STUMP » Articles » Mortality with Meep: Estimated U.S. Total Deaths for 2020, All Causes » 4 January 2021, 04:21

Where Stu & MP spout off about everything.

Mortality with Meep: Estimated U.S. Total Deaths for 2020, All Causes  

by

4 January 2021, 04:21

Key takeaway: I estimate about 3.3 million deaths for 2020 in the United States, with my range being 14% to 16% higher than total 2019 deaths.

Absolute floor for total U.S. deaths: 3.1 million

The CDC, as of December 30, 2020, had recorded over 3 million deaths for 2020.

If you go to their main table for total and excess deaths for 2020, you will see that they’ve recorded 2,913,144 deaths from all causes as of December 31, 2020. But that starts the week ending February 2, 2020. Also, it actually doesn’t include anywhere near all the deaths of December.

They don’t show you there the total number of deaths for 2020. They exclude all the weeks ending in January.

To get prior weeks, you have to go to a different data set. They actually have multiple sets one can use (and they’re all slightly different), but the one I pulled from is here: Weekly counts of deaths by jurisdiction and age group — these data are as of December 30, 2020.

Here are the totals for January, moving backwards:
January 25, 2020: 59,130 [cumulative deaths: 2,972,274]
January 18, 2020: 59,350 [cumulative deaths: 3,031,624]
January 11, 2020 : 60,703 [cumulative deaths: 3,092,327]

The week ending January 4, 2020, saw 60,145 deaths. However, that week included December 29-31, 2019. If we just prorate it to 4/7 of the deaths, 34,369, then we get 3,126,696 deaths.

These are deaths already reported to the CDC. Mistakes can be made, but usually, the mistake is the exact week in which the death occurred, not that a death occurred at all. Most of the shifts will be to other weeks within 2020. It is very unlikely that they counted a person who is not actually dead.

My floor estimate is 3.13 million deaths. This would require no deaths occurring at all in the last two weeks of December, which is absolutely false. The deaths reported for 2020 will not be lower than 3.13 million.

Historical comparison against 2019

The final 2019 death numbers were recently released: 2,854,838 total deaths in the U.S.

This floor of 3.13 million deaths would be a 9.5% increase over 2019 deaths, or 272K more deaths in 2020 than 2019.

This floor of 3.13 million would be a 10% increase over the 2,839,205 deaths in 2018 or 287K more deaths in 2020 than 2018.

Again, I’m treating this as an unrealistic lower bound. That’s pretending nobody died the last couple of weeks of December, which is obviously false.

First estimate: projected deaths for 2020 via CDC weighting and pre-COVID average

If you look at the actual table on this page, you see that while the total number of deaths through December 5 look plausible, they fall off rapidly after that, due to the lag in reporting (or even discovering) deaths.

The CDC has its own weighted predictions. For my first projection, I used the CDC’s weighted predictions through December 19, and then used the pre-COVID weekly average for weeks/days after that. When I look at that projection from the CDC, it may actually be undershooting. I get an estimate of 3.26 million using this approach.

That’s a 14% increase over 2019, or 409K more deaths in 2020 than 2019.

Second estimate: projected deaths for 2020 via CDC weighting and trending

In this approach, I’m going to capture that some states have much larger lags in reporting deaths than others.

I did an extremely simple trend for each state/jurisdiction (NYC and DC are separate), pushing the trend back to where the data last looked “solid”, and making sure that the total number of deaths by week was at least the average of the pre-COVID weekly deaths. This can bias the projection upwards.

That projection results in 3.33 million deaths, a 16% increase over 2019, or 471K more deaths in 2020 than 2019.

Third and fourth estimates: reversing the trend

So, what if there was a peak of deaths just after the last “solid” reported date? This will make for a smaller estimate for total deaths. I did this with both the unweighted data (reversed after week 47, but that’s definitely a lowball), and weighted data (reversed after week 49).

The unweighted projection, again, almost definitely a lowball, results in 3.24 million.

The weighted projection results in 3.30 million.

This still gives a range of 14% to 16% more deaths in 2020 than in 2019.

Comparison to other total deaths estimates

Unfortunately, I cannot find other people who have projected total deaths for the U.S. for 2020 (I’m sure they exist), but I can find plenty of projections for COVID deaths.

While that’s semi-useful, in the larger scheme of things, dead is dead. If a person died due to a drug overdose because they ODed while “self-medicating” to deal with pandemic lockdown stress, then that adds to the excess mortality.

The CDC has compiled the results from multiple projection models. Here are their results as of December 28, 2020:

Downloading their ensemble prediction set, they assume about 345K – 354K COVID deaths cumulatively through January 2, 2021.

This is less than my projections of 410K – 470K extra deaths in 2020 from all causes. Now, my projections may be a little high, but I think we’re going to be seeing extra mortality from non-COVID causes in 2020, which I will try to tease apart in upcoming weeks.

Mind you, even if only 345K out of 470K excess deaths are due to COVID, that still explains 73% of the extra mortality.

There was a lot of extra mortality in the U.S. in 2020, and most of it was due to COVID-19. [And yes, 14% increased deaths is a big deal. It is over three times bigger than the increase seen in the two prior large pandemics. Even 10% would be a huge increase historically.]

Underlying spreadsheet with projections


Related Posts
Mortality Monday: How Many People Will Die in the U.S. in 2017?
Mortality with Meep: How to Get Lots of Supercentenarians? Pension fraud!
Mortality with Meep: People Continue to Die in the Dominican Republic and the PR Problem Continues