Pro-religion attitude predicts lower vaccination coverage at country level

The study addresses global inequality in COVID-19 vaccination coverage and asks whether country-level attitudes toward science and religion help explain disparities in vaccination uptake beyond supply and income differences. Prior work shows trust in science is associated with vaccine acceptance, while stronger religious faith can correlate with hesitancy. The authors hypothesize that countries where more people prioritize religion over science in cases of conflict (science-religion choice) and where religious faith is stronger will exhibit lower vaccination coverage and slower vaccination speed. The purpose is to quantify inequality in coverage as of mid-2022 and test whether science trust, religious faith, and science-religion choice predict country-level vaccination outcomes, thereby informing strategies to address vaccine inequality by considering socio-cultural factors.

The paper reviews evidence that trust in science predicts positive vaccine evaluations and intentions across Italy, New Zealand, and multi-country surveys, highlighting distrust as a driver of hesitancy. It also surveys literature linking religious beliefs to lower vaccination uptake, including moral concerns, doctrinal conflicts, and the role of Christian nationalism, while noting complexities across denominations and contexts (e.g., inverted U-shaped relationships in some countries). Although science and religion can coexist, in vaccination contexts tensions arise. Cross-national differences in religiosity (e.g., low in China, high in Turkey and the US) suggest potential macro-level effects. The review also acknowledges instances where religious leaders’ endorsement increases vaccine acceptance, indicating that religious influences may both hinder and facilitate uptake depending on messaging and denomination.

Design and data sources: Observational, cross-national analysis integrating three public datasets. Outcomes were derived from Our World in Data (OWID) COVID-19 vaccination database; predictors and covariates from the 2018 Wellcome Global Monitor (WGM 2018) and World Values Survey Wave 7 (WVS 2017–2022). Countries were classified by income using the World Bank (2022) scheme. Countries in OWID not classed by the World Bank were excluded, yielding 214 countries covering ~99% of world population.

Time frame and outcome variables: OWID vaccination data up to July 1, 2022 were used, aligning with WHO’s mid-2022 70% target. For each country, vaccination coverage was defined as the maximum people fully vaccinated per hundred recorded up to 1 July 2022 (or nearest maximum around that date). Vaccination speed (daily growth) was computed as (maximum − minimum) / day interval, using only positive values to reflect valid progress. Values exceeding 100% (Gibraltar, United Arab Emirates, Samoa, Tonga) were capped at 100.

Predictors: Science-religion choice from two sources: (a) WGM 2018: country-level average response to “Generally speaking, when science disagrees with the teachings of your religion, what do you believe? Science or the teachings of your religion?” (excluding “don’t know/refused”). (b) WVS Wave 7: country-level average agreement with “Whenever science and religion conflict, religion is always right.” Religious faith (WVS): country-level average of a four-item index (belief in God, life after death, hell, heaven), with high internal consistency (Cronbach’s α=0.868; McDonald’s ω=0.881). Science trust (WGM): “In general, would you say that you trust science a lot, some, not much, or not at all?” averaged at country level. Higher scores indicate stronger science trust, stronger religious faith, and greater tendency toward religious choice.

Covariates: Income type (binary recode: high/upper middle=1; low/lower middle=0), GDP (from OWID), aging (share >65; OWID), stringency index (mean over time; from Oxford COVID-19 Government Response Tracker via OWID), population density (OWID), government trust (WGM: “How much do you trust the national government?”), vaccines trust (WGM: beliefs that vaccines are safe and effective, combined). Poverty share (OWID) was described but primary models report seven covariates listed above. Country names/ISO codes were harmonized across datasets; special cases (e.g., Palestine/Palestinian Territories; Morocco codes) were resolved. Mainland China, Hong Kong, Macao, and Taiwan were treated separately; “China” denotes mainland China.

Statistical analysis: Ordinary least squares (OLS) regressions with standardized variables were estimated predicting two outcomes: vaccination coverage (people fully vaccinated per hundred) and vaccination speed. Models were run with and without covariates. To assess robustness and avoid selective reporting, a multiverse (specification curve) analysis was conducted spanning four analytic decisions: outcome (coverage vs speed), predictor source (WGM vs WVS science-religion choice), regression type (OLS vs robust regression using M-type estimator, robustbase package), and covariate sets (all combinations of seven covariates, including none; 128 options). This yielded 1,024 specifications. The standardized coefficient of the focal predictor and its 95% CI were summarized; significance proportions and medians reported.

• Global vaccination status (as of mid-2022): At the country level, the average people fully vaccinated per hundred was 52.60 (SD=26.42); at the population level, 61.21% were fully vaccinated. For at least one dose, the country-level mean was 57.42 (SD=26.49); population-level 66.75%. These fell short of WHO’s 70% target by mid-2022.
• Inequality by geography and income: Lower coverage was concentrated in West and South Asia, Eastern Europe, Central America, and especially Africa. High-income countries started earlier and increased coverage faster than low-income countries, widening gaps over time.
• Top vs bottom countries: Examples among the top 20 coverage countries include Singapore (91.61%), China (88.40%), Spain (85.42%). Among the bottom 20: Sudan (9.94%), Haiti (1.37%), Burundi (0.12%).
• Vaccination speed differences: Top 50 countries had higher speed than bottom 50: M_top50=0.19 (SD=0.14) vs M_last50=0.04 (SD=0.02); Welch’s t(51.51)=7.80, p<0.001; Cohen’s d=2.17 (95% CI: 1.48–2.85); BF10=20.66. Bottom 50 in 2022 (M=0.04, SD=0.03) were still slower than top 50 initial speed before July 2021 (M=0.24, SD=0.19); t(46.14)=6.82, p<0.001; d=2.01 (95% CI: 1.29–2.71); BF10=16.24.
• Science-religion choice (WGM 2018): Predicts lower vaccination coverage (β=−0.526, p<0.001) and slower speed (β=−0.417, p<0.001) without covariates; associations attenuate and become non-significant with covariates (coverage β=−0.085, p=0.444; speed β=−0.077, p=0.565).
• Science-religion choice (WVS Wave 7): Consistently predicts lower coverage and speed with and without covariates. Coverage: β=−0.622, p<0.001 (no covariates); β=−0.558, p=0.015 (with covariates). Speed: β=−0.490, p<0.001 (no covariates); β=−0.557, p=0.027 (with covariates).
• Multiverse analysis: Across 1,024 specifications, 77.44% yielded a significantly negative coefficient for science-religion choice; median standardized coefficient β=−0.458, demonstrating robustness across analytic choices.
• Science trust (WGM): Positive predictor only without covariates (coverage β=0.428, p<0.001; speed β=0.340, p<0.001); becomes non-significant with covariates (coverage β=−0.068, p=0.471; speed β=−0.085, p=0.451).
• Religious faith (WVS): Robust negative predictor of both coverage and speed with and without covariates. Coverage: β=−0.533, p<0.001 (no covariates); β=−0.573, p=0.011 (with covariates). Speed: β=−0.427, p=0.001 (no covariates); β=−0.627, p=0.013 (with covariates).

The findings show that global vaccine coverage remained below WHO targets by mid-2022 and that disparities are pronounced across countries and income groups. While early inequality was partly driven by supply and rollout timing, persistent gaps in both coverage levels and vaccination speed suggest that non-supply factors play a key role. Country-level socio-cultural orientations toward science and religion help explain these disparities: a greater tendency to choose religion over science in conflicts and stronger religious faith are associated with lower coverage and slower speed. Science trust, strongly predictive at the individual level in prior studies, shows a weaker and specification-sensitive association at the country level, implying macro-level vaccination uptake is influenced by more complex and interacting factors. The results underscore that interventions aimed at increasing uptake must account for cultural and religious contexts. Engagement by religious authorities and framing vaccination as consistent with religious doctrine may mitigate hesitancy, while investment in science education could increase alignment with scientific guidance when it conflicts with religious beliefs. However, denominational differences and the potential for non-linear and higher-order interactions indicate that simple linear associations only partially capture the dynamics, warranting more nuanced modeling and targeted strategies.

The study documents substantial global inequality in COVID-19 vaccination coverage and speed as of mid-2022 and demonstrates that country-level pro-religion orientations—operationalized as science-religion choice favoring religion and stronger religious faith—predict lower vaccine uptake and slower progress. Science trust shows positive associations only absent covariates, suggesting more complex macro-level determinants. These insights highlight the necessity of integrating socio-cultural factors into strategies to address vaccine inequity. Policy implications include partnering with religious leaders to align pro-vaccination messaging with doctrine and strengthening science education to foster deference to scientific evidence in conflicts with religious teachings. Future research should examine denominational heterogeneity, explore higher-order interactions among cultural, political, and structural factors, and use innovative modeling approaches to better capture the complex determinants of population-level vaccination behavior.

• The analysis is observational and ecological at the country level; associations may be influenced by unmeasured macro-level factors and model specification choices.
• Science trust effects were not robust after adjusting for covariates, indicating sensitivity to included controls and potential confounding.
• Religious effects may vary by denomination and context; the study aggregates across denominations and countries, potentially masking heterogeneity.
• Differences in measurement between WGM and WVS (e.g., question wording) and missing data reduced sample sizes for some models, as noted in tables.
• Linear models may not capture higher-order interactions among socio-cultural and structural variables; authors note that true dynamics may be more complex than captured here.