Multilevel predictors of climate change beliefs in Africa

To improve the robustness of our analysis, we also include a measurement of drought for each administrative unit. We employ the 3-months standardized precipitation evapotranspira- tion index (SPEI-3) from the SPEI 2.6 database [58]. Since the SPEI is also presented as 0.5˚ x 0.5˚ gridded data, the data extraction procedure is identical to the other climate variables. The SPEI data are already presented as anomalies from a long-term baseline. We annualize these anomalies by taking the average value of the 12 previous months from the date the survey was conducted. In conclusion, we have four different climatic anomalies variables: precipitation, mean temperature, maximum temperature, and drought. All of them capture the anomaly of the previous year to the survey with respect to the long-term historical baseline. Results and discussion In this section we present and discuss the empirical results from our random forest analysis on what predicts climate change beliefs in Africa. Some common patterns emerge among the results from our five different models. We discuss these patterns by order of statistical rele- vance to each outcome variable. Fig 1 shows the main results for climate change awareness, belief anthropogenic climate change, and climate change-related risk perception. First, the importance of perceived worsening agricultural conditions stands out. Respon- dents’ who perceive worsening agricultural conditions show a greater awareness and perceived risk, more support for stopping climate change, and are more likely to believe in anthropo-

genic climate change. Indeed, Fig 1E shows that worsening perceptions of agricultural condi- tions are the main predictor for climate change risk perception. Arguably, the importance of the agriculture sector in terms of employment ―more than 50% of employees across Sub- Saharan Africa work in agriculture― makes agriculture a strong concern to African citizens [59]. Thus, perceiving that the impacts of climate change are bad for agriculture production may reduce the distance between climate change risks acuity and an amplified emotional response to climate-related events. These results suggest that for Africans climate change is not a problem for others across space and time, but rather a phenomenon that is happening here and now and with personal consequences [60]. This implies that policy makers should further highlight the global nature of climate change and its overall negative impacts on agriculture to raise CCBs, impulse individual adaptation, and mobilize public support. Second, attributing climate change to human activity is linked to higher risk perception, support for mitigation, and self-efficacy (Figs 1 and 2). If climate change is unnatural, it is extraordinary and thus riskier. Also, if climate change is human induced, its impacts can also be mitigated by human action. Besides, believing it is human caused can increase personal responsibility and, therefore, induce corrective responses [4, 61, 62]. This points to the conve- nience of spreading and highlighting the human origin of climate change to encourage behav-

ioral changes and mitigation strategies in Africa. Fig 2 shows that risk perception is positively associated with self-efficacy and the need to stop climate change. While some previous studies in the United States and United Kingdom point to fatalism or climate despair [2, 63, 64] ―where higher perceived risks discourage self- efficacy and action support, the opposite seems to be true for Africa. This could be the result of motivated control ―feeling more empowered in order to feel safe from a greater risk [65], or increased personal concern with climate change [66]. Either way, explaining why climate change is a critical risk for Africa will not discourage the African public, but it might encourage policy support and personal action [4]. As suggested by previous research, access to information and education are good predictors for believing in anthropogenic climate change and being aware of it. However, both variables show a slight negative effect when it comes to more affective dimensions such as risk PLOS ONE Multilevel predictors of climate change beliefs in Africa PLOS ONE | https://doi.org/10.1371/journal.pone.0266387 April 5, 2022 6 / 14