Skip to main content
Dr. Michael SchymuraEconomics of AI & Innovation
Europe, Measured

Methods / Evidence contract

What a survey can tell us. And what it cannot.

The European Social Survey is a repeated cross-section of probability samples. It is designed for structural comparison, not real-time polling, and it does not follow the same individuals from one round to the next.

The pinned baseline

The programme baseline is ESS Round 11, integrated edition 4.2, released on 2 July 2026: 50,116 respondents from 30 countries. The inspected integrated file records interviews from 11 March 2023 to 1 December 2024. The wider study documentation includes a separate Czech self-completion release and an earlier start date; these are not the coverage of this integrated file.

European Social Survey ERIC. ESS11, edition 4.2. DOI: 10.21338/ess11e04_2. Data use terms: CC BY-NC-SA 4.0. Earlier releases retain their own noncommercial conditions, citation and documentary record. Raw respondents are not redistributed.

Weights and uncertainty

Primary estimates use the ESS analysis weight, anweight, and available cluster and stratum variables. Earlier rounds require the documented construction of the analysis weight and, where necessary, a separate Sample Design Data File. An unweighted mean or a naive model-based interval does not pass the publication gate.

Population-weighted estimates and equal-country estimates answer different questions. The former gives countries weight according to their represented population; the latter gives each participating country the same influence. Every pooled result identifies which rule it uses.

Change requires a stable comparison

A trend is checked using all available country-round observations, a fixed set of countries observed in every compared round, and country-specific trajectories. A changing set of participating countries must not become an unmarked story about changing European attitudes.

Question wording, valid responses, missing-value labels and eligible populations are reviewed concept by concept. Similar variable names are not proof of comparability. Age profiles are not individual aging paths, and adjusted associations are not automatically causal effects.

Survey mode is part of the evidence

Czechia's Round 11 self-completion file is separate and is not included in ESS-01. The integrated recent releases contain face-to-face, video and unavailable mode codes; the face-to-face sensitivity retains only explicitly identified face-to-face interviews. This comparison does not identify a causal mode effect. Round 12 remains a future integration target until its exact release, coverage and questionnaire documentation have been reviewed.

Disclosure and missingness

Respondent records remain private and are never deployed. Public cells with fewer than 50 respondents are suppressed; cells with 50 to 99 are flagged for low precision. Weighted and unweighted missingness remain visible, and special missing codes are decoded from metadata rather than treated as substantive scores.

Every finding has a receipt

Published figures link to aggregate values, sample sizes, confidence intervals and an analytical receipt. Source editions and hashes, transformations, sample rules and calculation versions are retained. Data revisions, analytical corrections and editorial changes receive distinct version labels. Raw data and private repository metadata do not enter public downloads.

Official sources

What this evidence does not prove

The ESS can reveal a distribution, a difference or an association under stated assumptions. It cannot by itself establish why an institution is trusted, identify a policy's causal effect, or measure how the same person changed. Those limits apply to the headline as well as to the methods.

Image credits

Photographs provide editorial context, not evidence of survey responses. Local copies are cropped and resized; each image retains its source license.