Country-level conditions like prosperity, democracy, and regulatory culture predict individual climate change belief

Decades after the scientific community reached a consensus on human-made climate change, a substantial portion of the world's population remains either unaware or unconvinced of its anthropogenic origins. This lack of consensus is not uniform, varying significantly both within and across nations. While previous research has explored climate change attitudes primarily in Western, English-speaking democracies or focused on specific countries, analyzing individual attributes, this study aims to address the gap in understanding the cross-national variation in climate change belief by examining a more comprehensive, globally diverse dataset. Existing research reveals inconsistencies in the relationship between individual characteristics and climate change beliefs across different nations. Factors such as political orientation, often strongly correlated with climate change denial in Western countries, show weaker or even reversed relationships in non-English speaking nations. Studies focusing on climate change concern, while exhibiting similar cross-national variation, have not directly investigated the specific influence of country-level conditions on individual beliefs. This study addresses these limitations by analyzing a large dataset to determine how both individual and country-level factors influence climate change belief across 143 countries, accounting for non-linear relationships and complex interactions using advanced machine learning techniques. Understanding these factors is crucial for developing effective communication strategies and policies aimed at fostering global climate action.

Existing research reveals inconsistencies in the relationship between individual characteristics and climate change beliefs across different nations. Studies have primarily focused on either Western, English-speaking democracies or specific countries, examining the influence of factors like political orientations, cultural values, and demographics. While political orientation strongly correlates with climate change denial in the US and similar nations, this relationship is less significant or reversed in many non-English speaking countries. Studies on climate change concern show similar cross-national variations in the influence of individual traits. Political orientations and gender consistently predict climate change concern in Western democracies, but not elsewhere. Education and perception of local temperature changes are also influential, but their predictive effects vary widely across countries. These inconsistencies suggest the strong moderating effect of national circumstances on climate change attitudes but have not explored the specific role of country-level conditions in shaping individual belief. A study examining country-level climate change belief found correlations with wealth, education levels, political leaning, and vulnerability to climate impacts. However, it lacked the capacity to disentangle the effects of individual traits from societal circumstances and assumed linear relationships, which may be overly simplistic given the heterogeneity observed in comparative analyses.

This study utilizes data from the Gallup World Poll (GWP) collected between 2007 and 2010, representing approximately 95% of the world's adult population. Climate change belief is operationalized as being aware of climate change and believing that it is primarily caused by human activity. Two model specifications are used: the first predicts the likelihood of both awareness and belief, while the second focuses solely on belief among those already aware of climate change. A random forest model, a machine-learning technique, is employed due to its ability to handle mixed-type predictors, non-linear relationships, and high-dimensional interactions without bias from non-parametric distributions or endogenous relationships. The model includes nine country-level variables identified in prior research as potential influencers: presence of environmental NGOs and climate scientists; economic development (GDP per capita); civil liberties; exposure to climate impacts; fossil fuel dependency (per capita emissions and carbon intensity per GDP); environmental protection; and market liberalism. Individual-level variables include age, gender, education, internet access, place of living (rural/urban), and self-reported air and water quality. Regional effects are controlled for. To enhance interpretability, several knowledge extraction techniques are used, including calculating variable importance, significance levels, local prediction importance across countries, and partial prediction effects for different variable values. In addition to the random forest model, a recursive partitioning tree model is used to visualize the relationships between key variables and climate change belief. Partial variable predictions are computed to determine the global prediction direction of each variable.

The random forest model reveals that all examined variables significantly predict climate change belief. However, the predictive importance of each variable varies. Country-level conditions are strong predictors. In both model specifications, environmental protection, civil liberties, and world region are highly influential. Individual education and internet access significantly predict climate change awareness but less so for belief in anthropogenic causes. Exposure to climate impacts, carbon intensity, and market liberalism are more strongly associated with belief than awareness. Non-linear relationships were observed for economic development, market liberalism, and carbon emissions. A bell-shaped curve was found, indicating that moderate levels of these factors are associated with the highest likelihood of climate change belief, while both very low and very high levels are associated with lower levels of belief. The importance of predictors varied significantly across countries. Environmental protection was the most important predictor in many countries, followed by education and world region. World region was particularly important for predicting belief among those already aware of climate change, followed by GDP per capita and market liberalism. Regional effects were notable, with respondents from East Asia and Latin America more likely to believe in human-caused climate change than those from Sub-Saharan Africa or former Soviet countries. Partial variable predictions confirmed that younger individuals, urban dwellers, those with higher education and internet access, and those reporting air or water pollution are more likely to believe in climate change. Country-level environmental protection, civil liberties, and exposure to climate impacts also positively predict belief, while higher carbon intensity predicts lower belief.

The findings highlight the significant role of country-level factors in shaping individual climate change beliefs. This study is the first to demonstrate the substantial correlation between civil liberties, the presence of domestic NGOs and scientists, and individual climate change belief. While individual education and internet access are important for raising awareness, country-level conditions are more influential in shaping belief in human-caused climate change. The non-linear relationships observed for economic development, market liberalism, and carbon emissions suggest more complex causal processes than previously assumed, possibly involving factors like mass media saturation, system justification bias, and the influence of climate denialist groups. The varying importance of predictors across countries underlines the need for tailored communication strategies, acknowledging the diverse contextual factors that shape belief. This research contributes to a deeper understanding of the factors driving global climate change belief, providing valuable insights for policymakers and communicators seeking to promote climate action.

This study demonstrates that country-level conditions, particularly environmental protection, civil liberties, and economic development, are strong predictors of individual climate change belief. Individual-level factors like education and internet access primarily influence awareness. Non-linear relationships were found for economic development, market liberalism, and carbon emissions. The significant variation in predictor importance across countries highlights the need for context-specific strategies. Future research should investigate the interactions between individual and country-level factors and conduct more in-depth case studies across different regions to further refine our understanding of climate change belief formation.

The cross-sectional nature of the GWP data limits causal inference. The reliance on self-reported data may introduce biases. The time lag between country-level data and individual responses could affect the accuracy of the model. The specific operationalization of climate change belief may influence the results. The study's scope focuses on belief, not necessarily behavioral intentions or actions.