Data on COVID-19 fatalities
COVID-19 deaths: data from the European Centre for Disease Prevention and Control (Rosler et al. 2020) were obtained from ourworldindata.com. Government efficiency index is taken from the World Economic Forum's 2018 Global Competitiveness (WEF 2018); it is a composite measure that quantifies: (1) efficient public spending, (2) weak burdens on private companies, (3) efficient judiciary, (4) responsive to private sector and (5) transparent policy changes. GDP per capita is measured at purchasing power parity (constant 2011 international dollars) for most recent year available for each country (World Bank 2018). Government response index is a composite variable of comprising information on 17 policies thought to help mitigate COVID-19 spread (Hale et al. 2020), including containment (school closures, mobility restrictions etc.), economic policies (e.g., direct payments) and heath policies (e.g., testing regimes, extra healthcare spending).
Methods for counting COVID-19 deaths vary between countries (e.g. including deaths at home as well as hospitals, likelihood of less effective counting in low income countries) as there is no internationally accepted standard. Cultural and political variables, which are inherent to our regressions, may also affect how COVID-19 deaths are counted. We nevertheless believe the effect of reporting differences to be small in our results. First, a plot of cumulative data from 25 Nov 2020, from 167 countries (with at least 1,000 reported cases) follow a linear slope (i.e., case fatality ratio, CFR) of 2.0% (r2 = 0.89). Notably, cumulative figures from the four income categories for nations (low, lower middle, upper middle, high) fall on this same line. Hence, if deaths at lower income levels were significantly under-reported, cases would have to be under-reported by the same percentage. The CFR across these categories ranges from 1.56% in low income countries to 3.09% in upper middle-income countries, suggesting some under-reporting of deaths in lower income countries. In log-transformed numbers, however, this is only a 15% difference, very slight on a log–log plot of cases versus deaths.
Furthermore, the outlier nations seem to reflect actual pandemic situation rather than reporting irregularities. The lowest CFRs are Singapore (0.048%), Curacao (0.16%), Qatar (0.17%), Botswana (0.31%), UAE (0.35%), and Maldives (0.36%), each of which appears to have been genuinely strict in controlling COVID-19, such as requiring visitors to show a negative result from a certified COVID-19 PCR-test. The highest CFRs are in Yemen (29%), Mexico (9.7%), Sudan (7.4%), Ecuador (7.1%) and Bolivia (6.2%), all of which are countries with genuinely, tragically high COVID-19 death rates as opposed to outstanding administrative protocols for reporting them. In summary, while there is uncertainty in the COVID-19 death numbers, we do not believe these uncertainties are systematic or large enough to explain our regression results.
Control variables
Control variables were collated by ourworldindata.com (Rosler et al. 2020). These variables included percentage aged over 65 years from the World Bank's World Development Indicators, population sizes of nations in 2010 are from the United Nations Department of Economic and Social Affairs. Percentage urban population for nations in 2017 come from the World Bank's development indicators (Ritchie and Roser 2018). Obesity is measured as the percentage of the population aged over 18 that have a BMI greater than 30, using data from the World Health Organization (https://ourworldindata.org/obesity). We measured exposure to SARS during the 2002–2004 outbreak as a dummy variable, where a country is assigned a one if they had at least one case and a zero otherwise (WHO 2020). The countries effected by SARS are: China Hong Kong, Taiwan, Canada, Singapore, Vietnam', United States, Philippines, Thailand, Germany, France, Australia, Malaysia, Sweden, Great Britain, Italy, India, South Korea, Indonesia, South Africa, Kuwait, Ireland, New Zealand, Romania, Russia, Spain and Switzerland.
Cultural factors, including secular-humanism (RAT), openness to minorities (COS) and trust in institutions (INST), were derived from multivariate statistics and the World and European Values surveys (WEVS) data from 109 nations (EVS 2011; Inglehart and Welzel 2005; Ruck et al. 2018, 2020a, b; WVS 2020). The WEVS data are derived from the same 64 questions in the five waves of these surveys at 5-year intervals since 1990, administered to 476,583 participants from 109 different nations. These data were compressed into multivariate factors in two steps. The first used Exploratory Factor Analysis (EFA) to identify nine cultural factors underlying the WEVS data. From the EFA step, we summarized the common variance in the WEVS data and thereby remove the portion of the total variance that is likely to be measurement error or other forms of statistical noise. We then used the EFA factor loadings as weights for Principal Component Analysis (PCA), as the orthogonality of the principal components is advantageous for our subsequent regression modelling.
Here we have used the first three of these cultural components, labelled: Trust in Institutions, INST, Cosmopolitanism, COS, and Secular-Humanism, RAT (Ruck et al. 2020b). These components were interpreted based on the correlated cultural factors from the raw survey questions. Trust in Institutions, INST, was correlated with cultural factors such as confidence in institutions (r = 0.58) and interest in politics (r = 0.86). Individuals with high trust in institutions report high confidence in institutions like the media, the army and government and also have an active interest in politics. Secular-Humanism, RAT, is correlated with secularism (r = 0.76), political engagement (r = 0.62), respect for individual rights (r = 0.59) and low prosociality (r = − 0.45) (Ruck et al. 2018). High RAT reflects survey respondents who reported, for example, that religion is not important in their lives, that they are likely to attend protests or sign petitions, they only pay taxes when coerced and believe that homosexuality and divorce are justifiable (Inglehart and Welzel 2005; Ruck et al. 2018). Cosmopolitanism, COS, is correlated with the exploratory cultural factors for ‘openness to out-groups’ (r = 0.78), ‘openness to norm violators’ (r = 0.78) and ‘subjective well-being’ (r = 0.43). High COS implies willingness to have neighbours that are immigrants, from another race, homosexual or from other stigmatized groups; as well as self-reporting happiness and life satisfaction (Ruck et al. 2018, 2020b).
Principal component analysis
For principal component analysis (PCA), we use the ‘Factominer’ and ‘Factoextra’ packages in R to compute the contributions (Table S1) and loadings (Table S2) of the principal components. The PCA included all variables for 83 nations excluding Kosovo, Serbia and Montenegro, North Ireland, and Taiwan, which lacked urbanization data.