Politics & Global Warming, March 2021

Appendix II: Survey Method

The data in this report are based on a nationally representative survey of 1,037 American adults, aged 18 and older. Results are reported for the subset of 922 registered voters who participated in the survey. The survey was conducted March 18 – 29, 2021. All questionnaires were self-administered by respondents in a web-based environment. The survey took, on average, about 25 minutes to complete.

The sample was drawn from the Ipsos (formerly GfK) KnowledgePanel®, an online panel of members drawn using probability sampling methods. Prospective members are recruited using a combination of random digit dial and address-based sampling techniques that cover virtually all (non-institutional) resident phone numbers and addresses in the United States. Those contacted who would choose to join the panel but do not have access to the Internet are loaned computers and given Internet access so they may participate.

The sample therefore includes a representative cross-section of American adults – irrespective of whether they have Internet access, use only a cell phone, etc. Key demographic variables were weighted, post survey, to match US Census Bureau norms.

From November 2008 to December 2018, no KnowledgePanel® member participated in more than one Climate Change in the American Mind (CCAM) survey. Beginning with the April 2019 survey, panel members who have participated in CCAM surveys in the past, excluding the most recent two surveys, may be randomly selected for participation. In the current survey, 316 respondents, 272 of whom are registered voters included in this report, participated in a previous CCAM survey.

The survey instrument was designed by Anthony Leiserowitz, Seth Rosenthal, Jennifer Carman, Matthew Goldberg, Karine Lacroix, and Jennifer Marlon of Yale University, and Edward Maibach and John Kotcher of George Mason University.


Margins of error

All samples are subject to some degree of sampling error – that is, statistical results obtained from a sample can be expected to differ somewhat from results that would be obtained if every member of the target population was interviewed. Average margins of error, at the 95% confidence level, are as follows:

  • Total registered voters (n = 922): Plus or minus 3 percentage points.
    • Democrats (total; n = 403): Plus or minus 5 points.
      • Liberal Democrats (n = 213): Plus or minus 7 points.
      • Moderate/conservative Democrats (n = 188): Plus or minus 7 points.
    • Independents (n = 83): Plus or minus 11 points.
    • Republicans (total; n = 397): Plus or minus 5 points.
      • Liberal/moderate Republicans (n = 129): Plus or minus 9 points.
      • Conservative Republicans (n = 266): Plus or minus 6 points.


Rounding error and tabulation

In data tables, bases specified are unweighted, but percentages are weighted to match national population parameters.

For tabulation purposes, percentage points are rounded to the nearest whole number. As a result, percentages in a given chart may total slightly higher or lower than 100%. Summed response categories (e.g., “strongly support” + “somewhat support”) are rounded after sums are calculated (e.g., 25.3% + 25.3% = 50.6%, which, after rounding, would be reported as 25% + 25% = 51%).