We start by visualizing simple correlations that will subsequently be unpacked through mutivariate analysis. As Fig. 1 shows, the initial spread of COVID-19 was positively correlated with cosmopolitanism, COS (Adjusted r2 = 0.229, p < 0.0001) and rational-secularism, RAT (Adj. r2 = 0.217, p < 0.0001). Deaths correlate negatively with institutional confidence, INST (Adj. r2 = 0.029, p = 0.066). There is little correlation, however, with government efficiency, but it is positive (Adj. r2 = 0.011, p = 0.168).
The first two principal components (PCs) from the principal component analysis, which together explain 67% of the variance (Table 1), appear to capture two main sources of variation: people and institutions. The first component, PC1, which accounts for 46% of the variance, is positively loaded on all covariates except INST, and not strongly loaded on government efficiency (Table 1, Figs. S1 and S2). The second component, PC2, which reflects 21% of the variance, is primarily loaded on institutional confidence (INST) and government efficiency (Gov. eff.). To ensure that our choice of government variable did not determine the outcome, we tested the effects on the PCA using additional, similar variables (see Supplement).
Venturing to interpret the PCA, it appears that PC1 is a “people” component—cultural values, age and obesity—whereas PC2 captures the effectiveness of institutions, requiring effective governments and people with confidence in their institutions. The left side of Fig. 2 shows that COVID-19 deaths per capita correlate positively with PC1 (Adj. r2 = 0.469, p < 0.0001) and negatively with PC2 (Adj. r2 = 0.037, p = 0.045). A negative, but weaker, correlation with PC2 (Fig. 2, right) indicates that effective institutions predict reduced COVID-19 deaths per capita. As PC1 is also loaded on urbanism and GDP (Table 1), which themselves are underpinned by secular-rational and cosmopolitan cultural values (Ruck et al. 2018, 2020a), the comingling of factors motivate our regression analysis.
Next, we explore these hypotheses raised by the PCA with multivariate regression, which helps disentangle the joint effects of cultural factors, government efficiency, socioeconomic development and individual-level risks of age and obesity. In the regression results shown in Table 2, we have combined the institutional confidence and government efficiency variables as an interaction term, labelled as INST.Gov, as these combined variables are the essence of PC2, and also because we would expect governments to be more effective with public trust in their institutions. While Table 2 shows that INST.Gov has significant effect, the Supplementary materials show that institutional confidence and government efficiency have limited effect independently (Table S1), even with the independent effects included (Table S2).
The regression (Table 2) shows increased effect of several covariates on COVID-19 deaths per capita between Day 10 and Day 50, likely because so many fatalities occur weeks after infection. The z-scores of these effects between 5 and 55 days after the outbreak (Figure S3) increased over the first two months of the outbreak. By Day 50, the significant factors are more clear in the regression (Table 2). Not surprisingly, obesity and population size predict more deaths per capita. The Cosmopolitan (COS) cultural factor also predicts increased deaths. The combined variable, INST.Gov, predicts fewer deaths per capita (Table 2), which is broadly consistent with the negative correlation between PC2 and deaths (Fig. 2, right).
The regression results reveal similar patterns as the PCA. At the level of national populations, the cultural factor of cosmopolitanism, together with obesity, predict higher numbers of per-capita deaths in the first two months of COVID-19. At the government level, the complementary variables of government efficiency and public trust in institutions predict lower death numbers, but with weaker effect.
The measurable effect of the COS factor is larger after 50 days than after 10 days (Table 2), potentially because deaths increased by order(s) of magnitude during that time in many countries. We focus on Day 50 in Table 2, when the factor COS exhibits a larger effect than RAT (secular-rationalism). Strictly speaking, COS is a predictive factor in COVID deaths, not necessarily the ultimate cause. COS probably captures a cultural openness to human interaction, which facilitates COVID spread. Given the importance of government action (Hale et al. 2020)—in imposing travel restrictions, lockdowns and/or suspensions of educational, commercial and religious activities (Flaxman et al. 2020; Hsiang et al. 2020)—a reasonable counter-argument is that COS merely predicts the ability of national governments to control the pandemic. The problem with this is that authoritarian governments are better at imposing restrictions, and authoritarianism tends to correlate negatively with COS (Ruck et al. 2020a). Another counter-argument might be that nations with lower COS are less prone to accurately/honestly report their COVID statistics, and hence higher COS predicts higher reported COVID-19 deaths. This is possible and should be the subject of more granular, qualitative research.
Among demographic controls, population size and obesity predicted more deaths. By Day 50, obesity had the largest effect on COVID-19 deaths (Table 2). This was expected, as obesity increases the risk of fatality from COVID-19 (Hamer et al. 2020). Early in the outbreak (Day 10), higher GDP per capita predicts higher COVID-19 death rates (Table 2). While GDP correlates with obesity and life expectancy, this additional effect of GDP on COVID-19 may reflect economic incentives of wealthier populations to resist shutdown measures.
More broadly, our results support the case that open and tolerant societies, which tend to be democratic (Ruck et al. 2020a), may make it harder for governments to effectively mitigate the effects of a pandemic such as COVID-19. It might be related to the degree to which “survival” is a prominent factor in cultural values, in the sense that survival values prioritize economic and physical security over self-expression and quality of life, which tends to be more common in autocratic countries (Inglehart and Baker 2000). There is also a secondary effect of institutional confidence predicting lower deaths. All in all, this suggests that countries with high cosmopolitanism and low institutional confidence are in particular danger, such as many Latin American countries (Ruck et al. 2020a).
Finally, a larger question is the resilience of cultures and democracies to unprecedented challenges and events (Muthukrishna 2020). While trust in institutions predicted fewer COVID-19 deaths and ought to facilitate government action, this value has been declining for decades in many Western countries (Ruck et al. 2018). Cultural values of cosmopolitanism, which predict the economic prosperity and democracy of nations in the long term (Inglehart and Welzel 2005; Ruck et al. 2018, 2020a, b) may, in short-term crisis events, hinder a strategic, coordinated national response. Hence, while a multi-decade trend towards greater openness towards minorities around the world (Ruck et al. 2020a) is normally a good thing, governments should consider the role of cultural values in preparing for the next pandemic.