Non-transformative climate policy options decrease conservative support for renewable energy in the US

Motivated by ongoing partisan division in support of climate change policy, this paper investigates whether, among self-identifying liberals and conservatives, the mere presence of a non-transformative climate policy such as carbon capture and storage (CCS), lowers support for a renewable energy (RE) policy. To interrogate this question, we use a survey experiment asking 2374 respondents about their support for a RE policy when presented with the RE policy alone, and when presented alongside a CCS policy whose funding and implementation leverage independent funding sources. We find that among conservatives, the presence of a CCS policy lowers support for RE. Furthermore, despite the lack of apparent political party cues, when presented with the policy-pair, conservatives tend to view the RE policy in more partisan terms, specifically, less supported by Republicans. Additional experimental conditions with explicit party cues reinforce this interpretation. These findings suggest that the triggering of partisan perceptions even without explicit partisan cues—what we call political anchoring—might be a key impediment to bipartisan support of climate solutions in the U.S. context.


Introduction
Bolstered by broad support among the United States public [1], climate change legislation was a key agenda item for President Biden in 2021. Many of factors spurred on this initiative: over 21 billion dollars in damages from climate change related natural disasters [2], new urgent messages from the scientific community [3], and optimistic reports on the policy options for a transition to net-zero [4,5]. Despite this momentum, significant hurdles remain for sustained bipartisan action to promote a transition away from the use of fossil fuels [6,7], a necessary component of reaching climate goals by 2050 4 .
One such barrier to action is fossil fuel companies' marketing and lobbying activity. Historically, fossil fuel companies have worked to deny and cast doubt on the scientific basis of climate change to affect public opinion and provide cover to climate denialist politicians [8][9][10][11]. However, as outright denial has become increasingly untenable in the face of frequent climate disasters, economic realities, and legal liabilities [12,13], fossil fuel companies and their networks of supporting organizations have pivoted towards a politics of delay [14][15][16].
The goals of the politics of delay are twofold. Its first goal is to cast doubt on the efficacy of policy solutions to steer the economy away from fossil fuels. A second goal is to promote non-transformative solutions that allow for the continuation of current activities. For example, a 2021 report analyzing Facebook ads documented that fossil fuel organizations spent millions of dollars to convince young people that oil and natural gas were part of a 'pragmatic mix' of energy solutions to transition to a sustainable future [16,17]. Similarly, technologies like carbon capture and storage (CCS) have emerged with massive backing from fossil-fuel companies because they promise to permit industries reliant on fossil fuels to maintain current practices while reporting cuts in green-house gases [18]. However, studies on the effectiveness of CCS are sparse, with some indicating that current projects are under 50% effective and the increase in air pollution they would bring about may be even worse than doing nothing at all [19,20]. Additionally, the risks of committing to pathways reliant on unproven CCS technologies are skewed towards more vulnerable populations [21]. Therefore, understanding the effect that the public promotion of non-transformative, seemingly depoliticized solutions may have on public support for renewable energy (RE) is a pressing question.
Our work investigates whether in the United States media presentations of non-transformative solutions, a key element of delay strategies, has the potential to impact public support for RE. We randomly assign 2374 survey respondents to several experimental conditions and ask about their support for a RE policy. In the baseline condition, respondents see the RE policy alone. In the three main treatment conditions respondents see the same RE policy alongside a CCS policy that is signaled to be advanced independently in terms of its funding source. These three treatments vary in their inclusion of partisan cues. Unlike recent studies showing how combining policies into a single proposal may positively effect support [22,23], we design the experiment to signal to respondents that the RE and CCS policy are in no way bundled. Specifically, we signal to respondents that one's passing does not jeopardize that of the other.
We anticipate that despite this independence, the simple exposure to the CCS policy automatically activates a comparison, which we interrogate further. We capture evidence of a political anchor by measuring how perceptions of RE change when focusing on the comparison without explicit party cues. To foreshadow our results, we find that the CCS policy lowers support for RE compared to when conservative respondents see a RE policy alone. Our results suggest that political anchoring likely drives this change. Specifically, with the addition of the CCS policy, conservative respondents view RE as more partisan (i.e. supported by Democrats and not Republicans) compared to when they read the RE policy alone. Comparisons using additional experimental treatments that contain explicit party cues (for RE alone, or for the policy-pair) reinforce this interpretation.

Recruitment
We recruited respondents using the online survey platform Prolific 5 . Independent studies have found 5 www.prolific.co. that Prolific's panel of respondents are attentive and produce high-quality data [24,25]. From Prolific we recruited 1156 conservatives and 1218 liberals for a total of 2374 respondents. Prior to participation, Prolific collects demographic information, some of which we use in our analysis. Importantly, these prescreening procedures of Prolific allowed us to field our experiment to U.S. respondents who identify as liberal or conservative. Of the original 1173 conservative respondents, 17 reported that they were liberal, and only 2 liberal respondents of the original 1220 reported being conservative. We dropped the 19 respondents whose self-reports and Prolific classifications were misaligned. We did not use additional information provided through Prolific to create a quota sample by demographic characteristics or region. Sample summary statistics are available in the supplementary tables 2 and 3 with additional characteristics of the voting age population from the 2020 American National Election studies cumulative data file [26] and the Cooperative Election Study [27]. Our sample is younger, more likely to be female, and among liberals is more likely to identify as white. The survey was administered using the Qualtrics software. We restrict our sample to unique responses from individuals to ensure one exposure per respondent by capturing unique identifiers from Prolific and unique IP addresses by Qualtrics.

Experimental procedures and measures
We situate our study in the context of the United States for multiple reasons. Beyond practical considerations (Prolific covers the U.S. and U.K.), RE policy has been the subject of public debate and one of the centerpieces of the Biden agenda. Therefore we anticipated that U.S. respondents would be attuned to this issue. Additionally, we know campaigns in the U.S. to promote CCS are underway [28]. Finally, public polarization, especially affective polarization, is well documented [29] with less evidence for polarization over policy issues [30]. These observations make the U.S. a great candidate to answer our research questions and interrogate the mechanism of political anchoring.
We randomly assigned respondents to either a baseline condition or one of five experimental conditions. In the baseline condition, we present the RE policy alone without partisan cues. For the three main experimental conditions, we present the RE policy alongside a CCS policy that is presented as independent from the RE policy. As previously highlighted, independence means that one policy does not jeopardize the other from a budgetary perspective so that respondents do not think of these options as substitutes. Furthermore, these treatment conditions are differentiated by whether we show no party cues, party-cues in alignment with perceptions (RE supported by Democrats and CCS by Republicans), or party-cues that are misaligned (RE supported by Republicans and CCS by Democrats). In two final treatment conditions, the RE policy is presented alone but signaled as either a Democrat or Republican plan. Table 1 presents the distribution of respondents across all experimental conditions. Text from both policies was adapted from print media representations as well as an Exxon Mobil press release promoting CCS. While this design introduces added variation in terms of policy tools, we aimed to present realistic options that respondents could have been familiar with. This approach has the advantage of not deceiving participants, and leads to minimizing feelings of being lied to that could influence survey responses. The complete text introducing the RE and CCS policies can be found below. Party cues are included in italics, and treatments without party cues simply omit the text in rectangular brackets.

RE Policy (Democrat/Republican):
One policy measure (considered by Democrat/Republican lawmakers) expands tax credits and proposes new spending to transition the U.S. away from using oil, gas, and coal. The tax credits would incentivize investment in RE technologies such as wind and solar. In addition, individual purchasers of electric vehicles would benefit by receiving tax credits based on which vehicle parts are made in America. The rest of the bill would create programs to construct electric vehicle charging stations and implement updates to the electric grid to transmit wind and solar power better. Together, non-partisan researchers at the Rhodium Group project estimate that these and other policies focused on wind and solar could reduce U.S. CO 2 emissions by 737 million metric tons by 2030.

CCS Policy (Democrat/Republican):
(Democrat/Republican) [L]awmakers are also independently considering expanding tax credits for companies developing CCS technologies. CCS is the process of capturing CO 2 from industrial activity that would otherwise be released into the atmosphere and injecting it into deep underground geologic formations for safe, secure and permanent storage. CCS is one of the technologies that could enable heavy industry and manufacturing to limit their greenhouse gas emissions. Non-partisan researchers at the Rhodium Group project estimate that CCS could help industry reduce U.S. CO 2 emissions by 356 million metric tons by 2030.
Despite some variation, each policy description contains the same elements. First, the goal of the policy is presented, followed by a description of the measures included in the policy. Finally, a description of non-partisan research estimating the reduction of CO 2 emissions by 2030. The intent was to present these policies close to how respondents might encounter them as part of their everyday news consumption. When signaling efficacy we were intentionally vague while remaining truthful to minimize the impact of respondents' beliefs about efficacy (which we also measure and report). Respondents participate in the experiment first, answering questions about their support for and perceptions of the RE policy to avoid priming respondents with additional survey questions at the end of the survey.
After reading the policies, respondents answer comprehension check questions aimed at ensuring they understood the basic contours of the RE policy; when party cues were absent that one party over another was not described to support the RE policy; and when party cues were given, which party supported it. These measures help validate our assertion that respondents understood that no obvious partisan cues were present or that they recalled the partisan cues correctly when they were explicitly signaled. Respondents were given multiple attempts to revise their answers and were encouraged to read the text again if they failed. Only respondents who passed all of these checks could proceed with the survey.
Next we asked respondents about their support for the RE policy only. Our measure of support included a seven-point scale ranging from 'Strongly against' to 'Strongly in favor' with a 'Neutral' value at the center, which we analyze as a continuous measure. We also collect a second measure of support: respondents' willingness to pay for the RE policy, which we present as part of our robustness checks. Following these outcome measures we asked follow up questions aimed at gauging how respondents perceived the RE policy. Additional questions include environmental concern computed by averaging the responses to six questions from the General Social Survey covering different dimensions of environmental concern [31][32][33]. The full survey instrument is included in the supplementary information (SI).

Results
In November 2021, as congress was deliberating the 'Build Back Better' infrastructure bill containing over $550 billion in funding for clean energy and electrification, we fielded an omnibus, pre-registered 6 survey experiment to 2374 respondents, split roughly equally between conservatives and liberals. Figure 1 displays the distribution of support for the RE policy of liberal (panel A) and conservative (panel B) respondents separately presenting data on respondents in the first row who saw the RE policy alone, and in the second row who saw the policypair. Notably, regardless of which experimental condition respondents were exposed to, RE policy alone or the policy-pair, a majority of both liberals (98% or more) and conservatives (60% or more) were in favor of the RE policy that invested in RE infrastructure. However, the strength of support varied by ideology. Over 50% of liberal respondents were strongly in favor of the RE policy, with only a single liberal respondent expressing opposition. Conservatives on the other hand, were more measured: depending on whether they saw the RE policy alone or the policypair, 21% or 31% did not support it. These numbers mirror findings from nationally representative polls and reflect a broad support for RE [34].

Ideological differences in support of the RE policy
Looking at the difference between experimental conditions, we note that liberals did not change their viewpoints as a function of being presented with the policy-pair ( figure 1(A)). The average level of support 7 remained constant between conditions. Conservatives on the other hand, did display differences in support when exposed to the CCS policy alongside the RE policy ( figure 1(B)). Support declined from an average of 4.74 to 4.44 on a 7-point scale when exposed to the policy-pair; a marginally significant difference based on a pairwise t-test (p = 0.063). Considering the difference as a binary measure of support (5-7) or not (1-3, dropping the neutral category of 4), the observed shift represents a decline in support of the RE policy of nine percentage points when presented with a second, signaled to be financially unrelated, CCS policy.

The presence of a CCS policy affects conservative support for RE
Regression models shown in figure 2 substantiate the descriptive results of figure 1. Compared to respondents who saw only the RE policy, conservatives indicated less support when shown a second climate change policy signaled to be unbundled and complementary, rather than a substitute to the RE policy. With no additional control variables introduced, exposure to a CCS policy leads to a marginally statistically significant decline in support among conservatives (p = 0.062). This result is confirmed when we introduce environmental concern (p = 0.024) and additional controls (p = 0.042). On average, conservative support for a RE policy decreased by 0.30 points out of a 7-point scale-a 4% (0.30/7 = 4%) decline in support. Regression models also confirm regardless of the experimental condition, liberals showed no significant changes in their support for the RE policy. Because liberals already support the RE policy at very high levels, a ceiling effect limits our ability to detect whether including a CCS policy increases their support.
Together, the results suggest that conservatives exposed to the policy-pair lowered their support for RE. While the observed decline in support is not particularly large it results from a brief, politically neutral exposure. Therefore, when added to other media representations known to change support for climate policies, such as repeated exposures on social media [35], framing effects [36][37][38], elite cues [39,40], and outright climate change denial [9], we highlight a mechanism leading to an increase in political polarization of views over RE [34]. In sum, we are likely to document a lower bound for this effect and the impact of these aforementioned stimuli could be more pronounced. We turn next to understand why the addition of an independent CCS policy might produce this effect.

Perceived support by Republicans decreases the support of conservatives for RE
Prior work suggests two pathways for additional policy options driving the observed declines in support. First, although we present the RE and CCS policies as independent and therefore respondents are not asked to choose between the two, they could nevertheless infer the policies are options in a choice set and thus consider them against one another [41]. Thus the CCS policy could 'dilute' support for RE as respondents average, or spread, their support across their perceived options [42]. Second, the CCS policy may change perceptions about the RE policy, which could then cause the observed shift in support. For example, in a study measuring the perceived caloric content of different meals, Chernev found that the contrast between foods from different semantic  Multiple regression models of support for a RE policy across different experimental conditions comparing RE and RE + CCS. Each estimate with a 95% CI refers to the treatment effect in different models that are estimated separately for conservatives and liberals. Models differ in the additional covariates starting from no covariates (bottom), to adding environmental concern (middle), and additional demographic controls (top; income, age, sex and whether a person identifies as white or not). The reference group for all models is the RE policy alone. Complete regression tables can be found in supplementary tables 5 and 6.
categories-referred to as vice or virtue foodstended to make perceptions of their caloric intake more extreme [43]. They referred to this mechanism as a 'semantic anchor' because contrasting semantic categories provided a reference point for making judgments and, in this case, pushed judgments of adjacent foods away from one another. Given the documented wide variation between liberals and conservatives in support, perceived partisanship could take on such contrasting anchoring role, and presenting the policies side-by-side may activate this anchor.
That is, despite their presentation without party cues, respondents may perceive the CCS policy as more Republican and/or less Democrat, which in turn could shift their perception of the RE policy. Specifically, conservatives may be more sensitive to nonspecified political cues as they are more likely to have been exposed to previous messages supporting carbon capture [28]. Prior work on dilution effects would suggest a general decline of support across the political ideological spectrum, while anchoring would yield more polarization. Several measures of perceptions about the RE policy provide clues on whether dilution or anchoring drives polarization of support. Specifically, we measure perceptions in terms of efficacy, government involvement [37,44,45], speed of implementation [46,47], effects on jobs [22,36], and what percentage of Republican or Democrat respondents they think would support the policy [38,48,49] 8 . While we anticipated changes in support, especially among conservatives, it remains unexplored which of these factors that likely strongly correlate with support might be associated with the change.
To address this open question, we first establish the relationships between perceptions and support of the RE policy. Figure 3 displays the results of multiple regressions between support for the RE policy and our perceptual measures, estimated separately for liberals and conservatives in the baseline condition (i.e. when RE is presented alone). Mostly, a similar relationship exists between support and perceptions among conservatives and liberals. For example, conservatives and liberals who perceive the policy as more effective and creating more jobs show greater support. Alternatively, a perception that the policy would increase government involvement is related to 8 We thank an anonymous reviewer for pointing out that perceived beneficiaries of the RE policy may also be a driver behind support, and that these perceptions might also be altered by the experimental manipulation of including a policy-pair compared to the RE policy alone. Future work should investigate this avenue further. While we do not have a such a measure, we asked respondents whether they think the policy would help or hurt 'people like' themselves. We find that conservatives were more likely than liberals to expect the renewable energy policy to hurt people like themselves (t = 12, p < 0.001). a decrease in support. Furthermore, too much government involvement had a much stronger negative relationship to conservatives' support than to liberals' (z = 6.16, p < 0.001). Only the implementation speed had no statistical relationship with either group of respondents' support. However, conservatives and liberals differ in how their support associates with their perceptions of who else supports the RE policy. Perceptions of stronger Republican support among conservatives increase their support, whereas they decrease liberals' support. Similarly, perceptions of more Democrat support is related to more support among liberals but has no significant impact on conservatives.
Paying close attention to perceptions reported in figure 5 also reveals a large gap between observed and perceived support for RE. Despite a majority of conservative and liberal respondents saying they would support the RE policy (see figure 1), both assume that Republicans would generally not be in favor of it. We find that 73% of conservative respondents in the baseline condition thought that less than 40% of Republicans would support the RE policy. Similarly, 90% of liberal respondents thought less than 40% of Republicans would support the RE policy. Such low levels of perceived Republican support contrasts with the actual 69% support reported by conservatives respondents 9 . Conversely, the majority of both conservatives (73%) and liberals (76%) thought that more than 60% of Democrats would support 9 Our assumption that conservatives are likely to be Republicans and liberals are likely to be Democrats is bolstered by American National Election Studies data [26]. See SI table 1 for evidence of a strong correlation between party identification and political ideology. This correlation also exists in our data: 80% of conservatives identify as Republican, while 82% of liberals identify as Democrat. the RE policy. Both liberals and conservatives underestimate the observed level of support among both liberals (slightly), and conservatives (profoundly). As in other policy domains [50,51], these responses indicate that the RE policy was more popular than was commonly perceived and that the perception of polarization between ideological groups is more significant than it is. Now turning to changes in perceptions by experimental condition, conservatives significantly adjust their expectations of Republican support downward (t = 1.96, p < 0.05, see SI table 9) 10 . When presented with the RE policy alone, 27% of conservative respondents thought less than 20% of Republicans would support the policy. When presented with the CCS policy as well, however, 39% of conservatives thought that fewer than 20% of Republicans would support the policy-an increase of almost 50%. Conservative respondents who saw the policy-pair also significantly adjusted their perception of government involvement upward by an additional 8% (see SI figure 4(H) and table 9).
What do these results, taken together, suggest about dilution and anchoring as potential mechanisms? If dilution were operative here, a general adjustment in all perceptions would have been more likely. However, we find no evidence of this general change in perceptions, and in the case of liberals, we document no change in perceptions. Instead, the evidence suggests that the CCS policy led conservative respondents to view the RE policy in more partisan light. In turn, this perception likely acted as a semantic anchor that undermined their 10 We thank an anonymous Reviewer who asked us to produce these comparisons that were pre-registered as exploratory analyses. support. While we cannot establish a causal connection between this shift in perception and the decline in support for RE, this link is plausible and aligns with prior studies showing strong effects from minimal partisan cues [49].
To further evaluate whether conservative perceptions of partisanship might be anchoring their support for RE, we introduce two additional experimental conditions that add explicit political cues to the policy pair (see figure 4). In one new condition, we indicate that Democrats support the RE policy and Republicans the CCS policy (shown as 'Democrat RE + Republican CCS policy'). In the second new condition, we indicate that Republicans support the RE policy and Democrats the CCS Policy (shown as 'Republican RE + Democrat CCS policy'). We expect that when signals align with the perceptions described in figure 5, the effect will be similar to the policypair without partisan cues. Alternatively, when we present the policies with misaligned party cues, we expect to see a significant deviation in conservative support. We find evidence consistent with both of these expectations. When conservatives see policies that align with their perceptions of who supports the policies, support for the RE policy is indistinguishable from the presentation of the policies with no party cues. However, when cues are misaligned, conservative support is significantly greater than without cues (p = 0.03). Similarly, in treatments that include the RE policy alone with party cues we find strong evidence that conservatives have lower support for explicitly 'Democrat' policies, while liberals have less support for explicitly 'Republican' policies (see SI figure 7).
In addition to the results presented here, we conducted additional analyses that give us confidence in the robustness of our results. First, in addition to our assessment of support as a continuous variable, we also generated a dichotomous variable by classifying respondents as supportive (support of 5-7) or non-supportive (support of 1-3), dropping neutral respondents (support of 4). With these measures we conduct a series of logistic regressions (see SI tables 7-8). Our results for both conservatives (p = 0.024) and liberals are confirmed. Second, to evaluate how demographic bias in our sample might impact our results, we conducted our main analyses with survey weights targeting the demographic composition of conservatives and liberals separately in the nationallyrepresentative Cooperative Election Study, 2020 [27]. Again, our main results hold. We document stronger and statistically significant effect without control variables in our main effect on support among conservatives. Third, we replicate the main analyses using the willingness to pay as an alternative measure of support (SI section 'Replication with Willingness to Pay'). The results qualitatively align with those reported, albeit without statistical significance.
Going forward, research should seek to refine our findings to address some of the shortcomings of the present study. One limitation is that despite suggestive evidence of political anchoring, our design does not allow for conclusive identification of causal mechanisms driving our main effect. It is possible that the different policy tools or perceived beneficiaries may have played a role which we cannot address empirically in our study. Furthermore, we cannot rule out that the presentation of any second policy, even unrelated to climate would not elicit a similar result among conservatives. Moreover, without a comparative study taking place at another time it is challenging to assess how much the timing may have contributed to the effects we document. The fielding of the study took place during a national debate surrounding the Biden administration's 'Build Back Better' bill which may have amplified partisan divisions around climate policy. However, analysis in the supplementary tables 10-12 indicates low knowledge about the contents of the bill (with small differences in knowledge by ideology). As there are no differences across experimental conditions, we remain confident that variation in knowledge do no bias our results. In addition, future work may focus on different outcome measures that include behaviors (e.g. signing petitions). Last but not least, future research should take a more global perspective to analyze if these results hold outside of the politically polarized context of the U.S.

Discussion
In a survey fielded in November of 2021, we found that most liberal and conservative Americans support government policies to promote RE. However, among conservatives, this support was malleable. The introduction of a single, financially unrelated policy to promote CCS proved sufficient to produce a measurable decline in their support for RE. Furthermore, analysis of individual perceptions suggested that part of this decline was related to a shift in perceptions over partisan support for RE. That is, conservatives, when presented with the RE policy paired with a CCS policy thought that fewer Republicans would be in favor of the RE initiative. Our analysis of perceptions also revealed a significant gap between the observed and perceived support among Republicans and Democrats. Both conservatives and liberals thought that fewer Republicans would support RE than was observed; suggesting that policies to promote RE are more popular across the ideological spectrum than is commonly believed.
Nevertheless, evidence that a single exposure to the mere presence of a second policy proved enough to decrease support for RE has critical implications in the present moment. Outside of the lab, exposure to policies via traditional and social media is likely to be both more overtly politicized and repeated over time. While our results are only indirect evidence, they imply that a delaying strategy of promoting nontransformative climate change policies might effectively undermine public support for RE. This finding complements a growing literature investigating how the bundling of climate and other social policies in large policy frameworks such as the 'Green New Deal' may increase public support [22,23,52]. However, our work shows that presenting climate policies as decoupled from a funding perspective without bundling various goals, in our case RE and CCS, may undermine bipartisan support. Thus, even depoliticized media representations of multiple climate policy options could decrease support for RE by activating partisan perceptions, particularly among conservatives. As the timeline shortens for action to mitigate the worst effects of climate change [53], this work contributes to our understanding of the barriers to action.

Data availability statement
The data that support the findings of this study will be openly available following an embargo at the following URL/DOI: 10.17605/OSF.IO/8VRMD. Data will be available from 01 May 2023.