Not-so-straightforward links between believing in COVID-19-related conspiracy theories and engaging in disease-preventive behaviours

The study investigates how beliefs in different COVID-19-related conspiracy theories relate to individuals’ engagement in disease-preventive behaviours, and under what conditions these relationships may reverse. Two broad types of conspiracy theories are distinguished: risk-acceptance (acknowledging COVID-19 as a serious threat, e.g., bioweapon narratives) and risk-rejection (downplaying or denying the threat, e.g., “hoax” or “just like flu”). The authors hypothesize that RA-CT beliefs are positively associated with risk perception and negative emotions about COVID-19, which in turn predict greater preventive behaviours (H1a–H1b). Conversely, RR-CT beliefs are hypothesized to be negatively associated with risk perception and negative emotions, predicting fewer preventive behaviours (H2a–H2b). They further propose that outbreak severity moderates these links: RA-CT effects on risk perception/negative emotions are stronger in the early, less-severe stage (H3a), whereas RR-CT effects are stronger in the later, more-severe stage (H3b), leading to stronger positive (RA-CT) and negative (RR-CT) indirect effects on preventive behaviour at those respective stages (H4a–H4b). This work aims to clarify nuanced mechanisms linking conspiracy belief content, psychological responses, and health behaviours during evolving public health crises.

The paper situates COVID-19 conspiracy beliefs within broader definitions of conspiracy theories and their roles in sense-making during crises. Prior work has identified multiple pandemic-related conspiracy narratives (e.g., lab-leak/bioweapon, 5G, Bill Gates/vaccination, QAnon, and hoax narratives). Empirical studies have shown distinct behavioural correlates: belief that COVID-19 is human-made relates to greater prepping and some hygiene behaviours, whereas hoax beliefs relate to lower preventive behaviours and policy support (Imhoff & Lamberty, 2020; Sternisko et al., 2020). The authors argue that these beliefs reflect implicit appraisals of disease risk and emotions, consistent with research linking risk perception/threat feelings to preventive actions. They highlight a gap regarding mechanisms (risk perception and negative emotions) and context (outbreak severity) moderating these associations, which previous single-time-point studies during severe outbreak phases did not address.

Design: Repeated cross-sectional surveys at six time-points in the U.S. from early February to late March 2020 (about every 10 days), spanning mild (11 confirmed U.S. cases) to severe stages (up to 68,440 cases). Target n≈250 per time-point; total N=1,515 recruited via CloudResearch (MTurk). Three attention checks were included; seven participants failing all were excluded. Samples showed highly similar demographics across time-points. Measures: - Conspiracy beliefs: Items generated from January 2020 news/social media/fact-checks. Risk-acceptance (5 items; e.g., purposeful spread, bioweapon, cover-up; 0–10 scale); Risk-rejection (2 items; e.g., “similar to influenza, severity exaggerated”; 0–10 scale). Mean scores computed. Two-factor structure favored over one-factor in CFA. Coverage of circulating narratives validated via retrospective social media engagement. - Risk perception: 3 items (likelihood of infection for community, family, self; 1–7 scale), averaged. - Negative emotions toward COVID-19: 5 negative (afraid, threatened, anxious, tense, disgusted) and 3 positive (optimistic, calm, hopeful) items (1–7). Two-factor CFA fit well; negative emotions mean used as mediator. Positive emotions explored in supplements and treated as covariate in robustness checks. - Disease-preventive behaviours: 10 behaviours (e.g., mask use, handwashing, avoiding crowds/travel/public transport, staying home, hygiene, monitoring temperature; 1–5), averaged. Covariates: Institutional trust (4 items; confidence in scientists, medical experts, U.S. government, CDC; 1–7); political ideology (liberal and conservative items combined; higher scores = more conservative); demographics (gender, age, education, income; plus SES ladder, ethnicity, party for sample description). Analytic strategy: (1) Zero-order correlations in pooled and time-point samples; visualization via ggplot. (2) Parallel mediation (PROCESS) with RA-CT and RR-CT as predictors, risk perception and negative emotions as simultaneous mediators, preventive behaviours as outcome; covariates controlled; bootstrapped CIs for indirect effects. (3) Moderation by outbreak stage (coded 1–6, standardized for interactions) tested via multiple regressions predicting risk perception and negative emotions; moderated mediation (PROCESS) assessed stage-contingent indirect effects. Robustness checks included adding positive emotions as covariate; results unchanged.

- Bivariate associations with preventive behaviours (pooled N=1,508): - RA-CT belief: r=0.19, p<0.001; time-point range: −0.09 (Time 6, ns) to 0.28 (Time 3, p<0.001). - RR-CT belief: r=−0.01, p=0.752; time-point range: −0.15 (Time 6, p=0.018) to 0.12 (Time 2, p=0.064). - Parallel mediation (all paths p<0.001): - RA-CT → higher risk perception (b=0.16, SE=0.02) and higher negative emotions (b=0.12, SE=0.01) → more preventive behaviours. Indirect effects significant via risk perception (partial standardized IE=0.02, 95% CI [0.02, 0.03]) and via negative emotions (0.02, 95% CI [0.02, 0.04]). Direct effect to behaviours positive (b≈0.05, SE=0.01). - RR-CT → lower risk perception (b=−0.11, SE=0.02) and lower negative emotions (b=−0.12, SE=0.01) → fewer preventive behaviours. Indirect effects significant via risk perception (−0.02, 95% CI [−0.02, −0.01]) and via negative emotions (−0.01, 95% CI [−0.02, −0.01]). Direct effect to behaviours negative (b≈−0.03, SE=0.01). - Institutional trust unrelated to RA-CT/RR-CT but positively related to preventive behaviour. - Outbreak stage effects: Later stages predicted higher risk perception (b=0.85, SE=0.03, p<0.001) and negative emotions (b=0.32, SE=0.04, p<0.001); preventive behaviours increased over time (r=0.36, p<0.001). Trends aligned with external polling. - Moderation by stage: - RA-CT × stage: predicting risk perception b=−0.07, SE=0.04, p=0.089 (marginal); predicting negative emotions b=−0.12, SE=0.04, p=0.004. Stronger positive associations in early (less severe) stages (supports H3a, especially for emotions). - RR-CT × stage: predicting risk perception b=−0.15, SE=0.04, p<0.001; predicting negative emotions b=−0.01, SE=0.04, p=0.870. Stronger negative link with risk perception in later (more severe) stages (partial support for H3b). - Moderated mediation: - RA-CT → negative emotions → preventive behaviours: stronger indirect effect in less-severe than more-severe stages (index=−0.01, bootstrapped 95% CI [−0.02, −0.00]). - RR-CT → risk perception → preventive behaviours: stronger (more negative) indirect effect in more-severe than less-severe stages (index=−0.01, bootstrapped 95% CI [−0.02, −0.00]).

Findings reveal that the relation between COVID-19 conspiracy beliefs and preventive behaviour depends on conspiracy content and outbreak context. Believing risk-acceptance narratives elevated perceived risk and negative emotions, promoting preventive actions—especially early in the outbreak when normative risk perception was lower. In contrast, believing risk-rejection narratives suppressed perceived risk and negative emotions, reducing preventive actions—particularly during severe outbreak stages when normative risk perceptions were higher. These patterns persisted after controlling for institutional trust, political orientation, and demographics, clarifying mechanisms (risk perception and negative affect) that connect belief content to behaviour. The results underscore that conspiracy beliefs can function as sense-making frames that either acknowledge or deny threat, and that contextual severity modulates their impact on health-protective behaviour.

The study demonstrates nuanced, content- and context-dependent links between COVID-19 conspiracy beliefs and disease-preventive behaviours via risk perception and negative emotions. Risk-acceptance beliefs are associated with greater preventive behaviour, especially early in an outbreak, while risk-rejection beliefs relate to reduced preventive behaviour, especially when the outbreak is severe. Contributions include identifying psychological mediators, documenting contextual moderation across outbreak stages, and replicating/extending prior work. Future research should: (1) establish causality via experimental manipulations of conspiracy exposure; (2) use objective/observational measures of preventive behaviour; (3) broaden the set of conspiracy narratives as discourse evolves; and (4) examine potential adverse social consequences (e.g., prejudice and intergroup hostility) of risk-acceptance narratives and interventions to mitigate them.

- Repeated cross-sectional (not longitudinal) design with different participants at each time-point limits causal inference and may allow sampling differences to confound results. - Self-reported preventive behaviours may suffer from recall or social desirability biases. - Conspiracy belief measures captured a subset of narratives available early in 2020; some narratives (e.g., 5G, Bill Gates) were not widely circulating when items were generated. More RA-CT than RR-CT items; RR-CT internal reliability relatively low (results robust to single-item analyses). - Mean levels of conspiracy beliefs were largely stable across time-points, limiting inferences about temporal change in such beliefs. - Positive emotions toward COVID-19 were not the focus; although explored, their role remains less clear.