Given the uneven nature of progress at the national level to combat climate change, local environmental initiatives have taken on increased importance. For example, while the federal government under the leadership of President Trump withdrew from the Paris Climate Agreement in 2017—and again just hours after he took the oath of office in 2025—over four hundred mayors subsequently committed to upholding its emissions targets. This “We Are Still In” campaign includes cities and states with a combined gross domestic product of over nine trillion dollars and representing over 150 million Americans. As Salt Lake City mayor Jackie Biskupski said, “[w]e must lead where the White House refuses to” (quoted in Daalder Reference Daalder2017). The outgoing Biden administration explicitly urged “states and local governments to stay on course regardless of President Trump’s plans to purposefully ignore climate change” (Calma Reference Calma2024). Environmental initiatives adopted by local governments in the United States are even more salient following the passage of Trump’s One Big Beautiful Bill Act, which rolled back many policies to combat climate change in the Biden-era Inflation Reduction Act. Thousands of cities and towns have also joined international initiatives—such as Local Governments for Sustainability—and developed their own climate action plans to reduce greenhouse gas emissions (Betsill and Bulkeley Reference Betsill and Bulkeley2004). Many of these plans set highly ambitious climate goals, such as carbon neutrality. Consequently, although the traditional assumption has been that local governments focus on nonpartisan issues such as roads rather than politicized ones such as climate change, this hypothesis is increasingly being challenged by scholars (e.g., Benedictis-Kessner and Warshaw Reference Benedictis-Kessner and Warshaw2020) and activists (Gaffney Reference Gaffney2024), and progress on climate change has actually been a more bottom-up phenomenon compared with other policy issues (Selin and VanDeveer Reference Selin, VanDeveer, Selin and VanDeveer2009).
What factors impact whether local policy makers in the US decide to support climate plans? Answering this question is crucial for the future of environmental initiatives. Policy maker preferences are intrinsically meaningful because they are the decision makers who directly choose whether to support or oppose local climate initiatives (Freire, Mignozzetti, and Skarbek Reference Freire, Mignozzetti and Skarbek2021). Even when there is broad support among the public for certain policies—such as universal background checks for gun purchases—a lack of policy maker support can prevent these policies from being enacted or from enduring in the long term. Although previous studies have analyzed the views of American and global publics on climate change policy,Footnote 1 there has been much less research on the views of local policy makers themselves despite the substantive significance of their preferences.
The small number of studies that do survey local policy makers on environmental issues are often focused on descriptive relationships, assessing, for example, whether local policy makers were already implementing (Bae and Feiock Reference Bae and Feiock2013) or supported implementing (Einstein, Glick, and Palmer Reference Einstein, Glick and Palmer2020; Gerber Reference Gerber2013; Lee and Stecula Reference Lee and Stecula2021) different types of climate and sustainability policies, how local policy makers collaborate with environmental groups (Portney and Berry Reference Portney and Berry2014), and how local policy makers can use governance networks to try to address climate issues (Aylett Reference Aylett2015; Krause Reference Krause2011). Other work infers policy maker preferences on climate policies based on indirect data such as press releases (Boussalis, Coan, and Holman Reference Boussalis, Coan and Holman2019) or supporting legislation (Bromley-Trujillo, Holman, and Sandoval Reference Bromley-Trujillo, Holman and Sandoval2019; Hopper and Swift Reference Hopper and Swift2022) rather than directly assessing preferences in a survey instrument. While such descriptive work is valuable, it cannot identify the causal impact climate plan attributes have on policy makers’ views about which types of climate plans they would prefer to adopt in their communities.
Beginning with the well-founded assumption that elected policy makers will prioritize reelection above other factors (e.g., Mayhew Reference Mayhew1974), and drawing from prior work on policy feedback (Campbell Reference Campbell2012), on the public’s views toward specific policies, and on policy makers’ decision making, we developed a set of preregistered hypotheses about how climate policy design would impact the attitudes of elected US local policy makers.Footnote 2 Although increasing polarization likely reduces the relative impact that substantive policy design has on attitudes (Hacker and Pierson Reference Hacker and Pierson2019), we expected that partisan signaling about which political party or parties a climate plan is associated with is not the only factor that matters (though we indeed expect it to be a highly influential factor). Instead, various salient design-based factors, such as the visibility of benefits and costs to constituents (e.g., Mettler Reference Mettler2011), should determine exactly how different policy designs impact attitudes. The question we analyze is therefore not whether policy design matters, but exactly how it matters.
To address this important and previously unaddressed question, we leverage a national sample of elected local policy makers that includes mayors, county executives, and council members from across the US (Lee and Stecula Reference Lee and Stecula2021; Lee et al. Reference Lee, Nyhan, Reifler and Flynn2021; Malhotra, Monin, and Tomz Reference Malhotra, Monin and Tomz2019). Specifically, we conducted a preregistered conjoint experiment on over five hundred local policy makers that randomly varied seven attributes of a climate plan. We utilize a conjoint design because it allows us to simultaneously vary many attributes at once, better representing the multiple trade-offs policy makers face in choosing between complex policy instruments in the real world.
Overall, the results of our experiment illuminate how policy design impacts local policy maker support for climate plans. Here, we highlight the subset of our findings that we believe are the most significant. Per our preregistered hypotheses, plans with more hidden costs (such as tax subsidies or regulations) are favored compared with plans that impose more visible costs (such as taxes and penalties), and including policies with visible benefits (such as economic compensation to defray the costs of climate policy) is popular. Contrary to our preregistered expectations, economic compensation does not increase support for bolder policy action, universal compensation does not increase support relative to targeted compensation, and there is an effect whereby plans endorsed only by the Democratic Party backfire and are less likely to be supported than policies endorsed by neither party.Footnote 3 This latter finding differs from prior work on the public, which either finds the backlash effect only holds among Republicans (Guisinger and Saunders Reference Guisinger and Saunders2017; Merkley and Stecula Reference Merkley and Stecula2021), or that climate plans endorsed by the Democratic Party are more popular—even among Republicans—than plans endorsed by neither party (Scheuch Reference Scheuch2024). The impact of partisan endorsements is, in general, substantial, which contributes to debates about elite polarization and is in line with recent findings that elites are actually more polarized than the public (e.g., Enders Reference Enders2021). Moreover, in a sign of the impact of polarization and the strength of partisan signaling, we find that partisan climate plans endorsed by only the Democratic Party actually reduce support among policy makers more than policies such as taxes and penalties that impose highly visible costs on constituents. More optimistically, the positive impact of substantive policies such as tax benefits is not statistically different from the positive impact of bipartisan endorsement.Footnote 4
As an extension, we also explore in the online appendix whether there are elite–public gaps in climate change policy preferences (e.g., Dellmuth et al. Reference Dellmuth, Scholte, Tallberg and Verhaegen2022; Hertel-Fernandez, Mildenberger, and Stokes Reference Hertel-Fernandez, Mildenberger and Stokes2019; Kertzer Reference Kertzer2022; Lee et al. Reference Lee, Nyhan, Reifler and Flynn2021). If policy makers and the public, on average, hold contradictory views on the optimal structure of climate plans, then that makes it more difficult to design environmental policies with broad political support among elites and the public. This, in turn, increases the chances that such efforts will fail. We field a replication on a nationally representative sample of the US public, finding support for our preregistered expectation that the public and policy makers hold largely compatible preferences.
Overall, our project makes several contributions. First, it provides one of the only causally identified tests of US local policy makers’ climate policy preferences. Second, it contributes to a range of literatures and debates, from support for policies with more visible versus hidden costs and benefits to the role of polarization and partisanship among elites. Third, our findings will inform activists about what policies they should advocate for to achieve their goals, the public about what types of climate change policies their local officials are willing to support, and policy makers themselves about the kinds of climate policies they can successfully pursue with their colleagues. While climate change is global, our study provides a framework for how the probability of policy adoption at the local level in the US can be increased by strategic design.
Theory: Climate Plan Design and Support
How will climate policy attributes impact the willingness of local policy makers in the US to adopt environmental plans? To address this question, our foundational assumption is that elected local policy makers will prioritize reelection and political considerations above other factors (e.g., Mayhew Reference Mayhew1974). From this starting point, we generate theoretical expectations about specific climate change policy attributes based on three sources. First, we utilize prior studies that have been conducted on the US public to assess the impact of climate policy design on attitudes. Since elected policy makers—even at the local level—have political incentives to align their views with public opinion, we expect that studies on the mass public can shed light on which policy designs policy makers will likely favor as well.
Second, we draw from the literature on policy feedback and public policy. The strain of this research that is most applicable to the current study assesses the effect of policy design on public attitudes toward that policy. Particularly relevant are findings that policies with more visible benefits and benefits that are distributed more widely are likely to garner greater public support, while policies with more visible costs and benefits that are concentrated on particular groups are less likely to be supported (Campbell Reference Campbell2012).Footnote 5 The problem, for example, with benefits that are not visible (e.g., the home mortgage interest deduction) is that they “leave much of the public, even beneficiaries themselves, unaware of government’s role,” which makes the public less likely to support the government program (Mettler Reference Mettler2011, 32). On the other hand, government policies with visible costs (e.g., taxes) leave the public all too aware of who to blame. Thus, even if there are no studies on public opinion toward the specific climate change policy attributes this paper focuses on, we are still able to generate hypotheses for policy makers by considering the likely public response to the policies based on the visibility and distribution of their costs and benefits.
Third, where available, we draw from studies that evaluate policy maker preferences and behavior directly via observational, descriptive survey, and experimental research.
There is, of course, nearly an infinite number of attributes associated with climate plans that could be studied. We focus on three different categories of attributes that we believe are particularly salient and important: (1) specific policies, (2) politically relevant endorsements and participants, and (3) structural characteristics. From these categories, we derive seven attributes of climate plans and theorize how they impact support for the plan. We choose ideal-type attributes that are broad enough to be relevant to local policy makers and members of the general public across the US in both big and small cities and towns. Brutger and colleagues (Reference Brutger, Kertzer, Renshon, Tingley and Weiss2023) establish that a more abstract experimental design better enables researchers to identify whether an effect exists, which is the primary goal of our study. Though increasing contextual detail and nuance (e.g., exactly how much taxes would be raised) in our theory and design is certainly an avenue for future research, Brutger and colleagues (Reference Brutger, Kertzer, Renshon, Tingley and Weiss2023) also show that doing so does not typically result in substantively different results.Footnote 6 This means our theory and findings for relatively general and abstract climate attributes are likely to be externally valid, though a promising avenue for future research would be to test this directly. The seven attributes and their different levels are summarized in table 1, and we now proceed to discuss each in more detail. We also include the full list of our preregistered hypotheses and our findings in the online appendix.
Climate Plan Attributes
Table 1 Long description
The table consists of two columns: Attribute and Levels.
* Type of property tax: Benefits for clean/efficient energy use. Penalties for exceeding a certain carbon budget. A general tax increase to fund clean/efficient energy projects.
* Higher energy efficiency standards for: Newly constructed government buildings. All new construction. All new construction and existing buildings.
* Economic relief: None. Provided to all constituents. Provided to constituents hurt by the policy.
* Party endorsement: No party endorsements. Democratic Party. Republican and Democratic Party.
* International participants: Neither cities in China nor N A T O countries. Cities in China but not in N A T O countries. Cities in N A T O countries but not in China.
* Policy begins in: Two years. Four years. Six years.
* Cost-benefit projection: Low short-term costs; low long-term benefits. High short-term costs; high long-term benefits.
Specific Policies
The first category of attributes we consider is substantive climate-related policies. Actual policies are at the heart of any climate plan and thus are essential to study.
Property Tax Instruments
We analyze support for environmentally relevant property taxes because tax instruments are one of the principal tools that have been implemented and advocated for as a solution to climate change (Mildenberger et al. Reference Mildenberger, Lachapelle, Harrison and Stadelmann-Steffen2022). However, disagreement remains among scholars, activists, and policy makers about what specific type of tax instrument should be pursued. Since both Pigovian taxes and subsidies can potentially bring about the socially optimal level of greenhouse gas emissions by either putting a “price” on carbon (via taxes or penalties) or rewarding actors for using clean sources of energy (via subsidies), the optimal policy to push for may be the one with greater support. The different levels of this attribute will enable us to test exactly this: whether local policy makers are more supportive of environmental taxes or subsidies.
Our preregistered hypothesis was that policy makers should prefer climate plans with subsidies compared with plans involving taxes or penalties. Prior work shows that public policies with more visible benefits and less visible costs are generally more popular among the mass public (Kahneman and Tversky Reference Kahneman and Tversky1979; Mettler Reference Mettler2005). Although subsidies delivered via the tax code (e.g., Affordable Care Act health insurance subsidies) theoretically provide clear benefits to members of the public in the form of lower taxes, their benefits are actually relatively hidden, especially compared with programs that provide cash benefits like Social Security (Mettler Reference Mettler2011). Nevertheless, tax subsidies provide one significant advantage over tax increases: the costs they impose on the public are much less visible. For example, any costs subsidies entail in terms of lowered government revenue or increased borrowing are more obscure and easily hidden by policy makers (Jagers and Hammar Reference Jagers and Hammar2009). This logic helps to explain why previous literature finds that the public generally prefers subsidies to taxes and specifically prefers environmental subsidies to environmental taxes (Cherry, Kallbekken, and Kroll Reference Cherry, Kallbekken and Kroll2012; Drews and Van den Bergh Reference Drews and van den Bergh2016; Jagers and Hammar Reference Jagers and Hammar2009). Given their political incentives to align with public opinion, we expect that policy makers will also be more likely to prefer subsidies over taxes and penalties. Instead of penalizing their constituents with taxes, they can reward them for good environmental behavior with subsidies. In accordance with this expectation, one observational study of California policy makers found increased support for energy reform—even among Republican politicians—when the legislation involved tax subsidies for clean energy use (Hess et al. Reference Hess, Coley, Mai and Hilliard2015).
We focus on property taxes specifically because this is the largest single source of tax revenue directly collected by local governments in the US (Tax Policy Center 2024). Many local governments also offer property tax subsidies for clean and efficient energy use and some are experimenting with taxes as well. For instance, nine counties in the Bay Area, California; Orange County in North Carolina; and Iowa City, Iowa, all increased property taxes in recent years specifically to address climate change. Some local governments are even implementing penalties for carbon use. For example, the city of Boulder in Colorado adopted a carbon tax on electricity use, and Aspen and Pitkin counties in Colorado charge homeowners with energy-intensive amenities (e.g., heated pools and snowmelt driveways) a fee if they exceed a certain carbon budget. Consequently, this attribute is realistic and reflects the types of policies local leaders could consider adopting.
Energy Efficiency Standards
We study energy efficiency standards because regulations are another foundational policy tool that governments can employ to combat climate change. Instead of incentivizing good behavior through the use of economic carrots (subsidies) or sticks (taxes or penalties), governments can adopt a command-and-control approach and impose more direct limitations to combat climate change. Energy efficiency standards in buildings are an especially significant factor to study given that residential energy usage alone accounts for about 20% of greenhouse gas emissions in the US (Goldstein, Gounaridis, and Newell Reference Goldstein, Gounaridis and Newell2020). But are policy makers willing to support bold regulations that can make a substantial difference in the fight against climate change?
Our preregistered hypothesis was that policy makers should generally prefer plans with less ambitious climate regulations. Given that the costs of energy efficiency regulations that target people’s own houses should be quite visible to the mass public relative to regulations that solely involve government buildings, we expect that policy makers on average—especially Republicans and independents—will have incentives to avoid imposing those costs on their constituents. Nevertheless, we also preregistered an expectation that the opposite would hold true among policy makers with a greater level of concern about climate change (e.g., Democrats). Though climate regulations entail clear costs, they also involve a relatively visible benefit: command-and-control regulations guarantee some improvement in terms of clean and efficient energy use by forcing actors to behave a certain way (Krosnick et al. Reference Krosnick, Holbrook, Lowe and Visser2006). Consequently, for members of the public that are concerned about climate change, the benefits of climate regulations may outweigh the costs, giving like-minded policy makers who rely on the support of those citizens electoral incentives to support regulations as well. Our expectations with respect to this attribute therefore incorporate insights about cost, visibility, and certainty of benefits.
Indeed, one descriptive survey of US state legislators and city council members found that Democrats and those more concerned about air quality were more likely to support policies aimed at reducing emissions from buildings (Dorsey-Palmateer et al. Reference Dorsey-Palmateer, Sheikh, Shewmake, Thompson and Wang2026). Another study evaluating the voting behavior of state legislators found lower Republican support for laws that aid in financing energy efficiency enhancements in buildings (Coley and Hess Reference Coley and Hess2012).
Energy efficiency standards are particularly relevant to local climate policy making in the US because local governments typically have control over building codes. This means that energy efficiency standards are a domain where local governments can have a significant impact, making the inclusion of this attribute in our study realistic (Bae and Feiock Reference Bae and Feiock2013; Gerber Reference Gerber2013).
Economic Relief
We also explore the impact of providing economic relief to constituents because climate advocates and scholars have posited that such relief can help to defray the short-term costs of environmental policies and thus increase support for climate action (Bergquist, Mildenberger, and Stokes Reference Bergquist, Mildenberger and Stokes2020; Gaikwad, Genovese, and Tingley Reference Gaikwad, Genovese and Tingley2022), as well as the perceived fairness of climate plans (e.g., Bechtel and Scheve Reference Bechtel and Scheve2013). Studying this attribute can also shed light on two major arguments in the policy feedback literature: that policies with visible benefits should garner greater support and that policies with universal benefits will be more popular than those with more targeted benefits.
We preregistered a hypothesis that local policy makers will generally be in favor of plans with economic compensation. Following work in the policy feedback literature, which shows that visible cash benefits from the government such as Social Security checks can increase public support for policies (Campbell Reference Campbell2003), we expect that local policy makers will support economic compensation to defray the costs of climate policies to constituents and, in turn, reduce the chances of political blowback. This logic explains why several studies have analyzed the impact of economic relief and found that members of the public are more likely to support climate policies when they include some element of economic relief (Bergquist, Mildenberger, and Stokes Reference Bergquist, Mildenberger and Stokes2020; Gaikwad, Genovese, and Tingley Reference Gaikwad, Genovese and Tingley2022). Similarly, one study of congressional voting behavior found that support for carbon restrictions is greater when carbon-intensive employment is high and unemployment benefits are more generous (Kono Reference Kono2020). This makes sense since unemployment benefits can help to defray the economic costs of climate action. Relatedly, we also preregistered an expectation that economic relief would make policy makers more willing to support bolder policies by reducing the salience of those policies’ costs and providing political cover to support more ambitious policies elsewhere in the plan.
Finally, we also preregistered a hypothesis that support for universal economic relief to all constituents would be greater among policy makers than more targeted economic relief to those hurt by the climate plan. This expectation is derived from arguments and findings that universal programs such as Social Security that distribute visible benefits to all have greater positive feedback effects than means-tested programs that are more targeted (Campbell Reference Campbell2003). On the other hand, there may be salient differences between the climate and the healthcare and pension domains that make this hypothesis less likely to hold. Most notably, all citizens are likely perceived as vulnerable to health issues and many are susceptible to financial challenges in retirement. Government benefits in these areas may therefore be conceived of as public goods. By contrast, a much narrower slice of the population—such as those in polluting industries like coal mining—may be viewed as vulnerable to climate policy (Gaikwad, Genovese, and Tingley Reference Gaikwad, Genovese and Tingley2022; Gazmararian and Tingley Reference Gazmararian and Tingley2023; Wilson Reference Wilson1995). As such, there may be greater support for targeted relief in the climate realm than there is for other policy domains.
Politically Relevant Endorsements and Participants
The second broad category we analyze concerns politically relevant endorsements and participants. An essential element of politics is coalition building, which makes it critical to understand the actors that must be won over to pass climate policies.
National Party Endorsements
We include this attribute in our study because despite the maxim that “all politics is local,” in recent years local politics has become increasingly nationalized (e.g., Hopkins Reference Hopkins2018), meaning that national party endorsements may be particularly salient factors affecting local policy maker support for a climate plan. National party endorsements—or the lack thereof—may also provide a useful signal to local policy makers about how supporting a climate plan will impact their prospects for reelection to their current office or election to higher office in the future. Following previous studies that find that the public is more likely to support climate plans that have bipartisan support (Bergquist, Mildenberger, and Stokes Reference Bergquist, Mildenberger and Stokes2020; Scheuch Reference Scheuch2024), we expect the same will hold among local policy makers. If both parties support a climate plan, then that makes it less likely that a local policy maker’s support for a climate plan can be used against them as a political weapon. Moreover, observational research on legislatures shows that bills with bipartisan endorsements are more likely to pass, which could be another reason why policy makers would be more likely to support them (Harbridge-Yong, Volden, and Wiseman Reference Harbridge-Yong, Volden and Wiseman2023).
We also preregistered a hypothesis that among Republican policy makers, plans with no party support will be more popular than plans with only Democratic Party support. Given increasing polarization, Republican policy makers may worry that supporting plans endorsed by only the Democratic Party will be unpopular among their political base. Some evidence for this dynamic exists. For example, in an observational and experimental analysis, Merkley and Stecula (Reference Merkley and Stecula2021) find that Democratic elite cues in favor of climate action increase climate skepticism among Republican members of the public due to polarizing party cues communicated through the news media. Guisinger and Saunders (Reference Guisinger and Saunders2017) find evidence of a similar dynamic in a wide-ranging analysis of elite cues on international relations issues. If this dynamic holds among local policy makers, then that would have important implications for how climate activists and local policy makers themselves should think about how to build coalitions for environmental action. On the other hand, one experimental study on the US public found that Democratic-endorsed plans are more popular—even among Republicans and independents—than plans endorsed by neither party (Scheuch Reference Scheuch2024). Thus, whether our hypothesis will hold for Republican policy makers is an open empirical question.
We do not include plans endorsed by Republicans only because of (1) power constraints in our design and (2) their unlikely external validity.
International Participants
We also test the impact of foreign cities participating in a climate plan, since climate change is a global problem and action by only the US will be insufficient to fully address the issue. Nevertheless, there is a debate in the literature about whether reciprocity by foreign countries significantly impacts support for domestic climate action. On the one hand, research by Tingley and Tomz (Reference Tingley and Tomz2014) finds that reciprocity is not a significant factor impacting public support for climate policy, as most members of the public believe environmental action at home should not depend on the level of action abroad. On the other hand, some studies find that greater international participation in climate initiatives increases public support for climate plans (e.g., Bechtel and Scheve Reference Bechtel and Scheve2013; Rowan Reference Rowan2025). One reason for this is that a lack of reciprocity can make people believe that foreign countries are taking advantage of their efforts. Following this logic, we preregistered an expectation that plans with international participation should increase policy maker support.
Our research design also allows us to examine whether participation by cities in in-group countries, such as NATO members (as a shorthand for America’s European allies), matters more or less than participation by cities in out-group countries, such as China. From one perspective, cues from in-group countries may send a stronger signal about a policy’s appropriateness and necessity than cues from out-group countries (e.g., Duque Reference Duque2018). Nevertheless, support for climate action may actually be higher when Chinese cities participate if policy makers fear environmental policies adopted by the US and NATO countries—but not China—will erode America’s relative economic competitiveness. We preregistered both of these possibilities given the competing logics.
Structural Characteristics
The final broad category we analyze involves structural elements of climate plans. These kinds of institutional design elements have been shown to be critical factors impacting support for and the effectiveness of policy regimes and institutions (Bechtel and Scheve Reference Bechtel and Scheve2013).
Time to Implementation
We first analyze the impact of different timelines to policy implementation. Given the urgent necessity of climate action to stave off catastrophic global warming, the willingness of policy makers to act rapidly as opposed to kicking the can down the road is critically important. However, the “time-inconsistency” problem associated with climate change, whereby investing in reducing greenhouse gas emissions today may not fully pay off for many generations, incentivizes the current generation to underinvest in climate action (e.g., Hovi, Sprinz, and Underdal Reference Hovi, Sprinz and Underdal2009). This is exacerbated by the fact that climate change is still often depicted as a problem with consequences that are distant in time (Brügger et al. Reference Brügger, Dessai, Devine-Wright, Morton and Pidgeon2015) and, most importantly, by the findings that the public (Jacobs and Matthews Reference Jacobs and Matthews2012) and policy makers (Sheffer et al. Reference Sheffer, Loewen, Soroka, Walgrave and Sheafer2018) discount the future. Empirical work similarly suggests that the public specifically prefers some level of delay with respect to implementing certain climate policies (Rinscheid, Pianta, and Weber Reference Rinscheid, Pianta and Weber2020), and policy makers have been shown to delay the implementation of painful austerity policies if elections are approaching (Strobl et al. Reference Strobl, Bäck, Müller and Angelova2021). Drawing on these insights from prior work, we preregistered an expectation that policy makers therefore have political incentives to eschew climate action in the short term, though we also hypothesized that policy makers with greater concern about climate change (e.g., Democrats) would be willing to implement policies sooner. The theoretical rationale for this latter expectation is that from an objective, scientific point of view, delaying action is quite harmful for climate change mitigation efforts, and this argument has been prominently and publicly made by bodies such as the Intergovernmental Panel on Climate Change (Rinscheid, Pianta, and Weber Reference Rinscheid, Pianta and Weber2020).
Cost-Benefit Analysis
The last attribute we include is an estimate of the plan’s short-term costs and long-term benefits. Unsurprisingly, a plethora of studies find that climate plans with higher costs are less likely to be supported by members of the public (Bechtel and Scheve Reference Bechtel and Scheve2013; Bergquist, Mildenberger, and Stokes Reference Bergquist, Mildenberger and Stokes2020; Bernauer and Gampfer Reference Bernauer and Gampfer2015). However, the key question we examine is not whether local policy makers are less likely to support plans with higher costs and more likely to support plans with greater benefits—undoubtedly they are—but whether they are willing or unwilling to trade off higher short-term costs for higher long-term benefits.Footnote 7 This is a germane question given that prior descriptive survey research on elected policy makers shows that higher short-run costs reduce their support for long-term solutions to problems (Sheffer, Loewen, and Lucas Reference Sheffer, Loewen and Lucas2024). An example of policies with low short-term costs and low long-term benefits would be adding public recycling bins or planting shade trees, and an example of an initiative with high short-term costs and high long-term benefits would be electric public buses. Previous studies often examine the benefits and costs of climate action in isolation, but they are not independent and thus it is essential to consider them in combination.
Given previous findings in the policy feedback literature on cost aversion among the public, we preregistered a hypothesis that policy makers would generally prefer plans with low short-term costs for fear of political backlash. Nevertheless, we also expected that Democratic policy makers would (1) be more likely to support plans with high short-term costs and high long-term benefits than Republicans, and (2) be more likely to support plans with high costs and benefits than those with low costs and benefits. This follows from the logic that Democrats’ constituents may value the benefits of combating climate change more than the costs they will endure to a greater extent than Republicans’ constituents.
Though this attribute is admittedly abstract, it contrasts different ideal-type plans and therefore forces policy makers to consider whether high long-term benefits are worth high short-term costs. Given the findings of Brutger and colleagues (Reference Brutger, Kertzer, Renshon, Tingley and Weiss2023), we expect that increasing contextual detail would not lead to substantively different results. Furthermore, to make this attribute relevant across our sample, some abstraction was necessary since what is “low cost” for a resident of Los Angeles may be very high cost for a resident of a more rural town. Importantly, our results generally are consistent across policy makers in cities, municipalities, and townships, which suggests that the political dynamics across different sizes and types of localities are similar.
Differences in Support across Climate Plan Attributes
While the preceding discussion theorized about how differences within each climate plan attribute could impact support among policy makers, our design also enables us to hypothesize about differences across different attributes. Though many comparisons are possible, we focus only on a subset that we believe are the most theoretically motivated. None of the hypotheses in this section were preregistered, and thus this discussion and the corresponding empirical analysis should be considered exploratory.
First, we expect that support for plans with climate regulations will generally be greater than those involving taxes or penalties since the latter involve more visible costs on members of the public. Furthermore, regulations provide clearer benefits than tax incentives when it comes to combating climate change since the former guarantee increased use of clean and efficient energy, while the latter only nudges people in that direction (Krosnick et al. Reference Krosnick, Holbrook, Lowe and Visser2006). This hypothesis thus ties into debates about how the degree of coercion associated with different policies impacts support, and, more generally, what types of cost-imposing policy instruments can garner the greatest support. For example, Drews and Van den Bergh (Reference Drews and van den Bergh2016) argue that people prefer less coercive climate policies because of lower perceived financial and behavioral costs. A descriptive survey of state and local policy makers also finds suggestive evidence for this hypothesis, as renewable energy mandates—a type of command-and-control regulation—received greater support than a carbon tax (Lee and Stecula Reference Lee and Stecula2021).
Second, following work by Ashok and Huber (Reference Ashok and Huber2020) and Faricy and Ellis (Reference Faricy and Ellis2014) that finds the public prefers indirect spending via the tax code rather than direct spending, we expect policy makers will prefer plans that include tax subsidies rather than direct economic compensation. This expectation follows from the assumption that the public prefers indirect spending, which involves less government administration, and that subsidies are more likely to be viewed as helping “deserving” and “hardworking” taxpayers rather than creating the kinds of “government dependency” that direct benefits may cause (Ashok and Huber Reference Ashok and Huber2020). This hypothesis therefore intervenes in debates about the optimal mechanism with which to deliver benefits to constituents (Campbell Reference Campbell2012).
Finally, given recent arguments in the policy feedback literature that increasing polarization in the US could reduce the relative impact that substantive policy design has on attitudes (Hacker and Pierson Reference Hacker and Pierson2019), we believe it is relevant to compare the impact of political endorsements to other substantive policy instruments. In particular, we will assess whether the expected positive impact of bipartisan endorsement is greater than the expected positive effect of subsidies or economic compensation on policy maker support. If it is not, then that would imply that substantive policy design can be just as impactful as partisan signaling. Though we hypothesized that policy design will matter even in the face of partisan endorsement signals, we do not have any a priori expectations about the relative magnitudes of these two different signals.
Data and Methods
Design
To analyze the determinants of local policy maker support for climate plans, we utilize a conjoint experimental design that randomly varies the level of each attribute outlined in table 1. More specifically, this design involves presenting respondents with two hypothetical climate plans side by side and asking them to choose which plan they prefer, as well as to rate each plan individually.
As with any hypothetical experimental study, there are legitimate questions as to whether the results will be externally valid to real-world scenarios. After all, the choice-based conjoint design does not exactly replicate the process elected officials go through to make policy, and other salient factors—such as lobbying by private industry groups—can impact support for climate policy in the real world (e.g., Stokes Reference Stokes2020). Even so, there is evidence to suggest that our design does have high levels of external validity, especially relative to alternative experimental designs. For example, many local governments have balanced budget requirements, which mandate the adoption of revenue-generating measures (such as taxes) to offset new spending—which is more accurately reflected by our bundled plan design than by assessing preferences individually across different plan attributes. Additionally, while some local governments have passed individual climate policies in separate pieces of legislation, there are real-world examples of more bundled climate plans similar to our conjoint. For instance, Washington DC’s Clean Energy Omnibus Amendment Act of 2018 (1) increased fees on sales of heating and fuel oil, as well as natural gas consumption; (2) imposed new climate-related regulations, including on housing; and (3) provided additional energy assistance to low-income residents. This mirrors our conjoint, which also involved increases in (1) taxes or penalties, (2) regulations, and (3) economic compensation.
Most relevantly, Hainmueller, Hangartner, and Yamamoto (Reference Hainmueller, Hangartner and Yamamoto2015) directly test whether the results from conjoint and vignette experiments correspond closely with real-world behavioral benchmarks. They find that paired conjoint designs—such as ours—match behavioral benchmarks quite closely and perform better than the more common single vignette design. Theoretically, this is likely the case because conjoints have several attractive properties. First, they simultaneously vary many attributes at once, better representing the multiple trade-offs policy makers face in choosing between complex policy instruments in the real world compared with more simplified vignette designs. For example, policy makers do not decide whether to support a climate package based exclusively on the particular policies embedded in it, but also based on factors such as who is endorsing the plan and when the policies will be implemented. Our design captures this real-world complexity. Second, though a conjoint experiment—like any survey experiment—measures stated rather than revealed preferences, prior work finds evidence that fully randomized conjoint designs can significantly reduce social desirability bias compared with other types of experimental designs (Horiuchi, Markovich, and Yamamoto Reference Horiuchi, Markovich and Yamamoto2022). One explanation for this result is that respondents are less likely to fear sanction for expressing their true views, since the complexity of conjoint designs enables them to rationalize their choices based on other nonsensitive attributes.
Despite the advantages of conjoint designs in terms of internal validity, realism, and diminishing social desirability bias, readers should note that experiments in general have limitations, especially when it comes to external validity. In real-world settings, policy makers—on both the political Right and Left—face a host of additional pressures not explicitly modeled in our experiment. For example, business interests often lobby and hold particular sway with Republicans, and labor unions are major players in Democratic politics (Mildenberger Reference Mildenberger2020). Constituents, the media, and party leaders also hold influence with both. This means that our results will not fully predict the public stances policy makers will take or how they will vote. Instead, our study, like all other survey experiments, measures underlying climate policy preferences and how the specific design factors we manipulate impact them. This is valuable, as it enables us to understand policy makers’ baseline attitudes, or starting points, before additional factors like lobbying impact their behavior (Stokes Reference Stokes2020). Further, it is also plausible that since policy makers are strategic actors, they have already, to some extent, anticipated these pressures and integrated them into their responses in the experiment. This dynamic, if present, would increase the external validity of our results. Yet even if it is not operative, the controlled experimental setting still enables us to isolate the impact of policy design and make credible causal inferences, whereas it is more challenging to do so with observational data precisely because of the multifaceted pressures policy makers face.
The context that respondents receive at the beginning of the experiment is that these are climate plans proposed by a nonpartisan international organization. This mirrors real-world international climate initiatives aimed at local governments and makes the inclusion of the international participants attribute more credible.
Sample
In partnership with CivicPulse, a nonprofit research organization, we leverage a sample of 573 elected local policy makers that includes mayors, county executives, and council members from across the US (Arias and Schwartz Reference Arias and Schwartz2026). This sample has been utilized in prominent research, including on environmental issues (Lee and Stecula Reference Lee and Stecula2021; Lee et al. Reference Lee, Nyhan, Reifler and Flynn2021; Malhotra, Monin, and Tomz Reference Malhotra, Monin and Tomz2019). CivicPulse recruited respondents in April and May 2021 by inviting via email a random sample of all US town, municipal, and county elected officials serving populations above a thousand. We show summary statistics for the policy maker sample in table 2.Footnote 8 We also utilize survey weights, provided by CivicPulse, in our main analyses to increase sample representativeness. These weights include the locality’s college education rate, population size, level of urbanization, and 2020 Democratic vote share. They are designed to increase the sample’s representativeness with respect to the full population of local governments in the US.Footnote 9 In total, respondents in our elite experiment completed over 2,100 paired-choice tasks.
Summary Statistics
Table 2 Long description
The table contains five columns: Variable, Minimum, Maximum, Mean (weighted), and Mean (unweighted).
* Partisanship: Minimum 1.00, Maximum 5.00, Weighted Mean 3.21, Unweighted Mean 3.28.
* Ideology: Minimum 1.00, Maximum 6.00, Weighted Mean 2.98, Unweighted Mean 3.14.
* Climate change belief: Minimum 1.00, Maximum 4.00, Weighted Mean 3.45, Unweighted Mean 3.50.
* Gender: Minimum 1.00, Maximum 2.00, Weighted Mean 1.51, Unweighted Mean 1.33.
* Age: Minimum 1.00, Maximum 15.00, Weighted Mean 8.74, Unweighted Mean 7.31.
* Education: Minimum 1.00, Maximum 7.00, Weighted Mean 4.25, Unweighted Mean 5.45.
* District Democratic 2020 presidential vote share: Minimum 1.00, Maximum 3.00, Weighted Mean 2.65, Unweighted Mean 2.63.
* District college proportion: Minimum 1.00, Maximum 3.00, Weighted Mean 2.09, Unweighted Mean 2.32.
* District urban proportion: Minimum 1.00, Maximum 3.00, Weighted Mean 2.11, Unweighted Mean 2.26.
* Government type: Minimum 1.00, Maximum 3.00, Weighted Mean 2.03, Unweighted Mean 2.08.
* Election ambition: Minimum 0.00, Maximum 3.00, Weighted Mean 0.60, Unweighted Mean 0.50.
* Local fossil fuel or auto industry: Minimum 1.00, Maximum 2.00, Weighted Mean 1.39, Unweighted Mean 1.37.
* Local green industry: Minimum 1.00, Maximum 2.00, Weighted Mean 1.30, Unweighted Mean 1.33.
* Extent that local community is impacted by climate change: Minimum 0.00, Maximum 5.00, Weighted Mean 1.48, Unweighted Mean 1.42.
* Extent that supporting climate change policy is politically harmful: Minimum 1.00, Maximum 5.00, Weighted Mean 3.03, Unweighted Mean 3.02.
Note: † Higher = stronger Democrat; ‡ higher = more liberal; § 4 = yes/mostly humans; ∥ 1 = male, 2 = female; # 7 = born 1956–60, 8 = born 1961–65; †† 4 = some college, 5 = college degree; ‡‡ 2 = second tercile; §§ 2 = second tercile; ∥∥ 2 = second tercile; ## 1 = county, 2 = municipal, 3 = township; ††† 0 = none, 1 = state, 2 = federal; ‡‡‡ 1 = no, 2 = yes; §§§ 1 = no, 2 = yes. For measures with more than two levels, the levels surrounding the mean are noted in the variable description.
Results
Following Leeper, Hobolt, and Tilley (Reference Leeper, Hobolt and Tilley2020), we estimate marginal means (MMs) for each attribute level.Footnote 10 MMs reflect the probability that a respondent will support a particular climate plan with a particular attribute level, averaging over the distribution of the remaining climate plan attributes. Values over 50% indicate that the attribute level increases the overall favorability of the plan, while values under 50% indicate it decreases favorability. While many studies employ average marginal component effects (AMCEs), MMs are more readily interpretable, and in subgroup analyses AMCEs can be misleading because they are sensitive to the choice of baseline category. Results are robust to AMCEs as the estimand, which are described in greater detail in the online appendix. Figure 1 displays the main results for all policy makers in our sample. In the online appendix, we include a summary table of the results for each preregistered hypothesis that we discussed above in the theoretical section.
Results for All Policy Makers
Note: Bars are 95% confidence intervals based on respondent-clustered standard errors for the MM of each attribute level.
Figure 1 Long description
A horizontal dot plot with error bars. The X-axis represents the Probability a respondent will select a climate plan with a particular attribute level, ranging from 30 percent to 60 percent with a dashed vertical baseline at 50 percent. The Y-axis lists seven categories with specific levels.
* Type of property tax. Benefits for clean energy is above 50 percent. General tax increase and Penalties for exceeding carbon budget are both below 50 percent.
* Higher energy efficiency standards for. Newly constructed government buildings is below 50 percent. All new construction and All new and existing buildings are both above 50 percent.
* Economic relief. None and Provided to all constituents are near 50 percent. Provided to constituents hurt by the plan is above 50 percent.
* Party endorsement. No party endorsements is above 50 percent. Democratic Party is the lowest point on the graph at approximately 37 percent. Republican and Democratic Party is the highest point at approximately 60 percent.
* International participants. Neither cities in China nor N A T O countries and Cities in China but not in N A T O countries are below 50 percent. Cities in N A T O countries but not in China is above 50 percent.
* Plan begins in. Two years is above 50 percent. Four years is at 50 percent. Six years is below 50 percent.
* Cost-benefit projection. High short-term costs with high long-term benefits and Low short-term costs with low long-term benefits are both below 50 percent.
We find that many different categories of climate policy attributes—from the specific policies included in them and in political endorsements to their structural characteristics—impact local policy maker support. At a high level, this indicates that the chances of local climate policy adoption can be increased by strategic design.
Property Taxes
In accordance with our preregistered expectations, we find that local policy makers prefer plans that include subsidies compared with those involving taxes or penalties.Footnote 11 Specifically, local policy makers are about nine to 11 percentage points less likely to support a climate plan that includes some kind of tax or penalty instead of benefits for clean or efficient energy use (p < 0.01). Furthermore, tax increases and penalties are not just less popular relative to subsidies, but also reduce support for the climate package in general. Although carbon and other related taxes have received much attention as a policy tool to address climate change, this result implies that it is wiser from a political perspective to advocate for subsidies that encourage clean energy usage. This is especially the case because taxes and subsidies are policy substitutes in that they can have similar substantive impacts on climate change. These findings also have broader implications for the policy feedback literature: they provide evidence that politicians prefer policies that avoid imposing visible costs on constituents.
Energy Efficiency Standards
Optimistically for the future of climate change efforts, and in contrast to our preregistered expectations, we find that local policy makers are likely to support climate plans with expansive energy efficiency standards. Compared with plans that set higher energy efficiency standards only for newly constructed government buildings, policy makers are about six to seven percentage points more likely to support plans with higher standards for all new construction or all new construction and existing buildings (p < 0.01). Given that the latter two policies would be significantly more invasive, disruptive, and potentially costly for constituents relative to higher standards only for government buildings, this result suggests at least some level of willingness on the part of local policy makers to impose costs on their constituents to make progress on combating climate change. Contrary to our expectations, we optimistically also do not find a significant difference between the two more expansive energy efficiency attribute levels. Since residential buildings in the US have a huge carbon footprint, this is an especially significant substantive finding. Future work could disentangle whether these findings are driven by cost, visibility, or the certainty of benefits that regulations provide.
Economic Relief
Per our expectations, local policy makers are generally supportive of plans with economic compensation to defray the costs of climate policies. For example, relative to no economic compensation, policy makers are over 7.5 percentage points more likely to support plans with relief to constituents harmed by climate policy (p < 0.01). Economic compensation may also be attractive for its potential to reduce the chances of political blowback. This finding accords with work in the policy feedback literature that finds that programs providing visible benefits to the public tend to garner greater support, and is consistent with existing work on climate policy making and the importance of using policy benefits to create climate coalitions, specifically when targeting affected groups (e.g., Bechtel, Genovese, and Scheve Reference Bechtel, Genovese and Scheve2017; Gaikwad, Genovese, and Tingley Reference Gaikwad, Genovese and Tingley2022).
In contrast to our theoretical expectations and findings in the policy feedback literature (e.g., Campbell Reference Campbell2003), we do not find any significant evidence that plans with universal economic relief are favored compared with those involving targeted relief. In fact, the opposite holds: narrow relief is favored by almost five percentage points compared with universal relief (p ≈ 0.01). There are several potential explanations for this somewhat surprising finding. First, policy makers may perceive narrow relief to those hurt most by the plan to be fairer than giving relief to all, especially since respondents may hold a mental model that the losers of climate policy are a relatively small group, reducing the need for universal relief. Second, this finding may be explained by a desire for local politicians to target rewards to key constituencies, such as their supporters or swing voters (e.g., Golden and Min Reference Golden and Min2013). Third, and related to the above two points, policy makers may believe that constituents who are economically hurt the most by a climate plan will have the greatest motivation to take political action against them. If this is the case, then concentrating relief to those harmed may be perceived as the optimal electoral strategy. Future work should follow up on this finding to explicitly test its mechanisms. Specifically, it could evaluate whether greater support for universal benefits in the healthcare and pension domains—in the form of Medicare and Social Security—is driven by differential perceptions of how large a population is vulnerable.
National Party Endorsements
In accordance with our theoretical expectations, support for a climate plan is 6.5 percentage points greater among policy makers when there is bipartisan endorsement of the plan than when neither party endorses it (p < 0.01). More interestingly, and in line with the hypothesis that an endorsement from only the Democratic Party can be counterproductive (Guisinger and Saunders Reference Guisinger and Saunders2017; Merkley and Stecula Reference Merkley and Stecula2021), support is significantly lower—over 16 percentage points—when only the Democratic Party endorses the plan compared with when neither party does (p < 0.01). Although we preregistered a hypothesis that this dynamic would hold among Republican respondents, we did not anticipate it holding among our full sample. Given that the MM for endorsement by only the Democratic Party is the largest in absolute value (relative to the 50% threshold) in our entire study, this suggests that supporters of climate action must take care to avoid the perception of overly partisan environmental plans. For example, local policy makers may do well to encourage national parties to avoid endorsing their climate plans unless bipartisan support can be gained.
International Participants
While previous studies have found that international participation in climate agreements has mixed effects (Bechtel and Scheve Reference Bechtel and Scheve2013; Rowan Reference Rowan2025; Tingley and Tomz Reference Tingley and Tomz2014), we find some evidence, in accordance with our preregistered hypothesis and the theory of conditional cooperation, that policy makers are more likely to support a climate plan with greater international participation compared with no international participation. This is likely due to beliefs about fairness and conceptions of climate change as a collective action problem. Specifically, we find that international participation from in-group actors (i.e., NATO) increases support by six percentage points relative to no international participation (p < 0.01), whereas we do not find similar evidence for participation by out-group actors (i.e., China) compared with no international participation. This suggests that transnational climate initiatives among friends might be more likely to succeed compared with initiatives that are solely domestic in nature. Nevertheless, in terms of the external validity of this finding, theories of distributive politics would expect that international participation will become less impactful when climate initiatives become salient to key domestic interest groups, which have strongly held preferences that may not be conditional on what foreign actors do (Aklin and Mildenberger Reference Aklin and Mildenberger2020; Colgan, Green, and Hale Reference Colgan, Green and Hale2021). Future work could do more to directly test this possibility.
Time to Implementation
Optimistically and in contrast to our preregistered expectations, policy makers generally support climate plans with shorter implementation times compared with those with longer ones, suggesting that concerns about the effects of climate change outweigh “time-inconsistency” problems. For example, support is almost 6.5 percentage points less among policy makers for plans that begin in six years instead of two years (p < 0.01), and climate plans that will not be implemented for six years actually have lower levels of support in general. Given the time pressure humanity is under to avoid the worst effects of climate change, these are promising findings overall.
Cost-Benefit Analysis
On average among our full sample of policy makers, we do not find any statistically significant differences between plans with high short-term costs and high long-term benefits compared with plans with low short-term costs and low long-term benefits. This optimistically implies, in contrast to our expectations, that climate plans with high short-term costs are not viewed less favorably than plans with low short-term costs, as long as the expectation is that they will also yield high long-term benefits.Footnote 12
Between Attribute Comparisons
We highlight several salient differences across climate plan attributes here. First, per the logic in the policy feedback literature that policies with less visible costs will garner greater support, we find that support for plans with climate regulations on all new construction or on new construction and existing buildings is generally greater than support for plans with taxes or penalties. For example, support for plans with higher standards on all new construction is over seven percentage points greater than for plans with tax increases (p < 0.01). Thus, advocating for command-and-control climate regulations may be more politically palatable than climate-related taxes or penalties.
Second, in accordance with work by Faricy and Ellis (Reference Faricy and Ellis2014) and Ashok and Huber (Reference Ashok and Huber2020), we find that policy makers prefer plans with indirect spending via the tax code rather than those with direct economic relief provided to all constituents by over seven percentage points (p < 0.01). However, there is no difference in support between subsidies and targeted economic relief. Overall, then, our work does not suggest that one particular policy mechanism for delivering visible benefits to the public holds a large advantage over another.
Finally, we assess the significance of partisan signaling relative to other, more substantive, policy design attributes. In a sign of the impact of polarization and the strength of partisan signaling, we find that partisan climate plans endorsed by only the Democratic Party actually reduce support among policy makers 8.5 to 11 percentage points more than policies such as taxes and penalties that impose highly visible costs on constituents (p < 0.01). This provides suggestive evidence for Hacker and Pierson’s (Reference Hacker and Pierson2019) argument that increasing polarization may be reducing the relative impact of policy design. On the other hand, the positive impact of substantive policies such as tax benefits is not statistically different from the positive impact of bipartisan endorsement. This implies that while the effect of partisan signaling is large in our current environment, substantive design features can also have effects that are just as large.
Subgroup Analysis: Party Identification
In the online appendix, we present subgroup analyses based on a range of factors, including belief in climate change, political ambition, perceived effect of climate action on election prospects or local economic conditions, presence or lack thereof of carbon/green industries locally, government type (county, municipality, or township), local Democratic support, urbanness, and other demographic factors such as education and gender. Our results generally hold across these different factors. We do find that support for plans is higher overall in pro-climate (urban and Democratic) areas, but that such differences are modest. This may suggest that such localities could be the most promising venues for climate advocates to find partners for policy making. Here, we focus on a particularly salient respondent-level characteristic: political identification. Figure 2 plots the difference in MMs for Democrats and Republicans. The individual MMs for Democrats and Republicans used to calculate these differences are available in the online appendix.
Differences between Republican and Democratic Policy Makers
Note: Bars are 95% confidence intervals based on respondent-clustered standard errors for differences in MMs between Democratic and Republican local policy makers.
Figure 2 Long description
A horizontal dot plot with error bars. The X axis ranges from minus 20 to 20. Values below zero are labeled More preferred by Republicans and values above zero are labeled More preferred by Democrats. A red dashed vertical line marks the zero baseline. The Y axis lists seven categories with specific sub-options.
* Type of property tax. Benefits for clean energy, general tax increase, and carbon budget penalties all cluster near zero.
* Higher energy efficiency standards. Newly constructed government buildings leans Republican at minus 8. All new construction is at zero. All new and existing buildings leans Democratic at 8.
* Economic relief. None, provided to all, and provided to those hurt all cluster near zero.
* Party endorsement. No party endorsement leans Republican at minus 8. Democratic Party endorsement leans strongly Democratic at 15. Republican and Democratic Party endorsement leans Republican at minus 5.
* International participants. Neither China nor N A T O is at zero. Cities in China leans Republican at minus 4. Cities in N A T O countries leans Democratic at 4.
* Plan begins in. Two years leans Democratic at 9. Four years leans Republican at minus 4. Six years leans Republican at minus 7.
* Cost-benefit projection. High short-term costs with high long-term benefits leans Democratic at 9. Low short-term costs with low long-term benefits leans strongly Republican at minus 15.
Starting with our property tax attribute, we find no significant differences between Democratic and Republican local policy makers. Plans with subsidies are favored over those with taxes on a bipartisan basis, which only strengthens the political rationale for climate advocates to support the former over the latter. Similarly, no significant differences emerge between Democratic and Republican policy makers for the economic relief attribute. This suggests that economic relief in the context of climate plans has bipartisan support and thus could be a useful tool to increase local policy maker backing for environmental action.
We find larger differences in support for plans with higher energy efficiency standards by political identification. In accordance with our preregistered expectations, Democratic policy makers are eight percentage points more likely to support plans with higher standards on all new construction and existing buildings, while Republicans are seven percentage points more likely to support plans with higher standards on government buildings only (p < 0.01). This implies that it may be harder to gain Republican policy maker support for the most stringent energy efficiency standards. Nevertheless, we do not find a significant difference by party identification for higher standards on all new construction, the middle-ground policy. This suggests a political compromise is possible between the most narrow and most expansive energy efficiency standards.
For partisan endorsements, the subgroup analysis reveals substantial heterogeneity that is masked in a top-line examination. In line with our preregistered expectations, Republican policy makers are 14.5 percentage points less likely than their Democratic counterparts to support climate plans when they are endorsed by only the Democratic Party (p < 0.01). However, note that even Democratic policy makers are generally less supportive of plans with only Democratic Party support compared with no endorsements (i.e., the MM for Democratic policy makers is significantly below 50%). This is likely because Democrats do not want to be framed as overly partisan or extreme actors. By comparison, policies with bipartisan support are relatively popular among policy makers of both parties (i.e., the MM is significantly above 50% for Democrats and Republicans), though somewhat more so among Republicans (p < 0.07). These findings differ in several notable ways from prior work on the public. First, some past studies only find evidence of a backlash effect for Democratic-endorsed plans among Republicans (Guisinger and Saunders Reference Guisinger and Saunders2017; Merkley and Stecula Reference Merkley and Stecula2021), whereas we find the backlash holds among Democratic respondents too. Second, some previous work on the US public finds that plans endorsed by neither party are less popular than plans endorsed by only the Democratic Party (Scheuch Reference Scheuch2024), whereas local policy makers in our sample prefer plans endorsed by neither party. Third, we find evidence that Democratic-endorsed plans reduce overall support for the climate plan (i.e., the MM is under 50%), whereas some prior research on the public finds the opposite (Scheuch Reference Scheuch2024). On balance, our findings strongly suggest that local policy makers may do well to encourage national parties to avoid endorsing their climate plans unless bipartisan support can be gained, and that this imperative is stronger than studies on the public indicate.
We also uncover striking differences between Democratic and Republican policy makers with respect to the time-to-implementation attribute. As we hypothesized, Democratic officials are 6.7 percentage points less likely than Republican officials to support plans that begin in six years and almost 11 percentage points more likely to support plans that begin in two years (p < 0.01). Although Republican policy makers generally prefer longer implementation times, their preferences are less strongly held than those of Democrats, and there are no significant differences between Republican and Democratic policy makers for plans that begin in four years. Once again, this suggests that political compromise may be possible, though the urgent necessity of climate action might mean that climate advocates should still push for the shortest implementation times that are logistically feasible.
Finally, as we anticipated in our preanalysis plan, while Democratic policy makers are nine percentage points more likely to support plans with high short-term costs if these costs are associated with high long-term benefits, Republican policy makers are over 15 percentage points more likely to support plans that eschew long-term benefits to keep the short-term costs down (p < 0.01). This illustrates a fundamental disconnect between Democratic and Republican elites with respect to their outlooks on the severity of climate change.
In sum, these heterogeneous effects demonstrate that it is important to disaggregate preferences by salient individual-level characteristics—such as party identification—to have an accurate understanding of climate attitudes. Even so, there remains a surprising degree of agreement between Republicans and Democrats on many climate plan design features, and when partisan differences do exist, we observe that there is often space for compromise on middle-ground policies.
Robustness and Extensions
To verify the robustness of our core results and further validate our design, we take a number of steps, which we describe in detail in the online appendix due to space considerations. Across all the robustness tests, we find support for our core empirical findings.
First, we demonstrate that our results are substantively identical when sampling weights are omitted or when we utilize AMCEs rather than MMs as the estimand. Second, we show that our findings are likely not the product of false positives due to the high number of hypothesis tests we conduct. For example, in total, 75% (21 of 28) of the AMCEs we analyze in our elite study are significantly different from zero at the 5% level, even after using the Benjamini-Hochberg procedure to account for the possibility that some of our significant results are false positives (Kertzer Reference Kertzer2022).
Third, we illustrate that our results are substantively similar when we compare attribute levels using our rating outcome measure—which asked respondents to rate each plan individually on a five-point scale—rather than the traditional forced-choice outcome measure we utilize in our main models. While the binary measure used in our main models enables us to analyze how the manipulated factors impact relative levels of support for climate plans, the rating measure illustrates how the experimental factors affect absolute levels of support for climate plans (Bechtel and Scheve Reference Bechtel and Scheve2013, 13766). In other words, the former measure can tell us how different factors impact the probability of one plan being chosen over another, whereas the latter speaks to how various attributes affect whether policy makers would support taking climate action at all. This is significant because rather than choosing between two different climate plans, some policy makers may (1) oppose any climate policy or (2) prefer to build their own policy and thus oppose the particular policies we presented them with. Our ratings measure can capture these dynamics since respondents can register low support for both climate plans presented to them.
We also create a dichotomous version of our ratings measure that equals one if respondents “strongly support” or “somewhat support” a climate plan and zero otherwise. This variable captures the extent to which different levels of the attributes we analyze can move respondents from not supporting a climate plan to supporting a climate plan. About 46% of climate plans presented to respondents are either strongly or somewhat supported, meaning it is not the case that an overwhelming proportion of plans are opposed no matter the particular attributes associated with them. The robustness of our results across these different types of dependent-variable measures suggests we are capturing something real about policy makers’ preferences and our findings are not simply an artifact of the forced-choice nature of typical conjoint designs.
Fourth, climate plans endorsed by only the Democratic Party may be more likely in the real world than those with no party endorsements or bipartisan endorsement. To further evaluate the robustness of our findings, we subset our data to climate plans endorsed by only the Democratic Party. Almost all of the core results hold for this test, despite the decreased statistical power. For example, benefits for clean energy use and building regulations are still strongly supported over taxes and penalties, energy efficiency standards for only newly constructed government buildings receive the lowest level of support, international participation increases support for plans, and there is no significant difference between plans with low short-term costs and low long-term benefits compared with those with high short-term costs and high long-term benefits. We do find that support for economic compensation weakens, perhaps because Democrats are perceived as more dispositionally supportive of increasing spending, making this policy proposal less credible when it is endorsed only by them.
In the online appendix we examine which attributes policy makers said were most important in making their decision about whether to support the climate plan or not. Most interestingly, we uncover heterogeneity between Democratic and Republican policy makers. About 20% of Democratic local policy makers said the time-to-implementation attribute was the most important factor in making their choice of plans, compared with under 3% of Republicans. A review of the open-ended responses suggests policy makers who believe this is the most important attribute perceive climate change as a time-sensitive challenge that requires urgent action. For instance, one policy maker said, “The clock is ticking,” another argued, “Timing is critical!! Action must be taken quickly,” and a third insisted, “Really important to move quickly to avoid cataclysmic [destruction].” The disparity we observe between Democrats and Republicans on this question likely reflects a fundamental difference in belief about how urgent the climate crisis is.
We also extend our results in two different directions in the online appendix. First, we test for interaction effects, estimating the average marginal interaction effect (Egami and Imai Reference Egami and Imai2019) to assess whether political cover—either by reducing the salience of costs or signaling broad support—could create space for policy makers to support more ambitious plan designs. We find that political cover delivered via economic relief, delayed implementation time, party endorsement, or international support does not make policy makers more likely to support expansive climate policy proposals (i.e., plans with broader-based energy efficiency standards or high short-term costs and high long-term benefits). This null result may be due to the stickiness and polarized nature of climate change policy preferences. Democrats may believe relatively bold climate action is needed, with or without the kind of political cover policies like economic compensation can provide, given the significant threat climate change poses. On the flip side, Republicans may think bold climate action is unnecessary even if political cover is provided. Policy makers may also assume that even with marginally more political cover, their constituents will punish them if climate change mitigation policy produces short-term financial costs. Work by Mildenberger and colleagues (Reference Mildenberger, Lachapelle, Harrison and Stadelmann-Steffen2022), for example, shows that many citizens are unaware of policies like carbon tax rebates. A promising avenue for future work would involve more systematically analyzing the mechanisms underlying this null result.
Second, to assess the extent of elite–public gaps, we fielded an identical replication study on a nationally representative sample of 1,029 Americans recruited in December 2021 via Lucid. We show that members of the public and local policy makers largely hold aligned preferences on climate plans. For example, none of the MMs we estimate differ in sign (i.e., whether the factor increases or decreases support for the climate plan) between policy makers and the public, and over 75% do not significantly differ in size. These findings increase the ease of local climate policy making compared with if policy makers and the public held sharply contradictory views. In accordance with seminal work by Kertzer (Reference Kertzer2022), they also make a contribution by providing additional evidence that elite–public gaps are smaller than some fear, at least for this particular issue.
Still, some caution in interpreting these results is warranted, as we are unable to directly match policy makers with their constituents. CivicPulse does not provide data on which particular locality policy makers represent because, in conjunction with other data points such as age and gender, this information could be used to reveal the identities of policy makers and thus violate the terms under which policy makers agreed to participate in the survey. Nevertheless, since we have both a national sample of policy makers and a national sample of the public, our analysis enables us to assess whether elite–public gaps exist on average. This approach mirrors that of most prominent studies of elite–public preferences, which also do not directly match policy makers with their constituents and instead compare aggregate samples of the public and elites as we do (e.g., Kertzer Reference Kertzer2022; Lupton, Myers, and Thornton Reference Lupton, Myers and Thornton2015; Lupu and Warner Reference Lupu and Warner2022; Martin and Raffler Reference Martin and Raffler2021; Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018).
Conclusion
Despite the significant threat posed by climate change, limited progress at the federal level to combat it and the return of Donald Trump to the White House has increased the importance of initiatives at the local level. This paper addresses previously unanswered questions by examining the causal determinants of local policy makers’ support for climate policy. Our core results provide optimism about the future of climate action at the local level. We find that a range of climate policy attributes, including ambitious plan design elements, have a significant impact on local policy maker support, including for Republican policy makers. This means that the probability of policy adoption can be increased by strategic design. Most notably, climate plans with more hidden costs, such as tax subsidies and regulations, garner greater support than plans with more visible costs, such as taxes and penalties. This suggests that pushing for a carbon tax, as many activists, economists, and even policy makers have done, may not be the most effective political strategy in terms of obtaining support. Our findings also suggest that direct economic compensation to defray the costs of climate policy is popular, and that policy makers should make every effort to avoid the perception that climate plans are partisan, as this significantly reduces support.
Yet given that policy makers face pressures in the real world not modeled in our experiment, a promising avenue for future work would be to test the sensitivity of our results to additional experimental treatments. For example, would the preference for regulations over taxes be diminished if private industry lobbied against the former, and if so, by how much? Another valuable avenue for future work would be to more directly test the mechanism(s) driving elite–public congruence. Our lack of fine-grained data on electoral competitiveness means we cannot directly evaluate whether congruence is more likely for policy makers that have greater political incentives to pay close attention to public opinion. Whether it is the reelection motive that drives our findings, or whether politicians and citizens hold similar preferences in parallel, remains an open question.
Our findings may point to more generalizable preferences that local policy makers hold in different issue areas. For example, preferences for tax subsidies over taxes could translate into attitudes over financing public infrastructure, health, and services. Our results related to the importance of party endorsements and the backlash effect for Democrat-only endorsed plans may also have external validity for other similarly polarized policy areas, such as immigration and LGBTQ rights. Future work can address such questions by applying our research design to assess policy maker preferences across different issue areas while varying key elements of plan design.
These results also may speak to other local contexts, as policy makers around the world play a role in addressing climate change, though their scope for discretion might be narrower than that of subnational governments in the more decentralized US. For example, in the United Kingdom, local councils in cities such as Bristol have declared climate emergencies and adopted climate action plans (Bristol City Council 2025), and cities in Japan such as Yokohama have independently implemented renewable energy programs and zero-waste initiatives (Sato Reference Sato2025). Future studies can also assess which results hold in different country contexts. For instance, partisan endorsements may have weaker effects in countries with lower levels of polarization than the US, though many countries around the world (e.g., Canada) are also experiencing increases in polarization (Boxell, Gentzkow, and Shapiro Reference Boxell, Gentzkow and Shapiro2024). Given prior research showing that carbon taxes are generally unpopular cross-nationally (e.g., Umit and Schaffer Reference Umit and Schaffer2020), we expect our results related to taxes and penalties versus subsidies are likely to have external validity.
Overall, while climate change is a global problem and action can and should be taken at the national and international level to combat it, our project indicates that efforts at the local level also hold promise. “Think globally, act locally,” as the saying goes, is one avenue among many to combat climate change.
Supplementary material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592726106070.
Data replication
Data replication sets are available in Harvard Dataverse at: https://doi.org/10.7910/DVN/4OHGNA
Acknowledgments
This research was funded by the University of Pennsylvania’s Social and Behavioral Science Initiative, the Wharton Risk Management Center’s Russell Ackoff Doctoral Student Fellowship, the Penn Identity and Conflict Lab, and the Kleinman Center for Energy Policy. We thank Claudia Avellaneda, Devon Cantwell-Chavez, Jonathan Chu, Santiago Cunial, Fiona Cunningham, Aaron Deslatte, Julia Gray, Jessica Green, Daniel Hopkins, Rachel Hulvey, Jacqueline Klopp, Siyao Li, Diana Mutz, Michael Ross, Meghan Rubado, Robert Shaffer, Anton Strezhnev, Dustin Tingley, Alex Weisiger, Alan Zarychta, the staff of CivicPulse, and participants of the International Studies Association Annual Convention, the Southern Political Science Association Annual Conference, the Harvard Climate Pipeline Project, the Environmental Politics and Governance Conference, the Belfer Center’s Environmental and Natural Resources Seminar, the Wharton Risk Center Seminar for Early Career Research, and the University of Pennsylvania International Relations Workshop for helpful comments and advice.
Open access
