What Beliefs Are Central in the Climate Change Belief System Network?

Objective: Climate change is one of the most polarized issues among the US public, and the beliefs that divide the public –cultural worldviews, political ideology, environmental orientation, perceived scientific consensus, and climate policy preferences – form an interconnected belief system. Belief systems have historically been understood as having a hierarchical structure, in which abstract, deep core beliefs constrain policy core and secondary beliefs. Recent work suggests that such a system can be understood as a network, in which beliefs are nodes, edges represent the relationships between them, and centrality captures a belief’s structural importance. This paper asks which climate change beliefs are most central, and whether the answer depends on how the belief system is modeled. Methods: I use original survey data from a quota-based sample of 1,129 US residents collected in October 2017. Over the same 17 belief nodes, I estimate the belief system as both an undirected network, using a Gaussian graphical model, and a directed network, using a Bayesian network learned with hill-climbing, tabu search, and constraint-based algorithms. Structural importance is measured with bootstrapped betweenness, closeness, and strength centrality in the undirected model, and with out-degree and Markov blanket size in the directed model. Results: Structural importance cuts across the deep core, policy core, and secondary belief hierarchy that theory predicts. Individualism, perceived climate risk, and support for an international agreement are robustly central regardless of whether the network is modeled as undirected or directed, and support for an international agreement occupies the most upstream position in the directed network. Egalitarianism ranks highest on betweenness and closeness in the undirected network yet is peripheral in the directed one. Political ideology, partisanship, and perceived scientific consensus rank low on nearly every measure in both models. Conclusions: A belief’s structural importance likely depends on both the type of belief and modeling assumptions, complicating any single account of how climate change beliefs are organized. Rather than replacing the hierarchical approach common in policy process research, a network approach likely offers additional analytic leverage, particularly when undirected and directed models are estimated together.