Trumpism, climate and COVID: Social bases of the new science rejection

Huber/White/sandwich estimator. Stars summarize probabilities from z tests employing these standard errors. As expected, conservatives, Republicans, whites and evangelicals are all significantly more likely to approve of Trump (hypothesis H1). Trump support also is more common among peo- ple receptive to seemingly unpolitical conspiracy beliefs—that NASA faked the Moon landings, or the Earth is really flat (H6). Finally, as detailed in section 5.2 on interactions, Trump support is higher among Republicans who say that most of their friends belong to the same party (H5). Trump support in turn predicts both rejection of ACC and rejection of COVID-19 vaccina- tion. Although the positive signs of these direct Trumpism effects are not surprising, their sig- nificance is noteworthy because the analysis already controls for other characteristics including ideology, party, race and religion, as well as education×party and friends×party inter- actions. Trumpism, as hypothesized, exhibits a direct effect on science rejection for both issues, net of the effects from other identity indicators (H2). Identification as white or evangelical, on the other hand, does not directly affect ACC or vaccine rejection—contrary to hypothesis H3. Although Trumpism increases the odds of both climate and vaccine rejection (H1), its effects on climate are stronger. Several cross-cutting relationships may account for this differ- ence. Vaccine rejection is greatest among younger respondents (opposite to climate rejection),

presumably because older people have a sharper sense of their own mortality, are more accus- tomed to following medical advice, and face higher risks from COVID-19. Vaccine rejection also is greater among women. Finally, lower household income and openness to conspiracy beliefs increase the odds of rejecting vaccination, whereas neither income nor conspiracism directly affect climate views (supporting only the first part of H6). Trumpism plays a significant role as intervening variable in our analysis, consistent with the conclusions of Shao and Hao [19]. This means that background factors might exert indi- rect (mediated) effects through Trumpism, in addition to whatever direct effects they have (H2a). Such indirect effects are approximated as products of coefficients along a sequence of paths from exogenous to final endogenous variables. Where the sequence includes one positive Table 2. Logit coefficients (with robust standard errors), from GSEM using ordered logit regression. Estimation sample n = 1,061; variable definitions and coding are given in Table 1. Endogenous variables Exogenous variables Trumpism NoACC NoVax Age –0.022 (0.004)*** 0.018 (0.004)*** –0.024 (0.006)*** Female –0.226 (0.133) 0.091 (0.163) 0.556 (0.184)** White 0.611 (0.169)*** –0.136 (0.204) –0.068 (0.226) Income 0.006 (0.062) 0.006 (0.070) –0.214 (0.214)** Evangelical 0.542 (0.160)*** 0.207 (0.190) 0.029 (0.212) Education –0.111 (0.077) –0.321 (0.094)*** –0.428 (0.113)*** Ideology 0.304 (0.047)*** 0.213 (0.055)*** 0.305 (0.062)***

Education × ideology 0.006 (0.035) 0.119 (0.042)** 0.111 (0.048)* Friends 0.420 (0.125)*** –0.251 (0.168) 0.075 (0.180) Party 0.990 (0.096)*** 0.035 (0.128) 0.052 (0.150) Friends × party 0.678 (0.134)*** 0.576 (0.175)** 0.579 (0.193)** Conspiracist 0.427 (0.064)*** 0.060 (0.076) 0.239 (0.078)** Trumpism 0.181 (0.042)*** 0.096 (0.049)* *p < 0.05 **p < 0.01 ***p < 0.001 (two-tailed z tests). https://doi.org/10.1371/journal.pone.0293059.t002 PLOS ONE Trumpism, climate and COVID PLOS ONE | https://doi.org/10.1371/journal.pone.0293059 January 10, 2024 8 / 20