Covid 6/23/22: Under Five Alive

America has now approved the Covid vaccine for children as young as six months old. Should you vaccinate your very young child?

Yes, I believe that you should. I don’t think you absolutely have to do that. Children that young are at minimal risk from Covid. However, vaccination decreases the chance they spread the infection to others, including you, and there is some chance that there are long term effects from Covid, and this danger far exceed the trivial risk of any potential long term effects from the vaccine. Also, even if getting Covid isn’t that dangerous, it is still universally a stress ball and a giant pain in the ass. That’s worth avoiding on its own merits.

Either way, I consider it a small mistake. The kids will be alright.

We might not be, if the Responsible Authorities get their way. A commission to look at the failures of our Covid response are going to call for an overhaul of our public health system… that concentrates power to determine response with the Federal Government, because that’s going to go great. This probably won’t happen, and there’s always some chance the new job involves doing a semi-competent job of it rather than being captured by the same craziness as the FDA and CDC. But I am not holding my breath on that.

They have forgotten what actually happened. Federalism wasn’t the problem. Federalism saved us from a much worse problem, with states able to react in ways their populations demanded and could tolerate, and in the face of an administration bent on a Blinking Spirit approach to the pandemic – that if they didn’t look at it, maybe it would go away. It would similarly have saved us if, as I would have counterfactually expected, a Hillary Clinton administration had tried to impose extreme lockdowns over a long period and red states would have been having even less of it than they did in our timeline.

There are of course alternative scenarios where the correct response was extreme measures that would have suppressed a deadly virus, or that some states would have gone and destroyed their economies if left to their own devices. In extremis, the Feds do need to take control of such matters, and I am sure that in a true full emergency they would do so and be allowed to, but I notice that making it easier for them, making it a default outcome, is not making me feel better.

Another development is there is yet another Long Covid preprint making the rounds, causing the exact same causation vs. correlation confusions. I’ve concluded this simply is not a confusion that humans in our civilization are equipped to understand. There is no way to make it stick, so we’ll have to keep hearing about this, over and over. I’ve accepted this. Also, the data may well be royally screwed up as mortality numbers are crazy high including in the control group.

Executive Summary

  1. Covid vaccine available for children as young as six months.
  2. Meet the new Long Covid study, same as every other Long Covid study.
  3. They’re trying to give those who failed us more power again.

Let’s run the numbers.

The Numbers


Prediction from last week (made without taking into account Juneteenth so this should be adjusted by -10%): 650k cases (+1%) and 2,000 deaths (+4%).

Adjusted prediction for Juneteenth: 585k cases (-9%) and 1,800 deaths (-6%).

Results: 593k cases (-8%) and 1,768 deaths (-7%).

Prediction for next week: 675k cases (+14%) and 2,015 deaths (+14%).

Some common sense. I am all for Juneteenth, it is an excellent holiday, but we need to get it right. If we want to call it Juneteenth (as opposed to something like Emancipation Day) and it falls on a Saturday or Sunday, we should celebrate it on Friday, not on Monday. Because Friday is in fact a day compatible with the name ‘Juneteenth’, the 17th or 18th of June, whereas the Monday will not be compatible because it is the 20th or 21st. That’s simple math.

Also, because as the Talmud would have said, freedom cannot wait. It is always better to get your freedom one day sooner.

Florida’s case count is now big enough that when I forget to do the manual adjustment (in either direction) the numbers remind me to fix that.

Arkansas hasn’t reported a death in over a month, which seems rather implausible, but it isn’t that big an effect so I’m going to ignore it for now.

After all that, a -10% adjustment for Juneteenth seems to have been about spot on, or more likely I got lucky on the details. Lot of noise out there, luck is always welcome.

Things have been basically static for a while, so I’m mostly just going to say the -10% gets undone next week plus a little backfilling, and call it a day at +14%.



There continues to be a clear decline in the Northeast and Midwest, with the South and West less clear. Any fears of things spiraling out of control can be put to rest at this point.

Did You Know We Could Make Better Vaccines?

I mean, yes, of course you did. And yes, you figured out it was a good idea.

Someone please tell the Congress and the FDA.

So, Mr. Topol, why do you ask questions to which you already know the answers?

You know exactly how, and you know how I know you know. The key is deciding not to care about getting people a vaccine as much as you care about… other things.

You Know, For Kids

The vaccine’s been approved. Should you get it?

Here’s one take from a pediatrician.

Aside from the degree of Long Covid concern, this seems right to me. Getting Covid sucks even if everyone involved ends up fine, it’s a giant stress ball and pain in the ass. Children are unhappy, life and work are missed. Reducing the chances of that is valuable. And even though I discount Long Covid concerns, they’re still far bigger than the risks from the vaccine itself.

Scott Gottlieb points out that there are good reasons to think not catching a virus might turn out to be better than catching it, and the uncertainties favor vaccination.

Once again, though, I consider this a small mistake either way. There will be no hard sell here.

Let’s say you decide you want your kid to get it. What happens now, what to do?

With younger children, mass vaccination sites aren’t an option, so you’ll do this the same way we do other things, and it will go somewhat slower. Which is fine. This is a smaller (in multiple ways) cohort, and a smaller fraction of it is going to get the vaccine, and the situation is far less urgent.

Some places really have their act together, which is pretty great. Being outdoors is a nice touch.

Also, isn’t it funny that they couldn’t ship the vaccine until it was approved? Again, only a few days, but it outlines how much this very much isn’t an emergency. Speaking of which…

Florida Man

Everyone’s second favorite and/or second least favorite Florida Man, Governor Ron DeSantis, is reported to be at it again. There’s a long history of Florida Man doing things a little differently in the ‘Free State of Florida’ with varying degrees of crazy versus crazy like a fox, so let’s see what it is this time.

As Google’s non-sentient AI LaMDA might ask us if given the right prompt, does the press think?

What is going on?

Top reply to that first one, as usual making the permitted versus compulsory confusion:

The responses to that, in turn, are mostly not pointing out that no one is taking anyone’s rights away and are instead, shall we say, ‘misinformation’ about the vaccine.

Florida has a longstanding policy of actively recommending against vaccinating otherwise healthy children, and they logically extended this to younger children. They did not pre-order the new vaccine shots, both because they don’t especially think the shots are worthwhile, and also because ‘preordering’ is, according to them, more like transferring custody of the vaccines between agencies rather than a useful thing to be doing, and of course physicians are still free to order them.

I’m not sure if they’re right about that, but in practice I doubt this slows things down more than a few days and if slowing things down a bit is so terrible I suggest you first have a word with the FDA and CDC rather than Florida.

The Long Long Covid and Reinfection Discussions

Covid still exists, and everyone is taking the same attitude about it and being as reasonable about it as they are with everything else.


The Long Covid hypothesis keeps taking a licking, and it keeps on ticking.

This week had yet another Long Covid preprint, which of course is going to once again not use proper controls and confuse correlation with causation and get picked up by the media and terrify a bunch of people, when the main thing happening is that being unhealthy in various ways is the big risk factor for a Covid reinfection.

I grow tired and have a lot to do. What’s going on this time?


Getting a disease means you’re likely to be unhealthy for a bit, yo. Who knew?

Again, I grow tired. Dylan Smith and Zeynep Tufekci, you have the floor.


It’s not only that one line. Look at the rest of that chart.

But wait, they do weigh the sample, it’s all fine, right?

On Omicron’s ‘This is true’ comment above:

Here’s another Zeynep thread warning about the whole scary headline pipeline.

That all seems right. Getting infected with things isn’t great, and if you are sufficiently unhealthy in various ways that you handle it badly even the second or third time around (which is correlated with those times happening at all) then you can be in trouble.

I wrote all that yesterday. Today I woke up to the news that the numbers look rather borked and the whole thing might be worthless.

As in none of this makes any sense:

What these findings absolutely should not do, even if these concerns are addressed, is worry people in good health. It should do the opposite. First infections hit healthy people at rates not that different from unhealthy people. Reinfections clearly do not do this, and third or fourth infections very much do not do this. If you are doing fine in general, your risk is smaller, not larger, than it looks.

There is a renewed claim that some Long Covid symptoms are correlated with continued finding of spike protein within the body.

I highly approve of this framing by Arson. If what Long Covid we do see (that isn’t false attribution) is simply the virus hanging around some of the time, then the obvious solution – wipe out the virus properly this time – seems easy enough, and something we are getting much better at dealing with over time.

Paxlovid For the People

Paxlovid is like a bank loan. To get it, prove you don’t need it.

Yes, in some sense this represents a mismatch between who should get a limited supply of a wonder drug (those at highest risk) versus those who got the drug (those at less risk).

However, it also represents a very good match in the sense that it matches those who value Paxlovid the most (those who trust medical authorities, those who take steps to protect their health, those who saw themselves at high Covid risk) with the drug, while not giving it to the unvaccinated who are none of these things. It is well-documented that those who are not vaccinated don’t trust the vaccine and they also don’t trust Paxlovid because the lack of trust generalizes. And if not, then they’re showing they are not much worried about Covid, for reasons that may or may not be valid.

To the extent that those at high risk get vaccinated more often, especially the elderly, this could represent a good allocation even in the risk sense. Kids tend to be unvaccinated and also are still at very very low relative risk.

A worry is that perhaps some doctors are taking a ‘screw you, if you weren’t willing to take the vaccine I am not about to waste Paxlovid saving your ass’ attitude, but I doubt this is doing much of the work. At most, they’re not telling any patients about it.

In Other News

The Russian vaccine trial data was probably faked.

As usual, this wasn’t subtle data manipulation. This was something else.

This is unsurprising. Certain areas are less trustworthy than others, shall we say.

What is especially unsurprising is that, given there was a statistical fraud, the fraud was this obvious.

It is very easy to look at a bunch of numbers, decide ‘hey this would go well if these numbers looked better’ and then change the numbers. People do it all the time.

What they don’t do all the time is spend the extra effort to make those numbers stand up to even ordinary scrutiny, by which I mean ‘I looked at this and asked if it looked too smooth or blatantly violates Zipf’s Law or is otherwise super fishy and then did the work to make sure it didn’t.’ That’s much harder, especially if you are looking to get a good instinctual reaction from those not asking those questions as your top of mind goal, and also trying not to directly contradict what observations people have.

People round off numbers, they copy data or make it up, they p-hack, they do all sorts of things, but what they very rarely do is use a properly random process that simulates the types of distributions you’d naturally see so it actually stands up to ordinary and even careful scrutiny. They don’t let it look superficially fishy while they’re committing fraud so that an actual careful analysis won’t be able to show anything. They make sure it doesn’t look superficially fishy, and hope for the best.

Or they don’t, and Saddam Hussein reports he got 100% of the vote and dares you to tell him different.

So that’s the thing about outright fraud, at least in my experience. It’s never subtle. So the next time you’re thinking an election was fraudulent, or someone’s data is fake, don’t worry about something super subtle. Look for the obvious.

Not Covid

Mark Cuban is on a mission to reduce prescription drug prices, and is now selling a wide array of them super cheap at Cost Plus Drugs. I haven’t verified quality, but it being done by Cuban without me seeing any complaints is more verification than a lot of the foreign pharmacies a lot of people I know get drugs from, and the prices seem super cheap. They don’t take insurance at all, and don’t do paid advertising. Most importantly, we know why this is available, so we don’t have to go looking for the reason. If someone knows about something wrong (or right) with them, please share in the comments.

FDA Delenda Est this week is all about them once again going after vaping, and also looking to restrict the amount of nicotine in cigarettes. Neither of these things is either (1) any of their damn business or (2) helping. Reducing nicotine both creates a black market and potentially makes cigarettes more dangerous since it makes people smoke more of them to get the same hit (but it could also get people to reduce intake or quit, so it’s non-obvious) and making it harder to vape traditionally increases smoking, which is much worse than vaping. Reducing nicotine levels all the way to being non-addictive seems like a de facto ban on smoking. One guess how that would play out. When the FDA is sitting around not approving life-saving medicine and not allowing us to have access to baby formula, this is what it is doing instead. It has been pointed out that this was the FDA plan under Trump as well, but that doesn’t seem like it makes this any better (or worse).

AirBnB announces crazy house contest. They’re offering 100k to each of 100 winners to build crazy houses that can then be turned into wacky AirBnBs. For round one all you have to do is write a few hundred words describing your crazy idea, so this seems like great low-hanging fruit for anyone with a cool concept.

New study on later school starting times just dropped, I may write a full article later but for now I’ll leave this here, and here is a thread of studies on the question:


That is quite the effect size.

Essential Terms of the Authority Crisis, at Marginally Compelling. The model is that we’d like things to work like this:

For a long time, we all knew that the part where the popular media told us the institutional knowledge wasn’t great, but most of us mostly trusted the other steps and the resulting institutional knowledge. Now we don’t, because they’ve done too much spouting of obvious nonsense and been caught in too many lies that they defended by pointing to the need to maintain public trust. So now we can’t trust that part either.

I think there’s a lot of truth to that story, but there’s also another story to consider.

A tale I have been told about what is wrong with politics goes something like this. Trade and free trade agreements are good. Economists agree we should have a lot of trade and a lot of free trade agreements. Yay, trade. Politicians mostly know that the institutional knowledge says this, and they believe it too. So traditionally they go out and sign a bunch of free trade agreements.

Alas, the public doesn’t know economics and does not like free trade agreements.

Amazon Alexa will be able to mimic the voice of anyone after hearing a short clip. And you can already use it to read arbitrary text. It is being marketed as a way to remember a departed loved one. I… do not think that is going to be the use case.

This entry was posted in Uncategorized. Bookmark the permalink.

22 Responses to Covid 6/23/22: Under Five Alive

  1. Anonymous-backtick says:

    “there is some chance that there are long term effects from Covid that far exceed the risks from the vaccine”

    Well, that’s a pretty jarringly weasel-worded discursion from rigor.

  2. Ninety-Three says:

    “They have forgotten what actually happened. Federalism wasn’t the problem. Federalism saved us from a much worse problem”
    A cynic would say that they are doing this because they remember *exactly* what happened and have correctly identified federalism as an obstacle to the imposition of their will.

    ” and are instead, shall we say, ‘misinformation’ about the vaccine”
    Typo: missing word, presumably “spreading”.

  3. gwern says:

    So that’s the thing about outright fraud, at least in my experience. It’s never subtle. So the next time you’re thinking an election was fraudulent, or someone’s data is fake, don’t worry about something super subtle. Look for the obvious.

    No, there’s lots of fraud that’s subtle and won’t go detected, it’s just that it’s “subtle” because the summary stats which get reported happen to not allow the slamdunk accusation of fraud rather than the fraudsters being sophisticated enough to set up plausible generative models for their fake data. There’s loads of areas where this is true; look at pretty much any psychology paper, where you get some means and some SDs and some p-values: there is almost no way to detect even the sloppiest frauds like a Diederik Stapel to the necessary standards from such impoverished reporting. (And I can think of several other authors who I am fairly sure are fabricating data sloppily, but I have no way of even beginning to prove this.) This is a selection effect: there are a rare few narrow slices of the literature where, like retrospective DNA testing of convicts, it just so happens that a handful of the most obvious and laziest frauds *can* be detected solely from the public material (whereas in all others, it’s pretty much impossible). And those slices of sunlight turn out to be so rife with cockroaches scurrying away from the light that you cannot have any faith that ‘oh yeah we’ve pretty much got this solved in all other areas too, look how lazy they all are, so QED, all fraud will be obvious, just check the obvious stuff and you’re fine’.

    Look at the stereotypical Western blot fraud: how many of those are *obvious*? Elizabeth Bik wouldn’t be famous if it was so easy to spot the copy-paste of images. Now, how many of Bik’s legions of fraud papers can you spot from the method used for the Sputnik papers? Or how many would GRIM detect as fraud? Very few to none. Remember the ivermectin paper fraud which turned out out be just a few patients copy-pasted repeatedly in the spreadsheet? Laughably obvious dumb stupid fraud of the are-you-even-trying sort… *if* you decrypted the patient-level spreadsheet and could actually look at it. (“Data available on reasonable request.”) Otherwise, it just looks like weird and somewhat implausible data but nowhere near the extraordinarily high bar required to shame a journal into retraction or possibly penalize the authors. There’s no Western plot you can eyeball for Photoshopped pixels in most studies. GRIM wouldn’t be able to detect that because with such a large ostensible n, almost all values become arithmetically possible, and GRIM only works well when the n is so small that there is no possible set of values which would average to ‘1.62’ etc. Or how about the Ariely honesty/mileage study where the mileage distribution is perfectly uniform and stops exactly at a particular mileage? Could not be more blatant when you look at the data he made the mistake of sharing with some more honest later co-authors instead of his usual argument that he no longer had the data for a result being questioned, yet, you can’t tell just how awful a dumpster fire it is from the original paper and indeed, no one did despite it being widely cited & read.

    • TheZvi says:

      Interesting contrast here, because we clearly agree that under the hood the fraud is basic and dumb in these spots, it’s just that you don’t see the data so you can’t prove anything. And in theory basically any result could happen at any time, so if you can avoid sharing any details then who can say if it’s real, really? But it still reads to me as ‘if they are doing a fraud they are doing a very basic fraud.’ The problem is that ‘look for the obvious’ is not so easy, but looking for something subtle wouldn’t work either, because these people are saying ‘no, I don’t have the data available, sorry’ or what not.

      • gwern says:

        Alzheimer’s is amusingly an example of that; a woman who was found to have doctored some Western blots had a drug candidate which just bombed – was that research fabricated too? Well, it’s Alzheimer’s, where nothing works, so how would you know the difference between a fake and a ‘real’ drug candidate…

        I would say the fraud goes well beyond that to cases where the fraud is a great deal subtler. It will still not look like ‘create a generative model to ensure that the fake data acts appropriately in every way downstream’, however, it just graduates from ‘copy-paste the same patient data row by row’ to even harder to forensically detect fraud like ‘make up or delete a few outliers to push it in the right direction’. The spectrum from a Stapel to a Wansink doesn’t have bright lines so much as yellow speedbumps, and further across the spectrum you go, the more common the fraud becomes as it becomes more subtle (and harder to do because it comes closer to being good research which requires a lot of work to create), and in aggregate, the damage done increases. There is, unfortunately, even in theory not much way to detect the low-grade kind of QRP and p-hacking fraud which is the ambient slightly-yellow warmth you swim in when you swim in the pool of science, other than very expensive large pre-registered replications to nail down the effect so precisely that the biases are collectively exposed as the drain is pulled and the pool empties.

    • Anonymous-backtick says:

      Or tying it back to the election case, all the things repeatedly called “audits” that turned out to be a light skimming of hand-selected ballots were the main thing making things so suspicious. Like a Magic player complaining if you look at his hands while he shuffles. Sure, it’s not ironclad evidence of fraud in and of itself, but come on…

      • TheZvi says:

        Elections are one place where you can’t hide the data, and I would expect any industrial-strength (as opposed to ordinary decent) fraud to be pretty easy to spot.

  4. Basil Marte says:

    On the study not making any sense (“5.5% of the control group dying in 30 days”): my guess is that they made a single-word mistake. Let me quote the relevant part:

    > We first identified users of the VHA with at least one positive [test] between March 1, 2020. and September 4, 2021, […]
    > We excluded those who died during the first 30 days after the first positive [test]. […]
    > We then constructed a non-infected control group.
    > We first identified 5,714,736 VHA users between March 1, 2020. and September 4, 2021 with no record of positive […] then randomly assigned a T_0 […]
    > We excluded those who died in the first 30 days after their T_0 […]

    This looks to me as if they copy-pasted parts of the text. Let me propose that the last sentence, corrected, would be:
    Use of the VHA was defined as having a record of use of outpatient or inpatient service, receipt of medication, or use of laboratory service with the VHA health care system in the two years prior to enrollment. <
    the implied <20y remaining lifespan, given mean age 63 +/- 16 y is fairly reasonable.

    • Basil Marte says:

      Either I made a stupid error of my own or WordPress ate things between too many and > and < symbols (testing that). It should go:

      Let me propose that the last sentence, corrected, would be:
      "We excluded those who died BY the first 30 days after their T_0 […]"
      As in, they took everyone who was in the database on 2020.Mar.01., and found that about half of those who died during the 1.5 year period did so before their randomly-assigned date, hence the need to remove 5.5% of the initial control sample. Since the sample was selected for not being completely healthy
      "Use of the VHA …

  5. myst_05 says:

    Would you recommend people to go get a 4th COVID shot if they’re healthy, not immunocompromised and never had COVID?

    • TheZvi says:

      I would say it is a small mistake either way. I am stopping at 3.

    • Thor says:

      Are you asking whether to get a 4th shot of the vaccine for the original strain that’s long out of circulation, or whether it’s worth it to get a shot of a vaccine updated to the current strain of the Virus, should the FDA ever deign to let that vaccine be sold?
      I think the latter is far more likely to actually be helpful, but I doubt a 4th dose of the original vaccine would cause you any serious harm

  6. lunashields says:

    I think you have a mistake in your recommendation about “yes kids should get the vaccine”. By now pretty much every kid had had covid. Ours had it before vaccine for adults were available, and we were reasonably strict about exposure until we got vaccinated ourselves back in february. Decided to test her for antibodies – turns out she’s positive.

    Every single survey comes up with at least 70 (and most over 90) percent positivity rate.

    With that in mind vaccination rec is pretty useless. Perhaps combines with an antibody test, and in case of negative, but not just blind.

  7. dtsund says:

    I’m skeptical that we’ll see long-term effects that only emerge years later from COVID infections. Unlike culprits like Epstein-Barr or Varicella, SARS-CoV2 is an RNA virus, not a DNA virus, and it lacks reverse transcriptase; this means it cannot integrate into the genome as DNA. And RNA catalyzes its own degradation. I won’t say it’s impossible that latent infections could resurface later, but it seems quite unlikely to me.

  8. Bldysabba says:

    “tale I have been told about what is wrong with politics goes something like this. Trade and free trade agreements are good. Economists agree we should have a lot of trade and a lot of free trade agreements. Yay, trade. Politicians mostly know that the institutional knowledge says this, and they believe it too. So traditionally they go out and sign a bunch of free trade agreements.

    Alas, the public doesn’t know economics and does not like free trade agreements.”

    This seems incomplete. Like a copy paste gone wrong

  9. ben says:

    Would you recommend that a child who already had and recovered from Covid get vaccinated?

  10. Zviva La Vida says:

    LG PuriCare is the only popular wearable air purifier (P100 filtering + fans reducing breathing resistance), and people still complain about it. Is it indicative of the efficient market hypothesis (e.g. for some reason, it’s really difficult to design a “supermask”), or should I exploit the opportunity, launch a startup, and strive to create a slimmer version with a better fit? This is basically the only device we need to stop worrying about the pandemic at the individual level.

    Also, how dangerous is Paxlovid? Common complaints: taste alterations, rebound, numerous potential drug interactions, muscle pains, elevated blood pressure, possible toxicity in <12yo and <40 kg patients, unknown unknowns. I can't vouch for these two videos, but they seem to raise important and concerning points.

  11. Yesterday the FDA’s advisory committee met to consider Omicron-specific vaccinations, either as bivalent replacements for existing vaccines or as monovalent boosters.

    I watched most of it and went through all the slides, pulling out some highlights and some statistical oddities.

    Summary: The vote to recommend the FDA proceed with Omicron boosters was 19 yes to 2 no. In general, Moderna’s bivalent (classic + Omicron/BA.1) looked to have broader application than Pfizer’s monovalent (Omicron/BA.1 only), unless you’d already had the classic vaccine anyway, in which case you’re fine.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s