• Keine Ergebnisse gefunden

3. Theoretical Framework

3.3 Analytical Tools

3.3.2 Networks

In order to use networks as an analytical tool we must assume that structures can be represented through networks and that networks matter in the policy process. The following chapter will extensively discuss the scholarly debate on networks and then elaborate a definition of network that will inform the subsequent analyses. Especially, the usefulness of network analysis in the European setting will be discussed in more detail.

There is an increasing trend in applying network analysis in social sciences. Up to 2007, over 1000 studies, making use of this concept as a metaphor, as a theory or as a method, have been published (Schneider, Lang, Leifeld, & Gundelach, 2007, p. 158). An updated glance at the literature confirms the trend. As can be seen in Figure 4, over the last 10 years research applying the concepts of networks has doubled.

Figure 4 Published Items per year using keyword “policy networks”31

It is thus not surprising that the scientific community has shown a particular interest in debating the usefulness but also pitfalls of network analysis. Schneider (2009) detects over twenty review articles that discuss the applications of network analysis across social sciences. Nonetheless, apart from the seminal article by Thomas Risse-Kappen (1996) in which he argued for applications of network analysis to “capture distinctive features of the EU” (Risse-Kappen, 1996, p. 53) and Tanja Börzel, who draw our attention to conceptual fuzziness of networks (Börzel,

31 The keyword “policy networks” was used in the database “Social Science Citation Index”. The results were

refined by: Web of Science Categories=(PUBLIC ADMINISTRATION OR POLITICAL SCIENCE OR SOCIOLOGY OR SOCIAL SCIENCES INTERDISCIPLINARY OR INTERNATIONAL RELATIONS OR SOCIAL ISSUES OR AREA STUDIES OR LAW).[Last accessed April 2013].

1997; Börzel & Heard-Lauréote, 2009), a systematic analysis of applications of network analysis as a method not as a concept in the European context is still missing and thus will be briefly summarized in this chapter. Such an in-depth analysis is particularly interesting given Risse-Kappen’s thesis that sui generis polities require sui generis methods.

Naturally, the applications of network analysis in the European context reflect the general methodological trend. After the first euphoric applications hailing the heyday of policy networks as a theory and as a method (Rhodes, 1990), a period of harsh criticism followed (Dowding, 1995; Kassim, 1994; Thatcher, 1998). In 1995 Keith Dowding warned that “to promise that network analysis will eventually go beyond demonstrating general features of networks will ultimately lead to disappointment” (Dowding, 1995, p. 158). In a similar vein, in the European context Kassim (1994) has argued that that network perspective is inappropriate to study EU processes because it is not able to capture the complexity of the EU institutions and EU processes. This critique came mainly from the fact that this approach was applied to describe certain structures and thus being very descriptive and consequently of little scientific relevance.

Nonetheless, what could be observed during the past years is that network analysis has celebrated renaissance and appeared in special issues of political science journals, such as Journal of Public Policy (2009, Special Issue 2) or PS: Political Science & Politics (2011, Issue 1). Moreover, the publication of the article “Network Analysis in the Social Sciences” by Borgatti and his colleagues in “Science” signals that network analysis is regarded as an established and useful method (Borgatti, Mehra, Brass, & Labianca, 2009). This sudden rebirth of network analysis can be explained with the advances in statistical methods and data collection in political science and the emergence of new research questions, stemming from the growing complexity of political processes. The concept of networks, which was first developed in sociology and anthropology (Lazer, 2011, p. 61), has added sociological aspects to the studies of policy processes. It seems to enable addressing more adequately the “agent-structure” problem and offer a more transparent method to identify the main actors of a policy process.

In the European context similarly to Risse-Kappen (1996), others have suggested that network analysis could be particularly useful to study the EU policy making (Héritier, 1993) and that contrary to Kassim’s (1994) critique “the European Union is an obvious focus for studying

policy networks as informal coordination mechanisms between state and non-state or public and private actors [increase]” (Kaiser, 2009b, p. 131).

Nonetheless, the answer to the question whether network analysis is particularly useful in the EU context should be separated from the analyses that utilize networks as a mode of governance. Its usefulness as an analytical tool should not depend on whether we expect the EU to be governed through networks or whether – as most recent research suggests – “the EU features far less network governance than the literature would have us believe” (Börzel, 2011, p. 53). Its usefulness depends on whether it helps us to obtain information that would not be possible to obtain by utilizing other methods.

It was already indicated that due to its polysemic nature, network analysis has been characterized by a certain degree of conceptual fuzziness (Börzel, 2011). Intuitively, networks could be viewed as “structure through which things circulate” (Lazer, 2011, p. 62). This definition comes close to the mathematical definition of a network through the notion of graphs. Graph is defined as “an abstract object formed by a set V of vertices (nodes) and a set of E edges (links) that join (connect) pairs of vertices” (Brandes & Erlebach, 2005, p. 7). The much more controversial is the question of how the vertices and edges are defined and consequently how the whole network should be conceptualized. Generally, in political science context a policy network is defined as a

“a set of political actors who engage in resource exchange over public policy (policy decisions) as a consequence of their resource interdependencies” (Compston, 2009, p. 11).

Apart from this broad definition, and particularly in the EU context, network analysis is used to describe certain types of governance. These applications are considered as qualitative applications of the network analysis. There have been several typologies suggested to describe the evolving interest intermediation structures between state and society (see Börzel, 2011), the most influential being “the network governance” introduced by Rod Rhodes (1996). Rhodes argued that what could be observed in the policy space is that state centric government was substituted by network governance, which included a wide range of public and private actors linked to each other across different polity levels.

A systematic literature analysis shows that nodes are mostly defined as public and private actors but the definition might also be much more restrictive and concentrating only on party elites

(Johansson, 2002) or on members at a certain level of governance (Beyers & Dierickx, 1998). In turn, much broader definitions are also practiced, if, for instance, also non-human actors are included by assuming that human actors can interact with non-human actors such as nature (Burgess, Clark, & Harrison, 2000). Schneider (2009) in his analysis of social science literature locates 7 different conceptualizations of the term “policy network”, which vary depending on their connection to policy, politics or polity aspects. These seven different definitions range from a small “issue network” to a policy network understanding, which seeks to include all relevant actors as for example was applied by Franz Urban Pappi (1993), who used snowball interview methods to inductively determine the most important actors in a policy process. Additionally, Ward & Williams (1997) emphasize the temporal typologies and distinguish between policy communities, durable established networks, issue networks and ad-hoc spontaneous mobilization of certain networks. Essentially, “the nature of the beast”, the definition of networks, depends on assumptions about actors involved in a policy process, the temporal scope of a policy process and the definition of the links between them.

Regarding the conceptualization of links, at first, there seems to be much more agreement in that it can be best understood as an exchange of information through different communication channels. However, operationalizations vary greatly. The major drawback is that most data is collected through interviews, asking “actors” to indicate with whom they exchange the most information or meet most frequently, and thus are often distorted by reporting biases (in particular when the respondents are asked to indicate their communication preferences for the past 5 years). More importantly, there is no clear threshold on what level of information exchange is sufficient for a link to be created and thus to sustain a network.

Given the plethora of possible definitions I suggest to rely on the definition that is closest to the mathematical understanding of a graph as it seems to be also the most straightforward definition.

Additionally, given that the thesis develops a dynamic setting, the networks should be restricted by a particular period of time. Based on these considerations I suggest the following definition:

A network is an object G at time t, which contains set of entities V at time t, which are connected together by the set of links at time t.