Covid 6/24: The Spanish Prisoner

The last scare is underway. Delta is an increasing share of Covid cases around the world, causing cases in many places to rise. Are enough people vaccinated? How bad are things going to get before we turn the corner one final time? 

The incremental news was not good. Calculations that looked comfortable last week look less comfortable now. I still expect things to mostly be fine, especially in areas with high mRNA vaccination rates.

Also, if you think John McAfee found dead in a Spanish prison. If you think he killed himself I have some computer security software and I’d like to sell you a subscription. Works great.

Let’s run the numbers.

The Numbers


Prediction from last week: Positivity rate of 1.8% (down 0.1%), deaths fall by 9%.

Result: Positivity rate of 1.8% (down 0.1%), and deaths fall by 9%.

Prediction for next week: Positivity rate of 1.8% (unchanged) and deaths fall by 8%.

Got this week on the nose. With the rise of Delta and the shift in tests from safe to unsafe regions, I no longer expect the positivity rate to continue to decline, and if anything an uptick is more likely than a downtick. For deaths, there’s no reason to think things won’t improve for a few more weeks. 


May 13-May 19592119412778113874
May 20-May 2661594812796313473
May 27-June 252783811704562991
June 3-June 97208179154312883
Jun 10-Jun 163686119613142254
Jun 17-Jun 235294438312632066

As discussed last week, I’ve shrunk the graph so we can see what’s happening recently, which was otherwise impossible to read. We saw progress this week, but the West’s number last week was indeed ahead of itself, so we saw only modest overall progress and hit the 9% decline target exactly. Things now seem like they’re back on the expected track and the orange New York line is down to 51 deaths last week. 

We should expect to see things continue to improve, but the increasing share of Delta infections does mean the fatality rate should now be rising, given the slow pace of additional vaccinations. 


May 6-May 1246,04559,94570,74046,782223,512
May 13-May 1939,60145,03063,52934,309182,469
May 20-May 2633,89034,69448,97324,849142,406
May 27-June 231,17220,04433,29314,66099,169
Jun 3-Jun 925,98718,26732,54511,54088,339
Jun 10-Jun 1623,70014,47225,7528,17772,101
Jun 17-Jun 2323,85412,80126,4566,46469,575

The lack of progress here this week is quite worrisome. Delta is only up to around 35% (the data says 30%, but it’s lagged to start with and three days old to boot, and makes me think the 35% I estimate extrapolating from last week is likely slightly low rather than high), so if that’s enough to get us treading water, there’s going to be trouble in the more vulnerable areas within a few weeks. The question then is, will it be contained the way the Alpha uptick was, or will it become a much more serious problem? Discussion in the Delta Variant section, but it looks much less optimistic than it did last week. 


Some substantial differences in Bloomberg vs. Washington Post, but same basic story. 

Last few days saw a dramatic drop to a new lower level, which does not bode well. Fully 75% of the last week’s doses were second doses, which also bodes quite poorly. Hopefully both have something to do with reporting or some quirk of the calendar. Juneteenth, our latest holiday, is a plausible culprit on both counts. Until I had to adjust for data distortions I never realized how many different holidays we have in America, and for official purposes this one’s all new. 

 Chances are this is some of it, but that a lot of it is that the recent uptick was not sustainable and we will continue our steady decline as we run out of willing arms in which to put shots. 

If not, we are going to be stuck not that far beyond our current level of 53.6% first doses for a long time, and the August 10 date for going from 66.4% -> 70% among adults will be far too optimistic, and more like the asymptote until something changes.

Here’s a state breakdown:

The correct answer to ‘when will the states listed as November or later get to 70% if nothing changes’ is very clearly never.

People have made their choice. If you didn’t want to get vaccinated back when there were lots of cases, why would you change your mind now, with everyone feeling safe and the country reopening? 

Delta. It’s time.

Delta Variant

I did a bunch of calculations last week to figure out if we could handle the Delta variant. There’s also a much simpler way to view the situation. Delta is the forward-looking pandemic. Anything less dangerous than Delta, including Gamma/P1, will soon be dominated by Delta, so it’s a reasonable approach to ignore the total numbers and look only at the Delta numbers in various places. Essentially, treat the red lines as the pandemic here, not the grey:

The problem with that approach is it requires a good appreciation of the rate of absolute growth of Delta rather than its current level, which is trickier to get right and causes errors to get compounded. Also, such a model would freak out about every variant we’ve ever seen emerge anywhere. It’s still clearly not going to be a comforting answer.

The best data source I found last week continues to be the best known source. It shows us up to 30% Delta for data reported as of Monday the 21st. If the estimates from last week are accurate, it should become a majority of infections within a week, and become dominant soon thereafter. Within a month it will effectively be the pandemic’s present, not only its future.

The only questions are, can we handle it? What happens if we can’t?

Last week I approximated 25% Delta, which if we take its reproduction rate seriously is now 32% Delta after an additional 5-day cycle, or 35% now. That change only adds a few percent to R0, so it doesn’t explain the lack of progress last week in case counts. As we’ve seen in the past, weekly numbers can be quirky.

The problem is that last week’s estimate of how fast we were vaccinating was too optimistic. Part of that was a bad estimate, part of that was bad news over the last week, but it looks like a more realistic guess is 0.5% off the population per week rather than 1%. Also, I have to increase my previous guess on the old R0 from 0.84 to more like 0.86 after this week’s numbers.

That would start us out at R0 = 1.13 with a gain each week of only 1% that is plausibly fading, if you ignore immunity from infections and further control systems entirely in all directions, which would take us three months to turn the corner. During that time, Delta would go up by a factor of 6 or so (and all other variants would presumably get mostly wiped out) which would mean we’d peak around double our current case rate. 

Basic sanity checks on these calculations look like they check out, and are at least in the right ballpark. 

That ignores that different regions are different, which likely makes things worse overall. The Northeast essentially ‘wins’ and shrugs things off, while the South could be in serious trouble. Which in turn leads to control systems, including an increased willingness to get vaccinated.

That’s not great. Is it too close for full comfort? Absolutely. But given that this excludes control system reactions and the effects of immunity from infection, both of which work in our favor, and I used a bunch of conservative estimates in various places, it’s a worst-case scenario I can live with. An all-Delta world, even with double or triple current rates, is still better than things used to be even for the unvaccinated, that’s the peak of trouble before things get better again, and the unvaccinated have a choice to change their status. 

The usual suspects are out in force, as one would expect, raising the alarm. On the lighter side, some of them cut more corners than others…

Did you know that you… can… just… put… lines… into… graphs? And label them however you like, with no relation to reality? Don’t wait. Use y = ax + b today!


This is the second year of Covid posts, so when we point to the South and say ‘this is what happens when people refuse to get vaccinated’ perhaps we should remember what the headline was a year ago?

It was 6/18/20: The Virus Goes South, followed by 6/25/20: The Dam Breaks. It looked like the South, plus Arizona, had lost control of the situation entirely. Then two weeks later, things there peaked and started declining again. Once again, I had underestimated the control system. It wouldn’t be my last time.

Exactly one year later, the exact same spots are showing the same pattern, to a lesser extent than last time. It hardly seems fair to ascribe this primarily to vaccination differences, even if the math says that those differences are a very big game. This is not our first rodeo, and I’m not about to repeat the same mistakes that easily. 

This goes both ways. We also shouldn’t get too cocky about things going well in the Northeast in the same way they went well last year, even if we have good reasons to be optimistic this time around.

In Other News

I remember what it was like to have this kind of failure of imagination. I miss those days.

Claim that vaccinated children in Los Angeles will be required to wear masks while at the places they are required to report to for the bulk of their day. On the plus side, if they opt for ‘distance instruction’ they only get ‘scant hours of instruction’ leaving them free to spend the rest of their time learning. 

Also worth noting is that I don’t think this is how math works, on multiple levels? Am I missing something?

Thread on vaccine situation in Taiwan, which is a disaster on multiple levels, and its interaction with the question of reunification with China. Especially after what happened in Hong Kong, I notice I am confused how so many could take such an indifferent attitude towards reunification.

MR once again hammers home that the Moderna dose of 100ug is clearly too big. I continue to think this undersells the case, because 100ug isn’t merely unnecessary overkill, it is likely actively worse than 50ug because the extra size makes the next-day side effects worse, without a meaningful increase in the level of protection.

Paper documents loss of grey matter in the brain after getting Covid-19, including for those who were not hospitalized – hospitalization did not seem to impact the magnitude of this effect. You do not want to get Covid-19. Given the timing this does not provide information on vaccinated people who then still got infected, nor does it differentiate between severity levels beyond whether someone was hospitalized. I do not have a good sense of what size impact one should expect from the effect observed here – it’s easy for this type of thing to be quite impactful, and also easy for it to sound scary while not having much impact at all.

Small study finds 39% risk of Covid-19 spread between roommates in hospitals. I didn’t get a chance to form an unanchored prior on this number but it seems well within the range of numbers I would have expected.

Mask wearing study with very large data sets finds that everyone wearing masks most or all of the time in public places leads to a 25% reduction in the rate of reproduction (R0). That is being interpreted as ‘masks work’ but I would caution both that the error bars on this are gigantic in both directions, and that this effect is actually smaller than I would have guessed, if it is indeed contrasting mostly wearing masks to never wearing masks. They couldn’t find any effect of mask mandates on rates of mask wearing, which is a methodological problem rather than a real observation of causal reality. Mask mandates absolutely change the rate of mask wearing, and the lifting of mandates has clearly reduced mask wearing in ways I have observed with my own eyes. I do think it’s fair to take results like this and update towards mandates having less impact than one would otherwise think, especially local mandates in a world where masks were being adopted generally anyway, and to put more of the ‘work’ involved on private reactions. The mistake would be either to translate the ‘masks work’ here as ‘mask mandates work,’ which they definitely did not find, or to translate the ‘mask mandates not found to increase mask wearing’ here as evidence that they don’t increase mask wearing. Two easy to make and opposite mistakes.

Someone Really Ought To (Somehow) Do a (Better) Study

Marginal Revolution highlights a NYT piece on the CDC and how broken it is, calls it one of the best pieces of the year. Giving the secondary link as a compromise with my NYT ban, as the content here is being represented by a credible source as unusually impressive and important.

A study on shelter in place orders and their impact on excess mortality. My conclusion is that the decision on when to issue such an order, and the counterfactual situation if orders weren’t issued (both in terms of people’s actions, and in the medium-term path of the pandemic), are sufficiently hopelessly confounded that this doesn’t provide much if any insight into what is happening here. 

Another study finds that lockdowns kill children, especially in developing areas, ‘potentially’ 1.76 child deaths per pandemic death averted in the least developed areas. As usual, they use an SIR model, which is a very poor way of creating a realistic counterfactual path of the pandemic in the absence of lockdowns, so I don’t take these results all that seriously beyond the headline fact that there’s very large downsides to locking down that are even larger in undeveloped areas. That part seems very true, and worth appreciating.  

Not Covid

Lying with statistics, CNN reversed x-axis edition:

John McAfee found dead in a Spanish prison while awaiting extradition to the United States. 

That ‘could have’ is doing a refreshingly large amount of work, but, I mean, no:

Everything we know about John McAfee, including getting a tattoo years in advance to indicate he didn’t kill himself in case he was found in exactly this situation, strongly indicates he didn’t kill himself. 

That’s a pretty strong move. If you did that and people still think you killed yourself that has to be a rolling-in-your-grave-thinking-literally-what-the-f***-did-you-want-from-me moment. Or it would be, if John was the type to care what other people think enough to bother rolling.

Still, while I find it unlikely, the ‘could have’ technically stands! Cause I gotta think: If McAfee explicitly set the stage to make us all think he was whacked by the government and would never kill himself, then killed himself anyway, would that be both a totally baller move and totally in character? Absolutely. And I’d have to put on a hat, so I could tip it.

In gaming news, while working on Emergents, I’ve had the chance to explore Roguebook, the new rogue deckbuilder from a team that includes Richard Garfield (the creator of Magic: The Gathering). Richard is awesome and takes lots of big swings, and this is effectively a spiritual sequel to Slay the Spire, which is so good it is the game I recommend to most people when they ask me what they should be playing, so I was hoping for greatness. What I got instead was… a solid game that does not swing for the fences and I don’t think is as good as Slay the Spire, but which offers enough new twists on the formula to be well worth playing if you’re up for another experience in the same style. I tentatively have it in Tier 3, a game worth playing if you like the genre.

The game I’m enjoying the most recently is a little thing called Slipways, a chill puzzle/strategy 3X game that’s definitely worth checking out. Not yet sure if it’s quite Tier 1 (Must Play), but definitely at least Tier 2 (Worth It). 

Finally, did you know that supplies are sufficiently backed up that I’ve been warned that if I want to order nice furniture for our new apartment in NYC, it might not get here until January? 

This entry was posted in Uncategorized. Bookmark the permalink.

28 Responses to Covid 6/24: The Spanish Prisoner

  1. Brockenborings says:

    Are U.S. officials in the habit of sending “subtle threatening messages” to ex-pats? I mean, I gotta say that I give the “totally baller move” you mention a pretty high probability.

  2. bmkr says:

    I’m not sure that the twitter statements of people who endorsed Vermin Supreme for president are really going to move my priors in any direction on the question of whether they killed themselves or were secretly murdered.

    Also, why does the Conspiracy Gang think he was killed? A cursory google search doesn’t really reveal anything.

  3. Everything I’m reading about John McAfee suggests that “totally baller moves” were his thing.

    • John Schilling says:

      Yes, high prior of McAfee doing “totally baller moves” in general, though not this one in particular. Low to moderate prior on people who swear they would never commit suicide, changing their mind when staring down the barrel of life in prison after 75 years of a life of totally baller moves. Low prior on McAffee in the past decade or so having done anything that would make a major power or the like go through the trouble of actually assassinating him. And if any of his older work had worth-killing-over aspects, moderate to high prior of his being simply killed when he was living weird in Belize, which didn’t happen.

      I haven’t tried to do even BOTE math on this, but this isn’t Jeffrey Epstein and I’m not jumping on the murder train.

  4. Matty Wacksen says:

    >Paper documents loss of grey matter in the brain after getting Covid-19, including for those who were not hospitalized – hospitalization did not seem to impact the magnitude of this effect.

    I’d bet large amount of money on this study being nonsense. Covid was not randomized to the participants. It could even be that “cognitive decline” causes covid, or it could be noise.

    Also, ” We further compared COVID-19patients who had been hospitalised (n=15) with those who had not (n=379),and while results were not significant, we found comparatively similar findings to the COVID-19vs control group comparison”

    • TheZvi says:

      Yeah, that was mostly my conclusion as well. Doesn’t mean it doesn’t happen but it does mean they didn’t prove anything.

    • Dave Baker says:

      Might be p-hacked as well? Good chance they looked at a bunch of numbers and this was the one where a significant effect showed up.

  5. Jake says:

    I think it’s reasonable to link to a NYT article provided that article is vetted by a credible journalistic institution like Marginal Revolution.

  6. myst_05 says:

    Its so strange that we’re still using Alpha-trained mRNA vaccines. Moderna announced in May that their booster has been a success against variants (, but still no follow up on when vaccine factories will switch over to focus on UK/India strains instead of the original strain. Makes me more and more pessimistic on how quickly we could pivot into a new vaccine if the current one starts to become useless against a potential new variant.

    • TheZvi says:

      It’s not strange to me at all given everything else that’s happened. Not saying it isn’t really stupid, but it’s not strange to me.

  7. waltonmath says:

    Re: “I don’t think this is how math works”, am [i]I[/i] missing something? Perhaps context? All the constraints in the quoted paragraph seem fittable to me, though it does look like class sizes outside of a narrow range will be unlikely under “the tentative agreement”.

    • Lambert says:

      I don’t know how the US funds its schools, but the economics tend to work out such that class sizes much smaller than the maximum permitted are not viable. (income is per-pupil and you’re hiring teachers who have the same salary regardless of whether they’re teaching 25 kids or 30.)
      Thus class sizes do tend to sit within a fairly narrow range.

    • TheZvi says:

      Biggest issue: There’s enough overlap between the categories that I don’t see how one could possibly satisfy all these requirements at once.

      • waltonmath says:

        I think I can read the categories as entirely non-overlapping, e.g. K-8 schools don’t count as “middle schools”. One thing the statement seems to leave ambiguous: are the class sizes averaged across grades, or grade categories? *shrug*

      • waltonmath says:

        It does seem like you could get inconsistencies by reading the statement as implying, for example, that averaging over third grade class sizes yields both 22 students per class and 30 students per class.

  8. Adam B says:

    This morning NPR was pushing a study, recently published in the BMJ, saying that COVID caused a 2 year drop in US life expectancy:

    That seems rather a lot, given that the rule of thumb is that *conditional on catching covid* it gives you 1 year’s worth of all-cause-mortality risk. As far as I can tell, this is because their definition of “life expectancy” assumes you are not only at risk of COVID-19, but also COVID-20, COVID-21, COVID-22, etc., with a new equally-deadly virus every year for the rest of your life. Surely somewhere around COVID-33 we’d finally get together to ban GoF.

    One thing that puzzles me is why if they use that definition they say they *don’t* see a similar drop in life expectancy in other peer countries, many of which had comparable COVID death rates.

    • Michael Hollander says:

      i think it’s too early to know this. a lot depends on how life expectancy is computed. if you don’t take into account how sick the people who died were previously (and we know covid kills sick people at a far higher rate), you could overestimate the change in life expectancy. and then afterward you’d have a kind of boomerang effect, where people who might have been dying (due to their existing illnesses) but slower, forcing up computed life expectancy.

      what i’ve read conforms with something like: conditioning on age alone the risk of dying from COVID if contracted equals about one year of risk from all other factors together. so i’d expect if everyone in the population got it, it would drop life expectancy by a year. so say 4 months in the US.

    • TheZvi says:

      I am willing to say that those numbers don’t reflect reality, and instead reflect the particular way they’re doing the calculation – they’re assuming that various bad things that happened in 2020 are at least semi-permanent I think?

      • Adam B says:

        Correct, the life “expectancy” is not the expectation value of any sensible quantity, unless you expect there to be a new pandemic every year so that every future year’s death statistics are the same as in 2020 and you have to survive ~10 covid-like infections over your lifetime in order to reach old age.

        However, this still doesn’t explain why they didn’t see a similar drop in peer countries that had similar pandemic death statistics to the US.

  9. Tim says:

    Any thoughts on Ivermectin being strongly supported with this recent meta-analysis?

    At this stage, I believe we should focus on three things:
    – global administration of mRNA vaccines
    – encouraging the adoption of FLCCC Protocol:
    – dealing with the mental health damage.

  10. Stephen Hayes says:

    I notice a conflict between “Everything we know about John McAfee… strongly indicates he didn’t kill himself” and “If McAfee explicitly set the stage to make us all think he was whacked by the government and would never kill himself, then killed himself anyway… totally in character”. It seems like we know something about his character indicating he might have killed himself.

  11. Triskele says:

    Two questions you may or may not know anything about…

    a) Is there research showing that Delta is actually case-by-case more dangerous or is it simply by virtue of being more infectious that the likelihood of a bad experience gets distributed over more candidates? I’ve seen two sort of competing claims – 1) the simple one is that with Delta you’re maybe 2x more likely to end up in the hospital, and 2) with Delta, more people pop on PCR tests experiencing light symptoms like runny nose and sneezes, which aren’t on most watch lists. So, is Delta subject to another denominator problem or, at the level of an individual, the real deal?

    b) Our regional health authority is alternating second shots between pfizer and moderna, depending on availability, saying “they’re basically the same vaccine”. Is this nonsense, reality, or a reasonable compromise given the situation?

    • Eric fletcher says:

      Even if Delta is exactly the same as prior strains once it gets in the body, but more infectious, that would lead to higher initial viral loads, which by itself causes more severe cases.

  12. TheZvi says:

    General comment: It warms my heart, in a strange way, to see such a high % of the comments be on McAfee, since that means things are going well enough in virus land for us all to be distracted by normal news again just because it’s fun and shiny. Not going to get into the weeds on it further – hopefully it was clear I think anyone ruling out just about anything without direct knowledge is acting overconfident, and enjoy the ride.

    • Michael Hollander says:

      I’m a bit taken aback, though, about your furniture news. I have a couch on order myself. They’re saying 8 weeks, but they’re also saying “8 weeks” is their inventory system’s euphemism for “unknown and indeterminately long”. We broke our old couch (metal frame broke due to metal fatigue) a few weeks ago, and i suppose we’ll be sitting on the floor a while.

  13. Quixote says:

    If you in in or near the NYC area, I strongly endorse getting used furniture. I don’t think used would be backed up the same way new is because there is no production pipeline.
    The quality of used furniture in NYC is extremely high, there are a lot of fairly high income / high net worth folks who both buy very nice stuff and remodel frequently. Things pop up on apt deco or craigslist that constitute absurd values, like a $19,000 rug for $850, or a china cabinet that runs $3,500 new at macy’s for $200. If you are not constantly chasing whatever today’s fashionable renovation is, you get extremely high quality stuff at a much lower cost and at a much lower environmental impact.

  14. Ivy Agnes Nguyen says:

    Regarding holiday data lags, I suppose we can assume similar for times when natural disasters/major weather events happen, like the Texas snow/deep freeze earlier this year & the ongoing western heat wave. I doubt nearly as many people will be getting tested or vaccinated on the days these events occurred. Obviously these effects would be more localised to the affected regions, but a large/long enough event or region could skew national stats.

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