The Advocacy Coalition Index: Identifying Coalitions by Simultaneously Taking into Account Coordination of Action and Belief Homophily

The Advocacy Coalition Framework (ACF) argues that policy actors who hold similar policy beliefs tend to coordinate their actions and form coalitions to achieve their policy goals. The first step for testing hypotheses related to the ACF is the identification of coalitions. The definition of an advocacy coalition is that “actors from a variety of positions who share a particular belief system” (the belief homophily condition) also “show a non-trivial degree of coordinated activity over time” (the coordination condition). Because of this two-fold definition, operationalization of coalitions is often difficult, especially for social network analysis. Some researchers start by identifying the existence of the coordination among actors and then examine the similarity of beliefs among the network subgroups that exhibit coordination. Others proceed in the opposite order: they first try to find subgroups based on belief similarity and then analyze them for coordination. We propose a new way of identifying coalitions that simultaneously takes into account both belief homophily and coordination: the Advocacy Coalition Index. This approach postulates criteria for theoretically “ideal” advocacy coalitions in a given policy domain and measures the divergence of the empirical coalitions existing in that domain from this ideal. The identification of coalitions takes three steps: (1) each pair of actors is plotted in a two dimensional graph where x-axis indicates the existence of a coordination relationhship and y-axis the similarity of beliefs; (2) the distance of the plotted pairs from the theoretically ideal advocacy coalitions are calculated; (3) the relationships that deviate from the “ideal” coalitions are discarded so that only relationships that meet the definition of advocacy coalitions remain. The major advantage of this approach is that it enables the researcher to compare different datasets and thus, assess the degree to which coalitions are present in each one of the policy domains analyzed. It also allows for a cut-off value to be set for defining whether meaningful coalitions exist in a given policy domain or not. The indexed score can also be aggregated from pairs level into actor and group level so that researchers can identify the different roles actors/groups play in networks (e.g. brokers). As an illustration of the approach, we analyze the climate change policy networks of Finland and Sweden. The data is based on policy network surveys where the respondents constitute the climate change policy subsystems in both countries. Although Finland and Sweden are often regarded as similar countries in terms of the political system and other socio-economic conditions, we find differences when it comes to the climate change advocacy coalitions. Finland exhibits a typical advocacy coalition constellation with opposing ecology and economy coalitions. The economy coalition is stronger and consequently, Finland’s climate change policy over the years has been relatively weak. In Sweden, we observe a large pro climate coalition and many actors with intermediating ties. Correspondingly, Sweden’s climate change policy has been very strong. We also compare the results obtained with the advocacy coalition index with other ways of analyzing coalitions, such as factions analysis and clustering analysis.