World Attitudes

World attitudes on climate change

In this research line we aim at developing an international, harmonized, and longitudinal comparative database on climate change attitudes. Existing large-scale studies have covered around 120–140 countries and provided valuable insights (Andre et al., 2024; Goldberg et al., 2021; Lee et al., 2015), yet they exhibit substantial gaps in geographic coverage, temporal continuity, and item comparability. Our objective is to overcome these limitations by integrating, expanding, and harmonizing heterogeneous survey sources into a single, coherent data infrastructure.

The following map displays initial evidence of an IRT model that combines dozens of harmonized survey items from major international and regional surveys into a single, comparable score of climate concern for each country. Because the underlying surveys use different questions, response scales, and samples, this Bayesian Item Response Theory (IRT) model treats each item as an imperfect signal of one latent trait — public concern about climate change — and pools information across countries and items to produce a score with an associated 95% credible interval, even where survey coverage is uneven. See Methodology below for details.

Methodology

Because different surveys ask different questions with different response scales, raw survey responses cannot simply be averaged together. The dataset instead applies a Bayesian ordinal Item Response Theory (IRT) model with partial pooling, which treats each survey item as an imperfect indicator of a single underlying trait: public support for climate action. Partial pooling lets the model borrow strength across countries and items, producing a comparable, uncertainty-aware climate support score for every country — even where direct survey coverage is sparse. Hover a country on the map for its score and 95% credible interval.

This first version shows a single cross-sectional snapshot pooling the survey waves fielded so far; a longitudinal view (tracking each country’s attitudes over time) is planned for a future update once enough repeated country-year coverage is available.

This map is based on an internal ATTCLIMPOLS working paper by Jordi Mas (UOC) and Marc Guinjoan (UAB), “Expanding global coverage of climate change support over space and time”, and on a preliminary run of the paper’s Bayesian ordinal IRT model, covering 165 countries. Coverage and estimates will be refined as further survey waves are incorporated and the model is finalised. The paper is a working paper and is not yet available for public distribution.