A registered report megastudy on the persuasiveness of the most-cited climate messages

Summary

This registered-report megastudy tests the ten most-cited climate messaging strategies against a common control on a large US national sample (N=13,544), preceded by five pilot replications (N=3,216) of three canonical strategies. Most messages produce small but real shifts on pro-environmental attitudes, concern, policy support, and political behavioural intentions (typically 1–4 percentage points), and these effects are remarkably similar across partisans — undercutting claims that Democrats and Republicans require distinct framings. Critically, no message reliably increased incentivized pro-environmental donations, and the most effective attitude-shifter (Scientific Consensus 2) actually reduced donations. The authors argue that short-form messages can move stated attitudes but have clear limits for inducing costly behaviour, and that the field needs commensurate, well-powered comparisons rather than further isolated demonstrations.

Key Contributions

• First systematic, head-to-head test of the 10 most-cited climate messaging strategies under identical measures and outcomes.
• Updates the temporal validity of canonical climate communication findings via preregistered replication-and-extension, with original authors permitted to revise treatments.
• Demonstrates the megastudy design as a corrective for underpowered, non-commensurate prior work in climate communication.
• Challenges the dominant assumption that audience-targeted (e.g., Republican-specific) framings are necessary.
• Documents a robust attitude–behaviour divergence, including a scientific-consensus backfire on donations.
• Releases a curated database of 157 climate messaging papers with citation-based selection metrics.

Methods

The authors identified 157 candidate papers via meta-analyses, reviews, expert input, and targeted searches, then ranked them by three citation metrics to select 10 strategies. After five pilot replications, the main Stage 1-accepted megastudy randomly assigned a demographically matched US sample to one of 10 treatment arms or a pooled control. Four preregistered primary outcomes (belief, concern, policy support, political behavioural intentions) plus an incentivized donation behaviour were measured pre- and post-treatment. Analyses used OLS with pre-treatment controls, partisan-subsample tests, Bayesian comparisons, and exploratory moderation/mediation. Power exceeded 99% for full-sample detection of d ≥ 0.10.

Findings

• 6/10 messages increased belief in climate change; Scientific Consensus 2 was largest (~3.12 pp, ~14% of the partisan gap).
• 7/10 increased concern (largest: Purity Framing, 2.47 pp); Warmth Framing produced a small backfire (-0.76 pp).
• 6/10 increased general policy support; 8/10 increased political behavioural intentions (largest: System Preservation Framing, 3.67 pp).
• No treatment increased donations; Scientific Consensus 2 backfired (-3.95 pp), and Gains Framing reduced Republican donations.
• Partisan heterogeneity was minimal: only 12/90 partisan interaction tests were significant.
• Attitude/intention effects were strongly intercorrelated but uncorrelated (or negatively related) with donation effects — a clear attitude–behaviour gap.
• Pre-treatment concern was the most consistent moderator (26/90 interactions significant); already-concerned respondents moved more.
• Effective messages shifted multiple mediators simultaneously (patriotism, threat to American way of life, disgust, personal relevance), obscuring single mechanisms.
• Megastudy effects exceeded pilot effects; extent of author revision correlated r=0.29 with effect size.

Connections

This paper is the methodological and empirical anchor for the current wave of large-scale climate messaging tests, and naturally pairs with Gauthier2026-iq, Szabo2026-rd, Dubey2026-bl, and Allcott2025-jb as related entries in the megastudy / commensurate-comparison tradition for political and environmental persuasion. Its documented attitude–behaviour gap and minimal partisan heterogeneity are especially relevant points of comparison for any of these works that report differential framing effects or behavioural outcomes.

Podcast

A research-radio episode discusses this paper: Listen