R. A. W. Rhodes
‘Tis all in pieces, all cohærence gone.’
(John Donne , ‘The First Anniversary. An Anatomy of the World’, 1985 edition, 335 line 213)
INTRODUCTION: THE UBIQUITY OF NETWORKS
Network analysis comes in many guises. It is common to all the social science disciplines. The vast literature ranges from social network analysis (Scott 2000) to the network society created by the information revolution (Castells 2000), from the actor-centred networks of technological diffusion (Callon, Law and Ripp 1986) to cross-cultural analysis (Linn 1999). This chapter focuses on that species of network analysis most common in political science - policy network analysis.
Few social science disciplines can ever agree on the meaning of an idea. So, a policy network is one of a cluster of concepts focusing on government links with, and dependence on, other state and societal actors. These notions include issue networks (Heclo 1978), iron triangles (Ripley and Franklin 1981), policy sub-systems or sub-governments (Freeman and Stevens, 1987), policy communities (Richardson and Jordan 1979) and epistemic communities (Haas 1992). I discuss these terms below. All are varieties of networks, so I use ‘policy network’ as the generic term.
This buzzing, blooming confusion of terms has not detained us for long. Defining policy networks will take no longer. Policy networks are sets of formal institutional and informal linkages between governmental and other actors structured around shared if endlessly negotiated beliefs and interests in public policymaking and implementation. These actors are interdependent and policy emerges from the interactions between them. There could be many qualifications to this definition, but it will do as a starting point for my exploration.
Part 1 of this chapter reviews the literature on policy network analysis, distinguishing between descriptive, theoretical and prescriptive accounts. It identifies three descriptive uses of the term; networks as interest intermediation, as interorganizational analysis and as governance. It then summarises the two main theoretical approaches - power-dependence theory and rational choice – before looking at the instrumental, interactive, and institutional approaches to managing networks. Part 2 looks at the debates and challenges in the literature. It focuses on the difficulties of synthesizing the findings from the proliferating case studies, and on the critics of the ‘new governance. It reviews the various answers to the question of why do networks change, looking at the advocacy coalition framework, the dialectical model, strategic relational theory, and the interpretive turn. It concludes with the observation that the study of policy networks mirrors general trends in political science in its concern with ethnographic methods and the impact of ideas. Finally, it looks at the problems of managing the institutional void, especially the difficulties posed by mixing governing structures, the diffusion of accountability, enhancing coordination, and devising new tools.
Part 1 – THE LITERATURE ON POLICY NETWORK ANALYSIS
The term policy network is used in three main ways in the literature: as a description of governments at work, as a theory for analyzing government policymaking, and as a prescription for reforming public management.
Networks as description
When describing government policy-making, the term policy network refers to interest intermediation, interorganizational analysis, and governance.
Networks as interest intermediation
The roots of the idea of a policy network lie, in part, in American pluralism and the literature on sub-governments. For example, Ripley and Franklin (1981, 8-9) define sub-governments are ‘clusters of individuals that effectively make most of the routine decisions in a given substantive area of policy’. They are composed of ‘members of the House and/or Senate, members of Congressional staffs, a few bureaucrats and representatives of private groups and organizations interested in the policy area’. The emphasis in this literature is on a few privileged groups with close relations with governments; the resultant sub-government excludes other interests and makes policy. Some authors developed more rigid metaphors to characterise this relationship. Lowi (1964) stressed the triangular nature of the links, with the central government agency, the Congressional Committee and the interest group enjoying an almost symbiotic interaction. This insight gave birth to the best-known label within the sub-governments literature, the `iron triangle' (see Freeman and Stevens 1987, 12-13 and citations).
The literature on policy networks develops this American concern with the oligopoly of the political marketplace. Governments confront a multitude of groups all keen to influence a piece of legislation or policy implementation. Some groups are outsiders. They are deemed extreme in behavior and unrealistic in their demands, so are kept at arms length. Others are insiders, acceptable to government, responsible in their expectations and willing to work with and through government. Government needs them to make sure it meets its policy objectives. The professions of the welfare state are the most obvious example. Over the years, such interests become institutionalised. They are consulted before documents are sent out for consultation. They don’t lobby. They have lunch. These routine, standardised, patterns of interaction between government and insider interests become policy networks.
There are many examples of the use of policy networks to describe government policymaking.1 Marsh and Rhodes (1992) define policy networks as a meso-level concept that links the micro-level of analysis, dealing with the role of interests and government in particular policy decisions, and the macro-level of analysis, which is concerned with broader questions about the distribution of power in modern society. Networks can vary along a continuum according to the closeness of the relationships in them. Policy communities are at one end of the continuum and involve close relationships; issue networks are at the other end and involve loose relationships (and on the influence of this approach see Börzel 1998, Dowding 1995, LeGales and Thatcher 1995, Richardson 1999).
A policy community has the following characteristics: a limited number of participants with some groups consciously excluded; frequent and high quality interaction between all members of the community on all matters related to the policy issues; consistency in values, membership and policy outcomes which persist over time; consensus, with the ideology, values and broad policy preferences shared by all participants; and exchange relationships based on all members of the policy community controlling some resources. Thus, the basic interaction is one involving bargaining between members with resources. There is a balance of power, not necessarily one in which all members equally benefit but one in which all members see themselves as in a positive-sum game. The structures of the participating groups are hierarchical so leaders can guarantee compliant members. This model is an ideal type; no policy area is likely to conform exactly to it.
One can only fully understand the characteristics of a policy community if we compare it with an issue network. McFarland (1987, 146), following Heclo's (1978) use, defines an issue network as ‘a communications network of those interested in policy in some area, including government authorities, legislators, businessmen, lobbyists, and even academics and journalists … [that] … constantly communicates criticisms of policy and generates ideas for new policy initiatives’. So, issue networks are characterised by: many participants; fluctuating interaction and access for the various members; the absence of consensus and the presence of conflict; interaction based on consultation rather than negotiation or bargaining; an unequal power relationship is which many participants may have few resources, little access and no alternative. The study of interest groups understood variously as issue networks, policy sub-systems and advocacy coalitions is probably the largest American contribution to the study of policy networks. They are seen as an ever-present feature of American politics (and for surveys of the literature see Baumgarten and Leech 1998 and Berry 1997).
Obviously the implication of using a continuum is that any network can be located at some point along it. Networks can vary along several dimensions and any combination of these dimensions; for example, membership, integration, resources. Various authors have constructed continua, typologies and lists of the characteristics of policy networks and policy communities (see for example,Van Waarden 1992). This lepidopteran approach to policy networks – collecting and classifying the several species – has become deeply uninteresting.
Networks as interorganizational analysis
The European literature on networks focused less on sub-governments and more on interorganizational analysis (see for example Rhodes (1999 ). It emphasises the structural relationship between political institutions as the crucial element in a policy network rather than the interpersonal relations between individuals in those institutions. At its simplest, interorganizational analysis suggests that a ‘focal organization attempts to manage its dependencies by employing one or more strategies, other organizations in the network are similarly engaged.’ A network is ‘complex and dynamic: there are multiple, over-lapping relationships, each one of which is to a greater or lesser degree dependent on the state of others’ (Elkin 1975, 175-6).2
The most impressive attempt to apply this variant of network analysis to politics and policymaking is the several collaborations of David Knoke, Edward Laumann and Franz Pappi (see especially Knoke 1990, Knoke et al 1996, Laumann and Knoke 1987). Their ‘organizational state’ approach argues that ‘modern state-society relationships have increasingly become blurred, merging into a mélange of interorganizational influences and power relations’. These interorganizational networks ‘enable us to describe and analyze interactions among all significant policy actors, from legislative parties and government ministries to business associations, labor unions, professional societies, and public interest groups’ (Knoke et al 1996, 3). The key actors are formal organizations, not individuals. In their analysis of national labor policy in America, Germany and Japan, Knoke et al 1996 compiled the list of key actors by, for example, searching public documents such as the Congressional Information Service volumes for the number of times they testified before the relevant Congressional or Senate Committee, including only organizations with five or more appearances. The individuals in these organizations responsible for governmental policy affairs were then interviewed on such matters as the informant’s perception of the most influential organization, the communication of policy information, and participation in the policy area. Knoke et al then use the techniques of network analysis to map the links between organizations, employing classic network measures such as centrality and density (for an introduction to such techniques see Scott 1991 and for a compendium see Wasserman and Faust 1994).
Knoke et al argue their data not only describes the power structure of their chosen policy area but also explains the different policy outcomes. The value of this species of network analysis lies in its use of the structural properties of networks to explain behavior and outcomes. Unfortunately, little work in this idiom is explanatory. Instead, it describes power structures and network characteristics. Moreover, ‘it has not yet produced a great deal that is novel’ (Dowding 2001, 89-90 and n. 2). It is hard to demur from this judgment when Knoke et al (1996, 210, 213) conclude that ‘the state clearly constitutes the formal locus of collective decision making that affects the larger civil society within which it is embedded’, or that ‘the more central an organization was in either the communication or the support network, the higher was its reputation for being influential’ (see also Thatcher 1998, 398-404).
Networks as governance
The roots of policy network analysis lie, finally, in the analysis of the sharing of power between public and private actors, most commonly between business, trade unions and the government in economic policymaking (Atkinson and Coleman 1989, Jordan 1981). Initially, the emphasis fell on corporatism, a topic worthy of an article in its own right (see Cawson 1986, Schmitter and Lehmbruch 1979). There was also the longstanding and distinctive Scandinavian analysis of ‘corporate pluralism’ (Rokkan 1966, Heisler 1979), which continues under such labels as ‘the segmented state’ (Olsen 1983, 118) and ‘the negotiated economy’ (Nielsen and Pedersen 1988). Latterly, the main concern has been with governance by (and through) networks, on trends in the relationship between state and civil society government rather than policymaking in specific arenas. Thus, governance is a broader term than government with public resources and services provided by any permutation of government and the private and voluntary sectors (and on the different conceptions of governance see Kjær 2004, Pierre 2000).
There are several accounts of this trend for Britain, continental Europe and the USA. Thus, for Britain, there has been a shift from government by a unitary state to governance by and through networks. In this period, the boundary between state and civil society changed. It can be understood as a shift from hierarchies, or the bureaucracies of the welfare state, through the marketization reforms of the Conservative governments of Thatcher and Major to networks and the emphasis on partnerships and joined-up government.3
There is also a large European literature on ‘guidance’, 'steering' and ‘indirect coordination’ which predates both the British interest in network governance and the American interest in reinventing government. For example, Franz-Xavier Kaufmann’s (1986) edited volume on guidance, steering and control is truly Germanic in size, scope and language. It focuses on the question of how a multiplicity of interdependent actors can be coordinated in the long chains of actions typical of complex societies (see also Bovens 1990, Luhmann 1982, Van Gunsteren 1976).
For the USA, Osborne and Gaebler (1992, 20 and 34) distinguish between policy decisions (steering) and service delivery (rowing), arguing bureaucracy is a bankrupt tool for rowing. In its place they propose entrepreneurial government, with its stress on working with the private sector and responsiveness to customers. This transformation of the public sector involves 'less government' or less rowing but 'more governance' or more steering. In his review of the American literature, Frederickson (1997, 84-5) concludes the word ‘governance is probably the best and most generally accepted metaphor for describing the patterns of interaction of multiple-organizational systems or networks’ (see also Kettl 1993, 206-7, Salamon 2002). Peters (1996, chapter 1) argues the traditional hierarchic model of government is everywhere under challenge. He identifies four trends, or models of governance, challenging the hierarchic model - market, participative, flexible and deregulated governance. Fragmentation, networks, flexibility and responsiveness are characteristics of flexible governance. In sum, talk of the governance transformation abounds even if the scope, pace, direction and reasons for that change are matters of dispute (and for a survey see Pierre 2000).