i mean by my count is 550,000 but YMMV
The math on this pretty simple.

First, you take CDC's reported deaths. Then you make an adjustment for underreporting of recent deaths. Various ways to do that I won't get into here, but all methods yield similar results.
Then you compare deaths in each week to deaths in each week in prior years.

Various ways to set a baseline. But one uncontroversial way to set a *maximum* expectation for "normal deaths" is to take:
1) maximum deaths in a given week 2014-2019
2) multiply by pop growth
Multiplying by population growth *itself* can create endogeneity and is an unknown parameter, so we can multiple instead by the average yearly change in deaths 2014-2019, about 1.7% increase each year due to population growth and aging.
So here's what actual deaths look like compared to if we assume every week of the year was as bad as the worst that week has been in any year 2014-2019, but then was actually 1.7% worse than that worst week.
The space between that "every week is the worst ever" line and the "what actually happened" line adds up to 450,000 deaths.

In other words, the *absolute minimum* excess death count since late February is 450,000.
However, this is a silly standard. Assuming that the "normal baseline" for 2020 would have been "the maximum we've observed for each week times 101.7%" is a bit odd.
It's not clear to me why you'd assume that the baseline expectation for deaths in 2020 was that every week would individually be the worst it's ever been.
Furthermore, it's pretty clear that COVID excess deaths were piling up from a below-zero baseline. Look where circled here. Normally, even in low-flu-death years, deaths decline during February. But they stayed flat! Flu deaths falling, but COVID rising, but from a below-0 start.
So if you make a more plausible baseline for 2020 based on not just arbitrarily assuming it was always going to be the worst ever all the time but imagine a world where it might have been sometimes-not-the-worst, you get this alternative baseline. Which yields 485,000.
But this is still very weird. Because it's still assuming a relatively severe flu season. What we actually observe in reality is that there's basically been zero flu season. Of course, we know why that is: because COVID precautions have reduced flu deaths!
But if we start getting into this thorny issue of changing cause of death, we would also need to adjust for things definitely not related to COVID, like, say, car accidents.
So, let's do it, and show once again as I have shown a million times that these excess deaths are overwhelmingly COVID, not suicide and drug overdose.
So here's major cause categories. What you can see is respiratory deaths are SUPER HIGH and other natural causes deaths are SUPER HIGH and mostly correlated with respiratory deaths and external causes deaths are kinda high too!
But the key thing to understand is that in the first two waves, those excess natural causes deaths ***are COVID deaths***. They just didn't get appropriately counted. They occur in the same times and states and in the same age groups as COVID excess mortality.
I'm not gonna bother demoing that here since I'm demo'd it about 10 other times. Suffice to say: the spike in non-COVID natural causes deaths IS COVID DEATHS which are being under-reported.
But comparing to 2019 is crude. Let's once again built a more plausible counterfactual accounting for pre-trends!
So now we're separately modeling excess deaths for respiratory, other natural, and excess deaths.
If we do that, we get an interesting result. It turns out the rise in external causes deaths is.... pretty normal based on the underlying time trends and the pre-COVID period! But trends for others are not.
Now, cause of death data, even with adjustments for incompleteness, is only viable thru December 12. So it turned out, through that period, I had estimated 391,000 excess deaths.

We can attribute 378,000 excess deaths to non-external causes.
But actually more if we allow our baseline to be a "low flu season" rather than a "high-ish flu season."
There have been more deaths since then. And if you suppose that 2020 was actually gonna be a low death year (plausible since life expectancy improved 2018-2019), then your baseline falls lower.
It is by assuming a baseline of continued life expectancy improvements similar to 2018-19 for 2019-20 that you get my very high figure of 550k. I know not everybody will buy that. But I think 2020 really was shaping up to be a good year in terms of health!
What little lab data we have even suggests our flu vaccine was going to be highly effective; i.e. that the labs picked the "right strains" of flu to target!
So I think the correct counterfactual is actually "a year of considerable life expectancy improvements." Which is why I get 550k. But as I said, YMMV. You can plausibly be from 450k to 550k.

But 400k is too low.
GUESS WHO JUST GOT HIS HANDS ON COLOMBIAN WEEKLY MORTALITY STATISTICS

THIS GUY

THIS GUY DID
gettin' a few others while im at it too
Anyways for those following the all-important "are we beating the Euros?" question....

.... in terms of weekly death rate, we are dying more than the Euros!!!!

.... but in cumulative terms we probably still have'em BARELY beat
My Latin American set includes Peru, Mexico, Brazil, Colombia, Chile, and Ecuador, by the way. Really would like Uruguay and Argentina but can't find the data. Costa Rica and Panama would be nice too.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Lyman Stone 石來民

Lyman Stone 石來民 Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @lymanstoneky

15 Jan
When people think of "extremely functional microstates that just work for hard-to-explain reasons" they always leave out Mongolia and it's infuriating.

Mongolia has had *negative* excess deaths in 2020 y'all.
When we give examples of successful introduction of participatory demcoracy, rapid economic development, etc, we should really be giving Mongolia as a textbook example.
I am partly pro-Mongolian because they have anomalously high TFR for their income level, because Mongolians love the babies.

Also, lowkey, this is, um, a thing that is happening:
Read 26 tweets
14 Jan
I get the desire to punish everybody associated with Trump but "a bunch of people who've been running the country for 4 years and have a bunch of political supporters and cash being angry and bored" is perhaps not actually the ideal outcome.
One of the issues to reckon with is that Reconstruction didn't go far enough, but attempts at strict Denazifiction were widely regarded as mistakes. De-Baathification has been seen as a catastrophe for IRaq.
I'm not sure what defines the differences. But study of authoritarian regimes most definitely does *not* support the idea that you should always purge as much as possible!
Read 7 tweets
14 Jan
listening to Dan Carlin's Supernova in the East and super irritated that he focused on Hiro Onoda instead of the 10-million-times-more-interested Teruo Nakamura.
The last Japanese holdout to ever surrender was not Japanese imperial fanatic from the home islands but...

.... a Taiwanese aborigene who decided not to surrender with his unit but just build himself a hut on a deserted island and lived quietly until accidentally discovered
Also I guess I shouldn't call him Teruo Nakamura since he did not actually speak Japanese (or Chinese!). His name in his own language was Attun Palalin.
Read 13 tweets
13 Jan
Saw @mattyglesias tweet about a poll of QAnon support (linked). I think it's a *bit* mistaken tho: the table he shared was QAnon support *among those who had heard of it*. But TONS more liberals have heard of QAnon than conservatives!
Here's the net favorability of QAnon (double-weighting the "very" folks) accounting for differences in who's even heard of QAnon.
Key to understand is Democrats and liberals are EXTREMELY anti QAnon not only because they are very unfavorable to it but because large shares of Rep/Con folks ***have never heard of QAnon***.
Read 22 tweets
13 Jan
I think it's fair to criticize US support of the Saudis in this war.

I also think it's unreasonable to argue that we should be providing even humanitarian aid in a region where we know both sides will steal it and use it to enhance their exploitation of the local people.
During the US Civil War, the Confederates starved because the Union starved them.

Food aid to the Confederacy would not have gone to slaves, folks. It would have gone to slave-owners.
It's possible to say both "we should not support the inhumane Saudi war effort" and also "if the Houthis want Iranian support, let Iran feed them."
Read 13 tweets
13 Jan
reading the self-important and vaguely-cultish textbooks by bayesians makes me want to go out and kill a sufficiently large number of bayesians that i can estimate the mean pitch of their screams
"when we have multiple models, we should choose one using Bayesian statistics"

no

you should do both and publish an appendix showing robustness tests, you cultist
"but my bayesian model takes 4 weeks to process i can't run 1,397 robustness tests"

yes

exactly
Read 33 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!