Abstract

In partnership with the Yale Program on Climate Change Communication, Facebook launched a Climate Change Opinion Survey that explores public climate change knowledge, attitudes, policy preferences, and behaviors across 31 countries and territories. Aggregated data is available publicly on Humanitarian Data Exchange (HDX). De-identified microdata is also available to nonprofits and universities under a data license agreement through Facebook’s Data for Good (DFG) program. For more information please email dataforgood@fb.com.

Analysis unit

Public Aggregate Data on HDX: country or regional levels De-identified Microdata through Facebook Data for Good program: Individual level

Universe

The survey was fielded to active Facebook users ages 18+

Kind of data

Sample survey data [ssd]

Sampling procedure

Sampled Facebook users saw an invitation to answer a short survey at the top of their Facebook Newsfeed and had the option to click the invitation to complete the survey on the Facebook platform. The sample was drawn from the population of Facebook monthly active users, defined as registered and logged-in Facebook users who had visited Facebook through the website or a mobile device in the last 30 days.

Within each country or territory surveyed, Facebook drew a sample in proportion to publicly available age and gender benchmarks. The sample population in the United States was drawn in proportion to the U.S. Census Bureau Current Population Survey 2018 March Supplement. All other countries and territories were sampled in proportion to data from the United Nations Population Division 2019 World Population Projections. Data were weighted separately for each country and territory using a multi-stage, pre- and post-survey weighting process based on census and nationally representative survey benchmarks, Facebook demographics, and Facebook engagement metrics, balanced to the total number of survey completions.

Mode of data collection

Internet [int]

Research instrument

The survey includes questions about people’s climate change knowledge, attitudes, policy preferences, and behaviors. The codebook with survey questions is available here.

Response rate

Response rates to online surveys vary widely depending on a number of factors including survey length, region, strength of the relationship with invitees, incentive mechanisms, invite copy, interest of respondents in the topic and survey design. Facebook provides survey weights to help make the sample more representative of each country or territory’s population.

Sampling error estimates

Any survey data is prone to several forms of error and biases that need to be considered to understand how closely the results reflect the intended population. In particular, the following components of the total survey error are noteworthy:

Sampling error is a natural characteristic of every survey based on samples and reflects the uncertainty in any survey result that is attributable to the fact that not the whole population is surveyed.

Other factors beyond sampling error that contribute to such potential differences are frame or coverage error and nonresponse error.