Multilevel predictors of climate change beliefs in Africa

efficacy. Overall, Black Africans show more concern and self-efficacy than other ethnic groups, among whom Arab Africans are the less concerned. Therefore, we cannot talk about a “white male” effect [50] because White Africans do not especially neglect risks (in contrast to Arab Africans) nor women show more concern, rather the opposite. Finally, age and agricultural experience are insignificant across all models, contradicting previous findings within the liter- ature [41, 49]. Our study has three clear limitations. First, is the correlational nature of our data. Although our results offer relevant insights about the predictors of climate change beliefs in Africa, cau- sality cannot be inferred from the available cross-sectional data. Second, random forest meth- odology involves a certain degree of stochasticity, which can lead to some level of randomness and uncertainty in the outcome. Finally, some variables in our dataset have less observations given that some questions within the Afrobarometer were only asked as conditioned follow-up questions. Conclusion There is a lack of consensus within the literature for what predicts climate change beliefs and most studies conducted on this topic are usually limited to Western developed nations. Our results show several relevant findings for what predicts climate change beliefs across 33 African countries, using a well powered and high-dimensional dataset. First, actual changes in local cli- mate conditions are stronger predictors of climate change beliefs in Africa than access to infor-

mation, political ideology, or demographics across different model specifications. Similarly, worsening perceptions of climate conditions for agriculture is a strong predictor for all five dimensions for climate change beliefs. Second, our analysis also reveals that authoritarian and intolerant individuals show overall limited climate change beliefs. Third, not speaking French, English or Portuguese is linked to a worse understanding of climate change. Finally, there is a small but significant gender gap across different dimensions. Combined, our results suggest that more different mechanisms underly climate change beliefs than previously assumed [10]. We strongly believe that in-depth case-studies should further examine the complex causal mechanisms that shape climate change beliefs across different African countries to improve our understanding in order to implement the best mitigation and adaptation strategies for the region. Supporting information S1 File. Operationalization. Descriptive statistics, coding, and corresponding questions from the Afrobarometer of all variables. (DOCX) Acknowledgments We want to thank Stefano Nembrini for his useful comments on our methodology section. We would also like to thank the editors of PLoS ONE and the reviewers for their thoughtful com- ments that helped us strengthen this article. Author Contributions Conceptualization: Alfonso Sa´nchez. Data curation: Juan B. Gonza´lez. Funding acquisition: Alfonso Sa´nchez. PLOS ONE Multilevel predictors of climate change beliefs in Africa

PLOS ONE | https://doi.org/10.1371/journal.pone.0266387 April 5, 2022 10 / 14