The "typical life expectancy" is not a terribly useful number (unless you are considering emigration). A much more useful measure is the life expectancy at a person's current age (which will always be greater than their current age).

And I note it is quite easy to figure out how many people you would expect to die from complications other than Covid---as a first order approximation it will be the same as in 2019 (unless you happen to know of a significant change other than COVID).

> And I note it is quite easy to figure out how many people you would expect to die from complications other than Covid---as a first order approximation it will be the same as in 2019 (unless you happen to know of a significant change other than COVID).

Correct, you could record the number of deaths due to heart disease, cancer, diabetes, etc. and subtract depressions in these death categories from Covid deaths. Subtract away reductions in non-covid deaths from the Covid death toll to account for the people that would have died absent Covid.

My point is the linked NYT article didn't do this correction and it's wrong to write that they did.

And how do those numbers compare?

Life expectancy is a population average. You don't die just because you reach that age.

There are quite a few studies showing that life years lost due to Covid are substantial even for the eldest.

True, but it's drastically more likely that someone over 65 would have died over the course of the pandemic (which is approaching several years at this point) than someone between 18-65. Years lost from elderly deaths may be "substantial" but it's an act of wilful ignorance to pretend that it isn't a lot smaller than years of life lost from a 20 year old dying.

But also, the subset of people who died from Covid is almost certainly not akin to a random sample from the broader population of the age cohort (and therefore not a subset which you could apply those life expectancy averages to).