Read on Twitter
In terms of YoY change in monthly deaths, the three biggest post-WWII death spikes in the United States are December 2020, April 2020, and August 2020.
Holiday interruptions mean recent data quality is low so I haven't been doing death days and still won't do a detailed thread. But y'all the high-level indicators are real bad.
April's peak weeks are worse than December's, and its low weeks are lower than December's, so it's a different kind of animal. But y'all December is shaping up to be worse than April, though we'll see how real values end up coming in.
Although on the plus side compared to the 1918 pandemic we actually genuinely did flatten the curve!
I don't know when I'll update my Europe data maybe today maybe tomorrow but I expect to find that Europe is doing horribly too, which is really encouraging because all that really matters is we beat the Euros. #America
best thing about this site is somebody is gonna give me @ s about that tweet
You are MORE likely to die of COVID conditional on getting it, than you are to die of a car accident **conditional on actually getting in a car accident**.
We've had 440k excess deaths from 70 million infections, compared to about 36k car deaths from about 9 mil ppl in crashes.
We've had 440k excess deaths from 70 million infections, compared to about 36k car deaths from about 9 mil ppl in crashes.
That's an IFR from COVID of about 0.6% and an IFR of "actually getting in a car accident" of about 0.4%.
We legally require car insurance, seatbelts, huge safety costs for car manufacturers, speed limits, police enforcement....
We legally require car insurance, seatbelts, huge safety costs for car manufacturers, speed limits, police enforcement....
Y'all you can go to jail for speeding on an empty road where there's nobody around.
I'm just saying, when people say "the COVID survival rate is 99%" or whatever they clearly do not understand what the survival rate is for, like, anything else on the planet
yall the survival rate from *getting shot* is >90% (getting shot in the hand/foot is a large share of gunshot wounds).
I have very strong not-getting-shot preferences! You probably do too!
I have very strong not-getting-shot preferences! You probably do too!
No but you can catch a car crash from walking outside https://twitter.com/the__dude98/status/1346526647651102727
By the way my official view of the Casey Mulligan deaths of despair paper is lol are you serious
Okay so I slept on it and I decided that dealing with this paper by @caseybmulligan was worth making time for.
Mulligan says that "the pandemic and recession" are associated with 10-60% higher deaths of despair.
What does this mean? https://www.nber.org/papers/w28303
Mulligan says that "the pandemic and recession" are associated with 10-60% higher deaths of despair.
What does this mean? https://www.nber.org/papers/w28303
Well, let's first note that he has no actual measure of "deaths of despair" here. He uses "non-COVID excess deaths" as his standard. So if you died of cancer, he says that's a death of despair.
This is obviously nonsense.
This is obviously nonsense.
"Deaths of despair" conventionally refers to deaths related to negative, self-destructive, or antisocial behaviors: suicide, alcoholism, drug abuse, violence, etc. It does not conventionally refer to, uh, septicemia?
So right off the bat, this paper's whole thesis is pretty shot to hell.
Let me show you how "official COVID deaths," "other natural causes deaths," and "external causes deaths" compare over time vs. the year prior. Almost all deaths of despair should be "external causes."
Let me show you how "official COVID deaths," "other natural causes deaths," and "external causes deaths" compare over time vs. the year prior. Almost all deaths of despair should be "external causes."
WHat I want you to notice is:
1) Non-COVID "natural causes" deaths are tightly correlated with COVID deaths. This is true at the state-week level as well.
2) Most non-COVID excess deaths are NOT external causes deaths (i.e. NOT deaths of despair like suicide)
1) Non-COVID "natural causes" deaths are tightly correlated with COVID deaths. This is true at the state-week level as well.
2) Most non-COVID excess deaths are NOT external causes deaths (i.e. NOT deaths of despair like suicide)
3) Non-COVID natural causes deaths run well BELOW trend in the latest period. That could be an insufficiency in my revision-adjustment method, but these are pretty back-dated datapoints: they shouldn't be so sensitive.
This means we've seen a long-run DECLINE in these!
This means we've seen a long-run DECLINE in these!
i.e. CONTRARY to the false claim that "hospitals stopped treating cancer so cancer deaths rose," we actually see that in the long run non-COVID natural-causes excess deaths have fallen BELOW 2019 levels in recent months!
Based on this, we can establish several key points:
1) It is probably not the case that cancer deaths are actually perfectly correlated with COVID. It is much more likely that a lot of cancer deaths are undiagnosed COVID.
2) Most non-COVID excess deaths are NOT deaths of despair
1) It is probably not the case that cancer deaths are actually perfectly correlated with COVID. It is much more likely that a lot of cancer deaths are undiagnosed COVID.
2) Most non-COVID excess deaths are NOT deaths of despair
The second point is of course a huge problem for @caseybmulligan , since he's counting all of these deaths as indicative of deaths of despair.
But the first is also a big problem! Which brings us to how he deals with COVID undercounting.
But the first is also a big problem! Which brings us to how he deals with COVID undercounting.
To his credit, @caseybmulligan does not repeat false claims of COVID overcounting as a major problem.
But his method of addressing undercounting is not very much explained. He appears to only identify undercounting when other causes on net rise above his excess threshold.
But his method of addressing undercounting is not very much explained. He appears to only identify undercounting when other causes on net rise above his excess threshold.
But this is problematic because 1) as he noted pre-official-pandemic excess deaths were running way BELOW the normal threshold, so your counterfactual has to be shifted to adjust for that and 2) the state-week correlation of other causes suggests cause-level aggregation
So for example, I'm pretty confident that very few car accident deaths are "actually COVID" but misdiagnosed.
But dementia patients who can't remember or describe their symptoms? That could totally be undiagnosed COVID!
But dementia patients who can't remember or describe their symptoms? That could totally be undiagnosed COVID!
And since excess dementia deaths are highly correlated with excess COVID deaths, uh, yeah, guys, a bunch of dementia patients are dying of undiagnosed COVID.
So as a reminder, the association between COVID deaths and other causes *really is* extremely strong, especially for categories where misdiagnosis is extremely plausible (other respiratory diseases, heart failure, dementia, etc).
So, what we've shown so far is basically that 1) @caseybmulligan underadjusts for COVID undercounts and 2) erroneously counts large numbers of non-deaths of despair (and even many COVID deaths!) as deaths of despair.
Okay, so what's the truth? What's REALLY happening with deaths of despair?
Well, one good metric is to look at deaths of external causes! Here's how respiratory deaths vs. external causes stack up.
Well, one good metric is to look at deaths of external causes! Here's how respiratory deaths vs. external causes stack up.
What I want to point out is that the change in respiratory deaths is VERY OBVIOUSLY above trend.
The change in external causes deaths.... is too, but way less so. Here's Ytd over Ytd change.
The change in external causes deaths.... is too, but way less so. Here's Ytd over Ytd change.
Respiratory deaths are up 120% YTD.
External causes deaths are up 16% YTD.
In both 2015 and 2016, these deaths rose by 8%, so 16% is a big increase obviously but given the declines in the last two years hard to parse.
External causes deaths are up 16% YTD.
In both 2015 and 2016, these deaths rose by 8%, so 16% is a big increase obviously but given the declines in the last two years hard to parse.
What I'm saying is, figuring out how much of the rise in external causes deaths is "the pandemic" vs. "the recession" vs. "external causes deaths have been rising for a long time" is not gonna be a trivial task!
So let's go ahead and figure out what those excess deaths really are.
First of all, the VAST majority of excess deaths are traffic and other accidents. A huge share of them are old people falling and dying as a result.
First of all, the VAST majority of excess deaths are traffic and other accidents. A huge share of them are old people falling and dying as a result.
Of course, the "other accident" category is also gonna include drug overdoses, also a very big share!
So let's get a handle on pieces we can figure out.
I have data from a handful of states on traffic fatalities for 2020, since highway departments track that pretty closely. 2020 traffic fatalities are up about 6% vs. 2019. Yes, UP!
I have data from a handful of states on traffic fatalities for 2020, since highway departments track that pretty closely. 2020 traffic fatalities are up about 6% vs. 2019. Yes, UP!
We also have decent statistics on murder. While it of course varies place to place, murder in 2020 was somewhere between 15% and 40% above 2019. (!!!!!!) Let's call it 20%.
Executions and police killings combined were probably not changed a lot vs. prior years. Call it no change.
Medical and surgical complications *most likely* declined, right? Fewer surgeries, especially elective, should mean fewer complications? I'm guessing here but call it a 10% decline.
Data from Massachusetts suggests suicides have been more or less unchanged at least in the first half of the year. https://www.medrxiv.org/content/10.1101/2020.10.20.20215343v1
A study of police reports in Queensland, Australia also found no suicide increase during lockdowns far more intense than what we had here in the US. https://www.thelancet.com/journals/lanpsy/article/PIIS2215-0366(20)30435-1/fulltext
There's a lot of medical research suggesting COVID made suicide risk factors worse.
But the two studies we have with actual suicide data find small *declines* in suicides. This is tricky to sort out!
But the two studies we have with actual suicide data find small *declines* in suicides. This is tricky to sort out!
Conservatively, let's say there's a 5% increase in suicides.
Let's also say there's a 10% increase in "things the cops aren't sure if it's suicide homicide or accident."
Let's also say there's a 10% increase in "things the cops aren't sure if it's suicide homicide or accident."
Gotta pause here as I have a call I have to take. brb
Okay, I'm back.
So this leaves us two categories of external causes deaths: overdoses and other non-drug, non-traffic accidents.
So this leaves us two categories of external causes deaths: overdoses and other non-drug, non-traffic accidents.
Where @caseybmulligan makes a really good contribution by hunting up two actual data series on drug related deaths. His data suggests they're up about 20-50%. Let's call it 40% for the year.
If we then assume that total external causes deaths must rise 16% (since that's what the CDC data seems to show), the residual must be "other accidents."
So how does this all shake out?
So how does this all shake out?
Here's the "best guess" of external causes deaths by type in 2019 and 2020.
The biggest category here, with one of the largest increases, is the residual: basically old people falling.
The biggest category here, with one of the largest increases, is the residual: basically old people falling.
So we can then add up suicide and drug overdoses along a longer time series. And here's how they compare over time and to respiratory deaths.
Here's yoy absolute change.
Key to remember here respiratory deaths include flu, which is way DOWN this year, and many deaths caused by COVID (i.e. misdiagnosed COVID deaths) are NOt counted as respiratory deaths. So that increase in respiratory deaths is 100% certain to be much lower than true COVID deaths
Okay, so, I said 28,000 deaths of despair in 2020. @caseybmulligan said 30,000 deaths of despair March-October, and he says this is pretty much entirely among people 15-54 because he shows zero NCED for old folks.
He also says 2018 had 66,000 deaths of despair. In reality, 2018 had 110,000 suicides and overdoses. What it had 66,000 of was drugs and overdoses ***among people aged 15-54***.
So @caseybmulligan , while a bit unclear at points, really does seem to be arguing that March-October suicides and overdoses among 15-54 year olds rose 30,000 above trend.
This is clearly wrong. As I've shown, deaths of despair probably only rose 28,000 *in total* for the year. In 2019, just 70% of these deaths were among 15-54s, which suggests a whole-year increase of probably 20,000 for this group, not 30,000.
And since I'm doing the whole year, and March-October is not the whole year, it's probably the case that deaths of despair March-October rose by about 10-18,000.
Okay.
That's not a wholesale rejection of this paper.
But, y'all. Saying, "drug overdoses are up this year" isn't exactly a super controversial point.
The question is WHY they are up.
That's not a wholesale rejection of this paper.
But, y'all. Saying, "drug overdoses are up this year" isn't exactly a super controversial point.
The question is WHY they are up.
If all you want to do is say, "deaths are up!" the exercise I did of traffic/drug/murder/external causes is the whole paper. You're done. You don't even need to deal with COVID at all. Just take the external causes category from the CDC data and decompose! (cc @caseybmulligan )
Of course, the reason to link to COVID is.... politics!
@caseybmulligan cites City Journal's piece "Death By Policy" twice. As best I can tell it's the only twice-cited piece? Maybe Case and Deaton is twice-cited in disagreement?
@caseybmulligan cites City Journal's piece "Death By Policy" twice. As best I can tell it's the only twice-cited piece? Maybe Case and Deaton is twice-cited in disagreement?
The introduction mentions the "cure is worse than the disease" idea. The conclusion mentions social isolation as a possible driver (though is careful not to overstate). The only reason we're talking about this is that there's a debate about lockdowns.
So let's see if lockdowns actually explain changes in deaths of external causes at the state level!
Here's the correlation between state-level changes in excess deaths and the extent to which their Google-measured mobility changed during 2020.
y'all that's a random scatterplot there's nothing there
SO here's the deal.
Deaths of despair rose in 2020. That's absolutely true. And it's probably somehow related to the pandemic.
But they likely rose by less than @caseybmulligan seems to say, and not because of lockdowns or reduced mobility.
Deaths of despair rose in 2020. That's absolutely true. And it's probably somehow related to the pandemic.
But they likely rose by less than @caseybmulligan seems to say, and not because of lockdowns or reduced mobility.
If @nberpubs would like me to write a response to @caseybmulligan I'm happy to do so but alas I am not an NBER Affiliated Scholar.
I should be clear in all of this:
I think that @caseybmulligan 's paper is *incorrect*. But I'm glad smart people are taking these questions seriously and trying to address them!
I think that @caseybmulligan 's paper is *incorrect*. But I'm glad smart people are taking these questions seriously and trying to address them!
State level unemployment changes are also not predictive of deaths of despair. https://twitter.com/mjmcin/status/1346835167206395904
Okay I'm gonna try to do my Europe data now
