Scholarly article on topic 'The Performance of Policy Networks: The Relation between Network Structure and Network Performance'

The Performance of Policy Networks: The Relation between Network Structure and Network Performance Academic research paper on "Sociology"

Share paper
Academic journal
Policy Studies Journal
OECD Field of science

Academic research paper on topic "The Performance of Policy Networks: The Relation between Network Structure and Network Performance"

The Performance of Policy Networks: The Relation between Network Structure and Network Performance

Annica Sandström and Lars Carlsson

The importance of policy networks has long been emphasized within the field of policy analysis. However, few attempts have been made to investigate the explanatory power of policy networks using the tools and theoretical concepts provided by social network analysis (SNA). This paper aims to address this need by determining if a relationship exists between the structural features of policy networks, their organizing capacities, and their performance. A comparative case study of four networks within the higher education policy sector confirms the assumption related to the existence of such a relation. It is proposed that an efficient and innovative policy network consists of a heterogeneous set of actors that are centrally and densely integrated. Furthermore, while the level of network heterogeneity is positively related to the function of resource mobilization in the process of policymaking, the level of centralized integration promotes the function of prioritizing. These findings are believed to contribute to our understanding ofpolicymaking in contemporary society. The current paper indicates that a significant explanatory power exists in the concept ofpolicy networks and that SNA is one way of advancing its possibilities.

KEY WORDS: policy analysis, policy networks, social network analysis, SNA, network performance, network structure

Treating Networks Seriously1

Networks and networking have become key concepts that are frequently applied in the world of authoritative policymaking and public administration, as well as in the scientific community analyzing and evaluating policy processes and their outcomes (Bogason & Toonen, 1998; Borzel, 1998; Kickert, Klijn, & Koppenjan, 1997; Koppenjan & Klijn, 2004; Marin & Mayntz, 1991). The emphasis on networks is truly driven by an increasing complexity characterizing the modern state. Many policy problems are considered far too multifaceted to fit the problem-solving structures of traditional government. For instance, although borders between both different levels of governmental units and different policy sectors are administratively defined, societal problems are characterized by their interdependent and cross-scale nature (Berkes, 2002; Hanf & Scharpf, 1978; Koppenjan & Klijn, 2004; Scharpf, 1991). An increasing aspect of the challenges that political decision makers face is the requirement for solutions that can only be obtained by the establishment of interor-

0190-292X © 2008 Policy Studies Organization Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington Road, Oxford, OX4 2DQ.

ganizational collaboration. The creation of networks crossing a multitude of formal organizational borders and hierarchical levels is considered a necessity whether policy concerns the task of achieving high-quality education and research, local economic development, or activities in any other policy area. A deepening of our understanding concerning how these cross-boundary networks operate and how their qualities might be related to success and failure of policymaking has long been—and still is—emphasized as vital for the task of improving the effectiveness of public management (Agranoff & McGuire, 2001; Hanf & O'Toole, 1992; Hanf & Scharpf, 1978; O'Toole, 1997). Since an essential aim of policy science is to provide insights that make it possible to improve this task, research on such networks and related processes is considered imperative to the field. It is a well-known fact that different ways of organizing a policy process are associated with different results. In this paper, the potential relationship between the structural features of policy networks on the one hand and their organizing capacities and performance on the other is examined using the tools and theoretical concepts provided by formal quantitative social network analysis (SNA).

The policy network approach has its scientific roots in organizational studies, political science, and policy science, and shows immense variety in regard to both how the concept of policy networks is comprehended and what a network approach to policy analysis really implies (Borzel, 1998; Carlsson, 2000; Kickert et al., 1997). The approach incorporates a rich variety of concepts, including advocacy coalitions (Sabatier & Jenkins-Smith, 1993), implementation structures (Hjern & Porter, 1993), iron triangles (Jordan & Schubert, 1992), issue networks (Heclo, 1978), policy communities (Jordan, 1990), and subgovernments (Rhodes, 1990). For policy researchers, particularly those engaged in implementation studies, the recognition of multiactor networks in the study of political activities as well as the need to "treat these networks seriously" is far from new. The understanding of policy and policy networks adopted for the purpose of this paper draws on this line of research. Policy is considered "a set of ideas and the practical search for institutional arrangements for their realization" (Hjern, 1987, p. 3). Adopting this definition, it becomes obvious that policies are not only decided, but they also emerge as people organize to realize their ideas. Thus, the making of policy reflects organizing processes that occur within networks of collective action; in other words, it is a series of "actions taken by members of a group to further their common interests" (Bogdanor, 1987, p. 113). Accordingly, policy networks can be perceived as "problem-specific entities, organizing a policy area by different forms of collective action" (Carlsson, p. 508). In acknowledging the fact that these networks might not correspond to formal hierarchical structures and therefore need to be empirically defined (Hjern, 1987; Hjern & Porter, 1993), a bottom-up approach to the study of these networks is emphasized. The bottom line proposed here is that differences in policy outcome can be attributed to and explained by the characteristics of these policymaking structures.

The perspective adopted in this paper is that policy networks are organized entities that consist of actors and their relations engaged in processes of collective action for joint problem solving. Consequently, these structures are the relevant analytical units to consider in the search for explanations to the success and failure of

policymaking. It should be acknowledged that the notion of policy networks adopted herein is somewhat more restricted than what is implied by some of the other concepts referred to earlier (advocacy coalitions, iron triangles, etc.). For example, divergences emerge in regard to the definition and delineation of policy networks (what the coordinating functions consist of), their durability, and their relation to formal policymaking and government authorities.

In this paper, the mere structure of policy networks—meaning the constellation of actors and the patterns of their interactions—is treated as the main variable in explaining policy outcomes and performance. The goal of this paper is to address a knowledge gap commented on by scholars engaged in the field. Critical voices have claimed that despite the significant attention given to the topic of networks, research has thus far proven unable to move beyond a metaphorical usage of the concept and toward a coherent theory (Carlsson, 2000; Dowding, 1995). The concept of networks has typically been used as an analytical device or as a descriptive metaphor illustrating a general phenomenon manifested as sets of related actors engaged in various kinds of political activities, often at the macro level. It is well accepted that although metaphors are heuristically helpful devices, they should not be treated as theoretical models capable of explaining policy change or outcome (Dowding, 1995). Research has been "insufficiently engaged in using network analysis to explain action and its consequences" (Raab, 2001, p. 554). This study is based on the assumption that theory development requires research that treats policy networks as independent variables rather than dependent variables, which has become common in policy research. Such research has the potential to outline what could be expected from different kinds of network arrangements. The metaphorical approach to policy networks is incapable of this assignment. Obviously, more is needed than to say that networks matter. The challenging question is, in what way do they matter? The formalized SNA, which is a method providing the means to map and quantify social relations, as well as visualize and analyze network structures mathematically, has previously been identified as having significant potential for this branch of research (see, for example, Dowding, 1985; Peters, 1998; Thatcher, 1998, in Adam & Kriesi, 2007).

Despite the indicated lack of studies, research has progressed in a way that we now know more about the structural effects on policymaking and how SNA might be applied in policy science (see, e.g., Daugbjerg, 1998; Human & Provan, 1997, 2000; Kriesi, Adam, & Jochum, 2006; Meier & O'Toole, 2001; O'Toole & Meier, 2004; Provan & Milward, 1995, 2001; Provan & Sebastian, 1998; Raab, 2002; Real & Hasanagas, 2005; Schneider, Scholz, Lubell, Mindruta, & Edwardsen, 2003). These studies relate, in various aspects, to the central undertaking in this paper. The seminal study by Provan and Milward (1995), for example, successfully moved the concept of networks beyond a metaphorical stage and policy research one crucial step forward. Applying SNA, the authors analyzed implementation structures within the community mental health system in the United States and brought forward the idea that certain aspects of network structure are decisive for the performance of these systems. Although additional studies raise similar issues, the amount of projects addressing this topic empirically remain limited, making the current influence on

policy research rather insignificant (Adam & Kriesi, 2007). SNA in the study of network performance has been more frequently applied in work associated with other scientific subjects, such as sociology, organizational studies, and business administration (see, e.g., reviews conducted by Borgatti & Foster, 2003; Burt, 2000; Flap, Bulder, & Völker, 1998). Many of these studies relate to the concept of social capital and deliberate on the idea that certain network structures generate higher social capital and are therefore better performing2 (Borgatti, Jones, & Everett, 1998; Burt, 2000). Contemporary policy research could potentially benefit from the underlying assumptions and propositions generated from this work. Hence, not merely the rise and existence of networks are important for the organizing process of policymaking, but also their structural qualities. In sum, the increasing importance that networks have been ascribed in public policymaking, coupled with the lack of systematic knowledge about these structures and the possibilities provided by SNA, constitutes the reasons for and justifies the scope and aim of this paper.

The aim of this paper is to elucidate the relationship between network structure and network performance of policy networks using SNA techniques. Does a relationship exist AMONG the structural qualities of policy networks, their organizing capacities, and their performances? Drawing on previous policy research, institutional theory, and the concept of social capital, an analytical skeleton framing the questions why and in what way structure might be related to performance is presented. For this purpose, a set of general hypotheses are formulated. The study sets out to test and refine these hypotheses—concerning structure and performance— empirically. The empirical data are based on a multiple comparative case study of four networks within the higher education policy sector that are related to the realization of a local strategic policy program aimed to establish new multidisci-plinary organizational units for education and research within a Swedish university. Although some of the studied networks have been very successful in creating these units, others have experienced severe problems or even failed in their mission. Why is this so and how do successes and failures relate to the structural qualities of these networks?

The reasons for choosing a case related to higher education policy are many. With its decisive features—namely, rapid expansion and characteristics of a multi-actor arena—the educational sector is, from many aspects, typical for the modern state (Bauer, 1999). In fact, the educational system was used as a frame of reference when Weick (1976) described the complexity of modern policy areas by labeling them as loosely coupled systems made up of numerous actors and widely spread responsibilities. The assumed relationship between structure and performance, which is the topic of the current case study, is likely to be applicable to other policy settings, encompassing similar features. Thus, it is believed that knowledge about the networks discussed in this paper will be of relevance for policymaking within other policy sectors as well, which strengthens the generality of the findings.


The next section describes the case study, defines the dependent variable (i.e., network performance), and discusses the procedure of data collection. Thereafter, the theoretical framework is presented and the arguments for why network structure presumably has implications for network performance are provided. This theoretical part of the paper ends with a section in which the independent variable (i.e., network structure) is operationalized and two hypotheses suggesting how these structural variables relate to performance are formulated. In the subsequent empirical part of the paper, the structural properties of the networks are analyzed and related to their performances, testing and refining the theoretical propositions. Finally, the paper ends with a discussion in which the findings are related to our understanding of contemporary policymaking.

Research Method

The policy networks analyzed in this paper are all related to the realization of a local strategic policy program at a Swedish university.3 Such local strategies have gained increased attention within the system of higher education in Sweden as a consequence of quite radical changes within the formal institutional framework governing the activities of research and education. During the last decades, the relationship between the state and the universities has been transformed, thereby redefining the institutional arena of higher education policy (Marton, 2000). For example, the decentralization of authority and amplification of competition—a reform with decisive implications implemented in the beginning of the 1990s (Bauer, 1999)—have resulted in local-level policymaking gaining increased strategic importance for managing universities to maintain their position in the competitive environment. Moreover, according to national regulations, all universities are required to adopt and implement local strategic programs (Law on Higher Education, 1992: 1434, § 4).

Such a policy process was initiated in the spring of 1999 at the university of focus in the current study, resulting in a program that suggested far-reaching changes. The networks of this paper are related to the realization of this local policy program. This program declared an intention to reorganize education and research into innovative multidisciplinary units organized around certain complexes of research problems (i.e., knowledge areas). The knowledge areas were identified in larger groups through seminars and workshops and were ultimately adopted by the university board of directors. For example, research and education related to the network economy and natural resources were identified as prioritized areas of concern. The task of effectuating these very broadly defined ideas was delegated to interested scholars within the organization. For each knowledge area, two persons functioned as coordinators, managing the work within the group of actors that evolved. Thus, a set of self-organized networks emerged, composed of actors presumably united by a

common aim to realize these new units, subsequently referred to as "arenas," by offering research and education in a multidisciplinary environment. These networks constitute the analytical units of this study. The paper addresses in what way the various performances among these processes could be explained by the structural network qualities.

Case study sampling is theory driven and should follow specific logic of replication for the sake of securing the external validity (Denk, 2002; Miles & Huberman, 1994; Yin, 1994). The networks in this study were chosen according to what was known about the variances in performances. For example, the work within some of the knowledge areas developed quickly, realizing new ideas and creating new structures, while others never managed to achieve common or joint action or simply codified already existing ideas. Thus, although situated in the same context and involving the same formal prerequisites, a significant variance in network performances exists. In this paper, the performance of these networks will be discussed in terms of their efficiency and innovativeness.

Efficiency refers to the internal organizing process of the networks and addresses the relationship between performance and the costs for performing (Abrahamsson, 1993; Vedung, 1997). In the present study, the establishment of arenas and the time required to do so are used as empirical indicators for separating the efficient networks from inefficient ones. Thus, networks that successfully managed to establish new organizational arrangements in a short period of time are considered as highly efficient, as time is considered as a reflection of how well the processes of collaboration and bargaining within these structures progress. The concept of efficiency essentially captures the ability to achieve a situation of mutual adjustments enabling collective action. The other aspect of performance (i.e., innovation) is frequently used in social sciences to capture "all the ideas, lines of actions or objects that are perceived as new" (Borell & Johansson, 1996, p. 33 [own translation]). Here, the application of the concept takes the qualitative aspects of collective action into account, addressing the innovativeness of the outcomes produced by the networks. The concept also reflects the level of overall goal fulfillment, since the intention of formal policymakers—in this case, the board of directors—was to establish new and innovative multidisciplinary units within the organization. The level of innovativeness defines the existing capacity to meet this aim and direction.

In this study, network structure constitutes the independent variable, which is conjectured to explain the discussed differences in performance. Undoubtedly, other variables might affect the networking processes. The opportunities given by the specific context in which the networks are formed make up a set of variables that may affect performance; the substance of the knowledge area and the strategies adopted by the involved actors are yet other variables. The threats to the external validity cannot be ignored, and the limitations given by the adopted case study should be acknowledged. However, since the overall contextual variables are held constant (i.e., all networks evolve in the same university context and are functionally equal), the case is believed to represent a sufficient case for the purpose of this paper.

Data Collection Procedure

Organizing basically implies the "division of labour into various tasks to be performed and the coordination of these tasks to accomplish the activity" (Mintzberg, 1979, p. 2). Previous research, adopting the perspective of policy networks as organized entities and using a bottom-up methodology, has demonstrated that policymaking might be analytically ordered in and empirically captured by four organizing functions, namely: problem definition, prioritizing, resource mobilization, and evaluation (Carlsson, 1993, 2000; Hull & Hjern, 1987). To exemplify, all types of organization require that the actors engaged in the process share an agreement regarding what the "problem" to be dealt with is. Given the number of alternatives concerning how to handle this problem, prioritizing is vital (i.e., deciding what kinds of action should be performed). Organizing requires resources, which must be mobilized in order to find solutions to identified problems or challenges. Finally, the whole process is dependent on the existence of internal judgments of performance, evaluation, or follow-up (Carlsson & Sandstrom, 2008). Presumably, organizing processes as well as policymaking depend on the performance and coordination of these four functions, which implies that policy networks can preferably be mapped with reference to how actors are involved in these functions. This is how the analytical units of the present study are empirically defined.

The networking processes were investigated from 2001 to 2003. A combination of qualitative and quantitative methods, snowballing interviews, and questionnaires were applied to map the networks and learn about their performances. The interviews were semi-structured and focused on the four organizing functions previously discussed. For example, the respondents were asked questions like: "How was the process of problem definition carried out, and who participated in that process?" and "How did you make the priorities, deciding what to do and how, and who took part in that process?" Based on the stories shared by the respondents concerning these practices, overall narratives describing the processes were constructed, and differences regarding how well these organizing functions were carried out could be identified.4 For example, it became clear that within certain networks, the issue of problem definition and prioritization were highly complicated tasks, ridden with numerous difficulties and related conflicts. In addition, the ability to mobilize resources significantly diverged among the networks.

The basic idea of the snowballing interview technique is to identify the networks inductively, letting the respondents nominate each other as being parts of the network (Miles & Huberman, 1994). Within each network, the interviews started with two presumably central actors (i.e., those persons assigned as coordinators) and continued until no new actor was identified as having any central role in the performance of any of the organizing functions. Based on the information given, the snowballing ends when the above criterion is met. In total, 24 respondents were interviewed. The interview sessions lasted from 30 to 90 minutes. A tape recorder was used, and each interview was transcribed for the purpose of qualitative analysis. The interviews provided data regarding the set of actors involved, how the organizing processes were carried out, and general information about the performance of

the networks—meaning their level of efficiency and innovativeness. The judgments concerning these issues as related to determining performance were made based on the aggregated information reported by the respondents.

For the purpose of collecting the sociometric data, the interview data were complemented by a questionnaire that listed the names of all actors mentioned during the interviews. Each respondent was asked to indicate the persons with whom they had discussed the development of arena X (knowledge area X) and mark the box that best described the frequency of such discussions. Respondents could choose between "occasional" and "several" contacts. In the SNA literature, the frequency of interactions is described as one suitable empirical indicator of "tie strength" (Friedkin, 1980), which is defined as "the combination of the amount of time, the emotional intensity, the intimacy (mutual confiding) and the reciprocal services which characterize the tie" (Granovetter, 1973, p. 1371). Acknowledging the fact that the notion of tie strength is far more complex, as well as incorporating aspects other than merely the frequency of the interactions, it is appropriate to say that the questionnaire captured one aspect of the phenomenon. However, only the more frequent contacts (several contacts) were coded and used as a basis for SNA, as these stronger ties better correspond to the theoretical ideas about policy networks. Another reason was to secure the validity of the data. Many respondents claimed that it was difficult to report on their occasional contacts accurately, and previous research has shown that information concerning stronger ties is less sensitive to information deficits, a common problem associated with social network data (Bell, Belli-McQueen, & Haider 2007; Freeman, Romney, & Freeman, 1987; Marsden, 1990; Wasserman & Faust, 1994). The respondents also received the opportunity to add new names to the list, as sociometric questions with open-choice designs are known to have a higher reliability (Wasserman & Faust, 1994). However, despite the precautionary actions taken, the risk that people forgot or are unable to report their relations in an accurate way is always evident. In total, 99 questionnaires were sent out, and 98 answers were received, resulting in a very high response rate. The data were imported into UCINET6 (Borgatti, Everett, & Freeman, 2002) to create a DL file type using the linked list format, nodelist1. The data set generated a matrix of asymmetric data (i.e., the ties were directed and might, or might not, have been reciprocated). Each relationship indicated by the respondents was given the value of 1. The relational data were subsequently processed and analyzed using the functions in UCINET6 (Borgatti et al. 2002) and Netdraw (Borgatti et al., 2002). The described procedure of data collection is thought to capture the targeted population (i.e., the links between actors involved in the organizing processes aimed at forming new institutional arrangements within the university—the so-called arenas).

The Theoretical Argument about Structure and Performance

Why should one assume that network structure is decisive for the performance of policy networks? The theoretical arguments for such an assumption might be found in a theoretical framework combining notions of networks, institutions, and social capital.

A common view within social network research is that networks emerge as a result of purposive action—that they are the result of individual actions taken with the aim to maintain or gain valuable resources in the form of material and/or symbolic goods (Coleman, 1990; Lin, 2001b). In other words, networks evolve because the participants need one another and share the advantages of joint action. This perspective is in line with the notion of policy networks as organized entities of collective action, described in the introductory pages of this paper. This prevailing state of mutual resource dependency links the actors both horizontally and vertically in networks. The state of resource dependency determines the level of hierarchy within the networks—a quality that might or might not correspond to existing formal hierarchies (Lin). Thus, not only political administrative entities can be hierarchical, but policy networks might also possess this quality.

Networking is frequently described as bargaining games (Elmore, 1993; Rhodes & Marsh, 1992; Thatcher 1998). As such, collective action is basically the result of a bargaining process in which actors adjust and adapt to the actions taken by others. However, a distinguishing feature of networks is that this convergence among actors—so vital for the process to continue—might occur despite the absence of formal hierarchical control (Chisholm, 1989; Lindblom, 1968). This bargaining feature is regarded as an important element characterizing the nature of the organizing processes of policymaking.

A point made by proponents of the network approach is that policy outcome is highly dependent on how these bargaining processes proceed. The influence of the actors, as well as their motives, expectations, and resources, is likely to affect network performance. However, it can also be argued that the outcome depends on how the interactions among these actors precede. "In network forms of resource allocation, individual units exist not by themselves, but in relation to other units" (Powell, 1990, p. 303). Following this stance, the web of interactions within policy-producing structures is an important aspect to consider when explaining policy outcomes, as these reflect the processes of bargaining and coordination. This perspective departs from a strict rational choice perspective, as the relational and structural aspects of policymaking are emphasized (Moberg, 1990; Udehn, 2002). Policy networks are believed to constitute something more than mere aggregations of individuals situated in a structural vacuum.

Granovetter (1985, 1992) has convincingly conceptualized the importance of considering these relational aspects in his argument about embeddedness. In his studies, he touched on the everlasting issue of action versus structure in social theory (Giddens, 1984; Hay, 1995). The controversy concerns whether policy outcomes or any other social phenomena could best be explained by choices made by autonomous individuals based on their interests and calculations or by structural variables, such as communities' rules, norms, cultures, or belief systems. This discussion has recently been reframed by theories of new institutionalism and is also viable in the debate concerning the theoretical basis of the policy network approach (Bogason, 2000; Evans, 2001; Koelble, 1995; Marsh & Smith, 2000, 2001; Peters, 1999; Raab, 2001). Although somewhat neglected, institutional theory does provide a useful theoretical foundation for policy network research—an observation that has

been both theoretically stressed and empirically demonstrated (Blom-Hansen, 1997; Heikkila & Isett, 2004; Provan, Isett, & Milward, 2004).

Broadly speaking, institutions "transcend individuals to involve groups of individuals in some sort of patterned interactions that are predictable, based on specified relationships among the actors" (Peters, 1997, p. 18). Using North's (1997) terminology, institutions should be understood as the rules of the game, or, as formulated by Ostrom (2005), as prescriptions that organize human interactions. Despite significant variety, the ideas constituting new institutionalism converge in the respect that they all consider institutions as the result of purposive action affecting human activities. However, they diverge regarding their nature, formation, and change, as well as the degree of constraints they put on human agency. Some theories put more emphasis on the individual, while others stress the cultural and historical aspects (Peters, 1997). For example, although mainstream versions of rational choice institutionalism perceive institutions as rules—an exogenous variable merely constraining the incentives and choices of fairly autonomous individuals—normative institutionalism focuses on the norms and values constantly framing, and being reframed by, the appropriateness of a situation to which the actors have to adapt (March & Olsen, 1989, North, 1997). In assuming the bounded rationality of self-interested actors, Granovetter (1985, 1992) suggests an analytical frame that combines elements of methodological individualism with a structural approach, acknowledging that all actions are socially situated and institutionally constrained. This frame is quite compatible with the central ideas of policymaking applied in this paper.

Granovetter's (1985, 1992) main contribution is his highlighting of the ongoing web of interactions—the networks—in the study of these institutional settings. The dialectical relationship between the institutional structure and individual action occur through network interactions. In addition, Hay (1995) stresses a relational apprehension of both structure and agency: "One person's agency is another person's structure" (p. 191). Hay further states that "[a] social or political structure only exists by virtue of the constraints on, or opportunities for, agency that it affects" (p. 189). Thus, social institutions such as norms, values, or rules at the structural level might be understood as the outcome of repeated interactions within policy networks. Meanwhile, this network interaction is both constrained and enabled by the broader institutional context.

This paper acknowledges the multilayered nature of institutions (Ostrom, 2005), a notion that affects the way the relationship between institutions and policy networks is understood. The process of policymaking is affected by both the formal and informal institutions governing the actors within their particular context. Acknowledging this fact, policy networks has been perceived as "organized entities that reflect specific types of institutional arrangements" (Carlsson, 2000, p. 58). At the same time, the policy networks that evolve as a result of the bargaining interactions and process of resource exchange are in themselves institutional processes forming the rules of the game. Therefore, policy networks should also be perceived as organized entities reflecting institutional processes. In this paper, the potential capacity of policy networks to form institutional rules that structure the behavior of the participating individuals in the process of organizing is assumed without

neglecting the impact that institutions have on this process on a higher level. This circumstance alone justifies the importance of treating networks seriously; they participate in the ongoing process of building and rebuilding institutional arrangements. This is important not only for understanding the conditions of policy making within universities, but in society as a whole.

However, the relationship between policy networks and their outcome is complex. It has been described as a dialectical relationship between network and agency, between network and context, and between network and outcome (Marsh & Smith, 2000). The goal of this study is restricted to exploring the relationship between network structure and performance. Within SNA, the idea that the structures of social networks are important explanatory variables is accepted. Proponents argue that the structure, which refers to how the actors within a network are connected and how relations are arranged, elucidate the underlying structure of the more stable interactions (Friedkin, 1981, p. 41). The network structure is assumed to impose both constraints and opportunities for action. "The structure of relations among actors and the location of individual actors in the network have important behavioural, perceptual, and attitudinal consequences both for the individual units and for the system as a whole" (Knoke, 1990, p. 9). Following this, the pattern of relations either enhances or restricts the process of organizing and performance. As such, the structural pattern presumably both reflects and affects the rise, maintenance, and substance of institutional arrangements. Some types of network configurations might be assumed to be more efficient and innovative than others in achieving collective action and crafting institutions governing the policy process. The idea that size and diversity of groups affect collective action is nothing new in social science (Olson, 1964)—an insight that has bearing on networks as well. Thus, it is likely to assume that certain structures enhance the ability to set a common agenda and to mobilize and allocate resources. Consequently, the question becomes one of what these networks look like in structural terms—an idea that will be discussed in more detail in the next section when bringing the notion of social capital into the theoretical framework.

Network Structure, Social Capital, and Performance

The idea that networks matter for the process of resource allocation is a basic implication of the notion of social capital (Burt, 2000; Lin, Cook, & Burt, 2001; Putnam, 1992, 2000; Rothstein, 2003). Despite the existence of a multitude of definitions, applied to different analytical units, and using a diverse set of empirical measurements, the social capital metaphor implies some kind of advantage attained by the social structure. "Social capital is productive, making possible the achievement of certain ends that would not be attainable in its absence" (Coleman, 1990, p. 302). Lin (2001a) defines social capital as "resources embedded in a social structure which are accessed and/or mobilized in purposive actions" (p. 12). Social capital is often viewed as having two main ingredients: resources and relations. Despite its popularity, the concept has been the subject of substantive critique (Kwon, 2004; Portes, 1998; Portes & Landolt, 2000). It has been stressed that interpretations of social

capital that hold more than one variable that could be correlated need to be analytically separated and given independent empirical measures (Portes, 1998). Consequently, researchers have handled the two aspects of social capital differently. Ostrom and Ahn (2003) distinguished between an expansionalist and a minimalist branch. Although some scholars have measured the amount or quality of embedded resources, others have focused more on the properties of the networks of relations, often by applying the concepts and tools of SNA (Burt, 1997, 2000, 2001). The aim of this study fits the latter kind of research.

Primarily two divergent propositions concerning how network structure relates to social capital exist: network closure and structural holes (Burt, 2000, 2001). Both might be applied at an individual level or on a group level. The first proposition is associated with Coleman's idea that a high level of interconnectedness within a network facilitates performance because of enhanced communication, the creation of common norms, and the possibility to restrain opportunistic behavior (Burt, 2000; Coleman, 1990; Lin, 2001b). It can be assumed that a well-connected network promotes collective action and enforces the organizing capacities of the policy network. By the same logic, the formation of institutions, as well as the creation and maintenance of rules, should be enhanced within such networks, thereby improving the policy process. However, the structural hole argument is more concerned with the importance of information dispersal among actors or sets of actors. This concept can be ascribed to Burt (1997, 2000, 2001) but truly draws on previous work, such as Granovetter's (1973) seminal study on the strengths of weak ties. Metaphorically, structural holes can be understood as some kind of break in the social structure that is identifiable by the absence of ties or the presence of weaker ties. The actors in a position to bridge such holes are supposed to have a strategic advantage as they have access to new and diversified information (or resources of any other kind) that can be used in the bargaining activities. Accordingly, a network that spans many structural holes is a network rich in social capital. The essential argument inherent in both the weak ties hypothesis and structural holes is that an individual's access to resources is determined by the characteristics of that person's social network relations. However, while Granovetter points to the relevance of tie strengths, Burt stresses the importance of nonredundant contacts.

The ideas underpinning the two perspectives—closure and structural holes—are seemingly contradictory. However, based on significant empirical material, Burt (2000) has combined the two into a hypothesis about the network structure of social capital. By separating the local structure (which refers to the in-group activities) from the global structure (which describes how networks are connected to other networks), Burt puts forth the proposition that "while brokerage across structural holes seems to be the source of added value, closure can be critical to realizing the value buried in the structural hole" (p. 398). Hence, both qualities are important for networking activities, as they enforce one another in the process of policymaking. These ideas, although sometimes differently framed using other terms, have been supported by other researchers, both empirically and theoretically (Lin, 2001b; Oh, Chung, & Labianca, 2004; Reagans & McEvily, 2003; Reagans & Zuckerman, 2001).

Thus, policy networks are not only organizing entities, but also institutional entities whose structural features presumably affect their performance. They reflect coordination processes in which actors, in their struggle for resources, form structures of collective action realizing their aims. With reference to the described theory, identifying how distinct patterns of interactions such as network closure and the existence of bridges over global structural holes are likely to affect performance is highly relevant for the understanding of the organizing process related to policy-making. The policy networks studied in this paper will therefore be scrutinized by focusing on these two network qualities. However, first, the concepts need to be operationalized, making them empirically measurable.

Measuring Network Closure

A network characterized by closure has in the literature been described as a well-connected network, either directly by virtue of the existence of many strong connections between network members or indirectly through a common contact. Network closure captures both the level of overall activity within a network and the general level of hierarchy (Burt, 2000). In the current study, this quality is captured through the use of two SNA measures: density and network centralization.

Density is calculated by dividing the actual number of connections within a network with the maximum number of possible connections (Scott, 2000). A higher density means a higher level of activity and closure. However, certain issues must be taken into account when using the density measure as an indicator of interconnectedness—for example, the size of the network (Friedkin, 1981; Moody & White, 2003). If networks of different sizes are compared, similar levels of density might in fact describe dissimilar levels of structural cohesion. "It requires a larger value of network density to achieve the same level of structural cohesion in a small [compared to] a large network" (Friedkin, 1981, p. 49). This refers to the classical "traveling problem," which occurs when the same empirical measure means different things in different contexts (Denk, 2002). This issue will be considered when analyzing the studied university networks. If the above condition is met, density might still be a good indicator of overall cohesiveness within these structures (Scott, 2000; Wasserman & Faust, 1994).

The second indicator of network closure (i.e., the level of hierarchy within the structures) is indicated by the overall network centralization. Higher centralization levels point to hierarchy, which in turn indicates a higher level of closure (Burt, 2000). Network centralization is calculated in two steps. Data describing the centrality of each individual actor are used as a basis for estimating the level of hierarchy for the network as a whole. First, the centrality scores of each of the other actors are subtracted from the highest centrality score (i.e., the centrality score of the most central actor). Thereafter, the differences are summarized. The result of this calculation is divided by the maximum possible sum of differences (Scott, 2000; Wasserman & Faust, 1994). In other words, the measure tells how "unequally well connected" the actors are.

Three principally different approaches to centrality exist: betweenness, closeness, and those based on degree. The notion of degree centrality is applied in this study, which means that the centrality score of an actor is determined by counting the number of direct links connecting this person to the other actors.5 The more connected a person is, the higher the centrality score is. The image of a star might serve as an illustration of the reasoning. A network formed as a star has one central actor to which all other actors are directly connected. Apart from these links, there are no relations present, which means that the organizing activities are solely dependent on this coordinating actor. Such a structure generates a centralization index of 100 percent; in other words, the centrality scores of the involved actors are as unequally distributed as they may be.

Both density and centralization address the issue of how well integrated policy networks are. A high-density level secures the flow of communication among the actors, facilitating bargaining and joint action. However, communication and collaboration might also be channeled through a central coordinating actor, which is why higher levels of hierarchy also point to integration. Both measures reflect aspects of a closed network; according to the theoretical assumptions, this network is likely to affect the internal process of organization and the performance of poli-cymaking. As mentioned earlier, Burt (2000) specified the importance of closure for "realizing the value buried in the structural hole" (p. 398). The four networks in this paper will be compared with one another and ranged in accordance to their level of closure. The logic for interpretation is that higher levels of density and centralization, respectively, indicate higher levels of network closure.

Measuring Global Structural Holes

While closure tells something about the local structure of a network, the existence of bridges over structural holes refers to the global structure and how the network in question is related to other constellations of actors. However, a delicate problem is that the snowballing method, which has been applied to collect the relational data, restricts the possibility of obtaining direct information about global structures, as the method only captures the "stronger" ties among a set of actors (Lin, 2001b). Therefore, the extent to which the networks in this study bridge global structural holes has to be investigated indirectly. Network heterogeneity has previously and successfully been used as an indicator of this quality (see, e.g., Reagans & Zuckerman, 2001). The basic idea is that networks constituted of actors from dissimilar backgrounds, representing different organizational units, etc., can be assumed to span many global structural holes. This is also how the problem is handled in this paper.

First, the diversity of the actors is considered by counting the number of organizational units represented in each network. The units are administratively defined, organized around scientific subjects constituting different divisions and representing various scientific disciplines within the university. This aspect is believed to capture the diversity of perspectives and other resources available in the process of policymaking. Further, the interorganizational exchanges within the networks are

considered. The extent to which the interactions within the networks are cross-boundary in their character is determined by calculating the percentage of ties crossing different administrative boarders.

Using a comprehensive analysis of these two measures, the level of heterogeneity among the four university networks will be compared. Hence, a network that consists of a diversified set of actors involved in many cross-boundary interactions is regarded as highly heterogeneous. As mentioned earlier, such networks secure "the source of added value" in the process of policymaking (Burt, 2000, p. 398). With the theoretical framework presented in this section as a frame of reference, the following general hypotheses about structure and performance can be formulated:

Hypothesis 1: The level of network closure (i.e., the level of density and centralization) affects the performance of policy networks.

Hypothesis 2: The level of network heterogeneity (i.e., the level of diversity among actors and the extent to which these are involved in cross-boundary interactions) affects the performance of policy networks.

While the hypotheses are generally formulated, they are firmly anchored in previous network research and herein fulfill the purpose of identifying the structural qualities that are likely to be decisive for policymaking. The next step is to refine these hypotheses. Thus, apart from testing the relevance of the two propositions, the aim of this paper is also of a more exploratory nature in that it will further specify the propositions' potential to explain organization and outcome in policymaking processes.

Structure and Performance in Empirical Terms

In this section, the empirical data concerning structure and performance are explored, and possible patterns are elucidated using a comparative approach. Are structural variations in the levels of closure and heterogeneity related to differences in efficiency and innovation within the studied university networks?

Network Performance: Efficiency and Innovation

The analysis and conclusions drawn regarding network performance are based on the information gathered through the interview study. The empirical data regarding the issue of performance are summarized in Table 1. The distinguishing features of each organizing process are described, followed by an overall estimation regarding their levels of efficiency and innovation, respectively.

When determining efficiency, the following two questions guide the analysis: Did the processes result in the realization of arenas? How long did the processes take? Depending on the answers to these questions, the networks are compared and numbered according to this first aspect of performance (see Table 1).

For example, the table implies that network A is the most efficient network. This knowledge area became the first arena to get the permission to start (i.e., to accept

Table 1. The Organizing Processes and Performances

Distinguishing Features of the Process Efficiency Innovativeness

Network A • Easy processes of (1) Arena established in (3) The arena was based

prioritization based on August 2001. on an already existing

a preexisting concept. educational program.

• No experienced need

for additional resources.

Network B • Easy process of (2) Arena established in (2) The concept of the

prioritizations partly January 2002. arena had its origin

based on existing in prevailing

collaboration. collaboration projects.

• Nonproblematic process

of resource


Network C • Difficult prioritization (3) Arena established in (1) A new structure was

process. August 2002. established that

• Nonproblematic process launched a new idea.

of resource


Network D • An unsuccessful (4) No result was _a

process of problem achieved.

definition and


• Serious lack of


aThis network never achieved any results; however, the ideas proposed within the group were highly innovative. Thus, if one were to adopt a strict process-oriented perspective of innovation (i.e., as the attempt to achieve something new), the network would be regarded as innovative. Thus, in this paper, the degree of realization is part of the definition.

students into its newly introduced program by the university's board of directors in August 2001). Network B started its arena in January 2002, followed by network C in August 2002. However, network D never managed to realize the desired result and is therefore considered an inefficient structure. Contrary to the organizing processes within the other networks, the actors involved in network D experienced a highly problematic problem-definition process with an ever-changing purpose. For example, the actors discussed whether the work should be directed toward research or educational matters. Likewise, they never decided what kind of research problems should be the focus or from what scientific angles the particular problems should be approached. No agreements were met concerning these vital issues within network D. Moreover, both the purpose of the work and the disciplines and scientific subjects to be included in the process were constantly being questioned and redefined, all reflecting a poor prioritization process.

The process of network A might serve as a contrast to network D's process, as no real controversies were attached to the prioritization function within this network. However, within network C, difficulties emerged in finding a common principle to structure the working process. Through bargains, compromises, and adjustments, the rules of the game finally emerged, and the work could progress in a satisfactory manner. A common comprehension regarding the importance of realizing the arena

outweighed the problems related to the challenge of achieving collective action experienced in this network.

Considering the second aspect of network performance (i.e., innovativeness), the outcomes are examined with the following question as a frame of reference: Do the arenas represent innovative and essentially new concepts? The level of innovation is determined by comparing the outcomes according to this important aspect (see Table 1). For example, network C developed—for the university and the involved actors—a new concept not previously implemented. The arena that became the result of this process united administrative units, scientific disciplines, and subject areas not previously involved in collaboration. No preexisting research program or known education program functioned as a role model for this organizing process. In contrast, network A's process concerned the implementation of a knowledge area based on an already existing education program. Consequently, nothing new was promoted within this organizing process, supporting the conclusion that the structure is characterized by a low-level of innovation. However, when it comes to network B, the work was originally built on an existing collaboration structure, although the original ideas and design of the knowledge area were significantly modified as the process continued.

Regarding network B and network C—both determined as innovative structures—the data in Table 1 describe active and successful processes of resource mobilization, with the focal aim of mobilizing new actors who carry appropriate resources. In these cases, the wide range of types of knowledge, as well as the scientific perspectives made possible through this line of action, could be related to the ability to achieve the highly innovative outcomes. No such process took place within network A, a circumstance that probably affected the ability to create something new and innovative within this group of actors.

Network Structure: Impacts of Closure and Heterogeneity

The structural qualities of the studied networks are presented in Table 2. In the first three columns of the table, the measures underpinning the analysis regarding network closure are presented—namely, network size, density, and degree centralization.

The relational data were imported into UCINET6 (Borgatti et al., 2002), where the functions for calculating density and degree centralization were run. Density was calculated based on the asymmetric set of raw data. For the sake of making valid interpretations, the level of density is interpreted by comparing networks of similar sizes. Accordingly, the density of network A, which contains 18 actors, is compared with network D, which contains 19 actors, while the two larger networks, network B and network C, are compared with each other. With reference to the data in Table 2, it can be concluded that networks A and B are more interconnected than the other two.6 This interpretation is also true when regarding the patterns of indirect ties as reflected in the degree centralization indices in Table 2. When calculating degree centralization, the choice of treating the data as symmetric was chosen, generating an index that reflects the overall communication structure of the underlying policy

Table 2. Structural Network Properties

Sizea (No. of Density (d) Degree of Diversity of Cross-Boundary

Actors) Centralization Actors Interaction

(%) (No) (%)

Network A 18 0.25 62 7b 29

Network B 42 0.16 51 17c 54

Network C 37 0.11 38 18d 61

Network D 19 0.15 40 10e 50

aThe questionnaire was sent out to 99 actors. The sizes of the networks in Table 2 reflect the number of actors reported as involved in "several" discussions concerning the development of the selected knowledge areas. Some actors were identified as isolates and removed from the analysis. The total number of actors in all four networks is 116, a figure that is explained by the fact that respondents were given the opportunity to complement the list with new names.

bDivisions/units: industrial design, mathematics, computer aided design, learning and educational processes, sound and vibration, school of music, and information and communication technology. cDivisions/units: structural mechanics, computer aided design, operation and maintenance engineering, system science, accounting and control, industrial design, industrial logistics, industrial marketing, industrial organization, industrial work environment, quality and environmental management, learning and educational processes, economics, architecture and infrastructure, manufacturing systems and engineering, structural engineering, and wood science and technology.

dDivisions/units: waste science and technology, industrial logistics, mining and geotechnical engineering, energy engineering, wood science and technology, chemical technology, quality and environmental management, mineral processing, economics, chemistry, process metallurgy, jurisprudence, political science, fluid mechanics, applied geology, sanitary engineering, traffic engineering, and water science and technology.

eDivisions/units: computer aided design, system science, gender technology and organization, industrial design, industrial work environment, mathematics, structural engineering, sound and vibration, languages and culture, and computer science and electrical engineering.

structures. In other words, the very existence of a link between two actors, disregarding its direction and whether or not it is reciprocated, was recoded and given the value of 1. Thus, when levels of activity and hierarchy are considered, the different structural properties of the four networks become apparent. For instance, networks A and B have higher levels of closure than networks C and D. What patterns of correlations are found when this information is compared with what is known about their performances?

When the information in Tables 1 and 2 is comprehensively analyzed, it becomes obvious that networks A and B, characterized by higher levels of network closure, also are significantly more efficient than the other two. In networks A and B, the most well-connected networks, the functions of problem definition and prioritization were quite easily carried out. In contrast, the actors within networks C and D experienced poor or complicated prioritization, and the levels of in-group closure were also remarkably lower within these structures. Thus, variations in efficiency and in the ability to establish a proper process of prioritization are seemingly related to variations in network closure. However, when it comes to innovation, the relationship is seemingly reversed, implying a negative correlation between closure and innovation.

Network heterogeneity is empirically determined by the measures presented in the last two columns of Table 2: the diversity of actors and the proportion of cross-

boundary interactions. Although the first reflects the number of work units represented in the organizing processes, the latter expresses the proportion of network relations crossing the organizational borders between these units. A comparative analysis reveals that network A is the homogenous network. The structures of networks B and C are essentially more heterogeneous in character, considering both the diversity of actors and the proportion of cross-boundary interactions, while network D can be placed somewhere in between. What patterns emerge when this information describing heterogeneity is compared with the identified differences in performances?

Relating the data on network heterogeneity (Table 2) to network efficiency (Table 1), a negative relationship is indicated. Network A, by far the least heterogeneous structure, is also the most efficient one in realizing the area. It is also intuitively likely that networks containing a less diverse set of actors are not as exposed to differences and competing interests, and therefore need less time to come to terms with collective action. Although the comparative analysis does not give full support to this assumption, it might still be an idea worth examining in future studies. Indeed, when network heterogeneity is related to innovation, the correlation is more robust. Network C, with the highest heterogeneity level, is also regarded as the most innovative structure; the rankings of the other networks follow the same pattern. The actors within these more heterogeneous networks also report active and successful resource mobilization processes, in which new actors possessing the proper resources and qualities were easily engaged in the work to establish the specific knowledge areas. However, despite involved actors' ambitions to launch new ideas and concepts, network D struggled unsuccessfully with this function, as its members were unable to involve identified key actors and gain the scientific expertise needed to ensure the work progressed. Hence, innovation—understood as the ability to promote new lines of thinking and develop new concepts—is seemingly promoted by heterogeneity. Variations in innovation and the ability to perform a successful resource mobilization process might indeed be related to variations in network heterogeneity. Thus, in order to create and introduce something new, networks that bring together actors possessing dissimilar resources must be formed.

In concluding this empirical section, the findings are discussed in the light of the hypotheses, emphasizing the particular impacts of closure and heterogeneity, formulated in the theoretical section of this paper. If a high-performing policy network is defined as a network that is both efficient and innovative, network B serves as a good illustration, as its structure is both heterogeneous and characterized by a significant level of local network closure. Thus, the idea that the two network qualities are important for policymaking has support in the empirical material. The proposition that "while brokerage across structural holes seems to be the source of added value, closure can be critical to realizing the value buried in the structural hole" (Burt, 2000, p. 398) is supported. These general ideas can be refined by relating structures to certain organizing functions that are proven important for the process of policymaking. The argument being proposed here is that although heterogeneity promotes the access to a diversified set of resources, closure enables the decisive process of prioritization within policy networks. The conjectured—and, as it seems

Figure 1. Network Structure, Organization, and Performance.

likely—relationship among network structure, organization, and performance found in this study is summarized in Figure 1, which illustrates the hypothesized relation between network closure and efficiency on the one hand and heterogeneity and innovation on the other. In the theoretical part the paper, these relations were specified in two general hypotheses:

Hypothesis 1: The level of network closure (i.e., the level of density and centralization) affects the performance of policy networks.

Hypothesis 2: The level of network heterogeneity (i.e., the level of diversity among actors and the extent to which these are involved in cross-boundary interactions) affects the performance of policy networks.

At this junction, it can be concluded that the empirical analysis supports these hypotheses. It should also be emphasized that the present study has clarified the link between structure and performance, demonstrating that qualities like closure and heterogeneity do not affect efficiency and innovation directly. Instead, they affect central organizing processes (i.e., prioritization and resource mobilization), which in turn affect the ability to generate innovative outcomes in an efficient way.

Conclusions, Implications, and New Issues

Although the network orientation is an integrated part of policy science, research is less often tailored to elucidate how and in what way the structure of policy networks might affect performance. Presumably, this study provides findings demonstrating that network structure is an important variable to consider, since the very structure reflects the interactions among actors involved in organization, which is in turn related to specific outcomes. The study supports the idea of "structural impacts" on policymaking previously proposed by other scholars. Based on a theoretical framework that combines central ideas of networks, institutions, and social capital, two general hypotheses regarding the relationship between structure and performance were formulated. These have been tested, applying SNA, in a comparative case study. The conjectured importance of network closure and network heterogeneity for the performance of policy networks has been supported. It is proposed that although heterogeneity is a necessity for the creation of innovative networks, the level of efficiency is positively related to the level of in-group closure. Further, the hypotheses were refined relating network structure to the performance of certain

organizing functions reflecting the process. Thus, the idea of centralized integration as a promoter of performance, launched in Provan and Milward's (1995) study, has been confirmed. In addition, the proposed importance of network heterogeneity (Lin, 2001; Reagans & Zuckerman, 2001) is supported by the analysis. It can also be added that the potential contribution of SNA for the field of policy science, suggested by other researchers (Adam & Kriesi 2007), has been demonstrated in this paper.

However, while the assumptions indicated in Figure 1 find empirical support, they still need additional testing. Limitations must be considered. First, the study is based on fairly limited empirical material. A second objection might be that there are difficulties in ruling out the possibility of hidden variables that could potentially affect performance. For example, the motivation and commitment shared by the actors involved in the network activities were found to be in common with the more successful networks in this study. The problematic issue of causality, which is an ever-pervading concern in social scientific research, must definitely be recognized. It should be noted that the relation between structure and performance could in fact, in some aspects, be reversed. The link among structure, organization, and performance is likely to be more complex than what is implied in Figure 1.7

This paper also has a methodical aim: to test the usefulness of SNA in policy research. Although this has proven successful, a relevant objection concerning the validity of the findings is related to known validity problems associated with the adopted network measures. For example, does density really capture the idea of closure? Do different density levels reflect different levels of network closure? It is believed that by combining quantitative measures with a solid qualitative analysis, such uncertainty has been effectively addressed. A clear correspondence between the picture gained from the interview data and the social network data exists. Another discovery is that it is important to study networks using a wide range of SNA measures. Although the present paper only presents a limited number of SNA measures, the material as such has been exposed for a more profound analysis (Sandstrom, 2004). The conclusion is that although SNA generates quantitative measures describing structural features, the need for a qualitative analysis is crucial for the validity of the analysis. It is important to ensure that the chosen SNA measures truly reflect the notions of the theoretical concepts they aim to capture. Accordingly, certain limitations are related to the empirical case study, the ability to control for hidden variables, and the issue of causality. While acknowledging these limitations, the purpose of this paper (to address the identified knowledge gap and contribute to our understanding of the explanatory power of policy networks) has been fulfilled.

However, the study triggers additional heretofore unanswered questions regarding the nature of the relationship between structure and performance. For example, is there actually a trade-off between network closure and network heterogeneity? What effects would such a relationship have on network performance? For example, would a policy process be promoted by a high level of in-group closure at the beginning of the process, thereby enhancing the function of problem definition and prioritization, or would a structure of this kind in fact hamper the possibilities

of mobilizing relevant actors, causing problem solving to suffer from a resource scarcity later on in the process? Research on these issues would require access to longitudinal SNA data. Unfortunately, this case study does not allow for a study of network dynamics or network evolvement, relevant issues that have been addressed by other researchers (Human & Provan, 2000; Provan, Isett, & Milward, 2004).

In this paper, network performance is discussed in terms of efficiency and innovation. Certainly, many other evaluation criteria are likely to be relevant for the evaluation of policymaking structures (e.g., economic efficiency, equity, adaptability, legitimacy, representation, and accountability) (Lundqvist, 2004; Ostrom, 2005; Sabatier et al., 2002). How do different structural properties relate to these variables? For instance, what kind of network would foster a deliberative discussion among the involved, and what about the issue of inclusion versus exclusion in public policy networks?

Implications for Policymaking

In contemporary research, the old Weberian model of a tightly coupled political-administrative hierarchy has been replaced by another image of the policy process—one characterized by policy networks, entities that are formed by private and public actors in pursuit of problem solving in a complex world. Presumably, findings concerning network structure and performance are important for our understanding of policymaking in modern-day society and for the challenges facing public administration management. The argument proposed herein is that many of these challenges could be rephrased into "network terms." Certain network structures are more favorable than others. The implications derived from this study are that if policy networks are too dense and homogeneous, they might be less innovative; if they are too heterogeneous, they will get little done. For example, when it comes to the existence of stagnated policy-producing structures, poor innovativeness is possibly explained by the lack of heterogeneity among the actors. Old and well-established policy structures containing actors with similar ideas, values, and approaches might in fact be effective in achieving desired goals; however, they may do so at the cost of innovation. Presumably, the level of innovation would be promoted by letting the process be more inclusive—in essence, by changing the configuration of the networks.

It is also likely that the inability to mobilize the necessary resources to solve particular policy problems could be referred to the absence of involvement of relevant actors (i.e., those actors who span important "holes" in the global structure). Resources should not be counted in only monetary terms; both information and legitimacy are regarded as important faculties. For example, considering legitimacy, could the inability to involve certain key actors be the main reason for the failure of the implementation of certain policies? On the other hand, could an overly extensive diversity among the actors explain why a certain policy network is incapable of prioritization? Alternatively, could a low level of accomplishment be attributed to the absence of a functioning coordinating unit? In agreement with this, poor prioritizing

processes could be counteracted by the creation of proper platforms for communication, bargaining, and mutual adjustments among the participants.

Policymaking is often riddled with conflict. This begs the question of whether policy issues, characterized by conflict, could be associated with certain network structures—for example, structures made up by many distinct subgroups. The assumption would be that such networks lack a consistency in opinions facilitated within closed networks. On the other hand, networks with a closed structure, presumably characterized by a lower level of conflict, might lack the innovative features promoting a needed policy change. However, most of these questions remain to be answered.

Certainly, explanatory power exists in the concept of policy networks, and the use of a formal network approach (e.g., SNA) is one way of exploring the possibilities of the concept. Undoubtedly, more theoretical and empirical work is needed to further develop our knowledge about how and why networks are decisive variables in policymaking processes. A research agenda of this kind would significantly advance the concept of policy networks, as well as benefit policy science and public management in general.

Annica Sandström is a Ph.D. in political science at the Department Social Science, Luleâ University of Technology. Research interests include policy networks and social network analysis.

Lars Carlsson is a professor in political science at the Department of Social Science, Luleâ University of Technology. Research interests include policy analysis, policy networks especially within the field of natural resources management. Professor Carlsson currently holds the position of president at Kristianstad University College in Sweden.

1. This quotation alludes to O'Toole's (1997) paper "Treating Networks Seriously."

2. It should be noted that how the term "performance" is understood and operationalized in these kinds of studies shows great diversity.

3. Lulea University of Technology (LTU).

4. A more comprehensive presentation of the information generated through the interviews is found in Sandstrom (2004).

5. Degree centrality takes the number of direct connections to and from an actor into account; closeness centrality regards how close an actor is to all the other actors while the measure of betweenness rests on the idea that the centrality of an actor is determined by how frequently an actor is situated between two other actors (Hanneman, 2004). Work aimed at clarifying the theoretical power of these measures has been conducted (Bonacich, 1987; Freeman, 1978 / 79; Freeman, Roeder, & Mulholland, 1979/80; Friedkin, 1991; Wasserman & Faust, 1994). Freeman et al. (1979/80) emphasize the particular advantage of applying the degree and betweenness measures for capturing network centralization. In this paper, the degree centrality measure will be used. Degree centralization is based on differences in "communication activity" while betweenness centrality reveals differences in the potential of "withholding" or "distorting" the flow of information (Freeman, 1979 /80). A network with a high degree of centralization corresponds to a network in which the communication flow is indirectly connected through a coordinating unit. Thus, the betweenness centralization index is very sensitive for the presence of "lines" or long rows of indirect communication among the actors. Such an imaginable network is regarded as being less compatible with the idea of network closure.

6. Regarding density as a proper indicator of structural cohesion, Friedkin (1981) argues that a dramatic increase of structural cohesion exists within the lower density interval—namely, between 0.0 and 0.5 d.

7. After finalizing the study on which this paper is based, two more cases studies in other policy areas were conducted. In these, the findings are further supported. See also Carlsson and Sandstrom (2008) and Sandstrom (2008).


Abrahamsson, Bengt. 1993. The Logic of Organizations. Newbury Park, CA: Sage.

Adam, Silke, and Hanspeter Kriesi. 2007. The Network Approach. In Theories of the Policy Process, ed. Paul A. Sabatier. Boulder, CO: Westview Press, 129-54.

Agranoff, Robert, and Michael McGuire. 2001. "Big Questions in Public Network Management Research."

Journal of Public Administration Research and Theory 11 (3): 295-326.

Bauer, Marianne. 1999. Transforming Universities, Changing Patterns of Governance, Structure and Learning in Swedish Higher Education. Higher Education Policy Series 48. London: Jessica Kingsley Publishers.

Bell, David C., Benedetta Belli-McQueen, and Ali Haider. 2007. "Partner Naming and Forgetting: Recall of Network Members." Social Networks 29: 279-99.

Berkes, Fikret. 2002. "Cross-Scale Institutional Linkages: Perspective from the Bottom Up." In The Drama of the Commons, ed. Elinor Ostrom, Thomas Dietz, Nives Dolsak, Paul C. Stern, Susan Stonich, and Elke U. Weber. Washington, DC: National Academy Press, 293-321.

Blom-Hansen, Jens. 1997. "A 'New Institutional' Perspective on Policy Networks." Public Administration 75 (4): 699-3.

Bogason, Peter. 2000. Public Policy and Local Governance. Institutions in Postmodern Society. Cheltenham, UK: Edward Elgar.

Bogason, Peter, and Theo A. J. Toonen. 1998. "Introduction: Networks in Public Administration." Public Administration 76: 205-27.

Bogdanor, Vernon, ed. 1987. The Blackwell Encyclopaedia of Political Institutions. New York: Blackwell Reference.

Bonacich, Phillip. 1987. "Power and Centrality: A Family of Measures." The American Journal of Sociology 92: 1170-82.

Borell, Klas, and Roine Johansson. 1996. Samhället som nätverk. Lund, Sweden: Studentlitteratur.

Borgatti, Stephen P. 2002. Netdraw. Included in UCINET 6 for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies.

Borgatti, Stephen P., and Pacey C. Foster. 2003. "The Network Paradigm in Organizational Research: A Review and Typology." Journal of Management 29 (6): 991-1013.

Borgatti, Stephen P., Candace Jones, and Martin G. Everett. 1998. "Network Measures of Social Capital."

Connections 21 (2): 28-36.

Borgatti, Stephen P., Martin G. Everett, and Linton C. Freeman. 2002. UCINET 6 for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies.

Börzel, Tanja A. 1998. "Organizing Babylon—On the Different Conceptions of Policy Networks." Public Administration 76: 253-73.

Burt, Ronald S. 1997. "The Contingent Value of Social Capital." Administrative Science Quarterly 42: 339-65.

--. 2000. "The Network Structure of Social Capital." In Research in Organizational Behavior 22, ed. Barry

M. Staw, and Robert I. Sutton, Greenwich, CT: JAI Press, 345-423.

--. 2001. "Structural Holes versus Network Closure as Social Capital." In Social Capital, Theory and

Research, ed. Nan Lin, Karen Cook, and Ronald S. Burt. New York: Aldine de Gruyter, 31-56.

Carlsson, Lars. 1993. Samhällets Oregerlighet. Organiseirng Och Policyproduktion i Näringspolitiken. Stockholm/Stenhag: Symposion Graduale.

--. 2000. "Policy Networks as Collective Action." Policy Studies Journal 28 (3): 502-20.

Carlsson, Lars, and Annica Sandström. 2008. "Network Governance of the Commons." International Journal of the Commons 2: 33-54. [Online]. Accessed August 4, 2008.

Chisholm, Donald. 1989. Coordination without Hierarchy. Informal Structures in Multiorganizational Systems. Berkeley: University of California Press, Ltd.

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

Daugbjerg, Carsten. 1998. "Linking Policy Networks and Environmental Policies: Nitrate Policy Making in Denmark and Sweden 1970-1995." Public Administration 76: 275-94.

Denk, Thomas. 2002. Komparativ Metod: Förstäelse Genom Jämförelse. Lund, Sweden: Studentlitteratur.

Dowding, Keith. 1995. "Model or Metaphor? A Critical Review of the Policy Network Approach." Political Studies 43: 136-58.

Elmore, Richard E. 1993. Organisational Models of Social Program Implementation. In The Policy Process. A Reader, 2nd ed., ed. Michael Hill. London: Harvester Wheatsheaf, 313-48.

Evans, Mark. 2001. "Understanding Dialectics in Policy Network Analysis." Political Studies 49: 542-50.

Flap, Henk, Bert Bulder, and Beate Völker 1998. "Intra-Organizational Networks and Performance: A Review." Computational and Mathematical Organization Theory 4: 109-47.

Freeman, Linton C. 1978 / 79. "Centrality in Social Networks. Conceptual Clarification." Social Networks 1: 215-39.

Freeman, Linton. C., Douglas Roeder, and Robert R. Mulholland. 1979/80. "Centrality in Social Networks: II. Experimental Results." Social Networks 2: 119-41.

Freeman, Linton C., A. Kimberly Romney, and Sue C. Freeman 1987. "Cognitive Structure and Informant Accuracy." American Anthropologist, New Series 89 (2): 310-25.

Friedkin, Noah E. 1980. "A Test of Structural Features of Granovetter's Strength of Weak Ties Theory." Social Networks 2: 411-22.

-. 1981. "The Development of Structure in Random Networks: An Analysis of the Effects of Increasing Network Density on Five Measures of Structure." Social Networks 3: 41-52.

-. 1991. "Theoretical Foundations for Centrality Measures." The American Journal of Sociology 96 (6):


Giddens, Anthony. 1984. The Constitution of Society. Cambridge, UK: Polity Press.

Granovetter, Mark S. 1973. "The Strength of Weak Ties. " American Journal of Sociology 78: 1360-80.

-. 1985. "Economic Action and Social Structure: The Problem of Embeddedness. " American Journal

of Sociology 91: 481-510.

-. 1992. "Economic Institutions as Social Constructions: A Framework of Analysis." Acta Sociologica

35: 3-11.

Hanf, Kenneth, and Laurence J. O'Toole Jr. 1992. "Revisiting Old Friends: Networks, Implementation Structures and the Management of Inter-Organizational Relations." European Journal of Political Research 21 (1-2): 163-80.

Hanf, Kenneth, and Fritz W. Scharpf. 1978. Interorganizational Policy Making: Limits to Coordination and Central Control. London: Sage.

Hanneman, Robert. 2004. Introduction to Social Network Methods [Online]. ~hanneman/SOC157/NETTEXT.PDF. Accessed February 2, 2004.

Hay, Colin. 1995. "Structure and Agency." In Theory and Methods in Political Science, ed. David Marsh and Gerry Stoker. New York: St. Martin's Press, 189-206.

Heclo, Hugh. 1978. "Issue Networks and the Executive Establishment." In The New American Political System, ed. Anthony King. Washington, DC: American Enterprise, 87-124.

Heikkila, Tanya, and Kimberly R. Isett. 2004. "Modeling Operational Decision-Making in Public Organizations: An Integration of Two Institutional Theories." American Review of Public Administration 34 (1): 3-19.

Hjern, Benny. 1987. "Policy Analysis: An Implementation Approach." Paper for the Annual Meeting of the American Policy Science Association, September 3-6, Chicago.

Hjern, Benny, and David O. Porter. 1993. "Implementation Structures. A New Unit of Administrative Analysis." In The Policy Process: A Reader, ed. Michael Hill. London: Harvester Wheatsheaf, 248-65.

Hull, Christopher J., and Benny Hjern. 1987. Helping Small Firms Grow: An Implementation Approach. London: Croom Helm.

Human, Sherrie E., and Keith G. Provan 1997. "An Emergent Theory of Structure and Outcomes in Small-Firm Strategic Manufacturing Networks." The Academy ofManagement Journal 40 (2): 368-403.

--. 2000. "Legitimacy Building in the Evolution of Small-Firm Multilateral Networks: A Comparative

Study of Success and Demise." Administrative Science Quarterly 45 (2): 327-65.

Jordan, Grant. 1990. "Sub-Governments, Policy Communities and Networks: Refilling Old Bottles?"

Journal of Theoretical Politics 2 (3): 319-38.

Jordan, Grant, and Klaus Schubert. 1992. "A Preliminary Ordering of Policy Network Labels." European Journal of Political Research 21 (1-2): 7-27.

Kickert, Walter J. M., Erik-Hans Klijn, and Joop F. M. Koppenjan, eds. 1997. Managing Complex Networks: Strategies for the Public Sector. London: Sage.

Knoke, David. 1990. Political Networks. The Structural Perspective. Cambridge, UK: Cambridge University Press.

Koelble, Thomas A. 1995. "Review: The New Institutionalism in Political Science and Sociology." Comparative Politics 7: 231-44.

Koppenjan, Joop F. M., and Erik-Hans Klijn. 2004. Managing Uncertainties in Networks. London: Routledge.

Kriesi, Hanspeter, Silke Adam, and Margit Jochum. 2006. "Comparative Analysis of Policy Networks in Western Europe." Journal of European Public Policy 13: 341-61.

Kwon, Hyeong-ki. 2004. "Associations, Civic Norms, and Democracy: Revisiting the Italian Case." Theory and Society 33: 135-66.

Law on Higher Education [Högskolelag]. 1992. [Online]. BHTML%7D=sfst_lst&%24%7BOOHTML%7D=sfst_dok&%24%7BSNHTML%7D=sfst_err&%24%7 BBASE°/o7D=SFST&°/o24°/o7BTRIPSHOW%7D=format%3DTHW&BET=1992%3A1434%24. Accessed August 4, 2008.

Lin, Nan. 2001a. "Building a Network Theory of Social Capital." In Social Capital. Theory and Research, ed. Nan Lin, Karen Cook, and Ronald S. Burt. New York: Aldine de Gruyter, 3-29.

--. 2001b. Social Capital. A Theory of Social Structure and Action. Cambridge, UK: Cambridge University


Lin, Nan, Karen Cook, and Ronald S. Burt, eds. 2001. Social Capital. Theory and Research. New York: Aldine de Gruyter.

Lindblom, Charles. E. 1968. The Policy Making Process. Englewood Cliffs, NJ: Prentice Hall.

Lundqvist, Lennart J. 2004. "Integrating Swedish Water Resource Management: A Multi-Level Governance Trilemma." Local Environment 9: 413-24.

March, James G., and Johan P. Olsen. 1989. Rediscovering Institutions. The Organizational Basis of Politics. New York: Free Press.

Marin, Bernd, and Renate Mayntz, eds. 1991. Policy Networks. Empirical Evidence and Theoretical Considerations. Boulder, CO: Westview Press.

Marsden, Peter V. 1990. "Network Data and Measurement." Annual Review of Sociology 16 (1): 435-63.

Marsh, David, and Martin Smith. 2000. "Understanding Policy Networks: Towards a Dialectical Approach." Political Studies 48, 4-21.

--. 2001. "Debates: There Is More than One Way to Do Political Science: On Different Ways to Study

Policy Networks." Political Studies 49, 528-41.

Marton, Susan G. 2000. The Mind of the State: the Politics of University Autonomy in Sweden 1968-1998. Ph.D. diss. Series Göteborg Studies of Politics No. 67. Göteborg, Sweden: BAS Publisher.

Meier, Kenneth. J., and Laurence J. O'Toole Jr. 2001. "Managerial Strategies and Behaviour in Networks: A Model with Evidence from U.S. Public Education." Journal of Public Administration Research and Theory 11: 271-93.

Miles, Matthew B., and A. Michael Huberman. 1994. Qualitative Data Analysis—An Expanded Sourcebook, 2nd ed. Thousand Oaks, CA: Sage Publications.

Mintzberg, Henry 1979. The Structuring of Organizations. Englewood Cliffs, NJ: Prentice Hall.

Moberg, Erik. 1990. Offentliga Beslut [Online] html. Accessed December 7, 2006.

Moody, James and Douglas R. White 2003. "Structural Cohesion and Embeddedness: A Hierarchical Concept of Social Groups." American Sociological Review 68, 103-27.

North, Douglas 1997. "Institutions, Economic Growth and Freedom: An Historical Introduction." In The

Economic Foundations of Property Rights: Selected Readings, ed. Svetozar Pejovich. Cheltenham, UK: Elgar, 87-108.

Oh, Hongseok, Myung-Ho Chung, and Giuseppe Labianca. 2004. "Group Social Capital and Group Effectiveness: The Role of Informal Socializing Ties." Academy of Management Journal 47: 860-75.

Olson, Mancur. 1964. The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press.

Ostrom, Elinor. 2005. Understanding Institutional Diversity. Princeton, NJ: Princeton University Press.

Ostrom, Elinor, and T. K. Ahn, eds. 2003. Foundations of Social Capital. Critical Studies in Economic Institutions 2. Cheltenham, UK: Elgar Reference Collection.

O'Toole, Laurence J., Jr. 1997. "Treating Networks Seriously: Practical and Research-Based Agendas in Public Administration." Public Administration Review 57: 45-52.

O'Toole, Laurence J., Jr., and Kenneth J. Meier. 2004. "Public Management in Intergovernmental Networks: Matching Structural Networks and Managerial Networking." Journal of Public Administration and Theory 14 (4): 469-94.

Peters, Guy B. 1999. Institutional Theory in Political Science: The New Institutionalism. London: Pinter.

Portes, Alejandro. 1998. "Social Capital: Its Origins and Applications in Modern Sociology." Annual Review Sociology 24: 1-24.

Portes, Alejandro, and Patricia Landolt. 2000. "Social Capital: Promise and Pitfalls of its Role in Development." Journal of Latin American Studies 32: 529-47.

Powell, Walter W. 1990. "Neither Market nor Hierarchy: Network Forms of Organization." Research in Organizational Behaviour 12: 295-336.

Provan, Keith G., and Brinton H. Milward. 1995. "A Preliminary Theory of Interorganizational Network Effectiveness: A Comparable Study of Four Community Mental Health Systems." Administrative Science Quarterly 40: 1-33.

-. 2001. "Do Networks Really Work? A Framework for Evaluating Public-Sector Organizational

Networks." Public Administration Review 61: 414-23.

Provan, Keith G., and Juliann G. Sebastian. 1998. "Networks within Networks: Service Link Overlap, Organizational Cliques, and Network Effectiveness." The Academy of Management Journal 41 (4): 453-63.

Provan, Keith G., Kimberly R. Isett, and Brinton H. Milward. 2004. "Cooperation and Compromise: A Network Response to Conflicting Institutional Pressures in Community Mental Health" Nonprofit and Voluntary Sector Quarterly 33: 489-514.

Putnam, Robert. D. 1992. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press.

-. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster.

Raab, Charles D. 2001. "Understanding Policy Networks: a Comment on Marsh and Smith." Political Studies 49: 551-56.

Raab, Jörg. 2002. "Where Do Policy Networks Come from?" Journal of Public Administration Research and Theory 12 (4): 581-622.

Reagans, Ray, and Bill McEvily 2003. "Network Structure and Knowledge Transfer: The Effects of Cohesion and Range." Administrative Science Quarterly 48: 240-67.

Reagans, Ray, and Ezra W. Zuckerman. 2001. "Networks, Diversity, and Productivity: The Social Capital of Corporate R&D Teams." Organization Science 12 (4): 502-17.

Real, T. Alejandra and Nicholas D. Hasanagas 2005. "Complete Network Analysis in Research of Organized Interests and Policy Analysis; Indicators, Methodical Aspects and Challenges." Connections 26 (2): 89-106.

Rhodes, R. A. W. 1990. "Policy Networks: A British Perspective." Journal of Theoretical Politics 2 (3): 293-317.

Rhodes, R. A. W., and David Marsh. 1992. "New Directions in the Study of Policy Networks." European Journal of Political Research 21: 181-205.

Rothstein, Bo. 2003. Sociala Fällor och Tillitens Problem. Stockholm: SNS Förlag.

Sabatier, Paul A., and Hank. C. Jenkins-Smith (eds.) 1993. Policy Change and Learning: An Advocacy Coalition Approach. Boulder, CO: Westview Press.

Sabatier, Paul A., Will Focht, Mark Lubell, Zev Trachtenberg, Arnold Vedlitz, and Marty Matlock. 2002.

Swimming Upstream. Collaborative Approaches to Watershed Management. Cambridge, MA: MIT Press.

Sandström, Annica. 2004. Innovative Policy Networks. The Relation between Structure and Performance. Licentiate Thesis, Lulea University of Technology, Sweden Accessed August 4, 2008.

Sandström, Annica. 2008. Policy Networks: The Relation between Structure and Performance. Doctoral Thesis, Lulea University of Technology. [online] Accessed August 4, 2008.

Scharpf, Fritz. W. 1991. "Games Real Actors Could Play: The Challenge of Complexity." Journal of Theoretical Politics 3 (3): 277-304.

Schneider, Mark, John Scholz, Mark Lubell, Denisa Mindruta, and Matthew Edwardsen. 2003. "Building Consensual Institutions: Networks and the National Estuary Program." American Journal of Political Science 47: 143-58.

Scott, John. 2000. Social Network Analysis. A Handbook. 2nd ed. London: Sage Publications.

Thatcher, Mark. 1998. "The Development of Policy Network Analyses: From Modest Origins to Overarching Framework." Journal of Theoretical Politics 10 (4): 389-416.

Udehn, Lars. 2002. "The Changing Face of Methodological Individualism." Annual Review of Sociology 28: 479-507.

Vedung, Evert. 1997. Public Policy and Program Evaluation. New Brunswick, NJ: Transaction Publishers.

Wasserman Stanley, and Katherine Faust. 1994. Social Network Analysis: Methods and Applications. Cambridge, UK: Cambridge University Press.

Weick, Karl E. 1976. "Educational Organizations as Loosely Coupled Systems." Administrative Science Quarterly 21: 1-19.

Yin, Robert K. 1994. Case Study Research, Design and Methods. 2nd ed. Thousand Oaks, CA: Sage Publications.