Leveraging machine learning to understand opposition to environmental tax increases across countries and over time

Abstract

Taxes targeting fuel, road usage, or carbon emissions for environmental protection often face public opposition. Can widely accessible machine learning methods aid in predicting and understanding opposition to environmental taxes? This study uses the random forest algorithm to predict opposition to increased environmental taxes based on 41 theoretically relevant respondent characteristics. Drawing on nationally representative surveys, we predict individual tax opposition across 28 countries in 2010 and 2020 (N = 70 710). Personal values and environmental evaluations tend to be more influential than demographics in predicting tax opposition, with key variables differing between countries and over time. A lack of commitment to pro-environmental behavior is the most important predictor in emerging economies. Conversely, concerns about environmental issues and prioritization of jobs and prices are influential in high-income countries, gaining prominence over the previous decade. Policymakers can leverage these insights to tailor communication of environmental tax increases in different contexts, emphasizing, for instance, job creation.