USA All-Cause Mortality Data Through January 2021

UPDATE: I have a post with the most recent data here.

It’s been a while since I talked about all-cause mortality data for the USA so I wanted to look at the most recent numbers. Just as a refresher, all-cause mortality data is so important because it’s unambiguous. Doctors may diagnose a particular death as due to one cause or another for a variety of reasons, but they all diagnose a dead person as having died. For example, did someone with COPD and COVID-19 die of the COPD or the COVID-19? Both answers are legitimate, and either might be preferred for a variety of reasons. No matter what the doctor who signs the death certificate puts down as the cause of death, though, he puts down the fact of death just the same. A trend for diagnosing death in a particular way might look like an underlying trend in the disease when it’s really just a trend in bookkeeping. Or a trend in diagnosis might make a new disease seem like it’s not killing people when it is. Looking at the absolute number of people who are dying, regardless of cause, can help us tell the difference. That’s why it’s so good to look at all-cause mortality. It doesn’t tell us much, but at least it can’t lie. So, without further ado, let’s look at the most recent data. (As before, you can get the data and see the latest data after this post becomes outdated here.)

Always ignore the data for the most recent week, and based on recent trends the second most recent week is likely to change a lot, too. It takes oddly long for mortality data to come into the CDC (up to two months to get all of it), so they have to do some guesstimating on more recent weeks. In the last few months they’ve tended to under-estimate rather than over-estimate it (I believe around six months ago they were over-estimating it a bit and over-corrected from that). Accordingly even a few weeks back may increase some. (Later on in this post I have a comparison to what this looked like a month ago so you can see for yourself.)

The orange line is the threshold for “excess mortality”, i.e. if the all-cause deaths/week is above that, this is more than we would expect from normal variation, based on previous years, and there might be something up. They also helpfully put a red plus sign above each week where the mortality exceeds the excess mortality threshold. Here is a zoomed-in shot to the last year or so:

Unfortunately this format, though good for having a sense of what’s going on with overall mortality, doesn’t make it easy to compare excess mortality across weeks since there is normal seasonal fluctuation in deaths. To help with this, I downloaded the data as a CSV file from their website and created a graph of excess mortality that’s much easier to compare (note that this is the full data, i.e. going back to January 2017):

This third wave is turning out to be the biggest yet, at least in terms of area under the curve. That said, since the definition of excess mortality is only a guess, though a guess made by applying statistics to historical data, I don’t think that we can put much stock in small differences. That is, the first wave having a very slightly higher peek in excess mortality probably doesn’t mean much. On the other hand, its peak was during the time when mortality is normally going down (heading into spring) while the peak in excess mortality in the third wave was close to the worst time (winter). Probably the best thing to do is to not worry about small differences and consider them equivalent in peak, but with the peak being sustained longer in this third wave.

The third wave looks like it has peaked but unfortunately the data isn’t really reliable enough to tell, yet. Here is the graph from my previous post, which only contained data through January 1st:

That had looked like it was peaking, too, and it turned out that even several weeks back on that graph were still incomplete. Again, for reference, here’s the graph I made from the CSV data at the time:

This may make it even clearer how far back the data can be revised. One thing that’s very clear is that, as a country, we need more timely mortality reporting. Having to wait two moths to get accurate data makes the data hard to use for any practical purpose. I do get that there will always need to be some revisions—someone who dies at home and is only discovered days or weeks later, for example—but it’s hard to believe that this is such a large fraction of deaths. It seems far more likely that antiquated reporting standards are to blame. That said, I don’t know that for sure; I think it is worth spending resources to find out why the data takes so long to be reliable.

So, given that the data is subject to revision, if diminishing revision, for up to two months, what can we make of this data? Not a whole heck of a lot, frankly. Just to refresh our memories, here’s the latest graph one more time:

It is inarguable that something is going on; it is equally inarguable that this is not the second coming of the black death (which killed off around a third of Europe). It is clear that we’ve had three waves of significant excess mortality, which do correspond pretty well to the three waves of COVID-19 infections we’ve detected as well as the three waves of deaths attributed to COVID-19. It is, of course, not possible just from these numbers to say what is up, however. It could be that COVID-19 is killing people. It could be that lockdowns meant to stem the flow of COVID-19 are killing people. It could be that there are homicidal maniacs running about killing extra people in three waves. This is the limitation of all-cause mortality. That said, if you compare the numbers of COVID-19 attributed deaths and excess mortality, they actually do tend to line up reasonably well. I don’t have the time to generate the side-by-side graph of that which would make it clear just how closely they do or don’t track, this is just me having done some quick approximations, so take it with at least one grain of salt, more being better.

Assuming that the numbers in December 2020 are not revised higher, I do find it interesting that the excess mortality was not significantly higher, on a weekly basis, than it was during the peak of when COVID-19 first hit the US. COVID-19’s first entry was only into some locations. Though it spread quickly, it was doing this spreading during the time when the weather was getting mild and people tended to be outdoors, which seems to correlate with when COVID-19 infections (and deaths) are at their lowest. During the third wave, COVID-19 was already well spread out through the country, such that when winter set in and infections began to spike, it didn’t need to spread in order to bloom everywhere.

Though infection numbers are not directly comparable across time because testing is so much more widespread and available as time progresses, testing was pretty available by the middle of the summer, so you can see COVID-19’s prepositioning for the third wave in the infection numbers:

In spite of these advantages that COVID-19 had in the winter of 2020 over the spring of 2020 we don’t see a similarly massive increase in all-cause mortality. There are various possible explanations for this but the data doesn’t really support any of them better than the others, so far as I can tell. It may be that the virus spread far more widely among the population during the first wave than we had any idea of. It may be that the most vulnerable people already died from COVID-19 in the first two waves, or it may be that treatment got better, or it may be that the virus became less deadly, or it may be that we got better about isolating the most vulnerable people. It may be a combination of all of these things, each contributing something to the outcome. Undoubtedly there are other possibilities I didn’t name, too.

I don’t have any grand conclusions to this post; I don’t make these posts about the all-cause mortality data to argue for any particular point. My goal is to highlight the little bit of highly reliable data we have, because I think we’re all better off if we’re at least familiar with it.

Masks, Outdoor Dining Indoors, And the Golden Calf

A friend of mine was recently telling me about some COVID-19 mitigations going on in Washington DC, where restaurants were permitted to winterize their outdoor dining areas, which means putting walls around them and a roof over them. I.e. they are putting their outdoor dining area indoors for the winter, and it apparently counts as being what we might call “ritually outdoors”.

My friend laughed at this, but it’s actually a really interesting example of how paganism works and the human urge to pagan superstition. I’d like to talk about it for the sake of understanding those better. (COVID-19 will eventually pass, the urge to paganism will always be with us.)

I maintain that one can see the most essential element of paganism in the Book of Exodus, when the Israelites turn to paganism while Moses is talking with God on Mount Sinai. They are wondering in the desert and are running low on food and water. They asked Moses whether he brought them out into the desert to die, because they could have done that with more convenience back in Egypt. Moses went up to Mount Sinai to ask God for help, and when he was gone for longer than the Israelites had patience for, they asked Aaron to make them a god. He gathered up their gold, and using the fire melted it and cast it into the form of a calf. Then the Israelites worshiped it, and shouted,

Israel, here is your God who brought you here from Egypt!

The key thing to notice is that they did not turn to worshiping someone else’s god. They were still trying to worship the one who brought them out of Egypt; they were just trying to worship him on their own terms. They did not want to worship him according to his instructions; that was taking too long and they couldn’t see Moses and didn’t know what became of him. They did not want to trust. They wanted something that they could take an active part in. They wanted control.

So they made themselves a statue. They made it of gold, to please the god they were trying to worship. They worshiped it, to please the god. In effect, they were trying to do what all pagans do—they were trying to engage in a transaction with the god. They had no real way of knowing what, exactly the god wanted. That wasn’t really the point, since the gods were, if you get down to it, unknowable. The point was that they were doing something, and about as much as a human being could be expected to do, and that really should be enough, shouldn’t it?

And here we come back to indoor areas which are ritually outdoors. COVID-19 has a lot in common with the ancient gods, especially if you bear in mind that the ancients generally had gods of sickness and pestilence that one would give offers to in order to be spared. We have some basic knowledge about how COVID-19 works, but it’s still mostly guesswork how it spreads. (It’s pretty clear that it’s airborn, I don’t mean that basic fact, but what actions actually spread it, and how far, etc are not known with certainty.) No one can ethically set up controlled experiments to see what does and does not spread it, so we have to remain in ignorance about most of the practical aspects of how it spreads. Still, we have to do something. We can’t all hide in individual burrows until the disease goes away (storing up weeks of food per person would massively overtax our food delivery infrastructure), so what do we do while we’re not doing the thing we know would actually work? It’s got to be something.

As the days with COVID-19 drag on, the amount people are willing to not do grows less and less. So, in the absence of knowing how to do the things we want to do while staying safe, we must do something to show that we have not grown proud, that we’re doing as much as a human being can be expected to do, and so we should be spared. So we wear masks while no one really knows their effectiveness. (There are some designs of mask which do a very good job at stopping respiratory droplets. There are some designs of masks that aerosolize respitory droplets more than wearing nothing. Whether stopping respitory droplets or aerosolizing them further affects transmission is, however, not known. Even apart from that, how masks are worn significantly affects their performance, making any sort of generalization near impossible. In medical settings people are trained in how to use N95 masks and are actually tested on it by being put in environments with aerosolized aspartame. If they can taste the aspartame, the mask doesn’t fit correctly. Without a proper fit, air tends to pull in from the outside through the bad seal because there’s less resistance there than through the filter medium, meaning that a poorly fitted N95 mask provides almost no protection.)

And when people want to eat at restaurants, the restaurants ritually purify themselves by building new buildings, dedicated to the god of disease, in order to placate the god of disease to show that we are not being hubristic but are taking it seriously and asking it to spare us.

This is a thing you will see any time danger intersects partial ignorance and necessity or desire. When people do not know what to do in order to avoid bad fates, they will make offerings to the gods in order to be spared.

And, as a bonus, if the one who made the offering was not spared, then at least he will not be shamed before other men; he did everything he could, and the gods are capricious.

Unfortunately, so are men. His fellow creatures will be very tempted to suspect that he did not make the offering in good faith, that he had some private vice that the god saw but men didn’t, and is being punished.

As COVID-19 has recently reminded us all, we human beings live in ignorance and so paganism will always temp us.

All-Cause Mortality Data for the USA

Update: I’ve got a post with the most recent data (through April 3, 2021) here.

After much searching and asking friends if they can find the data, I’ve finally found a source for all-cause mortality data for the United States of America. It’s the CDC’s excess mortality data page. There is a ton of data on this page, and I recommend checking it out. Before I show you a screenshot of what it looks like at the moment, I’ll explain why this data is so useful.

Consider the following hypothetical: a person suffers from COPD (chronic obstructive pulmonary disorder). This is a deadly condition where the lungs are deteriorating, and it just keeps getting worse until the person dies from it. Suppose the person with COPD probably is sufficiently advanced that they only have six months to a year left, and then they get COVID-19, and die. Did they die from COVID-19 or from COPD? Different doctors, hospitals, and medical systems will answer that question differently, and all in good faith.

(To see how it can all be in good faith, how much the COVID-19 pushed them over the edge is something God knows, but man can’t know with certainty. Had they gotten something else, like the flu or a common cold, that might have pushed them over the edge instead. If they got a common cold and died, we would call it a COPD death, not a common cold death. This is just one example, there are a lot of cases which are legitimately judgement calls on which people disagree.)

This disagreement is especially a huge problem internationally. There’s absolutely no reason why doctors in Russia, China, Kenya, and Paraguay would have the same standards for things; it’s not like they would ever talk to each other, or report anything to the same place.


Everyone diagnoses death in the same way, at least after a few minutes. Not being alive is a very difficult condition to miss, no matter what tests are common or what doctors are habituated to look for or what their beliefs or customs are on the primary cause of death. And what is true across countries is helpful across states, too. It doesn’t take much looking for find endless debate about whether COVID-19 deaths are being over- or under-counted; we can be fairly sure about deaths being accurately counted.

Or, rather, we can be after about eight weeks. One problem that we run into here is that the CDC has found that only about 60% of deaths are reported within 10 days of the death; it takes about 8 weeks to get completely stable numbers. This is a very long time to wait, so they have algorithms based on how long it typically takes each reporter that feeds into the CDC to report all deaths to predict, after 10 days, what the final count will be. It looks like lately they’ve been under-predicting deaths in the first week by (about) 15-20%, though it varies from week to week, and I don’t have enough data to say that with certainty. I’ve tracked it for 3 weeks now and the numbers seem to get reasonably stable (by which I mean changing by less than 5%) after a few weeks of being on the chart. That said, everything in the right-most 8 weeks does need to be taken as provisional, the further to the right, the more provisional, and bearing in mind that the provisional numbers have a bias towards under-predicting the final number of deaths per week:

As you can see, this spans a little more than 3 years. I’ve no idea what happened in January of 2018; I don’t recall any news items about excess mortality back then, nor anything that would have been an explanation for it.

I’m not, here, going to get into any sort of in-depth analysis. I think it’s a bit early for analysis, aside from a few observations. The first is that the excess deaths do, more or less, follow the same pattern as COVID-19 deaths reported by the CDC, which gives some confidence that those numbers on COVID-19 deaths aren’t wildly inaccurate.

The other observation is that COVID-19 is obviously not affecting mortality all that much. The worst weeks for excess mortality were about 40% excess deaths, but that only lasted a few weeks. Excess mortality quickly dropped to below 20% and often below 10%. Or you can just look at the area in blue under the yellow line versus the area in blue above it (that’s not quite perfect because there is a bit of uncertainty built into the yellow line, but not a lot). It did go back up again, in time with the second wave of COVID-19 cases, but it appears to have peaked. The peaking is within that 8 week window, but the CDC’s numbers on COVID-19 deaths show that they peaked back in august, so if all-cause follows COVID-19 deaths as well as it has in the past, it is likely that the peak in all-cause deaths we’re seeing is real. We’ll be able to be a lot more confident about that in November or December.

By the way, an interesting question, which we won’t be able to settle for months at the earliest, is whether there will be a discernible drop in all-cause mortality for a while. If there is, that would strongly suggest that COVID-19 mostly just hastened the deaths of people who were going to die soon anyway. It will be interesting to watch for this.