Collection and Measuring Data on Venue Participation

This section analyzes the participation within the policy process, by examining the partic-ipation in venues that deal exclusively or partially with the energy transition at the local level. To tailor the questionnaire to the respondents, the venues were accustomed to the four counties under study. Therefore, experts of the respective counties were asked in ad-vance if organizations come together to discuss issues relevant for the energy transition in formal or informal venues. From three of the four counties, at least one expert responded, and the venues were included as answers in the survey. In one county, no expert oered

insights on the venues. The experts' compilation of important venues was complemented with the city and county council meetings as well as with further commissions at the county level that targeted the local energy transition. Furthermore, in order to capture venues previously not named, the option of adding additional venues was oered in the survey. Since the venues were tailored specically for each county, only respondents choos-ing one of the analyzed counties as their primary area were asked this question. in total, 31 survey respondents answered the following question for all potential venues:

Please indicate how often your organization participates in each of the follow-ing venues. (1) Permanently / (2) Sometimes (e.g. as advisfollow-ing expert) / (3) Never / (4) Do not know this venue / (5) N/A

The results of this question can be visualized in a two-mode network in which the re-spondent's organizations are one node type and the venues are the second node type.

The links then represent the participation of one respondent in a given venue. These links have values according to the answer given: Permanently (1); Sometimes (e.g. as advising expert) (2); Never (3); Do not know this venue (4); N/A (5); and Did not answer (0).

While the values (1) and (2) indicate (at least partially) participation in the given venue, the values (3), (4), and (5) indicate that an organization does not participate in a given venue, however, the reasons for non-participation may vary between the answers. The value (0) on the other hand means, that the respondent did not select any of the given choices, and therefore, does not allow for such an interpretation.

The results are analyzed in two steps: In the rst step, weighted aliation networks are drawn, connecting organizations and venues. The links have the value 1, if the organization is a permanent member of the venue and the value 0.5, if the organization attends from time to time, for example as an advising expert. In the second step, co-occurrence net-works are derived from the aliations. Organizations are connected in the co-occurrence network, if they participate in the same venue(s). Both network types are then analyzed descriptively and with quantitative measures (see below) in order to understand, (1) who participates in the policy preparation, decision and implementation in local venues, (2) how inclusive the venues are, and (3) how the networks are structured.

The quantitative measures applied here are introduced in detail in section5.2. Here, it is briey outlined how the measures are applied to test the presented hypotheses. In order to understand the role of certain policy venues for the organizations involved in local energy policy-making, the centralities of the venues are analyzed in the aliation net-works of organizations and venues. The degree centralities, hereby, quantify how high the share of organizations that participate in a given venue is. Therefore, the standardized degree centrality in a two-mode networkC_D⁰⁰(v_i)is calculated for each venuev_i (as dened in equation 5.12). If non-institutionalized venues are created with the purpose of foster-ing communication between political and non-political actors, they should exhibit higher

betweenness centralities than institutionalized venues. This is due to the fact, that insti-tutionalized venues might create redundant paths, while alternative venues create paths between otherwise unconnected groups. Technically, the two-mode betweenness centrality C_B⁰⁰(v_i) (dened in equation 5.16) measures, on how many shortest paths between two organizationso, the venue vi lies. In order to make the degree and the betweenness cen-tralities comparable, both centrality measures are calculated in percent, respectively as (see also section 5.2.2):

C_D%⁰⁰ (v_i) = C_D⁰⁰(v_i) P

iC_D⁰⁰(vi) and C_B%⁰⁰ (v_i) = C_B⁰⁰(v_i) P

iC_B⁰⁰(vi)

In network terms, an inclusive policy subsystem can be understood as a network with or-ganizations able to reach each other quickly, e.g. by sitting at the same table. The speed of reachability can be quantied via the closeness centralityC_C⁰ (ni)of the actors i in the co-occurrence network (Beauchamp, 1965; Wasserman and Faust, 1994, 185), which is dened in equation (5.6 in section 5.2.1). The higher the closeness centrality of an actor, the faster this actor can reach out to all other actors. Based on the closeness central-ities of all actors, Freeman's (1979) standardized closeness centralization C_C (equation 5.8) characterizes the network as a whole (see also Wasserman and Faust, 1994, 186).

The resulting closeness centralization quanties the distribution of closeness centralities within one network in one value, and allows the comparison of this network characteristic between the dierent networks. One might argue that the smaller the closeness central-ization (i.e. small dierences in closeness centrality values), the better the communication within the network. The variance of the standardized actor closeness indicesSC (equation 5.9) captures the heterogeneity of the underlying closeness centralities. A mathematical introduction and detailed description of the centrality measures applied here is provided in subsection5.2.1.

Although closeness centralization is a good measure for understanding how closeness cen-tralities vary between the actors, the measure compromises a rened understanding of how well single organizations can be reached. A more rened way to study the speed of reachability, is to study the distribution of the geodesic distances within a network (Knoke et al.,1996, 110). Therefore, the geodesic distances for all pairs of actors are cal-culated and it is derived how many pairs are connected by a 1-path, 2-path, . . . , n-path.

The smaller the geodesic distances between the organizations, the faster the communi-cation between them, which in turn inuences the inclusiveness of the decision-making process. In the cases at hand, the length of the geodesic distances is highly dependent on whether sucient venues exist to include actors from outside the administrative and political sphere.

Hypothesis 6 suggests that dierences might be observable in the positions of

politi-cal (parties and administrative actors) as compared to non-politipoliti-cal actors. Hence, the networks are tested for core-periphery structures.³¹ In its simplest form, a test for core-periphery structure explores whether a core of actors, which is more densely connected than actors outside this core, exists. Since an a priori hypothesis is tested, it is justiable to work with the algorithm based on the block modeling approach proposed by Borgatti and Everett(1999). This approach maximizes the correlation between the permuted data matrix and an ideal structure matrix consisting of ones in the core block interactions and zeros in the peripheral block interactions (Borgatti and Everett, 1999). To understand whether a core-periphery structure exists, the density of the permuted matrix blocks is calculated (i.e. density within the core, density between core and peripheral actors, and density within periphery). A substantial dierence of these densities, and a high t score (measuring the correlation between the permuted data matrix and an ideal structure matrix (Borgatti and Everett, 1999)), would emphasize that a core-periphery structure is present.

6.4 Results