I didn't have time to do death day yesterday but it'll come in a few minutes.

It's going to be bad news.

One hump

Two humps

THREE HUMPS

Third wave.

Excess deaths around 300,000, R values rising over 1, case counts increasing: at this rate, winter will be awful. Nationally, probably <10% of people have been infected, so herd immunity is faaaaar away.

My very hopeful "normal by thanksgiving" projection is not going to happen. At this point, I'm not sure we can even project "normal by Christmas." Maybe "normal by Ash Wednesday"? But that's also when vaccines should begin to show up.

So really at this point maybe best to project "not normal until a vaccine arrives."

Now, of course, it's always possible that the revisions are over-adjusting. Maybe! I hope so!

But case counts and R values calculated from case counts are rising too, so, eh, I dunno, probably not a fluke?

So are excess deaths really COVID deaths?

Excess deaths are tightly correlated with COVID deaths by week across the country, by state in total, AND by state-and-week together. Also, the "unexplained excess" deaths are correlated with COVID. Ergo, EXCESS DEATHS ARE COVID DEATHS.

Correction to my <10% infection rate: it's still probably under 15%, but not quite as low as I thought; I hadn't updated my IFRs (thanks to @simon_bazelon for noting this).

Including a variable for stricter lockdowns in a simple regression does not yield any significant effect on mortality at all. They do not appear to have radically altered deaths in any way, positive or negative. https://twitter.com/_mengde_/status/1309877293796515841

Also, states with stricter lockdowns do not appear to have had *that* much bigger of economic losses, like rising unemployment, than other states? Seems like overall strict lockdowns just don't do all that much either way.

By the way, I got curious if the extent of under/over-reporting, as measured by excess deaths vs. official deaths, was politically correlated.

It's not. Trump vote share in 2016 is almost perfectly randomly distributed vs. implied death count accuracy.

There does not appear to be *any* measurable partisan bias in death classification rates by state. Stories of huge political interference one way or another are mostly just hack journos trying to make you feel scared because fear is what sells.

Last week, I finally took a look at cause of death data.

I concluded, quite strongly, that external causes of death like suicide and drug abuse simply COULD NOT be waaaay up based on the data we had.

I was wrong.

I made an adjustment for incompletion. I knew external cause reports could be delayed more than other causes. But I had no idea JUST HOW MUCH MORE delayed they were. Today I got an updated cause file allowing me to formally estimate revision timing.

Here's changes in deaths by cause, one graph which doesn't include COVID, the other graph which does.

Excess deaths are up CONSIDERABLY. However, respiratory deaths are still up by WAY more.

So it seems VERY likely that IN FACT deaths due to external causes are up. Many of those deaths are probably suicide or drug deaths. Homicides are probably up too. Traffic fatalities likely down.

This could be due to lockdowns, or it could be due to economic losses, or it could be random noise. Here's external causes deaths-to-date for each year since 2014. The increase this year is definitely large, but they have been rising for a while.

So how do external causes deaths stack up in the grand scheme of things?

Well, here's the *absolute number* change for each cause of death vs. 2019. Note that the heart disease excess deaths are probably COVID too, since COVID can cause heart failure.

Excess deaths are up.

But they are DWARFED by respiratory deaths.

So:
1) I was incorrect in my earlier assessment that changes in excess deaths precluded the possibility of a major suicide/drug OD spike

2) In fact, external deaths are up by a lot, which could be suicides or drug ODs

3) However, Respiratory deaths (COVID) are up by TONS more

A truth which holds in percentage or absolute terms.

Correct! Many deaths attributed to heart disease or dementia or diabetes are actually COVID deaths.

However, external causes deaths almost certainly are NOT, not least because they have a totally different time series! https://twitter.com/bdquinn/status/1309894594109353985

In total we have had 400,000 more respiratory deaths in 2020 to-date than we had in 2019 to-date.

We have had 50,000 more deaths due to external causes.

Here's external and respiratory deaths shown by week and year on identical scales, and by year on a single graph.

The rise in external causes deaths is definitely concern. But it is dwarfed by respiratory deaths (which aren't even all COVID deaths!).

Also, should not jump from "external cause deaths are up" to "lockdowns cause this."

Here's a scatterplot of external cause deaths 2020 vs. 2019, plotted against an index of mobility from Google's data.

Less mobility does NOT lead to more external causes death!

The state with the biggest mobility decline, Rhode Island, had very little increase in external causes deaths. The state with the smallest mobility decline, Mississippi, had a higher than average amount of external cause death increase.

The states with the biggest and smallest increases in excess deaths are mostly middle-of-the-pack in terms of mobility changes.

What we know is that external causes deaths are up.

We don't know for sure what sub-causes are driving this.

But we know it's uncorrelated with COVID deaths or cases, and ALSO uncorrelated with changes in actual mobility.

This tweet I said 400,000 more respiratory deaths.

That was incorrect on review. The correct number is 370,000. https://twitter.com/lymanstoneky/status/1309895219891122178

Regular readers know what's coming next is my comparison to other countries. May take a bit as I also need to eat lunch soon.

Good question, two answers, explained below:

1) Timing doesn't match
2) If that is what is happening, *it's still a COVID death*. https://twitter.com/jdavidwexler/status/1309938863033516040

1) Timing

If outbreaks cause people to avoid hospitals and thus die of other causes, then excess-but-not-officially-COVID deaths should spike AFTER COVID deaths, as people get word of rising cases and new outbreaks.

In reality, they spike BEFORE COVID deaths.

In general, non-COVID death spikes are simultaneous with or BEFORE COVID death spikes (except external cause deaths as outlined above), suggesting they are unlikely to be from hospital avoidance.

2) Still a COVID death

But suppose I'm wrong and this really is happening.

It's still a COVID death.

To understand why, let's imagine a different scenario, like a war.

How many people died in WWII?

Well, *military* deaths came to about 23 million.

But 30 million civilians died due to war crimes and war-related activities.

ANOTHER 30 million died of wartime famines and disease outbreaks.

So how many people died in WWII? Usually we say around 70-90 million-ish. We include even deaths that weren't war-caused *at all*, because the direct cause was a predictable societal response to war.

If a hurricane hits a city and kills 4 people from wins, but it knocks out power and then 12 people on dialysis machines die as a result, the death toll from that hurricane is not 4, it's 16, even though 12 of them were deaths of kidney failure, not "hurricane winds."

Likewise, people ARE avoiding hospitals. And that may have negative repercussions on their health.

If that avoidance is due to strict government rules, we might call those "avoidable" deaths and so not strictly COVID-19 caused.

But if they're just due to people rationally managing their own risk and avoiding being around tons of sick people, then, surprise, THOSE ARE COVID DEATHS. They're deaths that happened because, and indeed quite directly because, of the COVID-19 pandemic.

We can imagine another COVID-19 scenario: imagine that shutdowns made it hard for an area to get food. Imagine that a city starved, and had 25,000 starvation deaths, but successfully kept COVID out of the city.

I see 25,000 COVID deaths.

They are deaths which occurred BECAUSE OF COVID-19, and BUT FOR COVID-19 would not have happened, and did not arise from some centralized plan of response to COVID-19, but from distributed individual behaviors.

TO BE CLEAR:

There are people who *did not get COVID-19*, whose deaths *nonetheless are caused by COVID-19*, and whose deaths are *not caused by lockdowns or the government*. That's just reality!

Okay, got my international data updated.

US cumulative death toll ratio continues to rise slowly.

Europe's downtrend has stopped. They're now holding steady at elevated death rates. Doing better than the US, but not a ton.

Latin America MAY have peaked?

EuroStat got Albania and Romania (!!) added into the database this week, so here's the current map of countries we have some data for.

Technically Andorra is in the dataset too but they haven't actually reported for 2020.

So this re-affirms the view that the US and European experiences of COVID have been pretty similar in terms of general scale. It simply isn't true that the US is having massively worse death rates than Europe.