To create a detailed picture of the state of pluralistic ignorance for climate policy, we use a large stratified sample of US adults (N = 6,119) through the Ipsos eNation Omnibus nationally representative panel to compare public opinion on climate change to perceptions of popularity of those same opinions. We commissioned this panel to oversample less-populous states to assess the extent of pluralistic ignorance for each state with greater precision, and aiming for a 10% margin of error for all states. For the full sample, this sample size is more than 80% powered to detect small national levels of pluralistic ignorance (effects as small as d = 0.04), as well as being 80% powered to detect separate levels among Democrat, Republican, and Independent partisan groups (effects as small as d = 0.07), allowing for very granular comparisons. For all national-level analyses, we applied weights from the survey provider to ensure representativeness (e.g., down-weighting data from smaller states that we oversampled).
Actual levels of U.S. public support on climate policies were obtained from nationally representative public opinion data available from Yale’s Program on Climate Change Communication (YPCCC), who used the same survey provider who collected the data for the present study (which may help account for any surveyor-specific sampling or data practices)11,12. We also use YPCCC’s state level estimates of public opinion (which are estimated from their national survey data via multilevel regression with post-stratification). These data were collected during the same year as the data in the present study (see Methods). Further, to avoid any differences in policy support estimates due to item wording in comparing actual levels to responses from our panel, we used precisely the same policy descriptions as used by YPCCC. Thus, if wording for any of the policy items is subtly leading to inflated (or deflated) support numbers, this wording should also lead to inflated (or deflated) estimates of opinion estimates; thus, specific item wording would be unlikely to create discrepancies between actual and perceived policy support.
We asked participants to estimate the percent of Americans who were at least somewhat concerned about climate change (see Methods for full survey text, and a discussion about using the phrase “climate change” vs “global warming” for this item). We then chose a set of specific climate policies especially relevant to the decarbonization of the US and the attainment of climate mitigation goals such as the 2015 Paris Agreement. We intentionally selected a set of climate change mitigation policies that varied in core features such as utilizing market instruments as opposed to mandates, or those that facilitate investment and the creation of infrastructure. For each policy, we asked participants to estimate the percent of Americans who would support it. Our list of policies included support for a carbon tax levied against fossil fuel companies and redistributed to Americans through tax breaks. The list also included a renewable energy standard that mandates 100% electricity generated by renewable energy in the near term—an essential step in decarbonizing our energy production34. And, as decarbonizing our energy infrastructure will require rapidly siting of wind and solar across the US, we also included support for siting renewables on public lands34. Given the need to consider infrastructure, jobs, and social equity in transitioning to renewable energy, we also asked participants to estimate the support for the Green New Deal (GND). Notably, large environmental policy packages like the GND and the American Jobs Plan may play a key role in passing environmental legislation, as research shows that bundling more redistributive, social equity, and job-creating measures into major environmental policies makes them more popular35.
Finally, we asked participants about their news consumption, political affiliation, and demographic characteristics, which we used to conduct an exploratory cross-sectional analysis of possible sources of pluralistic ignorance in norm perceptions.
Prevalence and magnitude of pluralistic ignorance
Figure 1 shows that the vast majority of Americans greatly underestimate how many of their fellow Americans worry about climate change and support transformative climate policies to remedy the situation. While most Americans believe that less than half of the country is worried about climate change (Mest = 43.3), in actuality it is two-thirds (Mreal = 66), t(6118) = 70.9, d = 0.92, P < 0.001, 95% CIdiff = [22.0, 23.3] (see Methods for additional notes on these analyses). Americans’ estimates for major climate policy support is the same or even lower (Ms = 37–43%), when in fact two-thirds of the country or more support each of these policies (Table 1). The distributions of these estimates in Fig. 1 show two noticeable concentrations, one at around 50% and another around 25%, salient proportions that seem to serve as focal points for answering these questions, even though a similarly salient proportion—75%—would have been a far more accurate answer. The misperceptions in estimates are so robust that, for every item assessed, the estimates of the lowest 25% and of the middle 50% of respondents falls well below the true values. More precisely, between 79% and 88% of our national sample underestimate public concern or each policy support.
We also asked participants for estimates of support in their home state, and found these perceptions (when averaged across states) to not vary substantially from the national-level estimates and to have very similar distributions (see Supplementary Fig. 1 in the Supplementary Information File). Overall, this pattern of results suggests that people misperceive support for climate action broadly, having non-specific and robust misestimates for support for a variety of climate policies. Indeed, in an exploratory factor analysis of the five responses shown in Fig. 1, a single factor emerged (all other factors had eigenvalues <1).
Pluralistic ignorance across partisans and policies
Breaking these perceptions of national public concern and policy support down by partisan politics, we found that Democrats, Independents, and Republicans all estimate levels for climate concern and climate policy support below 50%, while actual values are much higher. However, Fig. 2 shows that Republicans’ estimates were reliably lower than Democrats’ by 5–12%, with Independents falling somewhere in between.
For the national policy items, contemporaneous polling was available broken down by partisans, so we can compare partisans’ estimates of nationwide support to actual partisan levels of support. Figure 2 shows that when Democrats, Independents, and Republicans estimate how the nation feels on these issues, their estimates of other Americans’ support for these policies only really resemble actual Republican levels of policy support. In fact, even if individuals’ estimates for the nation as a whole were, for some reason, based solely on Republican levels of support, all partisan groups would still be underestimating support for policies like a carbon tax and siting renewables on public lands. While differences between partisans are consistent with false consensus effects (e.g., Democrats—who are more likely to personally support climate policy—tend to provide relatively higher estimates of others’ policy support than do Republicans), these effects are dwarfed by the absolute levels of misperception held by all Americans that strongly underestimates climate policy support.
For all policies and concern for climate change at the national level, as well as state level support for a carbon tax and concern for climate change, we were able to compare each participant’s norm estimates to available contemporaneous data on actual support levels. Doing so, we create difference scores for each participant’s estimate, which can be aggregated into pluralistic ignorance levels that we analyze below.
Reflecting the lower norm estimates by Republicans, Fig. 3 shows that Republicans’ opinion misperceptions are stronger in magnitude than Democrats’ and Independents’ across all items. Further, we find that all partisan groups underestimate concern for climate change at both the national and state level by roughly 20–30%. In policy support, we find that the magnitude of misperception is highest for support to site renewables on public lands, with underestimates closer to 35–40%. Underestimation is smaller for support for 100-percent renewable energy mandates, which is still between 20–25% lower than actual levels. Support for a carbon tax and a Green New Deal fall in between these levels.
We can also directly compare state and national pluralistic ignorance levels for the two items for which we have data for both (a carbon tax and worry about climate change) to test if estimates are more accurate for state than national items. Using a mixed model to predict pluralistic ignorance levels across these four items using a dummy-coded fixed effect for item location (state = 0 vs nation = 1) and random intercepts for participant and item type (carbon tax vs worry), we find no difference between the two, t(21762) = −0.94, P = 0.350, suggesting that people are equally inaccurate at estimating opinions of fellow denizens of their own state, relative to the entire US public.
Aggregating levels of pluralistic ignorance by state, we can map the magnitude of false norm perceptions across the country. Figure 4 shows that residents of all states underestimate how much the nation is worried about climate change and support climate policy (for separate maps for each policy, see Supplemental Fig. 2 and Supplemental Fig. 3). For both perceived popular worry and climate policy support, we see that the southern gulf states (e.g., Mississippi) tend to show the highest pluralistic ignorance. But, underscoring the ubiquity of this misperception, even liberal states such as California and New York underestimate climate policy support as much as many conservative states. In fact, no state was less than 20% off in their estimates of climate policy support. These errors are robust for the more proximal state-level estimates as well, where participants of virtually every state underestimated how concerned their fellow state residents were about climate change, and how much they supported a carbon tax (see Supplemental Fig. 4).
Variation by demographics
We regressed pluralistic ignorance across items in a linear mixed model weighted to be nationally representative, with random intercept for participant and item, on the full battery of demographics assessed (see Methods). As shown in Supplementary Table 2, we find a number of statistically significant factors. Consistent with false consensus effects, we find participant’s political orientation has a notable effect (with 22% underestimation for those who are very liberal to 33% for those who are very conservative). Race also has a notable impact, with 25% underestimation for white respondents to 35% underestimation for black respondents, and other races falling in between. Other demographic characteristics had smaller, but still statistically significant effects. For instance, those living in urban areas were about 29% off, while their suburban counterparts were 26% off (and rural respondents falling in between).
Notably, there was no demographic group for which the estimated range reached accurate levels—instead all groups assessed were at least 20% off. Further, some demographics which might have been anticipated to predict reduced misperceptions did have statistically significant effects, but were small shifts in absolute terms: Those who attended 12 years of schooling but never obtained a GED or diploma were 28% off, while those with a doctorate were still 27% off, just a single percentage point better.
Variation by local norms
In exploratory analyses, we assessed two state level predictors for their relationship to pluralistic ignorance levels across items with both known real and perceived levels: the voting margin for Biden in the 2020 election (used as a state-level proxy for prevalent political ideology), and the logged number of climate or environmental protests per capita (see Methods). The effects of these predictors were assessed in a multiple regression mixed model with random intercepts for participant and item, and controlling for the top five demographic variables shown to have an effect and likely vary by state (personal political orientation, race, employment status, age, and income). Consistent with an availability heuristic, we find that both indicators of local norms influence norm estimates: there is a significant effect for state political ideology b = −0.02, t(39760) = 2.35, P = 0.019, such that states with the highest margin for Biden had pluralistic ignorance levels of 25.5%, while those with highest margin for Trump were 28.1% off. We also find a significant effect for state-level environmental protests b = −0.47, t(39980) = 2.32, P = 0.020, such that states with the highest level of protest were estimated to have pluralistic ignorance levels of 25.9%, while those with the fewest protests were 28.3% off.
Variation by media consumption
In exploratory analyses, we assessed the relationship between news media consumption and pluralistic ignorance levels across items. Using a dummy-coded variable to compare consumers of each news outlet (those who view it at least weekly) to those who do not, we assess the effect of media consumption for each outlet in a multiple regression mixed model with random intercepts for participant and item. And as media consumption may vary based on demographics, we control for the full battery of demographics assessed here, including personal political orientation, education, age, race, and income. We find that consumers of all news media outlets underestimate climate concern and policy support by around 25–30% (see Fig. 5, Panel A). Contrasting the differences between viewers and non-viewers of each outlet (see Fig. 5, Panel B), we see relatively lower levels for consumers of public broadcasting (National Public Radio), and mainstream news outlets including major national papers (e.g., New York Times), major cable news outlets (e.g., CNN), national broadcast news networks (e.g., ABC). We see relatively higher levels for those who consume news from major conservative outlets like Fox News and other conservative outlets (e.g., Breitbart), as well as for other liberal outlets (e.g., The Nation).
This pattern generally suggests that media exposure to outlets with less favorable coverage of climate change policy correspond with lower estimates of public support, with the increase in pluralistic ignorance for those consuming “other liberal outlets” as the exception. Notably, “other liberal outlets” was also the least consumed outlet, with only 15.5% regularly consuming it (95% CI = [14.6, 16.4]), while all other outlets were consumed by 20–63% of Americans. One possibility is that consumers of these liberal outlets recognize their news source both liberal and niche, and therefore presume others do not share their more liberal, pro-climate attitudes.