Various actions and decisions in wars, and their efficacy, tends to get debated decades and centuries to come. That still doesn't mean countries don't have to take actions when they get into wars, though.
On the last point, isn't the cost to utility just the excess utility? The $1 of lost GDP isn't actually a loss to utility, because I kept the $1 instead of getting the first dollar of benefit, and the seller both forewent the dollar of sales but also forewent a similar $1 of operating costs.
Agreed. Why not look at cases like Vietnam, Taiwan, South Korea, China, Australia, New Zealand etc. where initial strong lockdowns and eradication strategies and then continuing travel restrictions /quarantines brought back normality early on and were then punctuated by localized lockdowns to address outbreaks where they passed the travel barriers? Those are the ultimate examples of lockdowns preventing future lockdowns.
That was mine and many observers’ preferred policy in terms of what to do in spring 2020, and this was followed early on in Denmark for instance, and also in places like Senegal. But then travel opened up without quarantines in the summer of 2020… One can argue this was unreasonable or unrealistic in the US and Europe — but the US still today has a travel ban for Europe since spring/early summer 2020. Just not for other places, due to weird politics.
Why do you only focus on cases where this strategy succeeded? Many other countries pursued similarly aggressive early approaches that later failed miserably, e.g. many South American countries. (That region is also interesting to study because we have Brazil as a natural control.)
Vietnam is a police state, China's lying, and the rest of your list are island nations with secure borders who started with a low infection count. South Korea isn't technically an island but they have the DMZ. All of those countries are unfair comparisons.
Any particular comparison is "unfair". That's the point of a statistical analysis. Are there relevant differences among these countries? Do any of the "unfairnesses" turn out not to matter?
Come on. The antecedent probability that we'd find such a high proportion of islands in a world where the vast majority of countries aren't islands, if being an island was casually unrelated to success in handling the pandemic, is minuscule. And, of course, we have an obvious mechanism that would explain the effect if being an island was in fact the explanation.
Yes, it's clear that being an island in this sense is extremely helpful. Being a police state may or may not be. Even among island nations, we might be interested in what policies (either governmental, or pursued independently by residents) were helpful or not. That's my overall point. We shouldn't just ignore them.
Would be nice to have a larger dataset that includes places like Indonesia, Philippines, Seychelles, Cyprus, Caribbean nations, etc, to get a better sense of what does and doesn't work for island nations.
Why would China be lying? When there is a COVID outbreak in China, it's very noticeable, and the government takes great efforts to inform the people. China also has numerous neighbors that quarantine Chinese travelers; they would notice if China had a pandemic.
China might be lying to some extent (especially about the countryside) but not by three orders of magnitude. It's not North Korea. There are quite a few expats there and they know how to get around the Great Firewall. If the situation in places like Shanghai or Beijing had ever been remotely as bad as in Europe/the Americas/India we'd all know by now.
Those are also countries that are widely separated by oceans and where border controls can be effective. Africa has also done relatively well due to its relative lack of transportation connectivity. In some ways China also benefits from the same dynamics due to its suspicion of foreigners. Certainly their coercive quarantine measures left nothing to chance. New Zealand discovered this when some tourists did not follow the rules and caused an outbreak.
Australian states also closed borders to Victoria's during its second wave, which was doable. Granted there are fewer border communities than in the US or Europe, but it's not impossible to close land borders.
(Australian state borders are currently closed to New South Wales - who are having an outbreak which is small by global standards but potentially threatens our COVID-zero status)
The "island" thing sounds like a pretty red herring to me, unless people believe that the virus reached Italy by sneaking through land borders along the Silk Road, or that flying over the Philippine Sea magically heals all passengers on a plane but flying over Kazakhstan doesn't. Western countries got 99 reasons why they couldn't achieve what e.g. Australia could but land borders ain't one.
By July 2020 Italy had managed to bring case numbers far enough down that it could have implemented an actual test-and-trace program without breaking a sweat, but actually it only implemented a perfunctory one because "muh privacy". In addition, it refused to implement border quarantines -- there were superspreader events when people went to dance clubs right after getting off planes from Brazil or Croatia. Sure, a handful of the seeds of the second wave might have sneaked through the land border with Slovenia, but very clearly that wasn't the main issue.
The government can implement restrictions/tests at every airport and port in a way that would be impractical to do at every single road crossing between two schengen countries.
While that is certainly true, and it means that Italy doesn't have a path to get Australia like pandemic numbers, airport and seaport lockdowns could plausibly have meant numbers more in line with the rest of Europe.
Certain Schengen countries had shut down their borders altogether during the first wave, keeping them open with restrictions can't be much harder than that. (Of course there might still be people crossing them illegally, but probably not enough to go from "test-and-trace programs feasible" to "test-and-trace programs unfeasible". Most illegal immigration to Italy comes from the sea, not from land borders, and Australia isn't that much farther away from Indonesia/New Guinea then Italy is from North Africa.)
>there might still be people crossing them illegally, but probably not enough to go from "test-and-trace programs feasible" to "test-and-trace programs unfeasible".
Presumably this depends a lot on the nature of those land borders. Do we have any idea how many people have been illegally crossing between Detroit-Windsor or Tijuana-San Diego or in smaller places like Basel or Konstanz?
> The "island" thing sounds like a pretty red herring to me, unless people believe that the virus reached Italy by sneaking through land borders along the Silk Road
Cases got through to Italy faster through other routes (air) than they would have through land borders. However, this doesn't disprove the argument that *even if* Italy had prohibited entry from abroad as much as feasible, cases would have eventually snuck in through land borders, while the same is not the case for Australia or NZ.
The infection spread Wuhan > Milan > London > NYC before ANYONE knew it existed. There were effectively four virus origins, because those four cities are connected by the fashion industry. Wuhan is where the garments are made.
At the time that Milan was blowing up, we also had blow-ups in Qom/Tehran and Daegu/Seoul, and by the time London and NYC had big blowups we also had Madrid, and could have had Seattle.
Is Wuhan a larger garment manufacturing center than other cities in China, Southeast Asia, or Bangladesh?
Isn't the U.S. getting about 180,000 illegal immigrants every month, which are dispersed around the country immediately? How could any travel restrictions possibly be effective under those circumstances?
This. Anyone who thinks a zero covid strategy could ever have been feasible in the US for this reason is basically either ignorant of the realities or illegal immigration or bonkers.
"Lockdown test and trace" only works if you can test every single person who crosses your border. Island nations were the only places in the world that could do "test and trace," and they needed a low infection rate to begin with to pull it off.
China's numbers can't be trusted, and Vietnam has border security Trump could only dream of and still caught a mid summer infection wave from illegal Chinese immigrants.
I see this a lot (and have posted about it a lot before on this forum and others), but I think people still don't realize how insanely behind the US New Zealand's COVID was at the time to lockdown. Generally speaking, most countries woke up to COVID around March 12th (Rudy Gobert + Tom Hanks). Assuming a 3 week infection to death period, we want to know deaths as of April 2 to infer infections.
On April 2nd, the US had a total of 8000 deaths, implying almost a million infections, and had 1500 people dying a day (and rising) implying 150k daily infections.
New Zealand had 1 death. New Zealand had 5 confirmed cases on March 12th. So the comparison to New Zealand is basically asking why couldn't the US have locked down 5-6 weeks earlier - that's the only time we had a hope of NZ like outcomes. By March 12th the cat was out of the bag - our outbreak was already worse than it got in any of the Vietnam/Taiwan/South Korea ON the day we locked down.
Now Australia is a bit more of an interesting case, and probably the strongest pro lockdown one out there. But still - they had ~900 people die by the end of their summer wave, with a population of 25 million that means on a per capita basis their entire summer wave resulted in as many cases as the US already had by March 15th. It's just a completely different level of disease.
We didn't have a Summer wave - it is Winter in July in the southern hemisphere.
The July wave was entirely confined to a single city, so dividing by the total population makes no sense. Outside of Melbourne cases were ~0. On a per capita basis the height of the Melbourne outbreak were roughly comparable to the September lull in the US. If you were willing/able to implement right border controls you could have done the same. The US just lacks the state capacity to pursue an elimination strategy.
Ok, so Melbourne looks to be ~5 million people and NYC is ~ 8 million? But the greater NYC metro area is ~ 20 million, but a lot of those are in non New York States. So I may not exactly be comparing apples to apples but lets go with this.
Victoria has had 820 COVID deaths, let's just assume those are all in Melbourne.
On April 4th *alone* NYC had 819 COVID deaths. By 3/29 they had more deaths in New York (state, but mostly City at that point) than Melbourne did over the entire Pandemic. Even if we assume the 20 million number and 4x population, we hit that on 4/4.
Again, it was just a completely different level of disease burden in the US before lockdowns started. This matters a lot - if we assume a non-China type lockdown can get R down to 0.6, 12 weeks of lockdown turns a million cases into 2100. Which is still a decent amount of cases! R of 0.7 (an estimate I've seen for US lockdowns) puts you at 13,000 cases - very much not an under-control outbreak! If you only have 50k cases, you get down to 110 cases with a 0.6 lockdown, and 700 with a 0.7.
Yeah, it's pretty obvious when you do the back-of-the-envelope math that if some place in the US like NYC had tried to get to zero cases in the initial lockdown, they would have needed to enact a harsh lockdown for 6 months plus.
Yes, but I'm saying that such will work way less well if your starting point is USA 3/12 disease levels as opposed to Melbourne name literally any date disease levels.
The Atlantic provinces in Canada are another example of this strategy. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
The Atlantic provinces (Newfoundland, Nova Scotia, New Brunswick, Prince Edward Island) in Canada are another example of this strategy, and I haven't ever seen them mentioned. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
The Atlantic provinces in Canada are another example of this strategy, and one that is pretty underreported. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
Was added in response to this comment, and a couple of others. I definitely don't think it's comprehensive enough, even with the addition. I just feels like a lot of key data points are getting left out of the analysis (Not just Au/NZ, Taiwan and Korea as well to name a couple more)
Melbourne had a really harsh lockdown for *112 days*, which they did because "test-and-trace" completely failed for them (or they were insufficiently competent to implement it successfully), in order to get cases down from 100 per day (when lockdown started) to zero.
The fact that this happened is actually a really strong argument against "get to zero (or near zero) initially, then control by test-and-trace". If the latter part of the "optimal policy" isn't feasible, this dramatically reduces the payoff to the first half also.
Now, it took 112 days of pretty draconian lockdown in Melbourne to get from ~100 to 0. The initial lockdowns in the US and Australia happened more or less the same date (e.g. looks like March 20 in NYC, March 22 in NSW). How many days should New York have locked down for to get from the ~100k infections per day it was experiencing at the time to zero?
And would getting to zero a) be possible or b) make any sense as a goal in any location with substantial illegal migration? (E.g. the US and probably also Europe)
"make any sense as a goal in any location with substantial illegal migration?"
Probably not. But that's what internal border controls are for.
I don't think Melbourne eliminated COVID in the shortest possible time; China has regularly tested entire cities and eliminated the pandemic in numerous areas that way (e.g., look at its control of the recent Guangzhou outbreak).
Lockdown was announced on July 7 and appears to have started on July 9. From google (which just pulls the numbers for JHU), the 7dma for Victoria (they don't show Melbourne specifically) was 114 on July 8, and the worst recent day to that point had been 168 four days prior. And obviously the true number of infections was higher, although the positive test rates remained very low (about 0.5%, so hopefully catching the vast majority of positives). So yeah, I was a little under, but not dramatically. What few people seem to remember is that cases then rose for another 4 weeks after the start of lockdown, with the 7dma peaking in early August.
Porous land borders did end up causing Thailand's and Cambodia's current pandemic, but America and Britain never even tried so much as controlling flights. China (which is more ambitious than Thailand and Cambodia) so far has managed to avoid all waves of COVID despite vast land borders.
Britain relies on a regular flow of lorries in and out, with the same driver throughout their run, to supply it; controlling the border was never feasible.
Australia and New Zealand have been so successful because they trade by boat, and the crew of the boat don't have to have much contact with the people on the dock.
Strictly true, but both sides of Torres Strait are incredibly hard to reach by land. The Queensland end is a two-day drive in a winch-and-snorkel equipped 4WD from the nearest meaningful town, and the New Guinea end doesn't seem to be reachable by any sort of road at all.
Interesting to consider how this applies to yearly influenza. All the above things considered, would pro-lockdown people suggest yearly lockdowns until universal flu vaccines? It's unfortunate RNA viruses mutate so damn much.
I have heard a lot of people say that the lesson of this last winter is that we should newly emphasize hand-washing, normalize mask wearing (if sick or just if worried), and similar measures for flu season, given how small the flu season was this year. I haven't heard anyone propose lockdowns though. More in the vein of changing social norms and public awareness.
Not sure if official numbers are out yet, but I heard that the norm is about 20,000-30,000 and this year it was more like 600 - if true, something like a 97-98% drop in deaths. If it had happened without COVID around to rob the headlines and make large numbers of deaths seem trivial by comparison, it would be seen as a massive public health victory.
Actually it raises the question of, if lockdowns don't work then why did that happen? Or is flu different from COVID?
"Or is flu different from COVID?" Qualitatively but not quantitatively. Measures which reduce the R of COVID from 2.5 to 1.25 also reduce the R of influenza from 1.5 to 0.75
Globally, seasonal flu kills about 500k / year. I don't know about deaths, but flu incidence has gone down 99% globally this past season.
The US typically has 30m symptomatic cases, 440k hospitalizations, and 36k deaths each year. Deaths are hard to estimate last year because they are completely drowned out by un-tested covid deaths (which is in the 100s of thousands?). There is one known US pediatric death from the flu last season, compared to average about 200. Last season there were 1800 positive flu tests in the US, 200 of which were hospitalized (I don't know how many died). https://www.cdc.gov/flu/weekly/index.htm
There's nothing in your link indicating an increase in influenza deaths. The official stats listed there are "influenza and pneumonia" deaths, which will almost certainly include some of the untested covid deaths. Covid notably causes pneumonia.
Mortality stats generally uses the category "influenza-like illness" (ILI) because it is hard to reliably distinguish flu from a variety of other respiratory diseases. Generally most ILI is the flu, but obviously one wouldn't assume that is still true for 2020.
If you have any statistics about distinguishing influenza from ILI please do share. I have linked the official CDC data above which has information based on actual flu testing, showing a 99% decline in lab-tested flu.
I just see someone else pointed out that that paper is about deaths in the 2019-2020 flu season, which was significantly worse than usual. The paper you linked says nothing about the 2020-2021 flu season, which reliable data shows was about 99% smaller than usual.
As well as anti-covid measures cutting flu transmission, consider the possibility that a substantial number of the most vulnerable people who in other years might have died of flu might have died of covid first instead.
Well, influenza has an IFR depending on age non-monotonically, with a minimum in your early 20s (IIRC) -- infants have a pretty much negligible chance of dying from COVID but not from the flu. (But in first-world countries other than e.g. Israel or Ireland I guess there aren't anywhere near enough of them to sway the statistics either particular way, anyway.)
That paper is only about the US and as far as I can tell about the calendar year 2020. In the Northern Hemisphere most flu deaths are in January and February, so before anywhere outside Hubei implemented COVID-related restrictions (other than bans on direct flights from China).
Flu deaths in the 2020 southern hemisphere winter (July/August) and the 2021 northern hemisphere winter were down by several orders of magnitude from usual, IIRC
I have written an opinion piece advocating for increasing measures against the flu: https://ermsta.com/posts/20210527 "Defeating influenza: lessons from covid"
However I don't advocate for lockdowns, as I believe they are wholly overkill. I believe vaccinations, paid sick leave, contact tracing, border screening, situational masks, and self-administered home-testing are the way to go, in vaguely decreasing order of importance. I put a low emphasis on measures that require high compliance from individuals (physical distancing, hand washing, masks). Also hand washing has little impact on flu transmission.
The ambitious goal for flu is to vaccinate south and east Asia (<5% of India and China get flu vaccine), where flu spreads year round and may be the source of flu elsewhere in the world.
Sounds wonderful. I suggest you supplement this with a comprehensive dose of surveillance to catch anyone who is noncompliant. You could setup something like camps for those who do not show enough solidarity to join, or give them a "compliance helper" who watches them to motivate them to do better.
Currently, I've seen no indication of a significant decrease anywhere else with reliable mortality numbers, though I'm sure it exists, but you'd need to at least tally up well-monitored countries and their number before making this claim.
Given that the US accounts of a large % of the population where COD can be monitored well, I'd say it's very likely your source are wrong.
Before the pandemic, I did used to wonder how many sick days would be saved by people following basic hygiene - e.g. hand-washing, not going to work when ill and potentially contagious.
I've been pretty consistently pro-lockdown but wouldn't support them for flu because the threat of flu is much lower. But there are plenty of other things we should do to reduce flu deaths - masks on public transport (if you're ill but have to travel); work from home (if you can) when potentially contagious; improve ventilation in buildings; get vaccinated.
The data in the Scientific American hot take is from the WHO. The data in the JAMA article is from the CDC and their dedicated flu report seems to think the flu season is milder:
The numbers you're talking about are flu and pneumonia combined, and include a number of codes that aren't influenza. Unless I'm misunderstanding the actual data, the "flu and pneumonia" deaths can include lots of non-flu deaths. Given that the CDC and the WHO both produce reports saying that the flu has been minimal this year, I'll go with those.
I don't want to go into how diagnosis on small samples might be wrong, but I'd say it's fairly intuitive to think that they are more prone to manipulation and misinterpretation than country wide mortality, which is the most uncontroversial number available.
Yes, various forms of pneumonia and influenza are combined, in part because differentiating between which of them was causal to the patient's death is hard, but I've got little reason to believe there was a giant spike in covid unrelated pneumonia this year, though it may be the case.
Have you looked at the actual number of people the WHO and CDC are basing their estimstes on ? I.e the people being tested for the presence of various flu strains in the analysis they use.
Probably not lockdowns, since the flu is significantly less dangerous than COVID. But, as others had said, normalizing mask-wearing at the very least does make sense.
I think the number 1 problem with that strategy is simply that compliance would be very low. Many people already were dubious of following COVID lockdowns, even when they felt they might be necessary. I think very very few people would be able to be convinced that we suddenly need to be having flu lockdowns when we’ve survived without them basically forever.
It should be noted influenza deaths increased significantly in the US, ~20% accord to the JAMA end-of-year mortality report I read (death being the closest proxy for infection rate in this case, since people aren't neurotic enough to confirm it)
Of course you could come up with a just-so story about how covid-19 and hospital closure somehow causes this (seems unlikely, since there's no correlation with death from influenza after covid, and common sense could well help us infer it to be otherwise, a strong immune reaction remains protective once the pathogen is gone + hospital treatment for influenza is mildly effective to very ineffective to harmful depending on who you ask)
But it still seems like the increased sanitary measures did didly-squat to limit the spread of influenza. Which is rather surprising, but not so much if your working hypothesis is that of a conspiratorial nut (e.g. covid-19 is real but "what mass-media and most review literature portrays to be covid-19" is not)
What is the purpose of performing multiple high and sub-level analyses of changing definitions of what constitutes a case, what constitutes a hospitalization and what constitutes a death? First of all, more testing initially means a spike in cases. There is already a high false positive rate for testing regurgitating dead nucleotides and amplifying it many times just to get a match. Threshold sensitivities were modified for non-vaccinated vs vaccinated patients. Case counts were modified from 'clinical x-rays' to '# of symptoms' to 'if you had a positive test within 28 days, inclusive if you died from a ladder or car accident'. Testing was eliminated for those who were already vaccinated or partially vaccinated. Older patients were forcibly locked up together in elder care homes with positive cases, even though the # of co-morbidities for each case due to death is already high, and has an average age in the 80s... The cofounding incentives of diagnosing patients as positive or getting treatments, getting payoffs either by medical practitioners or hospitals. The ad nausea footage of the Chinese falling ill and locking up of doors with piling dead bodies when the elephant in the room is that countries like Russia/China and some states in the United States have been fully opened for more than one year yet the cognitive dissonance when none of this materializes in the West. The initialization of pandemic by case counts, not by death rates as changed per WHO's guidelines a decade ago. Moving targets, changing definitions, high compliance rates yet no difference whatsoever or strong correlations when compared to different nations that didn't even bother reacting. Flu cases mysteriously disappearing or causes attributable to other causes of death being removed, when people are denied services to hospitals for testing/treatments because of a focus solely on this .. illness. The alignment of recurring seasonality of influenza/flu season with spike in cases, also post-mortem death rates following 'variants' / high vaccination rates. There are a lot of confounding variables that make this whole point moot. Placebo groups are entirely vaccinated.. already or they weren't treated with a negative control but an alternative vaccine for measles. No baseline studies for animal mortality. Multiple stakeholders/shareholders in multiple corporations/media companies/fact checkers/social media platforms incestuously having a strong profit incentive for booster shots or stronger lockdowns.
What exactly is the point you are trying to make here? Because it feels to me like you're trying to imply a conspiracy without outright saying that COVID-19 was a hoax in the US. Replying to all of these claims would be difficult, but there's at least one I know is absolutely wrong.
> Case counts were modified from 'clinical x-rays' to '# of symptoms' to 'if you had a positive test within 28 days, inclusive if you died from a ladder or car accident'... Flu cases mysteriously disappearing or causes attributable to other causes of death being removed
We can be pretty sure that the deaths attributed to COVID-19 were actually caused by COVID-19. The CDC has a graph of excess deaths here (https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm) which shows that for the duration of the pandemic more people than usual died. This number of excess deaths aligns with the number of COVID deaths that hospitals reported, and if anything, indicates that hospitals under-counted. That is to say, unless you can think of some other reason for six hundred thousand people to die in the United States, the only thing they can be attributed to is the one thing that was different in 2020 versus 2017 - COVID.
Likewise, we can be sure that these aren't just flu cases being misreported because that would imply that the Flu was suddenly many times more effective in 2020. This would be especially confusing given that people were self-isolating many times more strictly than before. More importantly, if it WAS some kind of hyper-deadly variant of the normal flu, the reaction by the government and people should have been exactly the same - whatever level of lockdowns they deem appropriate to save the most lives.
2020 was not only different in terms of COVID. You're assuming all the measures didn't themselves create any deaths, which is unreasonable given that a key effect of lockdowns is restricted access to healthcare.
The point the guy is making is that unfortunately this analysis is not very good and actually quite typical of the stuff you find in the epidemiology literature. I normally quite enjoy SA's essays but this one just had me whacking my head against the table quite a few times.
• Taking model simulations as credible despite that those models are consistently askew? Check.
• Assumption that germ theory is complete and the behavior of respiratory viruses is well understood, even though real world data contains many things that it cannot explain? Check.
• Assumption that "COVID deaths" actually means "died of COVID" although it doesn't? Indeed!
Still, at least the evaluation of the Flaxman study is reasonable.
Look at Swedish excess death for 2020. Depending on how you select the base rate you can get anywhere between zero and 9000 extra deaths. That is not a serious number, the absolute level is the same as in 2012. Modelling predicted 90,000+ COVID deaths if they didn't lock down hard. That is not a success for lockdown nor modelling.
I also have a bone to pick with the attempt to separate voluntary behavior from "lockdown". People's behavior changed because they were lied to by modellers and other assorted pseudo-scientists like Flaxman and Ferguson, and that's also what drove formal legalized lockdowns. Those two things can't be separated. The real question is not "do lockdowns work" - they don't and this can be easily seen in the raw data sans models, it does not require complicated analysis (the economist was on the right track!). The real question is what effect reduction in social interactions and mobility had, versus the costs of that, independent of whether it was compelled or not.
Finally, we should also count not just the economic costs but the destruction of the social capital of science. When you see that the totally fraudulent Flaxman study was cited over 1000 times and is just an exercise in self-delusion, it destroys people's confidence in all institutions but especially scientific institutions. That story has been repeated ten thousand times throughout this era and will have large and unclear effects on society going forward. Certainly, society seems to be splitting into people who take academic modelling and government data seriously, vs those who write it off as the work of the systemically biased.
> Assumption that "COVID deaths" actually means "died of COVID" although it doesn't
I don't think Scott literally assumed this. I think his assumption here is that whatever the discrepancy is between these two numbers, it isn't systematically biased in one direction or another. Is there any reason to think that Sweden systematically over-classified people as "died of covid" while Norway systematically under-classified people as "died of covid"? Unless you've got something like that, the analysis should still stand, with just slightly larger error bars.
Also, the bit about sending deaths back to the level of 2012 - that seems to me like a huge cost, no? Or do you think that a decade of medical progress is negligible?
Firstly, 8 years isn't a decade (2020 vs 2012) and you're assuming that falling mortality rates over time are all to do with medical progress, vs things like healthier lifestyles. I don't think anyone really knows exactly how the progress there breaks down.
You're right that the systematic over-counting would be relatively consistent, however, lockdowns are ultimately justified on *absolute* terms, not relative. Absolute numbers and the exact values on the Y axis are critical. Look at your reply for what can happen otherwise - you're asserting that going back to the 2012 level is a "huge cost" with no idea of the actual absolute numbers involved (they are quite tiny on the scale of a country).
Was anyone in Sweden freaking out in 2012 about the horrible amounts of death that year? No. Nobody noticed nor cared. It was a normal year. This should be a clue that something has gone seriously wrong with our perception of risk and mortality.
(note: the comparison of 2020 vs 2012 is one of population adjusted mortality).
I feel like there's a very common reasoning error amongst highly educated people that's been visible throughout COVID times, but also I've seen it in earlier years too. It's something I keep meaning to write about on my own blog but other things always take priority. The error I mean is reasoning exclusively in relative terms, and forgetting to look at absolutes. The most inglorious example of this with COVID is the fixation on R and "exponential" growth. R(t) is just the derivative of the graph of positive test results over some very short time window. It doesn't itself contain any insight or meaning, yet experts routinely say things like "we must lock down because R is above 1" and politicians then parrot them. Lockdowns were originally justified on *absolute* grounds where the absolute value was hospital capacity, because justifying them on relative grounds would have been obviously insane, but the moment they were implemented the activation criteria smoothly shifted to being merely "positive test results are trending up" with the absolute values forgotten.
The fixation on exponential growth is another example of this. Viruses don't grow exponentially, they have S-curve type growth in which their "exponential" phase stops very quickly and then growth starts slowing down again, until they reach a peak and then cases start to fall. And this can all occur very quickly as is well known from outbreaks of other respiratory viruses like flu, common cold, etc. Yet people get stuck on the exponential part and act as if left unchecked COVID cases would constantly double until the entire population is infected. Left unsaid is what would happen after that if the scenario were true: the day before "terminus day" half the population would get infected at once and then infections would suddenly drop to zero (yeah yeah, modulo a final day of mop-up because the population isn't a square number, exponential growth is not always a clean doubling etc - you get what I mean). This is all totally wrong and no disease in history works this way, in fact recognizing that all epidemics are bell curve shaped is the founding observation of the field of epidemiology itself. It means the absolute numbers involved are very important, yet, nobody cares. They're all fixated on relative rates of change.
Outside the world of COVID I've seen this in my own industry (software). It's extremely common for people to make statements like "X cannot scale" when talking about software that solves a problem where the maximum sizes are already well bounded e.g. "number of customers we can reasonably be expected to have in the next 5 years" in a mature bank is not going to be 100,000x higher than it is today, simply because that would require the bank to completely take over all of its competitors and then take over most of another continent, which would never be allowed. Often reasoning in absolute terms as well as relative would let you engineer much better solutions but the last time I did a back of the envelope calculation showing that $SOME_SYSTEM being accused of not being scalable could actually work fine, jaws literally dropped and the customer got all excited, asking that I send him the calculations so he could show other people (they were trivial).
I don't know exactly what leads to this problem but it seems common. I think it's because reasoning in absolute terms ("it's a small problem so what's the big deal") sounds less clever, as if you aren't thinking ahead to what might be and thinking ahead is what clever educated people are expected to do.
There's a lot going on here, and it's very helpful discussion, so thanks for that. I think the discussion of relative vs absolute amounts is important for a lot of this, but I don't agree with what you think are the implications of all this.
> Was anyone in Sweden freaking out in 2012 about the horrible amounts of death that year? No. Nobody noticed nor cared. It was a normal year. This should be a clue that something has gone seriously wrong with our perception of risk and mortality.
I don't think that "freaking out" should be proportional to absolute amount of death. I think "freaking out" should be proportional to absolute amount of *death we can do something about*. No one was freaking out in 2012 because the fact that we were on trend indicated that we were probably doing about as well as we could do without drastic lifestyle changes. But the fact that we had the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we *could* do something about it.
> R(t) is just the derivative of the graph of positive test results over some very short time window. It doesn't itself contain any insight or meaning, yet experts routinely say things like "we must lock down because R is above 1"
> people get stuck on the exponential part and act as if left unchecked COVID cases would constantly double until the entire population is infected.
> recognizing that all epidemics are bell curve shaped is the founding observation of the field of epidemiology itself.
Maybe some people are making the mistake, but the founding observation of epidemiology is that outbreaks grow as a logistic curve, that is roughly exponentially increasing at first, roughly exponentially decreasing at the end, with the scale of its maximal point being given as a simple function of R. R is not just the rate of roughly exponential growth at the beginning, but 1-1/R is the absolute level at which the curve turns around absent any population-level behavioral changes that change R.
R is really important for understanding the absolute number of people that will be infected absent behavioral changes, and not just the relative number over the next few days.
> No one was freaking out in 2012 because the fact that we were on trend indicated that we were probably doing about as well as we could do without drastic lifestyle changes.
It's hard to see why drastic lifestyle changes weren't justifable in 2012 but they were justifiable in 2020, assuming the absolute numbers being discussed really are similar. I expect that's the point the OP was making.
> But the fact that we had the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we *could* do something about it.
We always knew we could do something about it. We always knew that widely deploying masks would cut down yearly influenza deaths. We always knew that vaccination, isolation and quarantine protocols could reduce these yearly deaths even further.
Heck, we even always knew that we could reduce traffic fatalities by improving car safety and lowering speed limits.
I think all of these facts were fairly common knowledge, but we didn't do any of these things, so I don't see how your reply actually addresess the OP's argument.
"the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we could do something about it."
That isn't a logical deduction at all. This is again a very common fallacy seen in the last year - the automatic assumption that the so-called "non pharmaceutical interventions" can affect the spread of a highly infectious respiratory disease. That has never been proven by anyone, it's just a modelling assumption, yet people act as if it's an axiomatic truth and then go ahead and make lots of other inferences on the back of it.
So: transient increases in mortality do not automatically imply that those increases could be eliminated via government action. And the data seems to show that in this case they can't.
Moreover, freaking out due purely to *relative* changes is exactly the fallacy I've just been criticizing. You haven't even put concrete bounds on what level of increase should cause society-wide freaking - just asserted that *any* increase in mortality vs prior baselines should do so. What if the rollback had been to 2015? Lockdowns still worth it then? 2018?
Consider this: if you sum deaths in Sweden from 2017+2018 and then 2019+2020, you get a nearly identical number! (the web page I already linked to shows this graphically). If you are willing to believe 2019 was an unusual outlier rather than the start of a new trend, you could therefore argue COVID didn't affect mortality at all. Now what?
This is why reasoning has to be grounded in absolute values. If it isn't then you can end up with wild over-reactions to noise in the data, which we see all the time now. E.g. newspaper stories claiming cases are "soaring" when looking at the graph shows the graph line barely visible above the the X axis.
"R is not just the rate of roughly exponential growth at the beginning ... but 1-1/R is the absolute level at which the curve turns around absent any population-level behavioral changes that change R."
This statement doesn't make any sense. R is a rate. You can't do an equation on it and get an absolute number of people. Again, you are demonstrating my point: there seems to be great confusion here between absolute numbers (people, cases, hospital beds) and mathematical abstractions.
I like this. It's nice of you to explicitly set out your points, because it makes them easier to reject.
First:
> You're assuming all the measures didn't themselves create any deaths, which is unreasonable given that a key effect of lockdowns is restricted access to healthcare.
It's implausible the lockdowns were directly responsible for the deaths of hundreds of thousands of people. It's doubly implausible that those excess deaths due to lockdown would directly mirror the spread of the infection according to our tests, especially in states where the government response to the pandemic was weaker. The number of confirmed cases goes up, so do the number of deaths two weeks later. I don't have time to investigate the correlation here, but you can see it just by comparing the two graphs visually.
> Taking model simulations as credible despite that those models are consistently askew? Assumption that germ theory is complete and the behavior of respiratory viruses is well understood, even though real world data contains many things that it cannot explain?
The models are constantly askew because they're sensitive to initial conditions, because they can't literally simulate the behavior of millions of humans, and because our understanding of the effectiveness of lockdown procedures is constantly in flux. We knew this at the start of the pandemic, every single credible scientist made a disclaimer on their models stating something to this effect. The reason we still use models despite all of that is because they're good at capturing the exponential growth we see in pandemic scenarios even if the precise details vary.
As for germ theory. What gives you reason to suspect that germ theory in particular is incomplete? And what specific behavior do you think that germ theory failed to model in this scenario? Keep in mind that germ theory is distinct from the SIR model and derivatives that most people use when doing naive, initial predictions.
>When you see that the totally fraudulent Flaxman study was cited over 1000 times and is just an exercise in self-delusion, it destroys people's confidence in all institutions but especially scientific institutions
Ah, yes, I remember the time the replication crisis happened and nobody ever trusted science again. Certainly, this singular highly cited study (which presumably had some fraction of those citations present counterarguments) which was only moderately influential, during a crisis where information was scarce, will irreparably damage the reputation of the institutions involved in publishing it and the enterprise of science as a whole forever.
Epidemiology and macroeconomics have *always* been on the list of the hardest sciences to trust, because they basically don't allow for experiments, and require masses of observational data, and deal with the complexities of large scale human society, in addition to the complexities of the individual interactions involved.
Social psychology used to be seen as somewhat more reliable than these sciences, but not since the replication crisis.
"It's implausible the lockdowns were directly responsible for the deaths of hundreds of thousands of people."
Nobody is saying *all* the deaths were caused by lockdowns, only that *some* were and fallacious logic assuming they cause none at all will lead to a conclusion of undercounting when we know that in reality the classification mechanism used must automatically lead to over-counting.
In the UK when lockdowns started, admissions to ER halved overnight. People were being told to avoid the hospitals. Moreover, elderly people who were already infected were flushed INTO care homes, where they proceeded to infect a lot of other people. All these are part of what I consider to be the consequences of model-driven mass hysteria, which we can usefully refer to as "lockdowns" as we don't have a better word for the whole enchilada right now.
"they can't literally simulate the behavior of millions of humans"
I'll echo WC in saying you should tell Neil Ferguson (nb: not Niall, that's a different academic). Because literally simulating the behavior of millions of humans is exactly what his model tries to do. Go read it - it's an agent based model that literally updates an array of people with their locations and health statuses.
Of course, it "can't" simulate the behaviour of the entire UK population so they just run it on a population slashed by some fraction and claim it's the same. By far not the worst assumption they make.
"every single credible scientist made a disclaimer on their models stating something to this effect"
I think you're agreeing with me then, just that there are no credible scientists. Because for sure if scientists were saying "our models will drastically over-estimate levels of death so please ignore them" then nobody got the memo.
Please. "Scientists" have throughout this entire event consistently claimed not only that models are highly accurate, despite overwhelming evidence to the contrary, but that politicians must take drastic action based on their predictions immediately, without even a few days of delay to let people outside the cosy little academic circle poke at them. The Report 9 from Ferguson et al that directly triggered lockdown in the UK didn't even give any uncertainty bounds.
Re: germ theory. We know germ theory must be incomplete because models that implement it are always wrong by miles regardless of what disease they're trying to model. For one it cannot explain why there's more than one peak - epidemiology has (amazingly) nothing to say on the topic of seasonality, why it happens or how that can be reconciled with the simple models they use. A SIR, SEIR, SEIR+ or whatever model is a direct implementation of germ theory in software. I don't know where you get this idea that they're distinct. SEIR means "susceptible, exposed, infected, recovered" which is the level of understanding of a virus that a child would have. These models are all based on the assumption that the only way people can get infected is by directly coming into contact with someone else who is infected (=exposure), and then rolling a 6 on some many-sided dice.
Yet there's lots of ways the known microbiology of viruses can lead to other views, for example, if infected people create viral aerosols that can hang around inside buildings for long periods then being "exposed" is about your contact with a place, not a person. SEIR models don't handle that at all. It would also imply that being outside is safe which we know from contact tracing studies is true, yet SEIR models have no concept of this. If those aerosols could also be moved around via atmospheric convection it would help explain why yearly influenza outbreaks all start everywhere at once instead of visibly spreading outwards, and why it's possible for (non SARS-CoV-2) coronavirus outbreaks to happen in entirely isolated populations - a topic that has been investigated briefly in the pre-modelling-era epidemiology literature but never solved.
Will Jones has been doing some good explorations of this topic if you want to read about it more.
As for trust in science - the Flaxman study isn't an outlier is it? The >1100 citations it accrued despite being an exercise in circular logic is unsurprising because stories of bogus studies becoming highly cited are a staple of Scott's writing. As for the citations arguing with it - nah. Academics hardly ever do that, especially in epidemiology. Go pick citations from Google Scholar and see for yourself. I just did it, the results are standard. The dominant theme you'll find for this paper and others is that citations virtually never argue with the paper they're citing. Gwern has some stats on this, negative citations are something like <2% of all citations. Real citations are usually almost random. Some of the ones I just spot checked cited Flaxman et al as evidence that "health care workers and seniors are at particularly high risk" (the paper has nothing to do with that), that "aggressive measures appear to be successful" (the paper takes this as axiomatic), that "stay at home orders were widely instituted" (the paper is a very indirect way to show that) and for assuming a serial interval of 6.5 days (the paper takes this as its own assumption).
So actually yes, watching scientists engage in a massive circular back-scratching campaign on the back of a completely fraudulent paper actually will irreparably damage the reputation of the institutions involved. Indeed it already has. Good luck finding average people in the UK who think Prof Ferguson and Imperial College aren't a bunch of illegitimate shysters.
If one good thing is to come from all of this it will be universal distrust of experts bearing mathematical models. It will probably just be the red tribe though.
Why is "universal distrust" a good thing? Shouldn't we have better calibrated moderate trust, rather than just flip from "expert always right" to "always ignore expert" or even "expert always wrong"?
I don't think people do calibrated trust. I also think that laundering conjecture through mathematics is really darned rampant and default to distrust is probably for the best. Note, not, "always distrust expert," rather, "always distrust expert with opinion in the form of mathematical model."
I think the problem is that detecting which models are good vs bad is very difficult, takes a lot of effort by trained people, journals will not propagate the resulting analyses (they end up only on blogs that you have to be in-the-know to find), and finally, my experience has been that the overwhelming majority of such modelling papers are bad. So just defaulting to distrust seems like a good heuristic.
Well it will certainly be good for those seeking investment in perpetual motion schemes. No more of that pesky faith in the Second Law of Thermodynamics, which is, after all, merely a mathematical model based on observation -- no underlying theory *at all*.
...although, the more I think about it, the more I agree with you. For example, if there is wider general mistrust that vaccines, seat belts, and household smoke alarms save lives -- all conclusions based on statistics and modeling -- then this will accelerate Darwinian natural selection a bit, and future generations will be a little smarter. I like it!
You read Scott's blog. So you're certainly capable of understanding the difference between empirical observations described mathematically and conjecture stated mathematically. Not that you're going to, of course.
Over time, responses to covid will undoubtedly cause some death: e.g., some patients skipped routine cancer screenings in order to social distance, which almost certainly will cause some people to die who would have survived if their cancer had been detected.
But deaths caused by lockdowns and other responses to covid would be more likely to accumulate over time: the ratio of total excess deaths to covid deaths in November-December 2020 would be higher than in March-April 2020 due to the bad effects of nine months of lockdown piling up compared to when lockdown was novel in March-April 2020.
But that's not what we see. Instead, we see total excess deaths being notably higher than official covid deaths in March-April 2020 (probably due to some doctors misclassifying covid deaths as something else before they became familiar with covid) and by November-December, the two figures correspond well.
Yes, the CDC's graph of incremental total deaths is convincing: the U.S. had extraordinarily more deaths in 2020 than would be predicted from deaths per year since WWII. Week-by-week excess total deaths match up extremely well with the usual estimates of covid deaths, and even suggest that covid deaths were undercounted in March and April 2020 and then were pretty accurately counted for the rest of 2020.
Using paragraphs would probably make more people willing to read what you write. Although if it's just conspiracy theorizing it might not be worth reading so maybe you're doing us a favor?
Yeah I think the main thing is that if you care a lot how other people perceive you you're both less likely to rant about conspiracy theories and more likely to type with careful spelling, punctuation, and paragraph breaks.
As a term, "conspiracy theory" is a lot like "criminal": there are certainly some non-central examples that aren't bad but those are rare enough that I don't feel the need to be too careful attaching negative connotations.
But you do have a point. In the future I'll consider trying to use the term "paranoid fantasy" instead, since that captures a little more precisely the meaning I'm trying to convey.
I live in Wisconsin, a state that had a judicial fiat against lockdowns for most of the pandemic but is otherwise very similar to its neighbors Minnesota and Michigan, which had fairly significant lockdowns. I've been hollering to the wilderness that the pandemic really didn't seem that much worse, or the economy any better, than these natural comparisons. Nice to see someone do the work to confirm my suspicions.
fwiw, as someone who lived in Wisconsin under voluntary lockdown, I saw very different outcomes (more cases) among my social group in Wisconsin then I did in comparative internet social groups in other states. Many (20%?) of the households on my block were infected, many (most?) of my housemate's coworkers, etc. Compared to my coworkers in NYC, none of whom were infected, the picture is very different. (Although I'll note that my job instituted a WFH policy about 7 to 10 days before the official NYC lockdown, so that probably is a maximally good lockdown case, compared to my housemate's job, which reopened as soon as the state did)
Better HVAC could be a game changer. Maybe someone needs to invent a ceiling fan with a filter which is continuously irradiated by UV light, which draws all the droplets up to the ceiling and neutralizes them. If this already exists, let me know.
That being said. I’m not sure about the whole analysis-by-country thing. I know that is how most of the statistics are framed. But population density plays a role (as well as people per household).
I think there is a “this is bad” threshold (set perhaps by the media) but when northern Italy and NYC had real crisis, I think that influenced people’s voluntary behavior. If there had not been epic level bad, people might have been going out more. Antimaskers never embraced the badness of it all, for whatever reason. In terms of drawing conclusions for future pandemics, “scare them so they obey” May be the moral. If you are right about the role of voluntary restrictions on individual behavior. Which appears to make sense.
There are so many things that need to be accounted for but are very hard to incorporate. One exception, the age demographics are easy to come by, and honestly I wonder why that isn't standard in all these studies. Another example is pre-existing health, both comorbidities everyone agrees about (obesity, diabetes, compromised immune systems) and those that are more controversial (eg vitamin D deficiency). An issue that seems almost criminal to ignore is the quality of healthcare patients received. Granted, it's hard to quantify and incorporate, but I don't think there's a controversy that initial treatment strategies like putting people on ventilators were between useless and counterproductive, and that the quality of care varied by location and improved over time.
You need much longer exposure to UV light than you can get with normally circulating air.[1] Or you have to use far more intense light for the short time each parcel of air is close to the UV source, to the point where the light becomes dangerous to people (at least their vision).
Hansonian hot take: lockdowns aren't about preventing COVID, they're about preventing dissatisfaction in conscientious distanced citizens, who want official validation that their non-distancing neighbors *should have* stayed indoors as much as them. In other words, staying at home causes lockdowns.
I have pretty much the opposite take: lockdowns cause unrest, which is why China ended theirs quickly (they didn't want people to have nothing to do all day but sit around thinking about how they were unhappy with the government) and blue states ended theirs slowly (they *did* want people to have nothing to do all day but sit around thinking about how they were unhappy with the government).
I think it's a cycle. People get scared and demand the government lockdown then they get bored and get angry so the government relaxes the lockdowns rinse and repeat.
I think that a big thing that causes lockdowns is a prior big wave.
To steal a point from my own comment, in the rest of the country people ask why people in NYC were so insistent on wearing masks for so long, even outdoors where it's extremely unlikely to matter. I think the answer is how bad it got in April 2020. The thing I remember people saying was that the sound of ambulances came from every direction, incessantly, day and night, for weeks on end. The news was talking about how there were emergency tent hospitals in Central Park, how the city had run out of space to store dead bodies. The government essentially told people to stay away from hospitals unless you were having trouble breathing, and even then every hospital in the city had people lining up out the door. The USNS Comfort came to provide relief to the city, and I'm sure it wasn't lost on anyone that the only other time that happened was September of 2001.
The week of April 11, 2020, the number of deaths in the city spiked to more than 7x the normal amount.
> I think the answer is how bad it got in April 2020. The thing I remember people saying was that the sound of ambulances came from every direction, incessantly, day and night, for weeks on end. The news was talking about how there were emergency tent hospitals in Central Park, how the city had run out of space to store dead bodies. The government essentially told people to stay away from hospitals unless you were having trouble breathing, and even then every hospital in the city had people lining up out the door.
> By now you’ve probably seen this NYTimes article describing the horrific state of affairs at a New York City hospital. Patients dropping like flies, nurses reduced to wearing trash bags, refrigerated trucks brought in to house the overflowing dead.
> my brother’s hospital across the city was nothing like that. His hospital admitted many COVID-19 patients, but few have been severe cases and none have died. Today Hillary Clinton (SHE’S TOTALLY RUNNING YOU GUYS) ordered pizza for everyone, and the hospital workers had a pizza party.
I would have to grant that March 27 predates April 11. But I tend to suspect that, given what was described by March 27, the next two weeks weren't much worse.
Were people lining up out the door of every hospital, or were they lining up out the door of some hospitals that made the news?
Don't know about this person's brother's hospital, but according to the CDC in NYC the week ending March 27 there were 2805 deaths in the city and the week ending April 11 there were 7862 deaths. Either way I don't think it changes my overall point.
The Office of the Chief Medical Examiner over on 1st Ave brought back the outdoor white tents for the overflow of bodies. Another sobering sight that was only precedented by 9/11.
That's the kind of thing that gets people's attention.
An emergency room doctor in NYC wrote an essay about how the incessant ambulance sirens suddenly died off around April 7, 2020 about 1 pm (IIRC), suggesting it took about 2.5 to 3 weeks for intense behavioral changes due to covid to have an effect on hospitalizations.
None of the emergency tent hospitals were widely used, no? The Javits Center was famously equipped and then never really used, the USNS Comfort was only used for a handful of people.
And morgue space is perennially tight; it’s a semi-regular story in Detroit that morgues cannot contain the amount of homicide victims. This is a statement on the amount of morgue space rather than the number of homicide victims.
I’m not in the corpse business, but I imagine it is optimized for an average amount, such that even a small increase would result overflow. It does seem reasonable that I would want my morgue always at capacity so I wouldn’t plan for space to sit empty waiting for a catastrophe. Moreover, since corpse handling in the US is well-regulated, maybe retrofitting is not practicable?
Kind of but you could also argue that lockdowns prevent free-riding: if conscientious citizens stay home for the purpose of preventing outbreaks but some free-riders ignore it and don't make any sacrifices, they're reaping some benefit (lower R0 and less coronavirus) without making any sacrifices. It's just coercion to prevent defection.
I'm a little skeptical of assigning some consistent purposeful motivation like this to an action independently taken by at least thousands, probably millions, of individual decision makers. Other than first movers, it probably becomes more about some form of inter-governmental inertia or peer pressure rather than any sort of calculated, considered action with well thought-out reasons. That likely applies to the other side of the coin, too. As soon as a few leaders denounce tyranny and exclaim loudly they're banning lower level governments from shutting anything down, now others that have branded themselves of a similar ideological bent feel like they need to do it too or face a backlash from the half of the 24 hours news cycle their voters pay attention to.
The number of conscientious citizens who honestly care one way or the other is drowned out by the pitchfork mob.
Well, it could also be social signaling. I mean, *almost everybody* doesn't commit murder because they feel it's wrong, and not because of the law against it. That is, if we repealed all laws against murder, I don't think the murder rate would rise by a factor of 100 because everyone would now think it's OK.
So the law against murder is to some extent a signal where we (as a society) tell anyone who is any kind doubt "we all agree this is wrong so DON'T DO IT in case you were at all unsure." In the same sense, you could consider lockdown measures as a version of social signaling, where the majority tells some wavering minority "we think this is much more serious than you do, and we're prepared to enforce that viewpoint on you."
Whether that's a good or bad thing depends not only on whether you think the majority in this case was actually making the correct call (as it does with murder), but also in whether you think the majority should be in the business of enforcing its will on dissenting minorities at all. A lot of room for philosophical disagreement.
1) "This is an absolutely beautiful graph. It’s showing how lockdown strictness (as of May 5) correlates with death rate over time. We find that early in the epidemic, the stricter your lockdown, the worse you're doing. This is the endogeneity... Later in the epidemic, the stricter your lockdown, the better you're doing - probably because the strict lockdown is giving good results."
While the graph indeed looks quite nice, the meaning is much less clear than implied. Unfortunately (from the perspective of being able to do easy causal inference using the time series variation), consider the following two stylised facts:
i) There is substantial spatial correlation in lockdown intensity - e.g. the hardest lockdowns occurred in the northeast and the Pacific west, the loosest restrictions were in the South and plains states.
ii) The various waves of Covid have exhibited substantial amounts of spatial correlation in outbreak levels. The first wave hit the northeast hardest, the second wave the south, the third wave really kicked off in the plains states, and so forth. It doesn't take that creative a thinker to work out plausible climatic patterns that could give rise to this sort of seasonality.
Accordingly, just how sure are we that the picture of "over time, the lockdown state advantage became stronger" is causal? One easy thought experiment is to extend the time horizon of the graph. And indeed, in the short-lived fourth wave (when b.1.1.7 finally broke through, before the vaccines murdered it), something like 9 of the 10 states with worst outbreaks (as measured by weekly cases per capita at that point in time) were states with stricter than average lockdowns. They also happened to be clustered in the northeast and midwest. Of course, this doesn't mean that the stricter lockdowns caused the fourth wave. But it should reduce our confidence in inferring much from that graph, as well.
2) "There was a significant negative correlation (-0.55) between the lockdown stringency index as of January 1, and the number of post-first-wave cases a state had. This was robust to... using all cases instead of just post-first-wave cases (although some of these changes slightly diminished the magnitude of the effect)."
I find it difficult to justify excluding the first wave. Given how rare reinfections are (call it the YOCO constraint), doing so mechanically advantages places that were hit hard early. Which also turned out to be pretty correlated with lockdown propensity.
As you say, the result gets weaker using all cases. But here *cases* have a glaring problem - tests were much scarcer in the first wave than subsequently. So a higher proportion of infections were missed in the first wave than in subsequent waves. So including the first wave but looking at cases amounts to (relatively speaking) throwing out much, but not all, of the first wave. Since deaths were measured much more consistently (both in level terms, and across time), better to just look at deaths. Which almost certainly will shrink the correlation again. (For example, NY and FL have basically identical cases per capita. NY has dramatically higher deaths per capita, though).
I am bothered that quality of healthcare is not a consideration. Doctors got better at treating Covid over time. And good lord in New York at first the hospitals were, hyperbolically stated, admitting people then sticking them on a ventilator until they died. As you pointed out the disease spread to different regions at different times. By the time patients started going to hospitals in Florida they had learned from New York's mistakes. I'm not claiming I know what specific difference this makes but it really bothers me to see it just kind of ignored.
By the sounds of it, yes, the hospitals in NYC were not adding a lot of value in the Spring 2020 wave. (And perhaps were even subtracting value: "hello, your measured oxygen is low, but you otherwise seem in good health. May we puncture a hole in your throat"). That presumably did raise IFRs early on, and to the extent that actual therapeutics are discovered over time (separately from merely learning to avoid bad treatment), this could cause IFRs to fall over time. Alas, pretty unclear how much effect this actually had. The only therapeutics I can think of, off the top of my head, were dexamethasone and monoclonal antibodies. Just how big the average real world effects of these were, and how frequently they were used, idk. (Also, while we probably would expect IFRs to fall over time for these above reasons, if, for example, a more dangerous strain arose, then a lockdown which delays infections can increase overall deaths. It's messy).
In principal, if lockdowns delay infections and IFRs fall, this is captured in the death counts and this thus gets accounted for as a "success" of lockdowns when doing the calculations. (E.g. Florida and California famously have ended up with pretty similar death rates. But CA had its major outbreak much later... it had much lower deaths than Florida for a while... until the dam broke. To the extent the IFR had fallen more by this time, that gets baked into CAs numbers).
If the concern is "states that got hit early tended to have more stringent lockdowns, but were hit early because of reasons X, and this gave them higher IFRs, and we shouldn't attribute this effect of X to the lockdowns", then there is a way of dealing with that (control for X, and preferably do the analysis at the county level). This is a reasonable point - one I would make myself - but is well outside the scope of Scott's data analysis here.
Another thing to remember is that different states had different policy decisions. Cuomo in New York decided to protect the hospitals and sacrifice the nursing homes, while De Santis in Florida chose the opposite priorities, which probably reduced Florida's death rate substantially relative to New York's. (Of course, from a Quality Adjusted Life Year's perspective, a focus on nursing homes isn't as effective as it is from a simple death's perspective.)
By the way, I'm not criticizing Cuomo for his making the wrong decision: all this stuff was brand new to the politicians at the time. (But it is kind of weird to win an Emmy for getting a bunch of oldsters killed.)
I fail to see the value in separating voluntary quarantining(VQ) vs. mandatory lockdown(ML). If everyone is already staying home what are the negative effects of ML we should be concerned about?
Additionally, VQ is caused by fear/concern. Accordingly, we could have increased VQ by increasing media coverage of the negative consequences and risks associated with Covid. Would this have been in any way superior? It seems this actually has negative effects even absent positive effects(people staying home) where mandatory lockdowns do not.
I was thinking a similar thing when reading the post. I know the general trend around here is to worry about government intervention qua government intervention, but I really think the stated economic and emotional costs here are very similar whether consider government mandated behaviors or "voluntary" behaviors caused by fear of the virus, so if we're measuring costs along those axis, then movement data is the most important factor to consider, not stated policy aims.
All else aside, voluntary quarantining takes account of varying personal costs.
My personal cost of quarantining is low, so I should do it whether it's required or not, as long as my chance * cost of getting Covid is higher. A counterfactual me who lived alone, found exercise essential to her mental health and had no space at home in which she could exercise (or any of a dozen other reasons getting out of the house might be essential) should start quarantining a lot later and stop a lot earlier than I should (and did). If you leave the gyms open, I won't go, she will (until it gets really bad), and people in between us will decide based on their personal cost/benefit calculations. If you close the gyms, she can't go even if it would be worth it to her, I and anyone like me weren't going anyway so can't benefit, so the only people you're forcing to stay home are the ones it's particularly high cost/low benefit for. (The guy who had it last month, and is naturally immune. The lady controlling her high cholesterol/depression/etc. with exercise. And so on.)
You can completely argue that that's still worth it - that A) those people don't have the incentive to count the cost to everyone else of them getting it appropriately, or B) humans are not actually that rational, and in either case you know better than them. But voluntary quarantining does use people's knowledge of their own cost/benefit, and mandatory lockdowns don't.
It isn't *everybody* who is staying home under Voluntary Quarantine, and it isn't a random 80% or whatever. It's the 80% who are at most risk of COVID *and* suffer the least from being isolated.
For an obvious non-COVID example, I'd estimate that 80% of the US population attends some sort of Christian religious service at least once a year, and does not visit a mosque, synagogue, temple, or whatever. But that voluntary state of affairs is vastly different from "All Americans are required to attend a Christian church at least once a year, and public practice of all other religions is forbidden".
For the COVID example, assume that in some nation 80% of the population lives in healthy multi-person households, 10% lives alone, and 10% lives in abusive households. In a sufficiently severe pandemic, 80% of the population stays home - and 20% *doesn't*, because both solitary confinement and being inescapably trapped in an abusive household both have huge psychological costs. And many of those people will be young and at approximately zero personal risk of dying from COVID.
If the government implements a strict home lockdown, as was apparently done in Spain and Wuhan and mandated but not enforced in California, you don't get to say "well, 80% of the population would have done that anyway, so the costs of the strict lockdown are only 25% higher than the costs of doing nothing and there's not really difference".
Also, some people still hold to the quaint notion that individual liberty is a thing of intrinsic value even if they don't choose to exercise every possible liberty at every possible moment.
I see great value in separating the voluntary from the mandatory, and I see the people who can't see the difference as a far greater long-term risk than COVID-19 ever was.
Staying at home is kind of a bad example, since it is quite unlikely that in jurisdictions where stay at home orders were really enforced (like presumably infamous French nightime curfew) people would not go out more without them. Also stay-at-home orders do not seem to be very effective at slowing covid covid when everything is already closed.
But I agree with your larger point that infringement on freedom caused by fear of covid is just as bad for people's quality of life as infringement on freedom caused by equally strict government restrictions. Relevant metric seems to me how much people alter their behavior because of the pandemic, and it matters little whether this change is caused by fear of infection or fear of getting punished by the government for violating a lockdown order.
I. e. temporary closing of pubs (bars for Americans), if it decreases covid spread, also decreases levels of VQ caused by a fear of covid, which is related (among other things) to levels of covid in the community, and thus it might in the end lead to smaller overall behavioral change caused by covid.
> since it is quite unlikely that in jurisdictions where stay at home orders were really enforced (like presumably infamous French nightime curfew)
Small data point: I live in a big city in France (Lyon), and I've never seen a cop during the lockdowns. My bother was controlled by one, but that's it. That's for a family of 5 people.
Same thing for the curfews, but since everything was closed, there weren't many incentives to go out at night (unless you're doing house parties). However, everyone I've talked to mentioned that the curfews were really exhausting (the first ones started at 6 pm!).
I've been making this point elsewhere, but note that the Swedish data clearly shows that the behavior of Swedes was radically different from the behavior of everyone else. Their deaths curve is broadly similar in shape to every other deaths curve. But their "new cases" curve looks nothing like anyone else's.
We are forced to conclude that either (1) the difference in official policy made a massive difference in overall behavior; or (2) a more restrictive official policy would have made no difference, because the Swedes are the most fractious, official-recommendation-resistant people in Europe, completely different in kind from any other Europeans.
I would suggest that official policy shows up not just in enforcement efforts, but also in the level of popular fear, and the channel from policy -> fear explains why restrictive policies saw high levels of compliance despite low levels of enforcement.
Honestly, if you are referering to spring 2020, I would be very cautious to inferanything about behavior from case numbers, since those are heavily contaminated by differences in testing regimes. Mobility data are much better indicator of behavioral changes.
If anything, once a lot of people are in VQ, ML may well have economic *benefits* as opposed to costs. If a small fraction of shoppers are out shopping, then businesses have to make a difficult decision to figure out whether the marginal cost of sending employees to go in to work will is more or less than the expected returns from selling to that small number of shoppers. (It's surely not enough to cover rent and other fixed costs.) Whereas if you move to a mandatory lockdown, no business will lose extra money in this way, and employees will be able to claim unemployment benefits.
I'm not seeing how this argument doesn't fully generalize to "denying people the ability to make their own choices may have benefits because otherwise they have to make their own choices and they might make bad choices".
When you put it as "may", it absolutely does generalize. In pure decision theory, when it's a person making a decision against an impersonal unknown of the state of the world, there's never advantage for a rational agent to have fewer options. But in game theory, when it's a person making a decision against another person who is making a decision with knowledge of the first person's options, there often can be an advantage to have fewer options. The classic example is that if two people are playing chicken, and both start out with the options "swerve" and "full speed ahead", if one player has the option "swerve" removed (and this is common knowledge), then that player ends up with a much better outcome.
Of course, it depends on *which* options are removed, and what the game scenario is. But this is a standard economic point - naively one thinks that more options are always better, but in situations involving interaction, this just isn't true.
> The classic example is that if two people are playing chicken, and both start out with the options "swerve" and "full speed ahead", if one player has the option "swerve" removed (and this is common knowledge), then that player ends up with a much better outcome.
Except that the obvious implication of this approach is that both players should remove their steering wheel before the match.
I'm not sure why this is an "except". This just shows that we can't simplistically assume that if pruning options for one person improves that person's situation, then pruning everyone's options the same way will improve everyone's situation. There are cases in which it can, say if we're playing a pure coordination game. I didn't mean this to show precisely what our policy should be - I just meant it as a counterexample to the proposed principle that was being used to argue that no limitation of options is ever good.
> If everyone is already staying home what are the negative effects of ML we should be concerned about?
I dunno, maybe blatant violation of our freedom to associate? As but one example, maybe people who live alone shouldn't have to suffer in isolation and should be able to decide a small group of people with whom they can associate without the government fining them for violating the mandatory lockdown.
> people who live alone shouldn't have to suffer in isolation and should be able to decide a small group of people with whom they can associate without the government fining them for violating the mandatory lockdown.
I would be surprised if there is anyone in this discussion who doesn't agree with this claim. I don't believe anyone here is advocating a mandatory lockdown that bans people who live alone from forming pods.
Probably not, but this is a property of mandatory lockdowns nonetheless, and some governments did try to enforce it (we did here in Canada). Official guidelines were that if you lived alone, you could have contact with only one other household.
Meanwhile, here in Washington State, the official guidelines were that people who lived alone couldn't have contact with any other household. Solitary confinement is understood to be torture when imposed on people in prison, so...
Sure, we can debate what sort of theoretical lockdowns might've been a good thing. But in that case, let's be sure to also loudly proclaim that the actual lockdowns as practiced were a horrible thing. In other words, please don't shrug it away when people point out the existing systems inexcusably violated people's rights to no good purpose.
As long as you don't shrug it away when people *also* point out that the existing systems inexcusably encouraged people to spread virus to no good purpose - particularly all the politicians who didn't stop at banning mask mandates, but also made all their public appearances mask-free in order to discourage people from wearing masks.
>I don't believe anyone here is advocating a mandatory lockdown that bans people who live alone from forming pods.
If you advocate lockdowns at all, then you are advocating that - because that's what people with actual legislative and regulatory power are actually going to implement, actually *did* implement, under the name "lockdown", and your advocacy of something you call "lockdown" will be counted as a vote for what they are going to actually do.
Advocating clever nerdy-technocratic policies that are never going to be implemented in the real world is sometimes fun and informative, sometimes a waste of time, but it's dangerously misleading when the thing you're advocating shares a name but not a reality with a thing other people are actually doing.
This is why no one should ever use the word "lockdown". I don't know how many governments used that word. Instead they advocated specific policies. I believe that some European governments banned people from visiting each other, but I don't believe that any US government ever banned individuals from entering the homes of other individuals.
We should be debating specific policies that specific governments introduced, not putting up a strawman of everything that any government ever did against a strawman of doing nothing.
"Most of the debate about whether lockdowns work centers on ideas within the Overton Windows of western countries, after the pandemic had started spreading - ie, given whatever level of lockdown your country had, is the marginal effect of more (or less) lockdown positive or negative?"
I feel this is missing the point. One need not appoint a president for life to implement lockdowns strict enough to actually work, and state of emergency laws are on the books in many Western countries that allow governments to implement lockdowns without violating various constitutions.
I would far rather spend a month or two in a China-style lockdown than a year plus in a Western one (or worse, if the Delta or some other variant spreads enough.) This strikes me as an area where Western governments had an absolutely massive policy failure with millions dead and their Overton window still isn't expanding wide enough to include "actually solving the problem".
It seems that a dictator is neither necessary nor sufficient, as most of actual dictatorships also failed to curb Covid. The real issue is that efficient government is really difficult, especially when a system accustomed to normalcy is suddenly faced with a novel crisis.
Interesting that, on naive estimates at least, the dollar-life or dollar-QALY trade-off rate of Covid restrictions is within a factor of 2 of other such tradeoffs determined by government regulation. I wonder how much of the lockdown debate is just people projecting their political values onto a novel factual question-- i.e. you probably think COVID restrictions were a bad deal if-and-only-if you think EPA restrictions are too.
Editing note…I had to go back to see what a QALY was. It might be good to follow the original mention with the acronym, I would have seen it more quickly, maybe others too. It made it difficult for me to follow the conclusion section at first.
My apologies, this is my first comment but a few 'What if's' come to mind. What if 85% of the people under 50 had the SARS-Cov-2 virus infection by Nov. 2020..and it was so mild they never noticed? What if the PCR test is the wrong test , has massive false positives, picks random protein and old viruses, is dialed up to a 45 'magnification' instead of 28 as recommended? What if the virus is 1 micron in diameter and the average mask (sieve) is 500 microns? What if the vapor travels not 6 feet but 100 feet? What if staying indoors is exactly against all the viral transmission rules?
What if the 'lockdown success' in the first wave is actually attributable to Sunshine and Fresh Air (aka Spring)?
What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?
What if .. for every 1% increase in unemployment 37,000 Americans die (as one prepandemic study showed)?
What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatoni, Iodine and dozens more EACH reduce the Absolute Risk(AR)
of hospitalization and death by 65-94% ? What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%? What if the Inventor if mRNA technology Dr. Robert Malone warned the FDA about the toxicity of the spike proteins (generated by the vaccine) a year ago and Dr. Byram Bridle found evidence in Pfizers Japan data and exModerna Dr. Luigi Warren says vaxxed people shed spike proteins and all 3 have been variously trashed, deleted, deplatformed and ghosted? What if mRNA Inventor Dr. Robert Malone
and his 16 patents have been scrubbed from Wikipedia? What if this unprecedented Orwellian fact of this pervasuve laundering is because Google and Jeff Skoll invested in GoF research starting in 2010? What if up is down and down is up, Alice - here in Wonderland? Scrubadubdub.
Friedrich Nietzsche said ,
"The strength of a person's spirit would then be measured
by how much 'truth' he could tolerate, or more precisely, to what extent he needs to have it diluted, disguised, sweetened, muted, falsified"
"What if 85% of the people under 50 had the SARS-Cov-2 virus infection by Nov. 2020..and it was so mild they never noticed?"
We know this isn't true because we can test people for antibodies. In the US these studies showed about 10% of people had had COVID in summer 2020, and about 25-30% of people have had it now. Antibody studies from other countries show concordant results.
"What if the 'lockdown success' in the first wave is actually attributable to Sunshine and Fresh Air (aka Spring)?"
This is what I tried to control for by comparing stricter-lockdown states and countries to weaker-lockdown states and countries. Both had spring at the same time. This is also what eg Brauner et al tried to control for by comparing countries that instituted restrictions at different times. I recommend you reread at least Evidence Section 1, maybe the whole piece, it's surprising that you missed this.
"What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?"
All vaccine studies have compared vaccinated to unvaccinated individuals, all of these people are experiencing spring at the same time.
"What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatoni, Iodine and dozens more EACH reduce the Absolute Risk(AR)"
See my piece here https://astralcodexten.substack.com/p/covidvitamin-d-much-more-than-you arguing Vitamin D is probably not very effective. I think the same is true of the other things you mention except corticosteroids, which have too many side effects to be given as a universal preventative.
"What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%"
This is true of everything. The decrease in absolute risk of death from lung cancer from not smoking is much less than 1%, not because smoking doesn't cause lung cancer, but because absolute risk statistics are misleading except in the context of knowing the preliminary absolute risk and thinking about it very clearly. I think this is a pretty serious mistake and I am skeptical you understand absolute risk.
I haven't heard of the specific doctors you mention, but I don't think they're relevant to lockdown statistics.
Spring in the north and spring in the south is different. It is better to compare spring in the northeast to winter in the south in many ways. What time of year are people driven indoors where spread is more likely due to being clustered in a climate controlled environment and vice versa? I find that this blunt attempt to normalize by season to be better than nothing, but pretty close to nothing if you don't understand the underlying mechanics that affect the transmission rate. I don't think they understand these affects enough that this normalization works very well. The regional timing of the outbreaks before there was country wide community dispersion is also very difficult to factor out.
The antibody test strains the meaning of what having Covid means. If a person's innate immune system was sufficient to prevent illness their antibody test will be just as negative as someone who never had the virus in their system.
What makes you say that? As far as I know, the adaptive immune system starts spooling up on the same signal that kicks off the innate immune system, and there are multiple crosslinks between the two. Even if every virus-infected cell gets slaughtered by NK cells within 24 hours, they will have produced cytokines to recruit macrophages and those will undoubtably eat at least a few infected cells and start presenting antigens, and then off we go to the B cell antibody factory. No?
I don't see how that would happen. If an innate response was sufficient the first time to shut down the infection before antibody production got very far, why would it not work just as well (or better) the second time?
I don't know how probable it is that someone could drive off an infection so fast through an innate response that undetectable levels of antibodies result. That seems to me a priori like a delicate question of whether there is any nontrivial delay between innate attack and antibody production kick-off, and whether there's a mechanism for shutting down antibody production early, and if so when it happens, and whether whatever level is produced is clinically detectable. I don't know nearly enough immunology to even know whether the answers to these questions are known.
But I'm just kind of a priori dubious that the innate immune system could entirely handle an infection without triggering adaptive components *at all*. The two systems are deeply intertwined, and anyway that doesn't seem very efficient and you'd think evolution would've knocked away a barrier to maximum efficiency like that long ago, back when we were coelenterates.
The mechanism that regulates the production of b-cells that will produce antibodies seems to depend on prolonged presence of the antigen. I-10 - I-13 describe the control mechanism.
I don't think it's spelled out 100%, but reading the first few sections it sure seems like successful innate immune response does not trigger the specific adaptive response that would lead to a positive antibody test.
Mostly agreed, but you're giving ivermectin too short a shrift. Check out the difference in the most recent COVID peak between Indian provinces that used ivermectin vs those that used remdesivir - it's almost an RCT, and the results are astounding.
Plus the guy who pioneered the use of corticosteroids in the treatment of COVID is all in on it, which isn't much but also isn't nothing.
Here's the thing that really bothers me about ivermectin:
The official recommendation is don't take ivermectin, instead... do absolutely ******* nothing. I have seen many a reasonable argument as to why ivermectin may not be all that useful in treating Covid. But damned if there's any indication trying it is somehow harmful. So it's not harmful, it may or may not be helpful. The official recommendation is do not use this. There's no making sense of that without resorting to some dark motivations or disgusting stupidity.
There are no indications that trying Lambert's Extra-Virgin Snake Oil* is somehow harmful. It sickens me that the medical establishment is just doing nothing.
The cool thing about Lambert's Extra-Virgin Snake Oil is that it's 40 year safety record justifies it being sold over the counter in a dozen countries. It also single handedly cured River Blindness and its inventor won a Nobel Prize. https://www.nobelprize.org/prizes/medicine/2015/press-release/
You say - "We know this isn't true because we can test people for antibodies". Antibodies are gone in a matter of weeks. They're kind of a useless measure. Much better T-cell, B-cell - memory cells in bone marrow. Cleveland Clinic and UBC found evidence of almost universal natural antibodies. UBC measured 8,000 proteins in their assay (I'm no virologist but looking for one single spike protein antibody vs 8,000 markers?). Natural immunity is better, broader and dramatically more protective than a narrow spike protein focused immune trigger. Example: SARS-Cov-1 which is what this one was 'made' from, protects against SARS-Cov-1 18 years later. It also works on SARS-CoV-2 and all the Variants and yet it's 20% different genetically than our new coronavirus while the scary variants are just .03% different.
You kinda skittered past these four "What If's"... "What if the PCR test is the wrong test , has massive false positives, picks random protein and old viruses, is dialed up to a 45 'magnification' instead of 28 as recommended? What if the virus is 1 micron in diameter and the average mask (sieve) is 500 microns? What if the vapor travels not 6 feet but 100 feet? What if staying indoors is exactly against all the viral transmission rules?" The reason you would do that is they negate the entire discussion. I'll skip over them too for the sake of brevity (not to concede any one of one iota).
And to counter your prior Vitamin D article, you didn't mention Vitamin D is related to 2,000 genes 300 of which are immune system related. SARS-Cov-2 is a wimpy virus just like HIV...unless it gets past your immune system and they wreaks health havoc.
Vitamin D deficiency increases Hypertension
Vitamin D deficiency increases Diabetes
Vitamin D deficiency concurrent with Obesity
Vitamin D deficiency occurs with Heart Disease-Stroke
Off on a vitamin D tangent. It's just these idiots have been causing the Vitamin D deficiency with scary health public service announcements for decades. Because of the tangent I gotta stop here but (see link later... not see Linkletter, that would be Art Linkletter and his show "Kids Say the Darndest Things" prolly before you born kid..in the 60s when we were building great immune systems under the sun)
I said -
"What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?"
osaid -
"All vaccine studies have compared vaccinated to unvaccinated individuals, all of these people are experiencing spring at the same time".
By that I mean you completely missed what I said which was the 24 drumbeat from the Miracle Vaccine Train networks is "Look! The Vaccines are Working!" They tout the plummeting infections to the vaccine. Rubbish. People going outside .. getting away from the viruses trapped in close quarters with people and getting fresh air (studies show zippo worries). 'Can you say Black Lives Matter'? 26 million Americans mostly mask-less AND in close proximity... Why didn't Everyone in America die? WTheck ? (I toned that down). We had the most important lessons of the pandemic to learn and we learned none of them?
I said -
"What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatonin, Iodine and dozens more EACH reduce the Absolute Risk(AR)"
You poo-pooed it and took the Pharma talking points verbatim on ... forget it.
Read up - there are many dozens of cheap and effective solutions.
WHO is trying to make Ivermectin "off limits" because they are concerned about parasitic infections. This, a drug available OTC everywhere costing .003 cent a pill with a 40 yr track record - zero deaths and won the Nobel Prize for it's inventor.
"What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%"
You said -
"This is true of everything. The decrease in absolute risk of death from lung cancer from not smoking is much less than 1%, not because smoking doesn't cause lung cancer, but because absolute risk statistics are misleading except in the context of knowing the preliminary absolute risk and thinking about it very clearly."
This is gobbledygook, start to finish. You compare Smokers with Smokers - a perfectly imperfect analogy.
We are comparing an experimental gene therapy to a placebo. People who smoke cigarettes are up to 3,000%* more likely to die from lung cancer than people who do not smoke. (* by that's just the CDC and we know they have licensed a lung cancer vaccine as 45% of their revenue comes from vaccines. Ironically, LESS likely to die from COVID-19 if you are a smoker and 45% less likely to get Parkinson's btw)
Just illustrative -
Relative versus absolute risk of comorbidities in patients with psoriasis
Conclusions: Presenting attributable risk in the form of the number needed to harm provides a clearer picture of the magnitude of risk and a basis for wiser medical decision making and patient education.
(Right, as everyone would agree)
I admit to not communicating Relative versus Absolute Risk well at all.
And also of making the mistake of thinking You Knew This - and you clearly did not.
In fact, you double down on the very problem. Whew.
So I hand the ball off to ...
Peter Doshi of the BMJ board he writes -
"Only full transparency and rigorous scrutiny of the data will allow for informed decision making, argues Peter Doshi
In the United States, all eyes are on Pfizer and Moderna. The topline efficacy results from their experimental covid-19 vaccine trials are astounding at first glance. Pfizer says it recorded 170 covid-19 cases (in 44,000 volunteers), with a remarkable split: 162 in the placebo group versus 8 in the vaccine group. Meanwhile Moderna says 95 of 30,000 volunteers in its ongoing trial got covid-19: 90 on placebo versus 5 receiving the vaccine, leading both companies to claim around 95% efficacy.
Let’s put this in perspective. First, a relative risk reduction is being reported, not absolute risk reduction, which appears to be less than 1%. Second, these results refer to the trials’ primary endpoint of covid-19 of essentially any severity, and importantly not the vaccine’s ability to save lives, nor the ability to prevent infection, nor the efficacy in important subgroups (e.g. frail elderly). Those still remain unknown. Third, these results reflect a time point relatively soon after vaccination, and we know nothing about vaccine performance at 3, 6, or 12 months, so cannot compare these efficacy numbers against other vaccines like influenza vaccines (which are judged over a season). Fourth, children, adolescents, and immunocompromised individuals were largely excluded from the trials, so we still lack any data on these important populations."
Peter Doshi gets complicated, what he's saying is 'It's a load of crap".
And for my part, I'm seeking a better way to convey this horrible statistical con.
And really I have to say to you "I think this is a pretty serious mistake and I am skeptical you understand absolute risk".
You say -
I haven't heard of the specific doctors you mention, but I don't think they're relevant to lockdown statistics.
I say -
They couldn't possibly be more relevant to Lockdowns.
If you are using faulty PCR tests, faulty infection rates, skewed hospitalization and death rates, AR\RR, social distances, vapor distances, efficacy of mask and on and on - you would never know and yet one can weave a story with faux precision.
These doctors are the glaring evidence that you and the people are not getting the full picture from CNN (who's ad revenue is 44% Pharma) of the NYTimes (owned Blackrock-Vanguard own 16,000 companies including Pharma)a nd FAANG and really everyone with a life & death vested interest. It's historic, unprecedented and permanent. Lockdowns are lunacy.
I though for sure someone, somewhere would see this Wikipedia wipe as the new Orwellian age we are in. The relentless layering of lies means the Truth we seek with aid of science and reason can never achieved. (So, I guess this is my last post. I envisioned this as something an outpost of reasonable truth-seekers) Take care and peace.
I live in New Zealand, and I feel this post is missing something without talking about us (or other countries that have pursued similar "Zero Covid" policies, like Australia).
With strict border controls and a strict lockdown (and probably significant assistance from our geography), we managed to eliminate Covid-19 in June 2020 and have mostly kept it eliminated since, with mandatory isolation of all arrivals, and a few short lockdowns when Covid managed to leak into the community. Australia's biggest wave (Victoria mid-2020) was eliminated after months of harsh lockdowns, and they've since had similar policies and outcomes. Our economies have also held up fairly well, helped by the ability to maintain minimal internal restrictions whenever Covid had been elmiminated.
I think there are good reasons (mainly geography) to think our experiences can't be just copy-and-pasted to the rest of the world. But I think the lesson specific to lockdowns is that there's a discontinuity when elimination is a plausible option. If e.g. Colorado had decided it was going to institute a seven-week-long harsh lockdown to try to eliminate Covid-19, it'd be at a great cost, but because Covid would be re-imported from another state so easily, the effect on Covid-19 deaths might not be that larger over the course of the pandemic. But in NZ's case, we were in a position where we could eliminate Covid-19 and remain that way.
That also changes the calculus on short "circuit-breaker lockdowns" whenever cases emerge in the community. NZ or Australian states locking down for 1 or 2 cases in a city may seem crazy to outsiders (and I do think there have a few overkill examples of this (looking at you, Western Australia)). But these can in fact have some of the highest benefit-to-cost ratios of all lockdowns. The reason is that they can be the difference between returning to a low restriction elimination state, or a phase transition to an endemic Covid state.
As such, I have a theory that optimal lockdown strategy may be bimodal. Lockdowns make little sense with zero cases, can be very good with a few cases (and strong border controls), may not make as much sense with a moderate amount of Covid, but may again start to make more sense with a healthcare-system threatening wave.
I feel like I covered this in the first paragraph of the second Preliminary Theoretical Issues: "If, the moment COVID had been reported in Wuhan, other countries had closed their borders tightly, that would have prevented the pandemic (at least for a while). In that sense, lockdowns definitely could have worked."
I don't quite agree with this new segment. Our policies weren't significant outliers in the world (i.e. NZ entered lockdown at the same time as the UK did). We did benefit a lot from having much less early undetected spread (probably helped by geographic isolation, and maybe the season). Other countries (like Peru) that tried this strategy failed badly.
Test-and-trace has helped but is more applicable to the e.g. South Korean containment model. The main strategy in both Australia and NZ is to keep Covid out entirely via border controls, and whenever that fails, often locking down as well as using contact tracing.
Have you read about the second wave in Victoria, Australia? That got pretty out of hand, but was still squashed when the government gave up on the softly-softly approach and locked down again
I'm going to have to agree with Honey here, that the second segment doesn't quite cover it, for the reasons they mentioned.
I would also like to bring a new point about New Zealand that I haven't seen raised when people talk about it in these parts.
Our wage subsidy during lockdown (although it obviously had some issues), was on the whole, very well done. I get the sense that people had to worry about the tradeoff between being able to buy food, make rent, ect in a lot of other places, if there was a lockdown, or they wished to remain indoors voluntarily.
I haven't really seen this addressed in a serious way, but surely this is one of the major factors, in compliance, voluntarily or otherwise.
I don't think your comment is correct. Melbourne had a decent amount of cases as late as July 2020. We eliminated with strict border controls, hard lockdowns and tbh very little test and trace.
East Asian countries should also be looked at, and treated separately from Australia and NZ. I don't think they have had as strict entry restrictions as AU/NZ, and they didn't eliminate COVID as successfully, nevertheless they maintained much lower rates than Western countries.
That made it sound as it if were a hypothetical scenario that never happened. And it wasn't even what happened in NZ/Australia. We did shut off to China by early February, and the rest of the world by mid-March, but it wasn't enough to stop enough cases being imported that would have likely lead to a large wave absent lockdowns.
And as for the second paragraph in that section: our lockdowns, while strict, weren't the "get shot for going outside" hypothetical you described. And even then, a lot of countries did try something really strict and still failed quite badly. (I like to point out Peru, which instituted a very hard and early lockdown, and still has one of the worst Covid death rates in the world). Lockdown-to-eliminate requires a number of other features, like sufficiently high compliance, financial support, and the ability to control new imported cases.
This assumes no aerosol transmission via atmospheric convection, which is a theory that has been taken seriously and investigated in pre-COVID times to explain seasonal influenza (look up "aerobiology").
Although I admire the effort that went into this analysis, all such analyses seem t over-estimate the robustness of both data and theory. Simple applications of germ theory give us epidemiological models which are to a first approximation always wrong by miles. Conclusion: germ theory is incomplete. Thus arguments like "lockdowns obviously work because theory says they must" are over-reaching. Will Jones has been doing some great research over at lockdownsceptics.org where he explores possibilities for what's going wrong there, and proposing some very logical explanations for the sudden start/stop nature of COVID outbreaks based on aerosol load and heterogenous susceptibility. No surprise that this work is being done by people outside the epidemiological establishment when you look at the Flaxman study
BTW re: Flaxman, the last-intervention-works issue wasn't a bug, the model was specifically designed that way. If you want to see buggy models look at Flaxman's colleague Ferguson and his COVID-Sim model - written in C! - and filled with memory corruption errors, race conditions and other genuine bugs. Flaxman's work is in some sense more sinister because they constructed a non-buggy model specifically to reach the conclusions they wanted to reach, using many subtle techniques to do so. Nic Lewis has got a good writeup of how they hid their subjective Bayesian priors in data tables on GitHub to hide that Sweden was being assigned a prior "country specific factor" wildly out of line with everywhere else, for instance.
I see this a lot (and have posted about it a lot before on this forum and others), but I think people still don't realize how insanely behind the US New Zealand's COVID was at the time to lockdown. Generally speaking, most countries woke up to COVID around March 12th (Rudy Gobert + Tom Hanks). Assuming a 3 week infection to death period, we want to know deaths as of April 2 to infer infections.
On April 2nd, the US had a total of 8000 deaths, implying almost a million infections, and had 1500 people dying a day (and rising) implying 150k daily infections.
New Zealand had 1 death. New Zealand had 5 confirmed cases on March 12th. So the comparison to New Zealand is basically asking why couldn't the US have locked down 5-6 weeks earlier - that's the only time we had a hope of NZ like outcomes. By March 12th the cat was out of the bag - our outbreak was already worse than it got in any of the Vietnam/Taiwan/South Korea ON the day we locked down.
Now Australia is a bit more of an interesting case, and probably the strongest pro lockdown one out there. But still - they had ~900 people die by the end of their summer wave, with a population of 25 million that means on a per capita basis their entire summer wave resulted in as many cases as the US already had by March 15th. It's just a completely different level of disease.
Re your final paragraph: substitute 'winter' for 'summer', but more importantly, note that that second wave happened in a single state (Victoria, population 6.7 million) and mostly in metropolitan Melbourne (population 5 million). So when confirmed new cases were running at several hundred per day (peaking above 700 a couple of times), that was a pretty significant outbreak.
This is what I getting for posting twice haha. Covered this in a followup:
"Sorry yeah used northern hemisphere seasons.
Ok, so Melbourne looks to be ~5 million people and NYC is ~ 8 million? But the greater NYC metro area is ~ 20 million, but a lot of those are in non New York States. So I may not exactly be comparing apples to apples but lets go with this.
Victoria has had 820 COVID deaths, let's just assume those are all in Melbourne.
On April 4th *alone* NYC had 819 COVID deaths. By 3/29 they had more deaths in New York (state, but mostly City at that point) than Melbourne did over the entire Pandemic. Even if we assume the 20 million number and 4x population, we hit that on 4/4.
Again, it was just a completely different level of disease burden in the US before lockdowns started. This matters a lot - if we assume a non-China type lockdown can get R down to 0.6, 12 weeks of lockdown turns a million cases into 2100. Which is still a decent amount of cases! R of 0.7 (an estimate I've seen for US lockdowns) puts you at 13,000 cases - very much not an under-control outbreak! If you only have 50k cases, you get down to 110 cases with a 0.6 lockdown, and 700 with a 0.7."
I've long thought there are way too many confounding variables to really be able to draw any strong conclusions about this sort of thing. Just looking at how wildly different the outcomes are in different countries, or in states that have similarly strong/weak lockdowns.
We have heard, for example, that the weather affects transmission. People who argued this accurately predicted changes in case numbers in many cases, most notably the big winter wave in the US and Europe, predicted months in advance. So ... how the hell do you account for that?
Given all that, I am inclined to just evaluate what policies work the best given our understanding of how the virus works and of human behavior. I know we all like this sort of empirical analysis, scientific method and all that, but on some level if it's this hard to measure you just end up measuring your own blind spots. "We can figure this out using empirical analysis" reminds me of the High Modernism thing - if you can't really figure out a good way to approach it scientifically, just kinda do something science-y.
A few other points:
<ul>
<li>An underrated part of lockdown politics is really about the financial assistance to laid off workers. For many people early on the understanding was that lockdowns lead to people losing jobs and so political support for unemployment benefits, etc. And that ending lockdowns would mean an earlier end to those benefits, so whatever people thought of the money stuff drove views on lockdowns.</li>
<li>I wonder how much the <i>public messaging</i> matters. In the US, it wasn't a particularly right-wing thing to be unconcerned about the virus, or left-wing to be concerned, at first. It only started that way after Trump and Fox News went around talking about how it's all overblown and how "this" is "the Democrats' new hoax" and all that. You mentioned how actual behavior is often loosely connected with government rules - it seems that it is more strongly correlated with the partisan messaging.</li>
<li>I think we got locked into a not-very-helpful pro- or anti-lockdown debate early on that prevented any discussion of <i>what are the best lockdown policies</i>. Early on there was discussion of centralized quarantine - people said NYC was hard hit because of overcrowding in small apartments, where a sick person would stay at home and infect everyone else - and widespread testing and tracing. None of which ever happened. At the same time, as far as I can tell "six feet apart" has been known for awhile to not have any basis in fact, and people still follow it. And restrictions on outside activity lasted a long time after it was clear that it was a way lower risk.</li>
<li>I think the medical establishment did a bad job of communicating at many times. Big examples include the initial anti-masking message, and the failure to give more realistic guidelines that people who don't follow the strictest guidelines might follow. To steal a point from someone, all non-COVID CDC advice is <i>also</i> super-strict rules that nobody follows (IIRC they recommend using a dental dam during cunnilingus. And 4 drinks in a day for a woman is "binge drinking").
And some political bullshit like saying it was OK to go to racial justice protests because it's a more important issue than COVID.</li>
<li>That said I'd be remiss if I didn't add that the medical establishment's bad job pales in comparison to the aggressively terrible job from the administration. People forget it now but before Fox News was against vaccines or lockdowns or masks, Trump was even against testing people, on the basis that it would increase the number of cases and make him look bad. He said this out loud on TV repeatedly! The main guiding principles seemed to be the stock market, and a mystical belief that nothing could stop the "Trump train". And people on his Coronavirus "task force" have alleged that they slow-walked the response because it was seen as a blue state problem.
All the arguments about calling it the "China Virus" or whatever came about because they needed new talking points to replace the "it's a media hoax" talking points once it became clear that it was, in fact, a big deal.</li>
<li>I think it's clear that earlier waves cause lockdowns (and voluntary lockdown-ish behavior). People in other parts of the country ask why everyone in NYC wore masks so long - probably because how bad it got in April of 2020.</li>
Lots of really important points here. Especially that first one, where one main point of a lockdown is that it enables economic relief policies, and so may well be part of an economic *positive* rather than a negative. I also have a quibble about the messaging around protests (mainly because protests were outdoor activities that turned out to have overly harsh messaging elsewhere), but otherwise largely think these are the right points to be considering.
Maybe Scott can fix the HTML formatting that didn't work?
I think the point about the protests was that at a time when people were not allowed to have outdoor weddings, funerals, or even protests about other things, the CDC and prominent scientists gave a free pass to a certain subset of protests. There were some anti-lockdown protests that had been heavily criticized by both government and media, but BLM-related protests were often neither criticized nor discouraged. Then Democratic governors (Tom Wolf did this prominently, but there were others) not only allowed the protests, but they actually joined and/or supported them. See this link: https://www.pennlive.com/coronavirus/2020/06/why-did-pa-gov-tom-wolf-join-a-protest-with-thousands-of-people-breaking-the-rules-in-a-yellow-phase-county.html
If the CDC can say that racism is a more pressing concern than COVID, then a lot of people are going to argue that [pet issue they are concerned about] is also more important than COVID. By telling them no, it creates the impression that the CDC is being political and also not scientific in their advice. It was a very substantial misstep and I believe one of the primary causes of a lot of conservatives/the right starting to reject lockdowns and restrictions.
Wait, did the *CDC* ever endorse these protests?! I thought it was just epidemiology Twitter that did that, but I admit I wasn't paying close attention.
(And my recollection of the anti-lockdown protests was that the most hyped ones occurred indoors without masks.)
There were a lot of health officials that did. I thought I remembered the CDC also saying some specifically positive things about the BLM protests, but they did refrain from condemning them. A former CDC director also came out in favor of the protests, which may be why I remember it that way.
The conservative rejection of lockdowns etc predates the George Floyd protests. IMO it really started because conservative pundits followed Trump on the "Coronavirus is no big deal" track off a cliff, and then needed to justify their previous statements about how everything was an overreaction. Remember Richard Epstein saying there would be no more than 500 deaths in the US from COVID or Rush Limbaugh saying it's no worse than the common cold (not the flu, a cold)?
That said I agree that the "protests are OK when it's something I agree with" attitude was a very damaging one, and only seems like a relatively small deal compared to everything else that happened.
There were certainly a lot of conservatives already bothered by the restrictions. The number who were vocal about it increased significantly after many public health officials either endorsed the BLM protests or refrained from condemning them, despite condemning other protests and non-protest gatherings of all kinds.
Protests ended up as an accidental evidence that outdoor spread is basically not a thing, BUT their real and imho substantial costs were that they damaged already low credibility of public health professionals, many of whom endorsed protests on the grounds that "white supremacy is worse than covid", among Republicans.
"Sweden had closed some public schools (other countries had closed all of them), restricted large gatherings (other countries had restricted all gatherings), “recommended” closing businesses and staying at home(other countries had mandated it), and closed public events and public transport. "
Actually public transport in Sweden never closed. At some point in late 2020 it was recommended that commuters wear a mask when using public transport during peak hours (which, anecdotally, not many did).
FWIW assuming there's only one of these I've seen people voice concerns about the stringency index before; accusations that data was retroactively re-classified to make it seem like low-lockdown areas had stronger lockdowns than they really did.
Thanks for the link about Belgium's death rate! I was very curious about that.
I hate to add yet another "theoretical concern" to your list, but shouldn't voluntary changes also effect the economic impact of any policy? If most people stop going out to eat voluntarily, then the associated reduction in R *and* the associated restaurants going out of business have little or nothing to do with the government policy.
For secondary effects, I think you missed a few that are potentially much larger than the ones listed. We had a lot fewer cars on the road, resulting in thousands of additional traffic deaths (yes, more, because the people left were driving faster--see https://www.nsc.org/newsroom/motor-vehicle-deaths-2020-estimated-to-be-highest); on the flip side, air quality may have improved substantially (https://www.cnn.com/2021/03/16/health/world-air-quality-report-intl-hnk-scn/index.html) and air pollution is actually a substantial health hazard, particularly in poor countries. Then there's the culture war and political effects: much-discussed increased politicization of science and policy, Trump would likely have coasted on incumbency effects and 4% unemployment to re-election, arguably growing inequality and the strength of stocks/big companies compared to small businesses and average peoples' salaries.
"We know this is possible in principle - some states tried things like closing parks and trails, which in retrospect probably wasn’t too useful since the virus doesn’t spread well outside."
Indeed, this particular intervention could even have been counterproductive. I saw an argument (probably on the SSC or Motte subreddits) that lockdowns past a certain point *increase* cases, because everyone is jammed together into the few places they can go, like grocery stores. I wasn't sure of that, but shutting down an outdoor space or preventing all but the largest and most densely-packed gatherings is likely to have 0 positive impact, and secondary effects like pushing people indoors or reducing physical activity very likely make it a net negative. I know that some outdoor areas like major national parks were and ski resorts (and still are) overwhelmed and probably couldn't have maintained standard services like ranger support without compromising employee safety, but local parks and most trails should have been kept open
"Maybe this was a dress rehearsal for a much worse pandemic later on, and the most important effect of our choices now will be setting the defaults and expectations for how we respond to that one."
And that's probably the most terrifying thing. There is no indication that the powers-that-be have realized any of their mistakes. The FDA is still going to ban tests early on. The CDC is going to be wrong about risk. The intelligentsia will sneer at the plebs for being scared, then sneer at them for being selfish, without a hint of self-reflection.
Talking about the US and comparing different states. My prior would be that red states would tend to lower the number of COVID deaths in the statistics, compared to blue states. Shouldn't we account for that? Also most of the big high density cities (aka places where virus spread should be the highest) are blue even in the red states. And they could have stricter local policies compared to the state.
On the messing with statistics point, I know it feels like something that might happen, but I haven't heard any convincing story as to *how* it could happen. Does the governor direct county coroners to say something different on death certificates? Do county coroners have a bias in how they fill out those certificates that doesn't show up in other years?
On the localities with stricter policies than the state, there were periods last summer where that was true, but as someone who lives in Texas, I remember much of June 2020 and March 2021 being spent with local authorities going to every judge and court they could to try to force the governor to allow them to enforce policies stricter than the state.
My prior would be that red states and blue states would tend to lower or increase the number of COVID deaths following restrictions being implemented or lifted in a way that made the decision look correct. An example would be that Florida might start lowering the death count after fully opening up, while NY might start lowering the death count after locking down. I have no evidence that any state did anything like this.
That's the sort of thing that they would do, if the elected officials were able to just plain falsify the statistics. But I think that in any contemporary rich country, one can't just completely falsify the statistics - at most, one could record deaths early or late, but one can't either delete deaths entirely (someone will notice the missing record at some point) or create deaths of people that don't exist. So I think one would test this by seeing if there are statistically significant jumps or drops in death rate just before some transition that go in the opposite direction.
Kenny, I think I agree with you. I guess I would clarify by saying that if I knew statistics *could be* manipulated appreciably in one direction, my prior in that instance would be that such manipulations would occur in whatever direction made the manipulators' policies look better, i.e. not just down for Red, up for Blue.
I have no opinion on whether this ever happened, but the more likely scenario is that the criteria for "this is a COVID death" would be manipulated, rather than outright fabrication of the statistics. If a local coroner says that the person died of heart disease, despite their COVID infection, but another local coroner would have ruled that a COVID death, that could play out over different localities and skew the numbers.
Again, I have no idea if this happened. It's been something I have been curious about since early on, as I heard a lot of people make that accusation with limited or no evidence.
How does the local coroner know he's supposed to do this? Because if some high-level state official calls all the local coroners and says "change your criteria for reporting deaths in support of my obvious political motive", many local coroners will call many local and national journalists and say "the State is trying to mess with the statistics".
By state agencies promulgating very boring criteria for what does or does not count as a COVID death. For instance, a policy to check whether the person tested positive for COVID within the last X days and registering that as a COVID death regardless of confounding reasons would come from the state, be pretty boring in terms of policy (it's an easy shorthand that helps coroners process cases consistently), but may in fact skew the data considerably.
There is also a reporting issue. Whether a state decides to categorize a death as "from COVID," "with COVID," or any other moniker, and what criteria it uses to categorize such deaths, doesn't necessarily mean they are trying to deceive the public as to how deadly the disease is--it's just a way they've decided to categorize things. The reporting issue comes in when the nuances of how the statistics are compiled and what they mean are distilled down to "COVID killed X people," even if the facts on the ground are that X people died but COVID caused or was associated with those deaths to varying degrees.
Then we shouldn't be talking about whether this is or is not the "more likely" scenario, we should be talking about the part where you (or, ideally, some reporter who had the same thought you did) published the actual Very Boring Official Criteria that were changed and highlighted the boring but important parts.
The two plots of "cases in Sweden vs Europe" and "deaths in Sweden vs Europe", considered together, seem like very strong support for the arguments "lockdown cannot pass a cost-benefit test" and "the Swedish non-lockdown really wasn't a lockdown".
We see a gargantuan bubble in Swedish cases, unrepresented anywhere else, that doesn't show up in Swedish deaths *at all*. Swedish cases soar to new heights at the same time Swedish deaths maintain a steady downward trend.
That's a huge number of people falling sick for whom falling sick is the right outcome -- they get all the benefits of whatever they were doing with negligible downside -- but who were, in every other country, prevented from achieving this correct outcome.
I think what might be happening there is that the UK had many more deaths than expected, skewing the graph a bit. If you remove the UK, Sweden looks pretty bad. The badness isn't really the same shape, but maybe this is because there's a lot of randomness in how long it takes COVID patients to die?
I did end up judging Sweden by deaths and not cases, so my conclusion should stay okay even if all those cases were an artifact of eg unusually good testing (which I haven't heard claimed)
But I'm saying the cases are important information that doesn't show up in the deaths.
Sweden sees an enormous increase in cases and a steep decline in deaths. That is exactly what you'd expect from a system of only voluntary measures, *if it were working perfectly* -- people who are at risk mitigate that risk, reducing deaths, while people who aren't at risk ignore their non-risk, resulting in a boom in cases. That second group reaps the benefits of whatever activities they were doing that led them to fall sick. But they don't suffer any real costs.
Compare that to any other country -- every other curve in the comparison looks basically the same. Sweden peaks at a high comparative level of deaths and declines steadily. Denmark peaks at a low level and declines steadily. Everyone, Sweden or no, sees a sharp uptick towards the end of the data.
But Denmark's *cases* peak and decline the same way their deaths do. That shouldn't happen -- the Swedish population is much better off, with their practice of infecting those who have nothing to fear from infection, than the Danish population is with their practice of preventing infections in everyone whether they're at risk or not. That is a huge cost that Denmark pays and Sweden doesn't.
Low-risk individuals getting sick is only "good" if it's going to prevent future waves by increasing the population's natural immunity. But Sweden's subsequent waves were worse, not better, than the rest of Europe's. So on a first glance, all those extra infections didn't actually make things better.
Besides, now that we have the vaccines, there's an even lower-risk way for a larger number of individuals to develop immunity.
> Low-risk individuals getting sick is only "good" if it's going to prevent future waves by increasing the population's natural immunity.
No, this is wrong. Low-risk individuals getting sick is good *because it means they were doing things that were valuable to them*. They received a significant benefit and paid an insignificant cost.
Low-risk individuals getting sick can only be considered bad in an all-else-equal sense. All else is not equal; the only way to become sick is by participating in activities that make you sick.
There's a standard example of GDP being a "bad target": if I pay you $100 to dig a ditch, and then you pay me $100 to fill the ditch in, everyone is worse off even as GDP rises by $200.
We measure the benefit of economic activity in terms of GDP anyway, because most transactions aren't like that. We assume that voluntary transactions leave the participants better off, and the expenditures which constitute GDP are an indicator of those gains. They aren't the gains themselves. They're an indicator. When a trade happens, it has two effects: the participants become better off, and GDP increases.
Sweden's increased covid cases are an exactly analogous indicator of benefits obtained by the Swedes.
That very logic cuts both ways, though -- just as we have no guarantee that GDP actually correlates to value, we also have no guarantee that sickness correlates to people doing activities they value and receive this so-called "significant benefit." The same "bad target" example applies: People could be choosing to intentionally get sick.
Only in this case, while we don't have numerous documented examples of ditch-digging-and-refilling, the literature is pretty clear that people are not rational about their choices around preventative health. I would bet that if you could identify the activity that got people sick and gave them the opportunity to reverse their decision to participate in it, many would take that trade.
Was that one dinner with friends really worth getting sick for two weeks? Even if you recovered, that's often going to mean more social opportunities lost than gained. Hardly a "significant benefit" and "insignificant cost," for the types of symptomatic illness we're discussing.
One dinner might not be worth getting sick, but that doesn't matter because I didn't know in advance which single dinner to avoid. The proper comparison is whether going out to dinner with friends every week (and other social events) is worth getting sick once. To me, it is.
Even if it worked in Sweden doesn't mean it would work everywhere. Sweden has unusually few people living with their parents. My grandma lives with my aunt who is a public school teacher; if schools hadn't been closed, there would have been pretty much no possible "voluntary measures" to reliably stop my grandma from getting COVID short of my aunt quitting her job.
I've seen it elsewhere that Sweden, the UK and also New York all made the same mistake in the first wave: they triaged by sending recovering elderly back to care homes, letting them spread the virus amongst the most vulnerable and drastically inflating death numbers.
Unfortunately you can't really measure such a thing but if it's true it would make these three areas very unreliable for comparisons.
Yes I think this is the flaw of the analysis of Sweden. Their deaths were driven by large nursing homes that other Nordic countries don't have. I thought it was strange that it is not mentioned that in the end, Sweden did fairly well relative to Europe, and also the Netherlands was not mentioned which also had lax policies and did well.
I think also that it's hard to generalize from Sweden's case/death ratio to other countries, since it has one of the best healthcare systems of developed country. So other countries would not have been able to deal with the same number of cases as well, so would have higher case/death ratios.
I'm extremely skeptical of the Sweden story. There's a problem with saying, "Look at this cool n=1 experiment!", then turning around and saying, "But the conclusions fall apart halfway through the experiment, so let's agree to ignore the data that don't fit the narrative."
If Swedes were still dying at high rates after they matched the lockdowns of their peers, the conclusions we should draw are obvious: the higher death/infection rates should NOT be attributed to the lockdowns that don't correlate with the claimed effect; we should be looking for some other factor instead. Maybe that factor is genetic, cultural, geographic, or completely unknown. But we should be able to reject the lockdown hypothesis for excess Swedish infections and death based on new data - instead of just ignoring clear evidence because it doesn't fit our hypothesis.
> That's a huge number of people falling sick for whom falling sick is the right outcome -- they get all the benefits of whatever they were doing with negligible downside -- but who were, in every other country, prevented from achieving this correct outcome.
This assumes that they're getting sick because they're correctly unconcerned and doing things they enjoy. Rather than going to work terrified of the virus because their unsympathetic boss said they'll be fired if they don't. I'd argue that the utilities from letting the later stay home (and do nice things like learn to make sourdough) outweigh multiple people's worth of utilities from having a night out.
Please try to keep in mind the millions of people who had no choice at any point during the pandemic about going to work. That includes the nice people involved in getting the sourdough ingredients to your home - farmers, drivers, food processing plants, public services, road maintenance, police, with many more.
I find myself not overly sympathetic to the people who were too scared to go to their job, regardless of whether it was actually dangerous for them or dangerous in general. I have a lot more sympathy for people with medical conditions that would put them in danger specifically. Not so much the 25 year old in perfect health who is freaking out and ordering all their food at home as if that doesn't put other people in the exact same danger they are too scared to deal with.
The 25-year-old would have to eat either way, and I can't see how ordering their food at home puts the involved workers into danger any more than going to grocery stores in person -- if anything, the other way around, given outdoors vs indoors and given much shorter contacts with delivery people than with supermarket cashiers.
The 25-year-old would have to eat either way, and I can't see how ordering their food at home puts the involved workers into danger any more than going to grocery stores in person -- if anything, the other way around, given outdoors vs indoors and given much shorter contacts with delivery people than with supermarket cashiers.
The bubble in reported Swedish cases was caused by a change in testing policy (https://medicalxpress.com/news/2020-05-sweden-ramp-coronavirus.html), bringing it closer to in line with the rest of Nordic countries/Europe. As I recall, there were regional outbreaks around that time (such as one in Norrbotten) but the true daily cases across the country likely were following a steady downward trend they had for a while.
Also missing: Taiwan, South Korea, Japan, Vietnam. (Despite what many people think, S.Korea and Japan citizenry have pretty low confidence in their governments - in Japan, people still remember the government mismanagement of the Fukushima nuclear reactors after the tsunami; for S.Korea: https://twitter.com/BluRoofPolitics/status/1399555956875296769)
Frustrating how often people dismiss those examples with vague stereotypes about Asian culture being more compliant/authoritarian. Without looking at the actual politics of these countries.
Vietnam is currently in an exponential phase. Taiwan had skyrocketing cases a few days ago but now they seem to be collapsing, not sure what the story is there. Japan did pretty well in spite of no apparent differences in policy.
Surely Japan is an example that should be studied more than it is. It seems far more relevant to EU or US than New Zealand, which received tons of media coverage. And of course Japan did various things, they are however poorly understood in the West (including by me) https://en.wikipedia.org/wiki/Timeline_of_the_COVID-19_pandemic_in_Japan
There are two other factors to consider: how open the economy is to the outside world, and how quickly information is collated and disseminated. The UK is very open, and had a government that was until recently very long to close down borders or institute mandatory quarantines, which is why the delta variant is spreading like wildfire. On the data collection front, originally it took them 2 weeks to collect and tabulate data due to an unfit-for-purpose health IT infrastructure, which obviously makes it very difficult to control either the government or private-actor control loop. In one particularly embarrassing incident they discovered that because they used an old version of Excel with a 32K row limit, they were undercounting cases.
I work as an MD in Sweden and have thought a lot about COVID-lockdowns. Many things that governments in other countries made illegal Sweden only recommended i.e. limit social contacts, don't have private parties etc. But the stringency index counts these recommendations as if they were mandatory: https://twitter.com/TTBikeFit/status/1359913436981903361/photo/1 . This makes Sweden seem more stringent than it was, especially during second and third wave. For example according to ourworldindata stringency Sweden had higher stringency than France in January -21, this at a time when it was illegal to be outside in France after 8 or 9pm without valid reason. So in the stringency index it may look like Sweden was a pretty average European country during second and third wave I don't think that was the case. Anecdotally French and German people in Sweden also told me it was much more relaxed here.
Also I did an estimate of the life years lost to COVID deaths in Sweden in 2020 a few months ago here: https://www.reddit.com/r/TheMotte/comments/msqqof/estimating_life_years_lost_due_to_covid_deaths_in/. Sweden had an excess mortality compared to previous years of 7.7 % and a 0.59 year decline in life expectancy. This translates to 63700 non QALY-adjusted life years lost to death or 6.37 years per COVID death or 2.32 days per person.
This is another weakness of the stringency indeces (and attempts describe the COVID policies of dozens of states under a sparse set of categories). Often times, the catgories available won't match precisely with a country's policy. Then, the maintainers of the index either subjectively choose which category is "closest" and put them there, or they write a detailed description of what each category actually means so that they can shoe-horn everyone in somewhere. Then, they summarize those detailed categories with pithy descriptions. This amounts to the same thing, as everyone continues using the short hand.
So in the end, you get a lot of gray area between categories which you hope averages out across the index but doesn't always.
I don't have a better way to do this kind of thing, but its important to recognize how flawed a description of actual policies these stringency scores are.
One complicating factor that doesn’t seem to be covered here is how many QALYs were lost due to things like:
1) Lack of exercise: gyms and parks were closed and group workout sessions were cancelled. Even when facilities were reopened, I didn’t use them because of restricted hours, fear of infection, and unwillingness to work out with a mask on. I’d guess many others didn’t either.
2) Increases in being sedentary: even if people managed to keep up their workout routine, people who worked from home just needed to move around less to do things compared to when they had a physical workplace to go to. Almost everyone I know with a step counter has reported a massive drop in their daily step count.
3) Delayed medical check ups and appointments. How many people will have cancers or other issues detected a year late or treatments delayed by a year? I know that me and many people I know delayed our appointments, figuring the odds are that we’d be fine. But a substantial number of people doing that means that many of us won’t be fine.
It’s possible it will take another year or more before this data is really available, but I’d be interested to know if we have any preliminary results or forecasted estimates on these. And if the strength of those results would overturn any of this post’s current conclusions.
For one counterexample, for me to work from home saves me two hours' train commute per day, allowing me to exercise at home which I otherwise wouldn't. (It also saves 1 h a day of walking from home to the train station and from the train station to the office and back, though.)
"Finally, voluntary behavior change and mandates are hard to separate. If you hear the government is thinking of a mandate, that might make you scared and cause you to do things voluntarily (even things not included in the mandate). Or if the government knows that most people are staying at home, it might feel more comfortable issuing a stay-at-home order to mop up the last few holdouts, whereas if no one had been staying at home it might not be willing to do that."
One factor to consider here and which might further confound is that lockdowns almost certainly don't have effects just in the countries where they're implemented - you can't just treat countries as closed individual units in an open and connected world economy. Of course, an important factor is that if some country is in a bad phase of epidemic, it increases the chances the virus spreads elsewhere - usually, of course, this leads to travel restrictions, but the travel restriction patterns are not always optimal. Here in Finland, a major issue for these weeks has been the spread of virus from Russia to Finland via football fans who were watching the UEFA European Championships in Russia while Russia has been going through a fairly harsh wave, and of course if Russia hadn't been going through that wave, it would have also reduced considerably the chances of spreading to Finland.
However, another thing is that while people will adjust their thinking about how bad the virus is and what precautions to take according to government messaging and actions, government messaging and actions may also have effect in other countries. Throughout the pandemic, the Finnish media has paid a close attention to what other countries have been doing - particularly Sweden, usually presented as a negative example of what not to do. For instance, a bunch of countries enacting lockdowns around Europe has generally been considered a pretty good signal that extra measures will also be coming in Finland, and might lead to voluntary behavior changes even before the actual measures come. Now, the medias - and accordingly, a fair number of people, assuredly only a minority but still probably the minority that acts as the opinion-makers - are paying close attention to UK's policies regarding reopening, and which way they will go, as well as what the real relationship between the Delta variant, vaccines, reopening etc. is vis-a-vis case and death numbers. All such matters may end up having effects on public behavior without the Finnish government necessarily doing anything major.
Of course, this also leads to something of a freerider problem - it's perfectly possible that one of the reasons why Sweden was able to have lenient restrictions at the start of the pandemic without the situation getting really out of hand *was* that the other Nordic countries had stricter restrictions, not only lessening the spread of pandemic from those countries but also potentially contributing to voluntary behavior changes within the Swedish society.
Minor notes:
- "Stringency index" may also be confounding due to the policy of countries being considered according to the strictest subnational unit (https://ourworldindata.org/grapher/covid-stringency-index). At least in Finland, restrictions in the greater Helsinki region have mostly been stricter than in the rest of the country, but the majority of people still live outside of greater Helsinki.
- One of the functions of the lockdowns might not only be that they serve as a government signal for "things bad", but their end might also serve as a government signal that "things good, get on with your lives", potentially allowing for a faster reopening than whatever inchoate signaling the government is able to do *without* having a lockdown, in case the people still choose to voluntarily stay home.
- regarding masks, their spread in society - or people giving up on mask-using if they've gotten used to it - really probably happens in slower waves than what can be measured in, say, three weeks, unless there's a formal mandate with teeth. At least in Finland, where there was next to no mask-using until late Autumn 2020 when the government finally issued a recommendation (not a strong mandate; there has only been a mandate in some public transportation in trains and some cities, and even that is not really enforced too strongly) to use masks, it took *months* for anything close to a 80 % mask compliance to happen in places where it was recommended. (Still, it did happen without formal mandates, eventually, which also tells something about the Nordic society!)
- as the single-person household map shows, Estonia has a high percentage of households, and it still had a big COVID wave a few months ago, having the highest case rate in Europe per capita at one point, if I remember correctly. https://en.wikipedia.org/wiki/COVID-19_pandemic_in_Estonia
Also forgot to mention that one reason why Nordic countries have managed to keep COVID generally in check with differing measures might simply be a robust welfare state - not only in how it generally means a fairly strong health care system (for instance, Finland's success probably depends on ability to genuinely maintain a pretty good test-and-trace system - not a lot of countries have been able to do that, which has also meant they've had to enact stricter lockdowns!) but also that a good welfare state increases people's trust in state and means the state can "get the message through" with recommendations and does not have to resort to formal mandates and punishments.
The former has daily national data, the latter daily US data by state and county. Both updated today.
While I agree with your analysis regarding "quasi-lockdowns" a la US and west Europe, I think you are missing the bigger picture with actual lockdowns a la China, east asia, oceania. I wrote about that on the reddit, but I see some people here covering similar points.
"Similarly, there was a really wide diversity of compliance levels with shelter-in-place orders. I know some people who didn’t see their closest friends for months, and others in the same cities who said “screw this” after a week or two and started having (small) parties again."
It's honestly difficult to say about how effective such measures are. The problems here in Ireland start after restrictions are eased, people come out in large numbers, and then we get a bump in the Covid cases. Part of that is economic, as businesses are pressuring government to relax restrictions (e.g. for Christmas as that is a major revenue generation period) and now for the summer - but we see that when restrictions are lifted, some people can't be sensible and then we get a spike once again in infection rates and then restrictions clamp down again.
Right now, where I live is one of the two areas in the country with the highest rates of Covid-19 (and we are not a large urban area). There was a spike in infections locally in the middle of June, and anecdotally that is alleged to be because in a local village, people who had been attending a match on the weekend went to the pub afterwards, where they mingled with people who were home from England for a funeral.
Given that the Delta variant was established in England at the time - well, there you go. This is also why many people are not happy with Boris Johnson and the relaxation of restrictions in Britain.
Just scanned the comments quickly, sorry if someone else has already pointed this out, but note that the same people who are likely to have high rates of voluntary behavior change are also likely to vote (or email their representatives, or...) for lockdowns. So if you check for the effects of lockdowns assuming nothing else is different between states that did/didn't implement them, you may be missing something.
My impression as a Bay Area Californian from listening to other people talk is that our voluntary measures, especially early on before the state took notice, were unusually extreme, our cooperation with the mask mandate once it did get instituted was unusually high, and the tendency for people to take precautions that weren't officially required (grocery delivery instead of shopping; washing groceries; refraining from taking walks, talking masked and distanced with one friend, or other activities that were explicit exemptions to mandates; etc.) was much more pronounced around here than many places. (Though take that with a grain of salt - the non-California part of the comparison is based on what people talk about, I wasn't actually in any other state until last week, and by that point differing mandates potentially affecting people's behavior had thoroughly confounded everything.) Still, I don't think the set of attitudes that get us the above set of behaviors are entirely unrelated to our state, and moreso our county, hitting the lockdowns early and hard. (At least, hard for a US State - anecdotally, compared to Chile or Spain or China we seem to have been quite mild.)
Maybe Bay Area was even more compliant, but Dallas was a ghost town until a couple months ago, and I don't think I ever saw anyone out without a mask. I tried to go see a movie for the first time since it all started last week and found there is still only theater even open within 10 miles of downtown. I don't think they've been legally required to be closed for nearly a year. They've just chosen not to reopen because there aren't enough people who want to go back to theaters even if they're legally able to.
It really seems to make these kinds of cross-state comparisons seem pointless. What difference can the extent of legal mandates really make if the behavioral changes in any place with enough population density to matter is nearly the same anyway?
It also seems hard to suss out why New York and Los Angeles were hit so hard having anything to do with local policies from the fact they're so much more exposed to international travel than other American cities. Restricting international travel isn't a local decision.
Los Angeles has a big chunk of population who couldn't care less what government tells them to do. Not from some highly academic libertarian philosophy, mind you, but because their family and sometimes personal history south of the border is that government is inherently corrupt and incompetent and only a fool pays attention to what it says if that conflicts with your mother wit, or your actual mother.
An effect not discussed (but which I don't know how you would handle)
In the UK the government said that lockdown was about avoiding hospital overload -- where overload is defined as a tipping point of serious cases per day the exceeds capacity and suddenly the standard of care drops and the mortality rate goes much higher.
I am interested in whether or not this turned out to be a thing? Early on, the consensus was that for serious cases a ventilator was required for recovery and countries were scrambling for ventilators; however, later on it the news told me that ventilators were irrelevant to recovery (that generally if you were put on a ventilator for covid, having it removed would kill you -- there was no recovery as such). So did it turn out that actually hospitals couldn't help you very much anyway if you had serious covid, which might make this capacity-issue not a thing. Just wondering because it's been a very long time since the UK government mentioned this issue -- nowadays they talk about balancing lives lost with keeping the economy running.
On the other hand, if the capacity issue does exist, then this changes the models above.
> So did it turn out that actually hospitals couldn't help you very much anyway if you had serious covid, which might make this capacity-issue not a thing.
There's a degree of covid severity, where intermediate patients are helped by supplemental oxygen but do not need mechanical ventilation. In early May, a shortage of medical oxygen in India was widely reported in the international press
This. If you need a ventilator, more likely than not omae wa mou shindeiru, but it's not just ventilators that there were shortages of in the hospital in northern Italy where my girlfriend works. (There even was a shortage of doctors, as half of them got COVID at the same time in April 2020 and the other half in November.)
As far as I can tell Austria definitely had a problem with its intensive care capacity, particularly Vienna. Intensive care wards are generally at 80-85% capacity in normal times, so a sudden influx of patients with covid can be quite dangerous. It's harder to care for patients with covid since you can't just put them in a normal bed and they need lots of care, but after the past year most nursing stations have been overworked for quite some time and don't have much capacity left.
For some reason the UK government and the NHS seem to have been downplaying how close hospitals were to capacity in January. At least in London, we had
- many hospitals keeping patients on less oxygen than recommended (due to some combination of a shortage of oxygen canisters + pipework in hospitals that couldn't handle such a high flowrate)
- intensive care was essentially full, and many people who would normally have been put on intensive care weren't
- each doctor/nurse had to look after more patients than normal
- most specialist doctors/nurses + wards were reassigned to covid
- all non-emergency surgery was cancelled (so there's now a massive waiting list)
By shutting down most non-covid activity + because of the fact that ventilators aren't as important as expected, hospitals managed to deal with significantly more patients than the maximum "surge capacity" that people were estimating at the beginning of the pandemic, but they couldn't have dealt with much more, and it had (and the coming wave will have) a very real cost to non-covid care.
I suggest the equipment is less an issue (in First World countries) than the burden on trained personnel. If the ER is jammed, there are people in the halls, and ambulances are stacked up 5 deep in the parking lot -- which happened at a few points in LA County -- and the RNs are working 16-hour shifts in moonsuits and the doctors falling asleep on their feet if they stop walking, you don't get first-class care. Mistakes are made. And not just with COVID, this is not the time to be walking into the ER with a curious abdominal pain that could be gas or could be an aortic aneurysm getting ready to blow.
This is obviously not something anyone likes to talk about, because we like to preserve the illusion that MDs and RNs are programmed robots delivering an identical unit of healthcare every time you insert the appropriate token, but for that reason it may be some of the talk about ventilators and such was sort of a cover story for the real truth, which is that some people were trying to ensure the healthcare *people* didn't get overwhelmed.
We also like t preserve the illusion that MDs and RNs can actually cure what ails us, or at least do *some* good. But how true was that of COVID e.g. last March? I read the OP's question as, basically, would "overloading" hospitals at that point have been a megadeath catastrophe, or would it have been the equivalent of "overloading" medieval bloodletters so they couldn't bleed as many people as they thought they ought to, or somewhere in between?
OK, I've stacked the deck by picking ridiculous endpoints; it's somewhere in between. But we need at least a rough estimate, or more precisely we needed a rough estimate last March and I don't recall seeing one.
I think it's always true. In infectious disease merely supportive care does a lot of good. The body is at rest, freed from the stress of activity or temperature variation or what not, it's getting proper nutrition and hydration, secondary complications and infections can be dealt with promptly, supplemental oxygen can be given, et cetera. For most of human history the bulk of what a good hospital did was pretty much just good supportive, i.e. nursing, care, and it definitely worked -- people were more likely to get better if they had excellent nursing care.
So I'm having a hard time seeing any plausible mechanism whereby a decline in the quality of supportive care, caused by professional staff exhaustion, would *not* increase the death rate. I'm perfectly willing to agree in the absence of data that the change might have not been very big, or even in principle negligible, although I would expect an unimportantly small change to accompany a disease that is either much less or much more lethal than COVID, so that the question of the optimality of the general physiological state of the patient was less relevant.
Don't have the link and can't remember how well researched the theory was, but remember reading about positive effects of simply being cared for. IIRC, it was ascribed to an atavistic mechanism of body allowing itself to dial down functions unnecessary for healing, now that somebody else watches for predators etc.
I do think a better pair of questions would be, "what is the optimum behaviour to respond to the pandemic?" and "what is the best way to get people to adopt that behaviour?". "Lockdown" confuses the two of these, and comparing lockdown vs voluntary measures is focussing on the second question at the expense of the first.
The part of your article I'm least convinced about is the attempt to model emotional impact. You summarise this as "52 months of stricter lockdown to save 1 month of healthy life", which does sound harsh, but it could also be phrased as "52 months of stricter lockdown instead of 52 months living through a pandemic with many voluntary limitations on behaviour", which is rather less harsh - albeit still not necessarily worthwhile. Also, some people may find it more emotionally traumatic to be living in a pandemic if the state and their fellow citizens behave in a non-optimal way. Early in the UK, there was a sense of collective pulling together in adversity that strongly mitigated against emotional impact. In other words, emotional impact is hard to measure.
On the other hand, you haven't mentioned the educational impact of disrupting children's education, which seems to be an important downside of the pandemic.
Can't find original link. But Sweden has very high asylum seeker numbers and these are hugely overrepresented in covid cases. Likely this explains quite a chunck of differences with other Scandinavian countries with much stricter immigration policy.
International migrants, in total, are 20 % of Sweden's total population and 16,1 % of Norway's total population, at least. A difference, but not one that can go anywhere near fully explaining the disparity in COVID numbers between countries.
If we look at the number of tourists, Sweden has more of them in total, but Norway has more per capita (Norway has half the population of Sweden, roughly.)
That might explain difference in cases, but I guess age profile of asylum seekers is such that they aren´t usually at high risk of death or hospitalization from covid (?)
I'd love to see "Ivermectin: Much More Than You Wanted To Know," given the absolutely baffling lack of main stream media coverage (and suppression by Facebook, Google, etc) on the topic.
From what I can tell, it's probably better and safer than vaccinating teenagers and under-12 children. It might be capable of actually eradicating COVID-19, if its usage is widespread enough. It might even cure long-haul symptoms, although so far the accounts are anecdotal and there hasn't be a formal study.
I don't see any compelling evidence / arguments to *not* to the claims being made by the Front Line COVID-19 Critical Care Alliance, while there are a lot of people who very obviously have a literal vested interest in a cheap medication like Ivermectin not being the solution to COVID-19.
I respect Bret Weinstein enough to believe that his current take on this topic is probably accurate, but it would help tremendously if Scott could weigh in on it, too.
That should be, "I don't see any compelling evidence / arguments to *not* BELIEVE the claims being made by the Front Line COVID-19 Critical Care Alliance..."
I too would like to see such an article, because I haven't yet had it explained to me how a treatment against various worm infestations and parasites is expected to work against a virus. And by that, I mean something more stringent than "well, it kills it in vitro, we have no idea of what the potential mechanism of action might be, but hey, why not give it a shot?"
If it works does it matter that we don't know the mechanism at first? I thought the sharp drop in India was due to ivermectin. But googling India covid and ivermectin you can find whatever story you want to hear.
"The study included a parasitological analysis of 256 stool samples in low socioeconomic areas of south Chennai with special attention to both intestinal protozoa and helminthes. This study has documented a high prevalence (75.7%) of intestinal parasites in the dwellers from south Chennai, India. Saidapet and Thiruvanmiyur dwellers were found to harbor the maximum number of positive cases."
So I am inclined to think this is rather like the vitamin D studies: take a population with a high incidence of parasitic diseases, dose them with a de-wormer, improvement in not having to fight off the parasite at the same time as you may be developing Covid-19 means better health outcome.
When/if we get studies on populations that don't have worms/parasites and that respond well when given ivermectin, then it will be more plausible that it is doing something specifically against the virus, rather than treating incidental sicknesses that make fighting off Covid-19 worse.
Huh, OK I think there are some studies from Oxford coming soon. Personally I know almost nothing. Did you look at the web pages from the FLCCC? These are doctors treating people for covid.
Don't overlook the more economical explanation of the fluctuations to be expected in small numbers. Any time you study *anything* in a small population, you find all kinds of weird correlations. These vanish when you do studies with sufficient statistical power that the spurious things finally average out. I think what people don't intuitively grasp is how *slowly* this happens. It tends to defy common sense that you really do need trials enrolling thousands and even tens of thousands to pick up on genuine effects that aren't huge, because you need to average away all the statistical junk. Even the pros are caught by surprise sometimes, so hard to avoid is the human tendency to see signal in noise.
There was at one time a thriving cottage industry among innumerate journalists that lay in discovering "cancer clusters" in certain small towns next to a tire plant, electrical substation, or farmer using ivermectin. Apparently it always escaped their notice that you never found such anomalies in big cities with a million inhabitants, even if the cities also had tire plants, electrical power lines, and traces of hormones or plasticizers in the drinking water.
The sharp drop in India is due to herd immunity, similar to that in Peru. By this point, a higher portion of India's population has been infected than any part of the United States, probably due to the spread of the Delta Variant.
Huh, how do you know that? (testing for antibodies is expensive) Ivermectin was used in some Indian provinces/ states. Not related to your comment, but I find it wierd/deja vu that ivermectin gets turned into a wacko conspiracy idea.
Thank you for the link to that article, it seems to chime with my instinctive notions. I may be jumping at an illusory mayfly here, but the fact that the quoted studies were done in Third World/Developing World countries makes me wonder if, in fact, the good results from ivermectin usage were *because* of its anti-parasite qualities and not any anti-viral ones, that is, that patients may have had exposure to worms, lice, scabies and other fun diseases and once ivermectin knocked those on the head, their immune systems were able to concentrate on fighting Covid.
I could be full of hot air on that, though.
Oh, well: if the ivermectin devotees are correct, then every farmer who's been out dosing their cattle for roundworm should have no worries about contracting the virus, if any accidental exposure to the drench has a prophylactic effect! 😁
That article is *very* far from a Scott Alexander MTYWTK post. It basically asserts that the data is weak, gives a cursory glance at a few recent studies, and refers you to the WHO for details. Lowe may be correct in his conclusion, but there's almost nothing there beyond his opinion.
I take your point about total deaths being sensitive to how old the country is, but cases numbers before autumn 2020 are an outright mess. Some countries (UK) had a policy of "Don't get tested, just self-isolate for 14 days" until mid-April.
Now that we have granular data, I like "Deaths 50-69" and also the slightly more complex "Deaths per 100k people in each age bracket". Anyway, if you use these it pretty much confirms that Sweden did much worse than Europe in the first wave, and relatively the same as Europe in subsequent waves.
Yep. When my girlfriend got COVID in April 2020 we were required to stay at home but I wouldn't get tested unless I got symptoms. (At a certain point I even consider lying about having symptoms to get tested, then decided that someone else would probably have needed that test more than me. I only found out I hadn't gotten COVID at all later on, when I got an antibody test out of my own initiative and pocket.)
At the end of Preliminary Theoretical Issue 2 I thought "WTH??? That kind of thinking only makes sense near the Pareto frontier, and almost all western countries very clearly aren't anywhere remotely near it!!!" until the middle Actual Evidence 1.1.
im not sure i buy that US data is as nice as we'd like to be. in the US, red states are fairly strongly clustered in the south, and blue states are very strongly clustered in the north, which means red states got much milder winters, most deaths occurred during winter, and covid is pretty seasonal, so we should expect lower-lockdown states to get some sort of fairly significant advantage, but its not clear how much, which confuses everything
What's driving the seasonality? If it's due to people being indoors more, that applies to both texans in summer hiding from the heat and new yorkers in the winter hiding from the cold.
My impression is that both the deaths from COVID, and the costs of lockdowns tend to primarily fall upon poor people because they have higher rates of comorbidities, in person jobs, less access to healthcare, less savings, don’t own a home as often, etc. While the $150K/QALY figure for a first world country’s government might make sense to the government, since they have access to a tax base of people and companies who make a ton of money, it might be overpriced in this case.
Note that poorer countries tend to price QALY’s lower. My guess is that that’s because if most of your tax base was making $20K a year, paying 7 times the average annual income to keep someone alive for a year might be more costly and impose higher opportunity costs.
In the case of COVID lockdowns and deaths, most economic damages are going to be born directly by the poor as are deaths. So (without redistribution actions) it’s as though the deaths and monetary costs are born by a lower income country, and thus we should discount the cost of a QALY appropriately so we can better judge tradeoffs.
Of course, some countries (like the US) had pretty massive aid packages that helped make it so the whole population helped bear the costs, which would raise the QALY price back up to the national pricing.
Overall my point is that maybe you should be either more for redistribution during the pandemic (to raise QALY price to where it’s worth it) or more against lockdowns (since QALY price might be lower than your normal estimate).
Obvious disclaimer that QALY prices are for assessing tradeoffs and not for valuing the moral worth of a person’s life.
I agree. I feel like the "economic costs are borne unevenly" argument was given short shrift. (As were most counter-lockdown arguments/data.) I don't know the data on this, but in my personal experience, people who had more education/income had much more job security during the pandemic; while people with less of both these did poorly. The costs of the lockdowns were strongly regressive along the income distribution, which is an ethical issue that shouldn't just be dismissed as a single factor among many.
I also noticed that those at the bottom of the income distribution (again in my subjective circles) were the most likely to complain about lockdowns and want them lifted. If the people you're hurting are saying, "don't hurt me anymore", I think you have an obligation to listen to them; and ultimately to seek other solutions.
As for the redistribution aspect, I feel like this is also regressive. I drive electric cars and have solar panels on my roof, so the gas price increases are something I sort of notice when people point them out to me. Meanwhile, my friends - who drive cheaper gas cars - are struggling to find a way to pay to get to work, while watching their dream of EVER buying a house someday disappear entirely.
I've heard the conservative argument that "actually, the top 10% pay a larger share of the taxes collected than the bottom 50%", or whatever, and I think, "Yeah, but that doesn't matter as much as it used to, does it?" Because when the government collects $3.4 trillion per year, but spends $4.8 trillion, the other $1.4 trillion comes from the value of the dollar. Meaning it's regressive to borrow money and then spend it - no matter how well-intentioned the idea. I have assets that grow with inflation, but my lower-to-middle income friends don't. I'm protected from inflation, but they get the shaft. Not sure "we'll fix the problem by letting the government pay for it" is a good idea, when it just means hurting the poor and struggling workers that much more.
You seem to be missing the psychological cost of not locking down. Lots of people are frightened by the lack of lockdowns. This isn't just that they are voluntarily locking themselves down, but the emotional / psychological cost of perceiving everyone else as being a (potential) plague-carrier.
If there is non-economic suffering from an absence of lockdown, then that makes the calculation rather more balanced.
I remember, myself, how utterly *relieving* it was when the first stricter-than-usual measures were announced here in March 2020, after a fair bit of terror of being in an exceptional situation and not knowing what was happening. Finally it felt like someone's in charge and that we're not just going in blind to a pandemic. Of course, that was related to the immediate state of expection, situation was a bit different once people started getting mentally adjusted to living in a pandemic.
UK is ending the last restrictions on July 19, and there's a lot of social media that is clearly people who are terrified and are going into stronger voluntary lockdowns as a result.
Lack of lockdown often makes concerned people overreact, cutting out hospital visits or playground visits or the like, so that even as the lockdown reduces activity for some people, it increases it for others.
I'm not aware of any comparable way that lack of churchgoing makes churchgoers overcompensate in some way.
A risk of death from any cause is (assuming rational people in some sense) accompanied a proportional amount of worry about death. So whatever cost we usually consider reasonable to save a live can really be considered to be the cost we are willing to pay for saving a life *and* reducing the amount we worry about dying. So it's reasonable to compare the cost we are willing to pay to prevent a COVID death to what we are willing to pay to prevent death from other causes. (Indeed, once you die, you don't feel anything, so arguably any cost we pay to reduce the risk of death is actually to reduce our worry about dying.)
But people aren't rational in that sense. In average they are more scared of flying than of driving even though the latter is ~4× more dangerous per passenger-mile than the former, and more scared of terrorism than of lots of other stuff even though the latter is almost always a negligible risk for the average person.
True (at least in general, I'm not sure about the specific example of flying vs driving), but I won't support altering policy to manage people's irrational fears.
"So every 52 months of stricter lockdown in counterfactual Sweden would have saved one month of healthy life. You will have to decide whether you think this is worth it, but it seems pretty harsh to me.... Even if this is true, that just means it’s 21 months of stricter lockdown to save one month of healthy life. Again, seems pretty harsh."
It's very important to note that this isn't the cost of lockdown being compared to life lost, but the emotional cost of an involuntary vs voluntary lockdown. In other words, assuming the wild guesstimate is right, the Swedish locking down as their neighbors did would have been worth it as long as the difference in quality of life between them and their neighbors during the lockdown was no more than 4.8-1.9 %. Now, I didn't live in Sweden during the months of their lockdown, but I highly doubt that they had a great time there, and it's well worth remembering that we were all very far from the Pareto frontier and there weren't many big easy gains to be had. Maybe life was 5% worse or 1% worse in Sweden compared to Denmark during that time, maybe it was even 2% better if life was much scarier and more uncertain for them! But both are swamped by a giant shared hit to quality of life in both countries.
"And even if Sweden had decided to double down and weaken their lockdown despite high case levels, people would probably have voluntarily stayed home more because the pandemic was so bad - inflicting most of the same costs that a state-mandated lockdown would.... This argument seems less convincing in the US, where red states mostly just consistently had weaker lockdowns than blue states did, and never really got stricter to compensate. "
This seems like it was definitely true in Europe in general, and arguably it is less obviously true in the US, but one thing you discount is that maybe the emotional hit got bigger in the non-lockdown states even if measures were never reintroduced. Maybe voluntary behavior change goes higher instead if the red states never got stronger lockdowns to compensate, so the hit is still larger - your evidence is only about legal restrictions, not mobility data.
You say it seems kinda harsh to save one healthy month through 51 months of stricter lockdowns. I think you are overestimating the difference in life quality between softer and stricter lockdowns. For one, voluntary behavior change makes the lived reality in these scenarios not all that different for sensible folks, and for another, a lack of lockdown causes suffering too, both psychologically (feeling like a sacrifice to capitalism in the face of a global crisis) and in terms of being subjected to avoidable risk (having to come into the office where people don't wear masks for no good reason). But I might be biased because introversion causes me to avoid large gatherings anyway, so I barely lose quality of life (and gain some, as remote work and studies are much more to my taste).
Another effect the lockdowns had were in suggesting appropriate voluntary behavior changes. Hearing about possible lockdown measures worked to drive home the idea that these are the things that should be avoided, prohibited or not. Sure, recommendations were made throughout, but a lockdown drives home that really, everyone should do these, it's not one of these things the teacher suggests but the kids ignore.
Indeed - I'm clearly in the minority here, and I feel a bit terrible saying it, but my quality of life has been substantially *better* under lockdown - I get to work from home, hang out with my dog all day, have dinner with my wife every day (even during busy periods at work), have a lot of mental space. I'm very lucky, my conditions are pretty optimal for lockdown life (I have a house right beside a park, no kids, work easily done from home, my family is thousands of miles away anyway, and although I miss in-person hangouts, a history of online friendships from moving around a lot).
Is there any known sensible estimate for a conversion of dollars into QALY, possibly assuming all the money to be subtracted directly from some specific healthcare budget or something like that?
There are broadly two methods of converting dollars into QALYs:
1) The first is by setting a healthcare budget, maximising QALYs within that budget and identifying what the most marginal QALY you can purchase on that schema is. The theory is that any healthcare intervention more marginal than this would displace a less marginal intervention (because you have a fixed budget) and therefore you should only purchase healthcare interventions less marginal than this number. To all intents and purposes, this is therefore the 'price' of a QALY
2) The second is by setting a value we are prepared to pay for a QALY and letting the healthcare budget float until the most marginal intervention we can fund is equivalent to this value.
To the best of my knowledge there isn't any sensible work on the 'true' price of a QALY in definition 2, because it derives from the 'gut feeling' of legislators and leads to a lot of confused and contradictory outputs. The private / insurance-based part of the US health system does something a bit similar to 2 where individuals decide how much their health is worth to them and pay accordingly, but this is confounded by a lot of other factors like individual ability to pay.
However, there has been a serious and systematic effort to identify the price of a QALY under definition 1, most notably in the UK where it is very important for setting spending priorities for the NHS there. This is an excellent summary of the 'state of the art' in estimating the price of a QALY. The methodology is to find the budget decrease that would cause the health system to produce one fewer QALY, and argue that at the margin this must be the price of a QALY. It finds that the value of a QALY is approximately $18,000:
Here's a critique of that approach (although to your point; everyone agrees the number is sensible in an order-of-magnitude kind of way, but data limitations mean it might still be off by quite a lot):
This is a really nice exchange because Claxton et al are famous advocates for low cost-per-QALY thresholds and Barnsley et al are famous advocates for the opposite - you can be pretty confident that anything they agree on is likely to be pretty close to the truth.
Conclusion 2 is doesn't answer the question it's trying to. The headline is "Quantify Emotional Costs Vs. Benefits", but you're not testing the costs of lockdown vs no lockdown. Instead you're testing the marginal benefits and costs of Sweden being stricter like Denmark. I would definitely have liked to see the former, as it is more relevant to the political debates I've seen in the wild; they tend to be "lockdowns are good and we should do them again next pandemic" vs "lockdowns are bad and we shouldn't do them next pandemic".
Regarding the question you did try to answer. I think it's flawed by the lack of a serious attempt to define how much suffering is involved. 51 person-months in stricter lockdown per QAYL saved. If the extra strictness is moving from 100 people per gathering to 10 people per gathering, that seems doable. If it's no school, then it's not worth it.
It's also worth considering positive benefits from the lockdown too. If the cost of a lockdown is that you can't go to a cafe. But the benefit is the waiter isn't forced to choose between bankruptcy and going into work every day terrified you'll catch covid. I would call that a net positive.
This probably applies when comparing the marginal costs/benefits of moving from Sweden to Denmark. But it *definitely* applies when comparing no-lockdown to lockdown. The point you made earlier about "The moral of the story is that everything not forbidden is compulsory, so you can’t always substitute voluntary behavior change for government mandates." absolutely needs to be considered when quantifying emotional costs vs benefits. My working class girlfriend had such rants about how unsympathetic employers around her were when it comes to covid and her daily fear of it reaching the family (they have medical vulnerabilities). Only the lockdown let them isolate, meanwhile my tech company sent everyone home to work in our spacious home offices with gardens outside. Any attempt to measure the emotional costs/benefits of lockdown has to cover the emotional upsides.
> Preliminary Theoretical Issues 6: The Pandemic As A Control System
I think that a complicating factor here is that people don't observe R. They observe (media reports of) cases and deaths.
All other things being equal, a given set of behaviours will result in a particular R. The "pandemic as a control system" theory suggests that R will hover around 1, which implies equivalent behavioural sets under both lockdown and non-lockdown counterfactuals. The potential "win" of lockdowns, therefore, is to induce the "bad covid, therefore panic" set of behaviours earlier than would otherwise happen, so the same R-value (1) can be achieved with a lower overall case and death rate.
This feels like a very complicated problem, in large part because people respond differently to lockdown orders. After a moderate level of restriction, the marginal case might be found in an "essential" business like a logistics warehouse or in a wilfully ignorant member of the population, and imposing further mandatory orders might have little effect. That "resistant margin" may itself vary from time to time and from place to place.
In the scatter plot of stringency on Jan first vs deaths per million, the correlation seems disproportionately driven by six states in the lower right. The rest just look like a blob of noise. Now, %12 of the dataset is obviously too large to be dismissed as an outlier. But which states are they? Is there any particular policy or timing they have in common?
I'm pretty sure those are Washington, Oregon, Vermont, New Hampshire, Maine, and one other - maybe Hawaii? (These are the states that systematically had lower case counts than almost all other states at almost all times from March 2020 to March 2021 - after that time, Washington and Oregon spent a while near the top of case counts, but still lower than most states had in the third wave.)
Have there been attempts to look at rates of all-cause mortality in locked down vs non-locked down areas? I could imagine this going either way. Hospital system collapse would increase all-cause mortality, which might not be accounted for by these studies, and in younger people.
IMO (for what it's worth) the critical issues for assessing the efficacy of NPIs are:
1) How do you control for precipitating conditions? In other words, the severity of the interventions was likely, in most cases, a function of the severity of the ongoing pandemic. The severity of the ongoing pandemic would be associated with the severity outcomes independently of the efficacy of the interventions, to some extent. Thus, reverse engineering from comparing outcomes across contexts, w/o controlling for the precipitating conditions, is of limited value.
2) Evaluating the efficacy of NPIs should necessarily address counterfactual assumptions about what would have happened absent the NPIs. You can't assess the differential effect of NPIs w/o addressing the issue of whether things might have been much, much worse if they weren't implemented - say if the pandemic raged more, whether more people would have stayed away from hospitals. That is the 300 lb. elephant in the "lockdown deaths" arguments.
3) Obviously, controlling for confounding variables and spurious associations when making comparisons across countries with vastly different conditions is VERY problematic. IMO, the data available is insufficient to do this. IMO, better is to look at patterns longitudinally in single countries. And even there, we don't really have a long-enough period to evaluate the effect of the NPIs - for example it may take years to really assess the impact of closing schools.
"Here, having US red states switch to a blue-state typical level of lockdown would save one month of healthy life per 51 person-months in stricter lockdown. Again, seems pretty harsh. Another way of looking at this is that each person who spent a month in slightly stricter lockdown would have saved someone else about 15 hours of healthy life"
Given most states had a period of strict early lockdown and then diverged in the pace of reopening how should I think about the hedonic delta between red-state a blue-state lockdown?
If I can save 15 hours of healthy life by wearing a mask, or going without in-person dining, for a month that's one thing. A month with the near total isolation of the early pandemic is another thing entirely.
It's also difficult to construct the hedonic wedge because of how voluntary the restrictions are. I'm young and healthy and aren't in contact with old people, so once most of my friends figured out that outdoor gatherings were okay my life was pretty normal post summer 2020. My uncle literally did not leave his house for any reason from April 2020 until he was vaccinated, but he's generally paranoid and risk averse. Would either of our experiences have changed under different state laws?
I spent March to December of 2020 in Bryan, TX, and then relocated to Austin, TX, from January to now (and will unfortunately have to go back next month). While I was in Bryan, I sort of felt like I was in a spaceship, where I was isolated from everyone and couldn't do much of anything, though I was able to go for bike rides around the neighborhood. While I've been in Austin, it's felt like things are nearly normal, because there are so many restaurants with outdoor patios and edible food, and outdoor activities to do with friends. (Though I had to develop a habit of bringing a mask with me when I left the house - in Bryan it was irrelevant because there was nowhere with other people that felt relevant to visit.)
Excellent post, which mirrors a lot of my thinking about lockdowns. But I think this:
2: If Sweden had a stronger lockdown more like those of other European countries, it probably could have reduced its death rate by 50-80%, saving 2,500+ lives.
Cannot be squared with this:
3: On a very naïve comparison, US states with stricter lockdowns had about 20% lower death rates than states with weaker ones, and about 0.6% more GDP decline. There are high error bars on both those estimates.
Unless Sweden was extraordinarily more open than even the reddest of states, but it was not:
An alternative explanation is that lockdowns in Sweden impacted behavior significantly more than in the US. (Perhaps US people were more inclined to self-isolate without the lockdown than Swedes, because COVID hit most of the US later, and people were more primed? Or perhaps US people were more likely to circumvent or ignore the mandated lockdown? Either would explain it.)
I think you mean 2020, but agree. If they got hit hard early, when the CFR was higher, that could explain both why Sweden has done poorly and why a better lockdown earlier on would have had a big impact.
I mean this as a third alternative to BM´s parenthesis - let´s say, arguendo, that people in the US and Sweden are equally likely to self-isolate without the lockdown, and they are equally willing to follow lockdown restrictions. But perhaps Swedish pre-covid lifestyle was more different from lockdown conditions than American pre-covid lifestyle.
I think that misstates the assertion, somewhat, which is that if Sweden locked down earlier, its lockdown would have 3-4 times the effectiveness of the lockdowns in the united states. That strikes me as a pretty big lift without any supporting data.
I read this great article about how certain systems have big swings based on small inputs (there is a name for this phenomena, but I can't recall it and did not bookmark the study), and speculates that COVID is like this. I.e., even if you start Sweden and Norway on the same basic course, odds are their experience will still deviate sharply. Think Plinko from the price is right, you drop the token in the middle and sometimes it ends up on the far right and sometimes on the far left, based on undiscernible differences.
It looks like the biggest thing a lockdown does is to "give cover" for things that people want to do anyway.
My employer (in Ontario, Canada) sent everyone home that could work from home on March 17. Was thinking about working from home anyway (I do IT stuff relating to cloud, so it's not like I need to be in a cubicle), but this meant I could just do it instead of having to arrange things.
From what I hear, the "return to office" will be based on how much of one's work needs to be done in a cubicle, so I suspect that I will be in the last cohort to return.
"The Honduran white bat is one of approximately 22 known species of bats that roost within leaf "tents." The Honduran white bat cuts the side veins extending out from the midrib of the large leaves of the Heliconia plant causing them to fold down to form a tent.
...It clings to the roof of its tent in small colonies of 1-15 individuals. The tent protects it from rain and predators. Rather than roosting in a single tent consistently, the Honduran white bat has a network of tents scattered across the forest; it alternates among these tents for roosting."
You don't think they're adorable? I admit, I have had my opinions towards bats changed to the point that now when games etc. have the usual "flurry of bats out of cave" as a jump scare, now I go "sweet flittermice" 😀
But tiny bats looking like balls of cotton wool all cuddled up together in their leaf tent don't make you go "awww?"
I know! Sell your diet pill with these types of cute otter videos and customers will be beating your door down thrusting fistfuls of fivers in your face demanding your product! 😁
One minor quibble about the SF Bay area Mobility Trends graph — people seem to forget that the SF Bay area counties were three weeks ahead of the rest of California in locking down. It's a mistake to say "It’s pretty evident from this graph that people were starting to decrease their mobility before any official government action." Mayor London Breed of San Francisco declared a state of emergency in SF on Feb 25th 2020, three weeks before Governor Newsom made his shelter in place announcement. Likewise, Santa Clara County asked employers to have their employees work from that last week February. And all the bay area counties coordinated their response to the early outbreak. I started tracking and graphing the case numbers between SF Bay area and LA, and the three-week head start that SF Bay area had made a big difference in the overall case loads between the two metro areas.
This applies to other jurisdictions and metrics as well. For instance, some regions of Italy required masks before the national government did. I don't know what date the chart marks as "when Italy mandated masks"; I presume the date when the national government did.
The premise behind lockdowns was not to reduce total infections, but rather to flatten the curve so that the hospitals were not overwhelmed. That appears to have succeeded in most places. If people are not vaccinated, lockdowns simply wont stop the spread of the virus since as soon as people are free, they will move about.
Different levels of lockdown had less of an effect on US states than European countries because there's no practical way to close a state's borders. While Norwegian customs officers could force foreign arrivals to quarantine at a hotel for ten days, there's no infrastructure stop someone from locked-down California from partying in Las Vegas for the weekend.
One potential confounding variable might be hospital availability. In a hypothetical country with an infinite number of hospital beds, we would expect the death toll to be significantly lower -- though obviously not zero; and thus the effect of lockdowns would be attenuated. In a place like NY or CA (to a lesser extent), where hospitals quickly got saturated, the opposite is the case.
That was a common enough belief in the early days of the pandemic, but then we learned that e.g. ventilators mostly just postponed the inevitable rather than saving anyone's life. For more hospital beds to mean *significantly* lower deaths, you need hospitals to be doing a significant amount of good and to a number of patients exceeding their existing supply of beds.
I'm skeptical that this was true in more than a few isolated and temporary cases, but if someone's got numbers, I'd be interested.
It's not just a question of ventilators, because if all your hospital beds are filled by COVID patients, then regular patients who could've been saved from non-COVID illnesses are now much more likely to die. In California specifically, this had in fact happened in winter.
Was that because patients with serious illnesses were being turned away from full-up ERs, or because patients with serious illnesses were afraid to call an ambulance because they didn't want to get The Coronavirus? I only remember hearing about the latter, but that's media-filtered anecdata so highly suspect.
In California and AFAIK New York, it was definitely the former. That is to say, both situations likely took place, but the limiting factor was definitely the number of hospital beds. Stacking patients on gurneys in the hallway can only take you so far.
Thanks for this. In regards to voluntary vs forced quarantine. I really liked the idea put forward by Po-Shen Loh (I heard on Lex Fridman https://lexfridman.com/po-shen-loh/) And explained some here. https://www.novid.org/ (though not a great web site.) The idea is to let you know when people 'near you' are getting infected and that allows you to decide when it's wise for you to voluntarily quarantine. The key idea to me is that this has all the right feedback.
Sorry all I know is from listening to the podcast and then wandering the website. It tracks local infections, and it (the software) knows your possible neighbors by tracking cell phone towers(?) (I've forgotten the details, he (Po-shen Loh) talks about it in the podcast. He said it worked best when linked with local covid data bases, I think to help control people 'spoofing' the system. (by reporting false positives)
OK looking at the lex podcast on youtube they talk about it for almost an hour starting at the 17:00 minute mark.
There are significant "blue" populations in red states (particularly in major metro areas) and vice versa. Maybe "states" are just too broad a brush to paint with?
According to that New Yorker article, you really have to look at communities rather than the states to get a truer picture. If 40% of COVID deaths were nursing home-related, how those communities reacted to the virus would be a compelling variable.
If "blue" counties had higher compliance with lockdown procedures, there is still the variable of high minority populations who worked in industries such as food processing, nursing home care, and other essential industries AND who lived in extended family situations, which may have significantly skewed the "blue" county numbers up.
So, I'd love to see a breakdown of "blue" versus "red" counties. Also, possible to examine COVID-related deaths by political affiliation? That may broadly indicate who was taking lockdowns seriously versus not? Obesity was a significant variable in worse outcomes - could obesity rates by county be correlated?
I'd personally like to see a lot less of blue vs red because it taints the outcomes with confirmation bias. The media has so desperately wanted to paint this in red vs blue that these analyses became unreadable for the most part. Cuomo the hero versus DeSantis the villain was actively being marketed while NY had 6x the death rate of Florida, which has evened out over time.
The framing is typically blue people are victims of a terrible virus and red people are victims of their own immoral behavior. You know what is actually correlated the most? Being old. I'm sorry to report that this isn't very interesting because red and blue people age the same every year according to my sources, ha ha.
It's not that they haven't tried to prove blue > red over and over and over. The data just doesn't support anything but very weak correlations and you pretty much have to throw out NY to get where you want to go. Covid is equal opportunity and that opportunity just doesn't change much across counties and borders and how you register to vote.
The biggest problem here is the assumption that there aren't external natural causes for increases and declines in case rates. There are no lockdowns for the flu and it is quite seasonal and people aren't voluntarily changing behavior for a flu outbreak. The timing of the worst US covid outbreak is nearly coincidental with the flu season. People change their behavior by season but this would still need to be factored out for a fair comparison that was judging either personal change from fear or government mandate.
The seasonality of the flu is not well understood and one assumes parts of the increase and decrease of covid are also unrelated to behavior (perhaps temperature, humidity, etc.).
This seemed to me a glaring omission in of the whole analysis. If the answer to the question "Do lockdowns work?" is a resounding YES*, but the magnitude is, "at about 0.1% of the effect that seasonality has", then there's this big asterisk, saying:
*The direction doesn't matter because the magnitude is so low.
I felt like that's the lesson everyone should have learned back in November, but then we forgot the obvious and went back to sifting through messy data looking to tease out minor effects from a whole lot of noise.
Think about it: after months of lockdowns, and despite no significant behavioral/policy change, the entire US experienced a massive, sustained outbreak of COVID-19. That outbreak didn't care about state borders or political parties. It's not that we were suddenly doing something across-the-board different in November that caused case rates to go up.
I predicted back in March 2020 that (if COVID-19 is seasonal) case rates would naturally go down on their own by May. They did. I then predicted that they would come back in November. They did. This isn't sophisticated analysis, but the magnitude of the effect dwarfs the squint-to-see-an-effect complex analyses for the lockdowns. Even I was surprised by how little impact lockdowns and other measures clearly had relative to the seasonality effect.
Incidentally, when my prediction in November came true, I further predicted that the following March/April we'd see case numbers fall again (which of course they did). However, I added a prediction that people would perceive the vaccine as the cause of the falling case rate, and that people would once again ignore seasonality. That last prediction is my most depressingly accurate, despite the evidence that the vaccine likely had very little impact on resolving the outbreak. (https://www.marklwebb.com/2021/04/is-it-over-yet.html)
I feel like this is a lesson we're incapable of learning at this point. Especially now that it's clear COVID-19 is endemic, we still have people who should know better (including Scott) saying they're hopeful the pandemic is at an end. Seasonality is not a hard lesson to learn, but I think there's this illusion of control we want to maintain, so we ignore the big things so we can focus on the smaller ones. We don't want to admit that all these heated debates were moot from the start, because we want to believe we can control a global pandemic. Why did we ever think that was possible in the first place?
It's not exactly a popular opinion but I am quite disappointed in science over the past 1.5 years (with the strong exception of vaccines). They struggle with the very basics of disease transmission and personal risk reduction even where there are millions of cases. I feel like I'm walking out of the dentist office after 40 years of treatment with a toothbrush and floss ... "this is really all you got here?". We are dumping trillions into this, is it really that hard to prove surface transmission wasn't a big factor? Airborne and aerosol transmission took how long? Over dispersion took a year to figure out?
How well do masks actually protect you? Nobody seems to really know. Does anyone care about HVAC at all? Seems important. There just seems to be a lot of arm waving, most of it in the right general direction, but not confidence inspiring and lacking specifics.
If homes are the biggest vector of disease transmission, why didn't anyone recommend wearing a mask in a house? Isn't riding in a car with someone like the worst case scenario? An on and on.
The deaths were a majority in old age homes that weren't able to cope. Something like 81% in Canada. Something like half in Sweden of all deaths were in old age homes. 42% in the US. Belgium 42%. 64% Norway. Spain 52% How the old were handled were likely the most important factor. How soon these were locked down and managed.
"If there had been looser lockdowns, more children would have had to go to school, which would have been either good or bad depending on how you feel about school."
This is a throwaway line in the essay and not the main point, but I think it undersells the likely importance of this decision.
The mainstream economic consensus is that each year of schooling increases lifetime income by 10%. In the United States this equates to an economic cost of ~100K (in net present value).
We have absolutely no idea how much worse virtual school is than in person education, but even a conservative estimate of 10% would equate to a cost of $700B. Dividing that in half (to only apply to the blue states) would mean that the costs of closing schools were more than 6 times higher than the direct economic impacts.
> The mainstream economic consensus is that each year of schooling increases lifetime income by 10%.
I'm not sure this is quite getting at the same thing. My understanding is that the economic consensus is about additional years of schooling vs no schooling at the end of one's education. That is, it's about the difference between dropping out as a junior and sticking around for senior year of high school, or the difference between going to a year of college or going directly to the workforce, or the difference between sticking for the second year of a PhD program rather than dropping out after the first.
But the covid school closures are more like no school for younger students than for older students. That is, a second grader who "attends school online" gets basically nothing compared to a second grader who attends in person. But it seems likely that a high school senior who attends school online is getting most of the same benefit as a high school senior who attends in person. And a second year PhD student is really getting almost exactly the same experience online as in person, unless their training involves a heavy lab component (which they probably still did this year, despite shut downs).
Well, you also need to assume people can't make it up later. Meaning, the education for the kids now continues at the same rate for the same length of time, regardless of the fact that a year has been chopped out (or made lower quality). That is, we need to assume nobody compensates by staying in school a little bit longer, or working a little harder for the next year or two. I guess we could assume that education is already delivered as fast as the human mind can accept it, so no catch-up is possible.
Those are reasonable assumptions, and by "reasonable" I mean "near-certain". We are not going to be creating a special Thirteenth Grade for just this cohort of students. Schools and teachers are going to be extra forgiving of students whose educational progress was disrupted by the lockdowns, not extra demanding. The parents who hypothetically would be extra demanding and make their kids work harder the next year of two, are mostly the ones who supplemented Zoom-schooling with a dose of home-schooling so their kids mostly don't need remedial education now.
And all the standardized tests will be rescaled for that cohort? So nobody gets bad news on the PSAT (if that's still a thing) and thinks -- geezus, I better buckle down if I want to get into college at all?
I tend to think most long-lived systems are in stable equilibrium a priori, which means they have restoring forces when perturbed. There's a whole lot of "thermometers" poked into the educational process these days, to see how things are going. I have a hard time believing either (1) they'll all be rescaled, or (2) nobody will pay any attention to the feedback and adjust.
I think it's very reasonable to believe there will be *some* educational loss -- indeed, I've seen it in my own school-age kids, and in some cases, like seniors in high-school with a lot of lab-oriented or otherwise hands-on/direct contact projects, these losses could be quite significant and personally derailing. But I think the analysis that says the system (schools, teachers, parents, students) will not constructively respond *at all* to the loss is too naive to be of much worth.
In your GDP vs lockdown analysis, what are the units for GDP? I know its decrease in GDP but is that a raw value or a percentage of GDP for some control day?
Pretty disappointed that you missed the most obvious confounder in all this discussion. Quote:
*** This is an absolutely beautiful graph. It’s showing how lockdown strictness (as of May 5) correlates with death rate over time. We find that early in the epidemic, the stricter your lockdown, the worse you're doing. This is the endogeneity - places (like NYC) that are doing really badly institute strict lockdowns to try to save themselves. Later in the epidemic, the stricter your lockdown, the better you're doing - probably because the strict lockdown is giving good results. ***
OR it could be that the states that got hit early had partial herd immunity because of getting hit early. This continually got missed by everyone all of 2020 and I'm sad you missed it today.
>Are there other explanations? I wondered if states that did worse in the first wave might have done better in the second because most of their vulnerable residents had already been infected/killed. But I don’t think this is a big problem; most places had only about 5% of the population infected during the first wave, and even the worst states only about 10%.
I'm pretty sure these 5% and 10% numbers involve antibody tests taken later, though it's possible that there was a methodological problem at precisely this point.
If you can point me to a quality population level serology study in June I'd love to read it, but I scoured the earth looking for one, presuming the CDC would have it in their best interests to generate something like that, and never saw it. The infection wave in Spring 2020 in Atlanta was way higher than 5%. More like 20%. It's just that nobody knew because nobody had tests, and even people who knew for sure they had it would go to the hospital and get turned away for a test because of test rationing.
If way more people had it than knew about it very early, when there were also few deaths, then later on, when testing *was* widely available and deployed, and there were also lots more deaths, then that would imply that the death-to-cases ratio significantly *rose* from the beginning to the end of the pandemic. That's a very counter-intuitive result, since we know treatments improved.
Also, could you define what you mean by "partial" herd immunity? I've never heard that term. The usual thinking behind "herd" immunity is that it's like a percolation problem, e.g. water seeping through cracks in concrete. If you have enough impermeable concrete, then the cracks don't form a continuous path and the water can't get through. If you don't, the cracks do, and the water gets through. But it's kind of an all-or-nothing thing, there's no such thing as the concrete being "partially" leaky. It's either leaky or it's not.
Same thing with pandemics: either the rate of general "impermeability" (immunity) is high enough that there aren't any continuous paths of transmission that connects almost every vulnerable person to an infected person, or there are. There doesn't seem to be any in-between "partial" state, which is why I'm puzzled by what you mean here.
The epidemiological models of pathogen spread are all fucked. They presume even mixing of a population in a Petri dish, when in fact it's a lot more like a virus spreading through a computer network. Ask yourself which case provides more net 'immunity' to a population:
Forced vaccination of a hermit who lives on a hill, or
The stripper at the titty bar catches covid and then gets over it
Early infection is most likely to run through the people who have the highest exposure to other people, so each infection and recovery among that network gives you more juice than vaxing some dude who lives in the sticks.
If you look at deaths/day, the March 2020 wave in the USA was about the same as the December 2020 wave. But the spring 2020 wave was undercounted because of test rationing, and the December wave was overcounted because testing was so ubiquitous. The two waves were comparable, and the spring wave was probably a lot worse in hot spots. The New York deaths curve was insane in April (1k/day) and only peaked out at 200/day in January. New York's rate over the summer of 2020 was almost exactly the same as it is today after mass vaccination. New York *definitely* had herd immunity in summer of 2020, and they got their new winter wave after the immunity wore off.
New York had 25k deaths by the end of May, and that was an undercount (I believe excess deaths in New York were probably ~50% higher, but you'll have to take my word for it right now). That implies somewhere in the neighborhood of 5-6 million infections in a state of 19 million people. That's a pretty chunky break on further outbreaks, though they did still get hit (albeit milder) over the winter.
Quick math - the big midwestern outbreaks in the winter seemed to grow for roughly 8 weeks. Let's say R was 2.0 during those outbreaks in the midwest. In NY with prior immunity of 30%, R is 1.4.
If both areas start with 1000 infections, at the end of the outbreak you'll have 250k infections and 1500 deaths a week in the Midwest, and 15k infections and 90 deaths a week in New York. And the only difference is it's the first real wave in the Midwest and the 2nd in New York
Be fair, part of the high NY count was the nitwits in charge throwing infected people back into nursing homes. But I generally agree with what you're saying - real infection rate in NY was much much higher than the numbers we have because of test rationing and lack of testing at all. Any analysis which doesn't take into account the lack of testing in spring 2020 is a bad analysis, which means all good analyses require a significant amount of guesswork.
The biggest problem with this post is that the range is too limited. The post does not mention the most obvious lockdown failures: India (which had the strictest lockdown) and Peru (which had the highest COVID death rate). It also does not mention the fairly lockdown-light East Asian success stories.
"In the end, Sweden still ended out with a death rate about double the European average. Seems pretty bad."
Objection: there's no such thing as a European average pre-second wave because Eastern and Central Europe was spared. In the end, Sweden ended up having a lower COVID death rate than most Eastern Europe.
"The real question we should be asking is what set of policies countries should have implemented."
Bingo.
"Contra Lemoine’s picture where Sweden just has an earlier start but eventually does no worse than everywhere else, here Sweden has the same (or better) start as everyone else, but clearly does worse afterwards."
Agreed.
"If Sweden had a stronger lockdown more like those of other European countries, it probably could have reduced its death rate by 50-80%, saving 2,500+ lives."
This is a dubious conclusion; it's only true for the first wave, if it's true at all. But there is a difference between preventing deaths and delaying them; Bulgaria had hardly any Spring 2020 wave at all, after its winter and spring waves it has the second-highest COVID death rate in the world.
Any analysis of lockdowns need to tease out nursing home policy. Sweden was blasted over it, so were New York and Pennsylvania. In my PA county to this day, 64% of deaths were nursing home residents. No amount of lockdowns was going to salvage a policy of stuffing confirmed COVID-positive patients back into poorly ventilated nursing homes
> Nobody agrees on exactly what “Scandinavian/Nordic” is
There appears to be a consensus that "Nordic" means exactly the five countries you mentioned, and it's just "Scandinavia" that's ambiguous, with it most often referring to Sweden, Norway, and Denmark, but sometimes also excluding Denmark, or being used synonymously with Nordic countries, as you do here. At least, that's what wikipedia says, and the maps I find by searching those terms on google images seems to support it.
Excess fatality tells a different story still. Personally, I'm content in knowing that I'll never know the effectiveness of various measures. Can you ever really know? If the deaths had been really bad, who knows we might have freaked out enough to start taking vaccines, try vit. D/ivm, let people test themselves, etc.
It's been unfortunate to see the discussion framed in terms of data rather than principles. If government can do X if data is Y then moral hazard naturally increases greatly as the measurement error in Y increases. I'd much rather attempt to reach consensus on the moral/philosophical framework.
Insofar as the presence of co-morbidities matters to policy assessments, why don't we see data (e.g. here), distinguishing #s of deaths, of folks with, vs. w/o, major co-morbidities?
The paper linked to here, on that "the average COVID victim *might* have lived another 8 QALYs", is quite a "might", seeing as it admits that "the SMR and reduction on remaining quality of life are not empirically estimated....".
Some data from the first wave (May 2020): 90% of fatalities were 70+, of which 50% people were living in retirement homes and 26% had home care services. For various reasons, Swedes are really old and fragile when they move to retirement homes. The average age of people moving to retirements home is 86, and about half die within a year (pre-covid). So I find it hard to believe that Swedish covid-19 victims would on average have lived another 8 years.
Can someone help me evaluate the Leech study on mask wearing? They detect a significant reduction in R from mask use, but they use a seemingly super-complex hierarchical bayesian model which always makes me suspicious. Does the data really support their conclusion or are they hacking their model somehow? I've seen other observational studies that don't correlate mask usage with R reduction, lab studies that show no reduction in viral cultures from mask wearers, and older data which indicate masks don't reduce influenza transmission. Can someone help me reconcile these data? I'm sure n95 masks help, but in my experience it doesn't seem like anyone wears those and I am highly skeptical that the average cloth mask worn by the average person could possibly do anything.
All the mask studies I have seen are overly complex which brings in greater opportunity of confirmation bias. One should ask why nobody simply asked people who tested positive or were hospitalized whether they were reliably wearing masks. 1000 interviews should get you some decent data. This would also have plenty of confounders but at least the starting point is more robust.
I finally just threw up my hands and assumed masks likely help but likely not as much as the preferred narrative would suggest, and they were not a huge burden so go ahead and wear them. My wild guess is they reduce accumulated risk by 10% to 50%.
The media seemed to be overly careful to never show or suggest that mask wearers were getting infected.
You came to the same conclusion I did, with a similar range. If I had to pick a real number instead of a range, it would probably be about 20% effectiveness. 20% is still huge, but a lot of people treated it like the most important factor, or singlehandedly enough to stave off the virus. It wasn't, and couldn't be.
The funny/sad thing is, that there are a lot of people both then and now that treat a mask like PPE, to protect the wearer of the mask. Masks were never billed as an actual way to keep from getting the virus. They were instead a means to not spread it to others. Mask wearers get the virus at about the same rate as everyone else (maybe some tiny percent less or something, but we would never be able to prove that). If you are only around other people wearing masks, you got some protection from them, if they had COVID.
My intuition has it that masks ought to be at least moderately effective based on theoretical reasons, some weak scientific evidence, as well as rational evidence. One such piece of evidence would be healthcare workers working with COVID-19 patients, as far as I can tell, using surgical masks pretty much across the globe even in places where you wouldn't expect this to be due to shortage of equipment. I can find studies that predictably show moving to aerosol-resistant PPE reduces infections by a large factor (from nonzero to zero), but I would expect unmitigated risk to be near-1, and apparently practical experience has it that surgical masks have been deemed "good enough" even in context that to me appears to be maximally risky and cost-benefit of greater protection being as high as it's going to be. Surely the hospitals in these countless naturally running experiments would notice if their workers were getting infected en masse? (Now, I do realize that you could interpret this the other way: after all, studies show that healthcare workers sometimes DO get infected by the patients, and I make assumptions like supposing treating patients is high-risk or assuming hospitals would notice if nurses were getting infected en masse, but by my read this is one of the higher-signal pieces of evidence and I interpret it as showing even mere surgical masks offering high degree of protection - if anyone can tell me why experience in hospitals does show surgical masks to lack any efficacy of note, I'd be interested to know. Perhaps hospitals tend to have such good ventilation that even working with patients is low-risk?)
Be that as it may, the key point here, I think, is that nobody wears masks all the time. People take them off when at home or with friends, when they eat or drink, and in most workplaces. Since the bulk of infections take place in these sorts of contexts, not in shops or public transport, even 100% effective masks worn with 100% compliance couldn't reduce the total number of infections by more than some low tens of percents, and as such finding real signal by conducting studies based on asking people if they had been wearing masks would already be incredibly difficult. Maybe if you only surveyed people whose infections had been contact-traced to situations where mask-wearing is more of a norm?
Your analysis of Sweden is flawed as you're using doctored stringency data. Go to an earlier version of the data and see how much they bumped Sweden's stringency index up by. They claimed they did this because of some vague guidance that was in place (rather than laws or mandates). Their data is silly as it suggests Norway and Finland were even less stringent than Sweden
Based on what I've heard and read from Twitter etc, Finland genuinely was less stringent than Sweden for at least some periods of time after last year's spring.
Here is a theory of lockdown: lockdowns can help people co-ordinate. If the virus is in the community and I don't think people will keep R far enough below 1 to eliminate it, then I have no reason to take avoidance measures beyond protecting my own health. Because infection risk increases with disease prevalence, everyone taking "personal risk measures" will always be insufficient on average to drive R below 1 if the prevalence is low enough (though they may be sufficient once the prevalence rises a lot) and so the disease will persist with cyclic pattern that depends on how quickly people re-evaluate personal risk based on changes in community prevalence.
However, with a lockdown aimed at elimination I can believe that other people will keep R down enough for elimination and I can now weigh the risks I take against the costly possibilities of compromising elimination or extending the lockdown instead of the less costly possibility of personally dealing with getting sick. Thus, with a lockdown, even under my own evaluation I should accept a much lower risk of infection. There is still a prevalence below which I won't take many precautions - but it is much lower, possibly lower than 1 case in the entire country in which case I can take precautions until elimination is probably achieved.
This is a very different theory of lockdown to "disease burden management", though! Also contradicts the view I advocated early last year which was that 12 months of suppression is a reasonable aim.
> nobody responds to the virus in any way and it spreads uncontrollably until 70% of the population has been infected.
70% is the threshold for herd immunity in a naive SIR model with an R of 3.33, but due to the phenomenon of "overshoot", the virus would infect 96% of the population before petering out. The "herd immunity threshold" means the reproduction number of the virus goes down to one (not zero). When that threshold is crossed, there is still a very large population of infectious people, each of whom infects one other.
The equation for the proportion (p) of a population infected by a virus with reproductive number R at the end state of an epidemic satisfies p = 1 - e^(-p*r). So once overshoot is considered, and assuming no intervention, the R of the disease which infects 70% of the population before sputtering out is about 1.7.
I'm not sure what to make of the mobility data. The rest of the data seems to suggest that lockdowns make a difference beyond what people did on their own, but the mobility data suggests that the government was basically telling people to keep doing what they were already doing. Maybe things like "ban on large gatherings" don't show up in mobility data (not that many people are going to conventions), but make a big difference in the spread? Maybe making the lockdown involuntary prevents things like "your boss tells you to come in even if you're sick"?
Your link to "value of one life" per the EPA doesn't link to the EPA. It instead links to a Statista article on "Percent change in Real Gross Domestic Product (GDP) of the United States from preceding period in 2020, by state", which I presume is an error?
The data you looked at, how you looked at it and the various ways of interpreting the data, and, of course, the questions you asked were fascinating and helpful. I can now think of this subject in more enjoyable ways. It's truly a breeze not having all the answers while maintaining methods of attaining them someday. It must be truly stressful needing an answer.
I believe serology tests in NYC suggest the number there was about 15%, I'm not sure why this shouldn't be a big deal because (to simplify massively) if all your super-spreader types are in that 15% of your population, you now have no more super-spreading events to worry about.
>This is a victory for lockdowns insofar as the correlation is significant, but strong proponents might be surprised by how small the effect was.
It would be even smaller if you just drew the straight line through the main cluster instead of calculating a least squares fit. Also "correlation is significant" isn't really appropriate here, you can't just feed your data into something that then spits out the words "significant" or "not significant" and then expect it to mean something; "significant" and "not significant" is *always* tied to a statistical model which you do not have.
Would it be arbitrary and sill and not at all mathematical to just draw the straight line by hand? Yes. Would it be mathematically any worse than what you are doing now? No. If you're doing a least-squares fit on data like this you already statistically in a state of sin.
> US states mostly had stable and predictable responses based on their internal politics - a nice exogenous factor!
Which correlates with the amount of voluntary changes people make; what do you think would happen if you tried to control for those?
a) In the discussion about lockdowns vs voluntary behaviour change the two are considered to have a complicated relationship but still to some degree independent. I think there's also a possibility that voluntary behaviour changes *because* people can hear their government debate lockdowns and can see it might be needed. I.e. the political discussion about lockdowns is also part of the pressure on voluntary change.
b) The death rate calculation is only looking at people dying of Covid. In the UK we've also seen that hospitals full of covid patients have to move resources away from other activities. This gives rise to an ever-growing waiting list for other operations. So you have lower standards of daily life for hundreds of thousands of people and many more deaths. The full death toll, and the 'how many good years saved,' calculations really ought to include people waiting more than a year for gallbladder operations etc.
2 good points, I want to add 2 examples for the first point.
_ Democrats tended to be more responsible than Republicans because their leader favored lockdowns and asked/implemented them while Republican leaders were more of "let us continue to live as if nothing has changed".
_ If you know your leader is going to institute a lockdown in a few days, you do some panic shopping and then put yourself in lockdown mode before the official date because this is a sign that the situation is dire.
In summary point 3, shouldn't that be 0.6%pts more GDP decline, not 0.6%? If the latter then that's almost nothing, but I gather from the data and discussion earlier it's a 0.6% point difference in the change in GDP.
Small note: I think looking at excess deaths is much better than relying on self-reported stats. Eurostat has a tool where one can see excess deaths (easily found via searchsites). It would not change Sweden's general position as doing worst in the Nordics, but would increase the distance between itself and other Western European countries.
One should also make note of the fact that Sweden is by far the most diverse Nordic country, and deaths in all Western countries have been concentrated among immigrant-heavy populations, partly for cultural reasons (high prevalence of multi-generational homes) but also socioeconomic (disproportionately the poorest groups).
At least early on, I think the Swedish focus was clearly more on mitigation than suppression. It seems quite clear that there is a link between that strategy and the high number of covid fatalities in the first wave, as you point out. However, one could argue that the strategy was _partly_ vindicated if you consider the cumulative deaths over the _entire period_ up until now. While Sweden was 2x above EU average by August 2020 in terms of cumulative confirmed covid-19 deaths per million people (as you pointed out), by now it's actually <10% _below_ EU average. Does this mean that Swedish chose a smarter strategy for 2nd and 3rd wave? That there is a natural "reversion to the mean" over time, as most governments were unable to protect the most fragile people for more than a year until the vaccines arrived? I don't know, but it would be interesting to hear your thoughts on it.
In a sense the analysis here is for "we tried X, it mostly didn't work, was it still worth it?" But that doesn't seem like a fair analysis to answer the question "is X worth it"?
In my memory the trade-off at the time was between "can we flatten the curve or do we need a lockdown?" Where "flattening the curve" meant staying at home if you're sick, social distancing when you're out, not shaking hands, don't touch your face etc, it was seen as the more gentle, cheaper approach that could get us through without overwhelming the hospitals. The alternative was a "lockdown" where everyone stays at home and you close businesses, which was more expensive, but was going to give us similar results as in China, where you completely get the virus under control, at least enough to go back to contact tracing.
Importantly this means that we didn't get the benefits that we were promised. We got the costs of a lockdown with the benefits of "flattening the curve."
If these were different terms at some point ("lockdown" and "flatten the curve" got mixed together very quickly after lockdowns started, as if they used to not be competing opposite alternatives...) then the calculation is very different, because we didn't expect the outcome that we got. In hindsight it's easy to say "we didn't get the Chinese results in the US and Europe, so overall it's an expensive measure that didn't get us many benefits" but you couldn't have known that at the time.
Another example may be the collapsed apartment building in Florida recently. Imagine you're one of the people who just bought an apartment in there a few months ago. You bought an apartment, the building collapsed, so now you either don't have an apartment or you're dead. Does that mean that buying apartments is a bad idea? No, because at the time you didn't know that the building would collapse. You have to judge the decision based on what you knew at the time, and usually buying apartments is a good idea.
Of course once people started sabotaging the lockdowns, (most obviously Trump, but also Cuomo keeping construction going...) and once the goal shifted to "we're just trying to flatten the curve" you could have argued that that doesn't make sense any more. And I never heard a good justification for the later lockdowns in Europe.
But before the sabotage we could only make the decision based on the results we saw in China. And some countries got the same benefits as China.
In the next pandemic, if people argue for another lockdown, what will we decide based on? Will we decide based on how lockdowns went in countries that messed it up last time, or based on the countries where the lockdown worked last time? I predict it'll be the latter. People will say "we should do a lockdown, and we should do one that works, because it can rapidly end the pandemic." If that's true, and if people aren't interested in dragging the next pandemic out like we did this one, (see the strong reaction in Asia this time, who had the prior experience of SARS) then they won't consult the analysis in this blog post, because it's only looking at cases where the lockdown mostly didn't work.
> If we had let the virus spread more, it would have gotten more chances to mutate.
Is this so certain? I'm thinking that, given the same number of total people infected, flattening the curve and keeping R on average around one for a longer time is much much worse regarding giving the virus maximum selective pressure, and it probably more than makes up for the reduced total number of people infected.
Also, let's irresponsibly assume stuff. Let's say that, doing nothing, 2% of the population would have died within the first 6 months and then herd immunity. And that each death killed 10 years of life. So the pandemic ended within 6 months, reducing life expenctancy by 2% of 10 years = 70 days. As a citizen of Spain (life expectancy of 83.2 years), you would need >23 covid pandemics in your lifetime to turn your country into the US (life expectancy 78.5).
The damage would not be spread out evenly among all people, any more than if my neighbors house burns to the ground all of us on the street experience it as an average slight singing of our furniture. Even the risk wouldn't be evenly spread, as the it is highly concentrated among the old.
So if nothing at all had been done, the actual outcome would be that almost nothing would happen to everyone under 40, those in their late 50 and early 60s would've experienced what people in their 80s do during a really bad flu season -- something like a 1 in 20 culling -- and among those over 70-75 it would have been close to the Black Death, something like a 20-35% mortality.
That is, deliberately doing nothing would be the functional equivalent of just writing off all of the very old folks, pushing them out on the ice floe.
I'm not making the case that it would have been a good choice to let it run. It's just another angle to put the whole thing into perspective.
But yeah, it's uneven. Though the extra deaths in normal US versus normal Spain also disproportionately affect the old. And, for a fixed number of life years lost, having them affect the old who have less years left to live means is actually a way to distribute it about as much as possible.
For US comparisons it would be nice to look only at what types of businesses are allowed to be open (also if we were able to know what fraction of eligible businesses were open as a measure of voluntary effect) since that's really the area the government has the most control over (since we did not use the police to break up house parties or give fines to individuals for walking outside etc)
I don't think Conclusion #2 (Quantifying Emotional Costs) gets enough attention. In my opinion, this is perhaps the most important trade-off. I lived in Idaho (with family in Spain) through this whole thing and I can't think of a better place to have been through it all... I sure was glad I didn't live like a prisoner as in Europe (for a whole year!); in rural parts you wouldn't know there was a virus even during the second wave in the winter.
And all this madness everywhere else just to save a few QALY (at the cost of sacrificing a bunch more QALY)? I think there is a lot more to live than avoiding death. And somehow, Idaho's deaths numbers are some of the better ones too!
The one aspect which only time will uncover: the economic ill effects from the lockdown are long term. Small businesses destroyed - the health and economic effects are decadal. Even people unemployed but on short term government subsidies - that impact is also decadal.
The entire economics analysis is contingent on a false assumption - that economic activity/growth is independent from one political entity to the next. If all but one country decide to lock down their borders and limit trade, how much of a difference does it make that the one holdout is still willing to play? If all your kid's friends tell him they won't play with him, you can't claim that he's outside "playing with friends" because he's still willing to get together. Economic activity isn't independent, and this assumption should place a giant red asterisk on the whole analysis.
Let's do a counterfactual reasonableness analysis. Let's say nobody knew about COVID-19, because we hadn't invented the technology to test for it, and hospitals just treated it like a particularly bad cold. Cumulative US deaths from COVID-19 for >1year currently stand at <0.2% of the population. Cumulative infections at close to 10%. Is it reasonable to attribute massive GDP changes to the virus alone? (Because <0.2% of people died and <10% got sick for two weeks?)
What if next year a bad variant of H2N3 flu started spreading across the US, but nobody was focused on it because we were too busy watching for more COVID-19. Say 10% of people got sick from it for 2 weeks, but nobody got scared because everyone who got tested showed they were negative for COVID-19. Of those infected, around 600k people died (out of a population of >350million), but they were mostly elderly, so nobody thought it was out of the ordinary. How likely is it that a drop in GDP would have even been measurable under those circumstances?
Now, maybe that thought experiment doesn't exactly encapsulate the difference between voluntary behavior and lockdowns. But it seems clear to me that nearly all of the economic impact from COVID-19 is downstream of behavior, not due to the virus itself.
I think this analysis missed out on the fact that increasing voluntary quarantining decreases the cost of mandatory quarantining.
This is maybe impossible to measure, but I think it would dramatically affect the cost calculations when looking at red vs blue states, as my assumption is that blue states had much higher rates of voluntary quarantining, and so stricter lockdowns were less costly there than they would have been in red states.
I played and earned Czechia probably $10B by killing all the old people who were going to use healthcare, draw from retirement, etc. Not to say that makes much sense, but neither does estimating the "economic costs" while ignoring the largest factor.
A very well written, sober and judicious piece. Nicely done indeed.
I have on quibble, which is with your conclusion #5, in which you compare the cost of emotional suffering by those who endured lockdowns with the benefit of the QUALMs gained by people who don't die of COVID.
I think if you are going to count the *emotional* suffering of those who endure lockdowns on the "cost" side of the ledger, then you have to count the (averted) *emotional* suffering of the friends and family of those who would have died on the "benefit" side.
The emotional cost of a death is not just, and perhaps ultimately not even mostly, borne by the person himself dying -- a death has savage emotional (and even practical, financial, childrearing et cetera) repercussions among the family of the deceased, and these go on for a long time, years often.
...and perhaps I should add that some of these emotional family costs were very heavy indeed during COVID, as often enough the family was not allowed any contact with their father, mother, uncle, grandfather, et cetera while the latter was dying.
No, you shouldn't count the emotional suffering of the family and friends of people who would have died.
Everybody dies. If someone dies from COVID and their family suffers emotionally, you can add in the emotional suffering, but then you need to *subtract* the emotional suffering because they didn't die later. You can only die once, after all; dying earlier just changes the time of death, and friends and family suffer either way. Now, you might argue that people suffer more when their friends and family die younger, which is true, but you can't count that as a whole death of suffering, you need to count it only as a fraction. And if someone dies at an older age, they're more likely to have someone *they* know die (since their life is longer, more time for other people to die in), which compensates for that.
Not to mention that COVID mostly kills old people anyway.
A strange point of view. So it wouldn't matter to you if your own daughter died at age 4 from Wilm's tumor, because she's just going to die anyway someday, perhaps at age 77 from a heart attack? Maybe it's even efficient, you get your mourning over with early on, and can move on with the rest of your life.
Always interesting to see what a wide variation in human nature there is, so thanks for this insight.
The point is that you can't save lives. You can only save life-years. If someone dies at 110 of COVID then they were probably going to die in the next year anyway. If someone dies at 15 of COVID then that's a bigger loss.
(Even biological immortality can't save lives. Accidents, murder, famine and war have a nonzero chance of killing you per year, and that inexorably adds up. It *can* increase the life expectancy to hundreds of years, but not infinity.)
Shouldn't that just be priced in as part of the cost of a death? I don't see why deaths from COVID should be treated differently from any other kind of death in this particular respect.
One point I'd make is that decisions by politicians have ramifications even where the leader's writ doesn't run. For example, on March 15, 2020, the mayor of Los Angeles shut all the movie theaters in Los Angeles, the movie capital of the world. I suspect that had impact around the world because you'd expect the mayor of Los Angeles to be in favor of people thinking of going to the movies as a fun and safe thing to do for the whole family, kind of like how the mayor of Amityville in "Jaws" is pro-going-to-the-beach. But suddenly Mayor Garcetti was announcing, in effect, that going to the movies could KILL you, so do not, by order of law, partake of Los Angeles' most famous product.
In contrast, the prime minister of Sweden did not shut Sweden's movie theaters. But very rapidly, nobody was going to Sweden's movie theaters so they shut themselves down for lack of business.
Ironically, the payoff from lockdowns in America is mostly due to the almost miraculously effective mRNA vaccines riding to the rescue faster than just about anybody other than the ever-optimistic Mr. Trump imagined. If instead we got, say, some 75% efficacy vaccines in late 2021 with supply chain problems keeping everybody who wanted a vaccine from getting one before 2023, then we in the US would have likely gone through 7 or 8 waves infecting maybe 75% of the population before the vaccines finally kicked in.
Under that scenario it's hard to see much point in kicking the can down the road with lockdowns other than to avoid overloading the hospitals (and hospital administrators proved pretty good at stretching the capacity of their hospitals in the pinch).
Of course, under that scenario, lots of people would have voluntarily locked themselves down: e.g., as I pointed out 15 months ago, movie theaters were legally free to be open in Sweden, but their owners shut them down anyway because everybody had stopped going to the movies. (I cite movie theater data a lot because it's easy to find only at BoxOfficeMojo.com.)
But because of the remarkable efficacy and speed of development of vaccines (which, amusingly, Biden and Harris were spreading Fear, Uncertainty, and Doubt about during their debates last fall), blue state policies come out looking pretty good.
Back on March 18, 2020, I wrote:
"The [British] forecasters argue that suppression would need to be practiced for the majority of the next eighteen months, at which point they hope a vaccine would be ready.
"Even worse, I’d point out, it’s possible that no vaccine will be found: Coronaviruses are hard to vaccinate against. They tend to mutate rapidly.
"On the other hand, kicking the can down the road for seven months or so on the prospect of mass death might possibly pay off in multiple ways, not all of which can be specifically anticipated at present.
"For instance, it seems possible for the English-speaking countries to have a South Korean-style testing regimen working by fall that could put out flare-ups before they turn into the apocalypse.
"There are many different methods to create a vaccine. Well-funded crash programs might get lucky.
"They could speed things up somewhat by cutting corners on safety. Today’s vaccine researchers are not used to taking heroic steps to rush because they typically work on the fairly obscure infectious diseases that still lack vaccines, so there isn’t much excuse for risking nasty side effects lately.
"Medical treatment of the infected ought to improve with time. Testing will certainly get better than it has been. (In America, it can’t get worse.)"
Apologies I only had time to skim this for now, although it all looks reasonable and I'll try to read it more closely when I have time. Apologies for the usual academic self-promotion, but I did want to mention that I just published a paper on this topic in Health Services Research:
The disadvantage compared to many of the ones discussed here is that ours is [much] less sophisticated modeling. The advantage is that by the same token we require [many] fewer assumptions.
What's the explanation for the Sweeden's huge peak around Jun 9-29 in infections not translating into analogous peak in covid deaths (which seem to continue to decline in this period)?
Is it that the first wave killed all voulnerable targets? Or perhaps voulnerable people stayed at home during the second wave? Or did something change in medical capacity or reporting? What's the official story?
My take: if this much analysis can only find a possible, small effect, then all but the most comically inadequate utilitarian (plausible) moral theory must condemn lockdowns vehemently.
By the way, since there's been people suggesting similar threads for masks, vaccines etc., can I suggest another Much More Than You Want To Know COVID thread for a topic that's interested me for quite a bit but which I can't really make heads or tails out of: Long COVID? How much is it a real physical thing, how much a psychological thing, and how much other conditions lumped with one moniker when they randomly crop up or flare out after COVID? What are the actual numbers of sufferers - and how many of the cases are serious, or have lasted more than a few months? There really is little "official" information of any kind on these.
It’s good that you made an effort to assess actions in a comprehensive way and the additional debate and discussion is also useful. The discussion HAS to begin somewhere. The mental health aspects of the lockdowns remain largely unrecognized and unreported. Anecdotally, therapists were nowhere to be found in the US after the child depression epidemic spiked. Appointments were available one year out in the Seattle area for example. Hopefully some best practices around how to handle a pandemic will emerge. Lessons of the Spanish Flu didn’t seem to provide much guidance except that children remained in school and didn’t have screens so they were already outdoors, which probably helped a great deal.
I'm not sure I absorbed all of the statistical analysis, every time I try to consider more than a few simple factors, my head spins. However, just to throw *more* info into your pile, here's some potentially interesting information that is not usually discussed when going over the Swedish infection rates/death rates and course of the pandemic in 2020.
A lot of what happened initially in the pandemic (ie Feb-Mar 2020) was the result of a ‘perfect storm’ of setup and subsequent events. The longest prior situational setup was that, prior to 8 years ago, there was a gap in the Social Democratic dominance of parliament for two terms, where “Moderates” (considered right-ish here, more like Obama-style democracy/neo-liberalism) had a coalition with the Liberals (market libertarians) and other right-leaning parties. Several waves of privatization and deregulation happened in this government, eg they sold the post office to the Danish company Postnord (who promptly shut down 2/3 of processing stations to function economically, you can imagine how the mail is now), most of the health clinics were sold to Capio (a CAPital Investment Organization who runs health care in France, yup, health clinics have immense wait times now and are understaffed )and more that lovely old Social Democratic Sweden will never recover from…
Now, old peoples’ homes used to be run by the state, but many had been farmed out to private organizations who either cited some obscure regulation (not enough space between a door and a window type of thing) to kick all the old folks out and then rent the places out to refugees/asylum seekers, charging the state massive rent as Sweden received more and more people fleeing war and famine etc in Iraq, Syria, Eritrea, Afghanistan, etc. Or they continued to run them as old-folks' homes, just more cheaply.
The remaining now-private old peoples’ homes were additionally no longer regulated with regard to what was legal or necessary training for the workers—nor pay levels for workers—so the äldreboende ended up losing all their previous staff who simply would not work for minimum wage, and hiring, yes, off-the-boat immigrants at low pay, and no training.
Ok, so, next step. Stockholm, being the “Capitol of Scandinavia” is a very wealthy city, by which I mean: it’s extremely bourgeois. One of the yearly parts of people’s lives is Sportlov, a “sporting vacation” that happens in late February, where everybody goes skiing! Except, if you’re rich and from Stockholm, you don’t go skiing in Sweden, it’s too flat, so you go to Northern Italy. Only the middle class skis within Sweden.
(personal note, I was playing music on tour in the US in Dec2019-Jan2020, we all got incredibly sick directly after shows on the west coast after Christmas, continued with shows on the East Coast with the aid of University sports doctors applying “Z-pacs” of steroids and antibiotics, so we were probably super spreaders of an as-yet-unknown virus! I got back to Stockholm at the beginning of Feb, had two more shows in Finland mid-Feb, but by then Covid-19 had been identified and the start of pandemic prep had begun. Not that my family goes anywhere for Sportlov anyway, we’re not that class of citizens, being musicians and schoolteachers.)
So. Rich people flying to Northern Italy in February 2020. Who drives them to and from the airports? Taxis. The cab drivers are, yup, mostly recent immigrants from the aforementioned war-torn regions, who, having escaped with families, often live in multi-generational households. Viral transmission begins in earnest, apparently a lot from London and to a lesser extent from Northern Italy. The taxi drivers are a big vector. This combines with in-country skiers flying up north to ski in Sweden.
In many of these households it is/was common to sit by the bedside of a person who is sick, and you can imagine the health and safety warnings not exactly reaching through the language barriers (honestly, even getting info through to me in English was not top notch, and I saw nothing in Arabic or Somalian/Eritrean at that time, and I live across the street from an asyl-boende, where they house teenage asylum seekers with no families.) Older people started getting Covid, the families sat with them, then the men go back to work driving the taxis, the women off to the old peoples’ homes, where viral transmission starts to take hold of the elder population in a big way. And the people working in the old peoples’ homes did not know how to deal with it due to a) no actual medical or assistance training and b) language barriers. The bulk of Covid infection in Stockholm, the bulk of the deaths in March and April 2020 were caused by these vectors, minimum wage jobs held by recent immigrants carrying the virus home from rich people traveling the world, to their families and then out to their jobs. So the thing is, there *are* a lot of multi-generational households here, they just aren’t the rich white peoples’ households.
It took most of April and into May before they figured that out. Regardless, I don’t think any regulation has yet been put in place regarding training for hospice workers—it takes a long time to get anything done in this sort of bureaucracy. At least this year (2021) I’ve seen vaccination posters and info in nearly every language available all over the place. I think there are people *not* getting vaccinated, still (I believe it’s just gotten to the under-30 year olds this month) because, yes people are idiots and especially in extremely entitled societies there will be people who think that nothing can harm them. With the lack of any real lockdown, most people (in Stockholm anyway, I didn’t go anywhere to observe anybody else this past year) ignore anything not explicitly mandatory—mask usage on the subways was like 5% max any time during the past year+. People just don’t pay attention, think “won’t happen to me!” etc. So you could consider 'entitlement' a possible factor as well.
Replying to myself to amend the "multi-gen household" idea in Sweden: summer is different than winter. Many people move to sommarstuga/summer houses for a month or more in the summers which both alleviates congestion in the cities and creates more multi-generational households for short periods of time.
Currently (July 2021) cases in Stockholm are up 44% (says SVT) and it's almost all people in their 20s, not hanging with the folks in the country.
At least that's where you have to look as well, if you see such a huge difference, but I guess you have to know say italians and scandinavians personally, otherwise it might not be a common knowledge.
Curious where these numbers come from, given that they are not at the link ...
> For example, when the EPA is determining how worth-it environmental regulations are, they value one life at $9.1 million; when the Department of Transportation is determining how worth-it road safety regulations are, they value a life at $9.6 million
Thank you, great analysis, for me the crucial point: "It's harder to justify strict lockdowns in terms of the non-economic suffering produced. Even assumptions skewed to be maximally pro-strict-lockdown, eg where strict lockdowns would have prevented every single coronavirus case, suggest that it would have taken dozens of months of somewhat stricter lockdown to save one month of healthy life."
and I would add that what you missed - you didn't address at all - is who is threatened by the disease and who is threatened by the restrictions, that here we are clearly shifting the burden from the sick and elderly to the children, that they are hardly threatened by the virus at all, but by the restrictions quite fundamentally, and again it is up for debate how such a manufactured shift is moral
and the second point, that you does not address at all that restrictions can still hurt after they have ended, as explained here:
Brilliant pondering. My take is the population that continues to wear masks and vaccinate will be generally fine as the variant breaks through the vaccinated. Those who don’t continue to wear masks or eschew vaccination will have a much higher viral load, and thus a tougher fight for recovery. For children under 12, they need to be isolated until a vaccination is available to them.
I remember people talking about non-official signals of seriousness as drivers of individual behavior; that there was a major uptick in taking the pandemic seriously when large organizations (e.g. the NCAA and NBA) started taking it seriously even though it lost them a lot of money and a major downtick when leaders (California, Britain, etc) were caught hypocritically flouting their own lockdowns.
Even if the confounding factors could all be untangled, isn't all of this granular analysis basically pointless because: (1) the "new infection" data are kinda garbage due to positive test rates being highly unreliable and, in any event, testing is not conducted on a randomized basis; and (2) the "covid deaths" data are also kinda garbage because they don't distinguish "died with" and "died from" covid.
Sweden comes 1000-4000 deaths 'worse' than it allegedly could in Scott analysis, but Scott makes the mistake of not looking at the big picture. If you plot overall mortality over the years per week (for example) you don't see anything interesting at all. Scott is just looking at noise. There is no signal there. There was no pandemonium of an epidemic. There was nothing but pointless lockdowns.
Also, Scott should compare with the predictions the modellers did. They expected 70000 excess dead in Sweden *with* the measures taken. The modellers were stupidly wrong and have never accepted their errors.
I didn't see where you dealt with the problem of risk. Unlike plague, dengue or cholera, COVID was a novel disease. We now know all kinds of stuff we didn't in early 2020. For example, we know roughly how dangerous and how contagious it is. We know how likely it is to kill someone, how long a typical infection will last, something about how to treat the disease, how quickly it mutates and so on. A lot of our knowledge is still sketchy. No one has had COVID and lived for even two years.
When dealing with a novel risk, it can make sense to be more cautious than when dealing with a well understood risk. Doing a retrospective analysis ignores this. It's called Monday morning quarterbacking in certain circles. Suppose COVID had been more deadly, less amenable to such treatment as we now have, more infectious, more likely to become more dangerous adn so on. Similarly, it could, as our president at the time suggested, have simply gone away on its own.
If you are a politician, you have to deal with risk. Sometimes they overreact. After 9/11 the political assumption was that the world was full of terrorists plotting new, similar attacks, and that it made sense to invade Afghanistan and Iraq, if only to send a message that the US was serious about fighting terrorism. Anyone who suggested that this was overreacting, perhaps arguing that it made sense to invade Afghanistan where the attack was organized but not Iraq which had nothing to do with it, they were noisily and sometimes viciously shouted down and sidelined. One politician argued that overreaction was proper in the face of "unknown unknowns".
Of course, it was much easier to understand and deal with 9/11 style terrorism. We had a lot of knowledge about how terrorist attacks could work, what was needed to make them less likely, conditions and potentials in the Mideast and existing terrorist organizations. It's not as if the intelligence community knew nothing about Al Qaeda and friends.
In contrast, COVID was something no one had ever seen before. The closest models were SARS and MERS which were much less infectious and could be stopped with conventional contact tracing. COVID was clearly more dangerous and much less well known. The only way to find out what the threat was was to let the threat play out and try to stay ahead of it. Needless to say, this did not work very well in Italy, NYC or elsewhere. It only worked in China because of massive reaction, shutting down entire cities, complete lockdowns that would have been impossible in less regulated nations.
Looking back and saying that we should have traded so many dollars for so many lives or illnesses or perhaps traded so many lives or illnesses for dollars is rather pointless.
As mentioned: Belgium covid high death rate partly explained by an unusual counting method. I'm surprised this analysis did not include excess death rate trends. Data for 2020 has come out, for example, here. This table is "age adjusted"; I'm not clear how that works exactly). Note that Sweden comes out lower/"better" than Germany by this metric. https://www.cebm.net/covid-19/excess-mortality-across-countries-in-2020/ )
I'm missing your ethical assumptions in this piece when you talk about what does and doesn't justify a lockdown. You seem to be implying in some parts that governments ought to act according to some utilitarian sum, but why shouldn't they (for example) maximize freedoms of the individual? Actually I don't think 'objective ethical theories' can be grounded at all and then government are just the results of complicated social contracts between citizens that act out of their (ir)rational self-interest, and from that perspective why be in favour of a lockdown if it doesn't benefit you more than not having a lockdown?
Late response here, but the Branch Covidians are plunging us into a dark age from which we may never recover. If you think this was about giving boomers 1-2 years of extra life, or saving fat people, this is very naive.
Humans aren’t meant to do lockdowns, better to have no humans left than a world where we are tightly controlled in this way.
I’d rather DIE than live in China, and am starting to think that everyone should rather than anyine live in such a subhuman state where you can be welded into your dwelling, removed into camps, injected involuntarily and be rendered nonessential
As a psychiatrist, if you don’t recognize the catastrophic effect of this on mental health, I suggest a new profession
Various actions and decisions in wars, and their efficacy, tends to get debated decades and centuries to come. That still doesn't mean countries don't have to take actions when they get into wars, though.
On the last point, isn't the cost to utility just the excess utility? The $1 of lost GDP isn't actually a loss to utility, because I kept the $1 instead of getting the first dollar of benefit, and the seller both forewent the dollar of sales but also forewent a similar $1 of operating costs.
No mention of Australia or New Zealand?
Agreed. Extremely stringent but comparatively short lockdowns in these countries were not discussed, but offer a different perspective in that they actively pursued and achieved elimination of the virus and a return to normal social and economic conditions (bar border restrictions). New Zealand’s economy actually grew in 2020.(https://www.bloomberg.com/news/articles/2020-12-16/new-zealand-economy-surges-out-of-recession-amid-spending-spree)
Agreed. Why not look at cases like Vietnam, Taiwan, South Korea, China, Australia, New Zealand etc. where initial strong lockdowns and eradication strategies and then continuing travel restrictions /quarantines brought back normality early on and were then punctuated by localized lockdowns to address outbreaks where they passed the travel barriers? Those are the ultimate examples of lockdowns preventing future lockdowns.
That was mine and many observers’ preferred policy in terms of what to do in spring 2020, and this was followed early on in Denmark for instance, and also in places like Senegal. But then travel opened up without quarantines in the summer of 2020… One can argue this was unreasonable or unrealistic in the US and Europe — but the US still today has a travel ban for Europe since spring/early summer 2020. Just not for other places, due to weird politics.
Why do you only focus on cases where this strategy succeeded? Many other countries pursued similarly aggressive early approaches that later failed miserably, e.g. many South American countries. (That region is also interesting to study because we have Brazil as a natural control.)
Cambodia scraped by due to luck; eventually its luck ran out when cases were imported from neighboring Thailand in February 2021.
Vietnam is a police state, China's lying, and the rest of your list are island nations with secure borders who started with a low infection count. South Korea isn't technically an island but they have the DMZ. All of those countries are unfair comparisons.
Any particular comparison is "unfair". That's the point of a statistical analysis. Are there relevant differences among these countries? Do any of the "unfairnesses" turn out not to matter?
Come on. The antecedent probability that we'd find such a high proportion of islands in a world where the vast majority of countries aren't islands, if being an island was casually unrelated to success in handling the pandemic, is minuscule. And, of course, we have an obvious mechanism that would explain the effect if being an island was in fact the explanation.
Yes, it's clear that being an island in this sense is extremely helpful. Being a police state may or may not be. Even among island nations, we might be interested in what policies (either governmental, or pursued independently by residents) were helpful or not. That's my overall point. We shouldn't just ignore them.
Third graph:
https://hwfo.substack.com/p/does-the-usa-have-the-worst-covid
Would be nice to have a larger dataset that includes places like Indonesia, Philippines, Seychelles, Cyprus, Caribbean nations, etc, to get a better sense of what does and doesn't work for island nations.
Why would China be lying? When there is a COVID outbreak in China, it's very noticeable, and the government takes great efforts to inform the people. China also has numerous neighbors that quarantine Chinese travelers; they would notice if China had a pandemic.
They lie about EVERYTHING, ask anyone who has ever seriously traded markets
China might be lying to some extent (especially about the countryside) but not by three orders of magnitude. It's not North Korea. There are quite a few expats there and they know how to get around the Great Firewall. If the situation in places like Shanghai or Beijing had ever been remotely as bad as in Europe/the Americas/India we'd all know by now.
It is also the optimal policy that is predicted from a wide range of models.
Those are also countries that are widely separated by oceans and where border controls can be effective. Africa has also done relatively well due to its relative lack of transportation connectivity. In some ways China also benefits from the same dynamics due to its suspicion of foreigners. Certainly their coercive quarantine measures left nothing to chance. New Zealand discovered this when some tourists did not follow the rules and caused an outbreak.
Australian states also closed borders to Victoria's during its second wave, which was doable. Granted there are fewer border communities than in the US or Europe, but it's not impossible to close land borders.
(Australian state borders are currently closed to New South Wales - who are having an outbreak which is small by global standards but potentially threatens our COVID-zero status)
The "island" thing sounds like a pretty red herring to me, unless people believe that the virus reached Italy by sneaking through land borders along the Silk Road, or that flying over the Philippine Sea magically heals all passengers on a plane but flying over Kazakhstan doesn't. Western countries got 99 reasons why they couldn't achieve what e.g. Australia could but land borders ain't one.
By July 2020 Italy had managed to bring case numbers far enough down that it could have implemented an actual test-and-trace program without breaking a sweat, but actually it only implemented a perfunctory one because "muh privacy". In addition, it refused to implement border quarantines -- there were superspreader events when people went to dance clubs right after getting off planes from Brazil or Croatia. Sure, a handful of the seeds of the second wave might have sneaked through the land border with Slovenia, but very clearly that wasn't the main issue.
The government can implement restrictions/tests at every airport and port in a way that would be impractical to do at every single road crossing between two schengen countries.
While that is certainly true, and it means that Italy doesn't have a path to get Australia like pandemic numbers, airport and seaport lockdowns could plausibly have meant numbers more in line with the rest of Europe.
Certain Schengen countries had shut down their borders altogether during the first wave, keeping them open with restrictions can't be much harder than that. (Of course there might still be people crossing them illegally, but probably not enough to go from "test-and-trace programs feasible" to "test-and-trace programs unfeasible". Most illegal immigration to Italy comes from the sea, not from land borders, and Australia isn't that much farther away from Indonesia/New Guinea then Italy is from North Africa.)
>there might still be people crossing them illegally, but probably not enough to go from "test-and-trace programs feasible" to "test-and-trace programs unfeasible".
Presumably this depends a lot on the nature of those land borders. Do we have any idea how many people have been illegally crossing between Detroit-Windsor or Tijuana-San Diego or in smaller places like Basel or Konstanz?
> The "island" thing sounds like a pretty red herring to me, unless people believe that the virus reached Italy by sneaking through land borders along the Silk Road
Cases got through to Italy faster through other routes (air) than they would have through land borders. However, this doesn't disprove the argument that *even if* Italy had prohibited entry from abroad as much as feasible, cases would have eventually snuck in through land borders, while the same is not the case for Australia or NZ.
The infection spread Wuhan > Milan > London > NYC before ANYONE knew it existed. There were effectively four virus origins, because those four cities are connected by the fashion industry. Wuhan is where the garments are made.
At the time that Milan was blowing up, we also had blow-ups in Qom/Tehran and Daegu/Seoul, and by the time London and NYC had big blowups we also had Madrid, and could have had Seattle.
Is Wuhan a larger garment manufacturing center than other cities in China, Southeast Asia, or Bangladesh?
Isn't the U.S. getting about 180,000 illegal immigrants every month, which are dispersed around the country immediately? How could any travel restrictions possibly be effective under those circumstances?
This. Anyone who thinks a zero covid strategy could ever have been feasible in the US for this reason is basically either ignorant of the realities or illegal immigration or bonkers.
Or ANYWHERE
Except I also hear the "but those are islands!!!1!" spiel from people in countries with a helluva lot less illegal immigration than the US.
"Lockdown test and trace" only works if you can test every single person who crosses your border. Island nations were the only places in the world that could do "test and trace," and they needed a low infection rate to begin with to pull it off.
Check the third graph here:
https://hwfo.substack.com/p/does-the-usa-have-the-worst-covid
China isn't an island nation; Vietnam is at least attempting test and trace. Thailand's, however, did fail due to the porous land border.
China's numbers can't be trusted, and Vietnam has border security Trump could only dream of and still caught a mid summer infection wave from illegal Chinese immigrants.
This poster is right.
I see this a lot (and have posted about it a lot before on this forum and others), but I think people still don't realize how insanely behind the US New Zealand's COVID was at the time to lockdown. Generally speaking, most countries woke up to COVID around March 12th (Rudy Gobert + Tom Hanks). Assuming a 3 week infection to death period, we want to know deaths as of April 2 to infer infections.
On April 2nd, the US had a total of 8000 deaths, implying almost a million infections, and had 1500 people dying a day (and rising) implying 150k daily infections.
New Zealand had 1 death. New Zealand had 5 confirmed cases on March 12th. So the comparison to New Zealand is basically asking why couldn't the US have locked down 5-6 weeks earlier - that's the only time we had a hope of NZ like outcomes. By March 12th the cat was out of the bag - our outbreak was already worse than it got in any of the Vietnam/Taiwan/South Korea ON the day we locked down.
Now Australia is a bit more of an interesting case, and probably the strongest pro lockdown one out there. But still - they had ~900 people die by the end of their summer wave, with a population of 25 million that means on a per capita basis their entire summer wave resulted in as many cases as the US already had by March 15th. It's just a completely different level of disease.
We didn't have a Summer wave - it is Winter in July in the southern hemisphere.
The July wave was entirely confined to a single city, so dividing by the total population makes no sense. Outside of Melbourne cases were ~0. On a per capita basis the height of the Melbourne outbreak were roughly comparable to the September lull in the US. If you were willing/able to implement right border controls you could have done the same. The US just lacks the state capacity to pursue an elimination strategy.
Sorry yeah used northern hemisphere seasons.
Ok, so Melbourne looks to be ~5 million people and NYC is ~ 8 million? But the greater NYC metro area is ~ 20 million, but a lot of those are in non New York States. So I may not exactly be comparing apples to apples but lets go with this.
Victoria has had 820 COVID deaths, let's just assume those are all in Melbourne.
On April 4th *alone* NYC had 819 COVID deaths. By 3/29 they had more deaths in New York (state, but mostly City at that point) than Melbourne did over the entire Pandemic. Even if we assume the 20 million number and 4x population, we hit that on 4/4.
Again, it was just a completely different level of disease burden in the US before lockdowns started. This matters a lot - if we assume a non-China type lockdown can get R down to 0.6, 12 weeks of lockdown turns a million cases into 2100. Which is still a decent amount of cases! R of 0.7 (an estimate I've seen for US lockdowns) puts you at 13,000 cases - very much not an under-control outbreak! If you only have 50k cases, you get down to 110 cases with a 0.6 lockdown, and 700 with a 0.7.
Yeah, it's pretty obvious when you do the back-of-the-envelope math that if some place in the US like NYC had tried to get to zero cases in the initial lockdown, they would have needed to enact a harsh lockdown for 6 months plus.
Yes, you need to do a Chinese/Australian style lockdown to achieve elimination.
Curfews. Work permits. Mask mandates. All enforced by police/army.
We had all of the above in Melbourne.
Yes, but I'm saying that such will work way less well if your starting point is USA 3/12 disease levels as opposed to Melbourne name literally any date disease levels.
The Atlantic provinces in Canada are another example of this strategy. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
The Atlantic provinces (Newfoundland, Nova Scotia, New Brunswick, Prince Edward Island) in Canada are another example of this strategy, and I haven't ever seen them mentioned. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
The Atlantic provinces in Canada are another example of this strategy, and one that is pretty underreported. They basically became a small, semi-isolated unit within Canada with fairly lax restrictions (a friend of mine in PEI has been playing Magic: the Gathering every week at his local game store for months now) once they got their initial outbreaks under control. When cases started to go up, they did short but fairly harsh "circuit-breaker" lockdowns and intensively used testing and contact tracing. It worked pretty well. The Maritimes IS fairly isolated and small, which works to its advantage, but it's also not very authoritarian, nor did the provincial governments have reputations for being very competent before the pandemic.
Victoria had the longest lockdown on Earth.
Now that Australia is putting people in fucking camps, are you finally willing to admit that it wasn’t a short no big deal lockdown?
Only a short stay in the camps.
Thats what they ALWAYS say...
He did mention both in passing, as examples of how to succeed if you move quickly enough and are capable of enforcing travel restrictions.
"No mention of Australia or New Zealand?"
Yes, 3rd paragraph in Preliminary Theoretical Issues 2
Was added in response to this comment, and a couple of others. I definitely don't think it's comprehensive enough, even with the addition. I just feels like a lot of key data points are getting left out of the analysis (Not just Au/NZ, Taiwan and Korea as well to name a couple more)
Sadly even this brief mention is incorrect. Our second lockdown in Melbourne July 2020 was quite successful and that was 4 months into the pandemic.
Melbourne had a really harsh lockdown for *112 days*, which they did because "test-and-trace" completely failed for them (or they were insufficiently competent to implement it successfully), in order to get cases down from 100 per day (when lockdown started) to zero.
The fact that this happened is actually a really strong argument against "get to zero (or near zero) initially, then control by test-and-trace". If the latter part of the "optimal policy" isn't feasible, this dramatically reduces the payoff to the first half also.
Now, it took 112 days of pretty draconian lockdown in Melbourne to get from ~100 to 0. The initial lockdowns in the US and Australia happened more or less the same date (e.g. looks like March 20 in NYC, March 22 in NSW). How many days should New York have locked down for to get from the ~100k infections per day it was experiencing at the time to zero?
And would getting to zero a) be possible or b) make any sense as a goal in any location with substantial illegal migration? (E.g. the US and probably also Europe)
"make any sense as a goal in any location with substantial illegal migration?"
Probably not. But that's what internal border controls are for.
I don't think Melbourne eliminated COVID in the shortest possible time; China has regularly tested entire cities and eliminated the pandemic in numerous areas that way (e.g., look at its control of the recent Guangzhou outbreak).
We were well above 100 a day in July. But yes you need strong border controls to make it work. Not just at the international border - but internally.
Obviously the US is just not capable of implementing such a policy.
Lockdown was announced on July 7 and appears to have started on July 9. From google (which just pulls the numbers for JHU), the 7dma for Victoria (they don't show Melbourne specifically) was 114 on July 8, and the worst recent day to that point had been 168 four days prior. And obviously the true number of infections was higher, although the positive test rates remained very low (about 0.5%, so hopefully catching the vast majority of positives). So yeah, I was a little under, but not dramatically. What few people seem to remember is that cases then rose for another 4 weeks after the start of lockdown, with the 7dma peaking in early August.
That is in part because the initial lockdown was relatively mild. Full stage 4 lockdown didn't come into place until 1 August.
Australia and New Zealand are islands with moats the size of the Pacific Ocean. Trump could only dream of a wall called the "Great Barrier Reef."
Porous land borders did end up causing Thailand's and Cambodia's current pandemic, but America and Britain never even tried so much as controlling flights. China (which is more ambitious than Thailand and Cambodia) so far has managed to avoid all waves of COVID despite vast land borders.
>despite vast land borders.
Are these the kind of land borders people commute across daily?
Britain relies on a regular flow of lorries in and out, with the same driver throughout their run, to supply it; controlling the border was never feasible.
Australia and New Zealand have been so successful because they trade by boat, and the crew of the boat don't have to have much contact with the people on the dock.
Moats, schmoats, the Arafura Sea is smaller than the Mediterranean and the Torres Strait is almost as narrow as the Strait of Sicily.
Strictly true, but both sides of Torres Strait are incredibly hard to reach by land. The Queensland end is a two-day drive in a winch-and-snorkel equipped 4WD from the nearest meaningful town, and the New Guinea end doesn't seem to be reachable by any sort of road at all.
Interesting to consider how this applies to yearly influenza. All the above things considered, would pro-lockdown people suggest yearly lockdowns until universal flu vaccines? It's unfortunate RNA viruses mutate so damn much.
I have heard a lot of people say that the lesson of this last winter is that we should newly emphasize hand-washing, normalize mask wearing (if sick or just if worried), and similar measures for flu season, given how small the flu season was this year. I haven't heard anyone propose lockdowns though. More in the vein of changing social norms and public awareness.
Not sure if official numbers are out yet, but I heard that the norm is about 20,000-30,000 and this year it was more like 600 - if true, something like a 97-98% drop in deaths. If it had happened without COVID around to rob the headlines and make large numbers of deaths seem trivial by comparison, it would be seen as a massive public health victory.
Actually it raises the question of, if lockdowns don't work then why did that happen? Or is flu different from COVID?
"Or is flu different from COVID?" Qualitatively but not quantitatively. Measures which reduce the R of COVID from 2.5 to 1.25 also reduce the R of influenza from 1.5 to 0.75
Unfortunately, I don't think the R0 of COVID-19 is 2.5 any more; the Delta variant's R0 seems to be at least twice that.
Globally, seasonal flu kills about 500k / year. I don't know about deaths, but flu incidence has gone down 99% globally this past season.
The US typically has 30m symptomatic cases, 440k hospitalizations, and 36k deaths each year. Deaths are hard to estimate last year because they are completely drowned out by un-tested covid deaths (which is in the 100s of thousands?). There is one known US pediatric death from the flu last season, compared to average about 200. Last season there were 1800 positive flu tests in the US, 200 of which were hospitalized (I don't know how many died). https://www.cdc.gov/flu/weekly/index.htm
I'm not exactly sure why false information seems to be actively spread in this thread (not blaming you, I assume some sort of media-induced biased in some of scott's demographics) but as best as we can tell, influenza deaths increased in the US in 2021: https://jamanetwork.com/journals/jama/articlepdf/2778234/jama_ahmad_2021_vp_210048_1620430592.84249.pdf
Number of cases are hard to measure due to the confounders already mentioned, but one can assume not that much fewer given the amount of deaths.
There's nothing in your link indicating an increase in influenza deaths. The official stats listed there are "influenza and pneumonia" deaths, which will almost certainly include some of the untested covid deaths. Covid notably causes pneumonia.
Mortality stats generally uses the category "influenza-like illness" (ILI) because it is hard to reliably distinguish flu from a variety of other respiratory diseases. Generally most ILI is the flu, but obviously one wouldn't assume that is still true for 2020.
If you have any statistics about distinguishing influenza from ILI please do share. I have linked the official CDC data above which has information based on actual flu testing, showing a 99% decline in lab-tested flu.
I just see someone else pointed out that that paper is about deaths in the 2019-2020 flu season, which was significantly worse than usual. The paper you linked says nothing about the 2020-2021 flu season, which reliable data shows was about 99% smaller than usual.
As well as anti-covid measures cutting flu transmission, consider the possibility that a substantial number of the most vulnerable people who in other years might have died of flu might have died of covid first instead.
Well, influenza has an IFR depending on age non-monotonically, with a minimum in your early 20s (IIRC) -- infants have a pretty much negligible chance of dying from COVID but not from the flu. (But in first-world countries other than e.g. Israel or Ireland I guess there aren't anywhere near enough of them to sway the statistics either particular way, anyway.)
I'm sorry to inform you, but you were lied to.
The numbers published by JAMA (as close as you can get to top tier consensus-focused journals) seems to indicate 7.5% increase in influenza deaths in 2020: https://jamanetwork.com/journals/jama/articlepdf/2778234/jama_ahmad_2021_vp_210048_1620430592.84249.pdf
But they lump influenza and pneumonia unassociated with covid-19 together, so somewhat hard to tell.
That paper is only about the US and as far as I can tell about the calendar year 2020. In the Northern Hemisphere most flu deaths are in January and February, so before anywhere outside Hubei implemented COVID-related restrictions (other than bans on direct flights from China).
Flu deaths in the 2020 southern hemisphere winter (July/August) and the 2021 northern hemisphere winter were down by several orders of magnitude from usual, IIRC
I have written an opinion piece advocating for increasing measures against the flu: https://ermsta.com/posts/20210527 "Defeating influenza: lessons from covid"
However I don't advocate for lockdowns, as I believe they are wholly overkill. I believe vaccinations, paid sick leave, contact tracing, border screening, situational masks, and self-administered home-testing are the way to go, in vaguely decreasing order of importance. I put a low emphasis on measures that require high compliance from individuals (physical distancing, hand washing, masks). Also hand washing has little impact on flu transmission.
The ambitious goal for flu is to vaccinate south and east Asia (<5% of India and China get flu vaccine), where flu spreads year round and may be the source of flu elsewhere in the world.
> border screening, contact tracing
Sounds wonderful. I suggest you supplement this with a comprehensive dose of surveillance to catch anyone who is noncompliant. You could setup something like camps for those who do not show enough solidarity to join, or give them a "compliance helper" who watches them to motivate them to do better.
In a last attempt to fight ignorance in this thread.
please consider reading the mortality statistics for the US in 2020 and note that influenza+pneumonia combined (sometimes hard to separate) but as-best-as-possible separated from covid-19 (assume erring on the safe side of attributing less deaths to them) increased in the US in 2020 in spite of measures: https://jamanetwork.com/journals/jama/articlepdf/2778234/jama_ahmad_2021_vp_210048_1620430592.84249.pdf
Currently, I've seen no indication of a significant decrease anywhere else with reliable mortality numbers, though I'm sure it exists, but you'd need to at least tally up well-monitored countries and their number before making this claim.
Given that the US accounts of a large % of the population where COD can be monitored well, I'd say it's very likely your source are wrong.
Before the pandemic, I did used to wonder how many sick days would be saved by people following basic hygiene - e.g. hand-washing, not going to work when ill and potentially contagious.
I've been pretty consistently pro-lockdown but wouldn't support them for flu because the threat of flu is much lower. But there are plenty of other things we should do to reduce flu deaths - masks on public transport (if you're ill but have to travel); work from home (if you can) when potentially contagious; improve ventilation in buildings; get vaccinated.
note, flu deaths increased in the US in 2020 based no the numbers aggregated by JAMA: https://jamanetwork.com/journals/jama/articlepdf/2778234/jama_ahmad_2021_vp_210048_1620430592.84249.pdf
This seems to contradict everything else I've seen on flu last year - and not just in the US. E. g.
https://www.scientificamerican.com/article/flu-has-disappeared-worldwide-during-the-covid-pandemic1/
I'm giving you the mortality data in the US as reported by the most reputable medical organization in the US
If you prefer CNN or scientificamerica's hottakes that's fine.
All I'm really trying to say is that you should consider looking at the actual data.
The data in the Scientific American hot take is from the WHO. The data in the JAMA article is from the CDC and their dedicated flu report seems to think the flu season is milder:
https://www.cdc.gov/flu/weekly/index.htm
The numbers you're talking about are flu and pneumonia combined, and include a number of codes that aren't influenza. Unless I'm misunderstanding the actual data, the "flu and pneumonia" deaths can include lots of non-flu deaths. Given that the CDC and the WHO both produce reports saying that the flu has been minimal this year, I'll go with those.
I don't want to go into how diagnosis on small samples might be wrong, but I'd say it's fairly intuitive to think that they are more prone to manipulation and misinterpretation than country wide mortality, which is the most uncontroversial number available.
Yes, various forms of pneumonia and influenza are combined, in part because differentiating between which of them was causal to the patient's death is hard, but I've got little reason to believe there was a giant spike in covid unrelated pneumonia this year, though it may be the case.
Have you looked at the actual number of people the WHO and CDC are basing their estimstes on ? I.e the people being tested for the presence of various flu strains in the analysis they use.
George's paper is about the calendar year 2020, whose flu wave was almost entirely before COVID restrictions even started.
Probably not lockdowns, since the flu is significantly less dangerous than COVID. But, as others had said, normalizing mask-wearing at the very least does make sense.
note, flu deaths increased in the US in 2020 based no the numbers aggregated by JAMA: https://jamanetwork.com/journals/jama/articlepdf/2778234/jama_ahmad_2021_vp_210048_1620430592.84249.pdf
I think the number 1 problem with that strategy is simply that compliance would be very low. Many people already were dubious of following COVID lockdowns, even when they felt they might be necessary. I think very very few people would be able to be convinced that we suddenly need to be having flu lockdowns when we’ve survived without them basically forever.
It should be noted influenza deaths increased significantly in the US, ~20% accord to the JAMA end-of-year mortality report I read (death being the closest proxy for infection rate in this case, since people aren't neurotic enough to confirm it)
Of course you could come up with a just-so story about how covid-19 and hospital closure somehow causes this (seems unlikely, since there's no correlation with death from influenza after covid, and common sense could well help us infer it to be otherwise, a strong immune reaction remains protective once the pathogen is gone + hospital treatment for influenza is mildly effective to very ineffective to harmful depending on who you ask)
But it still seems like the increased sanitary measures did didly-squat to limit the spread of influenza. Which is rather surprising, but not so much if your working hypothesis is that of a conspiratorial nut (e.g. covid-19 is real but "what mass-media and most review literature portrays to be covid-19" is not)
What is the purpose of performing multiple high and sub-level analyses of changing definitions of what constitutes a case, what constitutes a hospitalization and what constitutes a death? First of all, more testing initially means a spike in cases. There is already a high false positive rate for testing regurgitating dead nucleotides and amplifying it many times just to get a match. Threshold sensitivities were modified for non-vaccinated vs vaccinated patients. Case counts were modified from 'clinical x-rays' to '# of symptoms' to 'if you had a positive test within 28 days, inclusive if you died from a ladder or car accident'. Testing was eliminated for those who were already vaccinated or partially vaccinated. Older patients were forcibly locked up together in elder care homes with positive cases, even though the # of co-morbidities for each case due to death is already high, and has an average age in the 80s... The cofounding incentives of diagnosing patients as positive or getting treatments, getting payoffs either by medical practitioners or hospitals. The ad nausea footage of the Chinese falling ill and locking up of doors with piling dead bodies when the elephant in the room is that countries like Russia/China and some states in the United States have been fully opened for more than one year yet the cognitive dissonance when none of this materializes in the West. The initialization of pandemic by case counts, not by death rates as changed per WHO's guidelines a decade ago. Moving targets, changing definitions, high compliance rates yet no difference whatsoever or strong correlations when compared to different nations that didn't even bother reacting. Flu cases mysteriously disappearing or causes attributable to other causes of death being removed, when people are denied services to hospitals for testing/treatments because of a focus solely on this .. illness. The alignment of recurring seasonality of influenza/flu season with spike in cases, also post-mortem death rates following 'variants' / high vaccination rates. There are a lot of confounding variables that make this whole point moot. Placebo groups are entirely vaccinated.. already or they weren't treated with a negative control but an alternative vaccine for measles. No baseline studies for animal mortality. Multiple stakeholders/shareholders in multiple corporations/media companies/fact checkers/social media platforms incestuously having a strong profit incentive for booster shots or stronger lockdowns.
What exactly is the point you are trying to make here? Because it feels to me like you're trying to imply a conspiracy without outright saying that COVID-19 was a hoax in the US. Replying to all of these claims would be difficult, but there's at least one I know is absolutely wrong.
> Case counts were modified from 'clinical x-rays' to '# of symptoms' to 'if you had a positive test within 28 days, inclusive if you died from a ladder or car accident'... Flu cases mysteriously disappearing or causes attributable to other causes of death being removed
We can be pretty sure that the deaths attributed to COVID-19 were actually caused by COVID-19. The CDC has a graph of excess deaths here (https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm) which shows that for the duration of the pandemic more people than usual died. This number of excess deaths aligns with the number of COVID deaths that hospitals reported, and if anything, indicates that hospitals under-counted. That is to say, unless you can think of some other reason for six hundred thousand people to die in the United States, the only thing they can be attributed to is the one thing that was different in 2020 versus 2017 - COVID.
Likewise, we can be sure that these aren't just flu cases being misreported because that would imply that the Flu was suddenly many times more effective in 2020. This would be especially confusing given that people were self-isolating many times more strictly than before. More importantly, if it WAS some kind of hyper-deadly variant of the normal flu, the reaction by the government and people should have been exactly the same - whatever level of lockdowns they deem appropriate to save the most lives.
2020 was not only different in terms of COVID. You're assuming all the measures didn't themselves create any deaths, which is unreasonable given that a key effect of lockdowns is restricted access to healthcare.
The point the guy is making is that unfortunately this analysis is not very good and actually quite typical of the stuff you find in the epidemiology literature. I normally quite enjoy SA's essays but this one just had me whacking my head against the table quite a few times.
• Taking model simulations as credible despite that those models are consistently askew? Check.
• Assumption that germ theory is complete and the behavior of respiratory viruses is well understood, even though real world data contains many things that it cannot explain? Check.
• Assumption that "COVID deaths" actually means "died of COVID" although it doesn't? Indeed!
Still, at least the evaluation of the Flaxman study is reasonable.
Look at Swedish excess death for 2020. Depending on how you select the base rate you can get anywhere between zero and 9000 extra deaths. That is not a serious number, the absolute level is the same as in 2012. Modelling predicted 90,000+ COVID deaths if they didn't lock down hard. That is not a success for lockdown nor modelling.
I also have a bone to pick with the attempt to separate voluntary behavior from "lockdown". People's behavior changed because they were lied to by modellers and other assorted pseudo-scientists like Flaxman and Ferguson, and that's also what drove formal legalized lockdowns. Those two things can't be separated. The real question is not "do lockdowns work" - they don't and this can be easily seen in the raw data sans models, it does not require complicated analysis (the economist was on the right track!). The real question is what effect reduction in social interactions and mobility had, versus the costs of that, independent of whether it was compelled or not.
Finally, we should also count not just the economic costs but the destruction of the social capital of science. When you see that the totally fraudulent Flaxman study was cited over 1000 times and is just an exercise in self-delusion, it destroys people's confidence in all institutions but especially scientific institutions. That story has been repeated ten thousand times throughout this era and will have large and unclear effects on society going forward. Certainly, society seems to be splitting into people who take academic modelling and government data seriously, vs those who write it off as the work of the systemically biased.
Just to pick on one point:
> Assumption that "COVID deaths" actually means "died of COVID" although it doesn't
I don't think Scott literally assumed this. I think his assumption here is that whatever the discrepancy is between these two numbers, it isn't systematically biased in one direction or another. Is there any reason to think that Sweden systematically over-classified people as "died of covid" while Norway systematically under-classified people as "died of covid"? Unless you've got something like that, the analysis should still stand, with just slightly larger error bars.
Also, the bit about sending deaths back to the level of 2012 - that seems to me like a huge cost, no? Or do you think that a decade of medical progress is negligible?
Firstly, 8 years isn't a decade (2020 vs 2012) and you're assuming that falling mortality rates over time are all to do with medical progress, vs things like healthier lifestyles. I don't think anyone really knows exactly how the progress there breaks down.
You're right that the systematic over-counting would be relatively consistent, however, lockdowns are ultimately justified on *absolute* terms, not relative. Absolute numbers and the exact values on the Y axis are critical. Look at your reply for what can happen otherwise - you're asserting that going back to the 2012 level is a "huge cost" with no idea of the actual absolute numbers involved (they are quite tiny on the scale of a country).
Was anyone in Sweden freaking out in 2012 about the horrible amounts of death that year? No. Nobody noticed nor cared. It was a normal year. This should be a clue that something has gone seriously wrong with our perception of risk and mortality.
You can find a good analysis of all this here:
https://softwaredevelopmentperestroika.wordpress.com/2021/01/15/final-report-on-swedish-mortality-2020-anno-covid/
(note: the comparison of 2020 vs 2012 is one of population adjusted mortality).
I feel like there's a very common reasoning error amongst highly educated people that's been visible throughout COVID times, but also I've seen it in earlier years too. It's something I keep meaning to write about on my own blog but other things always take priority. The error I mean is reasoning exclusively in relative terms, and forgetting to look at absolutes. The most inglorious example of this with COVID is the fixation on R and "exponential" growth. R(t) is just the derivative of the graph of positive test results over some very short time window. It doesn't itself contain any insight or meaning, yet experts routinely say things like "we must lock down because R is above 1" and politicians then parrot them. Lockdowns were originally justified on *absolute* grounds where the absolute value was hospital capacity, because justifying them on relative grounds would have been obviously insane, but the moment they were implemented the activation criteria smoothly shifted to being merely "positive test results are trending up" with the absolute values forgotten.
The fixation on exponential growth is another example of this. Viruses don't grow exponentially, they have S-curve type growth in which their "exponential" phase stops very quickly and then growth starts slowing down again, until they reach a peak and then cases start to fall. And this can all occur very quickly as is well known from outbreaks of other respiratory viruses like flu, common cold, etc. Yet people get stuck on the exponential part and act as if left unchecked COVID cases would constantly double until the entire population is infected. Left unsaid is what would happen after that if the scenario were true: the day before "terminus day" half the population would get infected at once and then infections would suddenly drop to zero (yeah yeah, modulo a final day of mop-up because the population isn't a square number, exponential growth is not always a clean doubling etc - you get what I mean). This is all totally wrong and no disease in history works this way, in fact recognizing that all epidemics are bell curve shaped is the founding observation of the field of epidemiology itself. It means the absolute numbers involved are very important, yet, nobody cares. They're all fixated on relative rates of change.
Outside the world of COVID I've seen this in my own industry (software). It's extremely common for people to make statements like "X cannot scale" when talking about software that solves a problem where the maximum sizes are already well bounded e.g. "number of customers we can reasonably be expected to have in the next 5 years" in a mature bank is not going to be 100,000x higher than it is today, simply because that would require the bank to completely take over all of its competitors and then take over most of another continent, which would never be allowed. Often reasoning in absolute terms as well as relative would let you engineer much better solutions but the last time I did a back of the envelope calculation showing that $SOME_SYSTEM being accused of not being scalable could actually work fine, jaws literally dropped and the customer got all excited, asking that I send him the calculations so he could show other people (they were trivial).
I don't know exactly what leads to this problem but it seems common. I think it's because reasoning in absolute terms ("it's a small problem so what's the big deal") sounds less clever, as if you aren't thinking ahead to what might be and thinking ahead is what clever educated people are expected to do.
There's a lot going on here, and it's very helpful discussion, so thanks for that. I think the discussion of relative vs absolute amounts is important for a lot of this, but I don't agree with what you think are the implications of all this.
> Was anyone in Sweden freaking out in 2012 about the horrible amounts of death that year? No. Nobody noticed nor cared. It was a normal year. This should be a clue that something has gone seriously wrong with our perception of risk and mortality.
I don't think that "freaking out" should be proportional to absolute amount of death. I think "freaking out" should be proportional to absolute amount of *death we can do something about*. No one was freaking out in 2012 because the fact that we were on trend indicated that we were probably doing about as well as we could do without drastic lifestyle changes. But the fact that we had the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we *could* do something about it.
> R(t) is just the derivative of the graph of positive test results over some very short time window. It doesn't itself contain any insight or meaning, yet experts routinely say things like "we must lock down because R is above 1"
> people get stuck on the exponential part and act as if left unchecked COVID cases would constantly double until the entire population is infected.
> recognizing that all epidemics are bell curve shaped is the founding observation of the field of epidemiology itself.
Maybe some people are making the mistake, but the founding observation of epidemiology is that outbreaks grow as a logistic curve, that is roughly exponentially increasing at first, roughly exponentially decreasing at the end, with the scale of its maximal point being given as a simple function of R. R is not just the rate of roughly exponential growth at the beginning, but 1-1/R is the absolute level at which the curve turns around absent any population-level behavioral changes that change R.
R is really important for understanding the absolute number of people that will be infected absent behavioral changes, and not just the relative number over the next few days.
> No one was freaking out in 2012 because the fact that we were on trend indicated that we were probably doing about as well as we could do without drastic lifestyle changes.
It's hard to see why drastic lifestyle changes weren't justifable in 2012 but they were justifiable in 2020, assuming the absolute numbers being discussed really are similar. I expect that's the point the OP was making.
> But the fact that we had the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we *could* do something about it.
We always knew we could do something about it. We always knew that widely deploying masks would cut down yearly influenza deaths. We always knew that vaccination, isolation and quarantine protocols could reduce these yearly deaths even further.
Heck, we even always knew that we could reduce traffic fatalities by improving car safety and lowering speed limits.
I think all of these facts were fairly common knowledge, but we didn't do any of these things, so I don't see how your reply actually addresess the OP's argument.
"the death rate of 8 years ago when we've been on a trend of deaths decreasing by 0.5% per year indicates that this year we could do something about it."
That isn't a logical deduction at all. This is again a very common fallacy seen in the last year - the automatic assumption that the so-called "non pharmaceutical interventions" can affect the spread of a highly infectious respiratory disease. That has never been proven by anyone, it's just a modelling assumption, yet people act as if it's an axiomatic truth and then go ahead and make lots of other inferences on the back of it.
So: transient increases in mortality do not automatically imply that those increases could be eliminated via government action. And the data seems to show that in this case they can't.
Moreover, freaking out due purely to *relative* changes is exactly the fallacy I've just been criticizing. You haven't even put concrete bounds on what level of increase should cause society-wide freaking - just asserted that *any* increase in mortality vs prior baselines should do so. What if the rollback had been to 2015? Lockdowns still worth it then? 2018?
Consider this: if you sum deaths in Sweden from 2017+2018 and then 2019+2020, you get a nearly identical number! (the web page I already linked to shows this graphically). If you are willing to believe 2019 was an unusual outlier rather than the start of a new trend, you could therefore argue COVID didn't affect mortality at all. Now what?
This is why reasoning has to be grounded in absolute values. If it isn't then you can end up with wild over-reactions to noise in the data, which we see all the time now. E.g. newspaper stories claiming cases are "soaring" when looking at the graph shows the graph line barely visible above the the X axis.
"R is not just the rate of roughly exponential growth at the beginning ... but 1-1/R is the absolute level at which the curve turns around absent any population-level behavioral changes that change R."
This statement doesn't make any sense. R is a rate. You can't do an equation on it and get an absolute number of people. Again, you are demonstrating my point: there seems to be great confusion here between absolute numbers (people, cases, hospital beds) and mathematical abstractions.
I like this. It's nice of you to explicitly set out your points, because it makes them easier to reject.
First:
> You're assuming all the measures didn't themselves create any deaths, which is unreasonable given that a key effect of lockdowns is restricted access to healthcare.
It's implausible the lockdowns were directly responsible for the deaths of hundreds of thousands of people. It's doubly implausible that those excess deaths due to lockdown would directly mirror the spread of the infection according to our tests, especially in states where the government response to the pandemic was weaker. The number of confirmed cases goes up, so do the number of deaths two weeks later. I don't have time to investigate the correlation here, but you can see it just by comparing the two graphs visually.
> Taking model simulations as credible despite that those models are consistently askew? Assumption that germ theory is complete and the behavior of respiratory viruses is well understood, even though real world data contains many things that it cannot explain?
The models are constantly askew because they're sensitive to initial conditions, because they can't literally simulate the behavior of millions of humans, and because our understanding of the effectiveness of lockdown procedures is constantly in flux. We knew this at the start of the pandemic, every single credible scientist made a disclaimer on their models stating something to this effect. The reason we still use models despite all of that is because they're good at capturing the exponential growth we see in pandemic scenarios even if the precise details vary.
As for germ theory. What gives you reason to suspect that germ theory in particular is incomplete? And what specific behavior do you think that germ theory failed to model in this scenario? Keep in mind that germ theory is distinct from the SIR model and derivatives that most people use when doing naive, initial predictions.
>When you see that the totally fraudulent Flaxman study was cited over 1000 times and is just an exercise in self-delusion, it destroys people's confidence in all institutions but especially scientific institutions
Ah, yes, I remember the time the replication crisis happened and nobody ever trusted science again. Certainly, this singular highly cited study (which presumably had some fraction of those citations present counterarguments) which was only moderately influential, during a crisis where information was scarce, will irreparably damage the reputation of the institutions involved in publishing it and the enterprise of science as a whole forever.
Epidemiology and macroeconomics have *always* been on the list of the hardest sciences to trust, because they basically don't allow for experiments, and require masses of observational data, and deal with the complexities of large scale human society, in addition to the complexities of the individual interactions involved.
Social psychology used to be seen as somewhat more reliable than these sciences, but not since the replication crisis.
"It's implausible the lockdowns were directly responsible for the deaths of hundreds of thousands of people."
Nobody is saying *all* the deaths were caused by lockdowns, only that *some* were and fallacious logic assuming they cause none at all will lead to a conclusion of undercounting when we know that in reality the classification mechanism used must automatically lead to over-counting.
In the UK when lockdowns started, admissions to ER halved overnight. People were being told to avoid the hospitals. Moreover, elderly people who were already infected were flushed INTO care homes, where they proceeded to infect a lot of other people. All these are part of what I consider to be the consequences of model-driven mass hysteria, which we can usefully refer to as "lockdowns" as we don't have a better word for the whole enchilada right now.
"they can't literally simulate the behavior of millions of humans"
I'll echo WC in saying you should tell Neil Ferguson (nb: not Niall, that's a different academic). Because literally simulating the behavior of millions of humans is exactly what his model tries to do. Go read it - it's an agent based model that literally updates an array of people with their locations and health statuses.
Of course, it "can't" simulate the behaviour of the entire UK population so they just run it on a population slashed by some fraction and claim it's the same. By far not the worst assumption they make.
"every single credible scientist made a disclaimer on their models stating something to this effect"
I think you're agreeing with me then, just that there are no credible scientists. Because for sure if scientists were saying "our models will drastically over-estimate levels of death so please ignore them" then nobody got the memo.
Please. "Scientists" have throughout this entire event consistently claimed not only that models are highly accurate, despite overwhelming evidence to the contrary, but that politicians must take drastic action based on their predictions immediately, without even a few days of delay to let people outside the cosy little academic circle poke at them. The Report 9 from Ferguson et al that directly triggered lockdown in the UK didn't even give any uncertainty bounds.
Re: germ theory. We know germ theory must be incomplete because models that implement it are always wrong by miles regardless of what disease they're trying to model. For one it cannot explain why there's more than one peak - epidemiology has (amazingly) nothing to say on the topic of seasonality, why it happens or how that can be reconciled with the simple models they use. A SIR, SEIR, SEIR+ or whatever model is a direct implementation of germ theory in software. I don't know where you get this idea that they're distinct. SEIR means "susceptible, exposed, infected, recovered" which is the level of understanding of a virus that a child would have. These models are all based on the assumption that the only way people can get infected is by directly coming into contact with someone else who is infected (=exposure), and then rolling a 6 on some many-sided dice.
Yet there's lots of ways the known microbiology of viruses can lead to other views, for example, if infected people create viral aerosols that can hang around inside buildings for long periods then being "exposed" is about your contact with a place, not a person. SEIR models don't handle that at all. It would also imply that being outside is safe which we know from contact tracing studies is true, yet SEIR models have no concept of this. If those aerosols could also be moved around via atmospheric convection it would help explain why yearly influenza outbreaks all start everywhere at once instead of visibly spreading outwards, and why it's possible for (non SARS-CoV-2) coronavirus outbreaks to happen in entirely isolated populations - a topic that has been investigated briefly in the pre-modelling-era epidemiology literature but never solved.
Will Jones has been doing some good explorations of this topic if you want to read about it more.
As for trust in science - the Flaxman study isn't an outlier is it? The >1100 citations it accrued despite being an exercise in circular logic is unsurprising because stories of bogus studies becoming highly cited are a staple of Scott's writing. As for the citations arguing with it - nah. Academics hardly ever do that, especially in epidemiology. Go pick citations from Google Scholar and see for yourself. I just did it, the results are standard. The dominant theme you'll find for this paper and others is that citations virtually never argue with the paper they're citing. Gwern has some stats on this, negative citations are something like <2% of all citations. Real citations are usually almost random. Some of the ones I just spot checked cited Flaxman et al as evidence that "health care workers and seniors are at particularly high risk" (the paper has nothing to do with that), that "aggressive measures appear to be successful" (the paper takes this as axiomatic), that "stay at home orders were widely instituted" (the paper is a very indirect way to show that) and for assuming a serial interval of 6.5 days (the paper takes this as its own assumption).
So actually yes, watching scientists engage in a massive circular back-scratching campaign on the back of a completely fraudulent paper actually will irreparably damage the reputation of the institutions involved. Indeed it already has. Good luck finding average people in the UK who think Prof Ferguson and Imperial College aren't a bunch of illegitimate shysters.
If one good thing is to come from all of this it will be universal distrust of experts bearing mathematical models. It will probably just be the red tribe though.
Why is "universal distrust" a good thing? Shouldn't we have better calibrated moderate trust, rather than just flip from "expert always right" to "always ignore expert" or even "expert always wrong"?
I don't think people do calibrated trust. I also think that laundering conjecture through mathematics is really darned rampant and default to distrust is probably for the best. Note, not, "always distrust expert," rather, "always distrust expert with opinion in the form of mathematical model."
I think the problem is that detecting which models are good vs bad is very difficult, takes a lot of effort by trained people, journals will not propagate the resulting analyses (they end up only on blogs that you have to be in-the-know to find), and finally, my experience has been that the overwhelming majority of such modelling papers are bad. So just defaulting to distrust seems like a good heuristic.
Yes. And that was appropriate before.
Well it will certainly be good for those seeking investment in perpetual motion schemes. No more of that pesky faith in the Second Law of Thermodynamics, which is, after all, merely a mathematical model based on observation -- no underlying theory *at all*.
...although, the more I think about it, the more I agree with you. For example, if there is wider general mistrust that vaccines, seat belts, and household smoke alarms save lives -- all conclusions based on statistics and modeling -- then this will accelerate Darwinian natural selection a bit, and future generations will be a little smarter. I like it!
You read Scott's blog. So you're certainly capable of understanding the difference between empirical observations described mathematically and conjecture stated mathematically. Not that you're going to, of course.
Ah. So there's no empirical data underlying anyone's epidemiological models? An interesting assertion.
Over time, responses to covid will undoubtedly cause some death: e.g., some patients skipped routine cancer screenings in order to social distance, which almost certainly will cause some people to die who would have survived if their cancer had been detected.
But deaths caused by lockdowns and other responses to covid would be more likely to accumulate over time: the ratio of total excess deaths to covid deaths in November-December 2020 would be higher than in March-April 2020 due to the bad effects of nine months of lockdown piling up compared to when lockdown was novel in March-April 2020.
But that's not what we see. Instead, we see total excess deaths being notably higher than official covid deaths in March-April 2020 (probably due to some doctors misclassifying covid deaths as something else before they became familiar with covid) and by November-December, the two figures correspond well.
Yes, the CDC's graph of incremental total deaths is convincing: the U.S. had extraordinarily more deaths in 2020 than would be predicted from deaths per year since WWII. Week-by-week excess total deaths match up extremely well with the usual estimates of covid deaths, and even suggest that covid deaths were undercounted in March and April 2020 and then were pretty accurately counted for the rest of 2020.
Using paragraphs would probably make more people willing to read what you write. Although if it's just conspiracy theorizing it might not be worth reading so maybe you're doing us a favor?
I also bailed on this huge run-on paragraph.
I have no proof whatsoever, but IME, lack of white space strongly correlates with poor arguments and conspiracy theories.
It definitely implies "I'm writing this to scratch a psychological itch" rather than "I'm writing this to change people's minds"
Yeah I think the main thing is that if you care a lot how other people perceive you you're both less likely to rant about conspiracy theories and more likely to type with careful spelling, punctuation, and paragraph breaks.
As a term, "conspiracy theory" is a lot like "criminal": there are certainly some non-central examples that aren't bad but those are rare enough that I don't feel the need to be too careful attaching negative connotations.
But you do have a point. In the future I'll consider trying to use the term "paranoid fantasy" instead, since that captures a little more precisely the meaning I'm trying to convey.
The point is probably that even if absolutely everything you say is true, it wouldn't change the outcome. It's just statistical noise.
I live in Wisconsin, a state that had a judicial fiat against lockdowns for most of the pandemic but is otherwise very similar to its neighbors Minnesota and Michigan, which had fairly significant lockdowns. I've been hollering to the wilderness that the pandemic really didn't seem that much worse, or the economy any better, than these natural comparisons. Nice to see someone do the work to confirm my suspicions.
fwiw, as someone who lived in Wisconsin under voluntary lockdown, I saw very different outcomes (more cases) among my social group in Wisconsin then I did in comparative internet social groups in other states. Many (20%?) of the households on my block were infected, many (most?) of my housemate's coworkers, etc. Compared to my coworkers in NYC, none of whom were infected, the picture is very different. (Although I'll note that my job instituted a WFH policy about 7 to 10 days before the official NYC lockdown, so that probably is a maximally good lockdown case, compared to my housemate's job, which reopened as soon as the state did)
Better HVAC could be a game changer. Maybe someone needs to invent a ceiling fan with a filter which is continuously irradiated by UV light, which draws all the droplets up to the ceiling and neutralizes them. If this already exists, let me know.
That being said. I’m not sure about the whole analysis-by-country thing. I know that is how most of the statistics are framed. But population density plays a role (as well as people per household).
I think there is a “this is bad” threshold (set perhaps by the media) but when northern Italy and NYC had real crisis, I think that influenced people’s voluntary behavior. If there had not been epic level bad, people might have been going out more. Antimaskers never embraced the badness of it all, for whatever reason. In terms of drawing conclusions for future pandemics, “scare them so they obey” May be the moral. If you are right about the role of voluntary restrictions on individual behavior. Which appears to make sense.
There are so many things that need to be accounted for but are very hard to incorporate. One exception, the age demographics are easy to come by, and honestly I wonder why that isn't standard in all these studies. Another example is pre-existing health, both comorbidities everyone agrees about (obesity, diabetes, compromised immune systems) and those that are more controversial (eg vitamin D deficiency). An issue that seems almost criminal to ignore is the quality of healthcare patients received. Granted, it's hard to quantify and incorporate, but I don't think there's a controversy that initial treatment strategies like putting people on ventilators were between useless and counterproductive, and that the quality of care varied by location and improved over time.
You need much longer exposure to UV light than you can get with normally circulating air.[1] Or you have to use far more intense light for the short time each parcel of air is close to the UV source, to the point where the light becomes dangerous to people (at least their vision).
-------------
[1] https://www.ajicjournal.org/article/S0196-6553(20)30756-2/fulltext which suggests about 9 minutes of exposure to UVC light to wipe out SARS-CoV-2.
Hansonian hot take: lockdowns aren't about preventing COVID, they're about preventing dissatisfaction in conscientious distanced citizens, who want official validation that their non-distancing neighbors *should have* stayed indoors as much as them. In other words, staying at home causes lockdowns.
I have pretty much the opposite take: lockdowns cause unrest, which is why China ended theirs quickly (they didn't want people to have nothing to do all day but sit around thinking about how they were unhappy with the government) and blue states ended theirs slowly (they *did* want people to have nothing to do all day but sit around thinking about how they were unhappy with the government).
I think it's a cycle. People get scared and demand the government lockdown then they get bored and get angry so the government relaxes the lockdowns rinse and repeat.
I think that a big thing that causes lockdowns is a prior big wave.
To steal a point from my own comment, in the rest of the country people ask why people in NYC were so insistent on wearing masks for so long, even outdoors where it's extremely unlikely to matter. I think the answer is how bad it got in April 2020. The thing I remember people saying was that the sound of ambulances came from every direction, incessantly, day and night, for weeks on end. The news was talking about how there were emergency tent hospitals in Central Park, how the city had run out of space to store dead bodies. The government essentially told people to stay away from hospitals unless you were having trouble breathing, and even then every hospital in the city had people lining up out the door. The USNS Comfort came to provide relief to the city, and I'm sure it wasn't lost on anyone that the only other time that happened was September of 2001.
The week of April 11, 2020, the number of deaths in the city spiked to more than 7x the normal amount.
> I think the answer is how bad it got in April 2020. The thing I remember people saying was that the sound of ambulances came from every direction, incessantly, day and night, for weeks on end. The news was talking about how there were emergency tent hospitals in Central Park, how the city had run out of space to store dead bodies. The government essentially told people to stay away from hospitals unless you were having trouble breathing, and even then every hospital in the city had people lining up out the door.
Hmm.
I remember this post from March 27: https://elaineou.com/2020/03/27/racist-virus/
> By now you’ve probably seen this NYTimes article describing the horrific state of affairs at a New York City hospital. Patients dropping like flies, nurses reduced to wearing trash bags, refrigerated trucks brought in to house the overflowing dead.
> my brother’s hospital across the city was nothing like that. His hospital admitted many COVID-19 patients, but few have been severe cases and none have died. Today Hillary Clinton (SHE’S TOTALLY RUNNING YOU GUYS) ordered pizza for everyone, and the hospital workers had a pizza party.
I would have to grant that March 27 predates April 11. But I tend to suspect that, given what was described by March 27, the next two weeks weren't much worse.
Were people lining up out the door of every hospital, or were they lining up out the door of some hospitals that made the news?
Don't know about this person's brother's hospital, but according to the CDC in NYC the week ending March 27 there were 2805 deaths in the city and the week ending April 11 there were 7862 deaths. Either way I don't think it changes my overall point.
The Office of the Chief Medical Examiner over on 1st Ave brought back the outdoor white tents for the overflow of bodies. Another sobering sight that was only precedented by 9/11.
The total expected number of deaths in the week ending April 11 in New York City for an average year is 1053. In 2020, it was 7852.
https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm
That's the kind of thing that gets people's attention.
An emergency room doctor in NYC wrote an essay about how the incessant ambulance sirens suddenly died off around April 7, 2020 about 1 pm (IIRC), suggesting it took about 2.5 to 3 weeks for intense behavioral changes due to covid to have an effect on hospitalizations.
None of the emergency tent hospitals were widely used, no? The Javits Center was famously equipped and then never really used, the USNS Comfort was only used for a handful of people.
And morgue space is perennially tight; it’s a semi-regular story in Detroit that morgues cannot contain the amount of homicide victims. This is a statement on the amount of morgue space rather than the number of homicide victims.
I’m not in the corpse business, but I imagine it is optimized for an average amount, such that even a small increase would result overflow. It does seem reasonable that I would want my morgue always at capacity so I wouldn’t plan for space to sit empty waiting for a catastrophe. Moreover, since corpse handling in the US is well-regulated, maybe retrofitting is not practicable?
Kind of but you could also argue that lockdowns prevent free-riding: if conscientious citizens stay home for the purpose of preventing outbreaks but some free-riders ignore it and don't make any sacrifices, they're reaping some benefit (lower R0 and less coronavirus) without making any sacrifices. It's just coercion to prevent defection.
A more generous interpretation of that would be "people get pissed if the government let's their friends and family die unnecessarily"
I'm a little skeptical of assigning some consistent purposeful motivation like this to an action independently taken by at least thousands, probably millions, of individual decision makers. Other than first movers, it probably becomes more about some form of inter-governmental inertia or peer pressure rather than any sort of calculated, considered action with well thought-out reasons. That likely applies to the other side of the coin, too. As soon as a few leaders denounce tyranny and exclaim loudly they're banning lower level governments from shutting anything down, now others that have branded themselves of a similar ideological bent feel like they need to do it too or face a backlash from the half of the 24 hours news cycle their voters pay attention to.
The number of conscientious citizens who honestly care one way or the other is drowned out by the pitchfork mob.
Well, it could also be social signaling. I mean, *almost everybody* doesn't commit murder because they feel it's wrong, and not because of the law against it. That is, if we repealed all laws against murder, I don't think the murder rate would rise by a factor of 100 because everyone would now think it's OK.
So the law against murder is to some extent a signal where we (as a society) tell anyone who is any kind doubt "we all agree this is wrong so DON'T DO IT in case you were at all unsure." In the same sense, you could consider lockdown measures as a version of social signaling, where the majority tells some wavering minority "we think this is much more serious than you do, and we're prepared to enforce that viewpoint on you."
Whether that's a good or bad thing depends not only on whether you think the majority in this case was actually making the correct call (as it does with murder), but also in whether you think the majority should be in the business of enforcing its will on dissenting minorities at all. A lot of room for philosophical disagreement.
A few points, at the risk of repeating myself:
1) "This is an absolutely beautiful graph. It’s showing how lockdown strictness (as of May 5) correlates with death rate over time. We find that early in the epidemic, the stricter your lockdown, the worse you're doing. This is the endogeneity... Later in the epidemic, the stricter your lockdown, the better you're doing - probably because the strict lockdown is giving good results."
While the graph indeed looks quite nice, the meaning is much less clear than implied. Unfortunately (from the perspective of being able to do easy causal inference using the time series variation), consider the following two stylised facts:
i) There is substantial spatial correlation in lockdown intensity - e.g. the hardest lockdowns occurred in the northeast and the Pacific west, the loosest restrictions were in the South and plains states.
ii) The various waves of Covid have exhibited substantial amounts of spatial correlation in outbreak levels. The first wave hit the northeast hardest, the second wave the south, the third wave really kicked off in the plains states, and so forth. It doesn't take that creative a thinker to work out plausible climatic patterns that could give rise to this sort of seasonality.
Accordingly, just how sure are we that the picture of "over time, the lockdown state advantage became stronger" is causal? One easy thought experiment is to extend the time horizon of the graph. And indeed, in the short-lived fourth wave (when b.1.1.7 finally broke through, before the vaccines murdered it), something like 9 of the 10 states with worst outbreaks (as measured by weekly cases per capita at that point in time) were states with stricter than average lockdowns. They also happened to be clustered in the northeast and midwest. Of course, this doesn't mean that the stricter lockdowns caused the fourth wave. But it should reduce our confidence in inferring much from that graph, as well.
2) "There was a significant negative correlation (-0.55) between the lockdown stringency index as of January 1, and the number of post-first-wave cases a state had. This was robust to... using all cases instead of just post-first-wave cases (although some of these changes slightly diminished the magnitude of the effect)."
I find it difficult to justify excluding the first wave. Given how rare reinfections are (call it the YOCO constraint), doing so mechanically advantages places that were hit hard early. Which also turned out to be pretty correlated with lockdown propensity.
As you say, the result gets weaker using all cases. But here *cases* have a glaring problem - tests were much scarcer in the first wave than subsequently. So a higher proportion of infections were missed in the first wave than in subsequent waves. So including the first wave but looking at cases amounts to (relatively speaking) throwing out much, but not all, of the first wave. Since deaths were measured much more consistently (both in level terms, and across time), better to just look at deaths. Which almost certainly will shrink the correlation again. (For example, NY and FL have basically identical cases per capita. NY has dramatically higher deaths per capita, though).
I am bothered that quality of healthcare is not a consideration. Doctors got better at treating Covid over time. And good lord in New York at first the hospitals were, hyperbolically stated, admitting people then sticking them on a ventilator until they died. As you pointed out the disease spread to different regions at different times. By the time patients started going to hospitals in Florida they had learned from New York's mistakes. I'm not claiming I know what specific difference this makes but it really bothers me to see it just kind of ignored.
By the sounds of it, yes, the hospitals in NYC were not adding a lot of value in the Spring 2020 wave. (And perhaps were even subtracting value: "hello, your measured oxygen is low, but you otherwise seem in good health. May we puncture a hole in your throat"). That presumably did raise IFRs early on, and to the extent that actual therapeutics are discovered over time (separately from merely learning to avoid bad treatment), this could cause IFRs to fall over time. Alas, pretty unclear how much effect this actually had. The only therapeutics I can think of, off the top of my head, were dexamethasone and monoclonal antibodies. Just how big the average real world effects of these were, and how frequently they were used, idk. (Also, while we probably would expect IFRs to fall over time for these above reasons, if, for example, a more dangerous strain arose, then a lockdown which delays infections can increase overall deaths. It's messy).
In principal, if lockdowns delay infections and IFRs fall, this is captured in the death counts and this thus gets accounted for as a "success" of lockdowns when doing the calculations. (E.g. Florida and California famously have ended up with pretty similar death rates. But CA had its major outbreak much later... it had much lower deaths than Florida for a while... until the dam broke. To the extent the IFR had fallen more by this time, that gets baked into CAs numbers).
If the concern is "states that got hit early tended to have more stringent lockdowns, but were hit early because of reasons X, and this gave them higher IFRs, and we shouldn't attribute this effect of X to the lockdowns", then there is a way of dealing with that (control for X, and preferably do the analysis at the county level). This is a reasonable point - one I would make myself - but is well outside the scope of Scott's data analysis here.
Big thumbs up. Really hard to untangle all of this.
I am bothered at how COVID Uber Alles quickly became the worlds guiding philosophy
Good points.
Another thing to remember is that different states had different policy decisions. Cuomo in New York decided to protect the hospitals and sacrifice the nursing homes, while De Santis in Florida chose the opposite priorities, which probably reduced Florida's death rate substantially relative to New York's. (Of course, from a Quality Adjusted Life Year's perspective, a focus on nursing homes isn't as effective as it is from a simple death's perspective.)
By the way, I'm not criticizing Cuomo for his making the wrong decision: all this stuff was brand new to the politicians at the time. (But it is kind of weird to win an Emmy for getting a bunch of oldsters killed.)
I fail to see the value in separating voluntary quarantining(VQ) vs. mandatory lockdown(ML). If everyone is already staying home what are the negative effects of ML we should be concerned about?
Additionally, VQ is caused by fear/concern. Accordingly, we could have increased VQ by increasing media coverage of the negative consequences and risks associated with Covid. Would this have been in any way superior? It seems this actually has negative effects even absent positive effects(people staying home) where mandatory lockdowns do not.
I was thinking a similar thing when reading the post. I know the general trend around here is to worry about government intervention qua government intervention, but I really think the stated economic and emotional costs here are very similar whether consider government mandated behaviors or "voluntary" behaviors caused by fear of the virus, so if we're measuring costs along those axis, then movement data is the most important factor to consider, not stated policy aims.
> If everyone is already staying home what are the negative effects of ML we should be concerned about?
That people won't be able to stop staying home when that is what circumstances warrant.
All else aside, voluntary quarantining takes account of varying personal costs.
My personal cost of quarantining is low, so I should do it whether it's required or not, as long as my chance * cost of getting Covid is higher. A counterfactual me who lived alone, found exercise essential to her mental health and had no space at home in which she could exercise (or any of a dozen other reasons getting out of the house might be essential) should start quarantining a lot later and stop a lot earlier than I should (and did). If you leave the gyms open, I won't go, she will (until it gets really bad), and people in between us will decide based on their personal cost/benefit calculations. If you close the gyms, she can't go even if it would be worth it to her, I and anyone like me weren't going anyway so can't benefit, so the only people you're forcing to stay home are the ones it's particularly high cost/low benefit for. (The guy who had it last month, and is naturally immune. The lady controlling her high cholesterol/depression/etc. with exercise. And so on.)
You can completely argue that that's still worth it - that A) those people don't have the incentive to count the cost to everyone else of them getting it appropriately, or B) humans are not actually that rational, and in either case you know better than them. But voluntary quarantining does use people's knowledge of their own cost/benefit, and mandatory lockdowns don't.
It isn't *everybody* who is staying home under Voluntary Quarantine, and it isn't a random 80% or whatever. It's the 80% who are at most risk of COVID *and* suffer the least from being isolated.
For an obvious non-COVID example, I'd estimate that 80% of the US population attends some sort of Christian religious service at least once a year, and does not visit a mosque, synagogue, temple, or whatever. But that voluntary state of affairs is vastly different from "All Americans are required to attend a Christian church at least once a year, and public practice of all other religions is forbidden".
For the COVID example, assume that in some nation 80% of the population lives in healthy multi-person households, 10% lives alone, and 10% lives in abusive households. In a sufficiently severe pandemic, 80% of the population stays home - and 20% *doesn't*, because both solitary confinement and being inescapably trapped in an abusive household both have huge psychological costs. And many of those people will be young and at approximately zero personal risk of dying from COVID.
If the government implements a strict home lockdown, as was apparently done in Spain and Wuhan and mandated but not enforced in California, you don't get to say "well, 80% of the population would have done that anyway, so the costs of the strict lockdown are only 25% higher than the costs of doing nothing and there's not really difference".
Also, some people still hold to the quaint notion that individual liberty is a thing of intrinsic value even if they don't choose to exercise every possible liberty at every possible moment.
I see great value in separating the voluntary from the mandatory, and I see the people who can't see the difference as a far greater long-term risk than COVID-19 ever was.
Staying at home is kind of a bad example, since it is quite unlikely that in jurisdictions where stay at home orders were really enforced (like presumably infamous French nightime curfew) people would not go out more without them. Also stay-at-home orders do not seem to be very effective at slowing covid covid when everything is already closed.
But I agree with your larger point that infringement on freedom caused by fear of covid is just as bad for people's quality of life as infringement on freedom caused by equally strict government restrictions. Relevant metric seems to me how much people alter their behavior because of the pandemic, and it matters little whether this change is caused by fear of infection or fear of getting punished by the government for violating a lockdown order.
I. e. temporary closing of pubs (bars for Americans), if it decreases covid spread, also decreases levels of VQ caused by a fear of covid, which is related (among other things) to levels of covid in the community, and thus it might in the end lead to smaller overall behavioral change caused by covid.
> since it is quite unlikely that in jurisdictions where stay at home orders were really enforced (like presumably infamous French nightime curfew)
Small data point: I live in a big city in France (Lyon), and I've never seen a cop during the lockdowns. My bother was controlled by one, but that's it. That's for a family of 5 people.
Same thing for the curfews, but since everything was closed, there weren't many incentives to go out at night (unless you're doing house parties). However, everyone I've talked to mentioned that the curfews were really exhausting (the first ones started at 6 pm!).
I've been making this point elsewhere, but note that the Swedish data clearly shows that the behavior of Swedes was radically different from the behavior of everyone else. Their deaths curve is broadly similar in shape to every other deaths curve. But their "new cases" curve looks nothing like anyone else's.
We are forced to conclude that either (1) the difference in official policy made a massive difference in overall behavior; or (2) a more restrictive official policy would have made no difference, because the Swedes are the most fractious, official-recommendation-resistant people in Europe, completely different in kind from any other Europeans.
I would suggest that official policy shows up not just in enforcement efforts, but also in the level of popular fear, and the channel from policy -> fear explains why restrictive policies saw high levels of compliance despite low levels of enforcement.
Honestly, if you are referering to spring 2020, I would be very cautious to inferanything about behavior from case numbers, since those are heavily contaminated by differences in testing regimes. Mobility data are much better indicator of behavioral changes.
If anything, once a lot of people are in VQ, ML may well have economic *benefits* as opposed to costs. If a small fraction of shoppers are out shopping, then businesses have to make a difficult decision to figure out whether the marginal cost of sending employees to go in to work will is more or less than the expected returns from selling to that small number of shoppers. (It's surely not enough to cover rent and other fixed costs.) Whereas if you move to a mandatory lockdown, no business will lose extra money in this way, and employees will be able to claim unemployment benefits.
I'm not seeing how this argument doesn't fully generalize to "denying people the ability to make their own choices may have benefits because otherwise they have to make their own choices and they might make bad choices".
When you put it as "may", it absolutely does generalize. In pure decision theory, when it's a person making a decision against an impersonal unknown of the state of the world, there's never advantage for a rational agent to have fewer options. But in game theory, when it's a person making a decision against another person who is making a decision with knowledge of the first person's options, there often can be an advantage to have fewer options. The classic example is that if two people are playing chicken, and both start out with the options "swerve" and "full speed ahead", if one player has the option "swerve" removed (and this is common knowledge), then that player ends up with a much better outcome.
Of course, it depends on *which* options are removed, and what the game scenario is. But this is a standard economic point - naively one thinks that more options are always better, but in situations involving interaction, this just isn't true.
> The classic example is that if two people are playing chicken, and both start out with the options "swerve" and "full speed ahead", if one player has the option "swerve" removed (and this is common knowledge), then that player ends up with a much better outcome.
Except that the obvious implication of this approach is that both players should remove their steering wheel before the match.
I'm not sure why this is an "except". This just shows that we can't simplistically assume that if pruning options for one person improves that person's situation, then pruning everyone's options the same way will improve everyone's situation. There are cases in which it can, say if we're playing a pure coordination game. I didn't mean this to show precisely what our policy should be - I just meant it as a counterexample to the proposed principle that was being used to argue that no limitation of options is ever good.
> If everyone is already staying home what are the negative effects of ML we should be concerned about?
I dunno, maybe blatant violation of our freedom to associate? As but one example, maybe people who live alone shouldn't have to suffer in isolation and should be able to decide a small group of people with whom they can associate without the government fining them for violating the mandatory lockdown.
> people who live alone shouldn't have to suffer in isolation and should be able to decide a small group of people with whom they can associate without the government fining them for violating the mandatory lockdown.
I would be surprised if there is anyone in this discussion who doesn't agree with this claim. I don't believe anyone here is advocating a mandatory lockdown that bans people who live alone from forming pods.
Probably not, but this is a property of mandatory lockdowns nonetheless, and some governments did try to enforce it (we did here in Canada). Official guidelines were that if you lived alone, you could have contact with only one other household.
Meanwhile, here in Washington State, the official guidelines were that people who lived alone couldn't have contact with any other household. Solitary confinement is understood to be torture when imposed on people in prison, so...
Sure, we can debate what sort of theoretical lockdowns might've been a good thing. But in that case, let's be sure to also loudly proclaim that the actual lockdowns as practiced were a horrible thing. In other words, please don't shrug it away when people point out the existing systems inexcusably violated people's rights to no good purpose.
As long as you don't shrug it away when people *also* point out that the existing systems inexcusably encouraged people to spread virus to no good purpose - particularly all the politicians who didn't stop at banning mask mandates, but also made all their public appearances mask-free in order to discourage people from wearing masks.
>I don't believe anyone here is advocating a mandatory lockdown that bans people who live alone from forming pods.
If you advocate lockdowns at all, then you are advocating that - because that's what people with actual legislative and regulatory power are actually going to implement, actually *did* implement, under the name "lockdown", and your advocacy of something you call "lockdown" will be counted as a vote for what they are going to actually do.
Advocating clever nerdy-technocratic policies that are never going to be implemented in the real world is sometimes fun and informative, sometimes a waste of time, but it's dangerously misleading when the thing you're advocating shares a name but not a reality with a thing other people are actually doing.
This is why no one should ever use the word "lockdown". I don't know how many governments used that word. Instead they advocated specific policies. I believe that some European governments banned people from visiting each other, but I don't believe that any US government ever banned individuals from entering the homes of other individuals.
We should be debating specific policies that specific governments introduced, not putting up a strawman of everything that any government ever did against a strawman of doing nothing.
"Most of the debate about whether lockdowns work centers on ideas within the Overton Windows of western countries, after the pandemic had started spreading - ie, given whatever level of lockdown your country had, is the marginal effect of more (or less) lockdown positive or negative?"
I feel this is missing the point. One need not appoint a president for life to implement lockdowns strict enough to actually work, and state of emergency laws are on the books in many Western countries that allow governments to implement lockdowns without violating various constitutions.
I would far rather spend a month or two in a China-style lockdown than a year plus in a Western one (or worse, if the Delta or some other variant spreads enough.) This strikes me as an area where Western governments had an absolutely massive policy failure with millions dead and their Overton window still isn't expanding wide enough to include "actually solving the problem".
It seems that a dictator is neither necessary nor sufficient, as most of actual dictatorships also failed to curb Covid. The real issue is that efficient government is really difficult, especially when a system accustomed to normalcy is suddenly faced with a novel crisis.
Interesting that, on naive estimates at least, the dollar-life or dollar-QALY trade-off rate of Covid restrictions is within a factor of 2 of other such tradeoffs determined by government regulation. I wonder how much of the lockdown debate is just people projecting their political values onto a novel factual question-- i.e. you probably think COVID restrictions were a bad deal if-and-only-if you think EPA restrictions are too.
Editing note…I had to go back to see what a QALY was. It might be good to follow the original mention with the acronym, I would have seen it more quickly, maybe others too. It made it difficult for me to follow the conclusion section at first.
Fixed, thanks.
My apologies, this is my first comment but a few 'What if's' come to mind. What if 85% of the people under 50 had the SARS-Cov-2 virus infection by Nov. 2020..and it was so mild they never noticed? What if the PCR test is the wrong test , has massive false positives, picks random protein and old viruses, is dialed up to a 45 'magnification' instead of 28 as recommended? What if the virus is 1 micron in diameter and the average mask (sieve) is 500 microns? What if the vapor travels not 6 feet but 100 feet? What if staying indoors is exactly against all the viral transmission rules?
What if the 'lockdown success' in the first wave is actually attributable to Sunshine and Fresh Air (aka Spring)?
What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?
What if .. for every 1% increase in unemployment 37,000 Americans die (as one prepandemic study showed)?
What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatoni, Iodine and dozens more EACH reduce the Absolute Risk(AR)
of hospitalization and death by 65-94% ? What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%? What if the Inventor if mRNA technology Dr. Robert Malone warned the FDA about the toxicity of the spike proteins (generated by the vaccine) a year ago and Dr. Byram Bridle found evidence in Pfizers Japan data and exModerna Dr. Luigi Warren says vaxxed people shed spike proteins and all 3 have been variously trashed, deleted, deplatformed and ghosted? What if mRNA Inventor Dr. Robert Malone
and his 16 patents have been scrubbed from Wikipedia? What if this unprecedented Orwellian fact of this pervasuve laundering is because Google and Jeff Skoll invested in GoF research starting in 2010? What if up is down and down is up, Alice - here in Wonderland? Scrubadubdub.
Friedrich Nietzsche said ,
"The strength of a person's spirit would then be measured
by how much 'truth' he could tolerate, or more precisely, to what extent he needs to have it diluted, disguised, sweetened, muted, falsified"
"What if 85% of the people under 50 had the SARS-Cov-2 virus infection by Nov. 2020..and it was so mild they never noticed?"
We know this isn't true because we can test people for antibodies. In the US these studies showed about 10% of people had had COVID in summer 2020, and about 25-30% of people have had it now. Antibody studies from other countries show concordant results.
"What if the 'lockdown success' in the first wave is actually attributable to Sunshine and Fresh Air (aka Spring)?"
This is what I tried to control for by comparing stricter-lockdown states and countries to weaker-lockdown states and countries. Both had spring at the same time. This is also what eg Brauner et al tried to control for by comparing countries that instituted restrictions at different times. I recommend you reread at least Evidence Section 1, maybe the whole piece, it's surprising that you missed this.
"What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?"
All vaccine studies have compared vaccinated to unvaccinated individuals, all of these people are experiencing spring at the same time.
"What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatoni, Iodine and dozens more EACH reduce the Absolute Risk(AR)"
See my piece here https://astralcodexten.substack.com/p/covidvitamin-d-much-more-than-you arguing Vitamin D is probably not very effective. I think the same is true of the other things you mention except corticosteroids, which have too many side effects to be given as a universal preventative.
"What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%"
This is true of everything. The decrease in absolute risk of death from lung cancer from not smoking is much less than 1%, not because smoking doesn't cause lung cancer, but because absolute risk statistics are misleading except in the context of knowing the preliminary absolute risk and thinking about it very clearly. I think this is a pretty serious mistake and I am skeptical you understand absolute risk.
I haven't heard of the specific doctors you mention, but I don't think they're relevant to lockdown statistics.
Spring in the north and spring in the south is different. It is better to compare spring in the northeast to winter in the south in many ways. What time of year are people driven indoors where spread is more likely due to being clustered in a climate controlled environment and vice versa? I find that this blunt attempt to normalize by season to be better than nothing, but pretty close to nothing if you don't understand the underlying mechanics that affect the transmission rate. I don't think they understand these affects enough that this normalization works very well. The regional timing of the outbreaks before there was country wide community dispersion is also very difficult to factor out.
The antibody test strains the meaning of what having Covid means. If a person's innate immune system was sufficient to prevent illness their antibody test will be just as negative as someone who never had the virus in their system.
What makes you say that? As far as I know, the adaptive immune system starts spooling up on the same signal that kicks off the innate immune system, and there are multiple crosslinks between the two. Even if every virus-infected cell gets slaughtered by NK cells within 24 hours, they will have produced cytokines to recruit macrophages and those will undoubtably eat at least a few infected cells and start presenting antigens, and then off we go to the B cell antibody factory. No?
Wouldn't someone have to get re-infected to have enough antibody production going on to test positive?
I don't see how that would happen. If an innate response was sufficient the first time to shut down the infection before antibody production got very far, why would it not work just as well (or better) the second time?
I don't know how probable it is that someone could drive off an infection so fast through an innate response that undetectable levels of antibodies result. That seems to me a priori like a delicate question of whether there is any nontrivial delay between innate attack and antibody production kick-off, and whether there's a mechanism for shutting down antibody production early, and if so when it happens, and whether whatever level is produced is clinically detectable. I don't know nearly enough immunology to even know whether the answers to these questions are known.
But I'm just kind of a priori dubious that the innate immune system could entirely handle an infection without triggering adaptive components *at all*. The two systems are deeply intertwined, and anyway that doesn't seem very efficient and you'd think evolution would've knocked away a barrier to maximum efficiency like that long ago, back when we were coelenterates.
At the very least, this was interesting:
https://www.ncbi.nlm.nih.gov/books/NBK27090/
The mechanism that regulates the production of b-cells that will produce antibodies seems to depend on prolonged presence of the antigen. I-10 - I-13 describe the control mechanism.
This is my second reply to this comment. Found another article more specific to Covid:
https://www.pnas.org/content/118/21/e2101718118
I don't think it's spelled out 100%, but reading the first few sections it sure seems like successful innate immune response does not trigger the specific adaptive response that would lead to a positive antibody test.
Mostly agreed, but you're giving ivermectin too short a shrift. Check out the difference in the most recent COVID peak between Indian provinces that used ivermectin vs those that used remdesivir - it's almost an RCT, and the results are astounding.
Plus the guy who pioneered the use of corticosteroids in the treatment of COVID is all in on it, which isn't much but also isn't nothing.
Here's the thing that really bothers me about ivermectin:
The official recommendation is don't take ivermectin, instead... do absolutely ******* nothing. I have seen many a reasonable argument as to why ivermectin may not be all that useful in treating Covid. But damned if there's any indication trying it is somehow harmful. So it's not harmful, it may or may not be helpful. The official recommendation is do not use this. There's no making sense of that without resorting to some dark motivations or disgusting stupidity.
There are no indications that trying Lambert's Extra-Virgin Snake Oil* is somehow harmful. It sickens me that the medical establishment is just doing nothing.
*(now with 40% more colubrids)
The cool thing about Lambert's Extra-Virgin Snake Oil is that it's 40 year safety record justifies it being sold over the counter in a dozen countries. It also single handedly cured River Blindness and its inventor won a Nobel Prize. https://www.nobelprize.org/prizes/medicine/2015/press-release/
You say - "We know this isn't true because we can test people for antibodies". Antibodies are gone in a matter of weeks. They're kind of a useless measure. Much better T-cell, B-cell - memory cells in bone marrow. Cleveland Clinic and UBC found evidence of almost universal natural antibodies. UBC measured 8,000 proteins in their assay (I'm no virologist but looking for one single spike protein antibody vs 8,000 markers?). Natural immunity is better, broader and dramatically more protective than a narrow spike protein focused immune trigger. Example: SARS-Cov-1 which is what this one was 'made' from, protects against SARS-Cov-1 18 years later. It also works on SARS-CoV-2 and all the Variants and yet it's 20% different genetically than our new coronavirus while the scary variants are just .03% different.
You kinda skittered past these four "What If's"... "What if the PCR test is the wrong test , has massive false positives, picks random protein and old viruses, is dialed up to a 45 'magnification' instead of 28 as recommended? What if the virus is 1 micron in diameter and the average mask (sieve) is 500 microns? What if the vapor travels not 6 feet but 100 feet? What if staying indoors is exactly against all the viral transmission rules?" The reason you would do that is they negate the entire discussion. I'll skip over them too for the sake of brevity (not to concede any one of one iota).
And to counter your prior Vitamin D article, you didn't mention Vitamin D is related to 2,000 genes 300 of which are immune system related. SARS-Cov-2 is a wimpy virus just like HIV...unless it gets past your immune system and they wreaks health havoc.
Vitamin D deficiency increases Hypertension
Vitamin D deficiency increases Diabetes
Vitamin D deficiency concurrent with Obesity
Vitamin D deficiency occurs with Heart Disease-Stroke
Vitamin D deficiency increases Cancer rates
(Imost common comorbidities of COVID-19.
See also POC (people of colour) 2.5-3X rates)
Whatever the country search their Vitamin Status.
Then check their COVID status on https://www.worldometers.info/coronavirus/
Some countries had different D status within their borders.
There are 702 papers since Jan 2020 on Vit D and COVID-19
with health care people, researchers, clinicians, biologists
hemists - just every corner of science screaming uncharacteristically,
"Take Action!" Read 'em.
https://bit.ly/CovVitaminD
Off on a vitamin D tangent. It's just these idiots have been causing the Vitamin D deficiency with scary health public service announcements for decades. Because of the tangent I gotta stop here but (see link later... not see Linkletter, that would be Art Linkletter and his show "Kids Say the Darndest Things" prolly before you born kid..in the 60s when we were building great immune systems under the sun)
I said -
"What if the 'vaccine success' is again vastly attributable to Sunshine and Fresh Air ( and the CDC dialing down the PCR test at vaccine launch)?"
osaid -
"All vaccine studies have compared vaccinated to unvaccinated individuals, all of these people are experiencing spring at the same time".
By that I mean you completely missed what I said which was the 24 drumbeat from the Miracle Vaccine Train networks is "Look! The Vaccines are Working!" They tout the plummeting infections to the vaccine. Rubbish. People going outside .. getting away from the viruses trapped in close quarters with people and getting fresh air (studies show zippo worries). 'Can you say Black Lives Matter'? 26 million Americans mostly mask-less AND in close proximity... Why didn't Everyone in America die? WTheck ? (I toned that down). We had the most important lessons of the pandemic to learn and we learned none of them?
I said -
"What if the natural, cheap and off-patent solutions, Sunshine (Vit D), Zinc, Quercetin, Ivermectin, Corticosteroids, Melatonin, Iodine and dozens more EACH reduce the Absolute Risk(AR)"
You poo-pooed it and took the Pharma talking points verbatim on ... forget it.
Read up - there are many dozens of cheap and effective solutions.
WHO is trying to make Ivermectin "off limits" because they are concerned about parasitic infections. This, a drug available OTC everywhere costing .003 cent a pill with a 40 yr track record - zero deaths and won the Nobel Prize for it's inventor.
https://c19early.com/
I said -
"What if people think that vaccines 'effectiveness' is 95% when this citation is Relative Risk (RR) and vaccines Absolute Risk reduction hovers just over 1%"
You said -
"This is true of everything. The decrease in absolute risk of death from lung cancer from not smoking is much less than 1%, not because smoking doesn't cause lung cancer, but because absolute risk statistics are misleading except in the context of knowing the preliminary absolute risk and thinking about it very clearly."
This is gobbledygook, start to finish. You compare Smokers with Smokers - a perfectly imperfect analogy.
We are comparing an experimental gene therapy to a placebo. People who smoke cigarettes are up to 3,000%* more likely to die from lung cancer than people who do not smoke. (* by that's just the CDC and we know they have licensed a lung cancer vaccine as 45% of their revenue comes from vaccines. Ironically, LESS likely to die from COVID-19 if you are a smoker and 45% less likely to get Parkinson's btw)
Just illustrative -
Relative versus absolute risk of comorbidities in patients with psoriasis
https://pubmed.ncbi.nlm.nih.gov/27986396/
Conclusions: Presenting attributable risk in the form of the number needed to harm provides a clearer picture of the magnitude of risk and a basis for wiser medical decision making and patient education.
(Right, as everyone would agree)
I admit to not communicating Relative versus Absolute Risk well at all.
And also of making the mistake of thinking You Knew This - and you clearly did not.
In fact, you double down on the very problem. Whew.
So I hand the ball off to ...
Peter Doshi of the BMJ board he writes -
"Only full transparency and rigorous scrutiny of the data will allow for informed decision making, argues Peter Doshi
In the United States, all eyes are on Pfizer and Moderna. The topline efficacy results from their experimental covid-19 vaccine trials are astounding at first glance. Pfizer says it recorded 170 covid-19 cases (in 44,000 volunteers), with a remarkable split: 162 in the placebo group versus 8 in the vaccine group. Meanwhile Moderna says 95 of 30,000 volunteers in its ongoing trial got covid-19: 90 on placebo versus 5 receiving the vaccine, leading both companies to claim around 95% efficacy.
Let’s put this in perspective. First, a relative risk reduction is being reported, not absolute risk reduction, which appears to be less than 1%. Second, these results refer to the trials’ primary endpoint of covid-19 of essentially any severity, and importantly not the vaccine’s ability to save lives, nor the ability to prevent infection, nor the efficacy in important subgroups (e.g. frail elderly). Those still remain unknown. Third, these results reflect a time point relatively soon after vaccination, and we know nothing about vaccine performance at 3, 6, or 12 months, so cannot compare these efficacy numbers against other vaccines like influenza vaccines (which are judged over a season). Fourth, children, adolescents, and immunocompromised individuals were largely excluded from the trials, so we still lack any data on these important populations."
Peter Doshi gets complicated, what he's saying is 'It's a load of crap".
And for my part, I'm seeking a better way to convey this horrible statistical con.
And really I have to say to you "I think this is a pretty serious mistake and I am skeptical you understand absolute risk".
You say -
I haven't heard of the specific doctors you mention, but I don't think they're relevant to lockdown statistics.
I say -
They couldn't possibly be more relevant to Lockdowns.
If you are using faulty PCR tests, faulty infection rates, skewed hospitalization and death rates, AR\RR, social distances, vapor distances, efficacy of mask and on and on - you would never know and yet one can weave a story with faux precision.
These doctors are the glaring evidence that you and the people are not getting the full picture from CNN (who's ad revenue is 44% Pharma) of the NYTimes (owned Blackrock-Vanguard own 16,000 companies including Pharma)a nd FAANG and really everyone with a life & death vested interest. It's historic, unprecedented and permanent. Lockdowns are lunacy.
I though for sure someone, somewhere would see this Wikipedia wipe as the new Orwellian age we are in. The relentless layering of lies means the Truth we seek with aid of science and reason can never achieved. (So, I guess this is my last post. I envisioned this as something an outpost of reasonable truth-seekers) Take care and peace.
I live in New Zealand, and I feel this post is missing something without talking about us (or other countries that have pursued similar "Zero Covid" policies, like Australia).
With strict border controls and a strict lockdown (and probably significant assistance from our geography), we managed to eliminate Covid-19 in June 2020 and have mostly kept it eliminated since, with mandatory isolation of all arrivals, and a few short lockdowns when Covid managed to leak into the community. Australia's biggest wave (Victoria mid-2020) was eliminated after months of harsh lockdowns, and they've since had similar policies and outcomes. Our economies have also held up fairly well, helped by the ability to maintain minimal internal restrictions whenever Covid had been elmiminated.
I think there are good reasons (mainly geography) to think our experiences can't be just copy-and-pasted to the rest of the world. But I think the lesson specific to lockdowns is that there's a discontinuity when elimination is a plausible option. If e.g. Colorado had decided it was going to institute a seven-week-long harsh lockdown to try to eliminate Covid-19, it'd be at a great cost, but because Covid would be re-imported from another state so easily, the effect on Covid-19 deaths might not be that larger over the course of the pandemic. But in NZ's case, we were in a position where we could eliminate Covid-19 and remain that way.
That also changes the calculus on short "circuit-breaker lockdowns" whenever cases emerge in the community. NZ or Australian states locking down for 1 or 2 cases in a city may seem crazy to outsiders (and I do think there have a few overkill examples of this (looking at you, Western Australia)). But these can in fact have some of the highest benefit-to-cost ratios of all lockdowns. The reason is that they can be the difference between returning to a low restriction elimination state, or a phase transition to an endemic Covid state.
As such, I have a theory that optimal lockdown strategy may be bimodal. Lockdowns make little sense with zero cases, can be very good with a few cases (and strong border controls), may not make as much sense with a moderate amount of Covid, but may again start to make more sense with a healthcare-system threatening wave.
I feel like I covered this in the first paragraph of the second Preliminary Theoretical Issues: "If, the moment COVID had been reported in Wuhan, other countries had closed their borders tightly, that would have prevented the pandemic (at least for a while). In that sense, lockdowns definitely could have worked."
I'll add something about NZ and Australia in particular in there.
I don't quite agree with this new segment. Our policies weren't significant outliers in the world (i.e. NZ entered lockdown at the same time as the UK did). We did benefit a lot from having much less early undetected spread (probably helped by geographic isolation, and maybe the season). Other countries (like Peru) that tried this strategy failed badly.
Test-and-trace has helped but is more applicable to the e.g. South Korean containment model. The main strategy in both Australia and NZ is to keep Covid out entirely via border controls, and whenever that fails, often locking down as well as using contact tracing.
Have you read about the second wave in Victoria, Australia? That got pretty out of hand, but was still squashed when the government gave up on the softly-softly approach and locked down again
I'm going to have to agree with Honey here, that the second segment doesn't quite cover it, for the reasons they mentioned.
I would also like to bring a new point about New Zealand that I haven't seen raised when people talk about it in these parts.
Our wage subsidy during lockdown (although it obviously had some issues), was on the whole, very well done. I get the sense that people had to worry about the tradeoff between being able to buy food, make rent, ect in a lot of other places, if there was a lockdown, or they wished to remain indoors voluntarily.
I haven't really seen this addressed in a serious way, but surely this is one of the major factors, in compliance, voluntarily or otherwise.
For what it's worth, the U.S. had the most financial assistance of any country. Unemployed minimum wage workers were getting 2.5x their normal wage.
I don't think your comment is correct. Melbourne had a decent amount of cases as late as July 2020. We eliminated with strict border controls, hard lockdowns and tbh very little test and trace.
East Asian countries should also be looked at, and treated separately from Australia and NZ. I don't think they have had as strict entry restrictions as AU/NZ, and they didn't eliminate COVID as successfully, nevertheless they maintained much lower rates than Western countries.
That made it sound as it if were a hypothetical scenario that never happened. And it wasn't even what happened in NZ/Australia. We did shut off to China by early February, and the rest of the world by mid-March, but it wasn't enough to stop enough cases being imported that would have likely lead to a large wave absent lockdowns.
And as for the second paragraph in that section: our lockdowns, while strict, weren't the "get shot for going outside" hypothetical you described. And even then, a lot of countries did try something really strict and still failed quite badly. (I like to point out Peru, which instituted a very hard and early lockdown, and still has one of the worst Covid death rates in the world). Lockdown-to-eliminate requires a number of other features, like sufficiently high compliance, financial support, and the ability to control new imported cases.
This assumes no aerosol transmission via atmospheric convection, which is a theory that has been taken seriously and investigated in pre-COVID times to explain seasonal influenza (look up "aerobiology").
Although I admire the effort that went into this analysis, all such analyses seem t over-estimate the robustness of both data and theory. Simple applications of germ theory give us epidemiological models which are to a first approximation always wrong by miles. Conclusion: germ theory is incomplete. Thus arguments like "lockdowns obviously work because theory says they must" are over-reaching. Will Jones has been doing some great research over at lockdownsceptics.org where he explores possibilities for what's going wrong there, and proposing some very logical explanations for the sudden start/stop nature of COVID outbreaks based on aerosol load and heterogenous susceptibility. No surprise that this work is being done by people outside the epidemiological establishment when you look at the Flaxman study
BTW re: Flaxman, the last-intervention-works issue wasn't a bug, the model was specifically designed that way. If you want to see buggy models look at Flaxman's colleague Ferguson and his COVID-Sim model - written in C! - and filled with memory corruption errors, race conditions and other genuine bugs. Flaxman's work is in some sense more sinister because they constructed a non-buggy model specifically to reach the conclusions they wanted to reach, using many subtle techniques to do so. Nic Lewis has got a good writeup of how they hid their subjective Bayesian priors in data tables on GitHub to hide that Sweden was being assigned a prior "country specific factor" wildly out of line with everywhere else, for instance.
Posted above but I'll mention here as well...
I see this a lot (and have posted about it a lot before on this forum and others), but I think people still don't realize how insanely behind the US New Zealand's COVID was at the time to lockdown. Generally speaking, most countries woke up to COVID around March 12th (Rudy Gobert + Tom Hanks). Assuming a 3 week infection to death period, we want to know deaths as of April 2 to infer infections.
On April 2nd, the US had a total of 8000 deaths, implying almost a million infections, and had 1500 people dying a day (and rising) implying 150k daily infections.
New Zealand had 1 death. New Zealand had 5 confirmed cases on March 12th. So the comparison to New Zealand is basically asking why couldn't the US have locked down 5-6 weeks earlier - that's the only time we had a hope of NZ like outcomes. By March 12th the cat was out of the bag - our outbreak was already worse than it got in any of the Vietnam/Taiwan/South Korea ON the day we locked down.
Now Australia is a bit more of an interesting case, and probably the strongest pro lockdown one out there. But still - they had ~900 people die by the end of their summer wave, with a population of 25 million that means on a per capita basis their entire summer wave resulted in as many cases as the US already had by March 15th. It's just a completely different level of disease.
Re your final paragraph: substitute 'winter' for 'summer', but more importantly, note that that second wave happened in a single state (Victoria, population 6.7 million) and mostly in metropolitan Melbourne (population 5 million). So when confirmed new cases were running at several hundred per day (peaking above 700 a couple of times), that was a pretty significant outbreak.
This is what I getting for posting twice haha. Covered this in a followup:
"Sorry yeah used northern hemisphere seasons.
Ok, so Melbourne looks to be ~5 million people and NYC is ~ 8 million? But the greater NYC metro area is ~ 20 million, but a lot of those are in non New York States. So I may not exactly be comparing apples to apples but lets go with this.
Victoria has had 820 COVID deaths, let's just assume those are all in Melbourne.
On April 4th *alone* NYC had 819 COVID deaths. By 3/29 they had more deaths in New York (state, but mostly City at that point) than Melbourne did over the entire Pandemic. Even if we assume the 20 million number and 4x population, we hit that on 4/4.
Again, it was just a completely different level of disease burden in the US before lockdowns started. This matters a lot - if we assume a non-China type lockdown can get R down to 0.6, 12 weeks of lockdown turns a million cases into 2100. Which is still a decent amount of cases! R of 0.7 (an estimate I've seen for US lockdowns) puts you at 13,000 cases - very much not an under-control outbreak! If you only have 50k cases, you get down to 110 cases with a 0.6 lockdown, and 700 with a 0.7."
I've long thought there are way too many confounding variables to really be able to draw any strong conclusions about this sort of thing. Just looking at how wildly different the outcomes are in different countries, or in states that have similarly strong/weak lockdowns.
We have heard, for example, that the weather affects transmission. People who argued this accurately predicted changes in case numbers in many cases, most notably the big winter wave in the US and Europe, predicted months in advance. So ... how the hell do you account for that?
Given all that, I am inclined to just evaluate what policies work the best given our understanding of how the virus works and of human behavior. I know we all like this sort of empirical analysis, scientific method and all that, but on some level if it's this hard to measure you just end up measuring your own blind spots. "We can figure this out using empirical analysis" reminds me of the High Modernism thing - if you can't really figure out a good way to approach it scientifically, just kinda do something science-y.
A few other points:
<ul>
<li>An underrated part of lockdown politics is really about the financial assistance to laid off workers. For many people early on the understanding was that lockdowns lead to people losing jobs and so political support for unemployment benefits, etc. And that ending lockdowns would mean an earlier end to those benefits, so whatever people thought of the money stuff drove views on lockdowns.</li>
<li>I wonder how much the <i>public messaging</i> matters. In the US, it wasn't a particularly right-wing thing to be unconcerned about the virus, or left-wing to be concerned, at first. It only started that way after Trump and Fox News went around talking about how it's all overblown and how "this" is "the Democrats' new hoax" and all that. You mentioned how actual behavior is often loosely connected with government rules - it seems that it is more strongly correlated with the partisan messaging.</li>
<li>I think we got locked into a not-very-helpful pro- or anti-lockdown debate early on that prevented any discussion of <i>what are the best lockdown policies</i>. Early on there was discussion of centralized quarantine - people said NYC was hard hit because of overcrowding in small apartments, where a sick person would stay at home and infect everyone else - and widespread testing and tracing. None of which ever happened. At the same time, as far as I can tell "six feet apart" has been known for awhile to not have any basis in fact, and people still follow it. And restrictions on outside activity lasted a long time after it was clear that it was a way lower risk.</li>
<li>I think the medical establishment did a bad job of communicating at many times. Big examples include the initial anti-masking message, and the failure to give more realistic guidelines that people who don't follow the strictest guidelines might follow. To steal a point from someone, all non-COVID CDC advice is <i>also</i> super-strict rules that nobody follows (IIRC they recommend using a dental dam during cunnilingus. And 4 drinks in a day for a woman is "binge drinking").
And some political bullshit like saying it was OK to go to racial justice protests because it's a more important issue than COVID.</li>
<li>That said I'd be remiss if I didn't add that the medical establishment's bad job pales in comparison to the aggressively terrible job from the administration. People forget it now but before Fox News was against vaccines or lockdowns or masks, Trump was even against testing people, on the basis that it would increase the number of cases and make him look bad. He said this out loud on TV repeatedly! The main guiding principles seemed to be the stock market, and a mystical belief that nothing could stop the "Trump train". And people on his Coronavirus "task force" have alleged that they slow-walked the response because it was seen as a blue state problem.
All the arguments about calling it the "China Virus" or whatever came about because they needed new talking points to replace the "it's a media hoax" talking points once it became clear that it was, in fact, a big deal.</li>
<li>I think it's clear that earlier waves cause lockdowns (and voluntary lockdown-ish behavior). People in other parts of the country ask why everyone in NYC wore masks so long - probably because how bad it got in April of 2020.</li>
</ul>
+1
Lots of really important points here. Especially that first one, where one main point of a lockdown is that it enables economic relief policies, and so may well be part of an economic *positive* rather than a negative. I also have a quibble about the messaging around protests (mainly because protests were outdoor activities that turned out to have overly harsh messaging elsewhere), but otherwise largely think these are the right points to be considering.
Maybe Scott can fix the HTML formatting that didn't work?
I think the point about the protests was that at a time when people were not allowed to have outdoor weddings, funerals, or even protests about other things, the CDC and prominent scientists gave a free pass to a certain subset of protests. There were some anti-lockdown protests that had been heavily criticized by both government and media, but BLM-related protests were often neither criticized nor discouraged. Then Democratic governors (Tom Wolf did this prominently, but there were others) not only allowed the protests, but they actually joined and/or supported them. See this link: https://www.pennlive.com/coronavirus/2020/06/why-did-pa-gov-tom-wolf-join-a-protest-with-thousands-of-people-breaking-the-rules-in-a-yellow-phase-county.html
If the CDC can say that racism is a more pressing concern than COVID, then a lot of people are going to argue that [pet issue they are concerned about] is also more important than COVID. By telling them no, it creates the impression that the CDC is being political and also not scientific in their advice. It was a very substantial misstep and I believe one of the primary causes of a lot of conservatives/the right starting to reject lockdowns and restrictions.
Wait, did the *CDC* ever endorse these protests?! I thought it was just epidemiology Twitter that did that, but I admit I wasn't paying close attention.
(And my recollection of the anti-lockdown protests was that the most hyped ones occurred indoors without masks.)
There were a lot of health officials that did. I thought I remembered the CDC also saying some specifically positive things about the BLM protests, but they did refrain from condemning them. A former CDC director also came out in favor of the protests, which may be why I remember it that way.
https://www.politico.com/news/magazine/2020/06/04/public-health-protests-301534
https://www.axios.com/black-lives-matter-protests-coronavirus-science-15acc619-633d-47c2-9c76-df91f826a73c.html
The anti-lockdown protest that I remember most was outside, and involved a lot of people getting haircuts on the front lawn of government buildings.
https://en.wikipedia.org/wiki/COVID-19_anti-lockdown_protests_in_the_United_States#Michigan
Ah, I remembered most the April 30 protest, when armed demonstrators filled the capitol building - it looks like you're considering the May 15 one.
The conservative rejection of lockdowns etc predates the George Floyd protests. IMO it really started because conservative pundits followed Trump on the "Coronavirus is no big deal" track off a cliff, and then needed to justify their previous statements about how everything was an overreaction. Remember Richard Epstein saying there would be no more than 500 deaths in the US from COVID or Rush Limbaugh saying it's no worse than the common cold (not the flu, a cold)?
That said I agree that the "protests are OK when it's something I agree with" attitude was a very damaging one, and only seems like a relatively small deal compared to everything else that happened.
There were certainly a lot of conservatives already bothered by the restrictions. The number who were vocal about it increased significantly after many public health officials either endorsed the BLM protests or refrained from condemning them, despite condemning other protests and non-protest gatherings of all kinds.
Protests ended up as an accidental evidence that outdoor spread is basically not a thing, BUT their real and imho substantial costs were that they damaged already low credibility of public health professionals, many of whom endorsed protests on the grounds that "white supremacy is worse than covid", among Republicans.
"Sweden had closed some public schools (other countries had closed all of them), restricted large gatherings (other countries had restricted all gatherings), “recommended” closing businesses and staying at home(other countries had mandated it), and closed public events and public transport. "
Actually public transport in Sweden never closed. At some point in late 2020 it was recommended that commuters wear a mask when using public transport during peak hours (which, anecdotally, not many did).
Huh, I think the Stringency Index data said it did, but I've removed that mention until I can confirm.
Strange. This is the official guideline from the Swedish authorities https://www.krisinformation.se/en/hazards-and-risks/disasters-and-incidents/2020/official-information-on-the-new-coronavirus/domestic-travel-and-public-transport.
I doubt that public transport completely closed in any European country. There is probably a mistake somewhere.
FWIW assuming there's only one of these I've seen people voice concerns about the stringency index before; accusations that data was retroactively re-classified to make it seem like low-lockdown areas had stronger lockdowns than they really did.
Also the minor practical point: While Sweden never officially "Closed their borders" both Norway and Denmark Did close theirs, rendering it moot.
Thanks for the link about Belgium's death rate! I was very curious about that.
I hate to add yet another "theoretical concern" to your list, but shouldn't voluntary changes also effect the economic impact of any policy? If most people stop going out to eat voluntarily, then the associated reduction in R *and* the associated restaurants going out of business have little or nothing to do with the government policy.
For secondary effects, I think you missed a few that are potentially much larger than the ones listed. We had a lot fewer cars on the road, resulting in thousands of additional traffic deaths (yes, more, because the people left were driving faster--see https://www.nsc.org/newsroom/motor-vehicle-deaths-2020-estimated-to-be-highest); on the flip side, air quality may have improved substantially (https://www.cnn.com/2021/03/16/health/world-air-quality-report-intl-hnk-scn/index.html) and air pollution is actually a substantial health hazard, particularly in poor countries. Then there's the culture war and political effects: much-discussed increased politicization of science and policy, Trump would likely have coasted on incumbency effects and 4% unemployment to re-election, arguably growing inequality and the strength of stocks/big companies compared to small businesses and average peoples' salaries.
"We know this is possible in principle - some states tried things like closing parks and trails, which in retrospect probably wasn’t too useful since the virus doesn’t spread well outside."
Indeed, this particular intervention could even have been counterproductive. I saw an argument (probably on the SSC or Motte subreddits) that lockdowns past a certain point *increase* cases, because everyone is jammed together into the few places they can go, like grocery stores. I wasn't sure of that, but shutting down an outdoor space or preventing all but the largest and most densely-packed gatherings is likely to have 0 positive impact, and secondary effects like pushing people indoors or reducing physical activity very likely make it a net negative. I know that some outdoor areas like major national parks were and ski resorts (and still are) overwhelmed and probably couldn't have maintained standard services like ranger support without compromising employee safety, but local parks and most trails should have been kept open
"Maybe this was a dress rehearsal for a much worse pandemic later on, and the most important effect of our choices now will be setting the defaults and expectations for how we respond to that one."
And that's probably the most terrifying thing. There is no indication that the powers-that-be have realized any of their mistakes. The FDA is still going to ban tests early on. The CDC is going to be wrong about risk. The intelligentsia will sneer at the plebs for being scared, then sneer at them for being selfish, without a hint of self-reflection.
Talking about the US and comparing different states. My prior would be that red states would tend to lower the number of COVID deaths in the statistics, compared to blue states. Shouldn't we account for that? Also most of the big high density cities (aka places where virus spread should be the highest) are blue even in the red states. And they could have stricter local policies compared to the state.
On the messing with statistics point, I know it feels like something that might happen, but I haven't heard any convincing story as to *how* it could happen. Does the governor direct county coroners to say something different on death certificates? Do county coroners have a bias in how they fill out those certificates that doesn't show up in other years?
On the localities with stricter policies than the state, there were periods last summer where that was true, but as someone who lives in Texas, I remember much of June 2020 and March 2021 being spent with local authorities going to every judge and court they could to try to force the governor to allow them to enforce policies stricter than the state.
My prior would be that red states and blue states would tend to lower or increase the number of COVID deaths following restrictions being implemented or lifted in a way that made the decision look correct. An example would be that Florida might start lowering the death count after fully opening up, while NY might start lowering the death count after locking down. I have no evidence that any state did anything like this.
That's the sort of thing that they would do, if the elected officials were able to just plain falsify the statistics. But I think that in any contemporary rich country, one can't just completely falsify the statistics - at most, one could record deaths early or late, but one can't either delete deaths entirely (someone will notice the missing record at some point) or create deaths of people that don't exist. So I think one would test this by seeing if there are statistically significant jumps or drops in death rate just before some transition that go in the opposite direction.
Kenny, I think I agree with you. I guess I would clarify by saying that if I knew statistics *could be* manipulated appreciably in one direction, my prior in that instance would be that such manipulations would occur in whatever direction made the manipulators' policies look better, i.e. not just down for Red, up for Blue.
I have no opinion on whether this ever happened, but the more likely scenario is that the criteria for "this is a COVID death" would be manipulated, rather than outright fabrication of the statistics. If a local coroner says that the person died of heart disease, despite their COVID infection, but another local coroner would have ruled that a COVID death, that could play out over different localities and skew the numbers.
Again, I have no idea if this happened. It's been something I have been curious about since early on, as I heard a lot of people make that accusation with limited or no evidence.
How does the local coroner know he's supposed to do this? Because if some high-level state official calls all the local coroners and says "change your criteria for reporting deaths in support of my obvious political motive", many local coroners will call many local and national journalists and say "the State is trying to mess with the statistics".
By state agencies promulgating very boring criteria for what does or does not count as a COVID death. For instance, a policy to check whether the person tested positive for COVID within the last X days and registering that as a COVID death regardless of confounding reasons would come from the state, be pretty boring in terms of policy (it's an easy shorthand that helps coroners process cases consistently), but may in fact skew the data considerably.
There is also a reporting issue. Whether a state decides to categorize a death as "from COVID," "with COVID," or any other moniker, and what criteria it uses to categorize such deaths, doesn't necessarily mean they are trying to deceive the public as to how deadly the disease is--it's just a way they've decided to categorize things. The reporting issue comes in when the nuances of how the statistics are compiled and what they mean are distilled down to "COVID killed X people," even if the facts on the ground are that X people died but COVID caused or was associated with those deaths to varying degrees.
Then we shouldn't be talking about whether this is or is not the "more likely" scenario, we should be talking about the part where you (or, ideally, some reporter who had the same thought you did) published the actual Very Boring Official Criteria that were changed and highlighted the boring but important parts.
The two plots of "cases in Sweden vs Europe" and "deaths in Sweden vs Europe", considered together, seem like very strong support for the arguments "lockdown cannot pass a cost-benefit test" and "the Swedish non-lockdown really wasn't a lockdown".
We see a gargantuan bubble in Swedish cases, unrepresented anywhere else, that doesn't show up in Swedish deaths *at all*. Swedish cases soar to new heights at the same time Swedish deaths maintain a steady downward trend.
That's a huge number of people falling sick for whom falling sick is the right outcome -- they get all the benefits of whatever they were doing with negligible downside -- but who were, in every other country, prevented from achieving this correct outcome.
I think what might be happening there is that the UK had many more deaths than expected, skewing the graph a bit. If you remove the UK, Sweden looks pretty bad. The badness isn't really the same shape, but maybe this is because there's a lot of randomness in how long it takes COVID patients to die?
I did end up judging Sweden by deaths and not cases, so my conclusion should stay okay even if all those cases were an artifact of eg unusually good testing (which I haven't heard claimed)
But I'm saying the cases are important information that doesn't show up in the deaths.
Sweden sees an enormous increase in cases and a steep decline in deaths. That is exactly what you'd expect from a system of only voluntary measures, *if it were working perfectly* -- people who are at risk mitigate that risk, reducing deaths, while people who aren't at risk ignore their non-risk, resulting in a boom in cases. That second group reaps the benefits of whatever activities they were doing that led them to fall sick. But they don't suffer any real costs.
Compare that to any other country -- every other curve in the comparison looks basically the same. Sweden peaks at a high comparative level of deaths and declines steadily. Denmark peaks at a low level and declines steadily. Everyone, Sweden or no, sees a sharp uptick towards the end of the data.
But Denmark's *cases* peak and decline the same way their deaths do. That shouldn't happen -- the Swedish population is much better off, with their practice of infecting those who have nothing to fear from infection, than the Danish population is with their practice of preventing infections in everyone whether they're at risk or not. That is a huge cost that Denmark pays and Sweden doesn't.
Low-risk individuals getting sick is only "good" if it's going to prevent future waves by increasing the population's natural immunity. But Sweden's subsequent waves were worse, not better, than the rest of Europe's. So on a first glance, all those extra infections didn't actually make things better.
Besides, now that we have the vaccines, there's an even lower-risk way for a larger number of individuals to develop immunity.
I think Baltic countries and Eastern Europe had worse second waves than Sweden despite very strict lockdowns.
> Low-risk individuals getting sick is only "good" if it's going to prevent future waves by increasing the population's natural immunity.
No, this is wrong. Low-risk individuals getting sick is good *because it means they were doing things that were valuable to them*. They received a significant benefit and paid an insignificant cost.
Low-risk individuals getting sick can only be considered bad in an all-else-equal sense. All else is not equal; the only way to become sick is by participating in activities that make you sick.
There's a standard example of GDP being a "bad target": if I pay you $100 to dig a ditch, and then you pay me $100 to fill the ditch in, everyone is worse off even as GDP rises by $200.
We measure the benefit of economic activity in terms of GDP anyway, because most transactions aren't like that. We assume that voluntary transactions leave the participants better off, and the expenditures which constitute GDP are an indicator of those gains. They aren't the gains themselves. They're an indicator. When a trade happens, it has two effects: the participants become better off, and GDP increases.
Sweden's increased covid cases are an exactly analogous indicator of benefits obtained by the Swedes.
That very logic cuts both ways, though -- just as we have no guarantee that GDP actually correlates to value, we also have no guarantee that sickness correlates to people doing activities they value and receive this so-called "significant benefit." The same "bad target" example applies: People could be choosing to intentionally get sick.
Only in this case, while we don't have numerous documented examples of ditch-digging-and-refilling, the literature is pretty clear that people are not rational about their choices around preventative health. I would bet that if you could identify the activity that got people sick and gave them the opportunity to reverse their decision to participate in it, many would take that trade.
Was that one dinner with friends really worth getting sick for two weeks? Even if you recovered, that's often going to mean more social opportunities lost than gained. Hardly a "significant benefit" and "insignificant cost," for the types of symptomatic illness we're discussing.
One dinner might not be worth getting sick, but that doesn't matter because I didn't know in advance which single dinner to avoid. The proper comparison is whether going out to dinner with friends every week (and other social events) is worth getting sick once. To me, it is.
Even if it worked in Sweden doesn't mean it would work everywhere. Sweden has unusually few people living with their parents. My grandma lives with my aunt who is a public school teacher; if schools hadn't been closed, there would have been pretty much no possible "voluntary measures" to reliably stop my grandma from getting COVID short of my aunt quitting her job.
I've seen it elsewhere that Sweden, the UK and also New York all made the same mistake in the first wave: they triaged by sending recovering elderly back to care homes, letting them spread the virus amongst the most vulnerable and drastically inflating death numbers.
Unfortunately you can't really measure such a thing but if it's true it would make these three areas very unreliable for comparisons.
Yes I think this is the flaw of the analysis of Sweden. Their deaths were driven by large nursing homes that other Nordic countries don't have. I thought it was strange that it is not mentioned that in the end, Sweden did fairly well relative to Europe, and also the Netherlands was not mentioned which also had lax policies and did well.
I think also that it's hard to generalize from Sweden's case/death ratio to other countries, since it has one of the best healthcare systems of developed country. So other countries would not have been able to deal with the same number of cases as well, so would have higher case/death ratios.
I'm extremely skeptical of the Sweden story. There's a problem with saying, "Look at this cool n=1 experiment!", then turning around and saying, "But the conclusions fall apart halfway through the experiment, so let's agree to ignore the data that don't fit the narrative."
If Swedes were still dying at high rates after they matched the lockdowns of their peers, the conclusions we should draw are obvious: the higher death/infection rates should NOT be attributed to the lockdowns that don't correlate with the claimed effect; we should be looking for some other factor instead. Maybe that factor is genetic, cultural, geographic, or completely unknown. But we should be able to reject the lockdown hypothesis for excess Swedish infections and death based on new data - instead of just ignoring clear evidence because it doesn't fit our hypothesis.
> That's a huge number of people falling sick for whom falling sick is the right outcome -- they get all the benefits of whatever they were doing with negligible downside -- but who were, in every other country, prevented from achieving this correct outcome.
This assumes that they're getting sick because they're correctly unconcerned and doing things they enjoy. Rather than going to work terrified of the virus because their unsympathetic boss said they'll be fired if they don't. I'd argue that the utilities from letting the later stay home (and do nice things like learn to make sourdough) outweigh multiple people's worth of utilities from having a night out.
Please try to keep in mind the millions of people who had no choice at any point during the pandemic about going to work. That includes the nice people involved in getting the sourdough ingredients to your home - farmers, drivers, food processing plants, public services, road maintenance, police, with many more.
I find myself not overly sympathetic to the people who were too scared to go to their job, regardless of whether it was actually dangerous for them or dangerous in general. I have a lot more sympathy for people with medical conditions that would put them in danger specifically. Not so much the 25 year old in perfect health who is freaking out and ordering all their food at home as if that doesn't put other people in the exact same danger they are too scared to deal with.
The 25-year-old would have to eat either way, and I can't see how ordering their food at home puts the involved workers into danger any more than going to grocery stores in person -- if anything, the other way around, given outdoors vs indoors and given much shorter contacts with delivery people than with supermarket cashiers.
The 25-year-old would have to eat either way, and I can't see how ordering their food at home puts the involved workers into danger any more than going to grocery stores in person -- if anything, the other way around, given outdoors vs indoors and given much shorter contacts with delivery people than with supermarket cashiers.
The bubble in reported Swedish cases was caused by a change in testing policy (https://medicalxpress.com/news/2020-05-sweden-ramp-coronavirus.html), bringing it closer to in line with the rest of Nordic countries/Europe. As I recall, there were regional outbreaks around that time (such as one in Norrbotten) but the true daily cases across the country likely were following a steady downward trend they had for a while.
Also missing: Taiwan, South Korea, Japan, Vietnam. (Despite what many people think, S.Korea and Japan citizenry have pretty low confidence in their governments - in Japan, people still remember the government mismanagement of the Fukushima nuclear reactors after the tsunami; for S.Korea: https://twitter.com/BluRoofPolitics/status/1399555956875296769)
Frustrating how often people dismiss those examples with vague stereotypes about Asian culture being more compliant/authoritarian. Without looking at the actual politics of these countries.
Vietnam is currently in an exponential phase. Taiwan had skyrocketing cases a few days ago but now they seem to be collapsing, not sure what the story is there. Japan did pretty well in spite of no apparent differences in policy.
Surely Japan is an example that should be studied more than it is. It seems far more relevant to EU or US than New Zealand, which received tons of media coverage. And of course Japan did various things, they are however poorly understood in the West (including by me) https://en.wikipedia.org/wiki/Timeline_of_the_COVID-19_pandemic_in_Japan
I'm waiting for "Climate Change: Much More Than You Wanted To Know"
2nding this! I'd love an up-to-date survey of where things are at.
There are two other factors to consider: how open the economy is to the outside world, and how quickly information is collated and disseminated. The UK is very open, and had a government that was until recently very long to close down borders or institute mandatory quarantines, which is why the delta variant is spreading like wildfire. On the data collection front, originally it took them 2 weeks to collect and tabulate data due to an unfit-for-purpose health IT infrastructure, which obviously makes it very difficult to control either the government or private-actor control loop. In one particularly embarrassing incident they discovered that because they used an old version of Excel with a 32K row limit, they were undercounting cases.
I work as an MD in Sweden and have thought a lot about COVID-lockdowns. Many things that governments in other countries made illegal Sweden only recommended i.e. limit social contacts, don't have private parties etc. But the stringency index counts these recommendations as if they were mandatory: https://twitter.com/TTBikeFit/status/1359913436981903361/photo/1 . This makes Sweden seem more stringent than it was, especially during second and third wave. For example according to ourworldindata stringency Sweden had higher stringency than France in January -21, this at a time when it was illegal to be outside in France after 8 or 9pm without valid reason. So in the stringency index it may look like Sweden was a pretty average European country during second and third wave I don't think that was the case. Anecdotally French and German people in Sweden also told me it was much more relaxed here.
Also I did an estimate of the life years lost to COVID deaths in Sweden in 2020 a few months ago here: https://www.reddit.com/r/TheMotte/comments/msqqof/estimating_life_years_lost_due_to_covid_deaths_in/. Sweden had an excess mortality compared to previous years of 7.7 % and a 0.59 year decline in life expectancy. This translates to 63700 non QALY-adjusted life years lost to death or 6.37 years per COVID death or 2.32 days per person.
This is another weakness of the stringency indeces (and attempts describe the COVID policies of dozens of states under a sparse set of categories). Often times, the catgories available won't match precisely with a country's policy. Then, the maintainers of the index either subjectively choose which category is "closest" and put them there, or they write a detailed description of what each category actually means so that they can shoe-horn everyone in somewhere. Then, they summarize those detailed categories with pithy descriptions. This amounts to the same thing, as everyone continues using the short hand.
So in the end, you get a lot of gray area between categories which you hope averages out across the index but doesn't always.
I don't have a better way to do this kind of thing, but its important to recognize how flawed a description of actual policies these stringency scores are.
One complicating factor that doesn’t seem to be covered here is how many QALYs were lost due to things like:
1) Lack of exercise: gyms and parks were closed and group workout sessions were cancelled. Even when facilities were reopened, I didn’t use them because of restricted hours, fear of infection, and unwillingness to work out with a mask on. I’d guess many others didn’t either.
2) Increases in being sedentary: even if people managed to keep up their workout routine, people who worked from home just needed to move around less to do things compared to when they had a physical workplace to go to. Almost everyone I know with a step counter has reported a massive drop in their daily step count.
3) Delayed medical check ups and appointments. How many people will have cancers or other issues detected a year late or treatments delayed by a year? I know that me and many people I know delayed our appointments, figuring the odds are that we’d be fine. But a substantial number of people doing that means that many of us won’t be fine.
It’s possible it will take another year or more before this data is really available, but I’d be interested to know if we have any preliminary results or forecasted estimates on these. And if the strength of those results would overturn any of this post’s current conclusions.
For one counterexample, for me to work from home saves me two hours' train commute per day, allowing me to exercise at home which I otherwise wouldn't. (It also saves 1 h a day of walking from home to the train station and from the train station to the office and back, though.)
"Finally, voluntary behavior change and mandates are hard to separate. If you hear the government is thinking of a mandate, that might make you scared and cause you to do things voluntarily (even things not included in the mandate). Or if the government knows that most people are staying at home, it might feel more comfortable issuing a stay-at-home order to mop up the last few holdouts, whereas if no one had been staying at home it might not be willing to do that."
One factor to consider here and which might further confound is that lockdowns almost certainly don't have effects just in the countries where they're implemented - you can't just treat countries as closed individual units in an open and connected world economy. Of course, an important factor is that if some country is in a bad phase of epidemic, it increases the chances the virus spreads elsewhere - usually, of course, this leads to travel restrictions, but the travel restriction patterns are not always optimal. Here in Finland, a major issue for these weeks has been the spread of virus from Russia to Finland via football fans who were watching the UEFA European Championships in Russia while Russia has been going through a fairly harsh wave, and of course if Russia hadn't been going through that wave, it would have also reduced considerably the chances of spreading to Finland.
However, another thing is that while people will adjust their thinking about how bad the virus is and what precautions to take according to government messaging and actions, government messaging and actions may also have effect in other countries. Throughout the pandemic, the Finnish media has paid a close attention to what other countries have been doing - particularly Sweden, usually presented as a negative example of what not to do. For instance, a bunch of countries enacting lockdowns around Europe has generally been considered a pretty good signal that extra measures will also be coming in Finland, and might lead to voluntary behavior changes even before the actual measures come. Now, the medias - and accordingly, a fair number of people, assuredly only a minority but still probably the minority that acts as the opinion-makers - are paying close attention to UK's policies regarding reopening, and which way they will go, as well as what the real relationship between the Delta variant, vaccines, reopening etc. is vis-a-vis case and death numbers. All such matters may end up having effects on public behavior without the Finnish government necessarily doing anything major.
Of course, this also leads to something of a freerider problem - it's perfectly possible that one of the reasons why Sweden was able to have lenient restrictions at the start of the pandemic without the situation getting really out of hand *was* that the other Nordic countries had stricter restrictions, not only lessening the spread of pandemic from those countries but also potentially contributing to voluntary behavior changes within the Swedish society.
Minor notes:
- "Stringency index" may also be confounding due to the policy of countries being considered according to the strictest subnational unit (https://ourworldindata.org/grapher/covid-stringency-index). At least in Finland, restrictions in the greater Helsinki region have mostly been stricter than in the rest of the country, but the majority of people still live outside of greater Helsinki.
- One of the functions of the lockdowns might not only be that they serve as a government signal for "things bad", but their end might also serve as a government signal that "things good, get on with your lives", potentially allowing for a faster reopening than whatever inchoate signaling the government is able to do *without* having a lockdown, in case the people still choose to voluntarily stay home.
- regarding masks, their spread in society - or people giving up on mask-using if they've gotten used to it - really probably happens in slower waves than what can be measured in, say, three weeks, unless there's a formal mandate with teeth. At least in Finland, where there was next to no mask-using until late Autumn 2020 when the government finally issued a recommendation (not a strong mandate; there has only been a mandate in some public transportation in trains and some cities, and even that is not really enforced too strongly) to use masks, it took *months* for anything close to a 80 % mask compliance to happen in places where it was recommended. (Still, it did happen without formal mandates, eventually, which also tells something about the Nordic society!)
- as the single-person household map shows, Estonia has a high percentage of households, and it still had a big COVID wave a few months ago, having the highest case rate in Europe per capita at one point, if I remember correctly. https://en.wikipedia.org/wiki/COVID-19_pandemic_in_Estonia
Also forgot to mention that one reason why Nordic countries have managed to keep COVID generally in check with differing measures might simply be a robust welfare state - not only in how it generally means a fairly strong health care system (for instance, Finland's success probably depends on ability to genuinely maintain a pretty good test-and-trace system - not a lot of countries have been able to do that, which has also meant they've had to enact stricter lockdowns!) but also that a good welfare state increases people's trust in state and means the state can "get the message through" with recommendations and does not have to resort to formal mandates and punishments.
"My data source didn’t have the past few months, but if someone does, let me know."
Not sure if you read the reddit, I left some links there. These are the sources I have been using:
https://github.com/owid/covid-19-data/tree/master/public/data
https://github.com/nytimes/covid-19-data
The former has daily national data, the latter daily US data by state and county. Both updated today.
While I agree with your analysis regarding "quasi-lockdowns" a la US and west Europe, I think you are missing the bigger picture with actual lockdowns a la China, east asia, oceania. I wrote about that on the reddit, but I see some people here covering similar points.
"Similarly, there was a really wide diversity of compliance levels with shelter-in-place orders. I know some people who didn’t see their closest friends for months, and others in the same cities who said “screw this” after a week or two and started having (small) parties again."
It's honestly difficult to say about how effective such measures are. The problems here in Ireland start after restrictions are eased, people come out in large numbers, and then we get a bump in the Covid cases. Part of that is economic, as businesses are pressuring government to relax restrictions (e.g. for Christmas as that is a major revenue generation period) and now for the summer - but we see that when restrictions are lifted, some people can't be sensible and then we get a spike once again in infection rates and then restrictions clamp down again.
Right now, where I live is one of the two areas in the country with the highest rates of Covid-19 (and we are not a large urban area). There was a spike in infections locally in the middle of June, and anecdotally that is alleged to be because in a local village, people who had been attending a match on the weekend went to the pub afterwards, where they mingled with people who were home from England for a funeral.
Given that the Delta variant was established in England at the time - well, there you go. This is also why many people are not happy with Boris Johnson and the relaxation of restrictions in Britain.
Just scanned the comments quickly, sorry if someone else has already pointed this out, but note that the same people who are likely to have high rates of voluntary behavior change are also likely to vote (or email their representatives, or...) for lockdowns. So if you check for the effects of lockdowns assuming nothing else is different between states that did/didn't implement them, you may be missing something.
My impression as a Bay Area Californian from listening to other people talk is that our voluntary measures, especially early on before the state took notice, were unusually extreme, our cooperation with the mask mandate once it did get instituted was unusually high, and the tendency for people to take precautions that weren't officially required (grocery delivery instead of shopping; washing groceries; refraining from taking walks, talking masked and distanced with one friend, or other activities that were explicit exemptions to mandates; etc.) was much more pronounced around here than many places. (Though take that with a grain of salt - the non-California part of the comparison is based on what people talk about, I wasn't actually in any other state until last week, and by that point differing mandates potentially affecting people's behavior had thoroughly confounded everything.) Still, I don't think the set of attitudes that get us the above set of behaviors are entirely unrelated to our state, and moreso our county, hitting the lockdowns early and hard. (At least, hard for a US State - anecdotally, compared to Chile or Spain or China we seem to have been quite mild.)
Maybe Bay Area was even more compliant, but Dallas was a ghost town until a couple months ago, and I don't think I ever saw anyone out without a mask. I tried to go see a movie for the first time since it all started last week and found there is still only theater even open within 10 miles of downtown. I don't think they've been legally required to be closed for nearly a year. They've just chosen not to reopen because there aren't enough people who want to go back to theaters even if they're legally able to.
It really seems to make these kinds of cross-state comparisons seem pointless. What difference can the extent of legal mandates really make if the behavioral changes in any place with enough population density to matter is nearly the same anyway?
It also seems hard to suss out why New York and Los Angeles were hit so hard having anything to do with local policies from the fact they're so much more exposed to international travel than other American cities. Restricting international travel isn't a local decision.
Los Angeles has a big chunk of population who couldn't care less what government tells them to do. Not from some highly academic libertarian philosophy, mind you, but because their family and sometimes personal history south of the border is that government is inherently corrupt and incompetent and only a fool pays attention to what it says if that conflicts with your mother wit, or your actual mother.
An effect not discussed (but which I don't know how you would handle)
In the UK the government said that lockdown was about avoiding hospital overload -- where overload is defined as a tipping point of serious cases per day the exceeds capacity and suddenly the standard of care drops and the mortality rate goes much higher.
I am interested in whether or not this turned out to be a thing? Early on, the consensus was that for serious cases a ventilator was required for recovery and countries were scrambling for ventilators; however, later on it the news told me that ventilators were irrelevant to recovery (that generally if you were put on a ventilator for covid, having it removed would kill you -- there was no recovery as such). So did it turn out that actually hospitals couldn't help you very much anyway if you had serious covid, which might make this capacity-issue not a thing. Just wondering because it's been a very long time since the UK government mentioned this issue -- nowadays they talk about balancing lives lost with keeping the economy running.
On the other hand, if the capacity issue does exist, then this changes the models above.
> So did it turn out that actually hospitals couldn't help you very much anyway if you had serious covid, which might make this capacity-issue not a thing.
There's a degree of covid severity, where intermediate patients are helped by supplemental oxygen but do not need mechanical ventilation. In early May, a shortage of medical oxygen in India was widely reported in the international press
This. If you need a ventilator, more likely than not omae wa mou shindeiru, but it's not just ventilators that there were shortages of in the hospital in northern Italy where my girlfriend works. (There even was a shortage of doctors, as half of them got COVID at the same time in April 2020 and the other half in November.)
As far as I can tell Austria definitely had a problem with its intensive care capacity, particularly Vienna. Intensive care wards are generally at 80-85% capacity in normal times, so a sudden influx of patients with covid can be quite dangerous. It's harder to care for patients with covid since you can't just put them in a normal bed and they need lots of care, but after the past year most nursing stations have been overworked for quite some time and don't have much capacity left.
For some reason the UK government and the NHS seem to have been downplaying how close hospitals were to capacity in January. At least in London, we had
- many hospitals keeping patients on less oxygen than recommended (due to some combination of a shortage of oxygen canisters + pipework in hospitals that couldn't handle such a high flowrate)
- intensive care was essentially full, and many people who would normally have been put on intensive care weren't
- each doctor/nurse had to look after more patients than normal
- most specialist doctors/nurses + wards were reassigned to covid
- all non-emergency surgery was cancelled (so there's now a massive waiting list)
By shutting down most non-covid activity + because of the fact that ventilators aren't as important as expected, hospitals managed to deal with significantly more patients than the maximum "surge capacity" that people were estimating at the beginning of the pandemic, but they couldn't have dealt with much more, and it had (and the coming wave will have) a very real cost to non-covid care.
I suggest the equipment is less an issue (in First World countries) than the burden on trained personnel. If the ER is jammed, there are people in the halls, and ambulances are stacked up 5 deep in the parking lot -- which happened at a few points in LA County -- and the RNs are working 16-hour shifts in moonsuits and the doctors falling asleep on their feet if they stop walking, you don't get first-class care. Mistakes are made. And not just with COVID, this is not the time to be walking into the ER with a curious abdominal pain that could be gas or could be an aortic aneurysm getting ready to blow.
This is obviously not something anyone likes to talk about, because we like to preserve the illusion that MDs and RNs are programmed robots delivering an identical unit of healthcare every time you insert the appropriate token, but for that reason it may be some of the talk about ventilators and such was sort of a cover story for the real truth, which is that some people were trying to ensure the healthcare *people* didn't get overwhelmed.
We also like t preserve the illusion that MDs and RNs can actually cure what ails us, or at least do *some* good. But how true was that of COVID e.g. last March? I read the OP's question as, basically, would "overloading" hospitals at that point have been a megadeath catastrophe, or would it have been the equivalent of "overloading" medieval bloodletters so they couldn't bleed as many people as they thought they ought to, or somewhere in between?
OK, I've stacked the deck by picking ridiculous endpoints; it's somewhere in between. But we need at least a rough estimate, or more precisely we needed a rough estimate last March and I don't recall seeing one.
I think it's always true. In infectious disease merely supportive care does a lot of good. The body is at rest, freed from the stress of activity or temperature variation or what not, it's getting proper nutrition and hydration, secondary complications and infections can be dealt with promptly, supplemental oxygen can be given, et cetera. For most of human history the bulk of what a good hospital did was pretty much just good supportive, i.e. nursing, care, and it definitely worked -- people were more likely to get better if they had excellent nursing care.
So I'm having a hard time seeing any plausible mechanism whereby a decline in the quality of supportive care, caused by professional staff exhaustion, would *not* increase the death rate. I'm perfectly willing to agree in the absence of data that the change might have not been very big, or even in principle negligible, although I would expect an unimportantly small change to accompany a disease that is either much less or much more lethal than COVID, so that the question of the optimality of the general physiological state of the patient was less relevant.
Don't have the link and can't remember how well researched the theory was, but remember reading about positive effects of simply being cared for. IIRC, it was ascribed to an atavistic mechanism of body allowing itself to dial down functions unnecessary for healing, now that somebody else watches for predators etc.
Thank you for undertaking this detailed analysis.
I do think a better pair of questions would be, "what is the optimum behaviour to respond to the pandemic?" and "what is the best way to get people to adopt that behaviour?". "Lockdown" confuses the two of these, and comparing lockdown vs voluntary measures is focussing on the second question at the expense of the first.
The part of your article I'm least convinced about is the attempt to model emotional impact. You summarise this as "52 months of stricter lockdown to save 1 month of healthy life", which does sound harsh, but it could also be phrased as "52 months of stricter lockdown instead of 52 months living through a pandemic with many voluntary limitations on behaviour", which is rather less harsh - albeit still not necessarily worthwhile. Also, some people may find it more emotionally traumatic to be living in a pandemic if the state and their fellow citizens behave in a non-optimal way. Early in the UK, there was a sense of collective pulling together in adversity that strongly mitigated against emotional impact. In other words, emotional impact is hard to measure.
On the other hand, you haven't mentioned the educational impact of disrupting children's education, which seems to be an important downside of the pandemic.
The Sweden discussion should really consider the effect of its uniquely high immigrant population. Swedish virologists claim this is a huge factor.
Can't find original link. But Sweden has very high asylum seeker numbers and these are hugely overrepresented in covid cases. Likely this explains quite a chunck of differences with other Scandinavian countries with much stricter immigration policy.
One early link:
https://foreignpolicy.com/2020/04/21/sweden-coronavirus-anti-lockdown-immigrants/
International migrants, in total, are 20 % of Sweden's total population and 16,1 % of Norway's total population, at least. A difference, but not one that can go anywhere near fully explaining the disparity in COVID numbers between countries.
I believe Sweden also has much more international travel?
I'm not sure?
If we look at the number of tourists, Sweden has more of them in total, but Norway has more per capita (Norway has half the population of Sweden, roughly.)
https://data.worldbank.org/indicator/ST.INT.ARVL?locations=SE-NO
That might explain difference in cases, but I guess age profile of asylum seekers is such that they aren´t usually at high risk of death or hospitalization from covid (?)
I'd love to see "Ivermectin: Much More Than You Wanted To Know," given the absolutely baffling lack of main stream media coverage (and suppression by Facebook, Google, etc) on the topic.
From what I can tell, it's probably better and safer than vaccinating teenagers and under-12 children. It might be capable of actually eradicating COVID-19, if its usage is widespread enough. It might even cure long-haul symptoms, although so far the accounts are anecdotal and there hasn't be a formal study.
(https://covid19criticalcare.com/ivermectin-in-covid-19/)
I don't see any compelling evidence / arguments to *not* to the claims being made by the Front Line COVID-19 Critical Care Alliance, while there are a lot of people who very obviously have a literal vested interest in a cheap medication like Ivermectin not being the solution to COVID-19.
I respect Bret Weinstein enough to believe that his current take on this topic is probably accurate, but it would help tremendously if Scott could weigh in on it, too.
Damn lack-of-editing feature!
That should be, "I don't see any compelling evidence / arguments to *not* BELIEVE the claims being made by the Front Line COVID-19 Critical Care Alliance..."
I too would like to see such an article, because I haven't yet had it explained to me how a treatment against various worm infestations and parasites is expected to work against a virus. And by that, I mean something more stringent than "well, it kills it in vitro, we have no idea of what the potential mechanism of action might be, but hey, why not give it a shot?"
https://www.farmersjournal.ie/cattle-wormer-ivermectin-can-kill-coronavirus-538649
Right now, with all the shilling for "cattle drench for covid!", my attitude is hardening into "this is the Miracle Mineral Supplement cure for autism all over again" https://asatonline.org/for-parents/becoming-a-savvy-consumer/is-there-science-behind-that-bleach-therapy/
If it works does it matter that we don't know the mechanism at first? I thought the sharp drop in India was due to ivermectin. But googling India covid and ivermectin you can find whatever story you want to hear.
There again - high incidence of parasitic diseases in India: https://www.omicsonline.org/india/parasitic-disease-peer-reviewed-pdf-ppt-articles/
"The study included a parasitological analysis of 256 stool samples in low socioeconomic areas of south Chennai with special attention to both intestinal protozoa and helminthes. This study has documented a high prevalence (75.7%) of intestinal parasites in the dwellers from south Chennai, India. Saidapet and Thiruvanmiyur dwellers were found to harbor the maximum number of positive cases."
So I am inclined to think this is rather like the vitamin D studies: take a population with a high incidence of parasitic diseases, dose them with a de-wormer, improvement in not having to fight off the parasite at the same time as you may be developing Covid-19 means better health outcome.
When/if we get studies on populations that don't have worms/parasites and that respond well when given ivermectin, then it will be more plausible that it is doing something specifically against the virus, rather than treating incidental sicknesses that make fighting off Covid-19 worse.
Huh, OK I think there are some studies from Oxford coming soon. Personally I know almost nothing. Did you look at the web pages from the FLCCC? These are doctors treating people for covid.
Don't overlook the more economical explanation of the fluctuations to be expected in small numbers. Any time you study *anything* in a small population, you find all kinds of weird correlations. These vanish when you do studies with sufficient statistical power that the spurious things finally average out. I think what people don't intuitively grasp is how *slowly* this happens. It tends to defy common sense that you really do need trials enrolling thousands and even tens of thousands to pick up on genuine effects that aren't huge, because you need to average away all the statistical junk. Even the pros are caught by surprise sometimes, so hard to avoid is the human tendency to see signal in noise.
There was at one time a thriving cottage industry among innumerate journalists that lay in discovering "cancer clusters" in certain small towns next to a tire plant, electrical substation, or farmer using ivermectin. Apparently it always escaped their notice that you never found such anomalies in big cities with a million inhabitants, even if the cities also had tire plants, electrical power lines, and traces of hormones or plasticizers in the drinking water.
The sharp drop in India is due to herd immunity, similar to that in Peru. By this point, a higher portion of India's population has been infected than any part of the United States, probably due to the spread of the Delta Variant.
Huh, how do you know that? (testing for antibodies is expensive) Ivermectin was used in some Indian provinces/ states. Not related to your comment, but I find it wierd/deja vu that ivermectin gets turned into a wacko conspiracy idea.
Derek Lowe is the person you'd probably want to write that article, and in fact he has already done so:
https://blogs.sciencemag.org/pipeline/archives/2021/06/07/ivermectin-as-a-covid-19-therapy
Thank you for the link to that article, it seems to chime with my instinctive notions. I may be jumping at an illusory mayfly here, but the fact that the quoted studies were done in Third World/Developing World countries makes me wonder if, in fact, the good results from ivermectin usage were *because* of its anti-parasite qualities and not any anti-viral ones, that is, that patients may have had exposure to worms, lice, scabies and other fun diseases and once ivermectin knocked those on the head, their immune systems were able to concentrate on fighting Covid.
I could be full of hot air on that, though.
Oh, well: if the ivermectin devotees are correct, then every farmer who's been out dosing their cattle for roundworm should have no worries about contracting the virus, if any accidental exposure to the drench has a prophylactic effect! 😁
That article is *very* far from a Scott Alexander MTYWTK post. It basically asserts that the data is weak, gives a cursory glance at a few recent studies, and refers you to the WHO for details. Lowe may be correct in his conclusion, but there's almost nothing there beyond his opinion.
I take your point about total deaths being sensitive to how old the country is, but cases numbers before autumn 2020 are an outright mess. Some countries (UK) had a policy of "Don't get tested, just self-isolate for 14 days" until mid-April.
Now that we have granular data, I like "Deaths 50-69" and also the slightly more complex "Deaths per 100k people in each age bracket". Anyway, if you use these it pretty much confirms that Sweden did much worse than Europe in the first wave, and relatively the same as Europe in subsequent waves.
Yep. When my girlfriend got COVID in April 2020 we were required to stay at home but I wouldn't get tested unless I got symptoms. (At a certain point I even consider lying about having symptoms to get tested, then decided that someone else would probably have needed that test more than me. I only found out I hadn't gotten COVID at all later on, when I got an antibody test out of my own initiative and pocket.)
At the end of Preliminary Theoretical Issue 2 I thought "WTH??? That kind of thinking only makes sense near the Pareto frontier, and almost all western countries very clearly aren't anywhere remotely near it!!!" until the middle Actual Evidence 1.1.
im not sure i buy that US data is as nice as we'd like to be. in the US, red states are fairly strongly clustered in the south, and blue states are very strongly clustered in the north, which means red states got much milder winters, most deaths occurred during winter, and covid is pretty seasonal, so we should expect lower-lockdown states to get some sort of fairly significant advantage, but its not clear how much, which confuses everything
What's driving the seasonality? If it's due to people being indoors more, that applies to both texans in summer hiding from the heat and new yorkers in the winter hiding from the cold.
Mostly that, but partially also that the virus can survive outside the human body for longer at lower than at higher temperatures.
Just in terms of Scandinavia, Youyang Gu finds in the US that income inequality (measured by Gini coefficient) is the best predictor of deaths from covid: https://twitter.com/youyanggu/status/1407418434955005955
In that thread he cites a paper that finds this also for countries: https://link.springer.com/article/10.1007/s10198-021-01266-4
My impression is that both the deaths from COVID, and the costs of lockdowns tend to primarily fall upon poor people because they have higher rates of comorbidities, in person jobs, less access to healthcare, less savings, don’t own a home as often, etc. While the $150K/QALY figure for a first world country’s government might make sense to the government, since they have access to a tax base of people and companies who make a ton of money, it might be overpriced in this case.
Note that poorer countries tend to price QALY’s lower. My guess is that that’s because if most of your tax base was making $20K a year, paying 7 times the average annual income to keep someone alive for a year might be more costly and impose higher opportunity costs.
In the case of COVID lockdowns and deaths, most economic damages are going to be born directly by the poor as are deaths. So (without redistribution actions) it’s as though the deaths and monetary costs are born by a lower income country, and thus we should discount the cost of a QALY appropriately so we can better judge tradeoffs.
Of course, some countries (like the US) had pretty massive aid packages that helped make it so the whole population helped bear the costs, which would raise the QALY price back up to the national pricing.
Overall my point is that maybe you should be either more for redistribution during the pandemic (to raise QALY price to where it’s worth it) or more against lockdowns (since QALY price might be lower than your normal estimate).
Obvious disclaimer that QALY prices are for assessing tradeoffs and not for valuing the moral worth of a person’s life.
I agree. I feel like the "economic costs are borne unevenly" argument was given short shrift. (As were most counter-lockdown arguments/data.) I don't know the data on this, but in my personal experience, people who had more education/income had much more job security during the pandemic; while people with less of both these did poorly. The costs of the lockdowns were strongly regressive along the income distribution, which is an ethical issue that shouldn't just be dismissed as a single factor among many.
I also noticed that those at the bottom of the income distribution (again in my subjective circles) were the most likely to complain about lockdowns and want them lifted. If the people you're hurting are saying, "don't hurt me anymore", I think you have an obligation to listen to them; and ultimately to seek other solutions.
As for the redistribution aspect, I feel like this is also regressive. I drive electric cars and have solar panels on my roof, so the gas price increases are something I sort of notice when people point them out to me. Meanwhile, my friends - who drive cheaper gas cars - are struggling to find a way to pay to get to work, while watching their dream of EVER buying a house someday disappear entirely.
I've heard the conservative argument that "actually, the top 10% pay a larger share of the taxes collected than the bottom 50%", or whatever, and I think, "Yeah, but that doesn't matter as much as it used to, does it?" Because when the government collects $3.4 trillion per year, but spends $4.8 trillion, the other $1.4 trillion comes from the value of the dollar. Meaning it's regressive to borrow money and then spend it - no matter how well-intentioned the idea. I have assets that grow with inflation, but my lower-to-middle income friends don't. I'm protected from inflation, but they get the shaft. Not sure "we'll fix the problem by letting the government pay for it" is a good idea, when it just means hurting the poor and struggling workers that much more.
You seem to be missing the psychological cost of not locking down. Lots of people are frightened by the lack of lockdowns. This isn't just that they are voluntarily locking themselves down, but the emotional / psychological cost of perceiving everyone else as being a (potential) plague-carrier.
If there is non-economic suffering from an absence of lockdown, then that makes the calculation rather more balanced.
I remember, myself, how utterly *relieving* it was when the first stricter-than-usual measures were announced here in March 2020, after a fair bit of terror of being in an exceptional situation and not knowing what was happening. Finally it felt like someone's in charge and that we're not just going in blind to a pandemic. Of course, that was related to the immediate state of expection, situation was a bit different once people started getting mentally adjusted to living in a pandemic.
UK is ending the last restrictions on July 19, and there's a lot of social media that is clearly people who are terrified and are going into stronger voluntary lockdowns as a result.
What about the emotional suffering of devote Christians knowing that their neighbors aren't going to church? Or are engaging in sodomy?
At what point are people's own responses to their own situations their own responsibility?
Lack of lockdown often makes concerned people overreact, cutting out hospital visits or playground visits or the like, so that even as the lockdown reduces activity for some people, it increases it for others.
I'm not aware of any comparable way that lack of churchgoing makes churchgoers overcompensate in some way.
A risk of death from any cause is (assuming rational people in some sense) accompanied a proportional amount of worry about death. So whatever cost we usually consider reasonable to save a live can really be considered to be the cost we are willing to pay for saving a life *and* reducing the amount we worry about dying. So it's reasonable to compare the cost we are willing to pay to prevent a COVID death to what we are willing to pay to prevent death from other causes. (Indeed, once you die, you don't feel anything, so arguably any cost we pay to reduce the risk of death is actually to reduce our worry about dying.)
But people aren't rational in that sense. In average they are more scared of flying than of driving even though the latter is ~4× more dangerous per passenger-mile than the former, and more scared of terrorism than of lots of other stuff even though the latter is almost always a negligible risk for the average person.
True (at least in general, I'm not sure about the specific example of flying vs driving), but I won't support altering policy to manage people's irrational fears.
"So every 52 months of stricter lockdown in counterfactual Sweden would have saved one month of healthy life. You will have to decide whether you think this is worth it, but it seems pretty harsh to me.... Even if this is true, that just means it’s 21 months of stricter lockdown to save one month of healthy life. Again, seems pretty harsh."
It's very important to note that this isn't the cost of lockdown being compared to life lost, but the emotional cost of an involuntary vs voluntary lockdown. In other words, assuming the wild guesstimate is right, the Swedish locking down as their neighbors did would have been worth it as long as the difference in quality of life between them and their neighbors during the lockdown was no more than 4.8-1.9 %. Now, I didn't live in Sweden during the months of their lockdown, but I highly doubt that they had a great time there, and it's well worth remembering that we were all very far from the Pareto frontier and there weren't many big easy gains to be had. Maybe life was 5% worse or 1% worse in Sweden compared to Denmark during that time, maybe it was even 2% better if life was much scarier and more uncertain for them! But both are swamped by a giant shared hit to quality of life in both countries.
"And even if Sweden had decided to double down and weaken their lockdown despite high case levels, people would probably have voluntarily stayed home more because the pandemic was so bad - inflicting most of the same costs that a state-mandated lockdown would.... This argument seems less convincing in the US, where red states mostly just consistently had weaker lockdowns than blue states did, and never really got stricter to compensate. "
This seems like it was definitely true in Europe in general, and arguably it is less obviously true in the US, but one thing you discount is that maybe the emotional hit got bigger in the non-lockdown states even if measures were never reintroduced. Maybe voluntary behavior change goes higher instead if the red states never got stronger lockdowns to compensate, so the hit is still larger - your evidence is only about legal restrictions, not mobility data.
You say it seems kinda harsh to save one healthy month through 51 months of stricter lockdowns. I think you are overestimating the difference in life quality between softer and stricter lockdowns. For one, voluntary behavior change makes the lived reality in these scenarios not all that different for sensible folks, and for another, a lack of lockdown causes suffering too, both psychologically (feeling like a sacrifice to capitalism in the face of a global crisis) and in terms of being subjected to avoidable risk (having to come into the office where people don't wear masks for no good reason). But I might be biased because introversion causes me to avoid large gatherings anyway, so I barely lose quality of life (and gain some, as remote work and studies are much more to my taste).
Another effect the lockdowns had were in suggesting appropriate voluntary behavior changes. Hearing about possible lockdown measures worked to drive home the idea that these are the things that should be avoided, prohibited or not. Sure, recommendations were made throughout, but a lockdown drives home that really, everyone should do these, it's not one of these things the teacher suggests but the kids ignore.
Indeed - I'm clearly in the minority here, and I feel a bit terrible saying it, but my quality of life has been substantially *better* under lockdown - I get to work from home, hang out with my dog all day, have dinner with my wife every day (even during busy periods at work), have a lot of mental space. I'm very lucky, my conditions are pretty optimal for lockdown life (I have a house right beside a park, no kids, work easily done from home, my family is thousands of miles away anyway, and although I miss in-person hangouts, a history of online friendships from moving around a lot).
Is there any known sensible estimate for a conversion of dollars into QALY, possibly assuming all the money to be subtracted directly from some specific healthcare budget or something like that?
There are broadly two methods of converting dollars into QALYs:
1) The first is by setting a healthcare budget, maximising QALYs within that budget and identifying what the most marginal QALY you can purchase on that schema is. The theory is that any healthcare intervention more marginal than this would displace a less marginal intervention (because you have a fixed budget) and therefore you should only purchase healthcare interventions less marginal than this number. To all intents and purposes, this is therefore the 'price' of a QALY
2) The second is by setting a value we are prepared to pay for a QALY and letting the healthcare budget float until the most marginal intervention we can fund is equivalent to this value.
To the best of my knowledge there isn't any sensible work on the 'true' price of a QALY in definition 2, because it derives from the 'gut feeling' of legislators and leads to a lot of confused and contradictory outputs. The private / insurance-based part of the US health system does something a bit similar to 2 where individuals decide how much their health is worth to them and pay accordingly, but this is confounded by a lot of other factors like individual ability to pay.
However, there has been a serious and systematic effort to identify the price of a QALY under definition 1, most notably in the UK where it is very important for setting spending priorities for the NHS there. This is an excellent summary of the 'state of the art' in estimating the price of a QALY. The methodology is to find the budget decrease that would cause the health system to produce one fewer QALY, and argue that at the margin this must be the price of a QALY. It finds that the value of a QALY is approximately $18,000:
https://www.york.ac.uk/media/che/documents/papers/researchpapers/CHERP81_methods_estimation_NICE_costeffectiveness_threshold_%28Nov2013%29.pdf
Here's a critique of that approach (although to your point; everyone agrees the number is sensible in an order-of-magnitude kind of way, but data limitations mean it might still be off by quite a lot):
https://www.ohe.org/publications/critique-che-research-paper-81-methods-estimation-nice-cost-effectiveness-threshold
This is a really nice exchange because Claxton et al are famous advocates for low cost-per-QALY thresholds and Barnsley et al are famous advocates for the opposite - you can be pretty confident that anything they agree on is likely to be pretty close to the truth.
Thanks!
https://thefatemperor.com/published-papers-and-data-on-lockdown-weak-efficacy-and-lockdown-huge-harms/
Conclusion 2 is doesn't answer the question it's trying to. The headline is "Quantify Emotional Costs Vs. Benefits", but you're not testing the costs of lockdown vs no lockdown. Instead you're testing the marginal benefits and costs of Sweden being stricter like Denmark. I would definitely have liked to see the former, as it is more relevant to the political debates I've seen in the wild; they tend to be "lockdowns are good and we should do them again next pandemic" vs "lockdowns are bad and we shouldn't do them next pandemic".
Regarding the question you did try to answer. I think it's flawed by the lack of a serious attempt to define how much suffering is involved. 51 person-months in stricter lockdown per QAYL saved. If the extra strictness is moving from 100 people per gathering to 10 people per gathering, that seems doable. If it's no school, then it's not worth it.
It's also worth considering positive benefits from the lockdown too. If the cost of a lockdown is that you can't go to a cafe. But the benefit is the waiter isn't forced to choose between bankruptcy and going into work every day terrified you'll catch covid. I would call that a net positive.
This probably applies when comparing the marginal costs/benefits of moving from Sweden to Denmark. But it *definitely* applies when comparing no-lockdown to lockdown. The point you made earlier about "The moral of the story is that everything not forbidden is compulsory, so you can’t always substitute voluntary behavior change for government mandates." absolutely needs to be considered when quantifying emotional costs vs benefits. My working class girlfriend had such rants about how unsympathetic employers around her were when it comes to covid and her daily fear of it reaching the family (they have medical vulnerabilities). Only the lockdown let them isolate, meanwhile my tech company sent everyone home to work in our spacious home offices with gardens outside. Any attempt to measure the emotional costs/benefits of lockdown has to cover the emotional upsides.
> Preliminary Theoretical Issues 6: The Pandemic As A Control System
I think that a complicating factor here is that people don't observe R. They observe (media reports of) cases and deaths.
All other things being equal, a given set of behaviours will result in a particular R. The "pandemic as a control system" theory suggests that R will hover around 1, which implies equivalent behavioural sets under both lockdown and non-lockdown counterfactuals. The potential "win" of lockdowns, therefore, is to induce the "bad covid, therefore panic" set of behaviours earlier than would otherwise happen, so the same R-value (1) can be achieved with a lower overall case and death rate.
This feels like a very complicated problem, in large part because people respond differently to lockdown orders. After a moderate level of restriction, the marginal case might be found in an "essential" business like a logistics warehouse or in a wilfully ignorant member of the population, and imposing further mandatory orders might have little effect. That "resistant margin" may itself vary from time to time and from place to place.
In the scatter plot of stringency on Jan first vs deaths per million, the correlation seems disproportionately driven by six states in the lower right. The rest just look like a blob of noise. Now, %12 of the dataset is obviously too large to be dismissed as an outlier. But which states are they? Is there any particular policy or timing they have in common?
I'm pretty sure those are Washington, Oregon, Vermont, New Hampshire, Maine, and one other - maybe Hawaii? (These are the states that systematically had lower case counts than almost all other states at almost all times from March 2020 to March 2021 - after that time, Washington and Oregon spent a while near the top of case counts, but still lower than most states had in the third wave.)
Have there been attempts to look at rates of all-cause mortality in locked down vs non-locked down areas? I could imagine this going either way. Hospital system collapse would increase all-cause mortality, which might not be accounted for by these studies, and in younger people.
Areas, not that I know of, but countries, here are the ones who never exceeded 1.25× the usual total death rate during the pandemic https://ourworldindata.org/explorers/coronavirus-data-explorer?zoomToSelection=true&pickerSort=desc&pickerMetric=population&Interval=7-day+rolling+average&Relative+to+Population=true&Align+outbreaks=false&country=SYC~MAC~MUS~JAM~MNG~URY~NZL~NOR~FIN~DNK~SGP~CUB~TWN~AUS~MYS~KOR~THA~PHL~JPN&Metric=Excess+mortality and none strikes me as particularly lenient.
IMO (for what it's worth) the critical issues for assessing the efficacy of NPIs are:
1) How do you control for precipitating conditions? In other words, the severity of the interventions was likely, in most cases, a function of the severity of the ongoing pandemic. The severity of the ongoing pandemic would be associated with the severity outcomes independently of the efficacy of the interventions, to some extent. Thus, reverse engineering from comparing outcomes across contexts, w/o controlling for the precipitating conditions, is of limited value.
2) Evaluating the efficacy of NPIs should necessarily address counterfactual assumptions about what would have happened absent the NPIs. You can't assess the differential effect of NPIs w/o addressing the issue of whether things might have been much, much worse if they weren't implemented - say if the pandemic raged more, whether more people would have stayed away from hospitals. That is the 300 lb. elephant in the "lockdown deaths" arguments.
3) Obviously, controlling for confounding variables and spurious associations when making comparisons across countries with vastly different conditions is VERY problematic. IMO, the data available is insufficient to do this. IMO, better is to look at patterns longitudinally in single countries. And even there, we don't really have a long-enough period to evaluate the effect of the NPIs - for example it may take years to really assess the impact of closing schools.
"Here, having US red states switch to a blue-state typical level of lockdown would save one month of healthy life per 51 person-months in stricter lockdown. Again, seems pretty harsh. Another way of looking at this is that each person who spent a month in slightly stricter lockdown would have saved someone else about 15 hours of healthy life"
Given most states had a period of strict early lockdown and then diverged in the pace of reopening how should I think about the hedonic delta between red-state a blue-state lockdown?
If I can save 15 hours of healthy life by wearing a mask, or going without in-person dining, for a month that's one thing. A month with the near total isolation of the early pandemic is another thing entirely.
It's also difficult to construct the hedonic wedge because of how voluntary the restrictions are. I'm young and healthy and aren't in contact with old people, so once most of my friends figured out that outdoor gatherings were okay my life was pretty normal post summer 2020. My uncle literally did not leave his house for any reason from April 2020 until he was vaccinated, but he's generally paranoid and risk averse. Would either of our experiences have changed under different state laws?
I spent March to December of 2020 in Bryan, TX, and then relocated to Austin, TX, from January to now (and will unfortunately have to go back next month). While I was in Bryan, I sort of felt like I was in a spaceship, where I was isolated from everyone and couldn't do much of anything, though I was able to go for bike rides around the neighborhood. While I've been in Austin, it's felt like things are nearly normal, because there are so many restaurants with outdoor patios and edible food, and outdoor activities to do with friends. (Though I had to develop a habit of bringing a mask with me when I left the house - in Bryan it was irrelevant because there was nowhere with other people that felt relevant to visit.)
Excellent post, which mirrors a lot of my thinking about lockdowns. But I think this:
2: If Sweden had a stronger lockdown more like those of other European countries, it probably could have reduced its death rate by 50-80%, saving 2,500+ lives.
Cannot be squared with this:
3: On a very naïve comparison, US states with stricter lockdowns had about 20% lower death rates than states with weaker ones, and about 0.6% more GDP decline. There are high error bars on both those estimates.
Unless Sweden was extraordinarily more open than even the reddest of states, but it was not:
https://cdn.substack.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F72c14dca-59a1-4dfe-acff-66dbab0522b2_1157x787.png
(Unless even the bluest of states looked far more lax than Denmark or the UK.)
An alternative explanation is that lockdowns in Sweden impacted behavior significantly more than in the US. (Perhaps US people were more inclined to self-isolate without the lockdown than Swedes, because COVID hit most of the US later, and people were more primed? Or perhaps US people were more likely to circumvent or ignore the mandated lockdown? Either would explain it.)
Or R in Sweden of 2019 would be higher than R in the US of 2019.
I think you mean 2020, but agree. If they got hit hard early, when the CFR was higher, that could explain both why Sweden has done poorly and why a better lockdown earlier on would have had a big impact.
I meant 2019, as in "normal precovid life".
I mean this as a third alternative to BM´s parenthesis - let´s say, arguendo, that people in the US and Sweden are equally likely to self-isolate without the lockdown, and they are equally willing to follow lockdown restrictions. But perhaps Swedish pre-covid lifestyle was more different from lockdown conditions than American pre-covid lifestyle.
Gotcha. That makes sense.
I think that misstates the assertion, somewhat, which is that if Sweden locked down earlier, its lockdown would have 3-4 times the effectiveness of the lockdowns in the united states. That strikes me as a pretty big lift without any supporting data.
What did you think of the section "Pre-Conclusion I: Trying To Reconcile European And US Estimates"?
I definitely agree that changes early probably mattered more, as evidenced by the countries that managed to stave off a Corona toehold, and because the CFR dropped as the pandemic continued, https://www.cebm.net/covid-19/declining-covid-19-case-fatality-rates-across-all-ages-analysis-of-german-data/. But we still have the troubling second wave in Sweden, that other Nordic nations either avoided or minimized.
I read this great article about how certain systems have big swings based on small inputs (there is a name for this phenomena, but I can't recall it and did not bookmark the study), and speculates that COVID is like this. I.e., even if you start Sweden and Norway on the same basic course, odds are their experience will still deviate sharply. Think Plinko from the price is right, you drop the token in the middle and sometimes it ends up on the far right and sometimes on the far left, based on undiscernible differences.
It looks like the biggest thing a lockdown does is to "give cover" for things that people want to do anyway.
My employer (in Ontario, Canada) sent everyone home that could work from home on March 17. Was thinking about working from home anyway (I do IT stuff relating to cloud, so it's not like I need to be in a cubicle), but this meant I could just do it instead of having to arrange things.
From what I hear, the "return to office" will be based on how much of one's work needs to be done in a cubicle, so I suspect that I will be in the last cohort to return.
That bat is seriously cute.
If you want cute bats, try Honduran White Bats. They are excessively adorable.
https://en.wikipedia.org/wiki/Honduran_white_bat
"The Honduran white bat is one of approximately 22 known species of bats that roost within leaf "tents." The Honduran white bat cuts the side veins extending out from the midrib of the large leaves of the Heliconia plant causing them to fold down to form a tent.
...It clings to the roof of its tent in small colonies of 1-15 individuals. The tent protects it from rain and predators. Rather than roosting in a single tent consistently, the Honduran white bat has a network of tents scattered across the forest; it alternates among these tents for roosting."
Photo of roosting bats in leaf tent:
https://i.redd.it/yyy2p0liscz61.jpg
Yeesh. Chacon a son gout.
You don't think they're adorable? I admit, I have had my opinions towards bats changed to the point that now when games etc. have the usual "flurry of bats out of cave" as a jump scare, now I go "sweet flittermice" 😀
But tiny bats looking like balls of cotton wool all cuddled up together in their leaf tent don't make you go "awww?"
You're right, they are so cute it should almost be criminal. I want one!
I know! Sell your diet pill with these types of cute otter videos and customers will be beating your door down thrusting fistfuls of fivers in your face demanding your product! 😁
Whoops, wrong reply. But diet pill videos with cute otters and Honduran Leaf Bats will make a killing, I tell you, a killing!
One minor quibble about the SF Bay area Mobility Trends graph — people seem to forget that the SF Bay area counties were three weeks ahead of the rest of California in locking down. It's a mistake to say "It’s pretty evident from this graph that people were starting to decrease their mobility before any official government action." Mayor London Breed of San Francisco declared a state of emergency in SF on Feb 25th 2020, three weeks before Governor Newsom made his shelter in place announcement. Likewise, Santa Clara County asked employers to have their employees work from that last week February. And all the bay area counties coordinated their response to the early outbreak. I started tracking and graphing the case numbers between SF Bay area and LA, and the three-week head start that SF Bay area had made a big difference in the overall case loads between the two metro areas.
This applies to other jurisdictions and metrics as well. For instance, some regions of Italy required masks before the national government did. I don't know what date the chart marks as "when Italy mandated masks"; I presume the date when the national government did.
The premise behind lockdowns was not to reduce total infections, but rather to flatten the curve so that the hospitals were not overwhelmed. That appears to have succeeded in most places. If people are not vaccinated, lockdowns simply wont stop the spread of the virus since as soon as people are free, they will move about.
Different levels of lockdown had less of an effect on US states than European countries because there's no practical way to close a state's borders. While Norwegian customs officers could force foreign arrivals to quarantine at a hotel for ten days, there's no infrastructure stop someone from locked-down California from partying in Las Vegas for the weekend.
One potential confounding variable might be hospital availability. In a hypothetical country with an infinite number of hospital beds, we would expect the death toll to be significantly lower -- though obviously not zero; and thus the effect of lockdowns would be attenuated. In a place like NY or CA (to a lesser extent), where hospitals quickly got saturated, the opposite is the case.
That was a common enough belief in the early days of the pandemic, but then we learned that e.g. ventilators mostly just postponed the inevitable rather than saving anyone's life. For more hospital beds to mean *significantly* lower deaths, you need hospitals to be doing a significant amount of good and to a number of patients exceeding their existing supply of beds.
I'm skeptical that this was true in more than a few isolated and temporary cases, but if someone's got numbers, I'd be interested.
It's not just a question of ventilators, because if all your hospital beds are filled by COVID patients, then regular patients who could've been saved from non-COVID illnesses are now much more likely to die. In California specifically, this had in fact happened in winter.
Was that because patients with serious illnesses were being turned away from full-up ERs, or because patients with serious illnesses were afraid to call an ambulance because they didn't want to get The Coronavirus? I only remember hearing about the latter, but that's media-filtered anecdata so highly suspect.
In California and AFAIK New York, it was definitely the former. That is to say, both situations likely took place, but the limiting factor was definitely the number of hospital beds. Stacking patients on gurneys in the hallway can only take you so far.
Thanks for this. In regards to voluntary vs forced quarantine. I really liked the idea put forward by Po-Shen Loh (I heard on Lex Fridman https://lexfridman.com/po-shen-loh/) And explained some here. https://www.novid.org/ (though not a great web site.) The idea is to let you know when people 'near you' are getting infected and that allows you to decide when it's wise for you to voluntarily quarantine. The key idea to me is that this has all the right feedback.
This idea sounds intriguing - is there any more detailed summary of it?
Sorry all I know is from listening to the podcast and then wandering the website. It tracks local infections, and it (the software) knows your possible neighbors by tracking cell phone towers(?) (I've forgotten the details, he (Po-shen Loh) talks about it in the podcast. He said it worked best when linked with local covid data bases, I think to help control people 'spoofing' the system. (by reporting false positives)
OK looking at the lex podcast on youtube they talk about it for almost an hour starting at the 17:00 minute mark.
https://www.youtube.com/watch?v=6z1JwZbX4dQ
There are significant "blue" populations in red states (particularly in major metro areas) and vice versa. Maybe "states" are just too broad a brush to paint with?
https://www.newyorker.com/news/california-chronicles/what-the-san-francisco-bay-area-can-teach-us-about-fighting-a-pandemic
According to that New Yorker article, you really have to look at communities rather than the states to get a truer picture. If 40% of COVID deaths were nursing home-related, how those communities reacted to the virus would be a compelling variable.
If "blue" counties had higher compliance with lockdown procedures, there is still the variable of high minority populations who worked in industries such as food processing, nursing home care, and other essential industries AND who lived in extended family situations, which may have significantly skewed the "blue" county numbers up.
So, I'd love to see a breakdown of "blue" versus "red" counties. Also, possible to examine COVID-related deaths by political affiliation? That may broadly indicate who was taking lockdowns seriously versus not? Obesity was a significant variable in worse outcomes - could obesity rates by county be correlated?
I'd personally like to see a lot less of blue vs red because it taints the outcomes with confirmation bias. The media has so desperately wanted to paint this in red vs blue that these analyses became unreadable for the most part. Cuomo the hero versus DeSantis the villain was actively being marketed while NY had 6x the death rate of Florida, which has evened out over time.
The framing is typically blue people are victims of a terrible virus and red people are victims of their own immoral behavior. You know what is actually correlated the most? Being old. I'm sorry to report that this isn't very interesting because red and blue people age the same every year according to my sources, ha ha.
It's not that they haven't tried to prove blue > red over and over and over. The data just doesn't support anything but very weak correlations and you pretty much have to throw out NY to get where you want to go. Covid is equal opportunity and that opportunity just doesn't change much across counties and borders and how you register to vote.
"The framing is typically blue people are victims of a terrible virus and red people are victims of their own immoral behavior."
The correlation between partisanship and vaccination is pretty obvious, at least.
The birthday paper seems relevant. https://twitter.com/sangerkatz/status/1412426320835973132
"at the height of the pandemic, there were not very big differences in private behavior according to party, even if public behavior was different."
The biggest problem here is the assumption that there aren't external natural causes for increases and declines in case rates. There are no lockdowns for the flu and it is quite seasonal and people aren't voluntarily changing behavior for a flu outbreak. The timing of the worst US covid outbreak is nearly coincidental with the flu season. People change their behavior by season but this would still need to be factored out for a fair comparison that was judging either personal change from fear or government mandate.
The seasonality of the flu is not well understood and one assumes parts of the increase and decrease of covid are also unrelated to behavior (perhaps temperature, humidity, etc.).
This seemed to me a glaring omission in of the whole analysis. If the answer to the question "Do lockdowns work?" is a resounding YES*, but the magnitude is, "at about 0.1% of the effect that seasonality has", then there's this big asterisk, saying:
*The direction doesn't matter because the magnitude is so low.
I felt like that's the lesson everyone should have learned back in November, but then we forgot the obvious and went back to sifting through messy data looking to tease out minor effects from a whole lot of noise.
Think about it: after months of lockdowns, and despite no significant behavioral/policy change, the entire US experienced a massive, sustained outbreak of COVID-19. That outbreak didn't care about state borders or political parties. It's not that we were suddenly doing something across-the-board different in November that caused case rates to go up.
I predicted back in March 2020 that (if COVID-19 is seasonal) case rates would naturally go down on their own by May. They did. I then predicted that they would come back in November. They did. This isn't sophisticated analysis, but the magnitude of the effect dwarfs the squint-to-see-an-effect complex analyses for the lockdowns. Even I was surprised by how little impact lockdowns and other measures clearly had relative to the seasonality effect.
Incidentally, when my prediction in November came true, I further predicted that the following March/April we'd see case numbers fall again (which of course they did). However, I added a prediction that people would perceive the vaccine as the cause of the falling case rate, and that people would once again ignore seasonality. That last prediction is my most depressingly accurate, despite the evidence that the vaccine likely had very little impact on resolving the outbreak. (https://www.marklwebb.com/2021/04/is-it-over-yet.html)
I feel like this is a lesson we're incapable of learning at this point. Especially now that it's clear COVID-19 is endemic, we still have people who should know better (including Scott) saying they're hopeful the pandemic is at an end. Seasonality is not a hard lesson to learn, but I think there's this illusion of control we want to maintain, so we ignore the big things so we can focus on the smaller ones. We don't want to admit that all these heated debates were moot from the start, because we want to believe we can control a global pandemic. Why did we ever think that was possible in the first place?
It's not exactly a popular opinion but I am quite disappointed in science over the past 1.5 years (with the strong exception of vaccines). They struggle with the very basics of disease transmission and personal risk reduction even where there are millions of cases. I feel like I'm walking out of the dentist office after 40 years of treatment with a toothbrush and floss ... "this is really all you got here?". We are dumping trillions into this, is it really that hard to prove surface transmission wasn't a big factor? Airborne and aerosol transmission took how long? Over dispersion took a year to figure out?
How well do masks actually protect you? Nobody seems to really know. Does anyone care about HVAC at all? Seems important. There just seems to be a lot of arm waving, most of it in the right general direction, but not confidence inspiring and lacking specifics.
If homes are the biggest vector of disease transmission, why didn't anyone recommend wearing a mask in a house? Isn't riding in a car with someone like the worst case scenario? An on and on.
The deaths were a majority in old age homes that weren't able to cope. Something like 81% in Canada. Something like half in Sweden of all deaths were in old age homes. 42% in the US. Belgium 42%. 64% Norway. Spain 52% How the old were handled were likely the most important factor. How soon these were locked down and managed.
If we can batch half of them that way, they really should be broken out.
"If there had been looser lockdowns, more children would have had to go to school, which would have been either good or bad depending on how you feel about school."
This is a throwaway line in the essay and not the main point, but I think it undersells the likely importance of this decision.
The mainstream economic consensus is that each year of schooling increases lifetime income by 10%. In the United States this equates to an economic cost of ~100K (in net present value).
We have absolutely no idea how much worse virtual school is than in person education, but even a conservative estimate of 10% would equate to a cost of $700B. Dividing that in half (to only apply to the blue states) would mean that the costs of closing schools were more than 6 times higher than the direct economic impacts.
> The mainstream economic consensus is that each year of schooling increases lifetime income by 10%.
I'm not sure this is quite getting at the same thing. My understanding is that the economic consensus is about additional years of schooling vs no schooling at the end of one's education. That is, it's about the difference between dropping out as a junior and sticking around for senior year of high school, or the difference between going to a year of college or going directly to the workforce, or the difference between sticking for the second year of a PhD program rather than dropping out after the first.
But the covid school closures are more like no school for younger students than for older students. That is, a second grader who "attends school online" gets basically nothing compared to a second grader who attends in person. But it seems likely that a high school senior who attends school online is getting most of the same benefit as a high school senior who attends in person. And a second year PhD student is really getting almost exactly the same experience online as in person, unless their training involves a heavy lab component (which they probably still did this year, despite shut downs).
Maybe you don't know that Scott hated school. I think he's saying that he, as a kid, would have loved the lock down.
Well, you also need to assume people can't make it up later. Meaning, the education for the kids now continues at the same rate for the same length of time, regardless of the fact that a year has been chopped out (or made lower quality). That is, we need to assume nobody compensates by staying in school a little bit longer, or working a little harder for the next year or two. I guess we could assume that education is already delivered as fast as the human mind can accept it, so no catch-up is possible.
Those are reasonable assumptions, and by "reasonable" I mean "near-certain". We are not going to be creating a special Thirteenth Grade for just this cohort of students. Schools and teachers are going to be extra forgiving of students whose educational progress was disrupted by the lockdowns, not extra demanding. The parents who hypothetically would be extra demanding and make their kids work harder the next year of two, are mostly the ones who supplemented Zoom-schooling with a dose of home-schooling so their kids mostly don't need remedial education now.
And all the standardized tests will be rescaled for that cohort? So nobody gets bad news on the PSAT (if that's still a thing) and thinks -- geezus, I better buckle down if I want to get into college at all?
I tend to think most long-lived systems are in stable equilibrium a priori, which means they have restoring forces when perturbed. There's a whole lot of "thermometers" poked into the educational process these days, to see how things are going. I have a hard time believing either (1) they'll all be rescaled, or (2) nobody will pay any attention to the feedback and adjust.
I think it's very reasonable to believe there will be *some* educational loss -- indeed, I've seen it in my own school-age kids, and in some cases, like seniors in high-school with a lot of lab-oriented or otherwise hands-on/direct contact projects, these losses could be quite significant and personally derailing. But I think the analysis that says the system (schools, teachers, parents, students) will not constructively respond *at all* to the loss is too naive to be of much worth.
In your GDP vs lockdown analysis, what are the units for GDP? I know its decrease in GDP but is that a raw value or a percentage of GDP for some control day?
Pretty disappointed that you missed the most obvious confounder in all this discussion. Quote:
*** This is an absolutely beautiful graph. It’s showing how lockdown strictness (as of May 5) correlates with death rate over time. We find that early in the epidemic, the stricter your lockdown, the worse you're doing. This is the endogeneity - places (like NYC) that are doing really badly institute strict lockdowns to try to save themselves. Later in the epidemic, the stricter your lockdown, the better you're doing - probably because the strict lockdown is giving good results. ***
OR it could be that the states that got hit early had partial herd immunity because of getting hit early. This continually got missed by everyone all of 2020 and I'm sad you missed it today.
He more or less addresses it with:
>Are there other explanations? I wondered if states that did worse in the first wave might have done better in the second because most of their vulnerable residents had already been infected/killed. But I don’t think this is a big problem; most places had only about 5% of the population infected during the first wave, and even the worst states only about 10%.
The first wave went through most of the USA before we had tests to be able to track it.
I'm pretty sure these 5% and 10% numbers involve antibody tests taken later, though it's possible that there was a methodological problem at precisely this point.
If you can point me to a quality population level serology study in June I'd love to read it, but I scoured the earth looking for one, presuming the CDC would have it in their best interests to generate something like that, and never saw it. The infection wave in Spring 2020 in Atlanta was way higher than 5%. More like 20%. It's just that nobody knew because nobody had tests, and even people who knew for sure they had it would go to the hospital and get turned away for a test because of test rationing.
If way more people had it than knew about it very early, when there were also few deaths, then later on, when testing *was* widely available and deployed, and there were also lots more deaths, then that would imply that the death-to-cases ratio significantly *rose* from the beginning to the end of the pandemic. That's a very counter-intuitive result, since we know treatments improved.
Also, could you define what you mean by "partial" herd immunity? I've never heard that term. The usual thinking behind "herd" immunity is that it's like a percolation problem, e.g. water seeping through cracks in concrete. If you have enough impermeable concrete, then the cracks don't form a continuous path and the water can't get through. If you don't, the cracks do, and the water gets through. But it's kind of an all-or-nothing thing, there's no such thing as the concrete being "partially" leaky. It's either leaky or it's not.
Same thing with pandemics: either the rate of general "impermeability" (immunity) is high enough that there aren't any continuous paths of transmission that connects almost every vulnerable person to an infected person, or there are. There doesn't seem to be any in-between "partial" state, which is why I'm puzzled by what you mean here.
The epidemiological models of pathogen spread are all fucked. They presume even mixing of a population in a Petri dish, when in fact it's a lot more like a virus spreading through a computer network. Ask yourself which case provides more net 'immunity' to a population:
Forced vaccination of a hermit who lives on a hill, or
The stripper at the titty bar catches covid and then gets over it
Early infection is most likely to run through the people who have the highest exposure to other people, so each infection and recovery among that network gives you more juice than vaxing some dude who lives in the sticks.
If you look at deaths/day, the March 2020 wave in the USA was about the same as the December 2020 wave. But the spring 2020 wave was undercounted because of test rationing, and the December wave was overcounted because testing was so ubiquitous. The two waves were comparable, and the spring wave was probably a lot worse in hot spots. The New York deaths curve was insane in April (1k/day) and only peaked out at 200/day in January. New York's rate over the summer of 2020 was almost exactly the same as it is today after mass vaccination. New York *definitely* had herd immunity in summer of 2020, and they got their new winter wave after the immunity wore off.
New York had 25k deaths by the end of May, and that was an undercount (I believe excess deaths in New York were probably ~50% higher, but you'll have to take my word for it right now). That implies somewhere in the neighborhood of 5-6 million infections in a state of 19 million people. That's a pretty chunky break on further outbreaks, though they did still get hit (albeit milder) over the winter.
Quick math - the big midwestern outbreaks in the winter seemed to grow for roughly 8 weeks. Let's say R was 2.0 during those outbreaks in the midwest. In NY with prior immunity of 30%, R is 1.4.
If both areas start with 1000 infections, at the end of the outbreak you'll have 250k infections and 1500 deaths a week in the Midwest, and 15k infections and 90 deaths a week in New York. And the only difference is it's the first real wave in the Midwest and the 2nd in New York
Be fair, part of the high NY count was the nitwits in charge throwing infected people back into nursing homes. But I generally agree with what you're saying - real infection rate in NY was much much higher than the numbers we have because of test rationing and lack of testing at all. Any analysis which doesn't take into account the lack of testing in spring 2020 is a bad analysis, which means all good analyses require a significant amount of guesswork.
The biggest problem with this post is that the range is too limited. The post does not mention the most obvious lockdown failures: India (which had the strictest lockdown) and Peru (which had the highest COVID death rate). It also does not mention the fairly lockdown-light East Asian success stories.
"In the end, Sweden still ended out with a death rate about double the European average. Seems pretty bad."
Objection: there's no such thing as a European average pre-second wave because Eastern and Central Europe was spared. In the end, Sweden ended up having a lower COVID death rate than most Eastern Europe.
"The real question we should be asking is what set of policies countries should have implemented."
Bingo.
"Contra Lemoine’s picture where Sweden just has an earlier start but eventually does no worse than everywhere else, here Sweden has the same (or better) start as everyone else, but clearly does worse afterwards."
Agreed.
"If Sweden had a stronger lockdown more like those of other European countries, it probably could have reduced its death rate by 50-80%, saving 2,500+ lives."
This is a dubious conclusion; it's only true for the first wave, if it's true at all. But there is a difference between preventing deaths and delaying them; Bulgaria had hardly any Spring 2020 wave at all, after its winter and spring waves it has the second-highest COVID death rate in the world.
Any analysis of lockdowns need to tease out nursing home policy. Sweden was blasted over it, so were New York and Pennsylvania. In my PA county to this day, 64% of deaths were nursing home residents. No amount of lockdowns was going to salvage a policy of stuffing confirmed COVID-positive patients back into poorly ventilated nursing homes
> Nobody agrees on exactly what “Scandinavian/Nordic” is
There appears to be a consensus that "Nordic" means exactly the five countries you mentioned, and it's just "Scandinavia" that's ambiguous, with it most often referring to Sweden, Norway, and Denmark, but sometimes also excluding Denmark, or being used synonymously with Nordic countries, as you do here. At least, that's what wikipedia says, and the maps I find by searching those terms on google images seems to support it.
Thanks for article!
Extremely minor point: DOT currently uses $11.6 million as the value of a statistical life, not $9.6 million. https://www.transportation.gov/office-policy/transportation-policy/revised-departmental-guidance-on-valuation-of-a-statistical-life-in-economic-analysis
Excess fatality tells a different story still. Personally, I'm content in knowing that I'll never know the effectiveness of various measures. Can you ever really know? If the deaths had been really bad, who knows we might have freaked out enough to start taking vaccines, try vit. D/ivm, let people test themselves, etc.
It's been unfortunate to see the discussion framed in terms of data rather than principles. If government can do X if data is Y then moral hazard naturally increases greatly as the measurement error in Y increases. I'd much rather attempt to reach consensus on the moral/philosophical framework.
" Sweden... could have reduced its death rate by 50-80%, saving 2,500+ lives."
How many of those lives were not of folks w/ one foot already in the grave, e.g. w/ major co-morbidities?
Insofar as the presence of co-morbidities matters to policy assessments, why don't we see data (e.g. here), distinguishing #s of deaths, of folks with, vs. w/o, major co-morbidities?
The paper linked to here, on that "the average COVID victim *might* have lived another 8 QALYs", is quite a "might", seeing as it admits that "the SMR and reduction on remaining quality of life are not empirically estimated....".
Some data from the first wave (May 2020): 90% of fatalities were 70+, of which 50% people were living in retirement homes and 26% had home care services. For various reasons, Swedes are really old and fragile when they move to retirement homes. The average age of people moving to retirements home is 86, and about half die within a year (pre-covid). So I find it hard to believe that Swedish covid-19 victims would on average have lived another 8 years.
Sources (in Swedish):
https://www.socialstyrelsen.se/om-socialstyrelsen/pressrum/press/ny-statistik-om-smittade-och-avlidna-i-covid-19-70-ar-och-aldre/
https://www.dagensarena.se/innehall/darfor-skiljer-sig-dodligheten-nordens-aldrevard/
Can someone help me evaluate the Leech study on mask wearing? They detect a significant reduction in R from mask use, but they use a seemingly super-complex hierarchical bayesian model which always makes me suspicious. Does the data really support their conclusion or are they hacking their model somehow? I've seen other observational studies that don't correlate mask usage with R reduction, lab studies that show no reduction in viral cultures from mask wearers, and older data which indicate masks don't reduce influenza transmission. Can someone help me reconcile these data? I'm sure n95 masks help, but in my experience it doesn't seem like anyone wears those and I am highly skeptical that the average cloth mask worn by the average person could possibly do anything.
All the mask studies I have seen are overly complex which brings in greater opportunity of confirmation bias. One should ask why nobody simply asked people who tested positive or were hospitalized whether they were reliably wearing masks. 1000 interviews should get you some decent data. This would also have plenty of confounders but at least the starting point is more robust.
I finally just threw up my hands and assumed masks likely help but likely not as much as the preferred narrative would suggest, and they were not a huge burden so go ahead and wear them. My wild guess is they reduce accumulated risk by 10% to 50%.
The media seemed to be overly careful to never show or suggest that mask wearers were getting infected.
Thanks. Interestingly, LessWrong posted about this very study and was skeptical of the result: https://www.lesswrong.com/posts/fSNp942XBztB352fn/we-still-don-t-know-if-masks-work
My guess is cloth masks do very little but the political valance has created a strong publication bias whereby no one wants to criticize their use.
You came to the same conclusion I did, with a similar range. If I had to pick a real number instead of a range, it would probably be about 20% effectiveness. 20% is still huge, but a lot of people treated it like the most important factor, or singlehandedly enough to stave off the virus. It wasn't, and couldn't be.
The funny/sad thing is, that there are a lot of people both then and now that treat a mask like PPE, to protect the wearer of the mask. Masks were never billed as an actual way to keep from getting the virus. They were instead a means to not spread it to others. Mask wearers get the virus at about the same rate as everyone else (maybe some tiny percent less or something, but we would never be able to prove that). If you are only around other people wearing masks, you got some protection from them, if they had COVID.
My intuition has it that masks ought to be at least moderately effective based on theoretical reasons, some weak scientific evidence, as well as rational evidence. One such piece of evidence would be healthcare workers working with COVID-19 patients, as far as I can tell, using surgical masks pretty much across the globe even in places where you wouldn't expect this to be due to shortage of equipment. I can find studies that predictably show moving to aerosol-resistant PPE reduces infections by a large factor (from nonzero to zero), but I would expect unmitigated risk to be near-1, and apparently practical experience has it that surgical masks have been deemed "good enough" even in context that to me appears to be maximally risky and cost-benefit of greater protection being as high as it's going to be. Surely the hospitals in these countless naturally running experiments would notice if their workers were getting infected en masse? (Now, I do realize that you could interpret this the other way: after all, studies show that healthcare workers sometimes DO get infected by the patients, and I make assumptions like supposing treating patients is high-risk or assuming hospitals would notice if nurses were getting infected en masse, but by my read this is one of the higher-signal pieces of evidence and I interpret it as showing even mere surgical masks offering high degree of protection - if anyone can tell me why experience in hospitals does show surgical masks to lack any efficacy of note, I'd be interested to know. Perhaps hospitals tend to have such good ventilation that even working with patients is low-risk?)
Be that as it may, the key point here, I think, is that nobody wears masks all the time. People take them off when at home or with friends, when they eat or drink, and in most workplaces. Since the bulk of infections take place in these sorts of contexts, not in shops or public transport, even 100% effective masks worn with 100% compliance couldn't reduce the total number of infections by more than some low tens of percents, and as such finding real signal by conducting studies based on asking people if they had been wearing masks would already be incredibly difficult. Maybe if you only surveyed people whose infections had been contact-traced to situations where mask-wearing is more of a norm?
Your analysis of Sweden is flawed as you're using doctored stringency data. Go to an earlier version of the data and see how much they bumped Sweden's stringency index up by. They claimed they did this because of some vague guidance that was in place (rather than laws or mandates). Their data is silly as it suggests Norway and Finland were even less stringent than Sweden
Someone covered it on Twitter here: https://twitter.com/TTBikeFit/status/1359911243147018243
Based on what I've heard and read from Twitter etc, Finland genuinely was less stringent than Sweden for at least some periods of time after last year's spring.
Second to this. "Sweden is relaxed" was a meme, one repeated by me on the old blog, but that's different than it being true.
Here is a theory of lockdown: lockdowns can help people co-ordinate. If the virus is in the community and I don't think people will keep R far enough below 1 to eliminate it, then I have no reason to take avoidance measures beyond protecting my own health. Because infection risk increases with disease prevalence, everyone taking "personal risk measures" will always be insufficient on average to drive R below 1 if the prevalence is low enough (though they may be sufficient once the prevalence rises a lot) and so the disease will persist with cyclic pattern that depends on how quickly people re-evaluate personal risk based on changes in community prevalence.
However, with a lockdown aimed at elimination I can believe that other people will keep R down enough for elimination and I can now weigh the risks I take against the costly possibilities of compromising elimination or extending the lockdown instead of the less costly possibility of personally dealing with getting sick. Thus, with a lockdown, even under my own evaluation I should accept a much lower risk of infection. There is still a prevalence below which I won't take many precautions - but it is much lower, possibly lower than 1 case in the entire country in which case I can take precautions until elimination is probably achieved.
This is a very different theory of lockdown to "disease burden management", though! Also contradicts the view I advocated early last year which was that 12 months of suppression is a reasonable aim.
> nobody responds to the virus in any way and it spreads uncontrollably until 70% of the population has been infected.
70% is the threshold for herd immunity in a naive SIR model with an R of 3.33, but due to the phenomenon of "overshoot", the virus would infect 96% of the population before petering out. The "herd immunity threshold" means the reproduction number of the virus goes down to one (not zero). When that threshold is crossed, there is still a very large population of infectious people, each of whom infects one other.
The equation for the proportion (p) of a population infected by a virus with reproductive number R at the end state of an epidemic satisfies p = 1 - e^(-p*r). So once overshoot is considered, and assuming no intervention, the R of the disease which infects 70% of the population before sputtering out is about 1.7.
Source (with useful graph): https://twitter.com/CT_Bergstrom/status/1252008428542681088
Corona Game author here. I recommend using URL https://covidgame.info/ that loads the game in English by default. BTW for some more context, I commented on: https://news.ycombinator.com/item?id=27763873
I'm not sure what to make of the mobility data. The rest of the data seems to suggest that lockdowns make a difference beyond what people did on their own, but the mobility data suggests that the government was basically telling people to keep doing what they were already doing. Maybe things like "ban on large gatherings" don't show up in mobility data (not that many people are going to conventions), but make a big difference in the spread? Maybe making the lockdown involuntary prevents things like "your boss tells you to come in even if you're sick"?
Woohoo, I made the Pareto frontier! 1.8k deaths, 399B Kč: https://i.imgur.com/9DmHRuY.png
Strategy spoilers:
- Max 10 & masks/distancing stay on the whole time
- Never pay compensation
- Close high risk services & universities until the first summer
- Once summer comes, reopen services / schools
- When summer ends, close high risk services and universities again
- Once it's winter again, close all schools
- DO NOT reopen ski resorts
- Punish the anti-mask celebrity
- Don't invest in vaccination campaigns
- Do the cover up ¯\_(ツ)_/¯
Your link to "value of one life" per the EPA doesn't link to the EPA. It instead links to a Statista article on "Percent change in Real Gross Domestic Product (GDP) of the United States from preceding period in 2020, by state", which I presume is an error?
The data you looked at, how you looked at it and the various ways of interpreting the data, and, of course, the questions you asked were fascinating and helpful. I can now think of this subject in more enjoyable ways. It's truly a breeze not having all the answers while maintaining methods of attaining them someday. It must be truly stressful needing an answer.
>and even the worst states only about 10%
I believe serology tests in NYC suggest the number there was about 15%, I'm not sure why this shouldn't be a big deal because (to simplify massively) if all your super-spreader types are in that 15% of your population, you now have no more super-spreading events to worry about.
>This is a victory for lockdowns insofar as the correlation is significant, but strong proponents might be surprised by how small the effect was.
It would be even smaller if you just drew the straight line through the main cluster instead of calculating a least squares fit. Also "correlation is significant" isn't really appropriate here, you can't just feed your data into something that then spits out the words "significant" or "not significant" and then expect it to mean something; "significant" and "not significant" is *always* tied to a statistical model which you do not have.
Would it be arbitrary and sill and not at all mathematical to just draw the straight line by hand? Yes. Would it be mathematically any worse than what you are doing now? No. If you're doing a least-squares fit on data like this you already statistically in a state of sin.
> US states mostly had stable and predictable responses based on their internal politics - a nice exogenous factor!
Which correlates with the amount of voluntary changes people make; what do you think would happen if you tried to control for those?
A couple of thoughts:
a) In the discussion about lockdowns vs voluntary behaviour change the two are considered to have a complicated relationship but still to some degree independent. I think there's also a possibility that voluntary behaviour changes *because* people can hear their government debate lockdowns and can see it might be needed. I.e. the political discussion about lockdowns is also part of the pressure on voluntary change.
b) The death rate calculation is only looking at people dying of Covid. In the UK we've also seen that hospitals full of covid patients have to move resources away from other activities. This gives rise to an ever-growing waiting list for other operations. So you have lower standards of daily life for hundreds of thousands of people and many more deaths. The full death toll, and the 'how many good years saved,' calculations really ought to include people waiting more than a year for gallbladder operations etc.
2 good points, I want to add 2 examples for the first point.
_ Democrats tended to be more responsible than Republicans because their leader favored lockdowns and asked/implemented them while Republican leaders were more of "let us continue to live as if nothing has changed".
_ If you know your leader is going to institute a lockdown in a few days, you do some panic shopping and then put yourself in lockdown mode before the official date because this is a sign that the situation is dire.
In summary point 3, shouldn't that be 0.6%pts more GDP decline, not 0.6%? If the latter then that's almost nothing, but I gather from the data and discussion earlier it's a 0.6% point difference in the change in GDP.
Small note: I think looking at excess deaths is much better than relying on self-reported stats. Eurostat has a tool where one can see excess deaths (easily found via searchsites). It would not change Sweden's general position as doing worst in the Nordics, but would increase the distance between itself and other Western European countries.
One should also make note of the fact that Sweden is by far the most diverse Nordic country, and deaths in all Western countries have been concentrated among immigrant-heavy populations, partly for cultural reasons (high prevalence of multi-generational homes) but also socioeconomic (disproportionately the poorest groups).
At least early on, I think the Swedish focus was clearly more on mitigation than suppression. It seems quite clear that there is a link between that strategy and the high number of covid fatalities in the first wave, as you point out. However, one could argue that the strategy was _partly_ vindicated if you consider the cumulative deaths over the _entire period_ up until now. While Sweden was 2x above EU average by August 2020 in terms of cumulative confirmed covid-19 deaths per million people (as you pointed out), by now it's actually <10% _below_ EU average. Does this mean that Swedish chose a smarter strategy for 2nd and 3rd wave? That there is a natural "reversion to the mean" over time, as most governments were unable to protect the most fragile people for more than a year until the vaccines arrived? I don't know, but it would be interesting to hear your thoughts on it.
https://ourworldindata.org/explorers/coronavirus-data-explorer?zoomToSelection=true&time=2020-03-01..latest&pickerSort=asc&pickerMetric=location&Metric=Confirmed+deaths&Interval=Cumulative&Relative+to+Population=true&Align+outbreaks=false&country=USA~SWE~GBR~European+Union
Disclaimer: I'm Swedish ;)
Thank you for bringing this up. I hate when people take a tiny timeframe and act like its the whole picture.
In a sense the analysis here is for "we tried X, it mostly didn't work, was it still worth it?" But that doesn't seem like a fair analysis to answer the question "is X worth it"?
In my memory the trade-off at the time was between "can we flatten the curve or do we need a lockdown?" Where "flattening the curve" meant staying at home if you're sick, social distancing when you're out, not shaking hands, don't touch your face etc, it was seen as the more gentle, cheaper approach that could get us through without overwhelming the hospitals. The alternative was a "lockdown" where everyone stays at home and you close businesses, which was more expensive, but was going to give us similar results as in China, where you completely get the virus under control, at least enough to go back to contact tracing.
Importantly this means that we didn't get the benefits that we were promised. We got the costs of a lockdown with the benefits of "flattening the curve."
If these were different terms at some point ("lockdown" and "flatten the curve" got mixed together very quickly after lockdowns started, as if they used to not be competing opposite alternatives...) then the calculation is very different, because we didn't expect the outcome that we got. In hindsight it's easy to say "we didn't get the Chinese results in the US and Europe, so overall it's an expensive measure that didn't get us many benefits" but you couldn't have known that at the time.
Another example may be the collapsed apartment building in Florida recently. Imagine you're one of the people who just bought an apartment in there a few months ago. You bought an apartment, the building collapsed, so now you either don't have an apartment or you're dead. Does that mean that buying apartments is a bad idea? No, because at the time you didn't know that the building would collapse. You have to judge the decision based on what you knew at the time, and usually buying apartments is a good idea.
Of course once people started sabotaging the lockdowns, (most obviously Trump, but also Cuomo keeping construction going...) and once the goal shifted to "we're just trying to flatten the curve" you could have argued that that doesn't make sense any more. And I never heard a good justification for the later lockdowns in Europe.
But before the sabotage we could only make the decision based on the results we saw in China. And some countries got the same benefits as China.
In the next pandemic, if people argue for another lockdown, what will we decide based on? Will we decide based on how lockdowns went in countries that messed it up last time, or based on the countries where the lockdown worked last time? I predict it'll be the latter. People will say "we should do a lockdown, and we should do one that works, because it can rapidly end the pandemic." If that's true, and if people aren't interested in dragging the next pandemic out like we did this one, (see the strong reaction in Asia this time, who had the prior experience of SARS) then they won't consult the analysis in this blog post, because it's only looking at cases where the lockdown mostly didn't work.
> If we had let the virus spread more, it would have gotten more chances to mutate.
Is this so certain? I'm thinking that, given the same number of total people infected, flattening the curve and keeping R on average around one for a longer time is much much worse regarding giving the virus maximum selective pressure, and it probably more than makes up for the reduced total number of people infected.
Also, let's irresponsibly assume stuff. Let's say that, doing nothing, 2% of the population would have died within the first 6 months and then herd immunity. And that each death killed 10 years of life. So the pandemic ended within 6 months, reducing life expenctancy by 2% of 10 years = 70 days. As a citizen of Spain (life expectancy of 83.2 years), you would need >23 covid pandemics in your lifetime to turn your country into the US (life expectancy 78.5).
The damage would not be spread out evenly among all people, any more than if my neighbors house burns to the ground all of us on the street experience it as an average slight singing of our furniture. Even the risk wouldn't be evenly spread, as the it is highly concentrated among the old.
So if nothing at all had been done, the actual outcome would be that almost nothing would happen to everyone under 40, those in their late 50 and early 60s would've experienced what people in their 80s do during a really bad flu season -- something like a 1 in 20 culling -- and among those over 70-75 it would have been close to the Black Death, something like a 20-35% mortality.
That is, deliberately doing nothing would be the functional equivalent of just writing off all of the very old folks, pushing them out on the ice floe.
I'm not making the case that it would have been a good choice to let it run. It's just another angle to put the whole thing into perspective.
But yeah, it's uneven. Though the extra deaths in normal US versus normal Spain also disproportionately affect the old. And, for a fixed number of life years lost, having them affect the old who have less years left to live means is actually a way to distribute it about as much as possible.
I also don't agree
For US comparisons it would be nice to look only at what types of businesses are allowed to be open (also if we were able to know what fraction of eligible businesses were open as a measure of voluntary effect) since that's really the area the government has the most control over (since we did not use the police to break up house parties or give fines to individuals for walking outside etc)
Truly much more than I wanted to know. I lost interest midway through article and that doesn't happen often to me on ACT
I don't think Conclusion #2 (Quantifying Emotional Costs) gets enough attention. In my opinion, this is perhaps the most important trade-off. I lived in Idaho (with family in Spain) through this whole thing and I can't think of a better place to have been through it all... I sure was glad I didn't live like a prisoner as in Europe (for a whole year!); in rural parts you wouldn't know there was a virus even during the second wave in the winter.
And all this madness everywhere else just to save a few QALY (at the cost of sacrificing a bunch more QALY)? I think there is a lot more to live than avoiding death. And somehow, Idaho's deaths numbers are some of the better ones too!
Very respectiable stab at looking at the data.
The one aspect which only time will uncover: the economic ill effects from the lockdown are long term. Small businesses destroyed - the health and economic effects are decadal. Even people unemployed but on short term government subsidies - that impact is also decadal.
The entire economics analysis is contingent on a false assumption - that economic activity/growth is independent from one political entity to the next. If all but one country decide to lock down their borders and limit trade, how much of a difference does it make that the one holdout is still willing to play? If all your kid's friends tell him they won't play with him, you can't claim that he's outside "playing with friends" because he's still willing to get together. Economic activity isn't independent, and this assumption should place a giant red asterisk on the whole analysis.
Let's do a counterfactual reasonableness analysis. Let's say nobody knew about COVID-19, because we hadn't invented the technology to test for it, and hospitals just treated it like a particularly bad cold. Cumulative US deaths from COVID-19 for >1year currently stand at <0.2% of the population. Cumulative infections at close to 10%. Is it reasonable to attribute massive GDP changes to the virus alone? (Because <0.2% of people died and <10% got sick for two weeks?)
What if next year a bad variant of H2N3 flu started spreading across the US, but nobody was focused on it because we were too busy watching for more COVID-19. Say 10% of people got sick from it for 2 weeks, but nobody got scared because everyone who got tested showed they were negative for COVID-19. Of those infected, around 600k people died (out of a population of >350million), but they were mostly elderly, so nobody thought it was out of the ordinary. How likely is it that a drop in GDP would have even been measurable under those circumstances?
Now, maybe that thought experiment doesn't exactly encapsulate the difference between voluntary behavior and lockdowns. But it seems clear to me that nearly all of the economic impact from COVID-19 is downstream of behavior, not due to the virus itself.
Your comment could be a blog post itself. solid.
Great post
I think this analysis missed out on the fact that increasing voluntary quarantining decreases the cost of mandatory quarantining.
This is maybe impossible to measure, but I think it would dramatically affect the cost calculations when looking at red vs blue states, as my assumption is that blue states had much higher rates of voluntary quarantining, and so stricter lockdowns were less costly there than they would have been in red states.
I played and earned Czechia probably $10B by killing all the old people who were going to use healthcare, draw from retirement, etc. Not to say that makes much sense, but neither does estimating the "economic costs" while ignoring the largest factor.
A very well written, sober and judicious piece. Nicely done indeed.
I have on quibble, which is with your conclusion #5, in which you compare the cost of emotional suffering by those who endured lockdowns with the benefit of the QUALMs gained by people who don't die of COVID.
I think if you are going to count the *emotional* suffering of those who endure lockdowns on the "cost" side of the ledger, then you have to count the (averted) *emotional* suffering of the friends and family of those who would have died on the "benefit" side.
The emotional cost of a death is not just, and perhaps ultimately not even mostly, borne by the person himself dying -- a death has savage emotional (and even practical, financial, childrearing et cetera) repercussions among the family of the deceased, and these go on for a long time, years often.
...and perhaps I should add that some of these emotional family costs were very heavy indeed during COVID, as often enough the family was not allowed any contact with their father, mother, uncle, grandfather, et cetera while the latter was dying.
No, you shouldn't count the emotional suffering of the family and friends of people who would have died.
Everybody dies. If someone dies from COVID and their family suffers emotionally, you can add in the emotional suffering, but then you need to *subtract* the emotional suffering because they didn't die later. You can only die once, after all; dying earlier just changes the time of death, and friends and family suffer either way. Now, you might argue that people suffer more when their friends and family die younger, which is true, but you can't count that as a whole death of suffering, you need to count it only as a fraction. And if someone dies at an older age, they're more likely to have someone *they* know die (since their life is longer, more time for other people to die in), which compensates for that.
Not to mention that COVID mostly kills old people anyway.
A strange point of view. So it wouldn't matter to you if your own daughter died at age 4 from Wilm's tumor, because she's just going to die anyway someday, perhaps at age 77 from a heart attack? Maybe it's even efficient, you get your mourning over with early on, and can move on with the rest of your life.
Always interesting to see what a wide variation in human nature there is, so thanks for this insight.
The point is that you can't save lives. You can only save life-years. If someone dies at 110 of COVID then they were probably going to die in the next year anyway. If someone dies at 15 of COVID then that's a bigger loss.
(Even biological immortality can't save lives. Accidents, murder, famine and war have a nonzero chance of killing you per year, and that inexorably adds up. It *can* increase the life expectancy to hundreds of years, but not infinity.)
Shouldn't that just be priced in as part of the cost of a death? I don't see why deaths from COVID should be treated differently from any other kind of death in this particular respect.
One point I'd make is that decisions by politicians have ramifications even where the leader's writ doesn't run. For example, on March 15, 2020, the mayor of Los Angeles shut all the movie theaters in Los Angeles, the movie capital of the world. I suspect that had impact around the world because you'd expect the mayor of Los Angeles to be in favor of people thinking of going to the movies as a fun and safe thing to do for the whole family, kind of like how the mayor of Amityville in "Jaws" is pro-going-to-the-beach. But suddenly Mayor Garcetti was announcing, in effect, that going to the movies could KILL you, so do not, by order of law, partake of Los Angeles' most famous product.
In contrast, the prime minister of Sweden did not shut Sweden's movie theaters. But very rapidly, nobody was going to Sweden's movie theaters so they shut themselves down for lack of business.
Ironically, the payoff from lockdowns in America is mostly due to the almost miraculously effective mRNA vaccines riding to the rescue faster than just about anybody other than the ever-optimistic Mr. Trump imagined. If instead we got, say, some 75% efficacy vaccines in late 2021 with supply chain problems keeping everybody who wanted a vaccine from getting one before 2023, then we in the US would have likely gone through 7 or 8 waves infecting maybe 75% of the population before the vaccines finally kicked in.
Under that scenario it's hard to see much point in kicking the can down the road with lockdowns other than to avoid overloading the hospitals (and hospital administrators proved pretty good at stretching the capacity of their hospitals in the pinch).
Of course, under that scenario, lots of people would have voluntarily locked themselves down: e.g., as I pointed out 15 months ago, movie theaters were legally free to be open in Sweden, but their owners shut them down anyway because everybody had stopped going to the movies. (I cite movie theater data a lot because it's easy to find only at BoxOfficeMojo.com.)
But because of the remarkable efficacy and speed of development of vaccines (which, amusingly, Biden and Harris were spreading Fear, Uncertainty, and Doubt about during their debates last fall), blue state policies come out looking pretty good.
Back on March 18, 2020, I wrote:
"The [British] forecasters argue that suppression would need to be practiced for the majority of the next eighteen months, at which point they hope a vaccine would be ready.
"Even worse, I’d point out, it’s possible that no vaccine will be found: Coronaviruses are hard to vaccinate against. They tend to mutate rapidly.
"On the other hand, kicking the can down the road for seven months or so on the prospect of mass death might possibly pay off in multiple ways, not all of which can be specifically anticipated at present.
"For instance, it seems possible for the English-speaking countries to have a South Korean-style testing regimen working by fall that could put out flare-ups before they turn into the apocalypse.
"There are many different methods to create a vaccine. Well-funded crash programs might get lucky.
"They could speed things up somewhat by cutting corners on safety. Today’s vaccine researchers are not used to taking heroic steps to rush because they typically work on the fairly obscure infectious diseases that still lack vaccines, so there isn’t much excuse for risking nasty side effects lately.
"Medical treatment of the infected ought to improve with time. Testing will certainly get better than it has been. (In America, it can’t get worse.)"
My biggest takeaway from all this data is that I kept trying to move the sliders on each new image, and felt an increasing amount of shame each time.
Apologies I only had time to skim this for now, although it all looks reasonable and I'll try to read it more closely when I have time. Apologies for the usual academic self-promotion, but I did want to mention that I just published a paper on this topic in Health Services Research:
https://onlinelibrary.wiley.com/doi/full/10.1111/1475-6773.13688
The disadvantage compared to many of the ones discussed here is that ours is [much] less sophisticated modeling. The advantage is that by the same token we require [many] fewer assumptions.
-julian
What's the explanation for the Sweeden's huge peak around Jun 9-29 in infections not translating into analogous peak in covid deaths (which seem to continue to decline in this period)?
Is it that the first wave killed all voulnerable targets? Or perhaps voulnerable people stayed at home during the second wave? Or did something change in medical capacity or reporting? What's the official story?
My take: if this much analysis can only find a possible, small effect, then all but the most comically inadequate utilitarian (plausible) moral theory must condemn lockdowns vehemently.
Acts 16:31, 1 Corinthians 15:1-8, 1 Peter 1:17-21, Revelation 22:18-19
By the way, since there's been people suggesting similar threads for masks, vaccines etc., can I suggest another Much More Than You Want To Know COVID thread for a topic that's interested me for quite a bit but which I can't really make heads or tails out of: Long COVID? How much is it a real physical thing, how much a psychological thing, and how much other conditions lumped with one moniker when they randomly crop up or flare out after COVID? What are the actual numbers of sufferers - and how many of the cases are serious, or have lasted more than a few months? There really is little "official" information of any kind on these.
It’s good that you made an effort to assess actions in a comprehensive way and the additional debate and discussion is also useful. The discussion HAS to begin somewhere. The mental health aspects of the lockdowns remain largely unrecognized and unreported. Anecdotally, therapists were nowhere to be found in the US after the child depression epidemic spiked. Appointments were available one year out in the Seattle area for example. Hopefully some best practices around how to handle a pandemic will emerge. Lessons of the Spanish Flu didn’t seem to provide much guidance except that children remained in school and didn’t have screens so they were already outdoors, which probably helped a great deal.
I'm not sure I absorbed all of the statistical analysis, every time I try to consider more than a few simple factors, my head spins. However, just to throw *more* info into your pile, here's some potentially interesting information that is not usually discussed when going over the Swedish infection rates/death rates and course of the pandemic in 2020.
A lot of what happened initially in the pandemic (ie Feb-Mar 2020) was the result of a ‘perfect storm’ of setup and subsequent events. The longest prior situational setup was that, prior to 8 years ago, there was a gap in the Social Democratic dominance of parliament for two terms, where “Moderates” (considered right-ish here, more like Obama-style democracy/neo-liberalism) had a coalition with the Liberals (market libertarians) and other right-leaning parties. Several waves of privatization and deregulation happened in this government, eg they sold the post office to the Danish company Postnord (who promptly shut down 2/3 of processing stations to function economically, you can imagine how the mail is now), most of the health clinics were sold to Capio (a CAPital Investment Organization who runs health care in France, yup, health clinics have immense wait times now and are understaffed )and more that lovely old Social Democratic Sweden will never recover from…
Now, old peoples’ homes used to be run by the state, but many had been farmed out to private organizations who either cited some obscure regulation (not enough space between a door and a window type of thing) to kick all the old folks out and then rent the places out to refugees/asylum seekers, charging the state massive rent as Sweden received more and more people fleeing war and famine etc in Iraq, Syria, Eritrea, Afghanistan, etc. Or they continued to run them as old-folks' homes, just more cheaply.
The remaining now-private old peoples’ homes were additionally no longer regulated with regard to what was legal or necessary training for the workers—nor pay levels for workers—so the äldreboende ended up losing all their previous staff who simply would not work for minimum wage, and hiring, yes, off-the-boat immigrants at low pay, and no training.
Ok, so, next step. Stockholm, being the “Capitol of Scandinavia” is a very wealthy city, by which I mean: it’s extremely bourgeois. One of the yearly parts of people’s lives is Sportlov, a “sporting vacation” that happens in late February, where everybody goes skiing! Except, if you’re rich and from Stockholm, you don’t go skiing in Sweden, it’s too flat, so you go to Northern Italy. Only the middle class skis within Sweden.
(personal note, I was playing music on tour in the US in Dec2019-Jan2020, we all got incredibly sick directly after shows on the west coast after Christmas, continued with shows on the East Coast with the aid of University sports doctors applying “Z-pacs” of steroids and antibiotics, so we were probably super spreaders of an as-yet-unknown virus! I got back to Stockholm at the beginning of Feb, had two more shows in Finland mid-Feb, but by then Covid-19 had been identified and the start of pandemic prep had begun. Not that my family goes anywhere for Sportlov anyway, we’re not that class of citizens, being musicians and schoolteachers.)
So. Rich people flying to Northern Italy in February 2020. Who drives them to and from the airports? Taxis. The cab drivers are, yup, mostly recent immigrants from the aforementioned war-torn regions, who, having escaped with families, often live in multi-generational households. Viral transmission begins in earnest, apparently a lot from London and to a lesser extent from Northern Italy. The taxi drivers are a big vector. This combines with in-country skiers flying up north to ski in Sweden.
In many of these households it is/was common to sit by the bedside of a person who is sick, and you can imagine the health and safety warnings not exactly reaching through the language barriers (honestly, even getting info through to me in English was not top notch, and I saw nothing in Arabic or Somalian/Eritrean at that time, and I live across the street from an asyl-boende, where they house teenage asylum seekers with no families.) Older people started getting Covid, the families sat with them, then the men go back to work driving the taxis, the women off to the old peoples’ homes, where viral transmission starts to take hold of the elder population in a big way. And the people working in the old peoples’ homes did not know how to deal with it due to a) no actual medical or assistance training and b) language barriers. The bulk of Covid infection in Stockholm, the bulk of the deaths in March and April 2020 were caused by these vectors, minimum wage jobs held by recent immigrants carrying the virus home from rich people traveling the world, to their families and then out to their jobs. So the thing is, there *are* a lot of multi-generational households here, they just aren’t the rich white peoples’ households.
It took most of April and into May before they figured that out. Regardless, I don’t think any regulation has yet been put in place regarding training for hospice workers—it takes a long time to get anything done in this sort of bureaucracy. At least this year (2021) I’ve seen vaccination posters and info in nearly every language available all over the place. I think there are people *not* getting vaccinated, still (I believe it’s just gotten to the under-30 year olds this month) because, yes people are idiots and especially in extremely entitled societies there will be people who think that nothing can harm them. With the lack of any real lockdown, most people (in Stockholm anyway, I didn’t go anywhere to observe anybody else this past year) ignore anything not explicitly mandatory—mask usage on the subways was like 5% max any time during the past year+. People just don’t pay attention, think “won’t happen to me!” etc. So you could consider 'entitlement' a possible factor as well.
Replying to myself to amend the "multi-gen household" idea in Sweden: summer is different than winter. Many people move to sommarstuga/summer houses for a month or more in the summers which both alleviates congestion in the cities and creates more multi-generational households for short periods of time.
Currently (July 2021) cases in Stockholm are up 44% (says SVT) and it's almost all people in their 20s, not hanging with the folks in the country.
I cannot find anyone in comments interested in why Scandinavia did so well, so sorry if I'm repeating someone. Compare this: https://apnews.com/article/milan-health-ap-top-news-emmanuel-macron-virus-outbreak-b76b7e97cc6b3da0d2fa40a2e2b49503 to this: https://www.reddit.com/r/funny/comments/5blfam/finnish_bus_queue_personal_space_is_important/
At least that's where you have to look as well, if you see such a huge difference, but I guess you have to know say italians and scandinavians personally, otherwise it might not be a common knowledge.
Curious where these numbers come from, given that they are not at the link ...
> For example, when the EPA is determining how worth-it environmental regulations are, they value one life at $9.1 million; when the Department of Transportation is determining how worth-it road safety regulations are, they value a life at $9.6 million
Thank you, great analysis, for me the crucial point: "It's harder to justify strict lockdowns in terms of the non-economic suffering produced. Even assumptions skewed to be maximally pro-strict-lockdown, eg where strict lockdowns would have prevented every single coronavirus case, suggest that it would have taken dozens of months of somewhat stricter lockdown to save one month of healthy life."
and I would add that what you missed - you didn't address at all - is who is threatened by the disease and who is threatened by the restrictions, that here we are clearly shifting the burden from the sick and elderly to the children, that they are hardly threatened by the virus at all, but by the restrictions quite fundamentally, and again it is up for debate how such a manufactured shift is moral
and the second point, that you does not address at all that restrictions can still hurt after they have ended, as explained here:
Self-harm, suicide attempts, depression, eating disorders. Teenagers have filled the psychiatric wards of hospitals. It's like another epidemic has arrived - this time a wave of mental health problems. There was no relief for the kids as the lockdown ended. https://www.heroine.cz/rodina-a-vychova/5173-tezky-skolni-rok-je-za-nami-uleva-pro-teenagery-ale-neprichazi-co-ted-od-nas-potrebuji-ze-vseho-nejvic
The WHO's definition of health, includes well-being, was destroyed by the restrictions and again almost impossible to calculate.
Brilliant pondering. My take is the population that continues to wear masks and vaccinate will be generally fine as the variant breaks through the vaccinated. Those who don’t continue to wear masks or eschew vaccination will have a much higher viral load, and thus a tougher fight for recovery. For children under 12, they need to be isolated until a vaccination is available to them.
I remember people talking about non-official signals of seriousness as drivers of individual behavior; that there was a major uptick in taking the pandemic seriously when large organizations (e.g. the NCAA and NBA) started taking it seriously even though it lost them a lot of money and a major downtick when leaders (California, Britain, etc) were caught hypocritically flouting their own lockdowns.
Did that show up in the data?
Even if the confounding factors could all be untangled, isn't all of this granular analysis basically pointless because: (1) the "new infection" data are kinda garbage due to positive test rates being highly unreliable and, in any event, testing is not conducted on a randomized basis; and (2) the "covid deaths" data are also kinda garbage because they don't distinguish "died with" and "died from" covid.
Indeed, but there is more.
Sweden comes 1000-4000 deaths 'worse' than it allegedly could in Scott analysis, but Scott makes the mistake of not looking at the big picture. If you plot overall mortality over the years per week (for example) you don't see anything interesting at all. Scott is just looking at noise. There is no signal there. There was no pandemonium of an epidemic. There was nothing but pointless lockdowns.
Also, Scott should compare with the predictions the modellers did. They expected 70000 excess dead in Sweden *with* the measures taken. The modellers were stupidly wrong and have never accepted their errors.
I didn't see where you dealt with the problem of risk. Unlike plague, dengue or cholera, COVID was a novel disease. We now know all kinds of stuff we didn't in early 2020. For example, we know roughly how dangerous and how contagious it is. We know how likely it is to kill someone, how long a typical infection will last, something about how to treat the disease, how quickly it mutates and so on. A lot of our knowledge is still sketchy. No one has had COVID and lived for even two years.
When dealing with a novel risk, it can make sense to be more cautious than when dealing with a well understood risk. Doing a retrospective analysis ignores this. It's called Monday morning quarterbacking in certain circles. Suppose COVID had been more deadly, less amenable to such treatment as we now have, more infectious, more likely to become more dangerous adn so on. Similarly, it could, as our president at the time suggested, have simply gone away on its own.
If you are a politician, you have to deal with risk. Sometimes they overreact. After 9/11 the political assumption was that the world was full of terrorists plotting new, similar attacks, and that it made sense to invade Afghanistan and Iraq, if only to send a message that the US was serious about fighting terrorism. Anyone who suggested that this was overreacting, perhaps arguing that it made sense to invade Afghanistan where the attack was organized but not Iraq which had nothing to do with it, they were noisily and sometimes viciously shouted down and sidelined. One politician argued that overreaction was proper in the face of "unknown unknowns".
Of course, it was much easier to understand and deal with 9/11 style terrorism. We had a lot of knowledge about how terrorist attacks could work, what was needed to make them less likely, conditions and potentials in the Mideast and existing terrorist organizations. It's not as if the intelligence community knew nothing about Al Qaeda and friends.
In contrast, COVID was something no one had ever seen before. The closest models were SARS and MERS which were much less infectious and could be stopped with conventional contact tracing. COVID was clearly more dangerous and much less well known. The only way to find out what the threat was was to let the threat play out and try to stay ahead of it. Needless to say, this did not work very well in Italy, NYC or elsewhere. It only worked in China because of massive reaction, shutting down entire cities, complete lockdowns that would have been impossible in less regulated nations.
Looking back and saying that we should have traded so many dollars for so many lives or illnesses or perhaps traded so many lives or illnesses for dollars is rather pointless.
As mentioned: Belgium covid high death rate partly explained by an unusual counting method. I'm surprised this analysis did not include excess death rate trends. Data for 2020 has come out, for example, here. This table is "age adjusted"; I'm not clear how that works exactly). Note that Sweden comes out lower/"better" than Germany by this metric. https://www.cebm.net/covid-19/excess-mortality-across-countries-in-2020/ )
I'm missing your ethical assumptions in this piece when you talk about what does and doesn't justify a lockdown. You seem to be implying in some parts that governments ought to act according to some utilitarian sum, but why shouldn't they (for example) maximize freedoms of the individual? Actually I don't think 'objective ethical theories' can be grounded at all and then government are just the results of complicated social contracts between citizens that act out of their (ir)rational self-interest, and from that perspective why be in favour of a lockdown if it doesn't benefit you more than not having a lockdown?
It was painful reading this.
Late response here, but the Branch Covidians are plunging us into a dark age from which we may never recover. If you think this was about giving boomers 1-2 years of extra life, or saving fat people, this is very naive.
Humans aren’t meant to do lockdowns, better to have no humans left than a world where we are tightly controlled in this way.
I’d rather DIE than live in China, and am starting to think that everyone should rather than anyine live in such a subhuman state where you can be welded into your dwelling, removed into camps, injected involuntarily and be rendered nonessential
As a psychiatrist, if you don’t recognize the catastrophic effect of this on mental health, I suggest a new profession