Expected climate change consequences and their role in explaining individual risk judgments

France. The two coders in each country were later instructed to discuss whether they could resolve any of the coding disagreements. After solving disagreements to the extent possible, interrater reliability increased to 1.0 in the UK and 0. 999 in Norway, Germany, and France. The complete coding instructions are included in the supporting information, S1 File. Exam- ple responses can be seen in Table 1. Logistic regressions were used to investigate whether gender, age, and political orientation predicted the likelihood of mentioning specific categories while controlling for the country variable. We fit a total of six regression models, one for each category. In the next step, we used multiple linear regressions to explore the relationship between the content of the obtained responses (i.e., the six categories described above) and individual risk judgments (i.e., worry and risk evaluation), while controlling for the effect of several covariates (i.e., gender, age, political orientation, and country). The six categories used to code the open-ended responses were included as dichotomous variables (0 = category does not apply, 1 = category does apply). Additional multiple linear regressions tested possible effects concerning the structure of the obtained responses. An index for width was created by summarizing how many of the five causal levels from Bo¨hm and Pfister’s [11] framework were mentioned in each response (i.e., Level 1 + Level 2 + Level 3 + Level 4 + Level 5; see S2 Table for the numbering of the levels),

while excluding answers that mentioned either hardly any impacts or that did not relate to any of the categories (n = 134). Indexes for depth were calculated by looking at the level of specific- ity at the first sub-level (i.e., Level21 + Level22 + Level41 + Level42 + Level51 + Level52) and at the second sub-level (i.e., Level211 + Level212 + Level213 + Level221 + Level222 + Level223 + Level511 + Level521 + Level522 + Level523). Gender, age, political orientation, and country were included as covariates. The dataset was merged across countries for all regression analyses. All analyses were conducted in STATA 16.1 using the survey prefix command svy: to include survey weights in the regressions. Weights were added to ensure that the samples are representative of the adult population in each of the four countries. Using the svy: prefix gives coefficient p-values based on the Wald test. Effect sizes (McFadden’s Pseudo R2) for the logistic regressions were calculated from non-weighted versions of the regression models. Results Country differences in expected climate change consequences Most respondents (79% in the UK, 84% in Norway, 84% in Germany, and 88% in France) whose answers could be sorted into the six categories gave a response that reflected only one category. Table 1 shows that the most frequently mentioned climate change consequence Table 1. Frequencies for causal beliefs about climate change within the four countries (weighted). Categories Example responses UK Norway Germany France Total

Attitudes, motives, and goals (Level 1) «People will finally wake up» 14.95 % 0.49 % 1.28 % 1.24 % 4.68% Actions and activities (Level 2) «We will save more energy at home» 14.80 % 8.66 % 4.72 % 3.66 % 8.05% Emissions and pollution (Level 3) «CO2-emissions and CFCs» 12.92 % 6.17 % 2.50 % 7.92 % 7.53% Environmental changes (Level 4) «We hardly have seasons any more» 60.94 % 85.93 % 86.81 % 85.92 % 79.55% Impacts on humans (Level 5) «There will be severe consequences for us humans» 9.85 % 12.84 % 19.15 % 15.28 % 14.18% Hardly any impacts «Nothing will change» 11.86 % 2.93 % 1.63 % 0.28 % 4.29% Note. Each answer to the open-ended question could be sorted into several levels. Example responses are retrieved from the coding scheme available in the S2 Table. Missing responses were removed before calculating the percentages. https://doi.org/10.1371/journal.pone.0281258.t001 PLOS ONE Expected climate change consequences and individual risk judgments PLOS ONE | https://doi.org/10.1371/journal.pone.0281258 February 15, 2023 5 / 15