COVID-19 has led to a tragic loss of life. As of July 6, there have been over 128,000 confirmed deaths due to COVID-19 across Latin America and the Caribbean. Unfortunately, it is likely that this number is a vast underestimate of the true death toll from the pandemic. As the researchers at Our World in Data explain, there are several reasons why the number of confirmed deaths due to COVID-19 may differ from the pandemic’s true death toll. These reasons largely fall into two buckets: the way COVID-19 deaths are reported (for example, reporting only those deaths which occur in hospitals, reporting only those deaths for which a COVID-19 test was performed, or overall poor quality of death reporting systems) and the way COVID-19 impacted the number of deaths occurring due to other reasons (for example, increases or decreases in the number of deaths due to other reasons as a result of COVID-19 induced changes – such a decrease in road accidents as a result of lockdowns).

Data on “excess mortality” provide one way to approximate the pandemic’s true death toll. Excess mortality is defined by the WHO as “mortality above what would be expected based on the non-crisis mortality rate in the population of interest. Excess mortality is thus mortality that is attributable to the crisis conditions.” Excess mortality data thus relies on historical data to estimate the number of deaths we would have expected to take place for given a time period in a normal year—thus providing a counterfactual scenario to compare the number of actual deaths taking place.

In the context of COVID-19, by comparing the actual number of deaths (from all causes) recorded to the number of deaths we would have expected to see in the absence of the pandemic, we can shed some light on the pandemic’s true death toll.

This #GraphForThought uses the excess mortality data compiled by The Economist to zoom into four LAC countries (Brazil, Chile, Mexico, and Peru) and see how excess deaths (shown in grey) compares with the number of confirmed deaths from COVID-19 (shown in red). Depending on the country, the data is shown at the national level (Chile, Peru) or sub-national level (Brazil, Mexico) and is reported in weekly (Chile, Mexico) or monthly (Brazil, Peru) intervals. If the actual death toll were entirely accounted for by the number of confirmed COVID-19 deaths, we would expect the grey curve and the red curve during this period to be roughly the same. However, what we see is that in all locations shown here the number of confirmed deaths from COVID-19 vastly underestimates the actual death toll from the pandemic (the red curve is consistently much lower than the grey curve).

If we calculate the share of excesses deaths that are unaccounted for by confirmed COVID-19 deaths, this gives us a rough estimate for the possible degree of underreporting. The figure below reports this indicator for each country during the relevant reference period. As we can see, there is quite a bit of heterogeneity both across and within countries regarding the degree of underreporting. Within Brazil, for example, the cities of Manaus, São Paulo, and Rio de Janeiro have all been hotspots of the virus. However, in Rio de Janeiro only 32% of excess deaths are unaccounted for by confirmed COVID-19 deaths—whereas in Manaus, this share is over 80%. This suggests that in Manaus, the actual death toll from the pandemic may be as much as five times higher than reported. A similarly high rate of underreporting in observed in Peru and Mexico City.

There are many reasons why underreporting may be happening. One primary reason may be the limited capacity to test in many countries in the region. However, it is important to note that underreporting is not just a result of low testing rates. For example, despite having the highest rates of COVID-19 testing per capita in LAC, over 60% of excess deaths remain unaccounted for in Chile and Peru. We may think that in the case of Chile that this is driven by the reporting methodology for COVID-19 deaths, which during the indicated period did not account for people who died outside of the hospital (for example, those who died in their home) or those who died without a confirmed test. Note, however, that Chile recently announced changes in its reporting methodology.  In the case of Peru, we may think that the quality of testing is a barrier to more accurate reporting—as the expansion of lower quality “rapid tests” may have reduced the likelihood that deaths due to COVID-19 were correctly diagnosed.

Moreover, we must be cautious in interpreting these numbers as the true degree of underreporting—as the accurate reporting of COVID-19 deaths is not the only reason that the grey and red lines may differ. As mentioned before, it is possible that COVID-19 has also affected the grey line by changing the rate at which people are dying from other causes. Indeed, COVID-19 containment measures may have led to changes in overall mortality rates due to impacts such as reduced road accidents (if mobility declined during lockdown), increased femicide (if domestic violence spiked during quarantine), or increased deaths due to other health-related issues (if hospitals became overwhelmed and health-seeking behavior changed).

In order to effectively combat COVID-19, it is critical that we know the true extent of the challenge that we are facing. The difficult and urgent policy decisions that governments are making are compounded by limited and imperfect information. Improving the quality of this information requires extensive testing, robust reporting systems, and transparent statistics. Some governments are more equipped than others to invest in rapidly developing and deploying this type of approach—and given need to act quickly, all governments are learning while doing. While data on excess deaths is not a perfect measure of the true COVID-19 death toll, it is one indicator that can help us to sharpen our understanding of the grave situation unfolding on the ground. It can help us to reflect on what official statistics may (or may not) currently be able to tell us, and how we might be able to improve upon existing information systems going forward. This virus has already claimed too many lives—many more than we know—and we must continue to do all we can to prevent further loss.