Nobody knows for sure what the Case Fatality Rate for Covid-19 is in the general population. (It changes based upon the demographics of the population and the quality of healthcare in the region, and can not be accurately calculated until a large number of people have fully recovered.) Currently the US estimate is hovering around 6% of those infected die, but this is likely inflated by the fact that many cases are asymptomatic and/or are never confirmed by a test.

Estimates from various countries have ranged from 0.07% in Singapore to 6% in the United States. Because many cases may be un-diagnosed this 6% estimate is likely to be high. (In Italy it was 14%.)

But due to the involuntary experiment currently spreading through the prisons of Ohio, along with good reporting by the state, we can generate an estimate of the CFR within the Ohio prison system.

On May 22nd 2020 when I downloaded DRCCOVID-19Information from the https://coronavirus.ohio.gov/ website, the corrections system had 4090 prisoners recovered from COVID-19, with 61 deaths (and 277 inmates still infected).   If we discount the inmates who are still infected (not yet recovered) and only look at those who have recovered (or died) we get a case fatality rate of 61 / (61 + 4090) = 61 / 4151 = .01469525 or 1.469 percent (you can round that up to 1.5%)

Of course, if 200 of those 277 “not yet recovered” are currently on ventilators, this CFR number will go up in the future….but this are the numbers we have to work with today.

So 1.5% of Ohio prisoners who were infected with Covid-19 have died.   (Ohio has 35,464 prisoners in isolation, so approximately 11.7% of them have been infected).

This CFR is 15 times higher than that of the seasonal flu (0.1%).   Of course, because prisoners are not generally representative of the US population (the demographics of prisoners skew young, while oldest adults are more frequently killed by covid-19) this CFR will not be the same as for the general population. But it does give us a ballpark number to think about for a lower limit.

If we take 1.5% CFA times 11.7% infected times the 330 million population of the US, we would get 579,150 deaths. If we let it spread to 60% of the population, that would be 2.97 million deaths.

Of course, other research has estimated the CFR in the US to be hovering around 6%.
But that is likely due to missing many un-diagnosed cases because of lack of testing.
So I would use the 1.5% as a lower bound on the CFR in the general population.

Luckily, most Americans are not housed in prisons and can social distance to more effectively prevent the spread of covid-19, so we have not gotten anywhere near 11.7% of the population infected. Yet.

It is also uncertain if prisoners get better healthcare in prison than the general population, but my guess would be no (at least for the 89% of people who could afford health insurance before the pandemic hit).

Related evidence:

• “Among the 3,711 passengers on the Carnival Corp-owned Diamond Princess cruise ship that was quarantined in Yokahama, Japan in February, 712 tested positive for the coronavirus, according to the CDC. Of the passengers who tested positive, 47% were asymptomatic at the time of testing; nine ultimately died. ”
  9 deaths / 712 confirmed positive = 1.26% case fatality rate. [Source]
• A related scientific paper with similar numbers (1.3%)   can be found here.
• According to the Marshal Project, on May 24th 2020, at least 415 inmates have died in the US, with at least 29,000 infected (which is a CFR of 1.43%).