Introduction

Given the size and global influence of the U.S., understanding the social bases for American citizens’ persistent skepticism toward climate science is of critical concern1,2,3,4. Beyond the central roles of partisan and ideological polarization, misinformation, and media echo chambers, religious characteristics are not only powerfully related to these factors, but may exercise their own unique influence on climate-change concern. Though numerous studies have identified links to particular religious identities (e.g., sectarian or evangelical Protestants5,6,7,8,9,10,11,12), practices (e.g., church attendance13,14), and broad theological orientations (e.g., belief in biblical literalism9,15,16), less clear is the role of specific theological beliefs regarding God, humanity, and the Earth. Might beliefs concerning God’s, versus humanity’s, relative influence upon the Earth’s environment serve to shape Americans’ views on climate change?

Though limited, extant research suggests theology plays a critical, if often ignored, role in shaping environmental attitudes. A number of studies, for example, have found “premillennial” or “dispensational” eschatological beliefs are robust predictors of Americans opposing government action to protect the environment or curb global warming17. These authors argue such “End-Times” theological beliefs might lead Americans to question the importance of preserving what will ultimately be destroyed in God’s due time. Related studies have shown that belief in a God who is either authoritative or benevolent in his involvement, rather than a mystical cosmic force, reduces support for environmental sustainability policies9,18, while others show belief in the “sacredness” of the environment diminishes beliefs in environmental risks such as pollution19. Yet recent scholarship has challenged the notion that belief in an involved deity with specific plans for the Earth necessarily leads to environmental apathy. Different studies by Veldman and Pogue, for example, instead argue more characteristically evangelical theological beliefs mark conservative political communities for whom climate skepticism has become a dominant norm20,21.

Another prominent line of research follows White’s thesis that “Judeo-Christian” influences have led Westerners to believe God has granted human “dominion” over the world to use as they please3,22,23,24,25. This is contrasted with a related Judeo-Christian idea of environmental “stewardship”26,27,28,29. Such beliefs have been shown to carry important implications. To the extent citizens believe the God of the Bible has granted humanity “dominion” over Earth, they exhibit less concern about their impact on the Earth. Alternatively, to the extent they believe God has appointed them as “stewards” responsible for Earth’s care, they show more concern30. On the one hand, studies that have explicitly measured the “dominion” concept find it negatively associated with environmental concerns or support for interventions, while experimentally priming a “stewardship frame” inclined Christians toward climate change belief29. Yet within contemporary political landscape, the climate attitudes of the most conservative Americans (e.g., evangelical Protestants) may be too rooted in partisan identities and norms for such theological beliefs to matter. Indeed, as early as 2012, Djupe and Burge found “dominion” beliefs only seemed to matter for non-evangelicals’ environmental attitudes22.

Importantly, the evidence for a religion-environment link is most often in the form of cross-sectional correlations with conservative Christian affiliation, literalist orientation to the Bible, or other indirect measures9,23,27,28. Given the deeply interrelated nature of partisan identities, cultural norms, and theological beliefs, we aim to answer the recent call by Veldman and co-authors for experimental work to better understand how religious messaging shapes citizens’ attitudes30.

It is worth stressing that beliefs involving “dominion” and “stewardship” are ultimately normative in nature. That is, they are largely moral beliefs about how human beings should behave in relation to the Earth’s environment. Indeed, measures of these concepts often include such moral language. For example, Sherkat and Ellison measure stewardship belief by assessing agreement that “Human beings should respect nature because it was created by God28, while a component of “dominion” belief in Shin and Preston’s work is “Humans have the moral right to use nature as we choose”29.

The normative quality of these beliefs stands in marked contrast to the focus of the present study, which is the degree to which citizens believe that—in fact—God, or humanity, ultimately controls what will happen to Earth’s environment. Further, both dominion and stewardship beliefs could plausibly entail belief in divine control of Earth or human control.

We therefore propose that this more fundamental theological belief—one concerning the reality of divine, as opposed to human, control of Earth’s climate—is likely to have a powerful influence on attitudes about the Earth’s climate. Djupe and Burge document that, at the bivariate level, the more Americans agree God is in control of affairs on Earth, the less likely they are to attribute climate change to human activity. Extending this idea, we reason that if Americans believe the future of Earth’s climate is effectively “in God’s hands,” they should be skeptical of science claiming that human behavior is leading Earth toward environmental catastrophe. Such a belief should, therefore, correspond with less concern about climate change as an issue as well as weaker support for climate-related legislation. As a U.S. member of Congress stated at a town hall event while discussing climate change, “…I believe that there is a creator in God who is much bigger than us. And I’m confident that, if there’s a real problem, he can take care of it”31 (see Supplementary Note 7 for additional examples).

It is also worth stressing that much of the aforementioned evidence linking religious belief and climate attitudes has been correlational in nature. Yet we believe it is important to investigate whether causal evidence exists. Particularly in the US case, religious identities and beliefs have become strongly correlated with partisan and ideological identities, e.g., via widespread socio-demographic sorting into the Democratic and Republican parties32,33,34,35. Because of this sorting, any association between religious beliefs and climate change attitudes, for example, could potentially be confounded by partisan/ideological identities36 and/or exposure to different media frames and elite cues on the issue of climate change37.

Our general expectation is that those who believe more in divine (as opposed to human) control of Earth’s climate will exhibit less concern about climate change and be less inclined to support policies aimed at combatting it. The following two sections test this expectation, first with nationally representative (observational) data and then with a pre-registered survey experiment. Our experiment, in particular, tests the following pre-registered hypotheses:

H1 (Threat Hypothesis): Respondents in the “God in control” condition (compared to those in the “Humans in control” condition) will exhibit significantly less concern about climate change.

H2 (Policy Hypothesis): Respondents in the “God in control” (versus “Humans in control”) condition will be significantly less supportive of policies (and politicians) aimed at combatting climate change.

Previous work also leads us to expect belief in God’s (versus humanity’s) control over Earth’s future could shape the extent to which people might seek out information about climate change. Studies have shown information-seeking behavior regarding climate change is driven in part by affective responses to risk or threat38. If persons are made to feel uneasy or vulnerable, they will more likely engage in behavior—such as information-gathering—to mitigate that feeling39. Conversely, persons who feel future events are comfortably under control should be less likely to seek out additional information. Thus we also formulated the following pre-registered hypothesis:

H3 (Information Hypothesis): Respondents in the “God in control” (versus “Humans in control”) condition, will be significantly less willing to request climate change information.

Results

Study 1: divine control and climate change attitudes in nationally representative data

Our first test of H1 and H2 uses 2023 data from the Public Religion Research Institute (PRRI). The survey features a nationally representative sample of 5,540 adults living within the U.S. Our key independent variable is agreement with the statement that, “God would not allow humans to destroy the Earth,” with responses ranging from “Completely disagree” (1) to “Completely agree” (4). This statement, which nearly 30% of the sample agreed with, implies at least a tacit endorsement of the notion that God, not humans, will ultimately decide Earth’s environmental future.

Our dependent variables measure several distinct constructs: citizens’ concern about climate change, belief in humanity’s role in causing it, and/or support for legislation designed to mitigate climate change. In the case of concern about climate change, we use both a 4-point scale ranging from belief that climate change is (1) “not a problem at all” to (4) a “crisis”, as well as a dichotomized version of this measure to determine the probability that a respondent specifically believes climate change to be a “crisis.” For beliefs about climate change’s anthropogenic causes (a key indicator of climate change literacy40), we use a dichotomous measure of whether respondents endorsed the notion that, “Climate change is caused mostly by human activity such as burning fossil fuels.” Lastly, we used support for six different climate-related government policies to construct an additive scale. The scale included items such as support for federal restrictions on CO2 emissions, increased federal funding for renewable energy, and tax breaks for Americans who use renewable energy in their homes. The items had strong pairwise correlations (r > 0.50, p < 0.001 in all cases) and a Cronbach’s α equal to 0.9.

Using OLS and logistic multiple regression, we also control for a variety of theoretically relevant, and potentially confounding, covariates. These covariates include standard demographic (race/ethnicity, age, education, gender, and income), political (party identification and ideology), and religious variables (religious identification and religiosity) (see Supplementary Note 1 and Supplementary Note 2 details). All variables, including our independent and dependent variables, were rescaled to range from 0 to 1 for interpretive ease.

Figure 1 below presents the results of our analyses. Across all four models, we observe a statistically significant decrease in each climate change outcome as citizens move from less to greater agreement that “God would not allow humans to destroy the Earth.” Specifically, moving across the entire 4-point scale of our independent variable, we observe a 6 percentage-point decline in the scale measuring concern about climate change (p < 0.001; see upper-left panel); a 7.4 percentage-point decrease in the probability of specifically identifying climate change as “a crisis” (p < 0.001; see upper-right panel); and, a nearly 10 percentage-point decrease in the probability of believing that humans are causing climate change (p < 0.001; see lower-left panel). Further, the lower-right panel demonstrates that belief in divine (versus human) control of the Earth also predicts significantly less support for legislation aimed at combatting climate change (p = 0.02), though the magnitude of this effect was substantively smaller in size (a 2.3 percentage-point decrease) compared to the other estimated effects. For example, the effect observed in the upper-left panel is larger than the effect of changing one full point on the seven-point party identification scale (see Supplemental Note 2 for details).

Fig. 1: Greater belief in divine (vs. human) control predicts lower climate concern, belief in human responsibility, and support for climate policies.
figure 1

Panels (a) and (d) are OLS; (b) and (c) are logistic regression. CIs are 95%. All outcomes have been scaled to range from 0 to 1. All models control for race/ethnicity, gender, educational attainment, household income, age, party identification, ideological self-placement, religious identification, and religiosity. Panel (a) shows that greater agreement God is in control of Earth’s climate predicts lower climate concern. Panel (b) shows that greater agreement God is in control of Earth’s climate predicts lower probability of identifying climate change as a major concern or crisis. Panel (c) shows that greater agreement God is in control of Earth’s climate predicts lower belief that climate change is mostly caused by humans. Panel (d) shows that greater agreement God is in control of Earth’s climate predicts lower score on the combined policy scale, which measures support for various policies aimed at combatting climate change. Data from PRRI (2023). N = 5300.

Full size image

We further tested the robustness of these findings by re-estimating the models controlling for “dominion” and “stewardship” beliefs. (These variables exhibited only modest correlations with our independent variable (r = 0.39 and 0.29, respectively).) The results remain substantively unchanged (see Supplementary Note 2 for details). Collectively, these results provide strong initial support for the logic underlying both H1 and H2. (Notably, the models indicate that, with all other variables at their means, the predicted values of each outcome are (albeit slightly) above their midpoint, regardless of belief in divine control of Earth.)

Study 2: an experimental test

To more directly test our hypotheses by isolating the effect of believing in “divine control,” we fielded a survey experiment via Lucid Theorem in February of 2024 (N = 3345). The survey featured nationally representative quotas for gender, race-ethnicity, age, and geographic region.

We randomly assigned 80% of respondents to read one of two (fictitious) news articles. The articles were purportedly from an “independent news site” so as to minimize the possibility that respondents infer a partisan bias. Each article discussed the conclusions of a group of religious scholars. In one version of the article, the scholars conclude that humans will ultimately decide the future of Earth’s environment and climate, while in the alternative version, the scholars conclude that God will decide. The verbatim text for each condition is featured in Table 1, with shaded (unshaded) text indicative of manipulated (static) content. The decision to reference religious “scholars” was informed by existing research demonstrating the influential role that experts play in communicating controversial ideas to the public41, while reference to “a variety of major religions” was done to help ensure that the belief would not be perceived as applying to only one religion. The experimental vignette (ostensibly) featured text from a recent news article.

Table 1 Vignettes Used in Survey Experiment
Full size table

We again employ several outcome measures. The first outcome (Concern) closely resembles the PRRI question referenced above, and ranges from belief that climate change is “Not a problem at all” (1) to “A crisis” (4). As in the PRRI analysis, we also created a binary version of this measure given its ordinal nature, with 1 indicating a belief that climate change specifically represents a “crisis” or “A major problem but not a crisis” and 0 indicating any other response.

The third outcome (Action) is adapted from a Cooperative Election Study (CES) question, with responses ranging from belief that “Global climate change is not occurring; this is not a real issue” (1) to “Global climate change has been established as a serious problem and immediate action is necessary” (5). We then asked two policy-related outcomes, which gauged support for federal funding of renewable energy research (Renewables) and federal limits on carbon dioxide emissions (Emissions). Our fifth outcome (Candidate) gauges respondents’ support for a hypothetical political candidate who promises to make “stopping climate change” a top policy priority. Our final outcome (Information) is a binary measure of whether respondents requested that we provide them with a link to climate change information from the National Oceanic and Atmospheric Administration (NOAA). All outcomes were rescaled to range from 0 to 1.

Before turning to the main results, we first demonstrate that our manipulation succeeded in significantly shaping beliefs about divine (vis-à-vis human) control of Earth. We employed two subjective manipulation checks42: one to gauge the degree to which respondents believe that “God ultimately controls what happens to the Earth’s climate,” and one—based upon the PRRI measure—to gauge the degree to which respondents believe that “God would not allow humans to have a big impact on the Earth’s climate.” Though conceptually related, the former places emphasis on God’s direct control over Earth, while the latter places emphasis on God’s control over humans’ impact on Earth. Both of these items featured 6-point agreement scales. The two items were positively correlated to a moderately strong degree (r = 0.60, p < 0.001). Given the distinction in emphasis, however, we opted to analyze the two items separately in Fig. 2.

Fig. 2: Experimental manipulation increases belief in divine control of earth’s climate.
figure 2

Panel (a) shows that the experimental manipulation significantly increases agreement that God ultimately has control over the Earth’s environmental future. Panel (b) shows that the experimental manipulation significantly increases agreement that God would not allow humans to destroy the Earth. Both manipulation check items used 6-point agreement scales ranging from “Strongly disagree” (1) to “Strongly agree” (6). Red lines on bars in both panels indicate 95% confidence intervals. Data = Lucid Theorem. N = 2685.

Full size image

The figure reveals that changing from the Humans in Control to God in Control condition significantly increased agreement that, (1) “God ultimately controls what happens to the Earth’s climate,” and, echoing our predictor variable from the PRRI analysis above, (2) “God would not allow humans to have a big impact on Earth’s climate.” We also find that treatment assignment significantly correlates with responses to a factual manipulation check item, confirming that our key manipulation was perceived by respondents (see Supplementary Note 4).

Figure 3 displays the key findings from our experiment. We find that our manipulation (i.e., the change from “Humans” to “God” being in control of Earth’s climate) significantly affected respondents’ beliefs about climate change and desire for more information. In particular, the Concern, Crisis and Action outcomes were significantly reduced by approximately 4, 5 and 3 percentage points, respectively (p < 0.01 in all cases). The manipulation also decreased respondents’ demand for climate information from NOAA by 3 percentage points (p < 0.05). For comparison purposes, these effect sizes were comparable to shifting one point to the political right (i.e., more conservative) on our seven-point ideology scale (see Supplementary Note 5 for details).

Fig. 3: Divine (vs. human) control reduces climate concern and demand for information.
figure 3

All outcomes (displayed in vertical text on the left side of each panel) recoded to range from 0 to 1. Panels (a) and (c) estimated via OLS regression; panels (b) and (d) estimated via logistic regression. Panel (a) shows that the experimental manipulation significantly reduced climate change concern. Panel (b) shows that the experimental manipulation significantly reduced the probability that a respondent would identify climate change as a major problem or crisis. Panel (c) shows that the experimental manipulation significantly reduced agreement that climate change is a major problem that requires “immediate action.” Panel (d) shows that the experimental manipulation significantly reduced the probability that a respondent would voluntarily request to be given climate-related information from the National Oceanic and Atmospheric Administration (NOAA) at the end of the survey. CIs are 90 and 95%. Data = Lucid Theorem. N = 2685.

Full size image

Each treatment effect shown in Fig. 3 was also re-estimated controlling for various demographic and political covariates. Substantively and statistically, the results are virtually identical to those reported in Fig. 3 (see Supplementary Note 5 for details). Further, using a (pre-treatment) measure of attentiveness43, we find these effects to be substantially stronger among those most likely have attended to the experiment (e.g., the effect on Information is estimated to be nearly 5 percentage points). This serves as important corroborating evidence that the treatment was indeed efficacious.

We did not find any statistically significant effects on the policy-related or Candidate outcomes, though effects were more consistent with H2 among those likely to have attended to the treatment (see Supplementary Note 5 for details). For example, among the most attentive, we estimate a −3.2 percentage-point effect on the Emissions outcome (p < 0.05), which is consistent with H2 (see Supplementary Note 5 for details). Nevertheless, these overall non-significant effects on policy/political outcomes echo the weaker policy-related effects found in our PRRI analyses above, as well as other experimental work demonstrating the challenge of experimentally changing climate-policy attitudes44,45.

Finally, per the pre-registration document (see Supplementary Note 8) we also explored heterogeneous treatment effects by analyzing whether religious and political variables significantly moderate the treatment effect. Because religious and political identities overlap so significantly, either theological beliefs or climate attitudes may be so fixed for certain communities that our manipulations really only mattered for theoretically unique subsets. For example, having stronger (versus weaker) ties to conservative Christianity could mean possessing already stronger beliefs about God’s control over Earth’s future, thereby reducing the potential effect of the treatment. Similarly, affiliating as a Republican (versus a Democrat) could mean more crystallized skepticism toward climate change (i.e., for politico-ideological reasons), which could also attenuate treatment effects. The specific moderators we examined were as follows: (1) Born-Again/Evangelical Christian versus not Born-Again/Evangelical Christian, (2) religiosity (an additive scale featuring measures from PRRI involving religious attendance and perceived place of religion in one’s life, as well as a third item measuring perceived importance of religion in one’s life (α = 0.88)), and (3) political party identification (Democrat or Republican).

We find mixed evidence of heterogeneous treatment effects. For example, treatment effects tend to be noticeably weaker for Born-Again/Evangelical Christians (compared to non-Born-Again/Evangelical Christians) and those higher in religiosity, though it is the reverse pattern for the information outcome and none of the interaction terms, themselves, was statistically significant at conventional levels. Importantly, the manipulation check results were also weaker for Born Again/Evangelicals compared to the rest of the sample. This pattern is consistent with experimental findings by Djupe and Burge, who found “dominion” and “stewardship” beliefs were only manipulated among non-evangelicals, perhaps suggesting evangelicals’ theological beliefs are more anchored and their environmental views, consequently, less open to change22. We further investigated whether the patterns in the PRRI data set hold for Born Again/Evangelicals. Here, we also find substantially weaker effects, though all were at least directionally consistent with H1 (see Supplementary Note 2 for details). Collectively, these results suggest that, while belief in divine control of Earth varies among Born Again/Evangelical Christians, it does not inform climate-related attitudes to the same degree as we observe for other religious groups.

For the analysis featuring party identification as a moderator, there was again no clear pattern that Republicans (compared to Democrats) exhibited smaller effects. See Supplementary Note 6 for further details on these analyses.

In sum, our experiment finds substantial empirical support for both H1 and H3. This constitutes novel evidence that beliefs about climate change’s threat, the need to address it, and the desire for scientific information, can be causally affected by beliefs about whether God or humanity is ultimately in control of Earth’s future climate.

Discussion

Our study contributes to the growing literature on citizens’ climate attitudes several important ways. First, we identify a way in which a specific religious belief—i.e., belief in divine (vis-à-vis human) control of the Earth’s environmental future—is consequential for citizens’ beliefs about climate change. We theorize that this belief, which is not restricted to one single religion or religious denomination, should logically shape the degree to which climate change is perceived as a threat: if God is not in control of Earth’s future, it would mean that humanity is relatively unprotected and, therefore, in greater potential danger. Second, we show with nationally representative survey data that this belief is consistently predictive of citizens’ beliefs and attitudes about climate change, even after adjusting for potentially confounding factors. Third, using a novel experiment, we go beyond existing correlational analyses to show that belief in divine, versus human, control over Earth’s environmental future is capable of causally affecting citizens’ beliefs about climate change’s importance, perceived need for addressing it, and desire for scientific information.

Our study therefore speaks to both the (1) degree, and (2) manner, that a specific religious belief can shape citizens’ attitudes toward climate change. Further, our theoretical approach potentially helps explain previous work documenting how “End Times” beliefs involving Earth’s ultimate destruction predict lower environmental concern17,46: At the heart of such beliefs might exist a more fundamental sense that God, not humanity, controls Earth’s future, thereby rendering it futile for humanity to change its behavior toward the natural environment.

Our findings come with important implications. Religious and political identities are strongly interrelated, which can render it empirically difficult to determine whether religious factors exert unique, causal effects on climate attitudes apart from partisan factors. Yet our study indicates fundamental religious beliefs are capable of cutting across partisan lines and shaping perceptions about climate change’s causes and severity, even if partisan factors exert greater influence overall. More broadly, to the extent that religious leaders can influence citizens’ beliefs, our experimental results imply that religious leaders may be able to increase citizens’ concern about climate change via stressing humanity’s critical role in shaping Earth’s environmental future, though perhaps somewhat less so for Born Again/Evangelical Christians specifically.

It is important to stress that our results speak to the importance of one particular religious belief, not religion writ large. Future studies would do well to examine how belief in divine control is communicated by religious leaders, both between and within particular religious denominations. Additionally, we believe these effects would likely generalize to other contexts (e.g., developing countries) given evidence linking beliefs about God’s will and perceived causes of anthropogenic climate change47.

Lastly, given the relatively weak effects on specific policy attitudes, future work could delve deeper into why manipulations that affect perceptions of climate change’s severity do not necessarily translate into changes in support for mitigation policies. For example, might citizens be unclear (and/or unconvinced) about how these policies can successfully combat climate change? Given our findings, and the rising threat of climate change, we believe this is remains a crucial next question to investigate.

Methods

PRRI analyses

The PRRI data set was fielded online from June 8 to June 28, 2023 (n = 5,540 total respondents). Details regarding the sample, methodology, and survey questions can be found at the following link: https://www.prri.org/wp-content/uploads/2023/09/Topline-Climate-Change-V2.pdf.

The item we employ as our independent variable (level of agreement with the notion that “God would not allow humans to destroy the Earth”) was asked amid several other questions that explicitly dealt with environmental attitudes, providing some assurance that the question was interpreted to mean environmental destruction, specifically. We provide descriptive analyses of this and all other outcome measures in Supplementary Note 1. The distribution of responses was as follows: Completely disagree (45%); Mostly disagree (27%); Mostly agree (17%); and Completely agree (10%).

As noted above, the first outcome measure (Concern) and second outcome measure (Crisis) are based upon the same survey question, the latter being a dichotomized version of the former. We reasoned that while the Concern measure likely functions (approximately) as a continuum of concern about climate change, some respondents may not have interpreted the scale as a continuum. The Crisis measure, therefore, relaxes the assumption that respondents interpreted the response options as a continuous scale, and instead treats all responses as indicative that climate change is (1) a crisis, or (0) not a crisis.

The item measuring belief in climate change’s anthropogenic causes featured two alternative response options. These two options were as follows: (1) climate change is due to “natural” causes, and (2) climate change is not happening. Both of these options were combined and coded as 0, while belief that climate change is caused by humans is coded as 1.

The policy scale outcome measure featured six items in total, though two items were asked of only half the sample. Therefore, for any given respondent, five items were used. Each item had four response options ranging from “Strongly Oppose” to “Strongly Favor” with <2% of respondents skipping or refusing to answer the items. The six items measured support for the following policies: (1) federal funding for renewable energy, (2) stricter limits on the amount of CO2 released by industries, (3) taxing companies that produce fossil fuels, (4) stricter limits on CO2 emissions from vehicles (asked of half the sample), (5) a program to phase out gas-powered cars and replace them with electric cars (asked of half the sample), and (6) tax breaks for individuals who use renewable energy for their home (see Supplementary Note 1 for verbatim wording). While items 4 and 5 had similar means and standard deviations (on the 4 pt. scale, m = 2.27 (sd = 1.02) and m = 2.61 (sd = 0.94)), we re-estimated the model featured in the bottom-right panel of Fig. 1 controlling for which of these two items a respondent answered. Both substantively and statistically, these results were virtually identical to the result shown in Fig. 1.

Survey experiment

The survey experiment was fielded online via Lucid Theorem between March 9-March 12, 2024, to 3,345 total respondents. The pre-registration document details for the experiment can be found in Supplementary Note 8. The sampling quotas in place were for the following variables: (1) Age (categories: 18–24; 25–34; 35–44; 45–64; 65–99), (2) Gender (categories: Male; Female), (3) Race/ethnicity (categories: White (Non-Hispanic); Black (Non-Hispanic); Hispanic; Other Ethnicity), and (4) Region (Midwest; West; Northeast; South). See Supplementary Note 3 for demographic statistics.

The survey obtained approval from New York University’s Institutional Review Board (IRB): (IRB-FY2024-8644). Informed consent was obtained from all participants (all 18 years of age or older) before they could continue the survey. Respondents who consented to participate were then asked for an attention check at the start of the survey. Those who successfully answered this question were permitted to continue the survey. A fee of $1.50 per complete response was paid by the authors to Lucid. All respondents were debriefed about the survey experiment at the conclusion of the survey.

The survey was programmed to maximize sample size in the two experimental conditions (“God in control” and “Humans in control”). However, 20% of the sample was randomly assigned to a “pure control” group that received no experimental vignette. This design choice enabled us to gain insight into each vignette’s relative contribution to the main effects reported here. See Supplementary Note 6 for details.

A randomization check was performed to ensure that the two main experimental conditions were balanced on key demographic, political, and religious covariates. Specifically, a logistic regression model tested whether assignment into the “God in control” (versus “Humans in control”) was significantly predicted by respondents’ gender identification, age, religious identification, family income, educational attainment, party identification, ideological self-placement, and religiosity (see coding details for the religiosity measure below). The overall model was not statistically significant at conventional levels (χ2 (25) p = 0.16).

Supplementary Note 4 features the exact wording of all outcome measures. The Renewables and Emissions measures were adapted from the PRRI 2023 survey. As noted above, respondents were asked whether they wanted to be provided, at the end of the survey, with a link to factual information about global climate change from the National Oceanic and Atmospheric Association (NOAA). If respondents indicated “Yes”, they were provided with the following link: https://www.climate.gov/news-features/understanding-climate/climate-change-global-temperature. The corresponding article (published in January of 2024 on Climate.gov) provides respondents with textual and visual information about rising global temperatures and predicted temperature increases for the future, as well as a list of references for additional information.

All reported p-values are based upon one-tailed hypothesis tests, given that our hypotheses were directional in nature.