COVID-19 is an age-discriminating disease. So, all else equal, we would expect that countries with larger elderly shares in the population would be worse off during this pandemic and suffer higher mortality rates. That may appear to be the case, but the big proviso is that not all else is equal.
The chart above shows a scatterplot of the share of the population aged above 60 in the total population against a measure of the mortality rate. We use two different measures: the cumulative officially reported COVID-19 mortality rate and the cumulative estimated excess death rate, both per 100K people and where we use the mid-point estimate of the excess death model of The Economist.
The chart shows a clearly negative relationship between age structure and both measures of mortality. However, the relationship with the excess death rate is much less negative and appears almost flat. In other words:
But if we use the excess death data, the age-structure advantage disappears almost entirely.
This may look like an unsurprising result since we’d expect that countries with younger populations are generally poorer countries, where civil registration and vital statistics (CRVS) systems are generally weaker, and thus the pandemic may likely be considerably underestimated.
But there’s more to it than that.
Worse epidemiological odds
The difference between reported and excess deaths cannot be due to differences in data quality alone. The chart suggests that there’s something else going on than poor measurement. It shows the distribution of mortality across the World Bank income classification (high-income, upper-middle-income, lower-middle-income and low-income countries) as well as for the developing world as a whole (comprising all countries except high-income countries). It shows the mortality distribution based on two concepts: all-else-equal simulations and excess death estimates.
The first bar (in blue) shows the all-else-equal simulations. These are “post-pandemic” estimates of the mortality distribution under the assumption that everyone faces the same epidemiological odds in terms of getting infected and facing death once infected. In other words, infection prevalence rates (IPRs) are constant and identical across countries, age cohort and over time; age-adjusted infection fatality rates (IFRs) vary across age cohorts but are the same across countries and over time.
Because everything else is kept constant, the simulations capture the effect of demography. The epidemiological parameters interact with population size and age structure and that produces an estimate of the fatality toll. And because we look at a distribution, we divide that fatality toll by an estimate of the global toll. Through that division the level of the infection prevalence rate does not matter (it washes out in the numerator and denominator). The only thing that matters is demography. More details on the methodology underpinning this chart can be found in our paper.
The second bar (in red) then shows the mortality distribution through the lens of excess death estimates. These concern estimates of all-cause mortality over and beyond what one would expect in “normal times”. They do capture more broadly the effect of the pandemic as the estimates are not purely limited to COVID-19 mortality.
The fact that the excess death share of the developing world is well above the all-else-equal simulations of the mortality toll suggests a combination of two things: either epidemiological odds have been much worse for the developing world (i.e. higher infection prevalence rates and/or higher age-adjusted infection fatality rates) or the non-COVID-19, indirect effects captured by excess deaths have been more pronounced.
This then also explains the chart at the beginning of this post: developing countries enjoy a demographic advantage but this advantage has waned and mortality is higher than what demography would have suggested.
To conclude we show in the above chart how this discrepancy varies across World Bank regions. Notice how excess death shares are a lot lower relative to simulations for East Asia & Pacific (EAP), which resonates with the finding that epidemiological odds have been better there. The opposite is true for South Asia (SAR), which due to the outbreak in India saw a huge pick-up in excess deaths. This is the pattern for most developing regions, including Latin America & Caribbean (LAC), Sub-Saharan Africa (SSA) and the Middle East & North Africa (MNA).