3 min read
Country characteristics and Covid mortality
The severity of the Covid pandemic has varied sharply between countries. Stefan Gerlach and Joaquin Thul ask what factors may explain these differences. They find that as much as about half of the variation between countries in Covid mortality is explained by four variables: the median age of the population, tourist arrivals, inequality and a measure of government effectiveness.
Covid-19 has had huge social and economic consequences. The severity of the Covid pandemic, as captured by deaths per million inhabitants, has varied sharply between countries. What factors may explain these differences in mortality? Do they imply that some countries managed the pandemic better than others?
Structural determinants of mortality
To assess the determinants of mortality, a data set comprising 135 countries is studied. Variables are divided into three areas:
1. Susceptibility to the virus:
Reflected by the median age of the population. Mortality is likely to be higher in countries with an older population.
2. Risk of contagion:
Captured by population density and a measure of the extent to which a country may be disproportionally at risk from contagion. The rationale is that higher population density may be associated with more contagion and therefore higher mortality.
3. Policy and Governance:
Domestic policy choices and the ability of governments to manage a crisis may also have played a role in the pandemic. These factors may be encompassed in variables that capture income inequality, a series of governance metrics, the stringency of lockdowns and level of GDP.
Our analysis of the number of deaths per million inhabitants over the period April 2020 to August 2021 against the median age of the population and shows that there is a positive relationship between these variables (correlation = 0.64). However, there is substantial variation between countries and other factors will have been important too.
To explore the importance of other factors, a simple statistical exercise, stepwise regression, is performed. In this procedure, all the potential explanatory variables discussed above are considered and included if they satisfy a statistical criterion. While the relationship between mortality and stringency can go either way, it is included in the list of potential explanatory variables.
About half of the differences can be attributed to a set of unsurprising factors. Thus, the median age of the population, the extent of tourism, government effectiveness (as captured by their ability to combat corruption) and inequality all matter.
Assessing country performance
The statistical model accounts for about half of the variation in mortality, implying that other factors have mattered as well. Table 2 shows the top 20 and bottom 20 countries, ranked according to the percentage difference between actual and predicted Covid mortality. These differences are due to relevant factors that have been omitted from the models, such as a government’s skill or luck in managing the pandemic.
Reporting errors are also likely to play a role. Reporting errors seem large, in particular among low income or emerging economies, as suggested by the fact that some countries report much lower mortality than their neighbours.
Download the full edition of our Infocus publication here.