[JoeData]

Here is some more data.

Deaths from all causes has declined rapidly in the U.S. since January. The red dotted line is my best attempt to correct for reporting lag: https://i.imgur.com/z8M0k2h.png

Excess deaths exceeded COVID deaths every week of the pandemic: https://i.imgur.com/MeGMsRx.png

Total excess deaths in the U.S. were about 500,000 for 2020. The excess breaks down as 77% COVID, 16% non-COVID natural causes, and 7% external causes for the whole year. External causes include homicide, suicide, accidents, overdoses/poisonings.

The count of COVID deaths include COVID as underlying cause of death (91%) and COVID as a contributing cause (9%). Even when including the contributing cause, it looks like COVID deaths were significantly undercounted in the first wave March/April/May and somewhat undercounted in the later waves. https://i.imgur.com/YjZeI2E.png

Deaths from most major categories were higher or unchanged: https://i.imgur.com/TDGwVU9.png

Excess deaths by age group: https://i.imgur.com/oKda9Dl.png

Ages 0-14 had lower than normal deaths in 2020.

Non-COVID excess deaths by age group: https://i.imgur.com/8pJnauz.png

[gmiller123456]

>Excess deaths exceeded COVID deaths every week of the pandemic: https://i.imgur.com/MeGMsRx.png

Not entirely sure how to parse that statement, but I think most people would interpret that as "non-COVID excess deaths exceeded COVID deaths", which doesn't match what's on your graph. E.g. for week 15, expected deaths = 55k, deaths - COVID = 63k and all deaths=79k.

The only truthful way to interpret your statement is that "excess deaths (including COVID deaths) exceeded COVID deaths". Which isn't that interesting of a statistic.

[JoeData]

> "excess deaths (including COVID deaths) exceeded COVID deaths"

Yes, that's what I meant. I didn't think it was ambiguous. Lots of people think that COVID deaths were exaggerated. There are countries (with not very many COVID deaths) where the number of excess deaths is less than the number of COVID deaths. There are even countries with negative excess deaths for the whole year, even though they had some COVID deaths.

[extra88]

I'd expect lower or negative excess deaths in countries with very effective COVID responses because those responses can also prevent other causes of death. Masks, hand washing, social distancing also prevent the flu and other diseases. Lockdowns and working from home can reduce commuting and traffic-related deaths (though I heard in some places in the U.S. car deaths were up, possibly because drivers were driving faster and more recklessly on the lightly trafficked roads).

[joshuamorton]

It is very interesting. It means that either covid deaths were undercounted, or something else unique to 2020 also killed a large number of people, or both.

[gmiller123456]

It appears as though you're using the first interpretation, and assuming it's true. It is not. I agree it'd be very interesting if true.

The second interpretation (which is the one I called "not interesting") just means there was at least one non COVID death in each interval, and the total deaths never dropped below the average number of deaths. One could reasonably assume there was at least one non-COVID death per week, making that part of the statistic useless, and you could just state the total deaths.

[joshuamorton]

No I'm using the second interpretation. It doesn't mean that there was one non covid death on each interval, but that there was one non covid death above the baseline of the expected number of deaths. Or in other words, deaths were always higher than expected, even after accounting for deaths attributed to covid.

The number of unattributed excess deaths is fewer than the number of covid attributed excess deaths, bit it should be zero, because the null hypothesis is that the only things killing people are "the same stuff as last year" and covid.

[gmiller123456]

>there was one non covid death above the baseline of the expected number of deaths

While a true (and interesting) statement, that's not what the OP said. He said "Excess deaths exceeded COVID deaths", not "Excess deaths minus COVID deaths exceeded expected deaths". Again, the second statement is true (based on his graph), but not all all what he said.

[joshuamorton]

> He said "Excess deaths exceeded COVID deaths"

Yes, and this is the right statement!

Excess deaths minus covid deaths should be zero, because the null hypothesis is that covid is the only thing contributing to the excess deaths.

That there are additional excess deaths beyond covid deaths is interesting (and means, as I originally said, that covid deaths are undercounted, or that some other unknown force is killing people).

The statement you provide as an alternative "Excess deaths minus COVID deaths exceeded expected deaths", doesn't make sense. Excess deaths are already above the expected deaths.

Let me run an example.

Last year, 100 people died each month. This is the expected number of deaths. This year, 150 people died each month. This gives us 50 excess deaths. 40 deaths are attributed to Covid. So, 10 are of unknown origin. This is interesting!

If the excess deaths - covid deaths exceeded expected deaths, you'd have to have 40 covid deaths, 110 deaths of unknown origin, and 100 "expected" deaths, which would also be quite interesting, but abjectly terrifying too (the death rate would need to more than double).