Race, ethnicity, and support for climate policy

Addressing the increasing temperatures of the globe requires society-wide adaptation and mitigation efforts. One central challenge to these efforts is the resistance of groups to support broad policy efforts to reduce global temperatures, with particular resistance in the United States. While scholars have established that partisanship, ideology, demographic, and socio-economic characteristics shape support for climate policy, we do not yet understand how these factors might vary within and across racial and ethnic groups. In this paper, we use pooled data from the Cooperative Election Study (N total = 241 800) to examine differences in attitudes about climate policy between Asian, Black, Hispanic, and white Americans. Comparing across groups, we demonstrate that the many core findings of scholarship on support for climate policy apply nearly exclusively to white Americans, with varying correlational relationships for Asian, Black, and Hispanic Americans. Our efforts provide a much-needed examination of how racial identity shapes views on climate change and show that central, replicated results in scholarship on climate change apply largely to the views and behaviors of white Americans.


Introduction
Climate change poses severe threats and necessitates urgent mitigative and adaptive policy actions (IPCC 2022). Yet the United States has long struggled to advance effective policy measures, with over two decades of federal inaction until the recently passed Inflation Reduction Act, and increasingly partisan approaches to climate policy at the state level leading to policy retrenchment in several states (Bromley-Trujillo and Holman 2020, Mildenberger 2020, Stokes 2020. Addressing climate change requires both broad collective support for government policies (Stoutenborough et al 2014, Egan andMullin 2017) and an understanding of the needs and policy preferences of communities and demographics worst affected by climate change. It is thus essential to understand how support for such policies varies significantly across individuals and groups.
A broad body of scholarship has established that younger Americans, Democrats, liberals, women, people with higher educational attainment, and more income are more supportive of climate action and various mitigative policies (e.g. Egan and Mullin 2017, Ross et al 2019, Bergquist et al 2020. Yet much of the work on views of climate change draw on disproportionately white survey samples, adversely centering white responses as universal in the published scholarship. In this letter, we argue that the central characteristics documented by scholars as indicative of support for climate policy do replicate-for white people. But a broader understanding of support for climate policy in the population demands that we consider that these mechanisms play alternative roles for Asian, Black, and Hispanic Americans. Within each of these groups, we examine the variation in relationships between established correlates and support for climate policies using pooled data from four years of the Cooperative Election Study (CES) (N = 241 800).
People and communities of color bear the costs of climate change disproportionately, including greater economic consequences, local environmental and health risks, and climate-related displacements (Jesdale et al 2013, Cushing et al 2015, Sultana 2021, Hernandez-Cortes et al 2022. Given these inequalities, Black, Hispanic, and Asian individuals in the United States might value climate action more than white Americans, and may prioritize mitigative responses or adaptive responsibilities differently (Reckien and Petkova 2019). Indeed, gaps between racial and ethnic groups in their reported concerns over climate change are evident in survey data. For example, survey data from the past decade shows that Black and Hispanic Americans perceive greater risks from climate change than do white Americans Pearson 2016, Ballew et al 2021). Figure 1 presents aggregate levels of support for climate policy in the CES, which allow us to see how these trends extend to other dimensions of policy support. Here, we see differences both within and across race and ethnic groups. For each policy measure (CO2 regulation, state-level renewable mandates, and Environmental Protection Agency (EPA) enforcement of the Clean Water Act (CWA) and Clean Air Act (CAA)) and an aggregate measure of climate policy support, we see that support among Black, Hispanic, and Asian respondents exceeds white support, with smaller differences across these three groups.
But views about climate are not just driven by experiences with the consequences of climate change (Egan and Mullin 2012, Bergquist and Warshaw 2019, Bush and Clayton 2022, Howe et al 2019, but also because of social identities and sociodemographic characteristics (Fielding and Hornsey 2016, Ross et al 2019, Motta 2019, and different experiences with government services (Reckien and Petkova 2019). We find broad evidence for the importance of considering the ways that political, socio-economic, and demographic characteristics work in distinct ways across racial identities. While partisanship, ideology, and education are uniformly significant, all measures exert differing substantive effects. We also document variations in the direction and significance of socio-economic and demographic variables. The trend of older respondents being less supportive of climate action is limited to only white respondents. So too is the gender gap in environmental concern: white women are more supportive of climate policy, but Black men are more supportive than Black women.

Examining support for climate policy across racial and ethnic groups
To evaluate how support for climate policy varies across racial and ethnic groups, we use pooled data from the 2014. The large N of these datasets allow us to examine differences in attitudes about climate change policy across both race and ethnicity. We focus on four subgroups within our pooled sample: respondents who identify as Asian/Asian-American (N = 7077), non-Hispanic Black (N = 26 803), Hispanic (N = 27 109), or non-Hispanic white (N = 172 149). We use the terms Asian, Black, and white to refer to Asian or Asian-American or Pacific Islander, non-Hispanic Black, and non-Hispanic white respondents respectively through the rest of the paper. Hispanics in the paper are identified both through selecting Hispanic as a racial identity (i.e. from a set of racial categories) and as an ethnicity (i.e. as selecting it as an identity separate from the racial categories), given that Hispanics in the United States may define racial or ethnic identities in multiple ways (Masuoka 2008, Wong 2018. We omit other groups from our analysis given smaller sample sizes for other demographics, or because other response categories for the CES' racial identity question combine multiple racial or ethnic groups into categories such as Mixed or Other. We recognize this omits respondents who may hold multiple and possibly cross-cutting identities. For example, those who respond to the racial identity question as 'Mixed' account for 1.5% of the pooled CES sample. We also recognize that some of these racial and ethnic identities may overlap: for example, 2.7% of respondents identify as both white and Hispanic, and 2.5% of respondents identify as both Black and Hispanic. As the subset of respondents who report 'Mixed' racial identity have varying and distinct political experiences and views (Merseth 2018, Lemi 2021), we do not analyze them in the main manuscript, but we do present these results in appendix B4.
The CES (formerly known as the CCES) is a nationally representative survey that samples from YouGov panels, advertisements, and other survey draws with sample sizes of over 50 000 in each survey wave. Samples are weighted post-collection to match the distributions of the most recent American Community Survey demographics on age, gender, race, education, and Hispanic origin with additional poststratification measures based on previous national election votes, voter registration status, and to match state populations. See table 1 for information on the measures we use from the CES and the appendix for more information on the survey and sampling process.
We examine the average support across three climate policies as our dependent variable. In each survey wave, the CES asked respondents whether they support (1) or oppose (0) each of the following: (a) 'Give the EPA power to regulate Carbon Dioxide emissions.' (CO2 regulation); (b) 'Require that each state use a minimum amount of renewable fuels (wind, solar, and hydroelectric) in the generation of electricity even if electricity prices increase.' (Renewable mandate); (c) 'Strength the EPA enforcement of the CAA and CWA even if it costs US jobs.' (CAA/CWA enforcement). We average these three items into an aggregate indicator of climate policy support (α = 0.792; average interitem covariance = .129).
For each of our models (see table 2), we estimate separate ordinary least squares (OLS) models for each racial or ethnic group with fixed effects for survey year and survey weights; separate models for each survey year are available in appendix B2, and models without survey weights are available in appendix B3. We also estimate separate models using the categorical version of the partisanship and ideology variables and the effects for partisanship and ideology in separate models, following recommendations from Kalmoe (2020) as ideology may overlap with and capture several aspects of partisanship. A model including measures of both partisanship and ideology is available in appendix B1. After presenting a central set of models, we examine the effects of individual independent variables and present those results visually to aid in interpretation.

Correlates of climate policy support
Three patterns emerge from our central results that demonstrate the need to examine how factors vary across and within racial and ethnic groups (see table 2). First, the models for white respondents have a higher explained variance compared to other groups, but especially Black respondents. Second, the statistical significance and direction of effects for some variables are similar, but the size of the coefficients vary substantially across racial and ethnic groups. For example, we find that Democrats and liberals across all racial and ethnic groups are more supportive of climate policy than Republicans or conservatives, consistent with scholarship Mullin 2017, Guber 2017). However, the effect sizes vary considerably, with the largest effects among white respondents, and smallest among Black respondents. In comparison, while education is associated with more support across all groups, the effect sizes of increased educational attainment are higher among Black, Hispanic, and Asian respondents than white respondents. And third, some explanatory variables differ either in significance or direction across racial and ethnic groups. While age is negatively associated with support for climate policies among white, Asian, and Hispanic respondents, we see the opposite effect among Black respondents. Similarly, white respondents' income negatively correlates with climate policy support, but this relationship is positive among Black respondents, and non-significant among Asian and Hispanic respondents. We next engage in more detailed examinations of those variables where we find similar directional and significance effects but varying substantive effects, including partisanship, ideology, and education. We then examine those factors where significance or direction shifts across racial and ethnic groups, including age, gender, education, and income.

Partisanship
Partisanship exerts a much stronger influence on white respondents' climate policy preferences than on policy support among Black, Hispanic, or Asian respondents. Figure 2 presents predicted support for our aggregate measure of policy support among each racial or ethnic group across partisanship. Bolded lines represent post-estimation predicted probabilities for each of the racial or ethnic groups in our analysis, and dashed lines represent 95% confidence intervals.
Partisanship is meaningfully different for each ethnicity, with a move from Strong Democrat to Independent or from Independent to Strong Republican associated with declines in predicted policy support. The slopes of these effects are similar for Black, Hispanic, and Asian respondents, with a far steeper incline for white respondents whose predicted policy support drops by nearly 50% when shifting from Strong Democrat to Strong Republican.
One possibility is that lower political homogeneity among white Americans drive these results. Black voters in the United States overwhelmingly align with the Democratic Party in their political preferences and tend to have far weaker ties to the Republican Party (White and Laird 2020). Hispanic and Asian voters also align with the Democratic Party, albeit with greater heterogeneity across subgroups (Mishra 2016, Sadhwani and Mendez 2018, Lee and Ramakrishnan 2020. Racial threat and exclusion during the Trump Presidency have also led to Hispanic and Asian voters becoming more politically mobilized and engaged with the Democratic Party Arora 2018, Gutierrez et al 2019). In contrast, white voters' partisanship is far more heterogeneous and right-leaning (Cluverius et al 2020).

Ideology
Like partisanship, ideology is significantly correlated with environmental policy support for all groups, with stronger conservative identity associated with reduced policy support. And like partisanship, ideology's effects are strongest among white respondents and smallest among Black respondents. We plot this in figure 3, with bolded lines representing post-estimation predicted probabilities for each of the racial or ethnic groups in our analysis, and dashed lines representing 95% confidence intervals. One possible reason this effect size may be so much smaller among Black Americans is that conceptualizations of labels like liberal and conservative vary greatly across racial and ethnic demographics, with Black Americans viewing these terms and concepts differently than white Americans (Jefferson 2020). In both figures 2 and 3, we see the largest divides between white respondents who identify as Note: Post-hoc predicted probabilities from appendix A1. Solid lines are coefficients, dashed gray lines indicate confidence intervals. Year fixed effects, controls for age, education, gender, and income. Survey weights applied. Note: Post-hoc predicted probabilities from appendix A2. Solid lines are coefficients, dashed gray lines indicate confidence intervals. Year fixed effects, controls for age, education, gender, and income. Survey weights applied. either Strongly Republican or extremely conservative and all other groups. Even compared to non-white respondents who share these political identities, white respondents have lower levels of support for climate policy.

Education
We next examine differences across educational groups, separating our ordinal measure of educational attainment into a set of categories. Recall the significant relationship for education across all models, but substantively larger effects for Black, Hispanic, and Asian respondents. We examine predicted probabilities (figure 4) for each of these groups. We see differences in the slope for the effects of education across racial and ethnic groups. For Black and Hispanic respondents, the effects of education are large, significant, and positive. For these respondents, we see significant increases in support for climate policies when comparing respondents with a high school degree or some college to those with a 4 year college or post-graduate degree. Trends are smaller among Asian respondents, where the only significant difference is between Asian respondents with post-graduate degrees and those with high school education. Among white respondents, we see extremely small increases in the likelihood of policy support with increasing educational attainment.

Age differences
We now turn to those variables in the models presented in table 2 where we found differences in significance and direction across racial and ethnic groups. We start with age, where we use the model from table 2 and break down age into six categories. Figure 5 presents predicted probabilities (represented as bars) for each of these age groups.
Age exerts different patterns on support across race and ethnic groups. Among white respondents, we see a linear, negative relationship as older respondents are less and less supportive of climate policies. This pattern reflects findings from other scholarship that highlights how younger generational cohorts report both stronger agreement with the scientific consensus on climate change and greater support for related policies, relative to older cohorts (Ross et al 2019, Milfont et al 2021. Among Asian and Hispanic respondents, age and climate policy support are not significantly correlated. We see quite different patterns among Black respondents where age is positively correlated with support for climate policy, with the 65+ group supporting climate policy at the highest rate. There are several possible post-hoc explanations for this finding. One is that younger Black Americans may hold lower levels of concern for environmental and climate-related issues due to their prioritization of other policy issues relating to racialized economic inequities, policing and criminal justice, and similar issues that they may see as more closely impacting the Black community (Jefferson et al 2021, Peterson andRiley 2022). Alternatively, older Black Americans (who are strong Democratic voters) may be more receptive to party cues. Given both the ascendancy of climate change and increased prioritization of environmental issues within the Democratic Party in recent years (Egan et al 2022), older Black voters may be more exposed to, and more receptive to, partisan messaging highlighting the urgency of climate change and mitigative action. The well documented pattern of older Americans opposing climate action should be limited to older white Americans.

Gender differences
In examining gender differences, we again see different patterns for white and non-white respondents (see figure 6). A gendered gap among white respondents follows the common finding (McCright and Dunlap 2011a, Swim et al 2018): white men are significantly less likely to support environmental policies than white women. However, this pattern does not apply to other groups: Black men are more likely to express support for the policies surveyed than Black women and we see no gender differences between Asian or Hispanic men and women in their support for climate policies. This reinforces that the pattern of gendered differences in climate concern and policy support is a white male effect (e.g. McCright and Dunlap 2011b) and shaped by both the intersections of gender and race, rather than primarily gender.

Income
We finally examine differences by income category across racial and ethnic groups (see figure 7). Income is significantly associated with climate policy support for two groups-white and Black respondents, in opposite directions. While we see negative and significant effects for white respondents, Black respondents' support for climate policies increases with higher levels of income 4 . We find no relationship between income and policy support for either Hispanic or Asian respondents.

Discussion
In this note, we present an exploratory analysis of how the correlates associated with support for climate policy are highly dependent on the racial and ethnic group of an individual. We document several trends: both partisanship and ideology are correlated across groups, but that the substantive effects are much larger and consistent for white respondents. We also find that education, age, gender, and income all work in unique ways for different racial  and ethnic groups. Using data from four waves of the CES allows us to examine differences not just across racial and ethnic groups, but within these groups. This allows us to extend other work examining differences in environmental risk perceptions and beliefs across racial and ethnic groups Pearson 2016, Ballew et al 2021) by not only examining differences between these groups, but systematically examining how different factors such as age, gender, and political identity may behave within large and generalizable samples of Black, Hispanic, and Asian American respondents.
Our results draw attention to the racialized nature of political life in the United States (Azevedo et al 2022) and the consequences for environmentalism and climate action. The spillover of racial attitudes into climate views has been the focus of some work (Benegal 2018, Chanin 2018, Benegal and Holman 2021b, Benegal and Motta 2022; here, we examine how racial identity itself shapes climate views. In doing so, we build on work that demonstrates the importance of considering how central race is to shaping political views (DeSante and Smith 2020) and the variations within and across racial groups (Mohamed 2015, Raychaudhuri 2018, Bunyasi and Smith 2019.
While we know that some forms of messaging and policy bundling are more effective among different racial and ethnic demographics (e.g. Bergquist et al 2020) and that political divides on climate change are lower among Black and Hispanic Americans (Schuldt and Pearson 2016, Ballew et al 2021), we still know much less about the factors that may underpin and may mobilize climate policy support among different racial and ethnic groups. Studies of public opinion in other policy arenas documents how variation occurs both across and within racial and ethnic subgroups (Lien 1998, Fraga 2018, Fraga et al 2020, as well as by intersectional groups (Brown 2014, Brown and Gershon 2017, Silva and Skulley 2019. The work we have produced here contributes to these conversations by documenting differences in the factors associated with support for climate policy both across and within racial and ethnic groups. Most major surveys of American politics simply do not include a large enough sample of racial and ethnic groups to engage in this research. Even with more than 240 000 total respondents, we are constrained by a low n and are unable to examine patterns among Native Americans, Pacific Islanders, or individuals who identify as mixed race; these groups have distinct political experiences and policy views (Merseth 2018, Lemi 2021. We hope that future research might take advantage of specialized samples to evaluate patterns among these groups and may need to draw on alternate forms of sampling (e.g. respondent-driven or snowball sampling) to better understand public opinion among harder to reach populations that may be more underrepresented in national surveys. As surveys often overrepresent more politically engaged individuals (Keeter 2018), alternate sampling methods may be necessary to improve our understanding of environmental public opinion dynamics among different racial and ethnic demographics, particularly given how political efficacy and engagement may differ across these groups (Phoenix and Chan 2022). Previous studies document these trends using probability-based surveys Pearson 2016, Ballew et al 2021); that we find consistent results using larger, non-probability survey data from the CES provides additional evidence of these patterns.
In this paper, we document differences by the common correlates from the scholarship: partisanship, ideology, education, income, gender, education, and age. Future research might also consider the ways that psychological mechanisms and patterns that undergird preferences for climate policy also may vary across and within racial and ethnic groups. For example, researchers document how both policy framing and bundling may affect policy preferences (Stokes andWarshaw 2017, Bergquist et al 2020). We also know that senses of group efficacy and their relationship to political behavior and engagement differ across racial and ethnic groups (Phoenix and Chan 2022), and understanding how different group-specific mechanisms of public opinion and efficacy are related may help us better understanding environmental advocacy and engagement, particularly on environmental justice matters in Black and Hispanic communities. Others document how preferences for the status quo and system justification both correlate with climate views and policy preferences Jost 2021, Benegal andHolman 2021a), and how underlying psychological mechanisms vary across racial groups (Friesen andHolman 2022, Holman 2016). Understanding how these patterns might overlap and intersections offers a myriad of opportunities for future research. And, given that public opinion on climate shapes climate policy (Bromley-Trujillo and Poe 2018), understanding how these patterns might produce varying cumulative effects across geographic areas would be worth evaluating.

Data availability statement
The data and replication code that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.7910/DVN/7HLK04 .