Consistent effects of science and scientist characteristics on public trust across political regimes

The study investigates how characteristics of scientists (gender, nationality, public activism) and characteristics of science (discipline, societal impact) shape public trust, and whether these effects differ across political regimes. Motivated by debates about whether democracies foster higher trust in science, the authors note that cross-national surveys often treat “science” as a monolithic concept and ignore potential differences in how people conceive science and scientists across contexts. The purpose is to causally identify how specific attributes of scientists and research influence multiple trust dimensions, and to compare patterns across autocratic and democratic contexts. The study emphasizes the importance of public trust for science’s role in evidence-informed policy and highlights increasing politicization of scientific issues in some democracies as a critical backdrop.

Theoretical background distinguishes science as both method and social institution, involving researchers, disciplines, and supporting structures. Public understandings of science vary with educational systems, media, political interests, and cultural contexts. Stereotypes and biases in perceptions of scientists (e.g., gendered competence stereotypes) and disciplines (hard vs soft sciences) are well documented. National identity can shape in-group favoritism for co-nationals, and activism by scientists may affect perceived credibility depending on political polarization and issue salience. Political regimes may shape science education, autonomy, and public discourse, potentially affecting trust. The authors derive hypotheses: higher trust in genetics and materials science than in economics and education (H1), lower trust for female scientists (H2), lower trust when scientists engage in activism (H3), higher trust for high-impact research (H4), and higher trust when scientist is a co-national (H5). They also review expectations for covariates: education and science media consumption generally associate with higher trust, while political orientation predicts trust gaps primarily in democracies.

Design: Preregistered cross-national vignette survey experiment with additional pre-treatment covariates. OSF preregistration: https://osf.io/f4y7v. Sample: Online panel in 10 countries (United States, Germany, Poland, Turkey, China, India, Mexico, South Africa, Singapore, Brazil). Adults aged 18–55. Quota sampling on gender and age groups (equal male/female across four age bands). Fieldwork July–November 2023. Surveys in seven languages (English, German, Polish, Spanish, Portuguese, Turkish, Standard Chinese) via Qualtrics; recruitment by Bilendi. Initial N=13,407; exclusions for non-consent, duplicates, quota/geography, speeding (<3 min), failed attention/commitment checks, incomplete responses (<4 vignettes) or >90% missing. Final N=8,441 (min 790 India; max 875 Turkey; avg ~844 per country). Experimental stimuli: Each respondent evaluated four vignettes (one per discipline): genetics, materials science, economics, education research. Five dimensions: (1) discipline (fixed across four vignettes: each discipline shown once), (2) scientist gender (male/female), (3) scientist nationality (co-national vs non-national; outside the U.S. the foreign option was U.S. scientist; in the U.S. the foreign option was Chinese scientist), (4) scientist public activism (yes/no), and (5) research public impact (high: 20–25% affected vs low: 0.01–0.05% affected). Non-discipline dimensions randomized per vignette; 64 total unique vignette combinations across sets. Outcomes (per vignette; 1–4 scale: definitely not to definitely yes): Trust in scientist—competent (expertise), honest (integrity), responsible (benevolence). Trust in science—research trustworthy, should be publicly funded, should influence policy. Scientist trust items adapted from METI (Hendriks et al., 2015) selecting three high-loading items. Pre-treatment covariates: Demographics and science-related measures. Key covariates: Education (4-point: <HS, HS/further, college/university, doctorate); Political orientation (11-point left–right; reversed in China then recoded); Science media consumption (aggregate of frequency across 5 media types on 1–4 scale). Also included a general trust in science outcome (5-point Likert) for covariate analyses. Regime measure: V-Dem Liberal Democracy Index (LDI, 0–1); 0.5 threshold to classify non-democracies vs democracies; also used continuous LDI in interactions. Statistical analysis: Linear mixed-effects models (lme4). Pooled vignette analyses include respondent and country random intercepts; country-specific analyses include respondent random intercepts. Treatment-only models (no additional covariates) for experimental effects; discipline effect-coded (estimates represent mean outcome per discipline), other treatments dummy-coded (reference: female, no activism, non-national, low impact). Covariate analyses use linear models predicting general pre-treatment trust in science, interacting education, science consumption, and political orientation with continuous LDI; control for age and employment. Power and sensitivity analyses conducted via DeclareDesign; details in SI.

Pooled experimental effects (all countries):

• Gender: Male scientists rated lower than female across person dimensions and science outcomes: competence β = -0.020 (p < 0.01), responsibility β = -0.024 (p < 0.01), honesty β = -0.014 (p < 0.05). Their research rated less trustworthy β = -0.020 (p < 0.01), less support for public funding β = -0.030 (p < 0.01), and less policy influence β = -0.023 (p < 0.01).
• Activism: Publicly activist scientists perceived as more honest β = 0.028 (p < 0.01); their research favored to influence policy β = 0.020 (p < 0.05). No other significant detriments from activism.
• Nationality: Co-nationality increases perceived responsibility β = 0.015 (p < 0.05); no significant effects on competence, honesty, or science trust outcomes. In the U.S., framing foreign scientist as Chinese had no differential effect on honesty/competence.
• Impact: High-impact research (20–25% affected) increases all outcomes: scientist competence β = 0.043, honesty β = 0.027, responsibility β = 0.046; science trustworthiness β = 0.039; strong increases in support for public funding β = 0.145 and policy influence β = 0.125 (all p < 0.001).
• Discipline ranking: Genetics highest, then materials science, then economics and education (e.g., for honesty βgenetics ≈ 3.06 vs lower for others; all p < 0.01). Despite higher trust in genetics and materials science, economics often seen as more appropriate for policy influence than those fields. Education research lowest overall but similar honesty/trustworthiness to economics.

Cross-national patterns and regime differences:

• Broad consistency across countries and regime types in treatment effects. Gender effect primarily driven by Germany and Mexico (lower trust in male scientists/science; β ≈ -0.056, -0.059; p < 0.01) and China (male scientists rated less favorably; β = -0.041; p < 0.05).
• Activism effects heterogeneous: Positive in Singapore (higher trust in scientist; β = 0.047; p < 0.01) and Turkey (higher trust in science; β = 0.059; p < 0.01); negative in Poland (research rated less trustworthy; β = -0.042; p < 0.05); null elsewhere.
• Nationality effects essentially null across countries.
• Impact effects universally positive but stronger in democracies (e.g., Germany β = 0.195; U.S. β = 0.172; both p < 0.01) than in autocracies (e.g., China β = 0.095; Turkey β = 0.056; p < 0.01).
• Discipline ordering largely consistent across countries; education researchers least favored almost everywhere; in Germany, economists rated even lower as scientists though economics research not rated as poorly.

Covariate analyses (pre-treatment general trust in science):

• Education: Higher education associates with higher trust; relationship similar across regimes (predicted trust high-edu: 3.95 non-democracies vs 3.84 democracies). No evidence that democracy strengthens the education–trust link.
• Political orientation: Strong regime moderation. In democracies, right-leaning respondents show significantly lower trust than left-leaning (e.g., predicted trust: right 3.59 vs left 3.76). In non-democracies, left–right differences are not significant; both sides report higher trust overall (right 3.81; left 3.85). Interaction of ideology with democracy is significant (increasing democracy reduces trust among right-leaning respondents). Country specifics: negative right–trust association in most countries; exception India where left-leaning associates with lower trust.
• Science media consumption: Higher active consumption associates with higher trust in both regime types; regime does not significantly moderate the relationship. Trust levels higher overall in non-democracies at both low and high consumption (low: 3.62 vs 3.52; high: 3.96 vs 3.88). In China, India, and Turkey relationship is linear; in U.S., Germany, Brazil quadratic (highest trust among moderate consumers).

The findings directly address whether characteristics of scientists and their research differentially shape public trust across political regimes. Despite expectations of pronounced regime-based differences, experimental effects are largely consistent across democracies and autocracies: female scientists are trusted more than males; activism is neutral to slightly positive; high-impact research reliably boosts trust and support; and “hard” sciences (genetics, materials) garner more trust than social sciences. The only systematic regime-related heterogeneity is that high public impact yields larger trust gains in democracies. Pre-treatment covariate analyses show that education and science media consumption relate similarly to trust across regimes, whereas political orientation is strongly predictive of trust only in democracies, indicating greater politicization of science there. These results suggest that science’s credibility depends more on specific attributes of scientists and research than on regime context, with the notable caveat of issue politicization in democracies. The implications are that science communication and policy engagement can rely on broadly similar strategies across countries, emphasizing societal impact and transparent engagement without fearing credibility loss from activism, while addressing ideological polarization in democracies.

This cross-national preregistered experiment shows consistent effects of scientist and science attributes on public trust across diverse political regimes. Female scientists and publicly engaged scientists are trusted at least as much, often more, than their counterparts; high-impact research increases trust and support; and disciplinary stereotypes favor genetics and materials science over economics and education. Regime differences are modest, except that political ideology strongly structures trust in democracies but not in autocracies. The study contributes a nuanced, causally identified perspective by disaggregating perceptions of scientists and science, and by distinguishing person- from research-focused trust. Future research should: (1) unpack mechanisms behind higher trust in female scientists; (2) vary findings within disciplines and disciplines within findings to separate disciplinary reputation from content effects; (3) examine more polarized issue domains to test robustness under politicization; (4) use more representative sampling frames and broader age ranges; and (5) explore interventions to mitigate ideological polarization in democracies while safeguarding scientific autonomy in autocracies.

• Sampling: Online quota sampling (gender, age) in panels may limit population representativeness; respondents likely more digitally active; age restricted to 18–55.
• Experimental scope: Disciplines are confounded with vignette-specific descriptions; effects cannot fully disentangle disciplinary reputation from content.
• Topic selection: Avoided highly polarized topics (e.g., climate change, vaccines, COVID-19), which may limit generalizability to contentious domains.
• Effect sizes: Many effects are modest, though potentially impactful at population scale.
• Measurement and design: Nationality manipulation limited to U.S./Chinese non-co-national comparison; trust outcomes use reduced METI item set (three items) rather than full scale; vignette designs cannot capture all real-world context.
• Cross-national comparability: Despite translations and harmonization, cultural interpretation of scales may vary; samples not probability-based.