The COVID-19 pandemic presented unprecedented challenges, forcing governments worldwide to implement large-scale interventions like school closures and lockdowns. Prior research indicates a strong influence of partisanship on public attitudes toward these policies and perceptions of the crisis's severity. This study delves deeper into this relationship by exploring whether and how partisan differences in policy support connect to partisan gaps in beliefs about projected COVID-19 deaths. The researchers hypothesized two competing explanations: a 'mediation hypothesis,' where partisan differences in death forecasts drive policy preferences; and an 'independence hypothesis,' where partisanship independently influences both death forecasts and policy support. The study's significance lies in understanding the psychological mechanisms underpinning partisan responses to the pandemic and informing policymaking strategies.

Existing research highlights the role of political polarization in shaping public opinion and behavior during crises. Studies in the United States have shown a partisan divide in policy preferences and responses to the COVID-19 pandemic, with Republicans generally exhibiting less concern about the virus's severity and greater opposition to preventive measures compared to Democrats. However, the cognitive mechanisms behind these polarized attitudes remain unclear. The authors review prior studies indicating that partisan biases can stem from social identity motives and are often activated by collective threats. Other research demonstrates that policy preferences can be influenced more strongly by party affiliation than by individual beliefs. The current study aims to bridge this gap by investigating whether partisan differences in death forecasts mediate the relationship between partisanship and policy support.

To test the competing hypotheses, the researchers conducted four independent behavioral experiments across four countries: Argentina, Uruguay, Brazil, and the United States. Each experiment employed a similar methodology, involving an anchoring procedure where participants first considered either a very low or high number of projected COVID-19 deaths before making their own death forecasts. Participants then rated their agreement with a series of public health interventions. Experiment 1 (Argentina) used a convenience sample of university students (N=640). Experiments 2 (Uruguay, N=372) and 3 (Brazil, N=353) used samples recruited through Offerwise, a Latin American panel company. Experiment 4 (United States, N=615) utilized Prolific, an online platform for recruiting research participants, ensuring a sample representative of the US population in terms of age, gender, and ethnicity. The experiments differed slightly in sampling methods, anchors used, and the specific policies assessed to ensure relevance to each country's context. Data analysis for each experiment was conducted separately, and cross-country comparisons were avoided due to the differences in research design. Experiment 4 incorporated several additional conditions to further investigate the anchoring effects and included an incentive for accurate forecasting. This experiment manipulated anchors on both deaths and cases and included control conditions with no anchoring. Statistical analyses included t-tests to compare forecasts and policy preferences across anchor conditions, correlations to examine the relationship between forecasts and policy support, Bayes factor analyses to quantify evidence for or against the hypotheses, and structural equation modeling to examine the interplay between partisanship, forecasts, and policy support. The researchers used Bayesian Information Criterion (BIC) to compare the fit of different models (mediation, independence, and a full model).

Across all four experiments, the anchoring manipulation significantly affected death forecasts, demonstrating the effectiveness of the procedure. However, contrary to the mediation hypothesis, the anchoring manipulation did not significantly affect participants' support for COVID-19 policies. The correlation between forecasted deaths and policy support was consistently non-significant across all four experiments, indicating a lack of association between these variables. Power analyses demonstrated that the studies had sufficient power to detect even small effects. Equivalence tests confirmed that the differences between anchor conditions were not practically significant in terms of policy support. Bayes factor analyses provided strong evidence supporting the independence hypothesis (i.e., an absence of a relationship between forecasted deaths and policy support), with the data being much more likely under the independence hypothesis than the mediation hypothesis. Structural equation modeling further supported the independence hypothesis, showing that partisanship significantly influenced both death forecasts and policy support independently. In Argentina and Uruguay, greater support for the ruling party correlated with more optimistic death forecasts and greater support for policies. Conversely, in Brazil and the United States, greater support for the ruling party correlated with more optimistic death forecasts and less support for policies. This suggests that the effect is primarily driven by partisanship rather than ideology itself, since the ruling parties in Argentina and Uruguay held contrasting political platforms.

The findings challenge rational accounts suggesting that partisan differences in COVID-19 policy support stem from varying perceptions of the pandemic's severity. The consistent absence of a relationship between death forecasts and policy preferences across diverse political contexts suggests that partisan identity plays a dominant role, overriding individual beliefs about the crisis's intensity. This aligns better with tribal theories of partisan behavior, where party loyalty supersedes individual beliefs in shaping political attitudes and actions. The study's cross-national scope strengthens the generalizability of the findings, showing that the independence hypothesis holds across countries with diverse political systems and cultures. Although the magnitude of partisan effects may vary across countries, the overarching pattern of independent influences of partisanship on death forecasts and policy support is consistent. The results imply that simply providing information about the pandemic's severity may be insufficient to achieve widespread support for preventive measures. A unified message from leaders across the political spectrum could be critical for creating a more coordinated and effective public health response.

This study provides strong evidence that partisanship independently influences beliefs about the severity of the COVID-19 pandemic and support for preventive measures. The lack of association between death forecasts and policy preferences underscores the importance of partisan identity in shaping public responses to health crises. Future research could explore the interplay of additional factors, such as media consumption and social influence, in shaping these responses. The findings highlight the need for strategies that go beyond simply communicating risk information to overcome partisan divides and foster more effective public health responses.

Several limitations should be considered. First, direct comparisons of policy preferences across countries are not appropriate due to variations in research design and specific policy items measured. Second, the samples, while diverse in terms of political orientation, might not perfectly represent the general population of each country. Third, the relationships between partisanship and other variables are correlational, making it difficult to establish definitive causal links. Finally, the models explained only a portion of the variance in the data, suggesting that other factors influence policy preferences.