And here's a good long form . . .

discussion of the undercount of Covid deaths. And no, death certificates aren't going to let us get a true final tally. They are an excellent example of the reification of data. The cause of death may be given as pneumonia, or even just respiratory arrest, or it may be attributed to some comorbidity and the contribution of the virus to the death may not be noted. Doctors aren't trained to fill these out in any particular way and again, if the person doesn't have laboratory confirmed Covid-19 it will very likely not be on the death certificate even if everybody knows that's almost certainly what it was.

Outside of the U.S., in countries  with no adequately functioning public health system, we aren't getting any kind of useful count at all.

Update: Just to clarify, in response to a comment (I'm lifting this to the front page) part of the problem is that cause of death is often ambiguous. People have multiple morbidities. If someone who is otherwise not healthy gets finished off by flu or Covid, what do you put on the death certificate? It's a judgment call. The viral infection likely caused death sooner than it otherwise would have occurred, but heart disease or whatever is also a contributing cause. Doctors don't attribute death in a consistent way, and it's also a philosophical conundrum. After all we are all mortal and our life expectancy declines as we get older. You can't actually save anybody's life, you can only prolong it. How long is enough to matter is imponderable.