• Keine Ergebnisse gefunden

Evolving Structures of Local Discourse Networks - DNA ResultsResults

Studying Actors' Beliefs and Their Eect on Network Evolution

8.1 The Evolution of Discourse Networks

8.1.2 Evolving Structures of Local Discourse Networks - DNA ResultsResults

Chapter 6discussed the presence of actors in the discourse in terms of actor proportions and statement proportions. This section goes one step further by analyzing which actor constellations arise due to shared support for certain policy issues and their respective solutions.

Figures 8.1 to8.4 depict the co-occurrence networks of the four counties over the course of the four years. In these co-occurrence networks, links between two organization exist whenever they support a common statement. The more statements the organizations com-monly agree on, the more weight has the link between the two of them and the darker it is depicted in the graph. While specic actor groups might look strong because of their inter-nal density, they might not be very present within the discourse. Therefore, the size of the nodes depicts the statement frequency of specic actors, thus visualizing their presence within the discourse. In a rst step, these discourse networks are analyzed descriptively.

In a second step, dierences and developments in the discourse networks are analyzed by utilizing graph level measures.

Descriptive Analysis of the Co-Occurrence Networks. Figure 8.1 visualizes the co-occurrence network of the more successful urban county HA. The energy supplier Enervie dominates the discourse around the local energy transition inHA. Additionally, the community-owned local energy supplier SWS is very present as well. This dominance is counterbalanced by the presence of actors from the local administration (LA HA) and

Figure 8.1: Hagen (urban, successful): Co-Occurrence Network of Organizations Based on Common Categories

from the two state ministries: Ministry for Climate Protection, Environment, Agriculture, Nature and Consumer Protection [Ministerium für Klimaschutz, Umwelt, Landwirtschaft, Natur- und Verbraucherschutz] (MKULNV) and Ministry for Economy, Energy, Con-struction, Habitation, and Transportation [Ministerium für Wirtschaft, Energie, Bauen, Wohnen und Verkehr] (MWEBWV). In the graph 8.1 those three administrative actors as well as the energy supplier Enervie are bridging the gap between the otherwise almost remote groups. However, as the longitudinal visualization of the networks in gureB.17in the appendix shows, they cannot be understood as brokers, but are instead the few actors which are present in the discourse in all four years. Furthermore, the local administration as well as theMWEBWV share an important amount of categories with the energy sup-plier Enervie. This indicates that the statements which the energy supsup-plier makes in the discourse are backed by administrative actors in the discourse and vice versa, emphasizing the strong connection between these actors. FigureB.17in the appendix shows that what looks like a strongly clustered discourse network, when depicted over the course of the four years, are actually rather dense networks when studied over the course of one year only.

This indicates that dierent categories are mentioned in the dierent years, reecting a change in the content of the discourse.

Figure 8.2 shows the co-occurrence network of the less successful urban county BN.

Two observations are particularly noteworthy: First, the Municipal Energy Supplier Bonn

Figure 8.2: Bonn (urban, less successful): Co-Occurrence Network of Organizations Based on Common Categories

Figure 8.3: Alb-Donau-Kreis (rural, successful): Co-Occurrence Network of Organizations Based on Common Categories

[Stadtwerke Bonn] (SWB) is with 33 statements the most present actor in the discourse.

The organization uses a great variety of categories in their argumentation, thus connect-ing dierent groups of actors in the discourse. Second, plural administrative and political organizations are very present in the discourse as well (depicted by the turquoise nodes), clearly being the most present type of actor in this discourse. Further energy suppliers (orange) as well as environmental groups (green) and other organizations play only a mi-nor role in the discourse. The longitudinal depiction of the discourse networks in BN in gureB.18shows a strong increase in actors and density in the time period of 2012-2013.

Followed by a very scarce and strongly clustered discourse network in 2013-2014.

Summing up, the gures8.1and8.2, and even more the guresB.17andB.18, depict that both urban discourses are dominated by various energy suppliers as well as administrative and political actors. Further actors, such as environmental groups and initiatives for and against the transition, are far less present in the discourse.

Figure 8.3 shows the case of the successful rural county ADK. The network depicts a dense cluster of political and administrative actors. Actors from the local administra-tion (LA ADK ) are most present in the discourse. They share the support for certain categories with many other administrative as well as political actors, such as the well representedCDU and the regional association [Regionalverband] (RV D-I ). Additionally, the environmental NGOUnion for Environment and Nature Protection Germany [Bund für Umwelt und Naturschutz Deutschland] (BUND) is very present in the discourse. Its embeddedness into the cluster of political and administrative actors depicts that the state-ments theBUNDmakes are strongly supported by many of these actors. This presence of

Figure 8.4: Bodenseekreis (rural, less successful): Co-Occurrence Network of Organiza-tions Based on Common Categories

an environmental NGOis unique among the discourse networks. Figure B.19 shows that the coherence between political and administrative actors and the presence of theBUND remains stable over the course of the four years.

The co-occurrence network of the rural countyFN, which is less successful in implement-ing RE, is visualized in gure 8.4. The network depicts a coherent coalition of local political and administrative actors which is very present in the discourse, consisting of the Greens, the CDU, the local administration LA FN, and the regional planning asso-ciation RV B-O. Yet, this group is not well interlinked to other groups in the network.

While all organizations within this coalition seem to generally favor the energy transition, they agree up on two categories that might be responsible for the slow progress when it comes to implementing RE: First, they see the pondering of the environmental very critically and second, their majority speaks against the implementation of wind power within their own region. Beside those local political and administrative actors, the energy agency of the county EA FN, the municipal energy provider SAS, and the UM BW are present in the discourse as well. As gure B.19 visualizes, is the presence of the strong administrative and political coalition stable over the course of the four years. Lower link threshold values in the yearly networks depict furthermore a greater overlap of statements

with other actors in the network.

Comparing the co-occurrence networks of the four cases over time unveils that the lo-cal discourses around the energy transition had dierent peaks in the four counties. The amount of actors who participate in the local discourse as well as the network densities vary distinctivly in all four cases. While in BN and FN the discourse is most vital in the time period 2012-2013 and attening afterwards, in ADK most actors participated in the discourse in the time period 2011-2012, and in HA the discourse picked up inten-sity in 2013-2014. No clear patterns of increasing or decreasing discourse inteninten-sity are observed, suggesting that the local discourse around the energy transition follows local peculiaritiesfor example local decisions around new projectsrather than merely the national discourse. The overall networks show that in all four discourses, the political and administrative actors are very present and in the urban discourses the energy providers are very present as well. Comparing this observation with the longitudinal visualizations shows that dierences in the presence of political and administrative, energy suppliers, and other actors are less pronounced within a single year. Instead, the presence in the overall discourse reects the persistence of the political and administrative actors and the energy suppliers over all four years, while the participation of other discourse participants uctuates substantially.

The descriptive analysis of networkslike the one performed aboveis commonly applied in the network literature in general and with discourse networks in particular. A simple task in trivial networks becomes an unbearable task in more complex networks and when comparing multiple networks with dierent characteristics with each other. Therefore, the following evaluation complements the descriptive analysis with an analysis of graph level measures, which enables to explain variances between the cases and their timely evolution.

Comparing Graph Level Measures. The hypotheses H9a and H9b suggest that the discourses in successful and less successful counties might dier in their development.

Graph level measures which are able to capture segregation, modularity, and disagree-ment in the networks where introduced above, together with their theoretical meaning (table8.1). The hypothesis H9a expects that in successful counties, all ve measures will decrease over time, which would correspond with a more open discourse, less segregation between unlike actor types, less polarization in belief coalitions and policy issues, and less disagreement between actors. Contrary, the hypothesis H9b expects an increase of the measures in less successful counties, mirroring a discourse in which the actors get po-larized and segregated over arguments, and in which disagreement increases. In order to test these hypotheses, ve graph level measures were calculated for the four counties: The agreement modularity (a.mod), the agreement segregation (a.seg), the common modularity (c.mod), the common segregation (c.seg), and the disagreement density (d.density). In or-der to account for the longitudinal evolution, the analysis was performed for four one-year

Table 8.2: Graph Level Measurements (r: rural, u: urban, s:successful, ls: less successful) County timeperiod a.mod a.seg c.mod c.seg d.density ADK (r,s)

periods. Table 8.2 summarizes the results for all cases and time periods. The results for the successful rural case ADK show a decrease in agreement modularity, agreement seg-regation, common modularity, and common segregation. Thus, leading to a less clustered and more open discourse over the time of the analysis. In the successful urban county HA the tendencies are less pronounced. A negative tendency is observed for the agree-ment segregation and the common segregation, while both modularity measureagree-ments stay rather stable over time. Not in line with the hypothetical expectations is the development of the disagreement density, which is either stable (HA) or increasing (ADK), thus not showing the expected decrease of disagreement over time. In total, the ndings suggest that the discourse in successful counties becomes indeed more open, less segregated, and less modularized over time. However, disagreement between the actors seems to increase.

Thus, the results support the rst part of the hypothesis H9a while rejecting the second.

The discourse of the less successful urban county BN shows clear tendencies with in-creasing levels of all ve measurements. Indicating an increase in segregation, modularity, and disagreement. However, the rural county FNshows only minimal variance over time.

While the common modularity and the disagreement density clearly increase, the other three measurements are rather uctuating over time. Summing up, where tendencies are observable in the less successful cases, the measures increase over time, suggesting that actors in the discourses get more segregated over time regarding the concepts they sup-port, and that they increasingly talk about dierent policy issues.

Altogether, the variances in the values indicate that dierences between the development

of the discourse network structures in successful and less successful counties do indeed exist. While the discourse in successful counties becomes more open, less segregated, and less modularized, both, in terms of commonly supported concepts as well as regarding the policy issues that are mentioned, the opposite seems to be the case within the less successful counties. In combination with increasing disagreement in all four counties, this observation suggests that in less successful counties more disagreement is accompanied by a drifting apart of various groups in the discourse, while successful counties perform better in compensating this disagreement. However, the results call for further research to verify these tendencies with larger sample sizes.

The here applied measures of modularity, segregation, and disagreement density are graph level measures which allow to analyze network structures as a whole and their evolution over time. Another perspective to study the development of networks starts from an individual perspective and studies how individual preferences to build relations aect networks structures. The next section takes such a perspective by studying which eect beliefs, trust, agreement, and perceived inuence have on link formation of individual actors.

8.2 The Inuence of Shared Beliefs and Actors'