The Dark Side of Social Capital ^*

Research on social capital has stressed the advantages that networks can bring to managers and other economic actors. The enthusiasm with this "bright side" of social capital, however, neglects the fact that social bonds may at times have detrimental effects for a manager. This paper tries to correct the optimistic bias by looking at the "dark side" of social capital. Continuing benefits from social capital require that managers can adapt the composition of this social capital to the shifting demands of their task environment. This often implies the ability to create new ties while lessening the salience of some of the old bonds--if not severing them altogether. Available evidence, however, suggests that this ability may be encumbered by the same relationships purportedly responsible for the prior success of the manager. When and how this may happen is the central question we address in this paper. We argue that strong ties to cohesive contacts limit the manager's ability keep control on the composition of his network and jeopardize his adaptability to changing task environments. We test our ideas with data on managers operating in a special unit of a European high-technology firm.

"Social capital" has become a ubiquitous notion in the study of organizations. Despite differences in emphasis adapted to specific research agendas, most treatments coincide with the general definition of social capital as "Éthe sum of resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition" (Bourdieu and Wacquant 1992:119; see also Bourdieu 1980, for an early formulation). Behind this compelling metaphor, a rich stream of research highlights how social networks can enhance the ability of individuals and organizations to attain their goals. Scholars have identified two main mechanisms that account for the positive impact of social networks on individual and organizational action. First, networks facilitate access to information, resources, and opportunities (Granovetter 1974; Campbell, Marsden, and Hulbert 1986; Flap and de Graaf 1986; Coleman 1990; Burt 1992, 1997
; Podolny and Barron 1997). Second, social networks help actors to coordinate critical task interdependencies and to overcome the dilemmas of collective action (Blau 1958; Pfeffer and Salancik 1978; Kotter 1985; Gargiulo 1993; Gulati 1995).

Faithful to an agenda that vindicates the importance of social ties in modern economic transactions (Granovetter 1985), sociologists have largely stressed the advantages that embeddedness in social networks can bring to actors engaged in such transactions (e.g., Coleman 1990:300ss.). This enthusiasm with the "bright side" of social capital has led sociologists to disregard the possibility that social bonds may at times hinder, rather than help, an actor's ability to pursuit his interests (Portes and Sensenbrenner 1993). The instrumental value of social capital rests on the good match between the resources needed by an actor and the resources he could access through his contacts. To maintain this match, the actor should be able to adapt the composition of his social capital to fit his changing needs. Available evidence, however, suggests that an actor's ability to adapt his social capital may be encumbered by the particularistic demands posed by the same relationships purportedly responsible for the initial success of this same actor.

This paper tries to correct the optimistic bias of the existing research by looking at the "dark side" of social capital. Building on existing theory, we argue that the network that provided an actor with social capital may also limit his ability to change the composition of this network as he sees it fit. When and how this may happen is the central question addressed in this paper. Our answer focuses on the obstacles that strong, cohesive social bonds may pose to the adaptation of social capital to changing task environments. Evidence comes from managers operating in a special unit of a European subsidiary of a high-technology firm. The analysis emphasizes how the structural characteristics of the managers' social networks thwarted the adaptation of their social capital to a fluid organizational context.

The two sides of social capital

Managers bring to their job more than the skills they have accumulated through years of education and experience. They also bring the assets they can procure through their social networks. Economists refer to the assets embodied in the manager's skills as his "human capital" (Becker 1964). Sociologists coined the term "social capital" to designate the assets tied to the manager's social network (Bourdieu 1980; Coleman 1990).

The compelling metaphor embodied in the notion of social capital has brought together a number of research streams that focus on the beneficial effects of social networks in economic exchanges. Existing studies cover aspects such as individual success in labor markets (Granovetter 1973; Lin, Ensel and Vaughn 1981; Lin and Dumil 1986; Flap and de Graaf 1989); managerial careers (Burt 1992, 1997; Podolny and Barron 1997); and organizational performance (see Powell and Smith-Doerr 1994, for a comprehensive review). Networks can also help organizations seeking to form partnerships by providing efficient access to information on the availability, competencies, and reliability of potential partners, thus lowering searching costs and alleviating the risk of opportunism (Granovetter 1985; Powell and Brantley 1992; Gulati 1995). Closely related to this last point, scholars have also demonstrated how social networks have a crucial role in helping managers and organizations to reduce the uncertainty that results from task interdependence (Blau 1958; Pfeffer and Salancik 1978; Kotter 1985; Gargiulo 1993).

The interest in the role of social capital in overcoming coordination problems has been recently fueled by the growing importance of flatter structures, teamwork, entrepreneurial initiative, and decreasing reliance on authority relations to handle interdependence (Kanter 1984; Ditcher 1991; Baker 1992; Sheppard and Tuchinsky 1996). This new environment poses an important challenge for managers. While the new organizational blueprint gives managers considerable more latitude to perform their role, it also increases the uncertainty surrounding their task and careers. Managers have to handle a growing number of task interdependencies that link them to people outside their line of authority. Since the formal structure offers little assistance in dealing with these "lateral" interdependencies, managers have to find alternative ways of coordination to accomplish their task effectively. Social capital becomes a strategic tool to add value in this new organizational environment (Nohria and Eccles 1992).

By demonstrating the positive role of social relationships in modern economic life, sociologists have contributed to a time-honored research program revitalized by Granovetter (1985) with his seminal essay on the social embeddedness of economic transactions. While economists have typically ignored social relations or have treated them as an obstacle to attain economic rationality, sociologists have shown that economic rationality can be actually enhanced by embedding transactions in social networks that facilitate trust and diminish the risk of opportunism. In pursuing this research program, however, sociologists have often disregarded a well-known duality also stressed by Granovetter (1985) in his essay: social structures can be both a source of opportunities and a source of constraint for individual behavior. As one might expect from the positive tones of the metaphor, the bias towards the favorable effects of social structures is especially apparent in the discussions of social capital (e.g., Coleman 1992:300ss). This enthusiasm with the "bright side" of social capital has led sociologists to disregard the "dark side" of social relationships economists like to emphasize--that is, that social bonds could be an obstacle to pursuit economic interests.

Despite this pervading optimism, the evidence on the constraining effects of social bonds on individual action is not missing. While most of this evidence may come from studies that portray the supporting and yet oppressive effects of dense social networks in small communities (e.g., Fischer 1982), the dark side of social capital has been also stressed by sociologists analyzing economic behavior. Thus, Portes and Sensenbrenner (1993) review various studies describing ethnic entrepreneurs suffocated by the particularistic demands posed by the same strong social ties purportedly responsible for facilitating their initial access to essential resources. Summarizing their analysis of these findings, Portes and Sensebrenner (1993:1341) propose a direct relationship between the amount of social capital initially available to successful entrepreneurs on one hand, and the level of particularistic demands and restrictions to freedom of action placed on those entrepreneurs, on the other.

The theme behind these warnings is one familiar to economists that emphasize the negative effects of social bonds: the maintenance of social capital entails honoring obligations that may conflict with the pursuit of self-interest. Confronted with this problem, some sociologists have come to accept that actors may at times be better off with the impersonal, "arms-length" transactions glorified by economists than with the embedded exchanges advocated by their discipline (e.g., Uzzi 1997). Yet, the critique to the one-sided view of social capital goes beyond the dichotomy between "embedded" and "arms-length" transactions that typically confronts sociologists and economists. The restrictions that strong social bonds may pose on an actor may hinder not only his attempts to substitute arms-length transactions by previously embedded ones, but also attempts to initiate new embedded transactions. In other words, the network that provides the actor with social capital may also encumber his ability to change the composition of this social capital as he sees it fit. The tension here is not between embedded and arms-length transactions, but rather between the currently available social capital and the social capital required to cope with the new demands or opportunities perceived by the actor.

Perhaps one of the reasons for which scholars have largely ignored the dark side of social capital is related to the specific conditions in which its negative effects may become apparent. The dark side of social capital is not always consequential; rather, its effects will be noticed only after the actor reaches a point beyond which the resources available to him through his current contacts are no longer adequate. The entrepreneurs discussed by Portes and Sensenbrenner (1993) did benefit from the social capital initially available to them, but the obligations that resulted from those benefits curtailed their subsequent ability to pursue other business opportunities. It was the change in the entrepreneurs' opportunities--which may have resulted from their initial success--that made the constraining effects of social bonds consequential. Since the instrumental value of social capital lies on the match between the resources needed by an actor and the resources provided by the actor's contacts, changes in the actor's task environment may require changes in the composition of his social capital. In this respect, social capital is similar to human capital. Changes in task environments may render obsolete current skills and capabilities, forcing people to make further investments in education.

Investments in social capital, however, are substantially more complex than investments in human capital (Coleman 1990). While a person can typically acquire new skills without having to discard previous ones, the same is not always true for social capital. The maintenance of social capital requires investing time and energy in one's contacts. Since people--and managers are not an exception--have a limited amount of time and energy, an obligation to maintain relationships that are no longer advantageous may hinder the ability to cultivate other relationships necessary to renew the managers' social capital. The ability to withdraw from business relationships that are no longer advantageous has been often recognized as an important factor in the adaptability of managers and organizations to changes in their environments (Miles and Snow 1992). Paradoxically, the better the fit between an actor's social capital and his current environment, the harder will be to adapt this social capital to a new environment (Uzzi 1997). ₁ Changing the composition of social capital often implies creating new ties while lessening the salience of old bonds--if not severing them altogether. The more intense and productive the ties with the old contacts were, the more difficult will it be to part with those relationships. Such is the paradox of social capital: the brighter its bright side, the darker the potential effects of its dark side.

The mechanism behind this paradox is rooted in the very logic of reciprocity that turns relationships into the assets that form social capital (Coleman 1990). Reciprocity is more than repayment or the last favor received: it is a general norm that prescribes a certain type of behavior towards relevant others: takers ought to be givers. The norm of reciprocity may oblige a manager to assist a contact, even if he expects few benefits from future exchanges with this particular contact. This is especially the case if this contact has been rather helpful to the manager, or to somebody he is indebted to. Failure to reciprocate may result in strong sanctions and in a serious damage to his reputation as a trustful contact, a damage that can be consequential for the manager's ability to create new ties. It is precisely this failure to reciprocate what we often disapprove of the "instrumental" individual, the person who cuts the ties of obligation to the group once he sees no further benefit from exchanges with its members. In a culture of "generalized exchange" (Bearman 1997), reciprocity is not restricted to localized exchanges within cohesive clusters: favors can be repaid to people other than the ones in the group. Although most organizations probably aspire to have a culture of generalized exchange, the majority of them operate within a system of restricted exchange: favors ought to be repaid to the givers, or to the giver's friends. In these organizational contexts, the norm of reciprocity, combined with the capacity to impose sanctions to defectors, may force a manager to attend demands even when he expects no further benefit from the contacts posing those demands. ₂

The effects of reciprocity are likely to be compounded by a second mechanism that may keep managers tied to contacts that have lost their value as social capital, even without the managers being aware of the problem. This second mechanism is relational inertia. People get used to dealing with their long-term partners. This familiarity breeds strong bonds of mutual understanding and trust that greatly facilitates cooperation (Gulati 1995b). Unfortunately, the same strong bonds may also serve as a filter for the information and the perspectives reaching the actors, generating a "cognitive lock-in" that isolates them from the outer world (Grabher 1993). In addition, the easiness of cooperation with familiar partners raises the cost of making the investments that are necessary to initiate and to consolidate new relationships. This cost is even higher once the uncertainty associated with the formation of new ties is taken into account. These inertial factors can make established relationships extremely resilient to losses in their instrumental value due to changing task environments. If the change in task does not come with tangible signs such as a physical relocation or new a new reporting relationship, the manager may fail to recognize the need for adapting the composition of his social capital to fit the new task. Since failure to develop the adequate relationships is likely to have an impact on performance, the manager may eventually learn the hard way about the inadequacy of his network. Late adaptation, however, comes at a cost. In some cases, the cost may be failure in the new task, with potentially serious consequences for the manager and for the organization.

The tension between the forces of reciprocity and relational inertia on one hand, and the changing nature of the resources needed by a manager on the other is intrinsic to social capital. One simply cannot have the benefits of social capital without the corresponding obligations and the risk of relational inertia. This tension is particularly apparent for managers facing transitions that affect their task environment without simultaneously introducing clear-cut changes in the conditions under which they have to perform their jobs. This situation is rather frequent in the dynamic contexts that characterize today's flat, entrepreneurial organizations. Social capital is key to coordinate task interdependencies in these organizations. At the same time, the fluid character of the managerial task makes task interdependencies shift, following the changes in the resources and expertise managers have to coordinate to add value to the organization. Indeed, the fluidity of the interdependencies is one of the main arguments against attempts to coordinate them through formal mechanisms and standard procedures. In these contexts, the ability to adapt the composition of his social capital is perhaps one of the most valuable attributes of a good manager.

For a manager, the problem is how to have the benefits of the "bright side" of social capital while minimizing the potentially deleterious effects of its "dark side". How to enjoy the advantages of resourceful social networks while maintaining some degree of control over the composition of this network? Our approach to this question is to look at factors that may boost the impact of the forces of reciprocity and relational inertia, hence hindering the manager's ability to renew his network. We argue that the structure of the network that defines social capital is one such factor. More specifically, we argue that a manager's ability to renew the composition of the network that carries his social capital is a function of the structure of that network. The issue is to specify how network structure may curtail the manager's ability to decide how to allocate his time and energy across a set of actual or potential contacts. For this purpose, we turn to the control implications of structural hole theory.

The effects of network structure

Structural hole theory (Burt 1992) asserts that the ability of an actor to have control over his environment is a function of the structure of this actor's network. The more a manager depends on parties who are difficult to substitute and who can coordinate their behavior, the less his ability to negotiate his role in the network. Conversely, lack of contact among those parties creates "structural holes" that enhance the manager's ability to control his environment. ₃ Players who occupy brokerage positions between disconnected alters have comparative advantages in negotiating relationships, which allows them to secure more favorable terms in the opportunities they chose to pursue. According to Burt (1997:343), "managers with contact networks rich in structural holes know about, have a hand in, and exercise control over more rewarding opportunities. They monitor information more effectively than it can be monitored bureaucratically. They move information faster, and to more people, than memos."

The focus of the research inspired by structural hole theory has been on the outcomes of social capital. The information and control advantages that accrue to managers occupying brokerage positions in networks are deemed responsible for the managers' early promotions or comparatively higher bonuses, which are seen as a recognition to their ability to add value to their organizations (Burt 1992, 1997). Structural hole theory, however, may be also used to explain the process through which social capital is created and renewed by the managers. According to structural hole theory, a manager strongly tied to contacts that are in turn connected to one another is expected to have little autonomy to negotiate his role vis-à-vis his contacts. Likewise, we can hypothezise that such a manager should have difficulty in renovating the content of his network when the pressures of the environment may require him to do so.

The reasons behind our hypothesis lie on the interplay between social structures on one side, and the forces of reciprocity and relational inertia on the other. As Granovetter (1985) has argued and empirical studies have confirmed (e.g., Gulati 1995a) the probability of cooperation between two actors is enhanced by the presence of third parties. Burt and Knez (1995) dissect the mechanism behind this effect in their analysis of the impact of third parties on trust, which they define as "anticipated cooperation" (cf. Coleman 1990). Indirect connections create a reputational lock-in that ensures cooperation. Non-cooperative behavior by either partner may be reported to--or worse. observed by--common partners. This would typically have serious negative effects on future relationships entered by the defecting partner, hence serving as an effective deterrent to defection (Raub and Weessie 1990; Burt and Knez 1995). Consequently, managers linked through ties embedded in third-party relationships are more likely to conform to pressures of reciprocity. The strong ties to his contacts, and these contacts' ability to coordinate their demands on the manager, magnify the impact of the norm of reciprocity on the manager's behavior as well as the effects of relational inertia, hence affecting the manager's ability to invest in the development of new ties.

In line with our argument on the two sides of social capital, these potentially harmful effects may remain latent until the manager actually needs to develop those new ties due to changes in his task environment. Moreover, there is evidence that at least in some situations the short-term effects of strong ties to coordinated contacts may be actually beneficial to the manager. ₄ As Podolny and Barron (1997) have recently argued, dense ties among the key people a manager depends upon may actually facilitate performance, especially for low and middle-level managers. The manager with such a network is likely to face a well-defined and consistent normative framework within which to perform his role, thus avoiding the tensions of having to respond to conflicting demands. Our argument, however, suggests that this consistent framework may turn against the manager when he, prompted by changes in his task environment, needs to change the composition of the network. Although the peril is perhaps negligible when the new task is accompanied by changes in reporting line or physical location, this may not be always the case. If such tangible changes in the conditions under which the manager performs the new job are absent, the dense network that was hitherto a major source of support for the manager may now curtail his ability to develop the social capital needed to succeed in the new task. The old network is no longer a resource for the manager, but a constraint that encumbers his performance.

From a structural perspective, the negative effects of strong ties to cohesive contacts on managerial adaptability can operate in two different ways. First, restricted autonomy may result from the combined pressure posed by all contacts in the manager's network. The stronger the ties to those contacts, and the stronger the connections among the contacts, the more they will be in the position to collectively enforce reciprocity from the manager towards the members of the network. Group cohesiveness is also likely to drive members--including our focal manager--to a passive acceptance of the ongoing patterns of interaction as the normal way to carry out their business, hence promoting relational inertia. Second, restricted autonomy may also result from the dominant role of one or a few contacts who monopolize a large proportion of the time and energy the manager devotes to his network. The "hierarchical" role of these contacts in the manager's network puts them in the position to force their demands upon the manager, leaving him little time to invest in new relationships. Alternatively, the manager may perceive these prominent contacts as all the social capital he needs, thus discouraging further investments in developing this social capital.

To summarize our argument: we expect managers whose contacts networks lack structural holes to face more difficulties in adapting their networks according to the changing demands of their task environment, thus leading to a higher rate of failure in the renewal of their social capital. Lack of structural holes may result from strong ties to the various members of a cohesive cluster or from very strong ties to few players that are well connected within the rest of the manager's network. Managers lacking structural holes should be more likely to fail in adapting the content of their social capital in line with changing environmental demands. To test this proposition, we examine managers operating in a special unit of a European subsidiary of a high-technology firm. We focus on their failures to adapt the composition of their social capital to the changing demands placed on them by a flat organizational environment. We describe this organizational environment, then explain the data and methods used in our analysis.

The organization

Our study focused on a unit within the Italian subsidiary of a leading multinational computer company in the early nineties (see Benassi 1993, for detailed description). Like most firms in the industry, the company was dealing with difficult market conditions. Impressive price-cutting and accelerating competitive dynamics driven by dramatic advances in chip technology made profit margins plummet, forcing firms, and especially large firms, to reshape their activities. In this context, the search for more effective organizational configurations was a major endeavor for large computer manufacturers (The Economist 1993). Our firm was not an exception. At the time of our study, several initiatives of organizational change were under way. Headquarters explicitly initiated some of these initiatives, while others were emerging out of the everyday practice of organizational transformation. Our study focused on one of these emergent strategies. A small unit operating in one of the Italian plants was promoting alternative forms of voluntary cooperation among business units within and outside the organization. Although our unit was not originally conceived as a change agent, its style of work resulted in an emergent strategy that could lead towards a radical change in the processes and the culture of the Italian organization. A good part of that change was already observable inside the unit.

The Direzione Processi Industriali (Direction of Industrial Processes, or DPI) was a small unit formally created in January 1991 and staffed by 19 members, all but one of them male. Its origins can be related both to a process of organizational change and to individual initiatives. Early in 1991, the existing functional structure was substituted by a business unit organization and new centers of competence were created to deal with the emerging issues the company was facing. At the same time, an autonomous research initiative undertaken by a group of people working at the Italian plant focused on new technological strategies and on alternative models of conceptualizing the overall activity of the firm, both at the national and the international levels. The creation of DPI gave an institutional form to the new, disperse expertise that resulted from these previous initiatives. Out of the 19 members of DPI, 10 came directly from a unit that launched some of these initiatives and continued to work on their projects until these were completed or terminated. Another group of 6 people also came from a single unit within the firm. Thus, from the inception, the largest majority of DPI managers had worked before with some of their current colleagues. At least initially, several of them even continued with the same projects that they have started in their previous job, although new colleagues might have joined the teams.

DPI's scope of activity was very broad. The unit operated both inside and outside the company, providing solutions to internal business units, top management, and international functional managers, as well as to external clients. Its competencies included devising manufacturing strategies for the Italian plants, developing a market-driven quality approach, promoting marketing-manufacturing cooperation, and creating tools and methods to implement the different initiatives. It also coordinated activities of managers in charge of setting long-term strategies and represented the Italian plant in international company hearings. To an important extent, these tasks resulted from autonomous initiatives that became institutionalized in the new unit and were then formally recognized by the company. In this sense, DPI clearly resembles the entrepreneurial image associated with flat organization. This image also corresponded with the internal structure of the unit. Although formal hierarchy and task differentiation did exist, barriers among task groups and individuals were insignificant. Personal initiative was not only strongly encouraged within DPI: it was a requisite to do the job. The managers knew that solutions were driven by multiple contributions and that external and internal relationships were crucial for getting things done. They believed that traditional organizational mechanisms were no longer valid to accomplish the complex task they had at hand. Instead of relying on the formal hierarchical structure, the managers sought to leverage horizontal relationships and promote cooperation among internal departments. These initiatives often led to new business opportunities for the firm, both inside and outside the formal boundaries of the organization. Although DPI had the support of top management, its location in the formal structure made it a peripheral part of the organization. The head of DPI reports to the plant manager, who is two steps below the top management.

In line with the organizational blueprints adapted to its style of work, DPI favored working through project teams by which the unit pooled resources from within and outside the organization to provide specific solutions to both internal and external customers. Between January and October 1991, DPI was directly involved in 73 of such projects. At the time of our survey, 70 percent of the projects were still under way, 20 percent were recently completed, and 10 percent abandoned. One third of these projects was a continuation of those carried out by previous units. The remaining ones were either a direct initiative of DPI (43 percent) or were launched upon customers' demands (57 percent). In more than 60 percent of the projects, DPI acted as leading unit. Only 8 of the 73 project teams were formed exclusively with people from DPI. On average, project teams had 6.6 people representing 3.7 different units. DPI contributed an average of 2.8 team members per project. Typically, one of these people was the team leader, often paired with a leader from another unit. On average, each of the 19 DPI managers was responsible for more than 4 projects. In addition, he or she would participate in other project teams. Putting the two roles together, an average DPI manager participated in about 10 different projects. Although managers did have initiative regarding their participation in the projects, the composition of the teams was largely driven by reasons of technical expertise and experience, and was ultimately the responsibility of the head of the unit. In this sense, DPI managers were not truly self-selected into the project teams they participated.

The managerial approach and the internal structure of DPI makes the unit a good example of the fast pace, relationship-driven environment where adaptability of ones' social capital is key to individual and collective success. The managers developed and leveraged relationships both inside and outside the firm, actively seeking to promote cooperation ties that cut across the formal structure and the boundaries of the organization. To attain that goal, DPI managers must also coordinate the task interdependencies created by their joint participation in teams and by their combined attempts to mobilize resources from independent areas of the organization. This setting makes the unit a fruitful laboratory to analyze the factors that may thwart the renewal of the content of social capital.

Data and Methods

We argue that managers lacking structural holes should be more likely to fail to adapt the content of their social capital in line with changing environmental demands. To test this hypothesis, we examined mismatches between the level of task interdependence and the level of work-related consultation between the nineteen managers working for DPI. Specifically, we focused on failures to hold consultation at levels deemed adequate for the intensity of the respective task interdependence. Drawing on resource dependence theory (Pfeffer and Salancik 1978), we assume that effective managers should shape their social capital to maximize their capacity to handle the task interdependencies affecting their jobs. In an extreme case, this would mean a perfect mapping of the consultation network on the interdependence network. Individual managers, however, are likely to deviate from this extreme case in two ways. First, a manager may keep ties with people he does not depend upon--or people with whom he has only weak task interdependence. We may refer to these as cases of "excessive" consultation. Second, a manager may have a low level of consultation (or no consultation at all) with people he strongly depends upon. We call these situations of missing consultation "coordination failures."

From the standpoint of resource-dependence theory, the impact of each type of deviation on managerial performance should be different. Although ties with people a manager does not depend upon to carry his task may be seen as an inefficient way to allocate effort, these ties do not necessarily have a negative impact on the manager's performance. Moreover, such ties may be instrumental in two ways that go beyond the immediate task interdependencies affecting the manager. First, managers may have to keep weak ties with people that were key for their task in the past and who may be important again in the future. Indeed, Granovetter's (1973) analysis of labor markets vindicated the instrumental value of those ties. Second, managers may use those ties as bridges to access resources or information controlled by people the manager has been unable to reach directly, or to effect indirect political influence on people they depend upon but with whom they cannot establish good a working relationship (Gargiulo 1993). Thus, consultative ties not coupled with task interdependence may still perform an instrumental role for the manager. The consequences of the second type of deviation are different. Failure to consult with a colleague a manager clearly depends upon should make coordination more difficult and time-consuming, which in turn should affect the manager's performance. Our focus on coordination failure captures the theoretical difference between the two types of deviation. We expect managers whose networks lack structural holes be more prone to fail to adequately coordinate strong task interdependencies.

The data analyzed in this paper comes from a self-administered questionnaire distributed to all members of DPI in October 1991. This questionnaire was tailored using the extensive information gathered through ethnographic observation. From June 1991 to December 1991, one of the authors was allowed to engage in daily observation of the DPI operations. He also had easy access to the unit members for interviews, could consult written communication and formal documents, and attended team and unit meetings. All 19 managers working for DPI responded to the survey. Our questionnaire comprised two parts. First, a general part covered information on the manager's involvement in projects, as well as an evaluation of the functioning of the unit. Each member was also presented with a roster of his or her colleagues within DPI and asked to rank his or her level of consultation on problems relevant to his or her work with each of these colleagues, in a scale from 0 (no consultation) to 3 (strong consultation). This question yields a 19-by-19 matrix of consultation ties among DPI managers. Second, those managers who were responsible for the coordination of at least one of the 73 projects implemented by DPI were asked to fill out a booklet with information about each project under his supervision. The booklet surveyed detailed information on the project and on the people involved. It also included a matrix in which the project coordinator had to list all the people that participated in the project team, including him or herself, and to rate the level of cooperation between all pairs of members, on a scale from 0 (no cooperation) to 3 (strong cooperation). Based on this information, we constructed independent measures of task interdependence and consultation among the nineteen managers working for DPI.

Task interdependence. We measure task interdependence among DPI managers as a linear function of their joint involvement in the projects implemented by the unit. Two characteristics of our measure are worth noting. First, the measure is an objective indicator or task interdependence among our managers. It is also exhaustive, since the managers' task was largely confined to their participation in the projects initiated or facilitated by the unit. Second, the measure captures both direct and indirect task interdependence between the managers. By working with one another in a project, the managers' tasks become directly interdependent. Managers working in the same projects are also indirectly interdependent, since they are demanding resources from the same units in the firm, that is, from the ones that jointly participated in these projects. For any two managers {i, j}, we can define task interdependence t_ij as the number of projects in which managers i and j were jointly involved during the period covered by our research. This raw measure, however, is affected by the number of colleagues with whom a manager jointly participated in project teams. To control for this factor and to focus purely on the manager's pattern of interdependence - rather than on its volume - we use a proportional measure of task interdependence. Manager i's interdependence with manager j can be thus represented in proportional terms, as the ratio between the number of joint projects with manager j, t_ij, and the sum of all joint projects across all managers q, including j:

Consultation. This measure captures the strength of consultation for work-related issues among the same 19 managers. They were presented with a complete list of their unit's colleagues and asked for the extent to which they routinely consult with each colleague regarding matters that concerned their work in the unit. Managers could rate their answers from 0 (no consultation) to 3 (strong consultation). The raw response data generates a 19-by-19 matrix of consultative ties. The consultation network reveals a relatively flat structure. The score for the least prominent manager is only 60 percent of the score of the most prominent manager. ₅ The consultation network is also very dense: The average strength of the ties in the network is 1.32, with an effective range from 0 to 3. The average DPI manager reported some level of consultation with 11 different colleagues. Out of the 342 possible consultation ties, 62.6 percent are present; among these, 23.1 percent are strong, 23.7 percent medium and 15.8 percent weak. Any member of the unit could reach any other member in a maximum of three steps.

For a variety of reasons not related to task interdependence, managers may vary in their tendency to be the source or the target of consultation regarding work-related matters. To control for this idiosyncratic variation, with use a proportional measure of consultation rather than the raw scores; this allow us to capture the pattern of how a manager has to allocate time to the different people he actively consults or to the people who consult him for work-related matters. ₆ This allocation is a function of both the attention the manager seeks from others and the demands other colleagues pose on him. Thus, we measure manager i's consultation with manager j, S_ij, as the proportion of attention manager i has to allocate to manager j, both as a result of his seeking out j and of being sought out by j (being s_ij the self-reported raw measure and 0 lessthan/equals s_ij lessthan/equals 3):

The inclusion of the demands posed on the manager by his colleagues, even when the manager does not seek these colleagues out, is relevant in this context. Indeed, our reasoning assumes that the manager will have to pay some attention to these contacts, even if he does not seek them out to discuss his own task. The measure is sensitive to this difference. If the manager simply respond to demands on his time (s_ij > 0), but does not seek out this particular contact (s_ij = 0), his proportional consultative tie with the contact (S_ij) would be weak. As we shall see, such weak tie need not be detrimental. The tie, however, would become stronger if the manager also seek this particular contact out for consultation (s_ij > 0).

Structural holes. Lack of structural holes in the network is one of the two conditions we expect may curtail the managers' ability to adapt their social capital. Following Burt (1992), we measure lack of structural holes in the manager's network the as the sum of the constraint posed by each of the contacts in the network. This constraint is in turn a function of the direct consultation between i and j and of the extent to which j consults with the other contacts q in i's network (see Burt 1992: 50-71, for detailed discussion of these measures):

where S_ij is the proportional measure of consultation ties discussed before and _jc_ij captures the lack of structural holes in manager i's network.

Network hierarchy. Lack of structural holes can also result from a consultation network built around strong ties to one or few players who occupy a central position in the manager's network. We measure the level of informal "hierarchy" in the managers' consultation networks as a function of the distribution of the constraint across the manager's contacts. Concentration of constraint in dominant contacts is captured by the Coleman-Theil disorder index. ₇ The closer the index gets to 1.0, the more a single contact j dominates the manager's consultation network. Contact j's dominance is a function of the strength of the consultation between the manager i and j (Sij), and of the extent to which j is also consulted by the other contacts q with whom i also consults (Sjq Sqi).

Coordination Failures. This is the dependent variable in the analysis. We define as coordination failures those cases in which high (above average) task interdependence Tij between two DPI managers is coupled with low (below average) consultation Sij between these managers. ₈ As we noted before, DPI managers had a large number of consultation ties, which in all but one case exceeded the number of their task interdependencies. In this setting, the relative strength of the consultative tie is relevant. By focusing on the joint occurrence of high interdependence and low consultation, we capture cases of strong task interdependence that are coordinated by a weak consultative tie--that is, a tie weaker than the expected level of consultation among any two managers in the unit. Following this criterion, each of the 342 dyads in the network was coded as either failing (1) or not failing to coordinate (0). On the basis of this coding, we computed an individual failure measure for each manager, defined as the sum of coordination failures across all the people with whom he jointly participated in at least one project team. This measure may vary from 0 when the manager had successfully allocated above average consultation ties to all his strong interdependencies to a maximum of N when he fails to do so for all his strong interdependencies, being N the number of above average task interdependence ties within DPI. For example, one of the managers worked with 13 different colleagues, 7 of which were linked by strong task interdependence to this manager. His level of consultation with these nine colleagues, however, was below average in 4 of the cases, which we coded as coordination failures. On average, DPI managers committed 3.68 failures in coordinating 10.63 task interdependencies.

Coordination failures had a noticeable negative impact on the level of cooperation attained in the project teams, hence affecting DPI's ability to attain its goals. Failure to coordinate task interdependencies between DPI managers facilitating a project team hindered their ability to cooperate with other team members and jeopardized their role as catalyzers of cooperation. The main goal of the DPI managers within project teams was to promote cooperation among the other members of the team, who often came from different areas of the organization. Yet, coordination failures between DPI managers jointly participating in a team prevented them from attaining satisfactory levels of cooperation within that team. The proportion of coordination failures within DPI dyads in a project team was negatively correlated with the average level of attained cooperation among all team members (r = -.420; p < .001). The negative impact of coordination failures was also significant for both the average cooperation between DPI managers and other members of the team (r = -.476; p < .001), as well as for the cooperation among non-DPI team members (r = -.395; p < .001). ₉

Control variables. We have controlled by factors that may affect a manager's tendency to incur in coordination failures. Specifically, we looked at the number of weak consultative ties in a manager's network and to his workload. High failure scores may result from a manager's tendency to have a large number of "weak" consultative ties - that is, ties that are below the average strength of consultation in the unit. A larger number of weak consultative ties can result from having a large network of contacts, which in turn may translate into a larger number of coordination failures. Failure, however, was not significantly associated with the size of the manager's consultation network (r = .200; p = .206), nor with the number of ties to colleagues the manager did not work with in a project team (r = .197; p = .209). The effect of weak ties may be more troublesome in hierarchical networks. By definition, such networks have a large number of weak ties and a few very strong ones. Unless the distribution of task interdependence is equally skewed, this profile may automatically result in a large number of coordination failures. The mechanism at work, however, would not be related to the dominant role of the strong-tie contacts, as proposed in this paper. A significant association between hierarchy in the consultation network and the number of coordination failures could be simply an artifact of the number of weak ties in the consultation network, rather than a test of the impact of network structure on the manager's ability to freely allocate his consultative ties.

To rule out the possibility that the correlation between coordination failures and network structure was simply an artifact of the number of weak ties in the consultation network, we removed the effect of the number of weak ties from the original variables and re-estimated the models using the adjusted scores. We computed adjusted measures of coordination failures, lack of structural holes (constraint), and network hierarchy. The adjusted measures are the standardized residuals obtained by regressing failure, constraint, and hierarchy on the number of weak ties in the manager's consultation network. Consultation ties whose proportional strength (S_ij) was below the criterion level used to define coordination failures were coded as "weak." ₁₀ The adjusted measures are thus free from any spurious influence of the number of weak ties in a manager's consultation network. If our theory on the encumbering effects of network structure on the ability to allocate consultative ties is correct, both lack of structural holes and hierarchy in the consultation network should still have an impact on coordination failures once the effect of weak ties is removed. On the contrary, if the alleged impact of the structural variables on coordination failures is simply an artifact of the presence of weak ties, the impact of the structural variables should disappear after the effect of weak ties is removed from the measures.

A manager's failure to maintain adequate consultation ties with people he depends upon can also be affected by the particular workload conditions in which this manager has to perform his task. The larger a manager's workload, the higher the probability of failing to keep an adequate level of coordination for all the task interdependencies defining his role, and thus the higher the number of failures. Two workload dimensions are particularly relevant here: the number of project teams in which the manager participated and the number of different colleagues with whom the manager jointly participated in at least one project team. As one might expect, these two dimensions are strongly correlated (r = .744; p < .001), since the number of task interdependencies typically increases with the number of project teams in which the manager participates. This correlation creates multicollinearity problems that make it impossible to have both controls in the equation simultaneously. To circumvent that problem, we estimated two separated models. We expect both variables to have a significant positive impact on failure rates. We are not, however, interested in this impact, but only on the changes in the effect of the structural variables after controlling for workload conditions. Descriptive statistics and zero-order correlations between all these variables are included in the Appendix.

Results

Table 1 presents standardized regression coefficients measuring the impact of the structure of the managers' networks on their rate of coordination failures. The results furnish evidence supporting our predictions. Lack of structural holes in a manager's consultation network increases the number of coordination failures committed by this manager. The impact of lack of structural holes is still strong after removing the effect of weak ties, which suggests that the number of weak ties in the mangers' consultation network play a marginal role in accounting for the number of coordination failures. Rather, these failures are the result of how those weak ties are allocated to the different task interdependencies. Managers with network lacking structural holes were more likely to allocate weak consultation ties to strong interdependencies and strong consultation ties to weak or non-existent interdependencies, which prompted coordination failures. Similar results are obtained using hierarchy in the network. The distribution of constraint in a manager's consultation network affects the way in which the manager allocates his weak ties across levels of dependence. The more concentrated the constraint, the more likely were the managers to allocate weak consultation ties to strong task interdependencies. The increments in R-square for the models with control variables show that workload conditions were an important factor contributing to managerial failures. The effects of the structural variables, however, were not substantially affected by the introduction of the control variables. This suggests that those effects are largely independent of the managers' workload.

Both lack of structural holes and network hierarchy have comparable effects on coordination failures. The high correlation between these two variables (r = .661 for the standard measures and r = .465 for the adjusted measures; p < .01), coupled with the small number of observations, create multicollinearity problems that made it difficult to evaluate the relative impact of each of these structural properties on coordination failures. Attempting an approximation to such an evaluation, we obtained instrumental measures of both hierarchy and level of constraint by removing the effect of the level of constraint from hierarchy, and vice-versa. Regressing failure on constraint and on the instrumental hierarchy measure, we obtained estimates for the impact of each variable on coordination failures. A similar estimation was done using the instrumental measure of structural holes and hierarchy. The results, however, do not allow for an empirical evaluation of differences in the impact of each structural property on coordination failures. A comparison of the corresponding standardized regression coefficients shows that the impact of the instrumental measure of hierarchy is similar to the impact of the instrumental measure of constraint. ₁₁

Going one step further in the analysis, we explored the sources of variation in the shape of the manager's consultation networks. More specifically, we investigated individual and organizational sources of constraining relationships. To this end, we examined constraint scores at the dyad level and sought to identify correlates of highly constrained relationships. We used the marginal constraint posed by each contact relatively to the average constraint in the manager's network. Contacts posing high constraint are strongly tied to the focal manager and to this manager's other contacts. ₁₂

We found no significant association between relative constraint and similarity in background variables such as educational level, major, seniority, or age. The average DPI manager was neither more nor less likely to be constrained by a colleague like him or herself, when similarity is defined along these personal attributes. Relative constraint was also unrelated to formal reporting relationship. There was, however, an association between relative constraint and organizational background, with average relative constraint significantly higher among people who were together in the same unit before being assigned to DPI (1.054 vs .898; 2.66 t-test; p < .01). Similar results were obtained using raw constraint scores (c_ij) or the relative strength of the tie (S_ij). The dominant players in the managers' networks were more likely to be the people the managers used to work with in their previous assignment in the organization. The relatively smooth transition between those previous assignments and DPI, which in some cases even implied a continuation of the same project work, with the same people, did not always signal a change in the task environment of these managers. As DPI started its own projects, the task interdependencies of the managers shifted, but the change went unnoticed for a number of them, still faithful to consultative relations built over years in the firm.

This finding confirms some well-known ideas about the origin of strong relationships. Granovetter (1973: 1361) has pointed out that the strength of a tie " ... is a (probably linear) combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services that characterize the tie." Looking at the origin of ties, Feld (1981) stressed joint participation in similar organizational contexts as one of the main sources of relationships. Common organizational history put people in contact, prompts the exchanges of advice and services, and allows for repeated exchanges that are the basis for a strong relationship. Our analysis suggests that these relationships are likely to outlive the specific context in which they arise. If the change of context is not drastic enough--such as physical relocation--the old relationship may remain strong. This was particularly so in the case of DPI. A large number of the observed managers were previously working together in the unit that preceded DPI and initially continued to do so in the new unit, often in similar projects, even if new colleagues often joined the teams. The smooth transition did not signal a clear change in the task environment for these managers, who continued to consult with their old colleagues even when the new projects created interdependencies with managers coming from other parts of the organization.

To summarize: Our analysis of the DPI managers furnished evidence supporting the propositions discussed in this paper. Lack of structural holes in the network that defines social capital make it difficult for the manager to renew this capital as required in a changing task environment. The results are independent of the potentially spurious influence of the number of weak ties in the consultation networks and from differences in workload conditions across managers. The results, however, did not shed light on the relative importance of the two ways in which a constraining network structure may affect a manager's ability to renew his social capital. Both lack of structural holes and the presence of dominant contacts in the manager's network equally jeopardize his ability to adapt his social capital. An examination of the sources of the constraining relationships responsible for the lack of structural holes in the managers' networks revealed that those relationships typically corresponded to ties forged through years of work in similar organizational environments. The strength of those bonds, and the lack of tangible signs of change in the managers' task environment, fuelled the mechanisms of reciprocity and relational inertia that lead to coordination failures.

Discussion

This paper draws upon the control implications of structural hole theory to explore the dark side of social capital. We explain how the structure of a manager's contact network may affect his ability to adapt the content of his social capital. Faced with a growing number of task interdependencies that are mostly beyond his administrative control, a manager must develop informal mechanisms to effectively coordinate the uncertainties surrounding these tasks. In this context, a manager's social capital is a critical resource. Not any kind of social capital structure, however, will be equally useful in such an environment. Effective mangers should be able to strategically adapt the composition of their social capital to help coordinate shifting interdependencies. To do so, however, they must have the autonomy to develop, maintain, and leverage relationships, thus keeping control on how they allocate time and energy to their contacts. Factors that reduce such autonomy increase the likelihood of failures in coordinating critical task interdependencies, which in turn may have a negative impact on performance. Our research illustrates the impact of two such factors. A consultation network formed by strong ties with a set of cohesive contacts, or a network dominated by a small set of contacts, constrain the manager's ability to adapt the composition of his social capital according to the changes in his task environment. Conditions that facilitate the creation of strong ties with densely connected clusters, such as a common organizational history, increase the probability that managers will build constraining networks, hence affecting their subsequent ability to renew their social capital.

Our research has consequences for the debate on social capital, as well as for its effects on managerial and organizational performance. By focusing on factors that may hinder the renewal of social capital in dynamic task environments, we show how the network structure that carries social capital has an effect on the manager's ability to renew the composition of this social capital. Network structure, network composition, and outcomes are interdependent, but the relationship between these three variables is problematic. Early studies of social capital showed how the composition of an individual's contact network had an impact on the benefits accruing to this individual (Lin, Ensel, and Vaughn 1981; Lin and Dumil 1986; Flap and de Graaf 1989). More recently, scholars have focused on network structure as the motor behind those benefits, assuming implicitly or explicitly that network composition is a simple correlate of network structure (Burt 1992; Podolny and Barron 1997). By focusing on the renewal of the content of social capital, our research clarifies the relationship between network structure and network composition, elucidating the mechanisms through which this relationship operates. Our results suggest that both ties to a cohesive cluster and strong, exclusive ties to few contacts are conditions may turn the assets of social capital into liabilities that damange a manager's adaptability and performance.

The significance of our finding is enhanced by the fact that the relationship between network structure and favorable outcomes is not stable across contexts. If strong ties to a cohesive group were always detrimental for the performance of a manager, our results simply would add another spin to this deleterious effect. Yet, this is not always the case. Existing research suggests that the effects of strong ties to cohesive clusters on managerial performance may depend on the particular situation of the manager. In his analysis of managers in a U.S. high-tech firm, Burt (1992:147-153) found that strong ties to prominent contacts in a manager's network was the best route to fast promotion for women and entry-level men. He suggests that such ties may be a necessity for a manager in need of a legitimating sponsor. This positive effect of what amounts to an informal "network hierarchy" on promotion captures the well-documented influence of mentoring in managerial careers (e.g., Kram 1988). With a different emphasis, Podolny and Barron (1997) argued that strong ties to cohesive contacts linked to a manager by authority relations may also be beneficial in early stages of their career, when defining an identity and obtaining support from powerful players is critical for the manager. A manager with strong ties to a cohesive group may attain the necessary legitimacy to be recognized as a player. He may also profit from the additional social capital he can "borrow" from his sponsors. These benefits, however, may come at a price. Our evidence suggests that strong ties to a cohesive group may eventually curtail a manager's ability to adapt the composition of his social capital, even if those ties were initially beneficial for the manager. The difficulty in renewing the composition of his social capital affects the managers' ability to coordinate task interdependencies with and among other members of the organization, hence having a deleterious effect on his performance.

Our findings, therefore, pose an interesting dilemma for mangers as well as for organizations. On the one hand, managers at the early stages of their career may need to obtain decisive informal sponsorship to become legitimate players (Burt 1992) or to assert their identity in the organization (Podolny and Barron 1997). This legitimacy requirement is compounded by the fact that the lack of structural holes is less detrimental for entry level managers, where the benefits of a diverse contact network may be negligible (Burt 1997). A small, cohesive core of supportive contacts may be the best form of social capital for these managers. On the other hand, the managerial career eventually requires autonomy to allocate time and energy to different contacts to successfully cope with the shifting task interdependencies associated to the role. Yet, the same cohesive social capital that was instrumental in asserting the manager's position in the organization may be an obstacle to develop the type of social capital required to further his professional growth and his capacity to add value within this organization. Our argument suggest that this peril may be more consequential when the new task is not accompanied by tangible changes in reporting line or physical location that naturally weaken the manager's previous relationships.

Ideally, managers should be able to go through a smooth transition between these two stages in their careers, adapting both the structure and the composition of their social capital to the changes in their task environment. The human resource practices and the culture of the organization should also support this transition. In practice, the transition may not be always as easy. The initially supportive, legitimating sponsorship of a cohesive cluster or a strong mentor may translate into a liability that hinders the manager's ability to adapt his social capital. For the manager, there is no easy way out of this dilemma. For the organization, the solution lies in efficiently combining the benefits of sponsorship with the benefits of autonomy. Factors that make the transition more noticeable, such as physical or departmental relocation, may help avoiding the perils of the dark side of social capital discussed in this paper. Yet, these systemic solutions may not be always feasible or even desirable. Another, perhaps complementary way to deal with the problem is to clarify the role of organizational sponsors and to build in the ability to promote young "entrepreneurial" managers into their performance evaluation. Like a wise father, a wise sponsor must know not only how to help and defend his protŽgŽ, but also to allow him the necessary autonomy to grow independently.

Finally, the research reported in this paper has implications for the study of the managerial role in dynamic, flat organizational environments. The attenuation of authority embedded in the idea of "flat" organizational structures may not suffice to promote entrepreneurship within a traditional organization. This is especially so in context of rapid organizational change. Like the bird that remains inside the cage despite the open door, a manager adapted to a command-and-control structure may fail to use the freedom created by flatter structures. Scholars have recognized the role of previous socialization in this inability to take advantage of the new freedom. Our argument suggests that the freedom itself may be sometimes a mirage. Powerful control mechanisms embedded in informal structure and rooted in the organizational history can create bonds of interpersonal dependence that may severely curtail the organization's ability to move away from a command-and-control managerial style. Although necessary, the attenuation of authority is not a sufficient condition to promote managerial entrepreneurship. Hierarchies ingrained in the informal organization may still create an effective obstacle to flexibility. The unobtrusive nature of these ties may make their effect less apparent than the conspicuous impact of the formal hierarchy. Such an effect, however, may be equally consequential.

This warning is especially relevant in contexts of the transformation into a flat, entrepreneurial organization. Moving away from the well-established command-and-control structures is a task that requires more than administrative decisions. Top management decisions of reshaping the overall organizational through a general reduction of managerial levels are important. However, such decisions may not be enough to promote entrepreneurship. Past organizational characteristics can survive, even if formal structures have been revised or the chain of command reduced. A manager cannot act as a flexible network entrepreneur if the informal organizational structure does not allow it. Yet, the emergence of this structure is largely the result of effective managerial initiative. Informal structures that emerge out of established managerial practices can have a direct impact on the subsequent behavior of those managers, which in turn may affect the organization's ability to effectively enact change.

Conclusion

The research reported in this paper shed light on the "dark side" of social capital by discussing when and how social capital may become an obstacle to managerial and organizational performance. We show how the social capital embedded in strong, cohesive social bonds may hinder the necessary renewal of social ties to fit changes in the manager's task environment. We suggest that the risk can be higher if those strong ties have initially produced substantial advantages for the manager, a condition that defines the effectiveness of social capital. This paradox reminds us that social capital, like other properties linked to social structures, may have both positive and negative consequences for individual and organizational action. Like most things, social capital has a bright side, but also a dark one.

References

Baker, Wayne 1992 "The network organization in theory and practice." In Nitin Nohria and Robert G. Eccles (eds.) Networks and Organizations: Structure, Form, and Action. Boston, MA: Harvard Business School Press. Pag. 397-429.

Bearman, Peter 1997. "Generalized Exchange". American Journal of Sociology 102:1383-1415.

Becker, Gary S. 1964. Human Capital. New York: Columbia University Press.

Benassi, Mario 1993. Dalla gerarchia alla rete: modelli ed esperienze organizzative. Milano: Etas .

Blau, Peter 1955. The Dynamics of Bureaucracy. Chicago: The University of Chicago Press.

Blau, Peter 1964. Exchange and Power in the Social Life. New York: Wiley.

Bonacich, Philip 1987 "Power and centrality: a family of measures." American Journal of Sociology 92: 1170-1182.

Bourdieu, Pierre 1980 "Le capital social" Actes Sociologiques.

Bourdieu, Pierre and Loic J.D. Wacquant 1992. An Invitation to Reflexive Sociology. Chicago: University of Chicago Press.

Burt, Ronald S. 1992. Structural Holes. The Social Structure of Competition. Cambridge, Mass.: Harvard University Press.

Burt, Ronald S. 1997. "The contingent value of social capital." Administrative Science Quarterly 42:339-364.


Burt, Ronald S. and Marc Knez. 1995. "Kinds of third-party effects on trust." Rationality and Society 7:255-292.

Campbell, Karen; Peter Marsden, and Jeanne S. Hurlbert 1986 "Social Resources and Socioeconomic Status." Social Networks 8:97-117.

Coleman, James C. 1990. Foundations of Social Theory. Cambridge, MA: Harvard University Press.

Cook, Karen S. and Richard M. Emerson 1978. "Power, equity and commitment in exchange networks." American Sociological Review 43:721-739

Cook, Karen S. Richard Emerson, Mary Gillmore, and Toshio Yamagishi 1983. "The distribution of power in exchange networks: theory and experimental results." American Journal of Sociology 89:275-305.

Ditcher, Steven F. 1991 "The organization of the '90s." The McKinsey Quarterly, 1: 145-155.

Emerson, Richard M. 1962. "Power-dependence relations." American Sociological Review 27:31-40.

Feld, Scott L. 1981 "The focused organization of social ties." American Journal of Sociology 86:1015-1035.

Fischer, Claude 1982. To Dwell Among Friends. Personal Networks in Town and City. Chicago: University of Chicago Press.

Flap, Hendrik D. and Nan D. De Graaf 1989. "Social capital and attained occupational status." Netherlands Journal of Sociology, 22:145-161.

Gargiulo, Martin 1993 "Two-step leverage: Managing constraint in organizational politics." Administrative Science Quarterly 39: 1-19.

Grabher, Gernot 1993. "The weakness of strong ties. The lock-in of regional developmente in the Ruhr area" in Gernot Grabher, (Ed.): The Embedded Firm. London and New York: Routledge,

Granovetter, Mark 1973. "The strength of weak ties." American Journal of Sociology 78:1360-1380.

Granovetter, Mark 1974. Getting a Job. Boston: Harvard University Press.

Granovetter, Mark 1985. "Economic Action and Social Structure: A Theory of Embeddedness." American Journal of Sociology 91:481-510.

Gulati, Ranjay. 1995a. "Social structure and alliance formation pattern: A longitudinal analysis." Administrative Science Quarterly 40:619-652.

Gulati, Ranjay. 1995b. "Familiarity breeds trust? The implications of repeated ties for contractual choice in alliances." Academy of Management Journal 38:85-112.

Hannan, Michael T. and John Freeman 1989 Organizational Ecology. Cambridge, MA: Harvard University Press.

Kanter, Rosabeth M. 1984 The Change Masters. Innovation and Entrepreneurship in the American Corporation. New York: Simon amd Schuster.

Kotter, John 1982. The General Managers. New York: Free Press.

Kram, K. E. 1988. Mentoring at Work. Developmental Relationships in Organizational Life. Lanham, MD: University Press of America

Leifer, Eric 1988. "Interaction preludes to role setting: Exploratory local action." American Sociological Review 1988: 865-878

Lin, Nan and Mary Dumil 1986. "Access to occupations through social ties." Social Networks 8:365-385.

Lin, Nan, Walter Ensel, and John Vaughn 1981. "Social resources and strength of ties: Structural factors in occupational status attainment." American Sociological Review 46:393-405.

Marsden, Peter V. 1984. "Restricted access in networks and models of power." American Journal of Sociology 88:686-717

Marsden, Peter. V. 1982. "Brokerage behavior in restricted exchange networks." In Peter. V. Marsden and Nan Lin (eds.). Social Structure and Network Analysis. Beverly Hills, Ca.: Sage.

Miles, Raymond A. and Charles S. Snow 1992. "Causes of failures in network organizations." California Management Review, 34: 53-72.

Nohria, Nitin 1992. "Is a network perspective a useful way of studying organizations? In Nitin Nohria and Robert Eccles (Eds.). 1992. Networks and Organizations. Boston, MA: Harvard Business School Press.

Pfeffer, Jeffrey and G. Salancik 1978. The External Control of Organizations: a Resource Dependence Perspective. New York: Harper & Row.

Podolny, Joel m. and James N. Barron 1997. "Relationships and resources. Social networks and mobility in the workplace." American Sociological Review, 62 (in press).

Portes, A. and J. Sensenbrenner 1993. "Embeddedness and Immigration: Notes on the Social Determinants of Economic Action." American Journal of Sociology 98: 1320-1350.

Powell, Walter W., and Laurel Smith-Doerr. 1994. "Networks and economic life." Pp. 368-402 in The Handbook of Economic Sociology, edited by Neil Smelser and Richard Swedberg. Princeton, NJ: Princeton University Press.

Raub, Werner, and J. Weesie. 1990. "Reputation and Efficiency in Social Interactions: An Example of Network Effects." American Journal of Sociology 96(3): 626-654.

Sheppard, Blair H. and Marla Tuchinsky 1996. "Micro-OB and the network organization." In Roderick M. Kramer and Tom R. Tyler (eds.) Trust in Organizations. Thousand Oaks, CA: Sage.

The Economist 1993. "Within the wirldwind. A survey of the computer industry" (February 27, 1993).

Uzzi, Brian 1997 "Social structure and competition in interfirm networks: The paradox of embeddedness." Administrative Science Quarterly 42:35-67.

Appendix

Notes

*: We would like to thank helpful comments by Ron Burt, Ranjay Gulati, David Gibson, Andrej Rus, Harrison White, and the participants in the Network Analysis Workshop at the Lazarsfeld Center for the Social Sciences at Columbia University. Back.

Note 1: This phenomenon has been documented in studies of organizational adaptation, which show how the same processes that help organizations to be well adapted to their current environment can curtail their ability to adapt to a new environment (Grabher, 1993). A similar idea plays a central role in ecological models of organizations (Hannan and Freeman 1989). These models argue that "specialists" firms--that is, organizations highly adapted to specific environments--are less likely to survive drastic environmental changes than "generalists" organizations, since the latter are less dependent on specific resources for their survival. Back.

Note 2: Leifer (1988) and Bearman (1997) note how the norm of reciprocity, when applied to dyadic exchanges, may lock the players into endless exchanges due to ambiguity on the valuation of the "gifts" exchanged. A similar argument can be made regarding exchanges with a cohesive group. In this case, the presence of third parties reinforces compliance, thus increasing the likelihood of people entering such endless exchanges. Back.

Note 3: Structural hole theory builds upon a key intuition of exchange theory, which states that control in a relationship is a positive function of the availability of alternatives (Emerson, 1962; Blau, 1964). A series of simulation and experimental studies confirm the adequacy of this basic intuition. Players with access to several exchange partners who themselves lack such alternatives enjoy competitive advantage (Cook and Emerson, 1978; Marsden, 1982, 1984; Cook, Emerson, Gillmore, and Yamagishi, 1983). Back.

Note 4: It is worth noting that our treatment of social capital differs from that of structural hole theory, which makes social capital a sole function of the structure of the network: the more structural holes, the larger the social capital. Burt (1992:13) defends his analytical choice by assuming that actors who know how to structure a network to provide access to opportunities would also know whom to include in that network. This assumption provides an elegant solution to the tension between the constraint and opportunity aspects of social capital, but it also restricts the applicability of the concept. An individual strongly tied to a dense cluster of resourceful people has little social capital in Burt's sense. Yet, in some situations, strong connections with few strategic players may be actually beneficial for a manager (Burt 1992:147-153; Podolny and Barron, 1997). This effect is predicted by the more traditional definition of social capital adopted here, but it does not fit easily within structural hole theory. More importantly for this paper, Burt's assumption dissolves the problem of adaptation of social capital that is central to our discussion. Paradoxically, this drives the attention away from the fruitful control implications of structural hole theory developed here. Back.

Note 5: The prominence scores used here are Bonacich's (1987) eigenvector measure. They reflect a manager's tendency to be consulted by colleagues who are also preferred targets for consultation. The measure varies from a theoretical maximum of 1, for the most prominent actor(s), to near 0, for the least prominent actors. In the consultation network, prominence scores varied from 1.00 to .594, with an average of .791 and a standard deviation of .118. Back.

Note 6: The inclusion of both types of consultation is relevant in this context, since a manager may still have to allocate time to colleagues who consult him, even if he does not seek these people out. Back.

Note 7: The Coleman-Theil index is calculated as follows:

where C is the lack of structural holes (aggregate constraint) on actor i ( _jc_ij) and N is the number of alters in i's network. Note that the quotient C/N is the average constraint posed by i's alters, and thus c_ij/(C/N) expresses the constraint posed by actor j in terms of the average constraint in i's network. Back.

Note 8: he mean strength of task interdependence between managers is .056, with a median of .038 and a skewness coefficient of 2.035. By using the mean as a criterion to define strong task interdependence we leave aside about 60 percent of the cases. A low level of consultation with a sporadic project partner will not constitute a coordination failure. Consultation ties are normally distributed (.056 mean, .057 median; .033 skewness). The proportional measures of task interdependence (T_ij ) and consultation (S_ij) used in this paper removes differences in the volume of interaction of individual managers, making average figures a good indicator of the socially expected levels of interdependence and consultation between any two managers in the unit. Back.

Note 9: At the project level, we defined a coordination failure index as the ratio between the number of DPI dyads in the project team who failed to coordinate and the total number of DPI dyads in the project. Projects with a single DPI participant were coded as having zero failure rate. There were 65 projects in which at least a member of a unit other than DPI participated. Twenty-two of them have a single DPI participant. Cooperation is measured as the average level of cooperation reported by the manager coordinating the project. In principle, this measure may vary from 0 (no cooperation among the team members) to 3 (the cooperation among all team members was the strongest). Back.

Note 10: On average, DPI managers had 8.90 weak ties, ranging from 4 to 12. As expected, the number of weak ties in the consultation network was significantly correlated with the three variables, that is, number of failures (r = .533; p < .05), network constraint (r = .488; p < .05), and the hierarchy (r = .682; p < .01). These high correlations ruled out an alternative approach to control for the effect of weak consultative ties on coordination failures by simply including the number of weak ties into the regression. Back.

Note 11: These results were computed using the adjusted measures of our variables and thus are independent of the number of weak ties in the manager's consultation network. None of the coefficients for the instrumental measures were significant at the .10 level. Back.

Note 12: Relative constraint, given by the ratio c_ij/(C/N), is largely responsible for the level of hierarchy measured by the Coleman-Theil index (see footnote 7). Back.