Negative partisanship, positive partisanship, and variation in climate policy attitudes on the political right

Abstract

Conservatives are more likely than liberals to oppose climate policies, resulting in political polarization over climate change. Most research treats this gap as if it reflects two cohesive blocs on opposite sides of an issue. Drawing on original survey data from a probability sample of Canadians (n = 2,503), we find that while liberals are highly uniform in their orientation toward climate policies, conservatives are far more heterogeneous. Further analyses reveal conservatives' policy positions strongly correlate with their partisan affect—both the extent to which they dislike opposing liberals (negative partisanship) and the extent to which they like fellow conservatives (positive partisanship). These findings highlight the importance of considering variation within, and not just between, political sides. The results additionally suggest that reducing hostility toward the other side (particularly among conservatives) may facilitate cross-ideological climate coalitions.

Significance statement.

Many nations are reportedly stymied in their ability to enact climate legislation due to politically polarized views on climate change. Right-wing voters are often framed as a barrier to climate policy. Existing studies report on the “typical” conservative or liberal. We shift that focus to examine heterogeneity in attitudes toward climate policy. Using a probability sample of 2,503 Canadians, we found that while those on the political left are homogenous in their support for climate change policy, right-wing respondents are highly heterogeneous. Follow-up analyses reveal that conservatives' opposition to climate policy strongly correlates with the degree to which they like other conservatives and dislike liberals. These findings have implications for climate change communications and studying political polarization over climate change policy.

Introduction

Climate mitigation policy in democratic nations requires coordination and compromise between political elites, scientific experts, and citizens. Yet sound public policies are often rejected due to partisan or ideological divides. This dynamic is exemplified by a seemingly entrenched ideological gap in support of climate policy among the public. In a meta-analysis of studies from 33 nations, Bergquist et al. (1) found that self-identification with a conservative, or right-leaning, political ideology was negatively associated with support for climate change taxes and laws. Although the effect was small in Europe and Oceania, it was moderate in North America. Moreover, the polarization of climate policy support has widened over time, particularly in the United States, other Anglophone nations including Canada, and across Western Europe (2). An analysis of climate attitudes in 60 nations observed both a political divide, with weaker belief in climate change and support for climate policy among conservatives, and greater variation in attitudes among the political right than the political left (3).

In contrast to this focus on the “typical” left- and right-wing orientations to climate policy, some research emphasizes a range of positions on climate change. For instance, the Six Americas framework (4) documents six clusters of American orientations to climate change, ranging from highly engaged to dismissive. Similar analyses of Canadian data reveal five clusters of Canadians, with the same variation in range of orientations (5, 6). Other analyses call attention to variation in climate opinions within the political right: for instance, younger Republicans are much more likely to believe in global warming than older Republicans—and the difference between older and younger Republicans is almost as large as the difference between Republicans and Democrats (7). Mildenberger et al. (8) identify substantial heterogeneity in climate and energy opinions among both Democrats and Republicans, but note that “for some policy items, we see evidence of substantially less heterogeneity among Democrats as compared to Republicans.” We contribute to this stream of work by examining heterogeneity among those on the left and right, and the correlates of heterogeneity on the political right.

A structural reason to expect more heterogeneity on the right is that liberals' climate policy opinions tend to cluster near the high end of all available scales, whereas conservatives' responses are more dispersed across the range (9–11).^a While this pattern may in part be a scale effect, it is also likely reflecting substantive differences in opinion structure: liberals tend to converge on strong support for climate initiatives, while conservatives are divided among support, ambivalence, and opposition. That is, the right appears to have a wider distribution of policy positions than the left.

We consequently anticipate significantly greater heterogeneity in the climate policy opinions of conservatives than in the climate policy opinions of liberals. We expect little variance on the left. We evaluate our expectation by estimating the degree of—and identifying the correlates of—intra-ideological variance. Understanding such heterogeneity is crucial since addressing climate change requires coalition building (12, 13). Given our expectation of variance on the right but not on the left, we focus on potential correlates of intra-ideological variation among conservatives. We look at four classes of variables that might correlate with climate policy attitudes among those on the right.

First, a sociodemographic approach examines how attitudes vary across different subgroups. Prior work shows that opposition to climate change among both American (14) and Norwegian (15) conservatives is strongest among White, male respondents. Geographic context also might matter. Living in areas that experience substantial economic tolls from climate mitigation policies can vitiate policy support (16). All of that said, it is notable that existing work that includes multiple correlates often finds relatively small effects for sociodemographic variables (1, 17).

A second line of explanation centers on ideology, including worldviews, such as free-market individualism (18), intrinsic tendencies to see the world as fixed or malleable (19), and orientations toward government intervention (20). We do not explicitly probe these worldviews but, as a proxy, we investigate the role of ideology extremity among conservatives, with the expectation that more extreme ideologies generate stronger climate policy opinions.

Third, the expressive approach attends to social and/or emotional attachments (21, 22). In the context of climate attitudes, this approach is evident when individuals adjust their climate stances based on their perceptions of their in-group's preferences (23–26) and/or their sentiments toward the out-group (27, 28). Climate attitudes reflect social identities and group loyalties; for instance, a study in the United States showed that Republicans conditioned their climate change opinions on their perceptions of group norms among other Republicans (23). This approach coincides with partisan-motivated reasoning, where partisans evaluate evidence that supports their group's opinion as being more trustworthy and higher quality than evidence that contradicts the group consensus (29).

An extension of the expressive approach highlights the role of affective polarization in shaping policy positions. Affective polarization is typically defined as the gap between individuals' positive feelings toward their own side (positive partisanship) and negative feelings toward the opposing side (negative partisanship) (30). In terms of policy, the more one is affectively polarized, the more they will be inclined to follow group norms in taking particular policy positions. Higher levels of affective polarization can lead individuals to follow cues from elites on their side and/or reject cues from elites on the other side (31, 32). In the case of climate change, liberal elites typically endorse climate policies while conservative elites oppose them (13). Consequently, those higher in partisan affect will presumably hold more extreme views compared with those lower in partisan affect.

Recent work suggests it is important to analytically distinguish the components of affective polarization since positive and negative partisanship may capture somewhat different processes (33, 34). Following this distinction, we examine positive and negative partisanship separately. This allows us to speak to some disagreement in the literature about the relative salience of positive and negative partisanship (32, 35). We expect more positive partisanship (warmth toward one's own side) or negative partisanship (hostility toward the other side) to generate more extreme policy opinions (for conservatives, this means more opposition). Mayer and Smith (34) evaluated this expectation in the United States. Interestingly, they found that higher levels of positive partisanship were associated with increased climate policy support among liberals while greater negative partisanship was associated with decreased support among conservatives. This dynamic aligns with research highlighting the emotional foundations of climate-relevant attitudes, including how feelings of exclusion from environmentalism fuel conservative opposition to left-leaning policies (36, 37). This suggests that, on the right, negative partisanship will correlate more strongly with climate policy beliefs (than positive partisanship).

Finally, we examine distrust in science. Prior work shows that greater distrust in science underlies conservatives' (relative) opposition to climate policies (38–40). This follows given that the starting point for climate policy is recognizing the science underlying climate change.

The Canadian context

Our study focuses on the Canadian context. In Canada, divides over climate change map onto the broader left–right spectrum in ways generally analogous to the partisan divides in the United States (41). The Liberal Party and New Democratic Party anchor the center-left and left, respectively, and both generally support decarbonization policies. The Conservative Party represents the political right and has been more skeptical of climate policy (42). Although polarization in Canada is less entrenched than in the United States (43, 44), partisan divides are evident (45). Unlike the two-party system in the United States, Canada's multiparty system produces a wider range of partisan positions, but debates over climate policy are nonetheless divided between the aforementioned left-leaning parties and the conservatives (i.e. as two blocks). Since the United States is a global outlier in terms of the strength of the ideology-climate change link (44) and the rate of the increase in negative partisanship (46), the Canadian case allows us to evaluate whether variables like negative and positive partisanship and distrust in science account for heterogeneity in climate policy attitudes beyond the unique American context.

This case study is also of timely political importance. After positioning itself as a global leader on climate change following the 2015 Conference of the Parties, the federal Liberal government passed the ambitious “Pan-Canadian Framework on Clean Growth and Climate Change (2016),” which included both a consumer and industrial carbon tax. Despite opposition from some conservative provincial governments and consecutive leaders of the federal Conservative Party (47), the “Greenhouse Gas Pollution Pricing Act” (2018) passed, and the federal government spent $100 billion on climate measures between 2015 and 2020 (42). Yet multiple exogenous shocks—including the COVID-19 pandemic and its impact on inflation, shifts in global fossil fuel demand stimulated by the war in Ukraine, and United States tariff threats—contributed to a decline in public support for the consumer carbon tax, dropping from a high of 56% support in 2021 down to just 45% in early 2025 (48). While support declined more among supporters of the Liberal Party (13% decrease) than among supporters of the Conservative Party (5% decrease), over this period, relative support for the tax remained highly partisan at 70 and 8%, respectively (48 ). In the spring of 2025, new Liberal Party leader Mark Carney called an election, won (a minority government), and on his first day in office, announced the end of the consumer carbon tax while reaffirming his government's commitment to the industrial carbon tax. These shifts raise pressing questions about the state of ideological divides in public opinion on climate policy, particularly the role of positive and negative partisanship in shaping climate attitudes, and emphasize the need to better understand how the public perceives decarbonization strategies that target industrial emissions.

Research objectives

Our study aims (i) to describe the nature of ideological polarization over climate policy attitudes and (ii) to identify the factors that account for variance in conservatives' attitudes. We focus on policy attitudes since policy coalitions between parties are an essential element of climate governance (49). We designed the survey to examine support for government climate policy generally, support for a carbon tax on industrial emitters, and support for replacing fossil fuels with renewable energies (decarbonization). To account for variables that might correlate with climate policy attitudes among conservatives, we examine the aforementioned approaches: sociodemographic measures (including gender, race/ethnicity, and geography), ideological extremity, negative and positive partisanship, and (dis)trust in science.

Results

We analyze data from an original survey delivered to a probability sample of Canadians (with poststratification weights to match the most recent Canadian census; n = 2,503, Summer 2024). Respondents were randomly recruited from the 90,000-person research panel of EKOS Research, each of whom was originally recruited into the panel via random-digit dial of cell phone and landline numbers. Live interviewers contacted participants via phone. Our dependent variables were constructed from three climate policy questions that asked respondents about their support for more government action on climate change, an emissions tax on industries, and replacing fossil fuels with renewable energies. Support for each ranged from 1 (strongly oppose) to 7 (strongly support; see Materials and methods).

Independent variables

The models of conservative climate policy support included 12 independent variables (see Materials and methods; Table S1). Sociodemographic variables include measures of gender, race/ethnicity, and region. We also include controls for respondents' age, sexual identity, education, ties to the oil and gas industry, and rural or urban residence. We assess political ideology for the full sample analyses (ranging from 0, very left wing, to 5, middle of the road, to 10, very right wing (45)). The subsample of conservatives includes respondents who self-identified with numbers 6 through 10, and we include degree (extremity) of conservatism as an independent variable to assess the ideology approach (30).

Most of the early work on positive and negative partisanship concerned the US context, with a straightforward operationalization of gauging evaluations toward the two main parties (Democrats and Republicans) (30). Scholars have used various measures when moving to a multiparty context like Canada (50). This includes looking at only the dominant parties (51), at all parties weighted by their relative size within the electoral system (52, 53), and at dyads of all party combinations (54). In the contemporary Canadian context, recent work suggests grouping parties into two ideological blocks: liberal or conservative (consistent with parties' climate stances, as discussed). Johnston explains that the party system has evolved such that voters no longer reward or punish individual parties for their performance but instead think about the victory of the two blocks (55, 56). He states, “The system can be conceived in continental European terms as comprising party families or blocks…the Conservatives and any party to its right as the ‘right’ block and the Liberals and all others as the ‘left’” (45, 55, 56). We thus asked separate feelings thermometers, about the left (“politically left-wing, or liberal”) and the right (“politically right-wing, or conservative”), with each ranging from 0 (cold) to 100 (warm) (30). We operationalized positive partisanship in terms of how warmly someone rates their in-group, and negative partisanship in terms of how cold they rate the out-group. (We thus took the inverse of the 0 to 100 scale for negative partisanship, so higher scores indicate negativity/coldness).

We assess the impact of distrust in science, where higher values indicate higher distrust (57).

Full sample: identifying heterogeneity in climate policy attitudes

Our first objective is to describe the nature of ideological polarization over climate policy attitudes. Figure 1 shows support for three climate policies across self-identified political ideology. The boxplots are bivariate (unadjusted for other variables) and therefore show starker ideological divisions. As expected, support for each climate policy measure is uniformly high on the political left. On the 1–7 scale, the median support among respondents selecting any range of liberal ideology (0, 1, 2, 3, 4) was 7 for each policy. As political ideology becomes more conservative, median support decreases. And, as anticipated, support on the right for each policy is highly heterogeneous. Levene's tests confirm substantially greater variance among conservatives than among moderates or liberals for all three policies (decarbonization: 505.79, P < 0.001; government action: 502.94, P < 0.001; industrial carbon tax: 501.00, P < 0.001; see Table S2). These metrics justify our subsequent focus on examining heterogeneity within the right.

Figure 1.

Boxplots showing distribution of support for (a) decarbonization, (b) government action on climate change, and (c) an industrial carbon tax by political ideology. Policy support ranges from 1 = strongly oppose to 7 = strongly support. Ideology ranges from 0 = very left wing to 10 = very right wing. Results are for the full sample.

Conservative subsample: explaining variation in climate policy support

We focus on explanatory variables within the right (n = 798), based on expectations from existing literature. This decision also makes empirical sense given there is virtually no heterogeneity in climate policy support on the left and little variation among moderates. We conducted three linear regressions (for each climate policy attitude) and dominance analyses (DAs). From the linear regressions, we present marginal estimates of climate policy support for key variables identified by the DAs.

Table 1 displays results from the three linear regression models (see Tables S4, S6, and S8 for full results, including CIs) (58).^b (See Table S3 for standardized coefficients and t test values.) The variance in support for decarbonization, government action, and support for an industrial carbon tax explained by variables in the models was substantial: 0.3344, 0.4007, and 0.3354, respectively. The results were fairly similar across models. Beginning with demographics, we see that, as expected, some non-White conservative Canadians, women, and those from certain regions (those less economically dependent on the fossil fuel industry) are more supportive of climate policies. Specifically, Asian Canadians were more supportive than White Canadians of decarbonization (coefficient = 0.77, P < 0.025), government action (1.00, P < 0.019), and an industrial carbon tax (0.85, P < 0.044). Black Canadians were more supportive than White Canadians of government action (1.66, P < 0.001), and Latinos were more supportive of decarbonization than White Canadians (1.92, P < 0.031). Indigenous Canadians and multiracial Canadians or those who selected another race were more supportive of an industrial carbon tax than White Canadians (1.09, P < 0.012; 1.42, P < 0.001). No other racial/ethnic differences emerged. Men were less supportive than women of government action and an industrial carbon tax (−0.44, P < 0.008; −0.68, P < 0.001). Quebecois were more supportive than people in the Prairies across all three measures (0.61, P < 0.031; 1.09, P < 0.001; 0.97, P < 0.001). People in Atlantic Canada were more supportive than people in the Prairies of decarbonization (1.34, P < 0.003). Additionally, age was positively associated with support in all three models (0.02, P < 0.001 or P < 0.002 for each). Gays/lesbians were more supportive of decarbonization than heterosexuals (1.02, P < 0.036). Education, ties to the fossil fuel industry, and rural/urban status were not associated with climate attitudes.

Table 1.

Linear regression of support for decarbonization, government action to address climate change, and an industrial carbon tax among the right.

	Decarbonization		Government action		Industrial carbon tax
	Estimate	SE	Estimate	SE	Estimate	SE
Positive partisanship	−0.02***	0.004	−0.03***	0.003	−0.02***	0.004
Negative partisanship	−0.02***	0.003	−0.02***	0.003	−0.02***	0.004
Degree of conservatism	−0.12	0.08	−0.02	0.06	−0.03	0.07
Distrust in science
A lot of trust	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Some trust	−0.49**	0.18	−0.69***	0.18	−0.45*	0.19
Not much trust	−0.99*	0.46	−1.30***	0.35	−0.70	0.37
Not at all trust	−1.57***	0.37	−1.37***	0.31	−0.89*	0.35
Gender
Woman	Ref.	Ref.	Ref.	Ref.
Man	−0.01	0.18	−0.44**	0.17	−0.68***	0.18
Nonbinary	0.60	0.69	−0.30	0.50	−0.31	0.77
Race/ethnicity
White	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Asian	0.77*	0.34	1.00*	0.42	0.85*	0.42
Arab	−0.04	1.18	0.85	0.48	0.51	1.07
Black	0.36	0.68	1.66***	0.46	0.98	0.73
Latino	1.92*	0.89	1.35	0.99	1.54	1.02
Indigenous	−0.10	0.45	0.17	0.29	1.09*	0.43
Other/multiracial	−0.13	0.38	0.35	0.60	1.42***	0.39
Age	0.02***	0.01	0.02***	0.01	0.02***	0.01
Sexual identity:
Straight	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Mostly straight	0.55	0.42	−0.15	0.61	0.26	0.42
Bisexual	0.22	0.53	0.63	0.43	0.86	0.54
Gay/lesbian	1.02*	0.48	0.65	0.75	0.79	0.56
Other	0.70	0.60	0.41	0.47	0.86	0.58
Don’t know/refused	0.47	1.21	−0.10	0.72	0.61	0.92
BA or above	0.29	0.20	0.18	0.19	0.11	0.20
Ties to the oil, gas, and coal industry
None	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Occupation and close contacts	−0.40	0.34	−0.30	0.29	−0.01	0.33
Occupation but no close contacts	0.14	0.35	−0.48	0.36	0.34	0.37
Close contacts but no occupation	−0.34	0.23	0.19	0.21	0.11	0.21
Rural	0.33	0.21	0.27	0.20	−0.11	0.22
Region
Prairies	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
British Columbia	0.00	0.30	0.01	0.29	0.04	0.29
Ontario	−0.15	0.25	0.22	0.23	0.23	0.25
Quebec	0.61*	0.28	1.09***	0.26	0.97***	0.30
Atlantic	1.34**	0.46	0.68	0.40	0.49	0.40
Constant	6.95***	0.70	6.54***	0.65	6.28***	0.67

^*** P < 0.001,** P < 0.01, *P < 0.05, two-tailed. R² statistics are 0.3344, 0.4007, and 0.3354, respectively. BA or above is compared with less than a BA. Rural residence is compared with urban. See Table S3 for standardized coefficients and t test values.

Counter to the expectation of a link between ideological intensity and support for climate policy on the right, degree of conservatism was non-significantly associated with climate attitudes in all models (−0.12, P < 0.13; −0.02, P < 0.700; −0.03, P < 0.705).

Both positive and negative partisanship were strongly linked to climate attitudes. Positive partisanship was associated with lower conservative support for decarbonization, government action, and an industrial carbon tax (−0.02, −0.03, and −0.02, all P < 0.001), as was negative partisanship (−0.02 and P < 0.001 in all models). In other words, greater warmth toward the right and greater coldness toward the left were associated with lower support for climate policy. While this affirms the role of affect, it is counter to the expectation that negative partisanship would display a stronger relationship than positive partisanship.

As expected, distrust in science was generally associated with lower support. Compared with people with a lot of trust in science, people selecting some (−0.49, P < 0.007), not much (−0.99, P < 0.032), and not at all (−1.57, P < 0.001) had far lower levels of support for decarbonization.

While linear regressions are helpful for identifying significant associations between variables, they cannot identify the substantive importance of variables in models. DAs do. They identify the proportion of explained variance attributable to variables in the models. In this study, DAs explain why levels of support differ for decarbonization, government action, and a carbon tax by pinpointing independent variables most responsible for variance in these dependent variables. Figure 2 shows the results for DAs for each model (see Tables S5, S7, and S9 for full results).

Figure 2.

DAs depicting variance explained within conservative support for decarbonization (green/left bar), government action on climate change (blue/center bar), and an industrial carbon tax (purple/right bar). Percentages refer to the percentage of explained variance attributable to a particular variable in the model.

The DAs collectively show that demographic variables, ideology, and distrust in science are responsible for a relatively small amount of explained variance (Fig. 2). While several variables were significant in regression models, most of them explained little variance. In contrast, both positive and negative partisanship were responsible for the most explained variance, at 21–29% depending on the model.

Given the substantive importance of positive and negative partisanship, we next present marginal estimates for both from the regression models. The pattern is similar: as conservative positive partisanship (i.e. warmth toward the right) increases, support for climate policy decreases, and as conservative negative partisanship (i.e. coldness toward the left) increases, support for climate policy decreases (Figs. 3–5). These findings underscore the substantial role affect plays in differentiating those on the right who support climate policy from those who oppose it. The results also differ from those in the US context (that shaped our expectation) where, as mentioned, among Americans on the right, only negative partisanship shaped climate policy opinions (34). Here, we see both positive and negative partisanship matter.

Figure 3.

Support for decarbonization by positive partisanship (warmth toward the right, in red/upper line at partisanship = 100) and negative partisanship (coldness toward the left, in blue/upper line at partisanship = 0).

Figure 4.

Support for government action on climate change by positive partisanship (warmth toward the right, in red/upper line and higher density at partisanship = 100) and negative partisanship (coldness toward the left, in blue/lower line and higher density at partisanship = 0).

Figure 5.

Support for an industrial carbon tax by positive partisanship (warmth toward the right, in red/upper line and higher density at partisanship = 100) and negative partisanship (coldness toward the left, in blue/lower line and higher density at partisanship = 0).

The results are robust to alternative model specifications, including unweighted models (Tables S10–S15), models that excluded respondents on the right who indicated greater warmth toward the left than the right (Tables S16–S21),^c and models that examined Conservative Party supporters (n = 610) rather than people who selected values of 6–10 on the political ideology measure (Tables S22–S27).

Discussion

This study contributes to research on beliefs about climate policy by identifying and analyzing heterogeneity in attitudes among Canadians and focusing on the role of in-group warmth and out-group coldness in explaining that heterogeneity. Little existing research estimates the amount of variance in climate policy support within political ideology groups, focusing instead on the differences between a “typical” liberal and conservative.

Unlike the political left that uniformly supported climate policy, there was only some heterogeneity among political centrists, and a high level of heterogeneity among right-leaning respondents. The relative uniformity on the left may reflect the incorporation of climate action into progressive identity over several decades, reinforced by elite cues, advocacy networks, and widespread acceptance of climate science in left-leaning constituencies (18, 19, 44, 59). The little heterogeneity among centrists is consistent with their weaker ideological anchors; it could be that they are more influenced by local context and economic concerns. The attitudinal heterogeneity within the right challenges the stereotype that conservatives wholesale reject climate action. Portraying conservatives as homogenously opposed to climate policy is inaccurate and may exacerbate negative partisanship and hinder the formation of policy coalitions. Moreover, while we found scant variance among those on the left, it is possible, if not likely, that in distinct policies and contexts, there may be variance (e.g. Pew Research Center 2017). In such cases, cross-ideology coalitions for climate action may involve subgroups of those from the left and the right (19).

When we examined the role of the four approaches discussed in the introduction (sociodemographic, ideological, expressive, and distrust in science) in explaining variance in right-wing climate attitudes, we found that the expressive approach is most influential. Combined, positive and negative partisanship account for roughly half of the explained variance in conservatives' attitudes toward climate policies. Looking more closely, we find that positive and negative partisanship play somewhat different roles across policies: positive partisanship explains more variance in support for government action and an industrial carbon tax (27 and 28%, respectively), while negative partisanship explains more variance in support for decarbonization (29%). One possible explanation is that support for government intervention and taxation more directly implicates small-government and free-market commitments central to conservative identity, amplifying the role of positive partisan attachment. In contrast, decarbonization policies can be more easily framed as a liberal agenda, heightening the salience of negative partisanship.

We offer several recommendations for those studying and working on climate policy. First, while much attention has been paid to sociodemographic and ideological correlates of climate attitudes, and a growing body of work examines the impact of distrust in science on climate denial and opposition among conservatives, it is crucial to account for expressive measures, such as negative and positive partisanship. We are not the first to call for such a focus, but the magnitude of the explanatory power of positive and negative partisanship in our study strengthens such calls. Second, for purposes of both accuracy and coalition building, it is important to avoid over-reliance on averages when reporting attitudes toward climate policy. As we showed, such averages mask important heterogeneity (particularly among the political right) that could pave the way for the emergence of policy coalitions to address climate change. Characterizing conservatives as anti-environmental could have the unintended effect of exacerbating distrust of science among those conservatives who feel this characterization is inaccurate (50).

Our findings stimulate three promising streams of research. First, many experimental studies testing interventions to increase climate policy support estimate average treatment effects for liberals and conservatives, implicitly treating ideological groups as internally uniform with respect to climate policy responses (60–63 ). That is, they recognize those from different political sides might react differently but pay much less attention to heterogeneity within sides. This approach may be theoretically appropriate when the primary goal is to test whether interventions operate differently across ideological groups. However, it provides limited leverage for understanding the substantial variation in climate policy attitudes observed within the political right. As a result, such designs often leave unexplained why some conservatives respond favorably to climate policy appeals while others do not, even under identical informational or framing conditions. Additionally, there has been a virtual cottage industry of interventions to reduce negative partisanship in particular (64); it would be useful to expand this work to document its downstream effects on policy attitudes (which is rarely done) and consider more tailored interventions aimed at specific ideological sides.

Second, this approach could extend beyond climate change to other polarized policies like immigration. Examining variation within seemingly homogenous groups, rather than relying on averages, may yield both social and scientific advantages. These inquiries could advance our understanding of public opinion and inform more effective strategies for tackling societal challenges.

Finally, given the explanatory power of partisanship in our study, even with a simple measure, we hope that future research will extend our work by employing more refined measures of partisan identity. For instance, the Negative Partisan Identity scale developed by Bankert (33) directly assesses the extent to which individuals define themselves in opposition to an out-party. The partisan sectarianism scale envelops multiple components of partisanship (or ideological) identity, including othering, aversion, and moralization (65, 66). Incorporating these measures alongside feeling thermometers may offer a more comprehensive picture of how partisan identities shape climate policy attitudes.

This article advances understanding of partisan differences in climate policy attitudes in two ways. First, rather than focusing on average differences between the political left and right, we demonstrate that climate policy attitudes are significantly more heterogeneous on the political right than on the left. Second, we identify the primary correlates of that heterogeneity among conservatives. We find that conservatives' feelings toward fellow partisans and toward liberals explains more variance in climate policy attitudes than trust in science, sociodemographic characteristics, or ideology. Approximately, half of the explained variance in right-wing climate policy attitudes is associated with positive and negative partisanship. Taken together, these findings indicate that affective partisan dynamics are strongly related to climate policy attitudes on the political right, with implications for efforts aimed at building cross-partisan support for climate action.

Materials and methods

Data and sample

The organization EKOS Research, in partnership with The Canadian Hub for Applied and Social Research (CHASR) at the University of Saskatchewan, collected responses from 2,503 Canadians. EKOS maintains a survey panel of approximately 90,000 Canadian adults, all of whom were recruited through random-digit dialing of cell phone and landline numbers. From this existing probability panel, EKOS selected respondents for this project through stratified random sampling based on age, sex, and region. Random sampling helps ensure more representative data than online opt-in panels, which can lead to inaccurate estimates (67, 68). EKOS fielded the survey between July 11 to September 9 in both English and French, and the response rate of participants who were contacted from the EKOS panel was 21.4%. Live interviewers contacted participants over the phone and recorded responses. The average length was about 22 min. The survey was pretested to ensure that respondents understood all questions. This project was approved by the research ethics board at the University of British Columbia (approval number H24-00972). Informed consent was obtained from all respondents.

Dependent variables

Three questions gauged respondents' support for climate policy: (i) “When it comes to Canadian governments addressing climate change, do you strongly support, somewhat support, somewhat oppose, strongly oppose, or neither support nor oppose this?” (adopted from the Stanford American Public Opinion on Global Warming survey); (ii) “One proposal to reduce the effects of climate change is to tax companies based on the amount of carbon emissions they produce. Do you strongly support, somewhat support, somewhat oppose, strongly oppose, or neither support nor oppose this?” (adopted from the Pew Research Center) (9); and (iii) “When it comes to replacing fossil fuels like oil, gas, and coal with renewable energies like wind and solar, do you strongly support, somewhat support, somewhat oppose, strongly oppose, or feel neutral about these efforts?” (developed by the authors). Respondents who chose “neither support nor oppose” to any of the three items were asked a follow-up question: “Do you lean more toward supporting it, opposing it, or neither?” Doing so resulted in a 7-point value for each question, including the five responses stated in each question above as well as “lean more toward supporting it” and “lean more toward opposing it” (69).

Key independent variables

Key sociodemographic variables are gender, race, and region. “Gender identity” included woman, man, and nonbinary. “Race/ethnicity” included White, Asian, Arab (including West Asian), Black, Latino, Indigenous, and other or multiracial. “Region” included British Columbia, Prairie provinces, Ontario, Quebec, and Atlantic provinces. “Political ideology” was measured through the item, “Where would you place yourself on a scale from 0 to 10 where 0 means very left wing, or liberal and 10 means very right wing, or conservative? 5 means middle-of-the-road. You can pick any number in between.” We adapted this question from Merkley (45). “Positive partisanship”: “In politics, people sometimes talk of left and right. How do you feel toward Canadians who identify as politically right-wing, or conservative, on the 0-to-100-point scale?” Greater values indicated greater warmth toward the right. “Negative partisanship”: “On the same scale, how do you feel toward Canadians who identify as politically left-wing, or liberal?” We reverse coded this response so that higher values indicated greater coldness. Prior to answering these thermometer questions, respondents were read these instructions: “For the following questions about your feelings toward different people, please use a scale from 0 to 100 where 0 means very cold or unfavorable, 50 means you are neutral, and 100 means very warm or favorable. You can pick any number between 0 and 100.” Using thermometers to measure partisan affect is widely used in work on affective polarization (70). We measured “distrust in science” through the question, “In general, would you say that you trust science a lot, some, not much, or not at all?” We incorporated this question from the Wellcome Global Monitor Survey (57).

Covariates

Five demographic controls that prior research has shown may be related to climate attitudes were included in regression models and are visible in Table S1. Age was measured in years. Sexual identity included straight, mostly straight, bisexual, gay/lesbian, other, and do not know or decline to answer. Education included under a bachelor's and bachelor's or higher. Residential location included urban and rural areas, as determined from the first three digits of respondents' postal codes. Ties to the oil, gas, or coal industry included no ties, current or former work in this industry and family or friends who work in the industry, current or former work in this industry but no family or friends who work in the industry, and no current or former work in this industry but family or friends who work in the industry.

Prior research has shown that each sociodemographic variable we include may be related to climate attitudes, including gender identity (14, 15, 71), race/ethnicity (14, 15), and region (72, 73), as well as our covariates: age (15, 74), sexual identity (75), education (14, 15, 76), urban or rural location (77), and ties to the fossil fuel industry (72). We did not include covariates without a clear relationship to climate attitudes because doing so can bias model estimates (78).

Analysis

Analysis proceeds in four stages, the first of which examines the full sample. We first present bivariate boxplots to show the distribution of support for climate policy attitudes by political ideology. The next three stages are exclusively among the political right in the sample (n = 798). In the second stage, we conduct three linear regressions examining the three attitudinal variables to determine which independent variables are significantly associated with each. Afterwards, we conduct a DA for each model to determine the relative importance of each independent variable. DA “is based on computing the reduction in prediction error that is associated with each IV in a statistical model” (79). DA helps determine the substantive importance of variables included in a model in terms of how they affect model fit. Lastly, we present marginal estimates, calculated from the linear regression models, for the variables for which the DA showed the best improved model fit.

Person weights are applied to all analyses. CHASR constructed this weight so that the sample is roughly representative of the Canadian population as of the 2021 census on the basis of sex, province, age, education, immigration status, and race/ethnicity. CHASR did so with the “pewmethods” package in R (80), which uses rake weighting (80, 81). Weights were trimmed at the 5th and 95th percentiles so that no respondents had an inappropriately outsize effect on the results. However, results were substantively similar regardless of whether the full weight or trimmed weight was used (or whether weights were not used).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Competing Interest

The authors declare no competing interests.

Funding

This research was funded by the Social Sciences and Humanities Research Council of Canada (Insight program; grant number: 435-2024-0171). This research was also supported in part by the UBC Open Access Fund for Humanities and Social Sciences Research, administered by the University of British Columbia Library.

Author Contributions

Emily Huddart (Conceptualization, Data curation, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Project administration, Writing—original draft, Writing—review & editing), Tony Silva (Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Visualization, Writing—original draft, Writing—review & editing), Parker Muzzerall (Methodology, Writing—review & editing), and James N. Druckman (Methodology, Writing—original draft, Writing—review & editing)

Data Availability

The data and code required to reproduce the results reported in this paper are available on the Open Science Framework (OSF). All results needed to evaluate the conclusions are provided in the main text and Supplementary material.