Public opinion data come from the Yale Climate Opinion Maps (YCOM), which are based on a statistical model that employs nationally representative Climate Change in the American Mind (CCAM) surveys conducted between 2008 and 2024. The model combines geographic, census, socioeconomic, and political data with CCAM survey data collected by the Yale Program on Climate Change Communication and George Mason University Center for Climate Change Communication (combined n > 35,000). For more information about the survey question wording and methodology, please visit YCOM
This page provides estimates of U.S. climate change beliefs, risk perceptions, and policy preferences at the state and local levels – a new source of high-resolution data on public opinion that can inform national, state, and local decision-making, policy, and education initiatives. The estimates are derived from a statistical model using multilevel regression with post-stratification (MRP) on a large national survey dataset, along with demographic and geographic population characteristics.
The modeling employs data from 31 nationally representative surveys of American adults (18+) conducted from November 2008 through December 2024 with a combined sample size of n>35,000. The surveys were administered by Ipsos and drawn from the Ipsos KnowledgePanel®, an online panel of members drawn using probability sampling methods. All questionnaires were self-administered by respondents in a web-based environment, and computers were loaned to individuals who were chosen to participate but did not have access. Respondents came from all 50 states and the District of Columbia, and 2,379 of 3,144 counties. Sample weights for all respondents were calculated by Ipsos to be nationally representative post-survey to match U.S. Census Bureau norms. For respondents who have taken the survey multiple times, only their most recent response was kept in the data, and all previous responses were removed. This resulted in 3,108 responses being removed.
Multilevel modeling with poststratification (MrP) was used to estimate the spatial distribution of climate opinions at state, county, congressional district, and metro area levels (Howe et al. 2015; Mildenberger et al. 2016, 2017). Dependent variables were first recoded into binary format, with positive response values grouped and coded as “1” (e.g., “Somewhat favor” and “Strongly favor”) and non-positive values coded as “0” (“Somewhat oppose”, “Strongly oppose”, “Don’t know”, Refused); see “Survey Question Wording” on the interactive map tool tab for a full list of questions and how each was recoded to binary). MrP modeling then proceeded in two phases. First, the multi-level model is constructed by predicting individual survey responses as a function of both individual-level demographics (gender, race, education, and a three-way interaction term among these variables) and geography-level covariates. Second, “post-stratification”, or spatial weighting, is performed using the fitted model, where population-weighted opinion estimates for each demographic-geographic subtype are aggregated based on the subtype population distribution within each geographic subunit.
The YCOM model uses four geography-level covariates: percent of people who drive alone to work, percent of same-sex households, percent of CO2 emissions per capita, and percent of people who voted Democrat in the most recent election. In order to ensure accurate and current estimates, two of the data sources used for these covariates in the current YCOM model (version 8) have been updated to reflect more current data.
The U.S. Census Bureau’s American Community Survey (ACS) variable which was previously the source of same-sex household data was discontinued in 2018. Therefore, we replaced this covariate with the ACS variable “Coupled Households by Type”, which includes an almost identical survey question most recently asked in 2020. CO2 data from the Vulcan Project, originally published in 2010, were replaced with updated 2024 estimates of CO2 emissions and supplemented with new data from Crosswalk Labs, which provides estimates of CO2 emissions at the census tract level. Presidential vote share was acquired from the Redistricting Hub for the 2020 election, as the vote share data for 2024 is not yet available. All new sources of data were compared with their outdated counterparts and found to be highly correlated.
Validating models is essential for producing accurate results. Our original YCOM model estimates were validated using three different methods. First, cross-validation analyses were conducted within the dataset. The dataset was divided into two sets of respondents, with one part used to run the model and the other kept aside for validation. The model estimates were then compared to the results of the set-aside respondents to directly quantify the percentage of correct answers the model predicted. These cross-validation tests were repeated multiple times using different sample sizes and dividing the data in different ways. Second, the model estimates derived from the full dataset were compared to the results of independent, representative state- and city-level surveys conducted in California, Colorado, Ohio, Texas, San Francisco, and Columbus, Ohio in 2013. The mean absolute difference between model estimates and validation survey results was 2.9 percentage points (SD = 1.5) among the four states (CA, TX, OH, CO) and 3.6 percentage points (SD = 2.9) among the two metropolitan areas (Columbus, OH, and San Francisco, CA), well within the margins of error for the survey results alone (at a 95% confidence level). Estimates have also been validated internally through a series of technical simulations. Third, some model estimates were compared with third-party survey data collected by other researchers in previous years.
Our current model estimates were validated by comparing modeled estimates with weighted survey averages at the national level and for the five most populous states. The mean absolute error between modeled estimates and weighted survey averages across all variables was 0.51 percentage points at the national level and 3.63 percentage points at the state level.
For the 2024 model estimates, uncertainty ranges are based on 95% confidence intervals using 99 bootstrap simulations. These confidence intervals indicate that the model is accurate to approximately ±7 percentage points at the state level, and ±8 percentage points at the county level. Such error ranges include the error inherent in the original national surveys themselves, which is typically ±3 percentage points.
Marlon, J. R., Wang, X., Bergquist, P., Howe, P., Leiserowitz, A., Maibach, E., Mildenberger, M., and Rosenthal, S. “Change in US state-level public opinion about climate change: 2008–2020.” Environmental Research Letters 17, no. 12 (2022). 124046.
Howe, P., Mildenberger, M., Marlon, J., & Leiserowitz, A. (2015) “Geographic variation in opinions on climate change at state and local scales in the USA,” Nature Climate Change 5. DOI: 10.1038/nclimate2583.