Power Imbalance, Mutual Dependence, and Constraint ... dependence.pdfpotential of resource...

Despite ubiquitous references to Pfeffer and Salancik’sclassic volume, The External Control of Organizations,resource dependence theory is more of an appealingmetaphor than a foundation for testable empiricalresearch. We argue that several ambiguities in theresource dependence model account in part for this andpropose a reformulation of resource dependence theorythat addresses these ambiguities, yields novel predictionsand findings, and reconciles them with seemingly contra-dictory empirical evidence from past studies. We identifytwo distinct theoretical dimensions of resource depen-dence, power imbalance and mutual dependence, whichin the original theory were combined in the construct ofinterdependence and yet have opposite effects on anorganization’s ability to reduce dependencies by absorb-ing sources of external constraint. Results from a study ofinterindustry mergers and acquisitions among U.S. publiccompanies in the period 1985–2000 indicate that, whilemutual dependence is a key driver of mergers and acqui-sitions, power imbalance acts as an obstacle to their for-mation. We conclude that our reformulation of theresource dependence model contributes to realizing thepotential of resource dependency as a powerful explana-tion of interorganizational action.•Resource dependence theory marked a watershed in organi-zational research by offering a unified theory of power at theorganizational level of analysis. Nearly three decades afterthe publication of Pfeffer and Salancik’s (1978) classic vol-ume, resource dependence theory is still widely cited byorganizational scholars. As of the spring of 2002, there hadbeen 2,321 citations of the book, 58 percent in the mostrecent ten years (Pfeffer and Salancik, 2003: xvi). As theseubiquitous references attest, the notions of power, depen-dence, autonomy, and constraint are inescapable in organiza-tional research. Despite the evident appeal of the resourcedependence imagery, however, “there is a limited amount ofempirical work explicitly extending and testing resourcedependence theory and its central tenets” (Pfeffer and Salan-cik, 2003: xvi). Consequently, resource dependence theoryhas acquired the status of a powerful general metaphor, butit has been marginalized as an engine for theoretical advance-ment and a basis for testable empirical research.

Why has such a foundational theoretical framework becomea ghost in organizational discourse, a lingering presence with-out empirical substance? Part of the answer may lie in ambi-guities in the resource dependence model that underminethe plausibility of some of the theory’s most distinctive pre-dictions and empirical findings. The theory’s central proposi-tion is that organizational survival hinges on the ability to pro-cure critical resources from the external environment. Toreduce uncertainty in the flow of needed resources, organiza-tions will try to restructure their dependencies with a varietyof tactics. Certain tactics are unilateral, in that they bypassthe source of constraint by reducing the interest in valuedresources, cultivating alternative sources of supply, or form-ing coalitions. Other tactics restructure dependencies by aim-ing directly at the constraining party in the relationship.Through cooptation, for instance, the dependent organization

© 2005 by Johnson Graduate School,Cornell University.0001-8392/05/5002-0167/$3.00.

•Thanks to Monica Higgins for her supportthroughout the development of thispaper. We are also indebted to MichaelHannan, Nitin Nohria, Jeffrey Pfeffer, JoelPodolny, Michael Tushman, and seminarparticipants at Harvard University, MIT,and Stanford University for their helpfulcomments on earlier drafts, as well as toDan Brass and the anonymous ASQreviewers for helping us greatly improvethe paper. Bill Simpson, Qingxia Tong, andSarah Woolverton provided excellentassistance with data collection and analy-sis. The financial support of the Divisionof Research at Harvard Business Schooland Stanford’s Graduate School of Busi-ness is gratefully acknowledged. Thispaper is dedicated to the memory ofGerald R. Salancik.

Power Imbalance,Mutual Dependence,and ConstraintAbsorption: A CloserLook at ResourceDependence Theory

Tiziana CasciaroMikolaj Jan PiskorskiHarvard University

167/Administrative Science Quarterly, 50 (2005): 167–199

#2430-ASQ V50 N2-June 2005—file: 50201-Casciaro

stabilizes the flow of valued resources by socializing mem-bers of the constraining organization or through the exchangeof other valuable goods, such as status, friendship, or infor-mation.

Whereas these tactics featured centrally in Selznick’s (1949)institutional theory and Emerson’s (1962) exchange theory,Pfeffer and Salancik (1978) uniquely theorized about organiza-tional responses to constraint that had not received explicitattention before. Of these, the most prominent is constraintabsorption. Absorbing constraint entails giving the rights tocontrol the resources that create dependencies to the depen-dent actor. Organizations can absorb constraint completelythrough mergers and acquisitions (Pfeffer, 1972; Pfeffer andSalancik, 1978). Partial constraint absorption can be achievedthrough formal long-term contracts, such as joint ventures(Pfeffer and Leong, 1977). Constraint absorption differs sig-nificantly from other responses to resource dependencies inthat it is the only tactic that gives the dependent organizationdirect control over valued resources. In contrast, with tacticslike cooptation, the more powerful organization obtainsanother valuable resource, such as a seat on the board ofdirectors of the dependent company, while continuing tomaintain direct control over resources critical to the depen-dent organization.

Despite the apparent difference between constraint absorp-tion and other methods of restructuring dependence rela-tions, resource dependence theory predicts that the associa-tion between dependence and constraint absorption isanalogous to the link between dependence and other tactics,namely, organizations characterized by a high degree ofdependence on others are more likely to absorb the sourcesof their dependence (Pfeffer, 1972; Pfeffer and Novak, 1976;Pfeffer and Salancik, 1978: 109–110). This prediction is puz-zling if one considers the motivation of the constraining partyto agree to a restructuring of the dependence relationship.Given that constraint absorption essentially gives the lesspowerful party the right to control the resource, agreeing to aconstraint absorption operation is equivalent to relinquishingone’s power and the favorable exchange conditions thataccompany it. This puzzle is further complicated by empiricaltests of resource dependence theory that have found supportfor a positive association between an organization’s depen-dence and its ability to absorb the organization that imposesthe constraint, despite the seeming absence of incentives forthe dominant organization (Pfeffer, 1972; Pfeffer and Novak,1976; Burt, 1983; Finkelstein, 1997).

We identify four sources of ambiguity in the resource depen-dence model that may account for these perplexing predic-tions and findings with respect to constraint absorption andpropose a reformulation of resource dependence theory thataddresses these ambiguities. First, the original discussion ofconstraint absorption did not clearly discriminate between thetwo dyadic power constructs that emerge from Emerson’s(1962) exchange theory, which yields two distinct theoreticaldimensions of resource dependence: power imbalance, orthe power differential between two organizations, and mutualdependence, or the sum of their dependencies. In the origi-

168/ASQ, June 2005


nal formulation of resource dependence, these were com-bined in the concept of interdependence.

Second, resource dependence is both a normative and a pos-itive theory, in which prescriptions are often confounded withpredictions. Yet what an organization should do to absorb itsconstraints and what it actually can do to absorb them oftendiffer dramatically, in that an organization’s motivation tomanage external dependencies does not necessarily coincidewith its ability to do so. A critical determinant of such abilityis the extent to which the dependence to be managed ismutual or imbalanced. Mutual dependence creates both theincentive and the ability to absorb constraint successfully.Hence, attempts to absorb constraint become increasinglysuccessful as the mutual dependence between two organiza-tions increases. Conversely, under conditions of power imbal-ance, the dependent organization is likely to be more motivat-ed but less able to absorb constraint. Thus, in contrast withthe predictions advanced in the original formulation ofresource dependence theory, power imbalance should actual-ly act as an obstacle to constraint absorption.

Third, the scope conditions of the resource dependencemodel are ambiguous. We specify the boundary conditionsfor the contrasting effects of power imbalance and mutualdependence by distinguishing among constraint absorption,other interorganizational operations aimed at restructuringdependencies, and operations aimed at using power giventhe dependence structure.

Finally, although resource dependence theory is dyadic,empirical tests of constraint absorption have largely focusedon the dependence of one actor on the other without consid-ering the reciprocal dependency. Because tests of the effectof power imbalance require simultaneous consideration ofdependence and its reciprocal in a single construct, priortests have not appropriately tested the effect of power imbal-ance on constraint absorption. In fact, as we explain theoreti-cally and show empirically, prior tests have generally cap-tured the positive effect of mutual dependence, rather thanthe effect of power imbalance, on constraint absorption. Incontrast, our empirical implementation of the hypothesesmaintains the dyadic nature of resource dependence theoryand explicitly employs the constructs of power imbalanceand mutual dependence. We provide a critical test of our pre-dictions with an empirical analysis of interindustry mergersand acquisitions among U.S. public companies during theperiod 1985–2000 using data on 468 industries constitutingthe entire American economy. In doing so, we expand con-siderably the scope and detail of prior empirical tests ofresource dependence.

A REVISED MODEL OF CONSTRAINT ABSORPTION

Power Imbalance and Mutual DependenceThe building blocks of organizational treatments of power anddependence (Thompson, 1967; Jacobs, 1974; Pfeffer andSalancik, 1978; Burt, 1983) can be traced to Emerson’s(1962) theory of power-dependence relations. In Emerson’sexchange framework, the power capability of actor j in rela-

169/ASQ, June 2005

Constraint Absorption


tion to actor i is the inverse of i’s dependence on j. In turn,dependence is a function of resource criticality and the avail-ability of alternative providers of critical resources. An actor i,therefore, is dependent upon actor j (1) in proportion to i’sneed for resources that j can provide and (2) in inverse pro-portion to the availability of alternative actors capable of pro-viding the same resources to i. Conversely, the dependenceof actor j on actor i varies (1) in proportion to j’s need forresources that i can provide and (2) inversely with the avail-ability of alternative actors capable of providing the sameresources to j.

Central to Emerson’s theory is the notion that “power is aproperty of the social relation; it is not an attribute of theactor” (Emerson, 1962: 32). This premise implies that anaccurate portrayal of power relations in a dyad calls for thesimultaneous consideration of the power capability of i inrelation to j and the power capability of j in relation to i. Thisdyadic approach to resource dependence yields two distinctdimensions of power in a dyad: power imbalance and mutualdependence. Power imbalance captures the difference in thepower of each actor over the other. Formally, this constructcan be defined as the difference between two actors’ depen-dencies, or the ratio of the power of the more powerful actorto that of the less powerful actor (Lawler and Yoon, 1996).The second dimension of dyadic power, mutual dependence,captures the existence of bilateral dependencies in the dyad,regardless of whether the two actors’ dependencies are bal-anced or imbalanced. Formally, this measure can be definedas the sum, or the average of actor i’s dependence on actor jand actor j’s dependence on actor i (Bacharach and Lawler,1981). Power imbalance and mutual dependence need to beconsidered simultaneously in order to produce a theoreticallyexhaustive portrayal of the power-dependence structure in adyad. This is because, for any value of power imbalance, apower-dependence relation can be characterized by varyinglevels of mutual dependence. Conversely, for any given levelof mutual dependence, there can be different levels of powerimbalance in the dyad.

Figure 1 illustrates this point. Given three possible levels ofdependence of each actor on the other, dyads shown inshaded boxes on the diagonal depict power-balanced relation-ships. Dyads in unshaded boxes are power-imbalanced, andthe levels of power imbalance are symmetric around thediagonal. Above the diagonal, power imbalance favors actor j;below the diagonal, power imbalance favors actor i. Althoughequal levels of power imbalance characterize many of thedyads, different levels of mutual dependence distinguishthem. For example, both Configurations 1 and 9 on the shad-ed diagonal are power-balanced, but Configuration 9 exhibitshigher mutual dependence than Configuration 1. In Configura-tion 1, actors i and j exert minimal power over each other,either because they do not depend on each other forresources critical to their survival or because both actorshave numerous alternative providers of needed resources. InConfiguration 9, both actors exert significant power overeach other because they depend on one another for criticalresources or have few alternative providers of needed

170/ASQ, June 2005


resources. Finally, the existence of mutual dependence is notrestricted to power-balanced dyads. As illustrated in Configu-rations 4 and 8, for example, two dyads can be characterizedby different levels of mutual dependence but identical levelsof power imbalance.

Figure 1 highlights two further important points. First, powerneeds to be considered dyadically by taking into accounteach actor’s dependence on the other. Accounts of powerdynamics that consider only the dependence of actor i onactor j cannot capture power imbalance, because a givenchange in i’s dependence on j can either increase ordecrease power imbalance. For example, if actor i is moredependent on actor j than j is on i, an increase in the depen-dence of i on j increases the power imbalance in the dyad.This can be exemplified by a shift from Configuration 4 toConfiguration 7. If, however, actor i is less dependent onactor j than j is on i, the same increase in actor i’s depen-dence on actor j actually reduces power imbalance in thedyad, as exemplified by a shift from Configuration 3 to Con-figuration 6.

The second important point is that changing one dimensionof dyadic power while keeping the other constant requiresaltering the dependencies of both actors on each other. Incontrast, changes in one actor’s dependence, holding theother actor’s dependency constant, bring about a change inboth power imbalance and mutual dependence (see shiftfrom Configuration 4 to Configuration 7). For this reason,analyses that consider power dynamics dyadically, butemploy individual-level measures of dependence rather thanthe dyadic measures, cannot separate the effect of changesin power imbalance from that of changes in mutual depen-dence. Power imbalance and mutual dependence determinethe structural conditions under which an actor will not onlybe motivated but will also be capable of restructuring depen-dencies by absorbing constraint.

Power Imbalance, Mutual Dependence, and theAbsorption of Constraint

Power imbalance. All off-diagonal configurations in figure 1are power-imbalanced to varying degrees. In Configuration 3,for instance, actor i is less dependent on actor j than j is on i,

171/ASQ, June 2005



Figure 1. Configurations of power imbalance and mutual dependence.

j’s Dependence on i

High (3)

Medium(2)

Low(1)

Low (1)

Configuration 7:Power imbalance: 2Mutual dependence: 4



Medium (2)




High (3)




i’s D

epen

den

ce o

n j

either because alternative exchange partners are more avail-able to i or because j provides fewer critical resources to i.Consequently, should the exchange between i and j fail, actorj would face greater uncertainty and worse exchange condi-tions than actor i because few alternative actors can provideresources that are critical to j, and the next best alternative isless desirable than i. As a result, actor i will find it easier thanj to dictate the terms of the relationship by threatening towithdraw from the exchange. The most likely result of thispower imbalance is that i will appropriate a larger portion ofthe overall benefits accruing from the exchange (Friedkin,1986; Piskorski and Casciaro, 2004). As the power imbalanceincreases, j faces increasingly undesirable exchange condi-tions and higher levels of uncertainty than i.

To obtain more favorable exchange conditions and reduceuncertainty in the procurement of needed resources, themore dependent actor in a power-imbalanced dyad willattempt to restructure its dependency by engaging in con-straint absorption operations with the power-advantagedorganization. A constraint absorption operation reducesuncertainty for the less powerful organization by granting itstable access to the resources on which it is dependent. Thedesire of the dependent organization to absorb the source ofconstraint, however, is not equivalent to its ability to do so.By stipulating a long-term contract, entering a joint venture,or, at the extreme, merging with the dependent organization,the dominant party would lose part or all of its discretion overthe allocation of its critical resource to the dependent party.For the dominant organization, this is equivalent to foregoingits bargaining power and the advantageous exchange condi-tions that accompany it (Gargiulo, 1993). The higher-powerorganization is therefore likely to resist the lower-power orga-nization’s attempt at constraint absorption. The less powerfulorganization is unlikely to overcome the resistance of thedominant organization, which is, by definition, in a betterposition to impose its will on the power-disadvantaged party.It follows that the power-disadvantaged organization’sattempts to absorb constraint are unlikely to succeed.

Although the more powerful actor is likely to resist the lesspowerful actor’s attempts at constraint absorption, it could beargued that the process is not symmetric, in that powerimbalance could prompt the more powerful organization toexercise its power over the less powerful organization andabsorb it. Similar reasoning has been put forward to explainthe process of infiltration through board interlocks (Pfefferand Salancik, 1978). There is a fundamental difference, how-ever, between infiltration and constraint absorption. The exer-cise of power through infiltration does not eradicate thestructural power imbalance between the organizations. Incontrast, the use of power through constraint absorptionundermines the more powerful actor’s advantaged position.

To illustrate this argument, consider the two possible scenar-ios that could occur if a power-advantaged organizationacquired the less powerful organization in the dyad. In thefirst scenario, the pre-merger dominant organization commitsto exchanging with the pre-merger less powerful organiza-tion, even if better choices are available from alternative

172/ASQ, June 2005


providers. Therefore, the pre-merger dominant organizationeradicates its own bargaining power to the benefit of the pre-merger dependent organization. In the second scenario, thepre-merger dominant organization chooses to exchange withthe best available partner, thus maintaining its bargainingpower. If that best available partner is outside the mergedentity, however, the pre-merger dependent organization willnot be able to procure critical resources from the pre-mergerdominant organization. Consequently, the pre-merger depen-dent organization will need to secure critical resources fromthe external environment and will therefore face the sameuncertainty that it faced before the merger. Before the merg-er, it was the responsibility of the less powerful organizationto secure these exchanges; now that the more powerfulorganization has acquired the less powerful organization, it isalso the responsibility of the more powerful organization tosecure these exchanges. To the extent that managing theuncertain exchanges of the less powerful organization iscostly, constraint absorption would entail substantial lossesfor the more powerful organization. In contrast, infiltrationthrough a board interlock would not result in such costs,because the success of the less powerful organization insecuring its resources does not directly affect the power-advantaged organization. These two scenarios suggest thatby absorbing the less powerful organization, the more power-ful organization suffers one of two sets of negative conse-quences: it either loses its bargaining power over the lesspowerful organization or it incurs the costs of managing theuncertainty faced by the less powerful organization. Forthese reasons, the dominant organization is unlikely to entera constraint absorption operation with the dependent partybecause the resources that the less powerful organizationhas to offer are not sufficiently scarce or critical to the domi-nant organization to overcome the losses it would sufferfrom a constraint absorption operation.

Given that the more powerful organization is both unwillingto absorb the less powerful organization and resists thepower-disadvantaged organization’s attempts to absorb con-straint, we hypothesize that constraint absorption operationsare unlikely under conditions of power imbalance.

Hypothesis 1: The greater the power imbalance between two orga-nizations, the lower the likelihood of constraint absorption opera-tions between them.

Mutual dependence. The effect of changes in mutualdependence on the likelihood of successful constraintabsorption are best described through a comparison of alower-mutual-dependence dyad, like Configuration 1 in figure1, with a higher-mutual-dependence dyad, like Configuration9. In both cases, power is balanced, implying that a failure ofexchange damages both actors equally, in that they faceequal uncertainty in the procurement of resources critical totheir survival. In the lower-mutual-dependence configuration,if i and j do not exchange with each other, they can still pro-cure resources from other actors on only slightly worseterms. As a consequence, when mutual dependence is low,there is no scope for negotiation between actors. If one actormakes excessive demands on the other, the target of the

173/ASQ, June 2005



excessive demands can quickly identify an alternativeexchange partner who is capable of providing similarresources.

In the higher-mutual-dependence configuration, however, ifactors i and j do not exchange with each other, they facegreater uncertainty and worse exchange conditions than doactors in the lower-mutual-dependence configuration becausethe resources they exchange are more critical to their sur-vival, and fewer alternative sources exist. Although highmutual dependence creates substantial incentives for actorsto exchange with each other, it also opens up significantscope for negotiations. If one actor, say i, makes excessivedemands on actor j, j has to enter into negotiations with ibecause j cannot easily find an alternative exchange partnerwho is capable of providing similar resources. But j can alsomake its own demands, aware that i will find it difficult tolocate an alternative exchange partner. Because actors i and jare in a power-balanced situation, no clear pattern of domina-tion will emerge, potentially leading to a lengthy and uncer-tain bargaining process (Emerson, 1962: 33–34). Becauseboth organizations depend on each other to provide criticalresources, this uncertainty has a substantial cost to both ofthem.1

The uncertainty and potential costs associated with recurrentnegotiated exchanges under conditions of mutual depen-dence suggest the use of long-term contracts such as jointventures or permanent interorganizational arrangements suchas mergers and acquisitions as tactics for both organizationsto ensure stable flows of the critical resources they can pro-vide to each other. Mergers, in particular, entirely eliminatethe need to renegotiate the terms of the exchange repeated-ly and the uncertainty about resource procurement. For thesereasons, we hypothesize a positive association betweenmutual dependence and constraint absorption operations:

Hypothesis 2: The greater the mutual dependence between twoorganizations, the higher the likelihood of constraint absorption oper-ations between them.

Power imbalance and mutual dependence. So far, we havediscussed the distinct effects of power imbalance and mutualdependence on the likelihood of constraint absorption opera-tions, but as off-diagonal configurations in figure 1 illustrate,power imbalance can exist under different levels of mutualdependence. The simultaneous presence of power imbalanceand high mutual dependence exerts two competing forceson the relationship. On the one hand, the higher-power actoris reluctant to agree to constraint absorption because doingso would eliminate its power advantage. On the other hand,the higher-power actor is still substantially dependent on thelower-power actor and is therefore motivated, to someextent, to stabilize the flow of resources provided by thepower-disadvantaged party.

When mutual dependence is high, however, power imbal-ance has an additional effect: it creates obstacles to the suc-cessful negotiation of exchanges in the dyad. Extensivemicro-sociological and psychological research documents thedifficulty of negotiated exchange under conditions of unequal

1The notion of mutual dependenceemployed in this paper has significantsimilarities with the concept of bilateraldependence associated with asset speci-ficity discussed in transaction cost eco-nomics (Williamson, 1975). But resourcedependence theory focuses on ex-antemutual dependence that arises due tostructural restrictions on exchange, whiletransaction cost economics usuallyassumes that two parties are initially inde-pendent but develop bilateral dependenceover the course of the relationship asthey invest in relationship-specific assets.Because resource dependence theoryfocuses on ex-ante mutual dependence,while transaction cost economics focuseson ex-post bilateral dependence, the twotheories are not incompatible, however,explicit comparison of their predictionsand mechanisms is beyond the scope ofthis paper.

174/ASQ, June 2005


power. In general, power imbalance reduces the frequency ofexchange among social actors by hindering conflict resolution(Lawler and Yoon, 1996). Under conditions of power imbal-ance, it is difficult for potential exchange partners to fosterthe information flow that is a precondition for the successfulnegotiation of an exchange (Giebels, De Dreu, and van deVliert, 1998). Even when information is available, higher-power parties do not attend to it as closely as lower-powerparties (Erber and Fiske, 1984; Keltner and Robinson, 1997).The lack of “domain consensus” (Levine and White, 1961)induced by such obstacles to information exchange increasesthe frequency of confrontational behaviors within unequal-power relationships, such as the tendency to make moredemands of, make fewer concessions to, and use coercivetactics against the potential exchange partner (Lawler andBacharach, 1987). Those with power advantages tend toargue for agreements that favor themselves, whereas disad-vantaged actors tend to argue for agreements that equalizebenefits. Unequal power thus introduces issues of legitimacyand fairness concerning the distribution of payoffs from theexchange, complicating the bargaining agenda and divertingattention from the structural dependencies the exchange issupposed to address (Lawler and Yoon, 1996). Taken togeth-er, these arguments suggest that actors are less likely todevelop mutually satisfactory exchange relationships underconditions of unequal power (Bacharach and Lawler, 1981;Hegtvedt and Cook, 1987; Lawler and Yoon, 1996).

When mutual dependence is low, the obstacles to negotiatedexchange generated by power imbalance are of little concernto the power-advantaged actor, who does not depend on thepower-disadvantaged organization for critical resources easilyprocured from alternatives suppliers. When mutual depen-dence is high, by contrast, the obstacles to negotiatedexchange induced by power imbalance expose both actors,not just the power-disadvantaged party, to losses ofexchange benefits and the risk of failing to come to an agree-ment at all. These potential costs encourage both exchangepartners to stipulate long-term contracts or to enter into per-manent organizational arrangements to avoid repeated expo-sure to problematic negotiated exchanges. But the confronta-tional bargaining environment that the exchange partners aretrying to avoid is precisely what undermines their ability tosuccessfully negotiate mutually satisfactory long-term agree-ments. As a consequence, the increased incentive to stabi-lize the relationship generated by the confluence of powerimbalance and mutual dependence may be insufficient to off-set the decreased likelihood that the trading partners will beable to engage each other productively. In fact, the obstaclesto negotiated exchange created by power imbalance may beparticularly severe under conditions of mutual dependence,when both actors aspire to appropriate the surplus from theexchange, and the power-disadvantaged actor becomestherefore particularly intolerant of the demands of the higher-power actor.

Thus, a priori, there are no theoretical bases to predictwhether the countervailing forces induced by power imbal-ance and mutual dependence will result in a positive, nega-

175/ASQ, June 2005



tive, or insignificant interaction effect between the twodimensions of resource dependence and the likelihood ofsuccessful absorption of constraint. We tackle this questionempirically in the context of our study.

Comparison with the Original Constraint AbsorptionHypothesis

Figure 2 summarizes the main similarities with and differ-ences between our model and the original specification ofthe constraint-absorption hypothesis proposed by Pfeffer andSalancik (1978). Both begin with the same assumption: thepower of actor i over actor j is the inverse of j’s dependenceon resources provided by i, and vice versa. On the basis ofthis definition, both models then proceed in largely similarfashion to treat power as a dyadic phenomenon. Our con-struct of power imbalance mirrors the original discussion ofpower asymmetry, which arises when “the dependent orga-nization lacks sources of countervailing power to control theattempts at influence of the power-advantaged organization”(Pfeffer and Salancik, 1978: 52–53). Although Pfeffer andSalancik did not explicitly address the construct of mutualdependence, their discussion of balanced dependencies cap-tures the possibility that power-balanced dyads can be char-acterized by different levels of mutual dependence (such asConfigurations 1, 5, and 9 in figure 1). Because they consid-ered mutual dependence only in relation to power-balanceddyads, however, the coexistence of power imbalance andmutual dependence in a dyad did not receive theoretical con-sideration.

Beyond the definitional similarities, the two models divergein numerous ways. First, they differ significantly in their appli-cation of the dyadic power constructs to the discussion ofconstraint absorption. The model proposed here examinesthe relationships between each of two dimensions of dyadic

176/ASQ, June 2005


Figure 2. Comparison of original and revised constraint-absorption hypothesis.

X

Power of actor i over actor j is the inverse of j’s dependence on i

Power imbalance

Mutual dependence

Effect of power imbalance onthe likelihood of successfulconstraint absorption

Effect of mutual dependence onthe likelihood of successfulconstraint absorption

Power asymmetry

Mutual dependence

Original specification

Assumptions

Yes

Constructs

Integrated into a singleconstruct, interdependence

Hypotheses

Positive

Positive

Empirical implementation

No

No

Revised specification

Yes

Yes

Yes

Negative

Positive

Yes

Yes

power and constraint absorption separately. In contrast, inthe original discussion of the constraint absorption hypothe-sis, there is no explicit distinction between power imbalanceand mutual dependence. Instead, the theory revolves aroundthe notion of interdependence, which, in Pfeffer and Salan-cik’s (1978: 41) words, is “not necessarily symmetric or bal-anced” but, rather, “can be asymmetric.” Because the con-cept of interdependence encompasses both asymmetric andsymmetric forms of dependence, the two dimensions ofdyadic power, mutual dependence and power imbalance, areeffectively conflated in a single construct.

The differences in the conceptualization of the main theoreti-cal constructs translate into different empirical predictions.We hypothesize that constraint absorption is negatively relat-ed to power imbalance but positively related to mutualdependence. In contrast, the original specification of the con-straint absorption hypothesis examined constraint absorption“as a correlate of organizational interdependence” (Pfeffer,1972: 387), without distinguishing between its constituentparts, power imbalance and mutual dependence. Forinstance, Pfeffer and Salancik (1978: 115–116) predicted, “Iforganizations merge to control interdependence, then theyshould acquire organizations in areas with which theyexchange resources. Moreover, they should make suchacquisitions more often when the exchanges are problemat-ic.” Similarly, in a study of joint ventures, Pfeffer and Novak(1976: 403) hypothesized, “Patterns of joint venture activitywill correspond to patterns of transactions interdependence;to the extent an organization in industry A is more interde-pendent with organizations in industry B, a higher proportionof its joint venture activities should be with firms in thatindustry.” In this view, interdependence, whether symmetricor asymmetric, creates uncertainty, and organizations man-age this uncertainty by increasing their coordination andmutual control through constraint absorption. That is, in theoriginal specification, the existence of either power imbal-ance or mutual dependence was hypothesized to increasethe likelihood of constraint absorption. Therefore, althoughboth specifications agree that higher levels of mutual depen-dence lead to a greater likelihood of constraint absorption,they offer competing predictions with respect to the effect ofpower imbalance.

The combination of power imbalance and mutual depen-dence within a single construct is not the only reason for thisdiscrepancy in predictions about the effect of power imbal-ance on constraint absorption. There is another fundamentaldifference between the two formulations of the resourcedependence model. Whereas the revised resource depen-dence model offers a purely descriptive account of organiza-tional responses to resource dependencies, the original spec-ification is both a positive and a normative theory. It haselements of a descriptive theory of power relations; however,in its predictions and empirical implementation, it offers a setof prescriptive statements focusing on the motivations of theless powerful actor, without considering the motivations ofthe more powerful one. From a prescriptive viewpoint, theless powerful organization should absorb the constraint in

177/ASQ, June 2005



order to reduce the uncertainty in procuring neededresources, but from a descriptive point of view that considersthe motivations of both actors in the dyad, the same increasein power imbalance should actually reduce the incidence ofsuch actions.

The focus on the motivations of the less powerful actor, with-out considering those of the more powerful one, is particular-ly salient in the empirical tests of the original constraintabsorption hypothesis. For example, Burt (1980: 919–920)concluded from his analysis of constraint absorption at theindustry level, “If firms in industry j suffer a constraint totheir structural autonomy from firms in industry i .|.|. the oddsof a significant merger relation .|.|. more than double. .|.|. Iffirms in industry j suffer negligible constraint from firms inindustry i, the odds are nine to one there will be no signifi-cant merger relations from j to i.” Though the extent of j’sdependence on i was analyzed in detail, no mention wasmade of the extent to which i is constrained by j. The analy-sis revolved entirely around the focal actor j, under the implic-it assumption that j’s motivation to manage its dependenceon i would be sufficient to induce a disproportionate numberof mergers and acquisitions between industry j and industry i.Other tests of merger as a response to constraint have simi-larly focused on j’s dependence on i, without taking into con-sideration the extent to which i is in turn dependent on j(Pfeffer, 1972; Finkelstein, 1997).

Such an empirical approach embodies a substantial discon-nect between the theoretical framework and the empiricalanalyses. The theoretical framework suggested a relationshipbetween dyadic interdependence, comprising both powerimbalance and mutual dependence, and constraint absorp-tion. The empirical analyses, however, considered only thedependence of actor j on actor i, without considering the reci-procal dependence. As a consequence, these analyses didnot test the effect of either mutual dependence or powerimbalance. At first sight, the finding that the dependence ofactor j on actor i is positively related to constraint absorptioncould be taken as evidence supporting the hypothesis thatpower imbalance is positively related to constraint absorp-tion, yet dependence and power imbalance are two distinctconcepts. Power imbalance requires both the dependence ofactor j on actor i and the dependence of i on j. As weexplained above, the effect of power imbalance cannot beestimated using only the dependence of one actor on theother without considering its reciprocal.

So what does the positive relationship between the depen-dence of actor j on actor i and constraint absorption actuallytest? Under a set of reasonable assumptions, it can beshown that the positive effect of one actor’s dependence onanother actually attests to the positive effect of mutualdependence, rather than power imbalance, on the likelihoodof constraint absorption. This counterintuitive claim can beeasily understood by considering the relationship betweenthe dependence of one actor on another and the two dimen-sions of dyadic power, as outlined in figure 1. First, anincrease in dependence of one firm, say i, on another, j, caneither increase or decrease power imbalance, depending on

178/ASQ, June 2005


j’s dependence on i. Because this increase has both a posi-tive and a negative effect, the two effects essentially cancelout in a regression. Second, an increase in dependence offirm i on firm j, holding constant the dependence of firm j onfirm i, always increases mutual dependence. Because theeffect on power imbalance of changes in i’s dependence on jis canceled out, but their effect on mutual dependence isalways present, the estimated effect of changes in i’s depen-dence on j is the effect of mutual dependence. A detailedexplication is provided in Appendix A.

Scope Conditions

Resource dependence theory provides predictions about awide range of actions that organizations take in response toresource dependencies. These actions can be classified intwo broad categories: power use operations, which exercisethe power that accrues to an organization given the extantpower-dependence structure, and power restructuring opera-tions, which aim at changing the power-dependence struc-ture. Restructuring operations can, in turn, be unilateral orbilateral. Unilateral restructuring operations aim at changingthe power-dependence structure by acting on elements out-side the focal dyadic relationship, namely, reducing the inter-est in a given resource, cultivating alternative providers of theresource of interest, or forming coalitions. In contrast, bilater-al restructuring operations restructure dependencies by aim-ing directly at the other party in the dyad. Bilateral restructur-ing operations can take two forms: cooptation and constraintabsorption. Below, we discuss the mechanisms throughwhich power imbalance and mutual dependence affect thelikelihood of different interorganizational operations. This dis-cussion points to the uniqueness of constraint absorptionamong all resource dependence predictions.

Power use. Resource dependence accounts of power use ininterorganizational relations have broadly posited that powerimbalance enables the dominant actor to influence thepower-disadvantaged actor and to extract a higher share ofthe exchange surplus. Empirical studies, most of whichexplicitly measure power imbalance, have provided broadempirical support for this hypothesis (Van de Ven, Delbecq,and Koenig, 1976; Pfeffer and Leong, 1977; Provan, Beyer,and Kruytbosch, 1980; Burt, 1983). Additional evidence forthe power-use hypothesis has also been provided by studiesof certain types of interorganizational relations, such as cor-porate board interlocks. Specifically, the infiltration model ofboard interlocks suggests that dominant organizations arelikely to seek seats on boards of less powerful organizations,with the intent of influencing the decision making of thedependent organizations to the dominant organization’s bene-fit (Pfeffer and Salancik, 1978: 164–165).

This brief examination of the effect of power imbalance onpower use provides an important contrast to the revised con-straint absorption model. First, whereas an increase in powerimbalance increases the likelihood of establishing an interor-ganizational relation motivated by power use, it decreasesthe likelihood of relations motivated by constraint absorption.The second point of difference lies in the empirical imple-

179/ASQ, June 2005



mentation of the key constructs of dyadic power. Most testsof the power-use hypothesis have used the construct ofpower imbalance. In contrast, existing empirical tests of theconstraint absorption hypothesis have only focused on thedependence of actor i on actor j without considering the reci-procal.

Although power imbalance has received substantial attention,very few studies have attempted to specify theoretically andtest empirically the effect of mutual dependence on poweruse. Perhaps the only exception to this general pattern isMizruchi (1989), who simultaneously considered both powerimbalance and mutual dependence in the context of intercor-porate influence. Mizruchi hypothesized that in a power-imbalanced dyad, the dominant organization is likely to useits position to influence the behavior of the less powerfulorganization. Similarly, in a mutual dependence situation, bothorganizations can use their positions to shape the behavior oftheir counterpart. This suggests that patterns of influenceshould be positively related to both power imbalance andmutual dependence. Mizruchi’s findings supported theseclaims in the context of contributions to political action com-mittees. This analysis suggests a substantial overlapbetween constraint absorption and power use with regard tomutual dependence: in both cases, the effect of mutualdependence on the likelihood of interorganizational action ispositive.

Unilateral restructuring operations. Because dyadic depen-dencies can result in substantial interorganizational influence,organizations are likely to attempt to increase their autonomyby restructuring their dependencies (Emerson, 1962; Blau,1964). One set of possible actions involves unilateral opera-tions. The overarching hypothesis relating power imbalanceto the likelihood of restructuring power imbalances by unilat-eral means again contrasts with the revised constraint-absorption model. Constraint-absorption operations requirethe consent of both parties to restructure the dependencies.Because the more powerful party is likely to resist power-restructuring operations, power imbalance should be nega-tively related to constraint absorption. Unilateral restructuringoperations, by contrast, do not require the consent of themore powerful party, thus the less powerful party can suc-cessfully engage in such operations. Of course, the morepowerful party can respond to these actions by pursuing thesame strategies vis-à-vis the less powerful actor, therebyneutralizing the latter’s attempts to balance the relationship.But because the power-advantaged organization cannotdirectly prevent the power-disadvantaged organization fromseeking to restructure the dependency, power imbalanceshould still have a positive effect on the likelihood of under-taking such operations. Similar reasoning applies to the effectof mutual dependence. To the extent that mutual depen-dence constrains both firms, it can be expected that both willwant to engage in unilateral restructuring operations. Conse-quently, as with constraint absorption, the likelihood of unilat-eral restructuring of dependencies should increase withmutual dependence.

180/ASQ, June 2005


Bilateral restructuring operations. Although unilateralrestructuring operations allow organizations to change theirdependencies without involving the other party, there can bestrong environmental limits on the extent to which they canchange their interest in a resource, cultivate alternativesources of supply, or organize a coalition. As a consequence,organizations will often have to engage in bilateral restructur-ing of dependencies. Bilateral operations can be differentiat-ed into cooptation and constraint absorption. Although bothare bilateral, power imbalance has opposite predicted effectson the likelihood of cooptation and constraint absorption. Incooptation, the less powerful actor offers the more powerfulactor another resource of value, such as information, friend-ship, or status, in the hope that the more powerful actor willchoose not to exercise its power (Emerson, 1962: 39). Themore powerful organization is likely to acquiesce to suchcooptation, as long as the value of the additional resourceoffered by the less powerful actor compensates it for notusing its power. In contrast to cooptation, constraint absorp-tion is not advantageous to the dominant organization in apower-imbalanced dyad. Thus power imbalance has oppositeeffects on the likelihood of cooptation and constraint absorp-tion. As with unilateral restructuring operations, studies ofcooptation have not explicitly theorized about or empiricallytested the relationship between mutual dependence andcooptation.

Our comparison of the revised constraint-absorption hypothe-sis with other predictions stemming from resource depen-dence theory reveals two important points. First, the revisedconstraint-absorption hypothesis makes predictions regardingthe effect of power imbalance that are directly at odds withthe other predictions of resource dependence theory. Ourmodel leads us to hypothesize a negative relationshipbetween power imbalance and constraint absorption, where-as prior work predicts a positive relationship between powerimbalance and the likelihood of power use and other power-restructuring operations. Second, in contrast to power imbal-ance, the effect of mutual dependence on constraint absorp-tion is identical to its effect on other phenomena of interestto resource dependence theorists. Although prior work hasvery seldom explicitly considered the effect of mutual depen-dence, simple extensions of resource dependence theorysuggest that the effect of mutual dependence is uniformlypositive across power use, unilateral restructuring operations,and both types of bilateral restructuring operations. Overall,this comparison of power use and unilateral and bilateralmethods of restructuring dependence provides a strong setof boundary conditions for the negative effect of powerimbalance advocated here and allows us to reconcile pasttheoretical treatments and empirical studies of resourcedependence with the revised model we propose.

METHODS

Setting, Data, and SampleMergers and acquisitions have been among the primarydomains of interorganizational action to be analyzed from aresource dependence perspective (Perrow, 1970; Pfeffer,

181/ASQ, June 2005



1972; Burt, 1983; Galbraith and Stiles, 1984; Palmer et al.,1995; Finkelstein, 1997). Mergers and acquisitions constitutethe purest form of constraint absorption. Through merger oracquisition (M&A), two organizations lose their status as dis-tinct social actors, and the rights to control resources thatgenerate dependencies are transferred to the merged entity.For this reason, mergers and acquisitions represent an idealdomain in which to test our revised model of constraintabsorption. To provide a direct comparison with the originalformulation of resource dependence theory, we tested ourhypotheses following the tradition established in pastresearch, which investigated the effect of resource depen-dence on merger formation at the industry level of analysis(Pfeffer, 1972; Burt, 1983; Finkelstein, 1997). These studiesassociated industry-level patterns of input-output transactionsto the incidence of mergers and acquisitions among firmsoperating within different industries in an economic system.

We drew the M&A data from Securities Data Corporation’s(SDC) Mergers and Acquisitions database. We searched thedatabase for mergers formed among all companies traded onthe U.S. stock exchange in every year between 1985 and2000. The search resulted in 8,249 reported deals. Of these,3,686 were internal operations such as self-tenders andrepurchases and thus were excluded from the analysis. Wealso excluded mergers and acquisitions involving companiesoperating within the same industry because the competitivedynamics that contribute specifically to the formation of intra-industry mergers may produce spurious results in tests ofresource dependence (Pfeffer, 1972; Finkelstein, 1997). Thefinal sample consisted of 1,907 interindustry deals, 61 ofwhich were hostile takeovers.

We relied on Burt’s (1980, 1983) seminal formulation of con-straint to operationalize the notion of dependence betweenfirms in different industries based on input-output patterns oftransactions across economic sectors. Data on interindustrytransaction patterns came from the Benchmark Input-Output(I-O) accounts for the U.S. economy developed by the Bureauof Economic Analysis (BEA). To insure a fine-grained defini-tion of industry boundaries, and thus reduce the incidence ofintra-industry deals in our analysis, we defined industriesusing six-digit I-O codes. Because SDC classifies M&A dealsby the four-digit Standard Industrial Classification (SIC) codesof the companies involved, we matched SIC codes to I-Ocodes based on matching criteria provided by the Bureau ofEconomic Analysis in its Survey of Current Business. The useof six-digit I-O codes allowed us to identify 468 distinct indus-tries, compared with the 51-industry classifications used inthe most fine-grained past study (Finkelstein, 1997). We gath-ered data on industry concentration in manufacturing fromthe concentration ratios published by the Census Bureau ofthe U.S. Department of Commerce. With the BEA’s Survey ofCurrent Business mapping of SIC categories to input-outputsectors, we identified the four largest firms in each sector,summed their sales, and divided the sum by the total volumeof sales for the sector reported in the input-output table.Both the Bureau of Economic Analysis and the CensusBureau release I-O accounts and concentration ratios every

182/ASQ, June 2005


five years. To obtain annual measures of exchanges betweenindustries for the period 1985–2000, we linearly extrapolatedthe measures over the four available accounts for 1982,1987, 1992, and 1997. Given that the input-output and con-centration data change very little over any five-year period(Burt, 1983), differences in the assumptions about the extrap-olation did not significantly affect either the annual measuresor the regression results. In addition, concentration ratios foragriculture, mining, construction, government enterprises,and special industries were not available in all time periods.To avoid a sizable loss of observations, we linearly extrapolat-ed the available concentration ratios across the missing timeperiods. Our findings did not significantly change whether weincluded or excluded these extrapolated observations.

Finally, we obtained historical data on the financial character-istics of these industries, such as size and debt structure,from COMPUSTAT by constructing industry averages of firm-level data on all U.S. public companies operating in the indus-tries of interest. The financial information available fromCOMPUSTAT to measure these variables is specified at anSIC level that does not always correspond directly to six-digitI-O codes. In all ambiguous cases, we converted SIC codesto I-O codes according to a variety of criteria to assess thesensitivity of our results to different approaches to variableconstruction. The results proved to be robust across all mea-sures. Another issue was the time structure of the financialdata. The historical file in COMPUSTAT provides financialinformation starting in 1987, so we collected these data on ayearly basis for the period beginning in 1987 and ending in2000. To calculate financial measures for 1985 and 1986, welinearly extrapolated the data available for 1987–2000. Aswith input-output and concentration data, performing theextrapolation using varying assumptions did not change theresults. Because of the intrinsically discretionary componentof extrapolation, however, we also ran all regressions exclud-ing 1985 and 1986. The results were not affected by thisalternative time structure of the data set. Given the robust-ness of the findings, we chose to report results for the entire1985–2000 period, to capitalize on all the available data.

Measures

Dependent variable. The dependent variable was mergerfrequency, defined as the proportion of the total number ofmergers and acquisitions formed by business units in indus-try i that involved business units in industry j. Business unitsin industry i and industry j were, respectively, M&A acquirersand targets.

Independent variables. The general concept of dependenceas the result of an interaction between resource criticalityand the existence of alternative providers of such resourceshas a direct manifestation at the level of a business unit in anindustry (Burt, 1982, 1983). The measures of dependence ofbusiness units in industry i on business units in industry jwere constructed in a three-step process. We began withmeasures of interindustry flows, zij, expressed as the totaldollar value of goods and services sold by industry i to indus-try j. Subsequently, we derived dependence of industry i on

183/ASQ, June 2005



industry j, which is high to the extent that industry i sells asignificant proportion of its goods and services to industry j,sij, or it buys a significant proportion of its goods and servicesfrom industry j, pij. To convert the measure of dependence ofindustry i on industry j to dependence of business units inindustry i on business units in industry j, we multiplied thedependence measure by four-firm concentration ratios inindustry j, Oj. Following Burt (1983), we formally define thismeasure of dependence of business units in industry i onbusiness units in industry j, as Cj→i:

2

Cj→i = (pij + sij) Oj

zji zijwhere pij = (Σq

zqi) and sij = (Σq

ziq)This measure is consistent with the notion that dependenceis determined by the joint effect of motivational investmentin an exchange and availability of alternatives, and not byobserved patterns of exchange (Emerson, 1962; Marsden,1983). This argument implies that the use of industry-leveldata to measure the dependence of one business unit onanother has sounder theoretical bases than the use of firm-to-firm transactions. Consider the following scenario involvingindustry i and industry j. Industry i purchases 30 percent ofits inputs from industry j, while industry j purchases none ofits inputs from industry i. Only two business units operate ineach of the two industries. Business units A and B operate inindustry i, and business units C and D operate in industry j.Business unit A purchases 30 percent of its inputs from busi-ness unit C and 0 percent from D, while B purchases 10 per-cent of its input from C and 20 percent of its input from D. Toinfer from this specific allocation of purchases and salesbetween firm-dyads that business unit A is more dependenton business unit C than on D, or that B is less dependent onC than A, is wrong. All that matters for the definition of A’sand B’s dependence structure is that both business unitsneed 30 percent of their input from industry j and that onlytwo firms, C and D, can provide those inputs. In spite of theirdifferent patterns of purchases from C and D, therefore,there is no difference in the degree of dependence of firms Aand B on C and D. Put differently, absorbing C or D wouldhave exactly the same effect on A’s and B’s dependencestructure. Pfeffer (1987: 44) captured this notion when henoted that “resource interdependence exists and is definedprimarily in terms of intersectoral, rather than interfirm, trans-actions.”3

The unit of analysis for this measure of constraint is individ-ual business units in an industry. When the unit of analysis isshifted to a dyad of business units in industries i and j, thedyad can be characterized by two constraint measures Cj→iand Ci→j, defined as:

Cj→i = (pij + sij) Oj

2We also ran our analyses using the con-straint measure defined as Cj→i = (pij

2 +sij

2) Oj. Burt (1982) used this functionalform in his analysis of the relationshipbetween dependence and profit. Ourresults are insensitive to the functionalform employed.

3In contrast, constraint absorption existsand is defined at the level of interfirmtransactions. Consider the example of theautomotive industry and its dependenceon the tire industry. The fact that tiremanufacturers and auto manufacturersmight merge to manage their depen-dence on each other only implies thatthese merged companies have success-fully restructured their own dependenceon a given industry at the firm level. Thisdoes not change the fact that, at theindustry level, tires are still needed tomake automobiles and therefore businessunits in the automotive and tire industrieswill have to transact with one another,with these transaction patterns definingthe underlying dependence structurebetween occupants of the tire and auto-motive industries. In our model, that is,the power-dependence structure isexogenous, and merger behavior isendogenous.

184/ASQ, June 2005


Ci→j = (pji + sji) Oi

Because the measure of constraint is directional, the con-straint of a business unit in industry i on a business unit inindustry j does not have to be the same as the constraint of abusiness unit in industry j on a business unit in industry i.Based on the directional measures of constraint, we con-structed a dyadic measure of power imbalance betweenbusiness units in industry i and business units in industry j,PIi↔j, as:

PIi↔j = Cj→i – Ci→j

We used the absolute value of the constraint differencebecause, according to our model of resource dependence,the direction of power imbalance is theoretically inconse-quential for the hypothesized effect of power imbalance onthe likelihood of mergers. Because this measure of constraintis not distributed normally, we used Stata’s lnskew0 function,which computes the natural logarithm of the original variablechoosing the exponent so that the skewness of the trans-formed variable equals zero. An alternative measure of powerimbalance is the ratio of the directional measures of con-straints. In this context, however, the use of the ratio wasprecluded by the large number of cases in which either Cj→ior Ci→j equaled zero. Adding a constant to the ratio variable toaddress this issue would have defeated the purpose,because doing so removes the ratio character of the mea-sure.

The two measures of constraint, Cj→i and Ci→j, can also cap-ture the mutual dependence between business units inindustries i and j. Business units in industry i and j are mutu-ally dependent to the extent that the constraint of businessunits in industry i on business units in industry j is high andsimultaneously the constraint of business units in industry jon business units in industry i is high. Formally, this mutualdependence is measured as:4

MDi↔j = Cj→i + Ci→j

As with power imbalance, mutual dependence, as we defineit, is not distributed normally. To address this issue, we againused Stata’s lnskew0 function to obtain a zero-skewnesstransformed variable.

Control variables. The bulk of the literature on mergers andacquisitions has focused on the price of purchase or on post-acquisition performance as primary objects of inquiry (e.g.,Hayward and Hambrick, 1997; Haleblian and Finkelstein,1999). The few studies that have investigated theantecedents of M&A formation have generally done so at thefirm level of analysis (e.g., Haunschild, 1993). Of the factorsemerging from this literature as explanations for M&A forma-tion, three general classes of firm characteristics can be ana-

4Although the additive specification ofmutual dependence flows directly fromthe theoretical framework advanced here,it may lead us to conclude that mutualdependence (MD) exists in a dyad, whenin reality it does not. For example, if adyad is characterized by Cj→i = 0 and Ci→j= 0.2, the additive specification of mutualdependence yields MD = 0.2. Yet it maybe argued that because one of the actorsis not dependent on the other, such adyad should be properly classified as fea-turing no mutual dependence. To accountfor this possibility, we also constructedMDi↔j = Cj→i * Ci→j. According to thisspecification, mutual dependence isgreater than zero only if both Cj→i and Ci→jare larger than zero. The empirical resultswere not affected by this change in func-tional form.

185/ASQ, June 2005



lyzed meaningfully at the industry level of analysis: profitabili-ty, size, and debt level. Industry profitability is an alternativeexplanation for the formation of mergers to the extent thatfirms in profitable industries are likely to be attractive acquisi-tion targets. We controlled for the profitability of both indus-tries involved in the merger. To operationalize industry prof-itability, we followed Collins and Preston (1968), whointroduced the idea of measuring average profits within anindustry as a price-cost margin for the purpose of makingcomparisons across industries. The dollar total of sales bybusiness units in an industry is the value of shipments for theindustry (VSi). This figure corresponds to the sum of the ele-ments in one row of an input-output table. The sum of ele-ments in a column of the table is the total cost of producingthe commodities sold to obtain the value of shipments. Thedifference between the value of an industry’s shipments andthe sum of direct costs, including materials, supplies, fuel,electric energy, cost of re-sales, and contract work done byothers, is the industry’s value added (VAi). This is the amountof income above and beyond the direct costs that businessunits in the industry obtained from the sales they made.Money paid to employees as wages and salaries for labor isconsidered part of the value added by an industry (Li). Thetotal net income for an industry can thus be computed as thedifference between dollars of value added and dollars givento labor costs. The price-cost margin for industry i (pcmi) isthe ratio of this net income over the value of shipments asthe gross industry income:

pcmi =VAi – Li

VSi

The industry price-cost margin is therefore a measure of theproportion of an average dollar of sales that can be treated asprofit.

Industry profitability is the only alternative explanation forM&A formation examined in previous resource dependencestudies of mergers and acquisitions (Pfeffer, 1972; Finkel-stein, 1997). We included additional controls for factorsknown or expected to affect acquisition activity, namely,acquirer’s free cash flow, acquirer-target size difference, andtarget’s debt. The expected impact of free cash flow on M&Aactivity is based on Jensen’s (1987) agency theory, accordingto which managers have incentives to invest excess cashflow in negative net present value operations, instead of pay-ing dividends to shareholders. We measured industry i’s freecash flow as the average ratio of debt (specifically, long-termdebt plus debt in current liabilities) to equity across businessunits in the industry, weighted by industry assets. As for size,acquiring companies tend to be larger than target companies.We controlled for size as the difference between industry i’sand industry j’s average total assets. Finally, because heavilyindebted firms tend to be undesirable acquisition targets, wecontrolled for the target’s indebtedness with industry j’s aver-age debt (long-term and short-term) over total assets. Toaccount for industry-level variation, we also included controls

186/ASQ, June 2005


for the industry-level standard deviation of free cash flow,size, and debt variables.

Board interlocks are another potential confound, as they canbe argued to operate as substitutes for constraint absorption.Theoretically, however, the omission of board interlocks isunlikely to undermine our findings, as explicated in AppendixB. Methodologically, controlling for the presence of boardinterlocks would be exceedingly data-intensive. Board inter-lock studies are generally based on samples of, at most, afew hundred companies. One of the richest studies of boardinterlocks (Haunschild, 1993) used a one-year cross sectionof interlock data for 949 companies. Our sample wouldrequire us to collect interlock data for over 10,000 companiesover a 15-year period—a monumental data collection effortfor one control variable without solid theoretical justification.

Modeling Approach

To account for variations of the intercept across the N cross-sectional units and the T time periods, and to address prob-lems of heteroskedasticity induced by the use of a proportionas the dependent variable, we tested our hypotheses usinggeneralized estimating equations (GEE) to fit grouped logisticpanel-data models of the form

logit(pijt) = �ij + xijt�, with nijt � Binomial (Nit,pijt)

where nijt is the number of mergers formed between industryi and industry j, Nit is the total number of mergers formed byindustry i with any industry, and pijt is the proportion of thosemergers that industry i formed with industry j (for a thoroughintroduction to GEE in the estimation of GLM, see Hardin andHilbe, 2003). An alternative correction for heteroskedasticityentails performing an arcsine transformation of the propor-tion, p, defined as 2 arcsin ( p) (Cohen and Cohen, 1983).This approach is less desirable than the use of grouped logis-tic generalized linear models, however, because it does notadjust for the frequency of the event (i.e., the denominator inthe proportion). Finally, we tested these models with themultivariate application of the quadratic assignment proce-dure (MRQAP). Compared with alternative regression tech-niques, MRQAP is robust to all forms of misspecification ofthe autocorrelation structure of dyadic data (Krackhardt, 1988)and therefore accounts for unobserved heterogeneity (e.g.,bandwagon effects) at the level of individual industries,industry dyads, industry triads, and any other possible combi-nation of industries. In fact, MRQAP provides a conservativetest of the hypotheses because it also controls for veryunlikely forms of autocorrelation, such as merger behavior inthe semiconductor industry being influenced by mergerbehavior in the chewing gum industry and the poultry andeggs industry. In our results, we report the significance levelsderived from the MRQAP procedure. We performed thesestatistical analyses using Stata’s GEE function for the estima-tion of generalized linear models, in conjunction with Simp-son’s (2002) implementation of MRQAP for Stata.

187/ASQ, June 2005



An additional modeling choice concerned the time structureof the panel. All the constraint-based predictors of mergerbehavior in the sample were time-invariant over a five-yearperiod, as input-output data were only available for the years1982, 1987, 1992, and 1997. Furthermore, the use of a pro-portion as the dependent variable, combined with the sparse-ness of merger observations in each of the 468*468 industry-dyads, implied that shortening the time structure of the panelwould inflate the number of times the denominator in theproportion equaled zero. Because binomial GLM models can-not estimate likelihoods for observations in which thedenominator is zero, excessively short time windows wouldhave increased the noise in the model. To minimize bothaggregation bias and noise, we aimed therefore to match thestructure of the panel as closely as possible to the structureof the measurements. To that end, we fit the regressionmodels on a panel composed of three time periods:1985–1990, 1991–1995, and 1996–2000. This time structuremirrors the one adopted by Finkelstein (1997).

To insure the robustness of the results, we also performedtwo sensitivity analyses. The first concerned the time struc-ture of the panels. We ran the GEE grouped logistic panel-data regression using both eight-year and four-year panels, aswell as a five-year moving-average panel. The results ofthese supplemental analyses were consistent with thoseobtained using a fixed three-period panel, which we reportbelow. The second sensitivity analysis concerned the risk set,that is, the definition of industry-dyads that were at risk ofentering mergers or acquisitions. The risk set can be definedin three ways. In the most restrictive approach, the risk setincludes only dyads in which both industries entered into oneor more M&A deals with any industry during the time periodconsidered. A second approach includes dyads in which atleast one industry entered into one or more M&A deals dur-ing the time period considered. The last and most inclusiveapproach allows all industry-dyads into the risk set. We per-formed our analyses using all three approaches and obtainedlargely comparable results. The only exception was the inter-action between power imbalance and mutual dependence,which was positive and statistically significant in some mod-els but not in others. We report the findings based on thesecond approach to specification of the risk set.

RESULTS

Table 1 presents descriptive statistics and correlation valuesfor all variables. Table 2a reports the results from the QAPGEE grouped logistic generalized linear panel models. Theregression coefficients reported are unstandardized. Standarderrors are noted in parentheses. Because the quadraticassignment procedure produces a non-parametric test of sig-nificance, coefficients in our models can be non-significanteven when standard tests would indicate statistical signifi-cance.

Model 1 includes all control variables. We found no evidencefor the effect of industry profitability on the probability ofmergers, consistent with prior studies (Pfeffer, 1972; Finkel-stein, 1997). We also found no support for other financial and

188/ASQ, June 2005


managerial agency explanations for M&A formation. In lightof past support for such factors at the firm level of analysis,these results suggests that financial and managerial agencytheories of M&A formation may not be meaningfully studiedat the industry level of analysis. This pattern contrasts sharplywith the significant effect of variables measuring the power-dependence structure, which indicates that the industry levelof analysis is at least appropriate, if not desirable in this con-text. Specifically, in model 2, the negative and significantcoefficient for power imbalance lends support to hypothesis1, which predicted a negative association between the powerdifference between potential partners and the likelihood of amerger. Model 3 tests the positive association betweenmutual dependence and the probability of merger formationpredicted in hypothesis 2. The positive and significant coeffi-cient for mutual dependence supports this hypothesis. Theeffects of power imbalance and mutual dependence remainsignificant when both variables are included in the samemodel (see model 4). In model 5, we introduce the interac-tion term and find no evidence for the moderating effect ofpower imbalance on mutual dependence.

In model 6, we pit power imbalance and mutual dependencedirectly against the traditional operationalization of powerdependence. Unlike the coefficients for power imbalance andmutual dependence, the effect for the constraint of industry jon industry i is not significant. When tested in the absence ofpower imbalance and mutual dependence (model 7), howev-er, the effect of the constraint of industry j on industry i issignificant, suggesting that past evidence for constraint mea-

189/ASQ, June 2005



Table 1

Mean, Standard Deviation, and Correlation of Variables

Variable Mean S.D. .01 .02 .03 .04 .05 .06 .07

01. Frequency of mergers of industry i 00. with industry j 0.01 0.1602. Power imbalance (normalized) 2.94 4.04 –.0103. Mutual dependence (normalized) –0.35 0.52 .07 .0304. Power imbalance � Mutual dependence –0.96 3.04 .03 –.68 .5805. Constraint of industry j over industry i 00. (normalized) –7.57 7.38 .03 –.15 .39 .3206. Profitability of industry i 0.15 0.10 –.01 –.01 .04 .04 .0607. Profitability of industry j 0.15 0.10 .00 –.01 .04 .04 .05 –.0108. Free cash flow of industry i 0.15 20.45 .01 .00 .00 –.01 .00 –.04 .0009. S.D. of free cash flow of industry i 0.25 16.89 .01 .01 .00 –.01 .00 –.04 .0010. Size difference of industry i and industry j 0.00 5.94 .00 .00 .00 .00 .04 –.06 .0611. S.D. of industry i’s size 2.54 6.44 .04 .01 .07 .03 .09 –.09 –.0112. S.D. of industry j’s size 2.54 6.44 .04 .01 .07 .03 .02 –.01 –.0913. Debt of industry j 0.32 0.12 .00 .02 .01 .00 –.05 .00 .0214. S.D. of debt of industry j 0.31 0.12 .00 .02 .02 .01 –.05 .00 .0215. 1991–1995 period 0.29 0.46 .00 –.05 .02 .04 .01 .00 .0016. 1996–2000 period 0.36 0.48 .01 .02 –.01 –.01 –.05 –.01 –.01

Variable .08 .09 .10 .11 .12 .13 .14 .15

09. S.D. of free cash flow of industry i .9910. Size difference of industry i and industry j .04 .0411. S.D. of industry i’s size .05 .05 .6512. S.D. of industry j’s size –.01 .00 –.65 .0013. Debt of industry j .00 .00 .03 .00 –.0614. S.D. of debt of industry j .00 .00 .04 .01 –.06 .9715. 1991–1995 period –.02 –.02 .00 –.02 –.02 –.06 –.0616. 1996–2000 period .00 .00 .00 .13 .13 .02 .05 –.48

sured at the individual level of analysis may have been anartifact of conflating power imbalance and mutual depen-dence. In addition, power imbalance and mutual dependenceyield a sizable improvement of fit over model 7, suggestingthat a dyadic specification of the mechanisms through whichenvironmental constraints affect merger behavior enhancesthe predictive ability of resource dependence.5

To ascertain whether power imbalance has a negative effecton the likelihood of constraint absorption irrespective of thedirection of imbalance, we ran models 8 though 13. Models8, 9, and 10 included only observations in which the con-straint of industry i over industry j was greater than the con-straint of industry j over industry i; models 11, 12, and 13include only observations in which the constraint of industry jover industry i was greater than the constraint of industry iover industry j.6 As shown in table 2b, in either set of mod-els, the coefficients for power imbalance and mutual depen-dence are, respectively, positive and negative, consistentwith the results obtained using the entire sample. This find-ing supports the notion that the greater the power imbalancein a dyad, the less likely it is that the dominant organization

5In supplemental models, we added con-trols for the concentration ratios of bothindustry i and industry j. The coefficientsfor both controls were negative and sig-nificant across all models but did notchange either the sign or significance ofthe coefficients for mutual dependence orpower imbalance.

6In these models, we excluded dyads inwhich the constraint of industry i onindustry j was identical to the constraintof industry j on industry i. Including these22,558 dyads did not change the results.

190/ASQ, June 2005


Table 2a

QAP GEE Estimation of GLM Panel Models on Proportion of Number of Mergers Formed by Industry i with

Industry j (N = 250,234)*

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Power imbalance –0.30••• –0.41••• –0.36••• –0.43•••(0.02) (0.02) (0.02) (0.02)

Mutual dependence 0.96••• 1.11••• 1.20••• 0.90•••(0.02) (0.02) (0.03) (0.03)

Power imbalance � Mutual dependence –0.06(0.01)

Constraint of industry j over industry i 0.10 0.37•••(0.02) (0.01)

Profitability of industry i 0.79 1.11 0.17 0.56 0.56 0.32 –0.59(0.26) (0.26) (0.25) (0.25) (0.25) (0.24) (0.24)

Profitability of industry j 0.93 1.16 0.47 0.75 0.78 0.68 0.51(0.24) (0.24) (0.23) (0.22) (0.23) (0.21) (0.24)

Free cash flow of industry i 0.03 0.04 0.06 0.04 0.04 0.04 0.06(0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)

S.D. of free cash flow of industry i –0.02 –0.02 –0.06 –0.03 –0.03 –0.03 –0.06(0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)

Size difference of industry i and industry j 0.0018 0.0023 0.0024 0.0028 0.0029 0.0016 0.0028(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

S.D. of industry i’s size/1000 0.02 0.02 0.02 0.01 0.01 0.01 0.01(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.005)

S.D. of industry j’s size/1000 0.02 0.03 0.02 0.01 0.01 0.01 0.03(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.005)

Debt of industry j –2.94 –4.22 –2.89 –3.98 –3.57 –3.68 –2.91(1.06) (1.03) (1.10) (1.01) (1.03) (0.98) (1.05)

S.D. of debt of industry j 2.59 3.70 2.51 3.59 3.24 3.38 2.62(1.04) (1.00) (1.07) (0.98) (1.00) (0.95) (1.03)

1991–1995 period –0.10 –0.16 –0.08 –0.03 –0.02 –0.05 –0.14(0.07) (0.07) (0.07) (0.08) (0.08) (0.08) (0.07)

1996–2000 period –0.35 –0.33 –0.34 –0.18 –0.16 –0.14 –0.26(0.07) (0.07) (0.07) (0.07) (0.07) (0.07) (0.07)

Constant –7.05••• –6.45••• –6.96••• –6.13••• –6.22••• –5.67••• –5.64•••(0.11) (0.12) (0.11) (0.11) (0.12) (0.12) (0.12)

Wald �2 232 555 1,999 3,742 3,679 3,795 1,066• p < .05; •• p < .01; ••• p < .001; two-tailed quadratic assignment procedure significance levels.* Standard errors are in parentheses. Because the quadratic assignment procedure yields a non-parametric test of sig-nificance, coefficients in our models can be non-significant even if standard tests would indicate statistical significance.

will acquire the dependent organization or that the dependentorganization will successfully acquire the dominant organiza-tion.

Finally, to shed light on the magnitude of the effects of themain variables in our model, we performed an analysis ofpredicted changes in the probability of a merger induced bymutual dependence and power imbalance. For power imbal-ance, the predicted probability of a merger at the median ofthe distribution is 0.0014. At one standard deviation from themedian, the probability is 0.00075, indicating that a one-stan-dard-deviation increase away from the median in powerimbalance reduces the probability of a merger by 53 percent.For mutual dependence, the predicted probability of a mergerat the median of the distribution is 0.0015. At one standarddeviation from the median, it is 0.0021, indicating that a one-standard-deviation increase away from the median in mutualdependence increases the probability of a merger by 42 per-cent.

DISCUSSION AND CONCLUSION

In this study, we aimed to tap the unrealized potential ofresource dependency as a powerful explanation of interfirmrelationships. We were motivated by the expectation that the

191/ASQ, June 2005



Table 2b

QAP GEE Estimation of GLM Panel Models on Proportion of Number of Mergers Formed by Industry i with

Industry j (N = 113,838)*

X Model 8 Model 9 Model 10 Model 11 Model 12 Model 13Variable Ci→j> Cj→i Ci→j> Cj→i Ci→j> Cj→i Ci→j< Cj→i Ci→j< Cj→i Ci→j< Cj→i

Power imbalance –0.25•• –0.36••• –0.34•• –0.46•••(0.03) (0.02) (0.03) (0.03)

Mutual dependence 0.86••• 1.00••• 0.94••• 1.09•••(0.03) (0.03) (0.03) (0.03)

Profitability of industry i 0.37 0.17 0.13 1.31 0.42 0.85(0.41) (0.40) (0.39) (0.31) (0.31) (0.30)

Profitability of industry j 1.33 0.82 1.05 –0.12 –0.27 –0.20(0.30) (0.29) (0.29) (0.39) (0.38) (0.34)

Free cash flow of industry i 0.09 0.09 0.08 –0.03 0.07 –0.02(0.02) (0.02) (0.02) (0.02) (0.03) (0.02)

S.D. of free cash flow of industry i –0.07 –0.08 –0.07 0.03 –0.08 0.03(0.02) (0.02) (0.02) (0.02) (0.03) (0.03)

Size difference of industry i and industry j –0.0114 0.0009 –0.0017 0.0238 0.0159 0.0213(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

S.D. of industry i’s size/1000 0.05• 0.04• 0.03 0.00 –0.01 –0.01(0.006) (0.006) (0.006) (0.006) (0.006) (0.006)

S.D. of industry j’s size/1000 0.002 0.000 0.000 0.052• 0.046• 0.033(0.006) (0.006) (0.006) (0.006) (0.006) (0.006)

Debt of industry j –3.78 –3.10 –4.17 –4.98 –4.21 –3.25(1.43) (1.59) (1.47) (1.50) (1.56) (1.47)

S.D. of debt of industry j 2.61 1.50 2.96 5.07 4.40 3.55(1.40) (1.55) (1.44) (1.44) (1.50) (1.42)

1991–1995 period –0.09 –0.06 –0.02 –0.10 –0.04 0.00(0.10) (0.10) (0.10) (0.11) (0.11) (0.11)

1996–2000 period –0.31 –0.39 –0.23 –0.15 –0.18 –0.03(0.10) (0.10) (0.10) (0.11) (0.10) (0.11)

Constant –6.11••• –6.44••• –5.78••• –6.38••• –7.06••• –6.16•••(0.16) (0.16) (0.16) (0.17) (0.16) (0.16)

Wald �2 708 1,333 1,861 562 1,262 1,934• p < .05; •• p < .01; ••• p < .001; two-tailed quadratic assignment procedure significance levels.* Standard errors are in parentheses. Because the quadratic assignment procedure yields a non-parametric test of sig-nificance, coefficients in our models can be non-significant even if standard tests would indicate statistical significance.

explicit recognition of power as an inherently dyadic phenom-enon would allow us to resolve several conceptual ambigui-ties in the resource dependence model and to reconcile pasttheoretical developments and empirical findings within acoherent framework. The study offers two distinct theoreticalcontributions to resource dependence accounts of interorga-nizational action. First, the revised resource dependencemodel demonstrates that both power imbalance and mutualdependence are necessary constructs in producing a thor-ough theoretical account of power and dependence at thedyadic level. Merely considering the reciprocal dependenciesof two actors is, in and of itself, insufficient to represent thepower-dependence structure in a dyad. The dependence of ion j and the dependence of j on i are building blocks of thetwo theoretically relevant dyadic constructs of power imbal-ance and mutual dependence. Similarly, considering only oneof the two dyadic constructs is also insufficient, becausepower imbalance and mutual dependence shape interorgani-zational action according to distinct mechanisms.

Second, the revised resource dependence model specifiesthe theoretical scope of the distinct effects of mutual depen-dence and power imbalance. We classified interorganizationalaction into operations aimed at restructuring dependenciesunilaterally and bilaterally and operations aimed at usingpower while leaving the dependence structure unaltered.Drawing clear boundary conditions for the revised resourcedependence model has two important implications. First, itallows us to resolve the seeming contradiction between thefindings of this study and those stemming from prior tests ofresource dependence. Second, it makes the revised modelapplicable across a wide spectrum of interorganizationalresponses to resource dependencies.

Based on these theoretical advancements, the revisedresource dependence model addresses the puzzling ques-tion, arising from the original version of the theory, concern-ing how and why the more powerful organization wouldenter into constraint-absorption operations with the depen-dent organization and thus surrender its power and theadvantageous exchange conditions it yields. This puzzleemerged from individual-level analyses of constraint absorp-tion that did not recognize the distinction between differentconfigurations of power imbalance and mutual dependence.These analyses found that an increase in the constraint suf-fered by firms increased their tendency to absorb firms in theconstraining industry. A fully specified dyadic model ofdependence demonstrates, instead, that constraint-absorp-tion operations are significantly less likely as the power differ-ence between potential partners increases and significantlymore likely between firms in mutually dependent industries.In this sense, we amend the notion that constraint-absorptionoperations, such as mergers and acquisitions or joint ven-tures, are fruits of dependence and its converse, power. Thecritical underlying driver of constraint absorption is mutualdependence; in fact, unbalanced power hinders constraintabsorption.

Seen as such, the concept of mutual dependence equips usto understand why firms seek long-term agreements and

192/ASQ, June 2005


when attempts to form such ties succeed. The concept ofpower imbalance helps us to understand the flip side of theconventional questions: that is, why do firms resist certaininterorganizational actions, and when do attempts to pursuesuch actions fail? Furthermore, compared with the traditionaloperationalization of constraint at the individual level thatcharacterizes most studies, our results based on two distinctmeasures of power imbalance and mutual dependence addsubstantial explanatory power to resource dependence theo-ries of constraint absorption.

Limitations and Future Research

Despite the theoretical contributions of this study, the empiri-cal analysis employed here is subject to limitations. The mea-sures of constraint we adopted in this study suffer from thesame problems that have plagued past investigations ofdependence in interfirm behavior (Pfeffer, 1972; Burt, 1983;Finkelstein, 1997). Although we used a more fine-graineddefinition of industry boundaries than previous studies did,we still followed the approach of scholars who preceded usby classifying merger partners according to the primaryindustry in which they operate. Our measure, therefore, isnot sensitive to the fact that most leading U.S. firms arediversified, and it is inaccurate to think of their behavior intheir primary industry without reference to how they are situ-ated in other markets. Although following the approachadopted in previous research allowed for a direct comparisonwith past studies that tested the conventional formulation ofresource dependence theory, the measurement error stem-ming from this methodological choice might have preventedus from fully appreciating our companies’ sources of attrac-tiveness as merger partners. While greater nuance in themeasurement of resource dependence is quite challenging,future research will be able to thoroughly clarify the role ofconstraint in interfirm behavior only by attending to thesemethodological issues.

Another interesting direction for future work concerns therelationship between resource dependence theories and the-ories of the embeddedness of economic exchanges in socialrelations, a phenomenon of demonstrated importance ininterfirm behavior (Granovetter, 1985; Uzzi, 1996, 1997). Therevised resource dependence model suggests that actors inhigh mutual dependence dyads are on average more likely toface uncertainty than actors in low mutual dependencedyads. Consequently, actors in high mutual dependencedyads are more likely to merge. The embeddedness viewsuggests that, to the extent that actors develop trusting rela-tions, they will be able to overcome problems of uncertaintyand opportunistic bargaining. Those successful actors willreduce uncertainty in the flow of needed resources by relyingon social norms of cooperation and reciprocity and henceshould rely less on formal long-term contractual arrange-ments. This notion suggests a testable hypothesis accordingto which the effects of embeddedness should be strongestunder conditions of high mutual dependence. In this sense,embeddedness does not exclude but, rather, complementsresource dependence mechanisms for interfirm action.Though the industry level of analysis employed here is not

193/ASQ, June 2005



appropriate for the study of embeddedness of personal rela-tionships, future research at the firm level can fruitfullyaddress this hypothesis.7 Indeed, past interorganizationalresearch conducted at the firm level of analysis—where therole of embedded personal relations can be analyzed mean-ingfully—suggests that both resource dependencies andsocial-structural factors help to explain patterns of interorgani-zational exchanges. Although the inclusion of social-structuralmechanisms does not diminish the explanatory power of theresource dependence determinants of interfirm behavior(Gulati, 1995), further research addressing the interactionbetween the two is needed (Piskorski and Anand, 2004).

Resource dependence theory is also highly complementarywith transaction cost economics, particularly with respect tothe bilateral dependence associated with asset specificity(see footnote 1). An integrated look at resource dependenceand transaction costs would allow us to explore the role ofex-ante power imbalance on the ways in which firms seek toprotect their ex-post specific investments. Specifically, wesuspect that firms with ex-ante power may resist integrationefforts designed to protect specific assets. Even though thepowerful actor would benefit from integration in order to reapthe benefits of asset-specific investments, such integrationwould also prevent the actor that is more powerful ex-antefrom exploiting the less powerful one. To the extent that thebenefits of integration are smaller than the losses associatedwith the inability to exploit, we should expect that actorswith ex-ante power will seek to avoid such agreements, eventhough such actions are inefficient from the point of view ofthe dyad. Future research should examine whether thishypothesis is borne out with data.

The dyadic approach adopted in this study can also beapplied to the analysis of triads of power relations. In aninteresting extension of resource dependence theory, Gargiu-lo (1993) showed that when obstacles exist to the formationof a direct tie with the constraining party, the constrainedactor can reduce uncertainty in the procurement of neededresources by building a cooptive relationship with a third play-er that controls the constraining party, thus using two-stepleverage. The support we provide for the negative effect ofpower imbalance on the ability to leverage mergers as aresponse to environmental dependencies points to two-stepmerger strategies as a potential tool to pursue desirable butreluctant merger partners. Specifically, for a merger to be aviable option, organizations in a position of dependence mayincrease their attractiveness as merger partners by acquiringa third-party organization that provides critical resources tothe targeted organization. Such a two-step strategy may helpreduce the power asymmetry that constitutes a sizableobstacle to the viability of mergers and acquisitions as instru-ments for reducing uncertainty in procuring critical resources.Future research documenting empirically the actual use orthe potential effectiveness of such indirect power strategiescould usher in a new set of tactics of intercorporate integra-tion.

Our analysis of power dependence is not necessarily restrict-ed to the study of firm dyads. The results of this investigation

7One of the reviewers suggested usinggeographic distance between industriesas a proxy for social mechanisms behindM&A behavior. From the COMPUSTATdatabase, we retrieved zip codes for allfirms across the 468 industries in ourdataset. We then obtained the latitudeand longitude for each of the firms andused them to calculate the distancebetween every firm in each of the109,278 industry dyads. The distancesbetween every possible combination offirms in each industry dyad were aver-aged in the variable “mean distance.”Including this control for interindustrygeographic distance did not change thecoefficients of other variables in any ofthe models we considered. In addition,the regression coefficient for mean dis-tance was positive, contrary to what thetheory of embeddedness postulates. Thisodd result concerning interindustry geo-graphic distance is a further indicationthat embeddedness—like financial andmanagerial agency explanations for M&Aformation—may not be meaningful at theindustry level of analysis.

194/ASQ, June 2005


Bacharach, S. B., and E. J. Lawler1981 “Power and tactics in bargain-

ing.” Industrial and LaborRelations Review, 34:219–233.

Blau, P. M.1964 Exchange and Power in Social

Life. New York: Wiley.

Burt, R. S.1980 “Autonomy in a social typolo-

gy.” American Journal ofSociology, 85: 892–925.

1982 Toward a Structural Theory ofAction: Network Models ofSocial Structure, Perception,and Action. New York: Acade-mic Press.

1983 Corporate Profits and Coopta-tion: Networks of MarketConstraints and DirectorateTies in the American Econo-my. New York: AcademicPress.

Cohen, J., and P. Cohen1983 Applied Multiple Regres-

sion/Correlation Analysis forthe Behavioral Sciences, 2ded. Hillsdale, NJ: Erlbaum.

Collins, N. R., and L. E. Preston1968 Concentration and Price-Cost

Margins in ManufacturingIndustries. Berkeley, CA: Uni-versity of California Press.

Emerson, R. M.1962 “Power-dependence rela-

tions.” American SociologicalReview, 27: 31–40.

Erber, R., and S. T. Fiske1984 “Outcome dependency and

attention to inconsistent infor-mation.” Journal of Personali-ty and Social Psychology, 47:709–726.

195/ASQ, June 2005



invite us in particular to extend our understanding of the rela-tionship between dependency and economic action to net-work positions. Two or more actors jointly occupy the samenetwork position when they have similar relations with eachactor in the network. The power of joint occupants of net-work positions is therefore a function of all dyadic ties thatlink one position to all other positions in the network. Pro-gressing from the dyadic to the positional level of analysiscan extend the applicability of power imbalance and mutualdependence constructs to a variety of phenomena, such asindustry profitability, which has been a primary arena for theapplication of power-dependence notions to the macro studyof structural inequality in economic systems (Burt, 1983).Past research has traditionally conceptualized industry prof-itability as a function of the constraint suffered by an industryvis-à-vis all other industries in an economic system, withouttaking into account the reciprocal constraint that the focalindustry imposes on all others. In a study of price-cost mar-gins in the American economy, however, Piskorski and Cas-ciaro (2004) provided support for the distinct positive andnegative effects of power imbalance and mutual depen-dence, respectively, on industry performance. Such findingsfurther encourage the adoption of a dyadic conceptualizationof constraint to illuminate behavior and outcomes of bothdyads and network positions.

Resource dependence theory rightfully stands as a classiccontribution to organizational discourse. Pfeffer and Salancik(1978) offered one of the richest, most comprehensiveaccounts of organizational life. Not only did they provide acomprehensive theoretical statement of power at the organi-zational level, they also brought it to life with nuanceddescriptions of an array of tactics through which organiza-tions manage the external environment. It is wasteful to rele-gate such foundational work to the margins of theoreticaldevelopment and empirical research. It is therefore theresponsibility of organizational scholars to develop an increas-ingly precise, theoretically sound, and empirically viable state-ment of the resource dependence model to ensure its contin-ued relevance. The reformulation advanced in this papersolidifies the theoretical and empirical bases on which to rein-state resource dependency explanations of interfirm action atthe core of organizational research.

REFERENCES

Finkelstein, S.1997 “Interindustry merger pat-

terns and resource depen-dence: A replication andextension of Pfeffer (1972).”Strategic Management Jour-nal, 18: 787–810.

Friedkin, N. E.1986 “A formal theory of social

power.” Journal of Mathe-matical Sociology, 12:103–126.

Galbraith, C., and C. Stiles1984 “Merger strategies as a

response to bilateral marketpower.” Academy of Manage-ment Journal, 27: 511–524.

Gargiulo, M.1993 “Two-step leverage: Manag-

ing constraint in organizationalpolitics.” Administrative Sci-ence Quarterly, 38: 1–19.

Giebels, E., C. K. W. De Dreu, andE. van de Vliert1998 “The alternative negotiator as

the invisible third at the table:The impact of potency infor-mation.” International Journalof Conflict Management, 9:5–21.

Granovetter, M. S.1985 “Economic action and social

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

Gulati, R.1995 “Social structure and alliance

formation patterns: A longitu-dinal analysis.” AdministrativeScience Quarterly, 40:619–652.

Haleblian, J., and S. Finkelstein1999 “The influence of acquisition

experience on acquisition per-formance: A behavioral learn-ing perspective.” Administra-tive Science Quarterly, 44:29–56.

Hardin, J. W., and J. M. Hilbe2003 Generalized Estimating Equa-

tions. Boca Raton, FL: Chap-man & Hall/CRC.

Haunschild, P. R.1993 “Interorganizational imitation:

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

Hayward, M. L. A., and D. C.Hambrick1997 “Explaining the premiums

paid for large acquisitions:Evidence of CEO hubris.”Administrative Science Quar-terly, 42: 103–127.

Hegtvedt, K. A., and K. S. Cook1987 “The role of justice in conflict

situations.” In E. J. Lawlerand B. Markovsky (eds.),Advances in Group Process-es, 4: 109–136. Greenwich,CT: JAI Press.

Jacobs, D.1974 “Dependency and vulnerabili-

ty: An exchange approach tothe control of organizations.”Administrative Science Quar-terly, 19: 45–59.

Jensen, M.1987 “The free cash flow theory of

takeovers: A financial per-spective on mergers andacquisitions and the econo-my.” In L. Browne and E.Rosengren (eds.), The MergerBoom: 102–143. Boston: Fed-eral Reserve Bank.

Keltner, D. K., and R. J. Robinson1997 “Defending the status quo:

Power and bias in social con-flict.” Personality and SocialPsychology Bulletin, 10:1066–1077.

Krackhardt, D.1988 “Predicting with networks:

Nonparametric multipleregression analysis of dyadicdata.” Social Networks, 10:359–381.

Lawler, E. J., and S. B. Bacharach1987 “Comparison of dependence

and punitive forms of power.”Social Forces, 66: 446–462.

Lawler, E. J., and J. Yoon1996 “Commitment in exchange

relations: Test of a theory ofrelational cohesion.” Ameri-can Sociological Review, 61:89–108.

Levine, S., and P. E. White1961 “Exchange as a conceptual

framework for the study ofinterorganizational relation-ships.” Administrative Sci-ence Quarterly, 5: 538–601.

Marsden, P. V.1983 “Restricted access in net-

works and models of power.”American Journal of Sociolo-gy, 88: 686–717.

Mizruchi, M. S.1989 “Similarity of political behav-

ior among large American cor-porations.” American Journalof Sociology, 95: 401–424.

Palmer, D., B. M. Barber, X. Zhou,and Y. Soysal1995 “The friendly and predatory

acquisition of large U.S. cor-porations in the 1960s: Theother contested terrain. “American SociologicalReview, 60: 469–499.

Perrow, C.1970 Organizational Analysis: A

Sociological View. Belmont,CA: Wadsworth.

Pfeffer, J.1972 “Merger as a response to

organizational interdepen-dence.” Administrative Sci-ence Quarterly, 17: 218–228.

1987 “A resource dependence per-spective on intercorporaterelations.” In M. S. Mizruchiand M. Schwartz (eds.), Inter-corporate Relations: TheStructural Analysis of Busi-ness: 25–55. Cambridge, UK:Cambridge University Press.

Pfeffer, J., and A. Leong1977 “Resource allocation in Unit-

ed Funds: Examination ofpower and dependence.”Social Forces, 55: 775–790.

Pfeffer, J., and P. Novak1976 “Joint ventures and interorga-

nizational interdependence.”Administrative Science Quar-terly, 21: 398–418.

Pfeffer, J., and G. R. Salancik1978 The External Control of Orga-

nizations: A Resource Depen-dence Perspective. New York:Harper and Row.

2003 The External Control of Orga-nizations: A Resource Depen-dence Perspective, 2d ed.Stanford, CA: Stanford Uni-versity Press.

Piskorski, M. J., and B. Anand2004 “Power, brokerage and sta-

tus: Creation of structuralinequality in exchange net-works.” Working paper, Grad-uate School of Business,Stanford University.

Piskorski, M. J., and T. Casciaro2004 “When more power makes

actors worse off: Turning aprofit in the American econo-my.” Working paper, Gradu-ate School of Business, Stan-ford University.

Provan, K. G., J. M. Beyer, and C.Kruytbosch1980 “Environmental linkages and

power in resource-depen-dence relations betweenorganizations.” AdministrativeScience Quarterly, 25:200–225.

196/ASQ, June 2005


APPENDIX A: Interpretation of Prior Tests of Constraint Absorption

Let yi→j be the likelihood of firm i entering a constraint absorption operationwith firm j, Cj→i a measure of dependence of firm i on firm j, and Ci→j a mea-sure of firm j’s dependence on firm i. According to the theory proposed inthis paper, yi→j should be empirically modeled as

yi→j = �(Cj→i – Ci→j) + �(Cj→i + Ci→j)We now seek to rewrite the equation in terms of the two variables Ci→j andCj→i. Because the equation contains an absolute value term, it has to berewritten into two equations: one when Cj→I > Ci→j and one when Cj→I < Ci→j.This separation yields:

for Cj→i > Ci→j

yi→j = �(Cj→i – Ci→j) + �(Cj→i + Ci→j)

yi→j = �Cj→i – �Ci→j + �Cj→i + �Ci→j

yi→j = (� + �)Cj→i + (� – �) Ci→j

for Cj→i < Ci→j

yi→j = �(Ci→j – Cj→i) + �(Cj→i + Ci→j)

yi→j = �Ci→j – �Cj→i + �Cj→i + �Ci→j

yi→j = (� – �)Cj→i + (� + �)Ci→j

When the two equations are considered in a single equation, regardless ofthe levels of imbalance between i and j, they are essentially summed togeth-er. This summation yields:

197/ASQ, June 2005



Selznick, P.1949 TVA and the Grass Roots.

Berkeley, CA: University ofCalifornia Press.

Simpson, W.2002 “Quadratic assignment proce-

dure in Stata.” Boston: Stataprogram, Harvard BusinessSchool.

Thompson, J. D.1967 Organizations in Action: Social

Science Bases of Administra-tion. New York: McGraw-Hill.

Uzzi, B.1996 “The sources and conse-

quences of embeddednessfor the economic perfor-mance of organizations: Thenetwork effect.” AmericanSociological Review, 61:674–698.

1997 “Social structure and compe-tition in interfirm networks:The paradox of embedded-ness.” Administrative ScienceQuarterly, 42: 35–67.

Van de Ven, A. H., A. L. Delbecq,and R. Koenig1976 “Determinants of coordina-

tion modes within organiza-tions.” American SociologicalReview, 41: 322–338.

Williamson, O. E.1975 Markets and Hierarchies.

New York: Free Press.

y = (� + �)Cj→i + (� – �)Ci→j

y = (� – �)Cj→i + (� + �)Ci→j

y = 2�Cj→i + 2�Ci→j

This implies that when the underlying equation is estimated using the con-stituent dependencies Ci→j and Cj→i, the coefficient estimates essentially cap-ture the causal effect of mutual dependence (given by �) and not of powerimbalance.

APPENDIX B: Board Interlocks and Constraint Absorption

If interlocks are substitutes for constraint absorption, their omission from themodel affects our findings, as illustrated in figures B.1, B.2, and B.3. In allfigures, the interlock entails i having a seat on j’s board, and constraintabsorption entails i acquiring j. It is assumed that the relationship betweenpower imbalance and the probability of i having a seat on j’s board is positiveif i is more powerful than j, and negative if i is less powerful than j.

Figure B.1 illustrates how the positive coefficient of mutual dependence as apredictor of constraint absorption is underestimated if interlocks are omitted.Similarly, figure B.2 shows that the negative impact of power imbalancefavoring j on the likelihood of i acquiring j would also be underestimated ifinterlocks were omitted. Figure B.3 illustrates, instead, how the negativeimpact of power imbalance favoring i on the likelihood of i acquiring j wouldbe overestimated if interlocks were omitted. That is, most of our findings(models 1–7 and 11–13) are actually conservative if one assumes that inter-locks are substitutes for constraint absorption. The only results that might beoverestimated are those in which the power imbalance favors the potentialacquirer (models 8–10). But because we find consistently negative coeffi-cients whether power imbalance favors the acquirer or the target, and thetheory suggests a negative relationship with constraint absorption regardlessof the direction of the power imbalance, it is difficult to argue that the sup-port for H1 is an artifact of interlocks being omitted from the analyses.

198/ASQ, June 2005


Figure B.1. Mutual dependence, board interlocks, and constraintabsorption.

Mutual dependence of i and j Interlocks(i has seat on j’s board)

+

+ –

(Underestimated ifinterlocks omitted)

Constraint absorption(i acquires j)

199/ASQ, June 2005



Figure B.2. Power imbalance favoring target j, board interlocks, andconstraint absorption.

Power imbalance favoring j Interlocks(i has seat on j’s board)



–

–

–

Figure B.3. Power imbalance favoring acquirer i, board interlocks, andconstraint absorption.

Power imbalance favoring iInterlocks

(i has seat on j’s board)



- -

+

Power Imbalance, Mutual Dependence, and Constraint ... dependence.pdfpotential of resource...

Documents

Transcript of Power Imbalance, Mutual Dependence, and Constraint ... dependence.pdfpotential of resource...