The conditions for policy-oriented learning: a dyadic analysis of the Finnish climate policy subsystem

Discussion and negotiation are key components of the democratic process. Actors often communicate each other with the hope that such communication can result in the change of opinions. The Advocacy Coalition Framework (ACF) has conceptualized changes in actor’s beliefs and behavior as policy-oriented learning and hypothesized that it is one important source of policy change. Despite its theoretical importance, there are still few empirical studies on how and when policy-oriented learning occurs among stakeholders. Moreover, many studies conclude that actors seldom change their beliefs, and that the polarization of beliefs is more prevalent than belief convergence. Consequently, to our knowledge, the conditions under which communication results in belief convergence are yet to be systematically analyzed. For analyzing the conditions susceptible for belief convergence, we propose two innovations with regard to the way of conceptualizing the idea. First, we differentiate between communication within coalition and between coalitions. Prior studies tend to treat communication among actors without taking their affiliated coalitions into account and therefore mix up learning effects within and between coalitions. The latter (i.e., communication between coalitions) is more theoretically important when studying policy-oriented learning and policy change. Second, we take dyads (i.e., communication channel between two actors) as our unit of analysis, while prior studies tend to focus on nodes (i.e., single actors). Our approach is more akin to the Weberian way of conceptualizing power. Max Weber defined power as the probability as an actor in a social relationship being in a position to carry out his/her will despite resistance by others. Accordingly, we assume that the effect of persuasion through communication is determined not only by the sender’s power for persuasion but also by the receiver’s ability to resist. We operationalize our framework through social network analysis and two-wave panel data of the Finnish climate policy subsystem. The data consists of main organizational stakeholders who reported their policy beliefs as well as their communication partners (i.e., source of scientific information and collaboration). The analysis takes three steps. First, each actor’s advocacy coalitions are identified with the Advocacy Coalition Index. Second, the conditions for belief change are analyzed through the multiple regression quadratic assignment procedure in which the degree of belief polarization/convergence of a dyad is set as the dependent variable and several sender and receiver effects of communication are set as independent variables. The relative reputation of the sender of scientific information is the main independent variable. Third, the association between change in beliefs and in reputational power is evaluated both descriptively and inferentially. Our study connects the conditions of policy-oriented learning, changes in reputational power, and the structures of coalitions. Accordingly, it contributes to advancing the conceptualization of subsystem dynamics through policy-oriented learning.