Indeed, because the risks from the virus are not distributed evenly among the population, we shouldn’t be trying to get a single fatality rate at all. We already know that the elderly are a highly vulnerable population — and we think we know how their mortality compares to those of younger cohorts. But we won’t really know until we know how widely the virus has spread through both age groups; if asymptomatic cases are more common among the young, for example, then we may still be overestimating the risks the virus poses to them directly. We need to be doing enough antibodies tests to do the same by race and by sex, to determine whether African Americans and men of all races are more likely to die of COVID-19 because of biological factors that make the virus more deadly (such as different rates of hypertension or differences in immune system function) or because of sociological factors that make them more likely to catch the virus (such as being more likely to still be working outside the home).

And we need to be doing serological testing by occupation — not so we can hand out “immunity passports” (which are likely a long way off, if they ever become plausible), but to properly assess risk in the first place. How widespread is the virus among transit workers? Among health-care workers? Among grocery clerks?Among agricultural and food-processing workers? Looking at elevated death rates tells us half the picture; the other half is knowing the degree to which that elevated death rate is due to the virus having already infected far more people in these jobs than among the general population, as opposed to conditions of the job (such as exposure to a higher viral load) that might make the virus more deadly.