Politics of problem definition: Comparing public support of climate change mitigation policies using machine learning

Abstract

Public support is a key contributor to successful policy adoption and implementation. Given the urgency of climate change mitigation, scholars have explored various determinants that affect public support for climate change mitigation policy. However, the relative decisiveness of these factors in shaping public support is insufficiently examined. Therefore, we deploy interpretable machine learning to understand which factors, among many previously investigated, are most decisive for structuring public support for various climate change mitigation policies. In this paper, we particularly look at the decisiveness of problem definition for shaping public support among various factors. Using U.S national survey data, we find that how individuals define the issue of climate change is more decisive for structuring public support for promoting renewable energy and regulating pollutants to mitigate the risks associated with climate change. However, the results also indicate that the most decisive factors associated with public support vary depending on the types of mitigation policy. We conclude that different strategies should be utilized to increase public support for various climate change mitigation policy options. Our findings contribute to a scholarly understanding of the specific politics of problem definition in the context of environmental and climate change policy.

Perception and performance in environmental policy

“…Only informing and communicating the challenges of climate change and the benefits of climate mitigation policy will not lead to successful adoption and implementation. Addressing this concern and turning toward the factors that influence public support for climate change mitigation policy, Choi et al (2024) argue that problem definition is the most decisive factor that shapes public support for these policies when compared to demographic characteristics, political and social predispositions, social trust, and individual experiences and attitudes. The methodological analysis is based on machine learning techniques that do not make as strong parametric assumptions as regression models and the like, for example, regarding multicollinearity of independent variables.…”

Uncovering the factors that affect earthquake insurance uptake using supervised machine learning

“…In addition, Abbas et al 12 found that in Pakistan, factors such as age, non-agricultural income, and preconceptions about the effectiveness of insurance can influence an individual’s willingness to pay for flood insurance, while Oral et al 13 discovered that in Turkey, earthquake preparedness is influenced not only by an individual's experience with earthquakes but also by cultural factors. Moreover, Greer et al 14 and Choi et al 15 argue that in the U.S., political ideology could significantly affect decision-makers’ perceptions of hazards, but its effects vary by the type of hazard.…”