual levelties, while strategic alliances occur across firms. Very littleempirical research has considered how social ties betweenindividuals can influence a different kind of network tie atthe interorganization level (for exceptions, see Galaskiewicz,1985b; Zaheer, McEvily, and Perrone, 1997). Thus, the find-ings have implications for cross-level perspectives on theformation of interorganizationai networks by showing howthe micro-content of network ties between individual leaderscan play an important role in the development of organiza-tion-level ties.

The findings of this study may also have important implica-tions for the corporate governance literature, as well as theliterature on interorganizational cooperation. Surprisingly littleempirical research has considered how boards of directorsmay influence strategic cooperation between firms. In thecorporate governance literature, resource dependence theo-rists have viewed interlocking directorates and joint venturesas two independent and alternative mechanisms for resolv-ing environmental uncertainty (Pfeffer and Salancik, 1978).Moreover, research on CEO-board relationships has also notconsidered how such ties may influence cooperative strate-gies. This literature has tended to examine how CEO-boardrelationships characterized by board independence frommanagement affect competitive strategy, such as diversifica-tion strategy, and the propensity for boards to disciplinemanagers for poor performance that affects shareholders(Walsh and Seward, 1990; Gibbs, 1993; Hoskisson, Johnson,and Moesel, 1994). The findings of this study suggest thatan unintended side effect of CEO-board relationships charac-terized by independent board control is to reduce the likeli-hood that firms will identify and pursue strategic alliance op-portunities with certain other companies (i.e., the homecompanies of outside directors on the board). The findingsare consistent with recent evidence that independent boardcontrol tends to politicize the CEO-board relationship,prompting CEOs to devote more time and attention to build-ing political support among directors rather than identifyingnew strategic opportunities (Westphal, 1998).

Research on alliances has primarily considered their forma-tion as resulting from critical resource contingencies and hasonly recently paid attention to some of the social structuralelements that may influence their formation (Gulati, 1998).Yet there is a growing stream of research in economic soci-ology that suggests that organizational actions are stronglyinfluenced by the social ties in which organizations are em-bedded (Gulati, 1995b; Gulati and Gargiulo, 1999). This studybuilds on the embeddedness perspective and specifies



some of the ties that may influence the alliance behavior offirms. It shows how organizational decisions, such as entryinto joint ventures, can be influenced in important ways bythe social networks of board interlocks in which organiza-tional decision makers are placed. Moreover, it demonstratesthat there may be a dark side to embeddedness that caninfluence the alliance decisions of firms.

The findings of this study suggest avenues for future re-search. One might examine, for instance, whether prior co-operative or controlling relationships between top managerson boards might affect not only the likelihood of formingstrategic alliances with specific other firms but also the sub-sequent success of those strategies {Gulati and Lawrence,1999). In an insightful study. Baker (1984) suggested thatdistinct social structural patterns in the stock options marketcan alter the direction and magnitude of option price volatil-ity. Similarly, in this instance, the social structure of boardinterlocks in which alliances are created can influence therelative terms of trade of the agreement and also dampenthe volatility that usually occurs in such partnerships. On onehand, good rapport between top managers resulting fromprior involvement in cooperative relationships on otherboards may lead to more successful alliances by facilitatingefforts to flexibly adjust the roles and responsibilities of alli-ance partners as environmental conditions change over time(Ring and Van de Ven, 1992; Doz, 1996). On the other hand,in-group biases resulting from CEO-board cooperation maylead to excessive levels of trust between top managers, sothat each party becomes overly optimistic about the capabili-ties and likely contributions of the other. Thus, empiricalresearch is clearly needed to determine how initial relation-ships between corporate leaders moderate the corise-quences of alliance formation.

Another area for future research would be to explore morethoroughly the full array of economic and organizational im-peratives that might motivate two firms to enter into a jointventure. There has been considerable work on this question,much of which has been incorporated in the control variablesincluded here, but there is still room to explore these issuesfurther. One such example is our measure of industry over-lap, which assesses the extent to which two firms in a dyadoperate in overlapping industries and captures the resultantopportunities for collaboration that may occur for two firms.This measure implicitly treats all industries as equivalent anddoes not take into account the characteristics of the indus-tries in which the overlap occurs. It would be useful to con-sider industry heterogeneity on important dimensions suchas concentration ratio and geographical dispersion of capabili-ties that could result in greater or fewer alliances. Along thesame lines, it would also be fruitful to consider the role ofother possible determinants of alliance formation, some ofwhich may even be associated with board cooperation andcontrol. Our comprehensive efforts to include a variety ofeconomic and strategic variables that have been consideredto predict alliance formation, while also using the Heckmanselection model to control for the effects of sample selec-tion bias, is a first step in this direction.


This study confirms that board interlocks can be influentialfor the creation of new strategic alliances between firms.This effect is strongly conditioned, however, by the behav-ioral processes that underlie the connection between firms,and these cohesion effects are moderated, in turn, by thethird-party ties in which they are situated. Thus, the findingsshow how the creation of alliances between firms is shapedby the social context in which they are embedded, contribut-ing to a richer, socially informed account of organizationalaction.

REFERENCES

Aiken, Michael, and Jeratd Hage1968 "Organizational interdepen-

dence and Jntraorganizationalstructure." American Socio-logical Review. 33: 912-930.

Alderfer, Clayton P.1986 "The invisible director on cor-

porate boards." Harvard Busi-ness Review. 64: 38-52-

Aldrich, John H., and Forrest D.Nelson1984 Linear Probability, Logit, and

Probit Models- Beverly Hills,CA: Sage.

Alster, Norm1986 "Strategic partners: Seeking

the right chemistry." Elec-tronic Business, May 15: 50-54.

Baker, Wayne E.1984 "The sociai structure of a na-

tional securities market,"American Journal of Sociol-ogy, 89:77&-811.

Barney, Jay B., and Mark H,Hansen1994 "Trustworthiness as a source

cf competitive advantage."Strategic Management Jour-nal, 15: 175-190.

Baron, Robert A., and David A.Kenny1986 "The moderator-mediator vari-

able distinction in social psy-chological research: Concep-tual, strategic, and statisticalconsiderations," Journal ofPersonality and Social Psy-chology, 51: 1173-1182.

Beatty, Randolph P., and EdwardJ. Zajac1994 "Top management incentives,

monitoring, and risk sharing; Astudy of executive compensa-tion, ownership and boardstructure in initial public offer-irjgs." Administrative ScienceQuarterly, 39: 313-336.

Belliveau, M. A., Charles A.O'Reiltv' III, and James B. Wade1996 "Social capital at the top: Ef-

fects of social similarity andstatus on CEO compensa-tion." Academy of Manage-ment Journal, 39: 1568-1593.

Besnier, Niko1989 "Information vi/ithholding as a

manipulative and collusivestrategy m Nuku)ae-Jae Gos-sip." Language in Society, 56:315-341,

Bollen, Kenneth A.1989 Structural Equations vwith La-

tent Variables. New York:Wiley.

Borys, Bryan, and David B.Jemison1989 "Hybrid arrangements as Stra-

tegic alliances: Theoretical is-sues in organizational combi-nations." Academy ofManagement Review, 14:234-249.

Brass, Daniel J., and Marlene E.Burkhardt1992 "Centrality and power in orga-

nisations." }n M, Nohria and R.Eccles (eds,). Networks andOrganizations: Structure, Formand Action: 191-215. Boston:Harvard Business SchoolPress.

Brewer, Marilynn1979 "In-group bias in the minima}

intergroup situation: A cogni-tive-motivational analysis,"Psychological Bulletin, 86:307-324.

Brewer, Maritynn, and RoderickM. Kramer1985 "The psychology of intergroup

attitudes and behavior," An-nual Review of Psychology,36: 219-243. Palo Alto, CA:Annual Reviews.

Burgers, William P., Charles W. L.Hill, and W. Chan Kim1993 "A theory of global strategic

alliances: The case of theglobal auto industry." Strate-gic Management Journal, 14:419-^32,

Burt, Ronald S.1983 Corporate Profits and Coopta-

tion: Networks of Market Con-straints and Directorate Ties inthe American Economy. NewYork: Academic Press,

1987 "Social contagion and innova-tion; Cohesion versus struc-tural equivalence." AmericanJournal of Sociology, 92;1287-1335,

1992 Structural Holes: The SocialStructure of Competition.Cambridge, MA: Harvard Uni-versity Press.

Burt, Ronald S., and Marc Knez1995 "Kinds of third-party effects

on trust." Rationality and Soci-ety, 7: 255-292.

1996 "Trust and third-party gossip,"In Roderick M. Kramer andTom R. Tyler (eds.l. Trust inOrganizations: 68-89. Thou-sand Oaks, CA: Sage,

Byrne, Donn, Gerald L. Clore, andPhilip Worchel1966 "The effect of economic simi-

larity-dissimilarity as determi-nants of attraction." Journal ofPersonality and Sociai Psy-chology, 4; 220-224.

Cannella, Albert A., and MichaelLubatkin1993 "Succession as a sociopolitical

process; Internal impedimentsto outsider selection." Acad-emy of Management Journal,36; 763-793.

Cox, Bruce A.1970 "What is Hopi gossip about?:

Information management andHopic factions," Man, 5: 88-98-

Creed, W. E. Douglas, andRaymond E. Miles1996 "Trust in organizations: A con-

ceptual framework linking or-ganizational forms, managerialphilosophies, and the opportu-nity costs of controls." in Ro-derick M. Kramer and Tom R.Tyler (eds.). Trust in Organiza-tions: 16-38. Thousand Oaks,CA: Sage.



Daily, Catherine M.1996 "Governance patterns in bank-

ruptcy reorganizations." Stra-tegic Management Journal,17:355-375.

David, Parthifaan, Rahul Kochhar,and Edward Levitas1998 "The effect of institutional

investors on the level and mixof CEO compensation." Acad-emy of Management Journal,41: 200-208.

Davis, Gerald F.1991 "Agents without principles?

The spread of the poison pillthrough the intercorporatenetwork." Administrative Sci-ence Quarterly, 36: 583-613.

Davis, Gerald F., and Tracy A.Thompson1994 "A sociai movement perspec-

tive on corporate control."Administrative Science Quar-terly, 39: 141-173.

Demb, Ada, and F. FriedrichNeubauer1992 The Corporate Board: Con-

fronting the Paradoxes. Qx-ford: Qxford University Press.

Doz, Yves1996 "The evolution of cooperation

in strategic alliances: Initialconditions or learning pro-cesses?" Strategic Manage-ment Journal, 17: 55-83.

Doz, Yves, Gary Hamel, and C, K.Prahalad1989 "Collaborate with your com-

petitors and win." HarvardBusiness Review, 67 (1): 133-139.

Dyer, Jeff H.1996 "Specialized supplier net-

works as a source of competi-tive advantage: Evidence fromthe auto industry," StrategicManagement Journa}, 12:271-291.

Fama, Eugene F., and MichaelJensen1983 "The separation of ovwnership

and control." Journal of Lavyand Economics, 26: 301-325.

Fenigstein, Allan1979 "Self-consciousness, self-at-

tention, and social interac-tion." Journal of Personalityand Social Psychology, 37:75-86.

Fenigstein, AUan, and Peter A.Vanable1992 "Paranoia and self-conscious-

ness." Journal of Personalityand Social Psychology, 62:129-138.

Finkelstein, Sidney, and Donald C.Hambrick1988 "Chief executive compensa-

tion: A synthesis and reconcili-ation." Strategic ManagementJournal 9: 543-558.

Fleiss, Joseph L1981 Statistical Methods for Rates

and Proportions. New York:Wiley.

Fligstein, Neil1995 "Networks of power or the

finance conception of control?Comment on Palmer, Barber,Zhou, and Soysal." AmericanSociological Review, 60: 500-503.

Forsythe, J. B.1977 "Obtaining cooperation in a

survey of business execu-tives. " Journal of MarketingResearch, 14: 370-373.

Fowler, Floyd J.1993 Survey Research Methods.

Newbury Park, CA: Sage.

Gaertner, Samuel L., JeffreyMann, John F. Dovidio, Audrey J.Murrell, and Marina Pomare1990 "How does cooperation re-

duce intergroup bias?" Journalof Personality and Social Psy-chology, 59: 692-704.

Gaertner, Samuel L., JeffreyMann, Audrey Murrell, and JohnF. Dovidio1989 "Reducing intergroup bias:

The benefits of recategoriza-tion." Journal of Personalityand Social Psychology, 57:239-249,

Gaertner, Samuel L., John F,Dovidio, Mary C. Rust, Jason A.Nier, Brenda S. Banker, ChristineM. Ward, Gary R, Mottola, andMissy Houlette1999 "Reducing intergroup bias:

Elements of intergroup coop-eration/' Journal of Personal-ity and Social Psychology, 76:388-402.

Galaskiewicz, Joseph1985a "Interorganizational rela-

tions." In Annual Review ofSociology, 11: 281-304. PaloAlto, CA: Annual Reviews.

1985b Social Organization of an Ur-ban Grants Economy: AStudy of Corporate Contribu-tions to Nonprofit Organiza-tions, New York: AcademicPress.

Gibbs, Philip A.1993 "Determinants of corporate

restructuring: The relative im-portance of corporate gover-nance, takeover threat, andfree cash flow." StrategicManagement Journal, 14: 5 1 -68.

Granovetter, Mark1985 "Economic action and social

structure: A theory of embed-dedness." American Journalof Sociology, 91: 481-510.

1992 "Problems of explanation ineconomic sociology." In N.Nohria and R. Eccles (eds.).Networks and Organizations:Structure, Form and Action.25-56. Boston: Harvard Busi-ness School Press.

Groves, Robert M., Robert B,Cialdini, and Mick P. Couper1992 "Understanding the decision

to participate in a survey."Public Opinion Quarterly, 56:475-495.

Guiati, Ranjay1995a "Familiarity breeds trust?

The Implications of repeatedties on contractual choice inalliances." Academy of Man-agement Journal, 38: 85-112.

1995b "Social structure and allianceformation patterns: A longitu-dinal analysis." Administra-tive Science Quarterly, 40:619-652.

1998 "Alliances and networks."Strategic Management Jour-nal, 19:293-317.

1999 "Network location and learn-ing: The influence of networkresources and firm capabilitieson alliance formation." Strate-gic Management Journal, 20:397^20.

Gulati, R., and M. Gargiuto1999 "Where do interorganizational

networks come from?" Ameri-can Journal of Sociology, 104:1439-1493.

Gulati, Ranjay, Tarun Khanna, andNitin Nohria1994 "Unilateral commitments and

the importance of process inalliances," Sloan ManagementReview, 35 (3): 61-69.

Gulati, R., and P, Lawrence1999 "Organizing vertical networks:

A design perspective." Un-published manuscript, J. L.Kellogg Graduate School ofManagement, NorthwesternUniversity.


Gulati, R., and H. Singh1998 "The architecture of coopera-

tion: Managing coordinationuncertainty and interdepen-dence in strategic alliances."Administrative Science Quar-terly, 43: 781-814.

Hambrick, Donald C, TheresaCho, and Ming-Jer Chen1996 'The influence of top manage-

ment team heterogeneity onfirms' competitive moves."Administrative Science Quar-terly, 41; 659-684.

Hambrick, Donald C, and PhyllisMason1984 "Upper echelons: The organi-

zation as a reflection of its topmanagers." Academy of Man-agemenl Review, 2: 193-206.

Hansen, Gary S., and CharlesW. L. Hill1991 "Are institutional investors

myopic? A time series studyof four technology-driven in-dustries/' Strategic Manage-ment Journal, 12: 1-16.

Harman, Harry H.1976 Modern Factor Analysis, 3rd

ed. Chicago: University of Cht-cago Press-

Haunschild. Pamela R.1993 "Interorganizational imitation:

The impact of interlocks oncorporate acquisition activity."Administrative Science Quar-terly, 38: 564-592.

HaviJand, John B.1977 "Gossip as competition in

Zinacsntan." Journal of Com-munication, 27: 186-191.

Heckman, James1979 "Sample selection bias as a

specification error." Econo-metrica, 47: 153-161.

Hackman, James, and George J.Borjas1980 "Does unemployment cause

future unemployment: Defini-tions, questions, and answersfrom a continuous time modelof heterogeneity and statedependence." Econometrics,47: 247-283.

Hermalin, Benjamin E., andMichael S. Weisbach1988 "The determinants of board

composition/' Rand Journal ofEconomics, 19. 589-606.

Herman, Edward S.1981 Corporate Control, Corporate

Pov /er. Nev ' York: CambridgeUniversity Press.

Hill, Charles W. L, and Scott A.Snell1988 "External control, corporate

strategy, and firm perfor-mance in research-intensiveindustries." Strategic Manage-ment Journal, 9: 577-590.

Hirsch, Paul M.1982 "Netvi/ork data versus political

accounts; The normative cul-ture of interlocking director-ates." Paper presented at theannual meeting of the Ameri-can Sociological Association,San Francisco.

Hoskisson, Robert E., Richard A.Johnson, and Douglas D. Moesel1994 "Corporate divestiture inten-

sity in restructuring firms: Ef-fects of governance, strategy,and performance." Academyof Management Journal, 37:1207-1252,

Hosmer, Stanley, and David W.Lemeshow, Jr.1989 Applied Logistic Regression.

New York: Wiley.

Jaccard, James, Robert Turrisi,and Choi K. Wan1990 Interaction Effects in Multiple

Regression. London: Sage.

Jackson, J. Edward1991 A User's Guide to Pnncipal

Components. New York:Wiley.

Jensen, Michael C, and Kevin J.Murphy1990 "Performance pay and top-

management incentives."Journal of Political Economy.98: 225-263.

Johnson, Richard A., Robert E.Hoskisson, and Michael A. Hitt1993 "Board of director involve-

ment in restructuring: The ef-fects of board versus manage-rial controls andcharacteristics." StrategicManagement Journal, 14: 33-50,

Jones, Edward E., and Camille B.Wortman1973 Ingratiation: An Attributional

Approach. Morristown, NJ:General Learning Press,

Kacmar, tC. Michelle, and MatthewValle1997 "Dimensionality of the mea-

surement of ingratiatory be-haviors in organizational set-tings." Educational andPsychological Measurement,57: 314^328.

Kaplan, Marityn R., and J. RichardHarrison1993 "Defusing the director liability

crisis: The strategic manage-ment of legal threats." Organi-zation Science, 4: 412-432.

Kesner, Idalene F., and Roy B.Johnson1990 "An investigation of the rela-

tionship betvi'een board com-position and stockholdersuits." Strategic ManagementJournal, 11:327-336.

Khanna, Tarun, R. Gulati, and N.Nohria1998 "The dynamics of learning

alliances: Competition, coop-eration, and relative scope."Strategic Management Jour-nal, 19: 193-210.

Kipnis, David, and Stuart M.Schmidt1988 "Upward-influence styles: Re-

lationship vuith performanceevaluations, salary, andstress." Administrative Sci-ence Quarterly, 33: 528-542.

Kogut, Bruce1989 "The stability of joint ven-

tures: Reciprocity and com-petitive rivalry." Journal ofIndustrial Economics, 38: 183-198.

Kogut, Bruce, Weijan Shan, andGordon Walker1992 "The make-or-cooperate deci-

sion in the context of an in-dustry network." In Nitin No-hria and Robert Ecctes (eds.).Networks and Organizations:Structure, Form, and Action:348-365. Boston: HarvardBusiness School Press.

Kramer, Roderick M.1994 "The sinister attribution error:

Paranoid cognition and collec-tive distrust in organizations."Motivation and Emotion, 18:199-230.

1996 "Divergent realities and con-vergent disappointments inthe hierarchic relation: Trustand the intuitive auditor atwork." In Roderick M. Kramerand Tom R. Tyler (eds.). Trustin Organizations: 216-245.Thousand Oaks, CA: Sage.

Kreps, David M.1990 "Corporate culture and eco-

nomic theory." In J. Alt andKenneth Shepsle (eds.). Per-spectives on Positive PoliticalEconomy: 90-143. New York:Cambridge University Press.



Kumar, Kamaiesh, and MichaelBeyerlein1991 "Construction and validation

of an instrument for measur-ing ingratiatory behaviors inorganizational settings." Jour-nal of /^piied Psychology, 76:619-627.

Labianca, Giuseppe, Daniel J.Brass, and Barbara Gray1998 "Social networks and percep-

tions of intergroup conflict:The role of negative relation-ships and third parties." Acad-emy of Management Journal,41: 55-67.

LandfS, J. Richard, and Gary G.Koch1977 "The measurement of ob-

server agreement for categori-cal data." Biometrics, 33:671-679.

Laumann, Edward O., J.Galaskiewicz, and Peter V.Marsden1978 "Community structure as in-

terorganizational linkages." InAnnual Review of Sociology,4: 455-^84. Palo Alto, CA:Annual Reviev\/s.

Lewicki, Roy J., Daniel J.McAllister, and Robert J. Bies1998 "Trust and distrust: New rela-

tionships and realities-" Acad-emy of Management Review,23: 438^58.

Lorsch, Jay W., and ElaineMaclver1989 Pawns or Potentates: The Re-

ality of America's CorporateBoards. Boston: Harvard Busi-ness School Press.

Mace, Myles L.1971 Directors: Myth and Reality.

Cambridge, MA: Harvard Busi-ness School Press.

Mayer, Roger C. James H. Davis,and F. David Schoorman1995 "An intergrative model of or-

ganizational trust." Academyof Management Review, 20.709-734.

Messick, David M., and Dianne M.Mackie1989 "Intergroup relations." In

Mark R. Rosenweig and Ly-man W. Porter (eds.). AnnualReview of Sociology, 40: 45-81. Palo Alto, CA: Annual Re-views.

Miller, Norman, and Marilynn B.Brewer1996 Intergroup Relations. Bucking-

ham, U.K.: Open UniversityPress.

Mintzberg, Henry1983 Pov er In and Around Organi-

zations. Engiewood Cliffs, NJ:Prentice-Hall.

Mizruchi, Mark S.1992 The Structure of Corporate

Political Action: Interfirm Rela-tions and Their Conse-quences. Cambridge, MA:Harvard University Press.

1996 "What do interlocks do? Ananalysis, critique, and assess-ment of research on interlock-ing directorates." In AnnualReview of Sociology, 22: 271-298. Palo Alto, CA: AnnualReviev\'S.

Mowery, David C, Joanne E.Oxiey, and Brian S. Silverman1996 "Strategic alliances and inter-

firm knov^/ledge transfer."Strategic Management Jour-nal, 17: 77-91.

Murphy, Kevin J.1986 "incentives, learning, and

compensation: A theoreticaland empirical investigation ofmanagerial labor contracts."Rand Journal of Economics,17: 59-76.

NunnaHy, Jum C.1978 Psychometric Theory. New

York: McGraw-Hill.

Oliver, Christine1990 "Determinants of interorgani-

zational relationships: Integra-tion and future directions."Academy of Management Re-view, 15: 241-265.

O'Neal, Donald E., and HowardThomas1993 "Director networks/director

selection: Keys to the strate-gic role of the board of direc-tors," Paper presented at theStrategic Management Soci-ety Conference, Chicago.

O'Reilly, Charles A., Ill, Richard C.Snyder, and Joan N. Boothe1993 "Effects of executive team

demography on organizationalchange." In George P. Huberand William H. Glick (eds.).Organizational Change andRedesign: 147-175. NewYork: Oxford University Press.

Palepu, Krishna G.1965 "Diversification strategy, profit

performance, and the entropymeasure." Strategic Manage-ment Journal, 6: 239-255.

Palmer, Donald, Roger Friedland,and Jitendra V. Singh1986 "The ties that bind: Organiza-

tional and class bases of sta-bility in a corporate interlocknetwork." American Sociologi-cal Review, 51: 781-796.

Palmer, Donald A., P. DevereauxJennings, and Xueguang Zhou1993 "Late adoption of the multidi-

visional form by large U.S.corporations: Institutional, po-litical, and economic ac-counts." Administrative Sci-ence Quarterly, 38: 100-131.

Paulson, Steven K.1976 "A theory and comparative

analysis of interorganizationaidyads." Rural Sociology, 41:311-329.

Pettigrew, Andrew M,1992 "On studying managerial

elites." Strategic Management13: 163-182.

Pfeffer, Jeffrey, and Gerald R.Salancik1978 The External Control of Orga-

nizations. New York: Harper &Row.

Powell, Walter W., Kenneth W.Koput, and Laurel Smith-Doerr1996 "Interorganizational collabora-

tion and the locus of innova-tion: Networks of learning inbiotechnology." AdministrativeScience Quarterly, 41: 116-145.

Raub, Werner, and Jeroen Weesie1990 "Reputation and efficiency in

social interactions: An ex-ample of network effects."American Journal of Sociol-ogy, 96: 626-654.

Ring, Peter M., and Andrew H.Van de Ven1992 "Structuring cooperative rela-

tionships between organiza-tions." Strategic ManagementJournal, 13:483-498.

Siegel, Sidney, and N. JohnCastellan, Jr.1988 Nonparametric Statistics for

the Behavioral Sciences, 2ded. New York: McGraw-Hill.

Simmel, Georg1964 Conflict: The Web of Group

Affiliations. New York: FreePress.

Sitkin, Sim B., and Darryl Stickel1996 "The road to hell: The dynam-

ics of distrust in an era ofquality/' In Roderick M.Kramer and Tom R. Tyler(eds.), Trust in Organizations:196-215. Thousand Oaks, CA:Sage.


Smith, Ken G., Ken A. Smith,Judy D. Olian, Henry P. Sims, Jr.,Dougias P. O'Bannon, and JudithA. Scully1994 "Top management team de-

mography and process: Therole of social integration andcommunication." Administra-tive Science Quarterly, 39:

Stinchcombe, Arthur L.1990 "Weak structural data." Con-

temporary Sociology, 19: 380-382.

Useem, Michael1982 "Classwide rationality in the

politics of managers and direc-tors of large corporations inthe United States and GreatBritain." Administrative Sci-ence Quarterly, 27: 199-226.

1984 The Inner Circle. Oxford: Ox-ford University Press.

1993 Executive Defense: Share-holder Power and CorporateReorganization. Cambridge,MA: Harvard University Press.

Van de Ven, Andrew H.1976 "On the nature, formation and

maintenance of relationsamong organizations." Acad-emy of Management Review,1: 24-36.

van de Ven, WVnand P. M. M.,and Bernard M. S. van Praag1981 "The demand for deductibles

in private health insurance."Journal of Econometrics, 17:229-252.

Wade, James B., Charles A.O'Reilly 111, and Ike Chandratatt990 "Golden parachutes: CEOs

and the exercise of social in-fluence." Administrative Sci-ence Quarterly, 35: 587-603.

Walsh, James P., and James K.Seward1990 "On the efficiency of internal

and external corporate controlmechanisms." Academy ofManagement Review, 15:421^58.

Westphal, James D.1998 "Board games: How CEOs

adapt to increases in struc-tural board independencefrom management." Adminis-trative Science Quarterly, 43:511-537.

1999 "Collaboration in the board-room: The consequences ofsocial ties in the CEO/boardrelationship." Academy ofManagement Journal, 42:7-24.

Westphai, James D., and EdwardJ. Zajac1997 "Defections from the inner

circle: Social exchange, reci-procity and the diffusion ofboard independence in U.S.corporations." AdministrativeScience Quarterly, 42: 161-183.

Whisler, Thomas J.1984 Rules of the Game: inside the

Board Room. Homewood, IL:Dow-Jones Irwin.

Williams, Katherme Y., andCharles A. O'Reilly 1111997 "The complexity of diversity:

A review of forty years of re-search." In Deborah Gruenfeidand Margaret Neale (eds.I,Research on Managing inGroups and Teams: 3-28.Thousand Oaks, CA: Sage.

Zaheer, Akbar, William McEvily,and V. Perrone1997 "Does trust matter? Exploring

the effects of interorganiza-tional and interpersonal truston performance." Organiza-tion Science, 9: 141-159.

Zajac, E. J., and J. D. Westphal1996 "Director reputation, CEQ-

board power, and the dynam-ics of board Interlocks." Ad-ministrative Science Quarterly,41: 507-529.

Zenger, Todd R., and Barbara S.Lawrence1989 "Organizational demography:

The differential effects of ageand tenure distributions ontechnical communication."Academy of ManagementJournal, 32: 353-376.


cooperative or this study examines the influence of the...

Documents