Because most philosophies that frown on reproduction don't survive.

Tuesday, January 05, 2021

COVID Death Statistics: Looking Back at 2020

 It's been a while since I've written a post on COVID data, but the calendar year has ended and it's the time when people are looking back on 2020.  In particular, I recall a sort of debt of honor, in that back in April I promised Joseph Moore who blogs at Yardsale of the Mind that I would check back in at the end of the year to assess whether the pandemic had resulted in significant and measurable deaths in excess of what we would expect in a normal year.  

Joseph Moore, the author of the above linked blog Yard Sale of the Mind holds that in fact we're not really seeing any "excess deaths" due to COVID-19, as the people dying all would have died soon anyway. I think this is false and that the deaths in the US and elsewhere will be clearly visible in overall death data for the year once that becomes available. Since I think it's important to put one's credibility on the line when making statements like this, I'll lay out in a separate post my commitment to report back in January 2021, examine the data, and either rhetorically eat my hat or show clearly that this was in fact a serious matter causing a death toll in the US of over 75,000, at which point I'll invite Joseph Moore to rhetorically consume a hat of his own.

Though I'd never entirely forgotten this, I was reminded of it recently when I noticed that Joseph Moore had a year end review post up in which he looked back at COVID numbers.   His own analysis is perhaps a little incomplete.  He first spends some time discussing the difficulty of finding the data one desires on the CDC website, and then examines the comparison of the number of deaths it shows that involved COVID (shown as of then at 301,679) as compared to a percentage of the deaths of people in age categories under 34 that involved COVID (very low).  His wrap up is:

I note one last thing from the CDC data now, with more to come when I’ve gotten a chance to digest the data: the overall death rate in the US in 2020, pending, of course, updates that should roll in over the next couple weeks and push it ever so slightly up, was 0.884. That’s 884 deaths per 100,000 Americans. Back in May, I looked up the UN’s 2020 projected death rate prior to COVID, and it was 0.888. This number just comes from extrapolating from long-term tends, nothing special, but, barring a deadly pandemic or other disaster, such a method should produce a pretty accurate forecast. So it looks like the US, despite a raging pandemic that’s killed, they say, 350K people, will have had pretty much the same number of dead people in 2020 as projected before the pandemic. Huh.

This analysis did not strike me as particularly illuminating, so I decided to do my own analysis of how many more people have died than might otherwise be expected.  I pulled my data from the CDC FluView Dashboard, which is nice because it allows you to break the data down by week, age range, and region.   I then organized the data in Excel and produced some graphs.  The data for the last 4-6 weeks is not complete, but it's certainly complete enough for us to get a good picture.

As I mentioned, the FluView data is broken down by week.  This makes it easy to compare the period that you're trying to analyze with the same period in previous years.  To start with, I looked at this year compared to the data for 2016-2019.  As you can see, the deaths per week all cluster very closely around the 2016-2019 average, except for 2020, which beginning in Week 11 (mid March) rises above the norm and has not come down since, though the highest deviation from the average was in April.

Since all the other years stay pretty close to average and deviate from it only for brief periods, I decided it was acceptable to simply compare 2020 to the 2016-2019 average in the rest of my analysis.  Here I compared 2020 deaths by week to the 2016-2019 average.  Then I calculated the difference between 2020 and the previous four years as the "excess deaths".  I've charted those excess deaths as columns at the bottom.  As you can see, we peaked just under 25k in mid April, but we've had around 10k excess deaths nearly every week since.

If I total that excess death number from Week 1 to Week 49 (data for weeks 50-52 are still not remotely complete, as you can see by the "tail" at the end of the year -- they'll be revised as full data is collected over the coming weeks) I get a total of 384,857 excess deaths.  That's actually a higher number than the 365,170 US COVID deaths that appears on WorldofMeters as of today.  Of course, the excess deaths are not necessarily all COVID deaths.  Say, for instance, that it were the case that there were an elevated number of suicides due to lockdown loneliness.  Those suicides would be included in this number.  So we can't necessarily say that all of those deaths are due to the virus itself, but I think it's virtually impossible for someone to look at this data and argue that we are not experiencing a lot more deaths than we would normally expect, absent a pandemic.  How many is that 384k compared to the normal number of deaths?  Well, from 2016-2019 the average is 2.83 million deaths.  In 2020, there were 3.13 million deaths, and the last 4-6 weeks aren't even fully complete.  So the 384k additional deaths represent a 13.6% increase over the normal number.  

Obviously, another question someone might ask if whether the number of deaths is just higher because the population of the US grows ever year.  I think the first graph showing the deaths by week for the last five and a half years makes it pretty clear that the population effect is not the main cause here, but to give you another view of it, here's a graph of the total deaths by year for 2017 to 2020 (because this data set is based on a "flu season" which begins in September, the 2016 data is partial so I left it out.)

As you can see, the increase during the last couple years was 15k-25k.  Going up 300k+ is clearly something more than population growth.

Another question that often comes up is how these deaths are spread across different age groups.  Is it only very old people who are dying of COVID?  Obviously, the deaths of all people are important, but the argument that is made in this respect is that the only people dying of COVID are people who were on the brink of death anyway.

I think that the sheer magnitude of the death increase is some proof against this.  However, let's look at the age breakdowns, because there are some interesting things to see there.  This first graph looks at excess deaths for Americans who were 65+

That's a total of 294,595 excess deaths versus an average of 2,083,936, which is a 14.1% increase above the norm.  The FluView dataset doesn't cut the ages more finely than this, but as has been widely reported, a lot of these deaths are at the older end of the range.  The next graph surprised me a bit:

This above looks at the deaths among people aged 18-64.  It looks a lot like the previous one.  In total, it shows 93,388 excess deaths over a 2016 to 2019 yearly average of 703,566 per year, a 13.3% increase.  I would have expected the increase to be a good deal less.  Of course the number of deaths from people aged 18-64 compared to the number of people who are in the age group is much lower than the deaths per person aged 65+.  However, people aged 18-64 just aren't that likely to die in the first place, so apparently the increase in deaths to people 18-64 is almost as large on a rate basis as the increase for people aged 65+.

The next graph is interesting if your want a weird silver lining that people still haven't really figured out.  For people aged 18 and under, the number of deaths this year has been below what you would normally expect, and the periods when it's been most below have been the spring and fall, when the deaths for other ages have been highest.

People under 18 are very unlikely to die, so the numbers here are quote low.  2,914 fewer people under 18 have died than you would expect given the average of the prior years.  That represents a 8.2% decrease versus a base of 35,492.

So there you have it.  The due to the pandemic a lot of people are dying, significantly more than we would normally expect to die in a year.  Not only have almost 300k more people over 65 died than we would expect, nearly 100k people between 18 and 64 have died.  We're so far past the 75k number at which I offered to eat a hat, it seems shocking that anyone can still argue the point.  And yet, there are still people out there in my social media feeds claiming that it's all a hoax. 

One of the reasons I haven't written a much about COVID lately (or indeed about the election hoax true believers) is that it's become increasingly clear that many of these people's beliefs have little to do with the alleged evidence they cite.  A clear sign of this is how quickly people slide from one claim to another, or resort to simply kicking sand at the knowability of it all.  "Of course, it's gotten so hard to even know when THEY control everything."  Still, I think it's important to circle back every so often to look at the basic numbers.  These are knowable things.  It is not hard to go pull your own data and do your own analysis, and the conclusions are pretty obvious.


AmyRobynne said...

Thank you! The smaller kid numbers is a pleasant surprise! I always appreciate a good look at the data.

mandamum said...

I wonder if that smaller kid number is perhaps linked to kids being more likely to die in accidents than from disease, so fewer driving-hours is probably directly linked to fewer kid deaths in car accidents? I wonder if it's just car related, or if other types of accidents have gone up or down....

Michael said...

This is a good nuts-and-bolts analysis, though as you note it would be nice to see more granular age data. However, I suspect you'd get the same 13-15% increase for each bracket, albeit with different baseline mortality rates.

I enjoy a good conspiracy story as much as the next guy, but the loudest voices I've heard are not disputing the existence of virus mortality, but rather arguing against the heavy-handed and unevenly-applied lockdowns. What really rankles are stories such as the comic-book-store owner arrested for delivering online-orders, while Bezos balloons his marketshare, or cathedrals locked while shoppers mob costco. Then again, the cards were already stacked in that direction anyway; it has never been a good business proposition to open a restaurant. It all makes one so sad.

Ian said...

I guess it depends too on what threshold of excess deaths one regards as significant. For example, if someone had told me last February that we would have 70k extra deaths in 2020 from Covid, I would have regarded that as fairly insignificant and would have wondered what all the fuss was about, especially in light of what was being predicted at the time. My own rather subjective thresholds were over ~500k/600k being a major event (and 1M being a disaster), and between ~200-600k being significant, but not earth-shattering (200k is approximately how many die from flu+pneumonia each year) - that's a range of very roughly, maybe 10-20% excess deaths or so.

Back last February, and especially in April, I thought we would end up in the 500k range by the end of the year, so it has turned out to be not quite as bad as I had feared.

William Briggs has done similar analysis to yours, scroll down to the section titled 'The Numbers':

Darwin said...


Yeah, I would tend to think that the lower kid deaths are due to fewer accidents, which is the leading source of death for people under 18.


I agree that there's a lot of room to criticize stupid and ineffectual lockdown attempts. I do think that reduced gatherings have reduced spread, but a lot of that can simply be achieved through making people aware of the virus and how it spreads, and by very simple and commonsense restrictions. Some of the restrictions have been stupid and have just fed conspiracy theories because they're so obviously ideological, false, or self-serving.

My fire in this post is directed solely at the people who are still claiming that the virus itself is a fraud and that no one is actually dying of it.

Darwin said...


Yeah, looking back 70k was a low number. At the time, the official number of COVID deaths was around 50k, and because they were all so recent (although I believe the first excess death analysis was being done on the Italian outbreak) it was hard to sort the data out yet and a fair number of people were claiming that there were in fact no incremental deaths, but that these were all people who would have died within a couple weeks anyway.

I was rather less pessimistic in Mar-April. It seemed to me that if we were organized and successful in our response, we might keep the toll to 100-200k. Clearly, that didn't happen.

Anonymous said...

While I totally agree that some of the perspectives and mandates that favor lockdowns and restrictions are ridiculous if not harmful (one of my personal favorites was an Instagram story last April that lambasted anyone who went outside and "breathed public air"), I suspect your remark, Darwin, that "making people aware of the virus and how it spreads [would've helped reduce spread]" is a bit optimistic. Even with pressure to cut back on gatherings, not to mention awareness of the virus and how it spreads, many of my friends are having full-blown weddings, parties, gatherings, etc., events that strike me (even as someone who is currently planning a wedding that will be affected by the pandemic) as pretty optional. Rightly or wrongly, everyone judges their own events--their wedding, their baby shower, their bible study, their annual trip to Cabo or whatever--as too important to cancel or change. I don't know what sensible and commonsense restrictions would've ruled out enough of these gatherings that it made a difference but not so many that people didn't react against them in roughly the way they have.

I think public health policy is just tough because for it to be effective, masses of people have to get it, and for masses of people to get it, the policy has to be coarse enough that the gist of it can fit on a billboard. "Don't gather in large groups!" has a lot more impact than "Don't gather in large groups unless..." not because all gatherings are equally risky, but because if people operate under the latter, the policy's effectiveness depends not only on people remembering a fairly complicated message but also on their ability to apply the message adequately, an application which would require making a good judgement and having information (e.g. how many of your local hospital's ICU nurses are out of commission). That's a lot of moving parts when you consider the whole population.