Building Supranational Support: A Cross-Country Survey Experiment on EU Narratives

2025-11-16

2025-12-31

The key outcomes are all post-treatment measures collected in the harmonized cross-country survey following exposure to the informational videos.
They are grouped into seven pre-specified families of conceptually related variables:

Behavioral participation – indicator equal to 1 if the respondent voluntarily provides an email to be recontacted to share ideas and comments about the European Union (opt-in).

Emotions and message agreement – perceived neutrality and agreement with the video (1–5), and sentiment or emotional tone extracted from the open-ended recall question.

Issue salience – perceived importance (1–5) of major global issues: military conflicts, environment, economic growth, inequalities, migration, and trade wars.

Need for EU (with certainty) – desired degree of EU involvement (1–5) and self-reported certainty (1–5) for each issue.

Trust and political efficacy – trust in EU institutions, national institutions, parties, unions, and media (1–5); and perceived political efficacy (“my voice counts” in the EU and in Italy).

Policy support – agreement (1–5) with EU-level reforms and cooperation policies (common army, common migration policy, EU fiscal capacity, continued support for Ukraine, social and labor policies, import tariffs).

Identity – feeling of belonging as a local, national, and European citizen (1–5).

Within each outcome family, related survey items will be combined into standardized indices to summarize overall effects and to limit multiple-testing concerns. Individual items will also be analyzed separately for transparency.

This project is a cross-country randomized survey experiment designed to study how informational narratives about the European Union influence attitudes, beliefs, and engagement toward supranational governance. The study extends a previous Italian RCT (Programmatic Advertising and Electoral Participation, 2024) to a multi-country setting, focusing exclusively on survey-based treatments without programmatic ad delivery.

The experiment will be conducted in five EU countries: Italy, France, Germany, Poland, and Spain. A total of approximately 15,000 adult respondents (about 3,000 per country) will be recruited through the professional survey provider SWG S.p.A., using its proprietary CAWI platform. Within each country, participants will be stratified by gender, age group, and region and then randomly assigned at the individual level to one of four experimental arms:

T1 – External Threats (Defense Narrative): emphasizes external threats and the need for unity and protection.

T2 – Common Challenges (Cooperation Narrative): stresses shared challenges and the importance of joint European solutions.

T3 – Western Threats (Trump/Commercial Narrative): focuses on Western or commercial threats linked to global competition and Trump-era dynamics.

C – Control group: no video or neutral non-political content.

After exposure, all respondents complete a harmonized post-treatment questionnaire capturing trust, perceived EU effectiveness, identity, policy preferences, issue salience, and behavioral engagement. The control group receives the same questionnaire without video-related items. Data collection will take approximately 25 working days per country, with all survey instruments and videos fully localized by language.

Randomization was done by computer within the online survey platform (CAWI) managed by SWG S.p.A., using automated assignment algorithms with pre-specified random seeds and stratification by gender, age group, and region.

Individual-level randomization within countries, stratified by gender, age group, and region. Each respondent is independently assigned to one of four treatment arms within the CAWI survey platform (SWG S.p.A.).

No

15,000 individual respondents in total (approximately 3,000 per country: Italy, France, Germany, Poland, and Spain). Randomization is individual, not clustered, with participants evenly allocated across four treatment arms (T1, T2, T3, and Control).

15,000 individual survey respondents in total (approximately 3,000 per country: Italy, France, Germany, Poland, and Spain). Each observation corresponds to one individual who completes the online survey and is randomly assigned to one of four experimental arms.

Approximately 3,750 individual respondents per treatment arm:
– 3,750 Control (no video / neutral content)
– 3,750 T1 – External Threats (Defense Narrative)
– 3,750 T2 – Common Challenges (Cooperation Narrative)
– 3,750 T3 – Western Threats (Trump/Commercial Narrative)

Totals approximately 15,000 individuals across all five countries (≈3,000 per country), with balanced assignment within each country.

Assumptions: two-sided α=0.05, power=0.80, 4 arms with equal allocation; individual-level randomization (not clustered); stratification by gender/age/region; no design effect from clustering (DEFF≈1). If post-stratification weights introduce mild variance inflation, MDEs scale by √DEFF (e.g., DEFF=1.10 ⇒ ×1.05). Pooled (all countries combined; ~3,750 per arm): Standardized family indices (SD units): ≈ 0.06–0.07 SD (one treatment vs control). Binary “voluntary participation” (email opt-in): baseline 𝑝 0 ∈ [ 0.10 , 0.20 ] p 0 ​ ∈[0.10,0.20] ⇒ SD≈0.30–0.40. MDE ≈ 2–3 percentage points. Per-country estimates (~750 per arm per country): Standardized family indices: ≈ 0.14–0.15 SD. Binary opt-in: ≈ 5–6 percentage points when 𝑝 0 ∈ [ 0.10 , 0.20 ] p 0 ​ ∈[0.10,0.20]. Notes: MDEs are for a single treatment vs control contrast under ANCOVA with modest baseline covariate gains. Family-wise multiple testing is handled via pre-specified indices; item-level MDEs will be larger. If weighting inflates variance (e.g., DEFF=1.10), multiply the figures above by ≈1.05.