Greater traditionalism predicts COVID-19 precautionary behaviors across 27 societies

Abstract

People vary both in their embrace of their society's traditions, and in their perception of hazards as salient and necessitating a response. Over evolutionary time, traditions have offered avenues for addressing hazards, plausibly resulting in linkages between orientations toward tradition and orientations toward danger. Emerging research documents connections between traditionalism and threat responsivity, including pathogen-avoidance motivations. Additionally, because hazard-mitigating behaviors can conflict with competing priorities, associations between traditionalism and pathogen avoidance may hinge on contextually contingent tradeoffs. The COVID-19 pandemic provides a real-world test of the posited relationship between traditionalism and hazard avoidance. Across 27 societies (N = 7844), we find that, in a majority of countries, individuals' endorsement of tradition positively correlates with their adherence to costly COVID-19-avoidance behaviors; accounting for some of the conflicts that arise between public health precautions and other objectives further strengthens this evidence that traditionalism is associated with greater attention to hazards.

Figure 1

Results of a random effects, restricted maximum likelihood meta-analysis in which each study site was treated as a separate sample. Plot shows zero-order product-moment correlations between traditionalism and COVID-19 health precautions at each study site, ordered by effect size. For the individual country estimates, the location of the square along the x-axis corresponds with the correlation coefficient, the size of the square corresponds with the weight of that study site in the meta-analysis, and bands are 95% confidence intervals. At the bottom of the plot, an overall meta-analyzed point estimate is provided. The midpoint of the diamond corresponds with that point estimate, the width of the diamond corresponds with the 95% CI, and the dotted bands correspond with the 95% prediction interval. On the right side of the plot, weights, correlation coefficients, and 95% CIs respectively are numerically listed for both the site-specific correlations, as well as the overall estimate. Note that for the overall meta-analyzed point estimate, the 95% confidence interval does not overlap with zero, while the 95% prediction interval does.

Figure 2

Graphical visualization of the country-specific correlations listed in Fig. 1. Dotted lines are study site-specific product-moment correlations between traditionalism and COVID-19 health precautions. The solid thick line is the unweighted product-moment correlation in the pooled sample across all study sites. Dots show individual data points, jittered along the x- and y-axes to aid interpretability. Density plots along the x- and y-axes represent the raw distributions of the traditionalism and COVID-19 health precautions composites. Thin grey lines show density distributions at individual study sites, whereas the thick black lines show the overall distribution in the pooled sample across all study sites. Study sites are unlabeled to improve readability. For labeled study-site specific correlations and density distributions, see Figs. S2–S4 in the Supplement.

Figure 3

Results of a random effects, restricted maximum likelihood meta-analysis in which each study site was treated as a separate sample. The plot shows semi-partial correlations^, between traditionalism and COVID-19 health precautions at each study site, after adjusting for the effects of the five identified suppressor variables in multiple linear regressions where health precautions were regressed on traditionalism and each of those five variables. Covariates were identical across study sites. Note that the semi-partial correlations indicate the variance in health precautions uniquely explained by the aspects of traditionalism separate from the five suppressor variables, and the effect sizes can be interpreted using the same metrics applied to product-moment correlations. See Fig. 1 for a description of how to interpret the forest plot. For the overall meta-analyzed point estimate, neither the 95% confidence interval nor the 95% prediction interval overlap with zero.

Figure 4

Results of a restricted maximum likelihood moderated mixed linear regression in which COVID-19 health precautions were regressed on traditionalism, a health precautions indicator variable (e.g., either internal-facing or external-facing), and the interaction between those two variables in the pooled sample. The model included participants nested within study sites as random effects. To test this interaction, there were two observations for each participant; the first observation contained each participants’ internal-facing precautions score, and the second their external-facing precautions score. We simultaneously created an indicator variable specifying which health precautions subscale corresponded with each observation. Simple slopes were then plotted in the figure. There was an interaction between health precautions subscale and traditionalism (B = 0.16, SE = 0.01, t(7,535) = 12.76, p < 0.001). A simple slopes analysis revealed that the correlation between traditionalism and internal-facing precautions (B = 0.29, SE = 0.01, t(7,535) = 23.17, p < 0.001) was about twice as strong as the correlation between traditionalism and external-facing precautions (B = 0.14, SE = 0.01, t(7,535) = 10.84, p < 0.001). Note that these results were robust to the inclusion of demographic and COVID-19-related covariates, and they were not conceptually affected when the five suppressor variables were included as covariates (see Supplement page S26). Further, results did not conceptually change when using factor scores instead of averaged composites (see Supplement page S63). Finally, we considered the possibility that the presence—or lack of presence—of planning precautions may be confounding our interpretation of the external- and internal-facing precautions subscales. Specifically, the internal-facing subscale has more items related to planning precautionary behaviors (such as the importance of obtaining prophylactic supplies), whereas the external-facing subscale has more items related to actual precautionary behavior (such as wearing a mask when outside the home). To address this possibility, we created a modified internal-facing precautions composite that excluded all planning-related precautions. Using the planning-less internal-precautions composite did not conceptually affect these results (see Supplement S26), suggesting that planning behaviors versus actual behaviors are not confounding our explanation for the moderating effect of external- versus internal-facing precautions.